Phylogeny-based developmental analyses illuminate evolution of inflorescence architectures in dogwoods (Cornus s. l., Cornaceae)


Authors for correspondence:
Qiu-Yun (Jenny) Xiang
Tel: +1 919 515 2728
Robert G. Franks
Tel: +1 919 513 7705


  • Inflorescence architecture is important to angiosperm reproduction, but our knowledge of the developmental basis underlying the evolution of inflorescence architectures is limited. Using a phylogeny-based comparative analysis of developmental pathways, we tested the long-standing hypothesis that umbel evolved from elongated inflorescences by suppression of inflorescence branches, while head evolved from umbels by suppression of pedicels.
  • The developmental pathways of six species of Cornus producing different inflorescence types were characterized by scanning electron microscopy (SEM) and histological analysis. Critical developmental events were traced over the molecular phylogeny to identify evolutionary changes leading to the formation of umbels and heads using methods accounting for evolutionary time and phylogenetic uncertainty.
  • We defined 24 developmental events describing the developmental progression of the different inflorescence types. The evolutionary transition from paniculate cymes to umbels and heads required alterations of seven developmental events occurring at different evolutionary times.
  • Our results indicate that heads and umbels evolved independently in Cornus from elongated forms via an umbellate dichasium ancestor and this process involved several independent changes. Our findings shed novel insights into head and umbel evolution concealed by outer morphology. Our work illustrates the importance of combining developmental and phylogenetic data to better define morphological evolutionary processes.


Alterations in inflorescence architecture result in biological changes associated with pollination and reproduction that can drive new ecological adaptation and speciation (Wyatt, 1982; Fishbein & Venable, 1996; Friedman & Harder, 2005). During angiosperm evolution, innovation in inflorescence architecture has repeatedly occurred in different angiosperm lineages via alterations in the developmental pathways responsible for the ancestral forms (Tucker & Grimes, 1999; Soltis et al., 2005). Despite the importance of understanding how developmental and genetic changes have shaped inflorescence architecture, little progress has been made, in part because existing model organisms exhibit little variation in these traits and there have been few comparative studies between species from a phylogenetic perspective (Cronk et al., 2002; Kellogg, 2004, 2006).

The architecture of an inflorescence in part depends on its branching pattern and the relative position at which flowers are borne and is controlled by diverse developmental pathways (Benlloch et al., 2007). In an indeterminate (racemose) inflorescence, the apices do not end in a terminal flower, and are able to grow for a long period, generating a continuous main axis that laterally produces floral meristems or branches. In a determinate (cymose) inflorescence, the apices of the main and lateral axes terminate in a flower. Much additional diversity of forms is determined by variation in the duration of the ‘vegetative state’ of the inflorescence meristem (IM) during which the bud produces additional inflorescence branches instead of generating flowers (Prusinkiewicz et al., 2007) as well as the variation in the length of inflorescence branches and pedicels. Among the great diversity of inflorescence architectures, raceme, spike, corymb, cyme, umbel, and head are the most common simple forms. These simple forms may assemble in a variety of ways to form more complex architectures. Raceme, spike, and corymb are elongated indeterminate inflorescences. The cyme is an elongated determinate form, while umbel and head are condensed forms that can be either determinate or indeterminate. At the present, the development of raceme and cymes are well understood in a few model plants such as Arabidopsis, Antirrhinum, rice, maize, petunia and tomatoes (Coen et al., 1990; Alvarez et al., 1992; Weigel et al., 1992; Ingram et al., 1995; Blazquez et al., 1997; Bradley et al., 1997; Lee et al., 1997; Souer et al., 1998; Samach et al., 1999; Chuck et al., 2002; Komatsu et al., 2003; Satoh-Nagasawa et al., 2006; Lippman et al., 2008; Rebocho et al., 2008; Souer et al., 2008; Li et al., 2009; Li et al., 2010), but we know little regarding the developmental and genetic changes underlying the evolution of these inflorescence types, let alone the heads and umbels for which the developmental bases are poorly understood. Although an elegant model has been proposed to explain inflorescence diversity (Prusinkiewicz et al., 2007), this model does not explain well the formation of umbels and heads. It has long been hypothesized that determinate umbels and heads evolved from branched inflorescences (e.g. panicles) by the suppression of inflorescence branches to form umbels and by suppression of pedicels in umbellate forms to produce heads (Parkin, 1914; Stebbins, 1974; Wyatt, 1982; Harris, 1999). Recently, Endress (2010) raised questions on this hypothesis based on the evidence that basal angiosperm taxa do not bear panicles, and suggested that detailed comparative inflorescence studies combined with phylogeny are needed to test this hypothesis and to better understand inflorescence evolution. To our knowledge, there have been no previous comparative studies combining the analysis of developmental morphologies and phylogeny to elucidate the evolutionary trend and the underlying developmental basis between elongated branched inflorescences and condensed forms such as umbels and heads in any angiosperm lineages.

In this paper, the genus Cornus (dogwood) was chosen as our model group to investigate the evolutionary direction and underlying developmental basis among elongated compound cymes (panicles, sensuEndress, 2010), umbels (i.e. sciadioid in Endress, 2010) and heads (i.e. cephalioid in Endress, 2010). Cornus contains c. 50 species that are divided into four major subgroups (Eyde, 1988; Murrell, 1993; Xiang et al., 1998, 2006; Fan & Xiang, 2001). The phylogenetic relationships of these four groups are well resolved by phylogenetic studies based on multiple gene sequences and morphological data (Fan & Xiang, 2001, 2003; Xiang et al., 2006, 2008; Fig. 1). These four groups are divergent in inflorescence architectures, exhibiting a range of branch condensation patterns as well as petaloid bracts in some species (Fig. 1). The first inflorescence type, found in the blue- or white-fruited (BW) lineage, is characterized by paniculate cymes with rudimentary bracts. The second inflorescence type found in the dwarf dogwood (DW) group is a condensed form of a dichasium with fewer and shorter branches, referred to as a ‘minidichasium’ hereafter. In the dwarf dogwood (DW) group the entire inflorescence is subtended by four large petaloid involucral bracts. The third inflorescence type is found in the cornelian cherry (CC) group bearing a determinate umbel (ciadioid) with four unexpanded, nonpetaloid, involucral bracts. The fourth inflorescence type found in Cornus is a completely condensed structure, a determinate head (cephaloid) with four or six large petaloid, involucral bracts that is found in the big-bracted (BB) group. In an earlier analysis of inflorescence evolution in Cornus based on outer morphology, umbels were inferred to be derived from paniculate cymes while heads were inferred to have evolved either directly from paniculate cymes or via umbels (Xiang & Thomas, 2008). These hypotheses could not be distinguished because of a lack of knowledge of the development morphologies.

Figure 1.

Four major clades of Cornus L. showing their phylogenetic relationships from Xiang et al. (2008) and the images of inflorescence types. Schematic diagrams of inflorescence structures are reproduced from Murrell (1993) with permission from American Society of Plant Taxonomists. BB, big-bracted; DW, dwarf dogwood; CC, cornelian cherry; BW, blue- or white-fruited.

In this paper, we conduct a comparative developmental morphological analysis of six Cornus species that represent the four major clades and cover the four major inflorescence types found in this genus. We characterized the developmental morphological progression for these six species via scanning electron microscopy (SEM) and histological analyses using materials collected over a 3-yr period. We describe a total of 24 developmental events, 13 correlate with alterations in inflorescence structure and size. We identify seven of these developmental events that are important to the development of umbels and heads. We trace these seven characters over the molecular phylogeny using methods that account for evolutionary time and phylogenetic uncertainty to uncover the evolutionary histories of umbels and heads. Our results suggest that heads and umbels in Cornus evolved independently from elongated forms via a small umbellate dichasium ancestor. Our results further reveal that the origins of umbels and heads each required multiple evolutionary changes that occurred independently in the CC and BB lineages in addition to the common changes that occurred in their common ancestor. These events include suppression of elongation of the rachis supporting the IM, suppression of inflorescence branch meristem (IBM) elongation as well as alterations in pedicel elongation. Additionally, changes in the pattern of primary IBM initiation and a reduction in the number of primary IBMs in early development support the formation of umbel and head forms, suggesting a much more complicated developmental evolutionary history than previously predicted from the analysis of the mature inflorescence morphology.

Materials and Methods

Sample selection and inflorescence bud collection

Six species spanning four groups of Cornus were chosen to cover the four different inflorescence forms in this genus: (Cornus controversa Hemsley, C. sanguinea L. and C. macrophylla Wall. from the BW group, producing paniculate cymes; C. officinallis Seib. & Zucc. from the CC group, producing umbels; C. florida L. from the BB group bearing heads (or cephaloid inflorescences); and C. canadensis L.f. from the DW group, bearing the minidichasium). Inflorescence bud samples of most of those species were collected every week after transiting to inflorescence growth (see the Results section) in 2007, 2008, 2009 and 2010 at J. C. Arboretum, Raleigh, NC, USA, except for C. florida and C. canadensis. The inflorescence buds of C. florida were collected from plants grown on the NC State University campus and those of C. canadensis were collected from plants grown in the phytotron of NC State University which were propagated from specimens collected in Cheshire, NH, USA. Two or more individuals were chosen for each species, if available, to cover potential variation within species. Usually 20–30 buds per individual were fixed in FAA (formaldehyde (37%) : acetic acid : 100% ethanol : ddH2O, 2 : 1 : 10 : 7, v/v) for histological study and c. 10–15 buds per individual in 3.0% glutaraldehyde in 25 mM potassium phosphate buffer (pH 7.0) at 4°C for at least 24 h for SEM study.

Observation of inflorescence developmental stages in Cornus

For SEM observation, preserved inflorescence buds were first washed in 25 mM potassium phosphate buffer (pH 7.0) at 4°C (three times at 1 h each), then dehydrated in 30, 50, 70 and 95% ethanol at 4°C (1 h each), followed by dehydration in 100% ethanol for 24 h at 4°C, which was repeated with two different temperature regimes: 4°C followed by room temperature and only at room temperature. Samples were then critical point dried using liquid carbon dioxide for 15 min using a Tousimis SAMDRI-795 (Tousimis Research Corp., Rockville, MD, USA) located in the Center for Electron Microscopy (CEM), NC State University. Buds were then dissected under the dissecting microscope, mounted on aluminum stubs with Pelco tape™ (Ted Pella, Inc., Redding, CA, USA) and finally sputter-coated with gold/palladium using a Hummer 6.2 Sputter System (Anatech, Union City, CA, USA) in the CEM. Prepared inflorescence buds were examined using a JEOL JSM-5900LV SEM (Jeol, Peabody, MA, USA) at 10, 15 or 20 kV. Images were captured digitally and colorized using GIMP (

The protocol for histological analysis followed Feng et al. (2009). For better penetration of dogwood inflorescence buds in FAA, hairy scales or hard bracts, which tightly protected the inflorescence buds, were removed and buds were vacuum-infiltrated for 30 min during fixation. After 24 h fixation, tissues were dehydrated and embedded in Paraplast® Plus (Paraplast Plus, Fisher Healthcare, Houston, Texas, USA). Sections were sectioned at 8 μm thickness using a rotary microtome and stained with toluidine blue (0.025%). Slides were observed under a Zeiss Axioscope2 microscope and images were captured using a Micropublisher 5.0 RTV digital camera and Q capture software (Q Imaging, Surrey, BC, Canada).

Analysis of sizes of IMs

Mature IMs were usually found to be dome-shaped or disk-like in Cornus species. The width and height of IMs were measured to compare the relative sizes of IMs in different species. The width of IM was defined as the distance between two attachment points of the two youngest leaves adjacent to the IM in the BW and DW groups in which bracts originated within the IM. In the CC and BB groups where bracts originated outside the IM, the width of the IM is the distance between the two attachment points of the second pair of the involucral bracts. The distance can be easily measured in the SEM images using ImageJ ( For histological images, only the medium longitudinal section through the pair of youngest leaves or bracts could be used. Height of the IM was defined as the height of the group of meristematic cells in histological sections in the inflorescence bud. The meristematic cells were identified based on their cellular morphology (smaller in size, more uniform in morphology and stain, typically rectangular with fewer and smaller vacuoles; see the Results section, Fig. 4). Pairwise comparison of the width and height of IMs among the four groups was performed using Student’s t-test in R-2.11.1 ( to determine whether there are significant differences in IMs among groups.

Tracing the evolutionary history of umbels and heads in Cornus

Seven developmental events (Table 1, characters in bold) were shown to differ between condensed and branched forms, suggesting that they are important to the formation of umbels and heads (Table 2). These characters were chosen for analysis of ancestral state reconstruction to identify the evolutionary developmental changes leading to heads and umbels. The characters were coded as binary data in the analysis. Character ancestral state reconstruction was performed using BayesTraits V1.0 (Pagel & Meade, 2006). The analysis used 1000 MrBayes trees from previous molecular phylogenetic study of Cornus (Xiang et al., 2008) to account for phylogenetic uncertainty. The phylogeny contains 22 species of Cornus distributed among the four major clades and two outgroup taxa, the sister genus of Cornus, Alangium and Diplopanax from Cornales. For species present in the phylogeny but not included in the development study, their character states were coded based on their adult inflorescence morphology in comparison to the morphology of the studied species, to determine the presence or absence of a developmental event. When character state was uncertain for a taxon, it was coded as missing state. For character state coding of Diplopanax, the inflorescence morphology of Curtisia and Grubbia, recently found to be the sister clade of Cornus-Alangium (Xiang et al., 2011), was considered. The structure of adult inflorescences in various species was determined from herbarium specimens or living collections in the J. C. Raulston Arboretum. Species of the genus that were not included in the phylogeny were all from the BW group, members of which have their inflorescence structure conserved as the elongated forms. A maximum likelihood (ML) analysis was first performed to obtain a sense of the average values of the rate parameters for the BayesTraits analysis. An interval from 0 to 15 was then set as the prior of the exponential distribution of the rate parameters. The values of 70 and 90 were chosen for the ratedev parameter to get an acceptance rate between 20 and 40% as suggested by the manual. A total of 10 000 000 iterations were run for the analysis of each character. The ancestral state with the highest value of the posterior possibility was reported for the node connecting the major clades. The density distributions of the posterior possibilities were depicted by R-2.11.1 ( For nodes where the posterior possibility of reported ancestral state was < 0.70, the fossil command was used to test whether there is positive support of the reported state over other states by calculating the Bayes factor as described in the manual (Pagel & Meade, 2006).

Table 1.   Comparison of developmental events (DEs) that differ among species of Cornus
DECharactersBW groupCC groupBB groupDW group
C. san*C. mac*C. con*C. off*C. flo*C. can*
  1. Ch.A–G indicates characters A–G analyzed by character mapping, corresponding to those in Fig. 14 and Supporting Information Fig.S1.

  2. *Abbreviation of species names: C. san, C. sanguine; C. mac, C. macrophylla; C. con, C. controversa; C. off, C. officinallis; C. flo, C. florida; C. can, C. canadensis. Characters were defined based on differences of DEs at five developmental stages. Characters in bold text were selected for analysis of ancestral character state reconstruction on a phylogenetic framework with character coding matrix shown in Table 2 and Fig. 14. Detailed information for stages and DEs are referred to in the legend of Fig. 2 and the Results section. IM, inflorescence meristem; IBM, inflorescence branch meristem; BB, big-bracted; DW, dwarf dogwood; CC, cornelian cherry; BW, blue- or white-fruited.

2–3Bract initiation patternAfter maturation of IMsAfter maturation of IMsAfter maturation of IMsBefore maturation of IMsBefore maturation of IMsAfter maturation of IMs
4Size of inflorescence meristems (Ch.A)LargeLargeLargeSmallSmallSmall
7Elongation of rachis supporting apical inflorescence meristem (Ch.B)VisibleVisibleVisibleInvisibleInvisibleInvisible
5–9Initiation pattern of 10IBMs (Ch.C)Decussately, sequentiallyDecussately, sequentiallyDecussately, sequentiallyWhorled, simultaneouslyDecussately, sequentiallyDecussately, sequentially
5–9Number of 10IBMs (Ch.D)8 in four pairs8 in four pairs5 or 610 in four pairs6 in three pairs4 in two pairs
10–12IBM elongation (Ch.E)VisibleVisibleVisibleInvisibleInvisibleVisible
10–11Highest order of IBMs in distal pair of 10 IBMs20 IBMs30 or 40 IBMs40 IBMs20 IBMs20 IBMs30 or 40 IBMs
12Highest order of IBMs in inflorescence40 IBMs50 or 60 IBMs60 IBMs20 IBMs20 or 30 IBMs30 or 40 IBMs
14Pedicel initiation during floral organogenesis (Ch.F)No initiationNo initiationNo initiationInitiationNo initiationInitiation
19Awn development on petalsNo awnNo awnNo awnNo awnNo awnAwn developed
15–20Timing pattern of floral organogenesisNot all before winterNot all before winterNot all before winterBefore winterBefore winterBefore winter
21Pedicel initiation during anthesis (Ch.G)InitiationInitiationInitiationNo initiationNo initiationNo initiation
23Bract petaloidyNot petaloidyNot petaloidyNot petaloidyNot petaloidyPetaloidyPetaloidy
Table 2.   Character state matrix of seven characters important to the development and evolution of umbels and heads in Cornus
  1. Characters and states (character number corresponding to Table 1):

  2. A, size of inflorescence meristems: 0, large; 1, small.

  3. B, elongation of rachis supporting apical inflorescence meristem: 0, visible; 1, invisible.

  4. C, initiation pattern of primary (10) inflorescence branch meristems: 0, decussately and sequentially; 1, close to whorled and simultaneously.

  5. D, number of 10 inflorescence branch meristems: 0, 8 or 10 in four pairs; 1, 5 or 6; 2, 6 in three pairs; 3, 4 in two pairs; 4, fewer than four.

  6. E, inflorescence branch meristem elongation: 0, visible; 1, invisible.

  7. F, pedicel initiation during floral organogenesis: 0, no initiation during floral organogenesis; 1, initiation during floral organogenesis.

  8. G, pedicel initiation during anthesis: 0, initiation during anthesis; 1, no initiation during anthesis.

C. nuttallii Audubon1102101
C. florida L.1102101
C. kousa Hance1102101
C. disciflora Moc. & Sesse ex DC.1102101
C. capitata Wall.1102101
C. oligophlebia Merr.0000000
C. hongkongensis Hemsley1102101
C. alternifolia L.f.0001000
C. controversa Hemsley0001000
C. peruviana J.F.Macbr.0000,2000
C. walteri Wangerin0000000
C. racemosa Lam.0000000
C. oblonga Wall.0000000
C. chinensis Wangerin1110111
C. sessilis Torr. Ex Durand1110111
C. eydeana QY Xiang & YM Shui1110111
C. mas L.1110111
C. officinallis Seib. & Zucc.1110111
C. volkensii Harms11100,10,10,1
C. suecica L.1103011
C. canadensis L.f.1103011
C. unalaschkensis Ledeb.1103011

The results from BayesTraits were further compared with analyses using two alternative methods available in Mesquite 2.01 (Maddison & Maddison, 2007) using the best phylogenetic tree. Each character was traced over the phylogram of the best phylogenetic tree using the ML method with the mk1 model (one-parameter Markov k-state model, a generalization of the Jukes–Cantor model), following Xiang & Thomas (2008). For characters that are polymorphic for some species – for example, Cornus peruviana of the BW group and C. volkensii of the CC group are polymorphic for one and three characters, respectively (Table 2) – the parsimony method was used to trace the character evolution because the ML method does not accept polymorphism.


Inflorescence development in Cornus

Inflorescence buds for the next year start to develop immediately after anthesis, or 1 or 2 months later depending on the species. We describe here a total of 24 developmental events (DEs) that characterize the inflorescence developmental pathway of the Cornus species we examined (Fig. 2). Some of the DEs are shared among all Cornus species we examined while others are lineage-specific (Fig. 2, Tables 1, 2). The developmental events are grouped here into five main developmental stages as described in the following.

Figure 2.

Summary of inflorescence developmental pathways for the six species of Cornus examined in the study. Developmental events (DEs, y-axis) are plotted against the time when the events occur. Stage I – transition of vegetative meristem to inflorescence meristem (IM): DE1, vegetative meristem at the end of blooming season; DE2, initiation of the first pair of bract primordia in the peripheral zone of the IM; DE3, initiation of the second pair of bract primordia in the peripheral zone of the IM; DE4, mature IM ready to generate primary (10) inflorescence branch meristems (IBMs). Stage II – initiation of primary (10) IBMs: DE5, initiation of the first group of 10 IBMs in the peripheral zone of the IM with or without bract subtending; DE6, initiation of the second group of 10 IBMs in the peripheral zone of the IM with or without bract subtending; DE7, elongation of rachis supporting the apical IM; DE8, initiation of the third group of 10 IBMs in the peripheral zone of the IM with or without bract subtending; DE9, initiation of the fourth group of 10 IBMs in the peripheral zone of the IM with or without bract subtending. Stage III – initiation of higher-order IBMs: DE10, initiation of secondary (20) IBMs in the peripheral zone of the 10 IBMs; DE11, initiation of tertiary (30) IBMs in the peripheral zone of the 20 IBMs; DE12, initiation of other higher-order IBMs in the peripheral zone of the 30 IBMs. Stage IV – floral organogenesis: DE13, initiation of floral meristems; DE14, initiation and elongation of pedicels during floral organogenesis; DE15, initiation of sepal primordia; DE16, initiation of petal primordia; DE17, petal primordia fully enclose floral bud; DE18, initiation of stamen primordia; DE19, development of awns on petals; DE20, initiation of gynoecium (carpel primordia). Stage V – inflorescence branch elongation and bract expansion: DE21, elongation of inflorescence branches; DE22, additional pedicel elongation during anthesis; DE23, expansion of bracts; DE24, floral bud open exposing mature stamen and release pollen.

Stage I: transition from vegetative meristem to IM – DE1 to DE4

The first stage in the inflorescence developmental pathway is the transition from a vegetative meristem (DE1) to an IM (DE4). During the vegetative to reproductive transition, the vegetative meristem enlarges and becomes more dome-shaped as it transitions into an IM that generates IBMs. In some, but not all, of the species we examined, the initiation of two pairs of bract primordia (DE2 and DE3) occurred during the vegetative to reproductive transition. The vegetative meristems of the Cornus species examined in this study are usually narrow and flat-topped (Figs 2, 3a,d,g,j,m,p, 4a,c,e,g). However, the morphology of the IMs varies among the different inflorescence types. They differ in size and in the organization of the meristematic cell layers (Figs 3–5).

Figure 3.

Stage I of the inflorescence developmental pathway: vegetative meristem to inflorescence meristem transition (DE1 to DE4). (a–c) Cornus sanguinea; (d–f) C. macrophylla; (g–i) C. controversa; (j–l) C. officinallis; (m–o) C. florida; (p–r) C. canadensis. Blue colored area indicates bracts. lf, leaf; SC, scale. For a description of DE1–DE4, see Fig. 2. Bars, 100 μm.

Figure 4.

Longitudinal sections of vegetative meristems and inflorescence meristems in four species of Cornus. (a) Vegetative meristem of C. sanguinea; (b) inflorescence meristem of C. sanguinea; (c) vegetative meristem of C. officinallis; (d) inflorescence meristem of C. officinallis; (e) vegetative meristem of C. florida; (f) inflorescence meristem of C. florida; (g) vegetative meristem of C. canadensis; (h) inflorescence meristem of C. canadensis. lf, leaf; br, bract; Bars: (a–f) 100 μm; (g) 200 μm; (h) 50 μm.

Figure 5.

Comparison of average width (closed bars) and height (open bars) of mature inflorescence among five species of Cornus. Significant difference: *, < 0.01. Error bars are ±SD.

In general, the IMs from species with paniculate cymes (C. macrophylla, C. sanguinea and C. controversa) are significantly larger than those in species from other three groups (Figs 3c,f,b, 5; < 0.01; Tables 1, 2). In addition, meristematic cells in these species are more densely packed and the IMs have more corpus or mantle layers (Fig. 4b) than the condensed forms in C. officinallis (umbels, Fig. 4d) and C. florida (heads, Fig. 4f).

Another difference noted between the different species during the vegetative to reproductive transition is whether bract primordia initiate during the transition. In C. florida and C. officinallis, two pairs of bract primordia arise during the vegetative–reproductive transition, so they form outside of (abaxial to) the IM (Fig. 3k,l,n,o; Tables 1, 2). By contrast, in the branched forms (C. canadensis and the BW species), pairs of bracts initiate within the IMs subtending IBMs and initiate after the vegetative–reproductive transition (Fig. 6a,e,i,u; Tables 1, 2).

Figure 6.

Stage II of the inflorescence developmental pathway: initiation of primary inflorescence branch meristem (10 IBM, DE5–DE9). (a–d) Cornus sanguinea; (e–h) C. macrophylla; (i–l) C. controversa; (m–p) C. officinallis; (q–t) C. florida; (u–x) C. canadensis; Blue, bracts; purple, first group of 10 IBMs; salmon, second group of 10 IBMs; orange, third group of 10 IBMs; yellow, fourth group of 10 IBMs; green, central floral meristem. lf, leaf; sc, scales; for a description of DE5–DE9, see Fig. 2. Bars, 100 μm.

Stage II: initiation of primary IBMs from IMs – DE5 to DE9

After the transition from vegetative to reproductive development, the IMs begin to initiate pairs of primary IBMs (10 IBMs). In this paper, we define a 10 IBM as any lateral branch meristem that arises from the periphery of the apical IM. The 10 IBMs have the ability to generate additional lateral meristems: either higher-order IBMs (i.e. secondary, tertiary, etc.) or floral meristems. In species that display paniculate cymes (C. sanguinea and C. macrophylla), the apical IMs give rise to four pairs of 10 IBMs (Figs 2, 6; DE 5, 6, 8, and 9; each pair was color-coded as purple, salmon, orange, and yellow, respectively). These four pairs of 10 IBMs arise sequentially at the periphery of the meristematic dome in a decussate fashion. Concomitantly there is an apically directed elongation of the rachis supporting the IM (Figs 2, 6, 7; DE 7). This initiation pattern of 10 IBMs matches the decussate phyllotaxy of leaves that occurs almost exclusively in the genus (two sister species C. controversa Hemsley and C. alternifolia L.f. are exceptions). In C. controversa, the initiation pattern of 10 IBMs is spiral and matches the spiral leaf phyllotaxy observed in this species. In C. controversa the first three groups of 10 IBMs arise singularly and spirally from the periphery of the relatively broad apical IM (Fig. 6i,j,k). The rachis subtending the apical IM continues to elongate and the apical meristem generates another two branch meristems to give rise to a total of five 10 IBMs (Fig. 6l).

Figure 7.

Longitudinal sections of young Cornus inflorescences. (a–c) Longitudinal section of C. macrophylla inflorescences, showing vertical elongation of central rachis supporting the inflorescence meristem (IM) and branch elongation in branches supporting inflorescence branch meristems (IBMs); (d) longitudinal section of one developing flower of C. macrophylla, showing no pedicel development during floral organogenesis; (e, f) longitudinal section of C. officinallis young inflorescences, showing no elongation of rachis supporting the central IM or branches supporting IBMs; (e) pedicel development in C. officinallis during floral organogenesis (f); (g, h) longitudinal section of C. canadensis young inflorescences, showing slight elongation of branches supporting IBMs and pedicel development during floral organogenesis; (i–l) longitudinal section of C. florida young inflorescences, showing no elongation of central rachis supporting the IM and branches supporting the IBMs and no pedicel development during floral organogenesis. lf, leaf; br, bract; sc, scales; pd, pedicel; se, sepal; pe, petal; st, stamen. Bars: (a–k) 100 μm; (l) 50 μm.

The apical IMs in C. florida and C. canadensis generate 10 IBMs with a similar pattern to those observed in C. sanguinea and C. macrophylla (decussate), but with a reduced number of 10 IBMs. In C. florida, three pairs were generated (Fig. 6s,t), and in C. canadensis only two pairs were produced (Fig. 6v,w,x). No elongation of the rachis supporting the IM was observed in these two species (Figs 6q–v, 7g–j).

The species that produces the umbel inflorescence (C. officinallis) shows a different initiation pattern of 10 IBMs. Unlike the sequential initiation pattern in other species, the first three groups of 10 IBMs (containing four, two and two IBMs per group, respectively) initiate almost simultaneously on the surface of the wide apical IM (Fig. 6m,n). Thus what may be a very compressed decussate initiation pattern of 10 IBM appears somewhat like a whorled pattern of initiation. Following the initiation of these eight 10 IBMs, another pair of 10 IBMs was formed from the center of apical IM (Fig. 6o,p) resulting in a total of 10 10 IBMs. Sometimes the second group 10 IBMs were observed to split into two IBMs and the fourth group of 10 IBMs could be aborted (Fig. 8).

Figure 8.

Variable splitting of primary inflorescence branch meristems (10 IBMs) in Cornus officinallis. Dotted line (lower) shows the splitting of one 10 IBM into two 10 IBMs in C. officinallis. br, bract. Bar, 100 μm.

Stage III: initiation of higher-order IBMs – DE10 to DE13

Each pair of 10 IBMs can give rise to a pair of secondary IBMs (20 IBMs, DE10) that initiate from peripheral portions of the primary IBMs. Secondary IBMs can lose IBM identity and differentiate into a floral meristem as observed in C. officinallis (Fig. 9j–l), for the third group of 10 IBMs of C. florida (Fig. 9n,o) and the most distal pair of 10 IBMs in C. sanguinea (Fig. 9a, c). Alternatively, the 20 IBMs can generate tertiary IBMs (30 IBMs, DE11). These 30 IBMs terminate as floral meristems in C. canadensis (Fig. 9q) and in most distal pair of primary branch meristems in C. macrophylla (Fig. 9e), or they can generate quaternary, quinary (fifth-order) or senary (sixth-order) IBMs (DE12) as in C. macrophylla and C. controversa. Senary IBMs are the highest order observed in the Cornus species we examined. All these high-order IBMs are arranged in dichasium structures. They initiate in a basipetal manner, starting from the most basal primary IBMs and continuing to more distal ones. The 10 IBMs in more distal positions produce fewer high-order IBMs. The IBMs in the BW species are elongated while only a slight elongation of IBMs was observed in C. canadensis (Fig. 7c,h), and no elongation of IBMs was observed in C. officinallis and C. florida (Fig. 7e,f,i,j).

Figure 9.

Stage III of the inflorescence developmental pathway: initiation of higher-order inflorescence branch meristems (IBMs) (DE10–DE12). (a–c) Cornus sanguinea; (d–f) C. macrophylla; (g–i) C. controversa; (j–l) C. officinallis; (m–o) C. florida; (p–r) C. canadensis. Blue, bracts; purple, first group of primary (10) IBMs or flowers developed from the first group of 10 IBMs; salmon, second group of 10 IBMs or flowers developed from the second group of 10 IBMs; orange, third group of 10 IBMs or flowers developed from the third group of 10 IBMs; yellow, fourth group of 10 IBMs or flowers developed from the fourth group of 10 IBMs; green, central floral meristem or central flower; lf, leaf; pd, pedicel. For a description of DE10–DE12, see Fig. 2. Bars, 100 μm.

Stage IV: floral organogenesis – DE14 to DE20

The central apical IM is the first to transition to a floral meristem in all species examined, as demonstrated by the initiation of sepal primordia (Fig. 9, DE 13). Flowers in the Cornus typically have tetramerous sepals, petals and stamens, two fused carpels and an inferior ovary (Fig. 10). Floral development is similar among different Cornus species. In all six species examined, four sepal primordia are the first floral organs to initiate at the margin of the floral meristem (DE15; Fig. 11a). Four petal primordia subsequently initiate adaxial to the sepal primordia to form the second floral whorl alternate with the sepal primordia (DE16; Fig. 11a). The petal primordia grow longer than sepals and grow to enclose the developing stamens and carpels (DE17; Fig. 11d). Stamen primordia initiate in the third floral whorl (DE18; Fig. 11e–g) followed by two carpel primordia from the remaining floral meristem apex (DE20; Fig. 11j).

Figure 10.

Mature flowers in bloom from the six species of Cornus studied. (a) C. macrophylla; (b) C. sanguinea; (c) C. controversa; (d) C. florida; (e) C. officinallis; (f) C. canadensis.

Figure 11.

Stage IV of the inflorescence developmental pathway: floral organogenesis (DE13–DE20), exemplified by Cornus canadensis. (a) Initiation of sepals and petals; (b, c) growth of sepals and petals; (d) petals covering the central meristem; (e) initiation of stamens; (f, g) growth of stamens; (h) initiation of awns on petals; (i) elongation of awns; (j) initiation of carpels; (k) mature inflorescence buds (two branches were taken off). For descriptions of DE13–DE20, see Fig. 2. br, bract; se, sepal; pe, petal; st, stamen; aw, awn; cp, carpel. Bars: (a–c, e) 10 μm; (d, f–k) 100 μm.

There are major differences of floral development among the Cornus species regarding the timing of floral organ initiation with respect to the calendar year (Fig. 2) and in the development of pedicels (DE14; Fig. 7d,f–h,k,l). In three species with condensed inflorescences, C. florida, C. officinallis and C. canadensis, floral organogenesis starts in the summer and all floral organs are initiated and well developed before winter. In two species with elongated inflorescences, C. sanguinea and C. controversa, floral organogenesis begins in the summer, but carpels do not develop until the following spring. In the third species with elongated inflorescences, C. macrophylla, which blooms later than C. sanguinea and C. controversa, floral organ organogenesis initiates in spring (Fig. 2).

Furthermore, pedicel development/elongation in C. officinallis (umbels) and C. canadensis (minidichasium) occurs during floral organogenesis (in the fall) (DE14; Figs 2, 7f–h). However, in species with elongated inflorescences, pedicel development/elongation occurs in the blooming season of the following spring, well after all floral organs are initiated (Fig. 12a). In C. florida (heads), flowers are sessile and pedicel development/elongation was not observed. In C. canadensis, awns (a unique feature of the DW group) develop on petals during this stage (DE19; Fig. 11h,i).

Figure 12.

Stage V of the inflorescence developmental pathway: branch elongation and bract expansion (DE 21–DE24). (a) Branch and pedicel elongation in Cornus controversa; (b) pedicel elongation in C. officinallis; (c) bract expansion and petaloidy in C. florida; (d) bract expansion and petaloidy in C. canadensis. For descriptions of DE21–DE24, see Fig. 2.

Stage V: brach elongation and bract expansion – DE21 to DE24

Species of the BW groups (C. sanguinea, C. macrophylla and C. controversa in this study) have a rapid increase in the length of branches and pedicels in the spring (DE21, 22; Figs 2, 12a), resulting in an expanded compound inflorescence. Longitudinal sections from developing branches at different stages indicated that the elongation of branches is supported by an increase of cell numbers in the pith and an increase in cell expansion in the cortex (Fig. 13). A minor elongation of the primary inflorescence branches and pedicels occurs in C. canadensis, but not in C. florida, in the spring before flowering. In C. officinallis, only pedicels elongate; branches remain rudimentary (DE22; Fig. 12b). In C. canadensis, C. florida and C. officinallis, involucral bracts expanded, but became petaloid only in the first two species (Fig. 2, DE23; Fig. 12c,d).

Figure 13.

Longitudinal section of inflorescence branches during branch elongation in Cornus macrophylla. (a) Inflorescence branches about to elongate; (b) elongating inflorescence branches; (c) elongated inflorescence branches. co, corpus; pi, pith; Bars, 100 μm.

Summary of inflorescence development pathways in Cornus

We identified 13 developmental events that differ among the species representing the four major inflorescence types we studied (Table 1; Fig. 2). The umbels and heads showed a combination of two unique developmental differences, respectively, while the minidichasium in C. canadensis display two unique differences and the elongated compound inflorescences in the BW groups showed four unique differences (Table 1; Fig. 2).

In addition, the relationship of the developmental events to the calendar year also varies among these species, especially for those in stages I, IV and V.

Reconstructing the developmental evolutionary history of umbels and heads in Cornus

The character matrix for the seven characters that are likely involved in the origin of umbels and heads is shown in Table 2. Results from ancestral character state reconstruction using BayesTraits were largely congruent with those inferred from ML and maximum parsimony (MP) methods (Fig. 14a–d; Supporting Information, Fig. S1), except for character A at node B (Fig. S1a). For this node, the BayesTraits suggested nearly equal possibility of state 0 and state 1, while ML prefers state 1. When comparing results between BayesTraits, ML, and MP methods, we found that ancestral character states with uncertainty were usually better resolved in BayesTraits (node C in Fig. S1a; nodes A, B, D in Fig. S1b; node A in Fig. 14a; nodes D, F in Fig. 14b; nodes D, F in Fig. 14c). Results from BayesTraits based on 1000 phylogenetic trees are described in the following.

Figure 14.

Evolutionary trends of four of the seven developmental characters (characters D–G) important to the development of umbels and heads in Cornus. Results for the first three characters (A–C) are presented in Supporting Information, Fig. S1. Character states shown in ball patterns (or colors) were derived from maximum parsimony analyses implemented in Mesquite 2.01 using the best phylogenetic tree (Diplopanax was removed manually from the outgroup to save space). Maximum likelihood analyses of these characters were not possible because of the character state polymorphism in some taxa. Character state and its posterior probability derived from analysis using a Bayesian method implemented in BayesTraits 1.0 using 1000 phylogenetic trees were shown at nodes. The distribution densities of posterior probabilities of character states estimated for the nodes of interest by BayesTraits are shown in the panels. For ancestral states with posterior probabilities < 0.7, a hypothesis test using Bayes Factors was performed to determine whether there is support for one state over the other state. *, Bayes factor is > 2, positive support for one over the other state; **, Bayes factor is > 5, strongly support one character state over the other; (a) character D; (b) character E; (c) character F; (d) character G; details of characters and states are provided in Tables 1, 2. PP, posterior probability.

Character A – size of IMs  The ancestral state of the crown group of Cornus (Fig. S1a, node B) was not clear, but state 1 (small IM 1) is strongly supported for the most recent common ancestor of BB, DW and CC groups (node D, PP (posterior probability) = 0.99 ± 0.03), and state 0 (large IM) for the ancestor of BW group (node C, PP = 1.00 ± 0.01). The results supported divergence of the IM size between the ancestor of the BB, DW, and CC lineage and the ancestor of the BW lineage.

Character B – elongation of rachis supporting IMs  Ancestral states at two deepest nodes (A and B in Fig. S1b) were resolved as state 0 (visible elongation) with high support (PP = 0.94 ± 0.10 and 0.81 ± 0.28, respectively). A switch to state 1 (invisible elongation) was highly supported at node D (common ancestor of BB-CC-DW groups, PP = 0.99 ± 0.04).

Character C – initiation pattern of 10 IBMs  The ancestral states of the root of Cornus (node B) and the most recent common ancestor of the BB-DW-CC clade (node D) were strongly supported to be 0 (PP = 0.99 ± 0.02 and 0.80 ± 0.23, respectively, Fig. S1c). There was a single evolutionary shift to state 1 (whorled and simultaneous) at the crown node of the CC lineage (node E, PP =1.00 ± 0.00).

Character D – number of 10 IBMs  State 0 (four pairs) was reconstructed for node B (crown of Cornus), node C (crown of the BW group) and node E (crown of the CC group) with strong support (PP = 0.95 ± 0.10, 0.93 ± 0.09 and 0.99 ± 0.01, respectively) and for node D (crown of BB-DW-BW clade) with relatively low, but significant support (PP = 0.60 ± 0.21) (Fig. 14a). There was a change in the number of 1o IBMs at the node L (uniting C. controversa and C. alternifolia) within the BW group, from eight in four pairs to five or six 10 IBMs (state 1). Evolutionary reduction in numbers occurred in the BB-DW clade. One scenario suggested a reduction from four pairs to three pairs (state 2) at the crown of the BB-DW clade (node F; PP: 0.48 ± 0.21), followed by a further reduction from three pairs to two pairs (state 3) at the crown node of DW (node G). The alternative scenario suggested a reduction from four pairs to two pairs at the node F followed by an increase from two pairs to three pairs at the crown node of the BB clade (node F; PP = 0.26 ± 0.18). The latter was supported by lower posterior probability, but the hypothesis test using Bayes Factor did not show a significant difference between these two scenarios.

Character E – IBM elongation  The ancestral states at nodes A and B were strongly supported as state 0 with elongated IBMs (PP = 0.94 ± 0.08 and 0.85 ± 0.17, respectively, Fig. 14b). The ancestors of the BB-DW-CC and BB-DW clades were also estimated to have elongated IBMs although with relatively low (but significant) supports (PP = 0.55 ± 0.48* and 0.70 ± 0.32, respectively). A switch from elongated IBM (state 0) to nonelongated IBM (state 1) occurred independently in the ancestor of the CC (node E) and BB (node H) clades.

Character F – pedicel initiation during floral organogenesis  The ancestral states were resolved as state 0 (no initiation during floral organogenesis) for all nodes with strong support, except for the CC and DW clades (Fig. 14c). The transition from state 0 to state 1 (initiation during floral organogenesis) was shown to have occurred independently in the CC and DW clades with high support (Fig. 14c, nodes E and G).

Character G – pedicel initiation during anthesis  Pedicel initiation during anthesis (state 0) was resolved for the crown nodes of Cornus and the BW group with strong support (Fig. 14d, PP = 0.81 ± 0.28 and 1.00 ± 0.00, respectively). Loss of pedicel initiation during anthesis (state 1) occurred in the common ancestor of the BB-DW-CC clade (node D, PP = 0.99 ± 0.04).

These evolutionary changes are summarized in Fig. 15. Based on the evolutionary changes of the branches, the ancestral inflorescence type reconstructed for the BB-DW-CC clade was a small umbellate dichasium based on the following character states: small IM size (character A), loss of rachis elongation (character B), decussate and sequential initiation of 10 IBMs (character C), four pairs of 10 IBMs (character D), visible elongation of IBMs (character E), no initiation of pedicels during floral organogenesis (character F), and no pedicel initiation at the anthesis (character G).

Figure 15.

Summary of inferred evolutionary developmental changes responsible for the origins of umbels and heads in Cornus. Ticks on branches indicate the evolutionary events (character A was not shown as the ancestral state, as the crown of Cornus was unclear). Numbers in parenthesis indicate character states. Schematic diagrams of inflorescence structures are reproduced from Murrell (1993, 1996) with permission from American Society of Plant Taxonomists. Detailed information about schematic diagrams is given in Fig. 1. BB, big-bracted; DW, dwarf dogwood; CC, cornelian cherry; BW, blue- or white-fruited.


Temporal divergence of inflorescence developmental stages among Cornus species – new insights for BW and DW species

Our comparative developmental study revealed that all inflorescence types in Cornus initiate their development soon after the anthesis and complete the architecture formation in the autumn. This finding is new to the DW and BW lineages that produce elongated, branched inflorescences (minidichasium and paniculate cymes, respectively). The inflorescences in these two lineages had been, for a long time, believed to develop in the spring of the year of blooming, because their floral buds look similar to the leaf buds until they expand in the spring, after leaves have fully developed. Only the umbels and heads in the CC and BB lineages were considered to be preformed in the autumn of the previous year because they are clearly visible as head-like buds in the autumn, different from leaf bud morphology (Murrell, 1993; Xiang et al., 2006). This temporal similarity in a broader scale among the inflorescence types indicates that the development of the branched inflorescence types in DW and BW clades is not as delayed as it appears to be based on the outer morphology.

We have also characterized at a finer scale the temporal divergence with respect to the calendar year of the developmental events among the species (Table 1; Fig. 2). Evolutionary shifts towards earlier inflorescence development and blooming occurred in the umbel (C. officinallis) and head (C. florida) lineage (Figs 2, 16). The link between the temporal shift in flowering and the origins of umbels and heads is unclear. It was proposed that inflorescence architectures affect pollinator behaviors and influence pollination efficiency (Wyatt, 1982). Observation in nature and experimental studies on artificial inflorescence architectures have shown that the flat and closely packed structures of umbels and heads increased the number of flowers visited by insects or birds per inflorescence as a result of the decrease in foraging cost in these flattened or planar inflorescence structures (Hainsworth et al., 1983; Cresswell, 1990; Jordan & Harder, 2006). By increasing floral visitation, umbels and heads could ensure a certain rate of pollination in early spring when the number of pollinators is limited. Therefore, selection may favor condensed inflorescence structures for early flowering plants or for plants occurring in places with scarce pollinators (e.g. in alpine habitat, as for the DW species). The evolution of condensed inflorescence structures in Cornus (e.g. umbels, heads, and minidichasium) could thus be the results of adaptation to early flowering or scarce pollinators. Inflorescence structures were also found to be correlated with types of pollinators in Arecaceae (Henderson, 2002). In this family, the condensed inflorescences were reported to be pollinated by beetles while the elongated inflorescences were found to be pollinated by bees, flies or wasps (Henderson, 2002). The dogwood species were reported to be pollinated by a wide range of insects (Eyde, 1988; Rhoades, 2010). Species producing condensed inflorescences (BB, DW, and CC groups) are pollinated by bees, flies, and beetles (Douglas, 1983; Barrett & Helenurm, 1987; Eyde, 1988; Rhoades, 2010), while species producing elongated, branched inflorescences (the BW group) were reported to be pollinated by flies and beetles (Robertson, 1928; Parmenter, 1956; Waldbauer, 1983). The available evidence is unclear regarding whether the evolution of condensed inflorescence structures in Cornus is linked to pollinator preference.

Figure 16.

Evolutionary trends of flowering time in Cornus inferred from analyses using BayesTraits 1.0 and Mesquite 2.01. Flowering time of species was determined based on herbarium specimen records, ‘A California Flora’ (Munz & Keck, 1959), ‘Manual of Vascular Plants of Northeastern United States and Adjacent Canada’ (Gleason & Cronquist, 1991), ‘Floral of China’ (Xiang & Boufford, 2005), and several other articles (Murrell, 1996; Xiang et al., 2003). Information on analyses and symbols are given in Fig. 14. PP, posterior probability.

Developmental evolutionary history of umbels and heads in Cornus

By tracking the evolutionary histories of developmental characters on the Cornus phylogeny, we have begun to illuminate the developmental basis of the origins of umbels and heads in the genus. Our results support the idea that umbels and heads were derived from the ancestral umbellate dichasium independently, each transition involving multiple developmental changes at different evolutionary times (Fig. 15), a much more complicated evolutionary history than can be deciphered from analysis of the mature external morphology (Xiang & Thomas, 2008). The initial developmental divergences that led to umbel and head forms occurred very early in the evolutionary history of the genus, before the BB-DW split from CC, dating back to the very early Tertiary based on molecular dating and fossil evidence (Xiang et al., 2008). The origin of umbels required at least five evolutionary changes. Two occurred earlier in the common ancestor of the BB-DW-CC clade: suppression of rachis elongation and loss of pedicel initiation/elongation at anthesis. Three changes occurred later in the CC lineage after it diverged from the BB-DW clade: a change from a decussate to a ‘whorled’ pattern of primary IM initiation, loss of IBM elongation, and initiation and elongation of pedicels during floral organ initiation (Fig. 15). It should be noted here that the ‘whorled’ pattern of primary IBM initiation could also be interpreted as a severe compression of the rachis elongation and reduction of the time between the initiations of successive IBM pairs. The origin of heads involved four evolutionary changes from the common ancestor of Cornus. The first two changes were common to those for umbels and the latter two changes occurred independently, one in the ancestor of BB-DW (reduction in the number of primary IBMs) and one in the BB clade (loss of IBM elongation, Fig. 15). These data suggested that the ancestor of the umbel and head forms in Cornus was a small eight- or seven-branched umbellate dichasium. This type of inflorescence is still observed in some extant species of Cornus, such as the male inflorescence of C. volkensii, a member of the CC group from tropical Africa, and C. peruviana in the BW group (Murrell, 1996). Our data also indicate that evolution of heads and umbels in Cornus did not occur through simple suppression of branches and pedicels as previously believed, but involved other important evolutionary changes that would not be apparent without a comparative analysis of the progression of the developmental morphologies. These new insights deepen our understanding of inflorescence evolution in Cornus. The evolutionary pathways of umbels and heads revealed in the study differ from, and are clearer than, those inferred from mature external morphology alone in Cornus (Xiang & Thomas, 2008).

Our study on Cornus is consistent with the hypothesis that inflorescence architectures evolved from elongated to condensed forms in angiosperms more generally (Parkin, 1914; Stebbins, 1974; Wyatt, 1982; Harris, 1999). The data from Cornus, however, do not support the hypothesis that umbels evolved from elongated inflorescences via simple suppression of inflorescence branches and heads evolved from umbels. It is as yet unclear if the developmental pathways of umbels and heads found in Cornus are conserved in other plant lineages that also bear determinate umbels and heads. Comparative studies of developmental pathways of inflorescence architectures in other plant lineages would be helpful for a better understanding of the evolution of umbels and heads in a variety of Angiosperm species.

By contrast, developmental studies have been conducted for indeterminate umbels and heads in a few plants, for example, Lotus and Coronilla in Fabaceae (Dong et al., 2005; Sokoloff et al., 2007) and Helianthus annuus in Asteraceae (Marc & Palmer, 1981). These studies found that, similar to the umbels and heads in Cornus, the IMs in these taxa also had a flat structure. The major developmental difference between these taxa and Cornus, however, is the lack of IBMs in these taxa. Initiation of IBMs was not observed in these taxa. Only floral meristem initiation was observed, which occurred in whorls directly from the expanded IM. This process was in contrast to the development of umbels and heads in Cornus, which formed IBMs first and subsequently suppressed their elongation. This suggests that the genetic controls in indeterminate heads and umbels are likely different from those in determinate heads and umbels. It further suggests that potentially fundamental differences exist in the formation of determinate and indeterminate umbels and heads.

Future work

Our phylogeny-based comparative developmental study clarifies the complex developmental pathways of the four inflorescence types in Cornus and identifies seven important characters differentiating the development of umbels and heads (Figs 14, 15, S1). These data suggest that the study of the genetic control of IM size, rachis elongation, IBM initiation and elongation, timing of pedicel formation and the transition of IM to floral meristem will be critical to understanding the evolution of umbels and head development in Cornus. We are currently working to establish a genetic transformation system for Cornus canadensis employing callus-regenerated plantlets (Feng et al., 2009), which will allow us to test the function of candidate genes involved in these developmental processes and to clarify the molecular genetic basis for the evolution of these developmental differences.

In conclusion, our results revealed a more complex developmental basis underlying the evolution of umbels and heads in Cornus than is predicted from external morphology. Our findings shed novel insights into head and umbel evolution in the genus. Our work illustrates the importance of combining developmental and phylogenetic analyses to better understand morphological evolution.


The authors thank the J. C. Raulston Arboretum for providing living plants for experiments, the phytotron at North Carolina State University for caring cultured plants, D. E. Boufford for assistance in collecting C. canadensis from the field, M. Mackenzier Jr and V. Knowlton at NCSU Electron Microscopy Center for technical assistance with the SEM analysis, and the herbaria at NCSU and UNC, Chapel Hill for use of the Cornus collections. We acknowledge the travel awards from the Developmental & Structural Section of Botanical Society of America to C-M.F. The research was supported by a NCSU Multi-disciplinary Research grant and a NSF grant (IOS-1024629).