Evolution of fruit and seed characters in the Diervilla and Lonicera clades (Caprifoliaceae, Dipsacales)

Abstract

Background and Aims

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.

Methods

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.

Key Results

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.

Conclusions

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.

INTRODUCTION

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.

Fig. 1.

Phylogenetic tree for the order Dipsacales with indication of families and informal clade names (based on that of Winkworth et al., 2008b, fig. 3a).

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).

Table 1.

Overview of classifications of Dipsacales

Present study		Hara (1983)			Backlund and Pyck (1998)	Donoghue et al. (2001)
Adoxaceae	Adoxa, Sinadoxa, Tetradoxa	Adoxaceae		Adoxa, Sinadoxa, Tetradoxa	Adoxaceae	Adoxaceae	Adoxoideae	Adoxa, Sinadoxa, Tetradoxa
	Sambucus	Caprifoliaceae	Sambucoideae	Sambucus				Sambucus
	Viburnum		Viburnoideae	Viburnum				Viburnum
Caprifoliaceae	Diervilla clade		Caprifolioideae	Diervilleae	Diervillaceae	Caprifoliaceae	Diervilleae
	Lonicera clade			Caprifolieae	Caprifoliaceae		Caprifolieae
				Triosteae
	Linnaea clade			Linnaeeae	Linnaeaceae		Linnaeeae
	Morina clade	Morinaceae			Morinaceae		Morinaceae
	Dipsacus clade	Dipsacaceae			Dipsacaceae		Dipsacaceae
	Valeriana clade	Valerianaceae			Valerianaceae		Valerianaceae

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.

MATERIALS AND METHODS

Taxon sampling

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.

Molecular methods

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).

Table 2.

Base composition of amplification and sequencing primers

Locus	Primers	Sequence 5′–3′	Reference
ITS	ITS5	GGA AGT AAA AGT CGT AAC AAG G	White et al. (1990)
	ITS4	TCC TCC GCT TAT TGA TAT GC	White et al. (1990)
trnK	trnK11	CTC AAC GGT AGA GTA CTC G	Young et al. (1999)
	matK510R	GAA GAG TTT GAA CCA AKA YTT CC	Young et al. (1999)
matK	matK-53F	CTT GTT TTG RCT NTA TCG CAC TAT G	Young et al. (1999)
	matK950R	CCA CAR CGA AAA ATR MCA TTG CC	Young et al. (1999)
	matK775F	TCT TGA ACG AAT CTA TTT CTR YGG	Young et al. (1999)
	matK1349R	CTT TTG TGT TTC CGA GCY AAA GTT C	Young et al. (1999)

Table 3.

Statistics for the aligned sequence data

	ITS	trnK	matK
Number of taxa	32	25	30
Aligned matrix length (sequence data)	673	776	954
Number of constant characters	313	476	609
Number of variable characters	360	300	345
Number of potentially parsimony-informative characters	265	144	204

Morphological methods

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.

Character state delimitation

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.

Table 4.

Description of morphological characters

1	Fruit type: (0) drupe, (1) berry, (2) achene, (3) capsule
2*	Maximum number of carpels: (0) 1, (1) 2, (2) 3, (3) 4, (4) 5
3	Number of sterile carpels: (0) no sterile carpels, (1) single sterile carpel, (2) two sterile carpels
4*	Number of seeds: (0) 1, (1) 2, (2) 3, (3) 4, (4) 5, (5) 6–20, (6) more than 20
5	Endocarp sclerification: (0) endocarp not sclerified, (1) endocarp sclerified
6	Number of endocarp layers: (0) 1, (1) 2, (2) 3
7	Endocarp part of diaspore: (0) endocarp not part of diaspore, (1) endocarp part of diaspore
8	Mesocarp anatomy: (0) mesocarp dry, (1) mesocarp fleshy
9	First mechanical layer: (0) epicalyx, (1) pericarp, (2) endocarp, (3) seed coat
10	Presence of epicalyx: (0) epicalyx absent, (1) epicalyx present
11	Exotesta anatomy: (0) exotesta compressed, (1) exotesta parenchymatous, (2) exotesta sclerified
12	Seed coat crystals: (0) crystals absent, (1) crystals present
13	Presence of sclerified endotesta: (0) sclerified endotesta absent, (1) sclerified endotesta present
14	Presence of anti-raphe: (0) anti-raphe absent, (1) anti-raphe present
15*	Embryo size: (0) embryo length less than 3/4 of seed length, (1) embryo length greater than 3/4 of seed length but not occupying entire seed, (2) embryo occupying entire seed and no endosperm present
16*	Seed coat thickness: (0) less than 25 µm, (1) greater than 25 µm and less than 100 µm, (2) greater than 100 µm
17	Winged seeds: (0) wingless seeds, (1) winged seeds
18	Sepal modification for fruit dispersal: (0) none, (1) developing into a plumose seed/fruit, (2) developing to seeds/fruits with awns/bristles, (3) enlarged and leaf-like aiming for wind dispersal
19	Carpel vascularization: (0) free adaxial and abaxial, (1) adaxial bundles only, (2) only free abaxial, adaxial not recessed
20	Ovule vascularization: (0) single bundle, (1) double or compound bundles
21	Ovule position with respect to the central axis: (0) marginal, (1) marginal above, median below, (2) median
22	Ovule reduction: (0) no traces of reductions, (1) sterile ovules, (2) vestigial archespore surrounded by nucellar tissue, (3) vestigial archespore
23	Integument number: (0) unitegmic, (1) bitegmic
24	Nucellus thickness: (0) tenuinucellate, (1) crassinucellate
25	Endothelium: (0) absent, (1) present, feebly differentiated, (2) prominent, with crystal layer
26	Embryo sac development: (0) Polygonum type, (1) Adoxa type, (2) Allium type
27	Embryogeny type: (0) Solanad, (1) Asterad, (2) Piperad, (3) Onagrad
28	Endosperm in seed: (0) absent, (1) scanty, (2) copious
29	Embryo development: (0) leucoembryote, (1) chlorophyllous

* Character state delimitation explained in detail in the Materials and Methods. A more extensive description of characters 18–29 can be found in Backlund (1996).

Phylogenetic analyses

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.

Table 8.

Morphological data set

Taxon	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29
Abelia chinensis	2	3	2	1	?	?	1	0	0	0	?	?	?	?	?	?	0	3	2	1	2	1	0	0	0	0	1	2	0
A. parvifolia	2	3	2	1	?	?	1	0	0	0	?	?	?	?	?	?	0	3	2	1	2	1	0	0	0	0	1	2	0
Adoxa moschatellina	0	4	0	4	1	2	1	1	2	0	1	0	0	1	0	0	0	0	0	0	2	3	0	0	1	1	1	2	0
Centranthus ruber	2	2	2	0	1	2	1	0	1	0	0	?	0	0	2	0	0	1	2	1	2	1	0	0	2	0	1	0	1
Dipelta floribunda	2	3	2	1	1	0	1	0	0	0	0	?	0	0	0	0	0	0	2	1	2	1	0	0	0	0	1	2	0
D. yunnanensis	2	3	2	1	1	0	1	0	0	0	0	?	0	0	0	0	0	0	2	1	2	1	0	0	0	0	1	2	0
Diervilla sessilifolia	3	1	0	6	1	1	0	0	3	0	2	0	1	0	0	1	0	0	2	0	0	1	0	0	0	0	1	2	0
Dipsacus fullonum	2	0	0	0	?	0	1	0	0	1	0	?	0	0	1	0	0	2	2	1	2	1	0	0	2	0	2	2	1
Heptacodium miconioides	2	2	2	0	1	0	1	0	1	0	0	1	0	0	0	0	0	0	2	?	?	1	0	0	0	0	1	2	0
Kolkwitzia amabilis	2	2	2	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0	2	1	2	1	0	0	0	0	1	2	0
Leycesteria crocothyrsos	1	4	0	5	0	0	1	1	3	0	2	0	1	1	0	1	0	0	2	0	0	1	0	0	0	0	1	2	0
L. formosa	1	4	0	5	0	0	1	1	3	0	2	0	1	1	0	1	0	0	2	0	0	1	0	0	0	0	1	2	0
Linnaea borealis	2	3	2	1	?	?	1	0	0	0	?	?	?	?	?	?	0	0	2	1	2	1	0	0	0	0	1	2	0
Lonicera dioica	1	2	0	5	0	0	1	1	3	0	2	1	1	1	0	2	0	0	2	0	0	1	0	0	0	2	1	2	0
L. involucrata	1	2	0	5	0	0	1	1	3	0	2	0	1	1	0	1	0	0	2	0	0	1	0	0	0	2	1	2	0
L. maximowiczii	1	2	0	5	0	0	1	1	3	0	2	1	1	1	0	2	0	0	2	0	0	1	0	0	0	2	1	2	0
L. xylosteum	1	2	0	5	0	0	1	1	3	0	2	1	1	1	0	2	0	0	2	0	0	1	0	0	0	2	1	2	0
L. maackii	1	2	0	5	0	0	1	1	3	0	2	1	1	1	0	2	0	0	2	0	0	1	0	0	0	2	1	2	0
Morina longifolia	2	2	2	0	1	0	1	0	0	1	0	?	0	1	1	0	0	0	2	1	2	1	0	0	2	0	0	2	0
Symphoricarpos albus	0	3	2	1	1	2	1	1	2	0	2	1	1	1	0	1	0	0	2	1	2	1	0	0	0	0	1	2	0
S. occidentalis	0	3	2	1	1	2	1	1	2	0	2	1	1	1	0	1	0	0	2	1	2	1	0	0	0	0	1	2	0
S. ebulus	0	4	0	4	1	2	1	1	2	0	0	0	0	0	1	0	0	0	0	1	2	3	0	0	0	1	1	2	3
S. racemosa	0	4	0	4	1	2	1	1	2	0	0	0	0	0	1	0	0	0	0	1	2	3	0	0	0	1	1	2	3
Scabiosa columbaria	2	0	0	0	0	0	1	0	0	1	0	?	0	0	1	0	0	2	2	1	2	1	0	0	2	0	2	2	1
Succisa pratensis	2	0	0	0	?	0	1	0	0	1	0	?	0	0	1	0	0	2	2	1	2	1	0	0	2	0	2	2	1
Triosteum perfoliatum	0	3	1	2	1	1	1	1	2	0	2	1	1	1	0	1	0	0	2	1	2	1	0	0	0	0	1	2	0
Viburnum acerifolium	0	2	2	0	1	1	1	1	2	0	1	0	0	1	0	1	0	0	0	1	2	2	0	1	0	2	1	2	1
V. plicatum	0	2	2	0	1	1	1	1	2	0	1	0	0	1	0	1	0	0	0	1	2	2	0	1	0	2	1	2	1
Valeriana officinalis	2	2	2	0	1	2	1	0	1	0	0	0	0	0	2	0	0	1	2	1	2	1	0	0	2	0	1	0	1
Valerianella locusta	2	2	2	0	1	2	1	0	1	0	0	?	0	0	2	0	0	0	2	1	2	1	0	0	2	0	1	0	1
Weigela florida	3	1	0	6	1	1	0	0	3	0	2	0	1	0	0	1	1	0	2	0	0	1	0	0	0	0	1	2	0
W. subsessilis	3	1	0	6	1	1	0	0	3	0	2	0	1	0	0	1	1	0	2	0	0	1	0	0	0	0	1	2	0

Charcters 18–20 (Table 2) were obtained from the study of Backlund (1996).

Comparing matrices and topologies

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).

Table 5.

ILD statistics from data set comparison

	Original aligned matrices			Lonicera clade excluded			Linnaea clade excluded			Heptacodium excluded
	trnK	matK	morphology	trnK	matK	morphology	trnK	matK	morphology	trnK	matK	morphology
ITS	0·00	0·45	0·00	0·18	1·00	0·97	0·00	0·02	0·00	0·00	0·86	0·02
trnK	–	0·00	0·00	–	0·02	0·17	–	0·00	0·00	–	0·01	0·00
matK	–	–	0·00	–	–	0·70	–	–	0·00	–	–	0·01

Incongruencies are indicated in bold.

Table 6.

Tree comparison using likelihoods and approximately unbiased (AU) and Shimodaira–Hasegawa (SH) tests

	ITS, trnK, matK ML tree			ITS ML tree			trnK, matK ML tree
	Δ ln L	AU	SH	Δ ln L	AU	SH	Δ ln L	AU	SH
BI consensus tree based on combined data	5·2	0·22	0·90	7·0	0·22	0·82	10·2	0·23	0·88
BI consensus tree based on molecular data	15·1	<0·01	0·81	17·9	<0·01	0·64	12·6	0·02	0·85
Shortest MP trees based on combined data	4·0–5·2	0·44–0·22	0·91–0·90	5·4–7·0	0·52–0·22	0·86–0·82	10·2	0·22–0·21	0·88
Shortest MP trees based on morphological data	403·1–528·5	<0·01	0·12	125·7–146·9	<0·01	<0·01	293·9–420·1	<0·01	<0·01
ML tree based on ITS data	126·0	<0·01	<0·01	best	0·70	0·93	139·4	<0·01	0·03
ML tree based on trnK data	1353·0	<0·01	<0·01	544·8	<0·01	<0·01	804·4	<0·01	<0·01
ML tree based on matK data	849·3	<0·01	<0·01	480·6	<0·01	<0·01	274·4	<0·01	<0·01
ML tree based on molecular data	best	0·69	0·97	12·7	0·08	0·72	4·5	0·85	0·98

Character evolution

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.

RESULTS

Seed shape and size

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).

Fig. 2.

Seed shape and size. (A, B) Diervilla sessilifolia: (A) lateral view; (B) TS, sclerified exo- and endotesta. (C) Weigela hortensis, lateral view. (D) Weigela floribunda, winged seed shaped by adjacent seeds. (E) Leycesteria formosa, lateral view. (F) TS, Leycesteria crocothyrsos, dorsiventrally flattened seed with sclerified exo- and endotesta. (G, H) Lonicera dioica: (G) lateral view; (H) TS, raphe and anti-raphe. (I) Triosteum perfoliatum, lateral view. (J) Triosteum hirsutum, TS, sclerified endocarp surrounding seed with sclerified exo- and endotesta, raphe and anti-raphe. (K) Symphoricarpos mollis, lateral view. (L) Symphoricarpos occidentalis, TS, sclerified endocarp surrounding seed, laterally flattened raphe and anti-raphe. (M–O) Heptacodium miconioides: (M) lateral view; (N) TS, pericarp with ribs surrounding seed with compressed, parenchymatous seed coat; (O) LS, fruit with one seed, minute embryo.

Table 7.

Seed and endocarp characteristics of studied species

Taxon	Tribe	Seed				Endocarp
		Length (mm)	Width (mm)	Thickness (mm)	Coat thickness(μm)	Thickness (μm)*
Diervilla sessilifolia	Diervilla clade	1·13 ± 0·05	0·79 ± 0·07	0·50 ± 0·06	79·77 ± 11·24	–
D. rivularis	Diervilla clade	0·98 ± 0·06	0·68 ± 0·04	0·50 ± 0·02	71·03 ± 6·72	–
Weigela florida	Diervilla clade	1·82 ± 0·22	0·91 ± 0·09	0·54 ± 0·05	74·38 ± 47·11	–
W. subsessilis	Diervilla clade	1·96 ± 0·18	1·10 ± 0·08	0·66 ± 0·07	72·58 ± 53·66	–
W. floribunda	Diervilla clade	1·52 ± 0·07	0·74 ± 0·10	0·54 ± 0·08	71·81 ± 34·61	–
W. japonica	Diervilla clade	1·19 ± 0·19	0·84 ± 0·09	0·58 ± 0·06	78·57 ± 33·08	–
W. hortensis	Diervilla clade	1·71 ± 0·16	1·13 ± 0·26	0·62 ± 0·03	54·91 ± 17·06	–
Leycesteria formosa	Lonicera clade	1·30 ± 0·05	0·96 ± 0·04	0·59 ± 0·04	56·23 ± 6·63	–
L. crocothyrsos	Lonicera clade	0·86 ± 0·04	0·67 ± 0·07	0·40 ± 0·01	66·28 ± 6·28	–
Lonicera dioica	Lonicera clade	3·60 ± 0·11	2·85 ± 0·18	1·96 ± 0·09	286·94 ± 55·66	–
L. canadensis	Lonicera clade	3·34 ± 0·38	2·13 ± 0·12	1·46 ± 0·09	124·82 ± 12·76	–
L. etrusca	Lonicera clade	5·02 ± 0·24	3·33 ± 0·22	1·70 ± 0·09	237·10 ± 24·29	–
L. implexa	Lonicera clade	3·93 ± 0·16	2·82 ± 0·11	1·44 ± 0·18	292·50 ± 62·73	–
L. caprifolium	Lonicera clade	4·63 ± 0·23	3·45 ± 0·21	1·37 ± 0·47	258·67 ± 40·02	–
L. vesicaria	Lonicera clade	5·37 ± 0·23	3·27 ± 0·17	1·26 ± 0·11	215·86 ± 34·00	–
L. alpigena	Lonicera clade	6·38 ± 0·22	5·08 ± 0·22	2·12 ± 0·08	432·33 ± 45·45	–
L. muscaviensis	Lonicera clade	3·17 ± 0·10	2·65 ± 0·09	1·08 ± 0·07	176·92 ± 31·69	–
L. involucrata	Lonicera clade	2·55 ± 0·07	1·87 ± 0·24	1·00 ± 0·08	71·17 ± 12·00	–
L. javanica	Lonicera clade	4·33 ± 0·34	2·76 ± 0·14	0·80 ± 0·06	101·64 ± 9·22	–
L. chrysantha	Lonicera clade	4·00 ± 0·20	3·38 ± 0·31	1·24 ± 0·10	155·70 ± 19·43	–
L. maximowiczii	Lonicera clade	4·02 ± 0·40	3·19 ± 0·43	1·85 ± 0·22	150·66 ± 41·81	–
L. xylosteum	Lonicera clade	3·88 ± 0·19	3·12 ± 0·19	0·88 ± 0·15	120·12 ± 21·33	–
L. maackii	Lonicera clade	4·63 ± 0·33	2·77 ± 0·25	1·58 ± 0·08	107·48 ± 18·58	–
Symphoricarpos albus	Lonicera clade	3·73 ± 0·21	1·90 ± 0·10	1·13 ± 0·03	51·18 ± 10·03	190·67 ± 26·27
S. oreophilus	Lonicera clade	3·37 ± 0·06	1·88 ± 0·08	1·07 ± 0·03	59·65 ± 28·71	240·35 ± 27·90
S. occidentalis	Lonicera clade	3·30 ± 0·08	2·24 ± 0·05	1·15 ± 0·00	59·54 ± 15·99	211·58 ± 27·12
S. mollis	Lonicera clade	3·35 ± 0·13	1·88 ± 0·03	1·05 ± 0·05	68·87 ± 23·52	267·50 ± 48·72
Triosteum perfoliatum	Lonicera clade	5·18 ± 0·60	2·50 ± 0·28	1·43 ± 0·25	60·84 ± 11·54	505·27 ± 326·00
T. angustifolium	Lonicera clade	5·07 ± 0·06	2·08 ± 0·03	1·27 ± 0·12	39·11 ± 3·96	456·35 ± 306·86
T. hirsutum	Lonicera clade	6·50 ± 0·10	3·20 ± 0·00	0·97 ± 0·06	60·99 ± 11·80	355·28 ± 54·89
T. sinuatum	Lonicera clade	6·12 ± 0·13	2·80 ± 0·22	1·46 ± 0·05	94·38 ± 28·76	875·42 ± 466·11
Heptacodium miconioides	–	9·17 ± 0·10	1·37 ± 0·10	1·40 ± 0·09	7·33 ± 2·30	26·92 ± 4·05

* Endocarp thickness only applies to fruits containing pyrenes, i.e. Symphoricarpos and Triosteum.

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).

In Heptacodium (Fig. 2M–O), seed shape ranges from spindle-shaped to slightly clavate (Fig. 2M, O). Seed shape in cross-section is circular to slightly elliptic (Fig. 2N).

Seed surface

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 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).

Fig. 3.

Seed surface. (A) Diervilla rivularis, outer tangential wall removed, exotesta with U-shaped sclerification pattern. (B) Weigela japonica, outer tangential wall removed, sclerified exotesta with pits. (C) Leycesteria crocothyrsos, sclerified exotesta with concave outer tangential wall and slightly undulating cell outline. (D) Lonicera dioica, sclerified, polygonal exotestal cells with distinct cell boundaries. (E) Triosteum perfoliatum, sclerified exotestal cells with concave outer tangential cell walls with undulating cell outline. (F) Symphoricarpos mollis, sclerified exotesta with polygonal cell outline. (G) Heptacodium miconioides, compressed, parenchymatous seed coat without distinct seed sculpture.

Fig. 4.

Seed coat anatomy. (A, B) Diervilla sessilifolia: (A) TS, exotesta with U-shaped sclerification pattern, compressed mesotesta, partially compressed, sclerified endotesta; (B) sclerified exotesta, outer tangential wall removed. (C, D) Weigela florida: (C) sclerified exotesta, compressed mesotesta, sclerified endotesta and adjacent endosperm with starch granules; (D) sclerified exotesta, outer tangential wall removed. (E) Leycesteria formosa, sclerified exotesta, remnants of compressed, parenchymatous mesotesta, sclerified endotesta reduced to thin cell layer. (F) Leycesteria crocothyrsos, sclerified exotestal cells with pits. (G) Lonicera canadensis, large, sclerified exotesta cells, compressed mesotesta, sclerified endotesta reduced to thin cell layer. (H) Lonicera alpigena, tall, sclerified exotestal cells with pits. (I) Triosteum hirsutum, moderately sclerified exotestal cells with concave outer tangential walls, compressed mesotesta, small sclerified endotestal cells. (J) Triosteum sinuatum, sclerified exotesta, large, parenchymatous mesotestal cells, thin, sclerified endotesta. (K) Symphoricarpos occidentalis, sclerified exotesta with inner tangential wall shaped by adjacent mesotesta. (L) Symphoricarpos oreophilus, sclerified exotesta, remnants of parenchymatous mesotesta. (M–O) Heptacodium miconioides, compressed, parenchymatous seed coat, adjacent endosperm. (P) Immature seed with mesocarp and sclerified endocarp, parenchymatous exotesta and multi-layered mesotesta, endotesta reduced to thin layer. Abbreviations: XT, exotesta; MT, mesotesta; ET, endotesta; EN, endosperm; EC, endocarp; MC, mesocarp; SC, seed coat.

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 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 4M–O).

Anatomy of seed coat

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.

Endosperm

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.

Fig. 5.

Endosperm anatomy, embryo morphology and vascular anatomy. (A) Diervilla sessilifolia, TS, thin-walled, isodiametric endosperm cells. (B) Weigela hortensis, TS, endosperm. (C) Leycesteria formosa, TS, endosperm. (D) Lonicera maackii, TS, slightly thickened, isodiametric endosperm cells. (E) Triosteum perfoliatum, TS, endosperm. (F) Symphoricarpos albus, TS, endosperm. (G) Heptacodium miconioides, TS, large, isodiametric endosperm cells. (H) Weigela floribunda, LS, winged seed with embryo. (I) Lonicera involucrata, LS, embryo less than 3/4 of seed length. (J) Triosteum sinuatum, small, drop-shaped embryo. (K) Diervilla sessilifolia, raphal bundle in mesotesta, rexigenous cavity (arrow) with spiral tracheids. (L) Lonicera xylosteum, raphal bundle with rexigenous cavity (arrows) in multi-layered mesotesta. (M) Triosteum hirsutum, large raphal bundle, rexigenous cavity (arrow) with spiral tracheids, multi-layered exotesta. (N) Heptacodium miconioides, immature seed, raphal bundle in multi-layered mesotesta, rexigenous cavity (arrow) with spiral tracheids surrounded by collenchyma.

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).

Embryo

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).

Vasculature

In the Diervilla clade, seeds of Diervilla (Figs 2B and 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 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 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 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.

Endocarp

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.

Fig. 6.

Endocarp morphology and anatomy. (A, B) Triosteum angustifolium: (A) dorsal view; (B) ventral view. (C) Triosteum perfoliatum, TS, sclerified endocarp with ribs surrounding seed. (D–F) Symphoricarpos albus: (D) dorsal view; (E) ventral view; (F) TS, sclerified endocarp surrounding seed. (G) Triosteum angustifolium, TS, endocarp rib with two layers of fibres and adjacent mesocarp. (H) Triosteum sinuatum, TS, endocarp rib. (I, J) Symphoricarpos occidentalis: (I) endocarp with three distinct layers, i.e. outer layer of fibres, sclereids, inner layer of fibres, seed with sclerified exotesta; (J) endocarp with adjacent seed. Abbreviations: EC, endocarp; SC, seed coat; EN, endosperm.

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).

Phylogenetic analyses (Figs 7 and 8)

Fig. 7.

Consensus trees from maximum parsimony (MP) analyses with bootstrap support values (>50). As the MP topologies based on sequence data (B) and combined data only differed in the Lonicera and the Linnaea clades, (B) only shows the latter two clades and the respective bootstrap values. (A) Topology based on combined data with bootstrap values of MP analysis of sequence data (left value) and bootstrap values of combined data (right value); bootstrap values with an asterisk (Lonicera and Linnaea clades) are based on combined data. (B) Topology of Lonicera and Linnaea clades based on sequence data with bootstrap values.

Fig. 8.

Majority rule trees of Bayesian inference (BI) analyses. Thicker branches have siginficant posterior probability scores (>0·95). As BI topologies based on sequence data (B) and combined data only differed in the Lonicera and Linnaea clades, (B) only shows the latter two clades. (A) Topology based on combined data. (B) Topology based on sequence data.

Molecular data

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).

Combined data

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).

Incongruence between data partitions and topologies

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.

DISCUSSION

Diervilla clade

The composition and monophyly of the Diervilla clade, Diervilla and Weigela are strongly supported in our analyses (Figs 7 and 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.

Lonicera clade

Whereas the monophyly of all four genera is well supported, intergeneric relationships are less clear (Figs 7 and 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).

Heptacodium

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.

Transfer of function

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.

Character evolution

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 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.

Fig. 9.

Parsimony optimization results for ten fruit characters (left) and seven seed characters (right). The upper number indicates the respective character, whilst the bottom number indicates the character state (Table 4).

Fruit type

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.

Carpel number and carpel sterility

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.

Seed number

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.

Number of endocarp layers

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.

Fleshy fruits

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.

Anatomy of seed coat

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.

Embryo size

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.

Conclusions

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.

Future prospects

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.

SUPPLEMENTARY DATA

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.

APPENDIX 1

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.

APPENDIX 2

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	–
Adoxa moschatellina	NA	U88194	EF490235	EF490235
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	–
Dipelta yunnanensis	NA	AY236180	AF446910	AY290042
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
Linnaea borealis	NA	AY236181	AF446911	AY290040
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	–
Lonicera xylosteum	NA	EU240714	AM503819	–
Morina longifolia	NA	AY236185	AF446915	AY290020
Sambucus ebulus	NA	DQ521256	EF490239	EF490239
Scabiosa columbaria	NA	AY236188	AF446918	AY290032
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
Sambucus racemosa	NA	AY236171	AY265204	AY265204
Triosteum perfoliatum	Botanical Garden Uppsala University, N. Pyck 1963-1028	AY236175	AF446905	FJ745409
Valeriana officinalis	NA	DQ180745	AY310467	AY794362
Valerianella locusta	NA	DQ354168	AF446922	AY794398
Viburnum acerifolium	NA	AY265114	AF446897	AY265160
Viburnum plicatum	NA	AY265143	AY265189	AY265189
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.