Evolution of bract development and B-class MADS box gene expression in petaloid bracts of Cornus s. l. (Cornaceae)


Authors for correspondence:
Qiu-Yun (Jenny) Xiang
Tel: +1 919 515 2728
Email: jenny_xiang@ncsu.edu
Robert G. Franks
Tel: +1 919 513 7705
Email: rgfranks@ncsu.edu


  • Despite increasing interest in the molecular mechanisms of floral diversity, few studies have investigated the developmental and genetic bases of petaloid bracts. This study examined morphological patterns of bract initiation and expression patterns of B-class MADS-box genes in bracts of several Cornus species. We suggest that petaloid bracts in this genus may not share a single evolutionary origin.
  • Developmental pathways of bracts and spatiotemporal expression of B-class genes in bracts and flowers were examined for four closely related dogwood species.
  • Divergent morphological progressions and gene expression patterns were found in the two sister lineages with petaloid bracts, represented by Cornus florida and Cornus canadensis. Phylogeny-based analysis identified developmental and gene expression changes that are correlated with the evolution of petaloid bracts in C. florida and C. canadensis.
  • Our data support the existence of independent evolutionary origins of petaloid bracts in C. canadensis and C. florida. Additionally, we suggest that functional transference within B-class gene families may have contributed to the origin of bract petaloidy in C. florida. However, the underlying mechanisms of petaloid bract development likely differ between C. florida and C. canadensis. In the future this hypothesis can be tested by functional analyses of Cornus B-class genes.


The petals of a flower can function in the attraction of pollinators and have significant ecological importance. Petals are generally larger than other floral organs, are bright in color and have papillate epidermal cells on the adaxial surface (Weberling, 1989; Irish, 2009). In Arabidopsis thaliana and its immediate relatives, these petal characteristics result from the specification of petal identity in the developing petal primordia. This petal identity specification requires the combined activities of related MADS-domain containing proteins (Coen & Meyerowitz, 1991; Pelaz et al., 2000; Honma & Goto, 2001; Theissen & Saedler, 2001). Based on studies in Antirrhinum majus and Arabidopsis thaliana, the genes required for petal identity specification have been placed into three functional classes (A-class, B-class and E-class). Studies on other flowering plants, however, indicate that the requirement of A-, B- and E-class gene functions for petal development is not conserved in all flowering plants and that the A-function hypothesized by the model is restricted to A. thaliana and its immediate relatives (Theissen et al., 1996; Munster et al., 2001; Theissen & Saedler, 2001; Stellari et al., 2004; Litt & Kramer, 2010; Rijpkema et al., 2010; Sablowski, 2010).

Heterotopic petaloidy (the occurrence of petaloid characteristics in nonpetal organs) has evolved many times during angiosperm diversification and contributes to the diversity of floral forms (Endress, 1994; Albert et al., 1998; Baum & Donoghue, 2002). Several authors have proposed that orthologous B-class petal identity genes play an important role in heterotopic petaloidy in a diversity of angiosperm species (van Tunen et al., 1993; Bowman, 1997). In this scenario, heterotopic petaloidy results from a shift of B-class gene expression to adjacent whorls (e.g. an outer shift to the first whorl to establish petal identity in sepals and a centripetal shift to the third whorl to establish petaloid stamens). This hypothesis is supported by studies on petaloid tepals of several monocot plants, for example, Tulipa in Liliaceae (van Tunen et al., 1993; Kanno et al., 2003) and Phalaenopsis in Orchidaceae (Tsai et al., 2004; Mondragon-Palomino & Theissen, 2008, 2009), as well as in a basal angiosperm Aristolochia (Jaramillo & Kramer, 2004). However, evidence from the studies of Aristolochia (Aristolochiaceae) and Aquilegia (Ranunculaceae) suggest that the function of the B-class gene homologs in heterotopic petaloidy in sepals is only required at a late developmental stage (i.e. during cell differentiation) rather than in the early stage during which organ identity is specified (Jaramillo & Kramer, 2004; reviewed in Kramer & Hodges, 2010). Furthermore, studies in Clematis (Ranunculaceae) and in Phalaenopsis (Orchidaceae) suggested that subfunctionalization following duplication of B-class genes explained heterotopic petaloidy in sepals (Kramer et al., 2003; Mondragon-Palomino & Theissen, 2008, 2009). On the other hand, a study in Asparagus found no evidence for the involvement of orthologs of Arabidopsis B-class genes in the petaloid sepals (Park et al., 2003, 2004). Thus, diverse molecular mechanisms may support heterotopic petaloidy within flowering plants.

At present, the role of B-class gene orthologs and paralogs in heterotopic petaloidy has been reported mostly for petaloid sepals and petaloid stamens. Few studies have investigated the roles of B-class gene orthologs in extrafloral petaloidy (e.g. petaloid bracts). A short list of lineages with petaloid bracts includes Saururaceae (Piperales), Nyctaginaceae and Amaranthaceae (Caryophyllales), Cornaceae and Davidiaceae (Cornales), Rubiaceae (Gentianales), Araceae (Alistmatales), Heliconiaceae (Zingerberales) and Euphorbiaceae (Malpighinales). Evidence of orthologs of Arabidopsis B-class gene expression in petaloid involucral bracts (bracts that appear in a whorl subtending an inflorescence) was recently reported by Vekemans et al. (2011) in the dove tree, Davidia involucrate (Cornales). These authors detected expression of two B-class genes and one C-class gene during an early developmental stage in bract primordia (Vekemans et al., 2011). The expression patterns of orthologs of the Arabidopsis B-class genes in petaloid bracts of other species remain uncharacterized.

An important goal of evolutionary developmental biology is to identify the molecular and developmental changes that occurred during cladogenesis that gave rise to descendant lineages differing in morphology (Kellogg, 2004). To our knowledge, there has been no analysis employing a phylogenetic approach to uncover the changes leading to the divergence between petaloid and nonpetaloid bracts among closely related species. The dogwood genus Cornus (Cornaceae, Cornales) has four closely related clades that exhibit variation in bract morphology, and thus provides a good model for understanding the molecular and developmental mechanisms of petaloid bract evolution. In Cornus, bract petaloidy evolved in two lineages that are sisters to each other, the dwarf dogwoods (DW group, e.g. C. canadensis) and the big-bracted dogwoods (BB group, e.g. C. florida). Phylogenetic studies have indicated that the petaloid clade is sister to the cornelian cherries group (CC group, e.g. C. officinalis) bearing nonpetaloid, involucral bracts. This DW-BB-CC clade is, in turn, sister to a large clade (BW group, bearing blue- or white-fruits; e.g. C. macrophylla, C. sericea; Xiang et al., 2006, 2008, 2011). Species within this BW group produce noninvolucral bracts that subtend inflorescence branches and are rudimentary and deciduous. The sister relationship of the DW and the BB in the phylogeny has led to the reconstruction of a common origin of the petaloid bracts in their common ancestor (Xiang & Thomas, 2008). Although reconstructing ancestral character states based on phylogenies can be a powerful tool to elucidate morphological evolution, an accurate understanding of morphological characters and their evolution requires data from developmental and genetic analyses. Differences in external morphology may be the result of multiple independent alterations of the developmental program that first occurred at different evolutionary times. On the other hand, even morphological similarities between closely related species may be the result of disparate developmental mechanisms that evolved independently.

In order to understand the origin (or origins) of petaloid bracts in the genus Cornus, we conducted a comparative developmental study in four species representing the four clades of this genus. We also examined the expression patterns of orthologs of Arabidopsis B-class MADS-box genes in these species to determine if evolutionary changes in their expression patterns are correlated with the origin of petaloid bracts in the genus. Furthermore, we examined the expression patterns of these genes in developing floral organs to assess if they might function in specifying during petal and stamen identity within the flower as reported in diverse flowering plant lineages.

Materials and Methods

Morphological analysis of bract development

Four species with divergent bract morphologies representing the four major clades of Cornus were chosen for analysis. For C. florida L., C. officinalis Seib. & Zucc. and C. macrophylla Wall. samples spanning the range of developmental stages were collected from plants grown on North Caroline State University (NCSU) campus or at the J. C. Raulston Arboretum (Raleigh, NC, USA). Samples of C. canadensis L.f. inflorescences were obtained from plants grown in the NCSU Phytotron. These C. canadensis plants were originally collected from wild populations at Spruce Knob, West Virginia, and Stoddard, New Hampshire. All materials were fixed in 3.0% glutaraldehyde solution overnight at 4°C. They were then dehydrated through an ethanol series, and processed for scanning electron microscopy as previously described (Feng et al., 2011).

Isolation of Cornus B-class gene orthologs

Inflorescence buds were collected at the J. C. Raulston Arboretum and NCSU main campus (or for C. canadensis from plants grown in the NCSU Phytotron) and stored in RNAlater (Ambion). RNAlater has been shown to prevent RNA degradation in microcrustaceans (Gorokhova, 2005), mammalian tissues (Mutter et al., 2004) and plant tissues (Harlow et al., 2006). Total RNA at different developmental stages was isolated following a modified CTAB RNA isolation method (Chang et al., 1993). RNA samples from the same species were pooled and cDNA was synthesized using the SMART RACE cDNA amplification kit (Clontech, Alameda, CA, USA) and SuperScript III first-strand synthesis system (Invitrogen). Twelve primers were used to amplify CorPI and CorAP3 fragments (Supporting Information Table S1) that were then subcloned using a TOPO TA cloning kit (Invitrogen). Subcloned fragments were sequenced in the Genome Sequencing and Analysis Core Facility at Duke University.

Cornus in situ hybridization

RNA in situ hybridization was used to examine the expression of PI and AP3 orthologs in Cornus during bract and floral organ development. RNA probes for in situ hybridization were in vitro transcribed following a protocol previously described by Franks et al. (2002) from species-specific subcloned fragments that were amplified using the following primer pairs: CorPIF29-AGTCTGGGAAGAGGTTGTGGGATGCTAAG, CorPIR29-CTTAGCATCCCACAACCTCTTCCCAGACT, CorAP3F29-GGACAGGTGAGAGTTTGAACGATCTGAGC, CorAP3R29-GCTCAGATCGTTCAAACTCTCACCTGTCC. These primer pairs amplified both of the paralogous CorPI-A and CorPI-B fragments (Results) that when subcloned allowed us to generate paralog-specific probes. The hybridization specificity of the CorPI-A and CorPI-B paralog-specific probes was tested using a RNA dot blot protocol (Sambrook & Russell, 2006). We observed a 10- to 100-fold higher hybridization signal from the perfectly matched probe relative to the imperfectly matched paralogous probe under our hybridization and wash conditions, indicating that the majority of the hybridization signal observed in our experiments is due to expression from the gene that is being assayed and not to cross-hybridization with the related paralog (data not shown). RNA probes were hydrolysed into fragments between 75 and 150 bp long through carbonate hydrolysis following the protocols of the Long lab (http://pbio.salk.edu/pbiol/in_situ_protocol.html). The in situ hybridization protocol for Cornus species was modified from protocols of Franks’ lab, Soltis’ lab, and Koes’ lab (Franks et al., 2002; Kim et al., 2005; Souer et al., 2008). Inflorescences were fixed in 3.7% formaldehyde for at least 8 h at 4°C, then dehydrated in an ethanol series and embedded in ParaplastPlus (Fisherbrand, Houston, TX, USA) as previously described (Franks et al., 2002). Hybridizations were repeated at least twice with multiple biological replicates in each experiment. Analyses of CorPI paralogs (CorPI-A and CorPI-B) were conducted in parallel to minimize bias from experimental variation.

Semi-quantitative reverse transcription-PCR

In order to prepare RNA for sqRT-PCR, fresh inflorescence buds (including bracts and flowers) were collected and immediately stored in RNAlater. For species with large, expanded bracts (C. canadensis, C. florida, C. officinalis), the bracts and flowers were manually separated. Total RNA was prepared using the CTAB method (Chang et al., 1993). cDNAs were synthesized using SuperScript III Reverse Transcriptase (Invitrogen) from total RNA. Primers used for sqRT-PCR of CorPI-A, CorPI-B and CorAP3 were designed from regions conserved between the Cornus species if possible. If a conserved site was not usable, species-specific primers were designed. Sequences of all primers are provided in Table S1. PCR conditions were optimized for each CorPI paralog and multiple pairs of PCR primers were tested; 35 cycles of amplification were used. The large subunit ribosomal RNA gene (26S rDNA) was used as the internal standard. Gel images were captured on a GelDoc XR System (Bio-Rad Laboratories, Inc. 2000, CA, USA) and expression levels were quantified with the Quantity One® software of the system. Normalized expression values for each lane were calculated by taking the signal value for the B-class gene amplicon and dividing by the signal value for the 26S ribosomal amplicon from that sample. The average normalized expression level of two independent biological replicates is reported in Fig. 5(b,d,f). The sqRT-PCR amplification products were sequenced to confirm their identity.

Character-mapping analyses

The character-mapping analyses were performed using the likelihood method with the mk1 model (one-parameter Markov k-state model) in MESQUITE v2.74 (Maddison & Maddison, 2007) via ancestral state reconstruction. The characters and character states are presented in Table 1. These characters were mapped onto the simplified phylogeny of Cornus reconstructed in several previous studies (Xiang et al., 1996, 1998, 2006, 2008, 2011). For the three BW species producing rudimentary bracts that fall off early in the spring, we did not have tissue for the gene expression analyses of expanding bracts. Thus, the character states for the expression of CorPI and CorAP3 transcripts in BW species bracts were coded as ‘not expressed’ (Table 1). An alternative maximum likelihood analysis, that instead coded the character states for the expression of CorPI and CorAP3 transcripts in bracts as ‘missing’ for the BW species, indicated that the ancestral states of most characters analysed were ambiguous (i.e. the character states at the root and internal nodes were equally likely and the evolutionary trends of characters became unclear; Fig. S1). Alternative scenarios shifted the relative timing of character state changes leading to C. canadensis and C. florida. The scenario consistent with the coding of the character states as ‘not expressed’ appears to be more parsimonious than alternative scenarios. It is noteworthy that the species included in this study represent the four major clades of the Cornus phylogeny. Other Cornus species not included in this analysis largely conserve the bract morphology of the clade. However, minor variation occurs in one species from the BB group, Cornus disciflora, that displays four involucral bracts that are not petaloid and not expanded. This variation represents an evolutionary reversal of the petaloid bracts to the ancestral state (Zhang et al., 2008; Fig. 1 of Xiang et al., 2011). Thus, missing the species in this analysis does not affect reconstructing ancestral states of the BB group.

Table 1.   Characters and character states of bract development and B-class gene expression that vary between species and were analysed for evolutionary history using Mesquite v2.74
SpeciesCharacter 1 bract initiation patternCharacter 2 CorAP3 expressionCharacter 3 CorPI-A expressionCharacter 4 CorPI-B expression
  1. IBM, inflorescence branch meristem; IM, inflorescence meristem; nd, no expression detected; DW, dwarf dogwood group; BB, big-bracted group; CC, cornelian cherries group; BW, blue or white fruits group.

  2. We did not have expression data from bracts in the BW group as they abscise early. For these species, Characters 2 through 4 were coded as ‘no expression detected’. We alternatively coded Characters 2 through 4 as ‘missing’ for these species (see Fig. 6 and Supporting Information Fig. S1).

Cornus canadensis (DW)0 (from IBM)1 (weak)0 (nd)0 (nd)
Cornus florida (BB)1 (from IM)2 (strong)1 (weak)2 (strong)
Cornus officinalis (CC)1 (from IM)1 (weak)0 (nd)1 (weak)
Cornus macrophylla (BW)0 (from IBM)0 (nd)0 (nd)0 (nd)
Cornus sanguinea (BW)0 (from IBM)0 (nd)0 (nd)0 (nd)
Cornus controversa (BW)0 (from IBM)0 (nd)0 (nd)0 (nd)


Temporal and spatial characteristics of bract initiation vary between Cornus species

Morphological descriptions of inflorescence development for Cornus florida, C. officinalis, C. macrophylla and C. canadensis have been previously reported (Feng et al., 2011). Briefly, in these four species, the inflorescence meristem produces a series of inflorescence branch meristems before generating floral meristems. The inflorescence branch meristems develop into the branches of the inflorescence, although in some of these species the elongation of the branches is very suppressed. The sequentially developing inflorescence branch meristems are false colored purple, salmon, orange and yellow in Fig. 2. All four species also initiate bract primordia; however, the timing and relative location of the initiating bracts relative to the development of the inflorescence branch meristems differ among the four species. In C. florida and C. officinalis, bract primordia are initiated from the periphery of the inflorescence meristem before it generates inflorescence branch meristems (Fig. 2d,e,g,h; Feng et al., 2011). The bract primordia are fairly well developed before the inflorescence branch meristems are morphologically discernable. In this sense, these bracts subtend the rest of the inflorescence and are true involucral bracts. In C. canadensis, the bract primordia arise from the outer portion of the inflorescence branch meristems. A key difference here is that the bracts in C. canadensis arise from the periphery of inflorescence branch meristem, not the inflorescence meristem directly. Thus the timing and positioning of bract initiation in C. canadensis is different from that in C. florida and C. officinalis. Bract initiation in C. macrophylla is similar to that in C. canadensis as the bract primordia arise from the outer portion of each inflorescence branch meristem. Additional bract primordia later initiate in positions subtending some of the higher order inflorescence branch meristems (Figs 1d and 2b,k).

Figure 1.

Bract diversity in dogwood species (Cornus). (a) Cornus canadensis (DW group); (b) C. florida (BB group); (c) C. officinalis (CC group); (d) C. macrophylla (BW group). Asterisk, involucral or pseudo- involucral bracts; white arrow, early deciduous rudimentary bracts. BW group, blue- or white-fruited dogwoods; DW group, dwarf dogwoods; BB group, big-bracted dogwoods; CC group, cornelian cherry. Bars, 1 cm.

Figure 2.

Initiation and early development of bracts in dogwood species (Cornus). (a–c) Cornus macrophylla; (d–f) C. officinalis; (g–i) C. florida; (j–l) C. canadensis. Bract primordia are false colored blue. The first group of inflorescence branch meristems are false colored purple; the second group of inflorescence branch meristems are false colored salmon; the third group of inflorescence branch meristems are false colored orange; the fourth group of inflorescence branch meristems are false colored yellow; the terminal flower is false colored green. The second pair of bract primordia is partially covered by young leaves (lf) in panel k. The arrow and black outline in panel f indicates the cryptic bract subtending the inflorescence branch meristem in C. officinalis that remains rudimentary or later aborts. For detailed information about inflorescence branch meristem development and nomenclature refer to Feng et al. (2011). IM, inflorescence meristem; sc, scales; lf, leaf. Bars, 100 nm.

Development of bracts progresses during the summer and fall before a period of winter dormancy. In C. officinalis and C. florida bracts enlarge substantially to enclose the young inflorescence buds, and thicken and harden into scale-like structures to protect the inflorescence bud during the winter. In C. canadensis and C. macrophylla, bracts from the inflorescence branch meristems expand slightly (Fig. 2c,l), but do not harden or enclose the inflorescence bud. In these two species, the young inflorescence and young bracts are protected by bud scales enclosing the inflorescence bud and adjacent leaf buds (Fig. 2a,j; marked ‘sc’ and ‘lf’, respectively). In all species examined, except C. canadensis, trichomes develop on the outer (abaxial) surface of the bracts (Fig. 2b,e,i).

Differences in bract growth and inflorescence branch elongation during spring growth period

The following spring, the two pairs of tightly wrapped, decussate bracts of C. florida and C. officinalis open up and expose the inflorescence. In C. florida the petaloid bracts expand significantly. The C. florida bracts gradually become yellowish and then become white in color c. 1 month before the anthesis (opening) of the true flowers of the inflorescence (Fig. 3a–d). By contrast, the bracts of C. officinalis do not exhibit a dramatic increase in organ size, or whitening; in fact, they appear to change very little during the spring growth phase. In C. canadensis – a rhizomatous herb – the winter bud emerges from the ground and elongates in the following spring into a leafy shoot bearing an inflorescence at the tip. The C. canadensis bracts expand significantly and gradually lose chlorophyll in a manner that is similar, at least superficially, to that of C. florida (Fig. 3l–o). In C. macrophylla bracts do not expand or whiten in the spring. Instead they remain rudimentary and later undergo abscission before the inflorescence is fully mature. It should be noted that the inflorescence branches of C. macrophylla elongate significantly in the spring into a paniculate cymous inflorescence. In C. canadensis there is a slight elongation of the branches of the inflorescence during the spring growth phase. This branch elongation is not observed in C. florida or C. officinalis.

Figure 3.

Development of bracts during spring growth period. Cornus florida (a–k); C. canadensis (l–v). (a) Phase I, after the inflorescence bud opens, but before bract expansion occurs; (b) phase II, mid bract expansion time point; (c) phase III, fully expanded, terminal bract morphology; (d) close up of (c); (e) adaxial epidermal cell morphology of phase I bracts; (f) adaxial surface of phase II bracts; (g) adaxial surface of phase III bracts; (h) higher magnification of adaxial surface of phase III bracts; (i) abaxial epidermal cell morphology of phase I bracts; (j) abaxial surface of phase II bracts; (k) abaxial surface of phase III bracts; (l) Phase I, bracts visible; (m) Phase II, mid bract expansion time point; (n) slightly later in Phase II; (o) phase III, mature bracts; (p) adaxial surface of phase I bracts; (q) adaxial surface of phase II bracts; (r) adaxial surface of phase III bracts; (s) higher magnification of adaxial surface of phase III bracts; (t) abaxial epidermal cell morphology of phase I bracts; (u) abaxial surface of phase II bracts; (v) abaxial surface of phase III bracts. Bars: (a, b, d, m–o) 0.5 cm; (c, l) 1 cm; (e–g, i–k, p–r, t–v) 100 nm; (h, s) 10 nm.

In summary, the mature inflorescences of these four species differ in two important characteristics: the location of the bracts relative to the branch meristems, and the extent of growth and petaloidy of the bracts. In C. florida, four large, white petaloid bracts subtend the entire inflorescence and can thus be considered true involucral petaloid bracts. The analysis of bract initiation in C. canadensis indicates that the large, petaloid bracts subtend the inflorescence branches rather than subtending the entire inflorescence. Because the rachis supporting each inflorescence branch does not evidently elongate, the four bracts appear like a whorl subtending the entire inflorescence. Thus, bracts in C. canadensis might be considered pseudo-involucral as developmentally they appear to arise subtending the inflorescence branches. In the two species with nonpetaloid bracts we also see both patterns of bract initiation. In C. officinalis, four small nonpetaloid, involucral bracts subtend the entire inflorescence while in C. macrophylla, a rudimentary bract is observed subtending each inflorescence branch.

Scanning electron microscopic examination of petaloid bract epidermal cell morphology

In order to further compare the morphological differences among species, we used scanning electron microscopy (SEM) to examine the epidermal cell morphology of expanding bracts during spring growth period. The bracts of two nonpetaloid species, C. officinalis and C. macrophylla, had no significant growth or expansion during the spring and were covered by thick trichomes. This made it impossible to discern the epidermal cell morphology under SEM. Thus, we were only able to observe the epidermal cell morphology of the petaloid bracts of C. florida and C. canadensis that were expanding during the spring growth period (Fig. 3). Besides, bracts before the new growth in the spring showed no apparent differences in the epidermal cell morphology of bracts among the petaloid and nonpetaloid species.

In C. florida and C. canadensis, at the beginning of the spring growth period, before the bracts began to increase in size (Fig. 3a,l), the epidermal cells of the adaxial surface in both species are similar, exhibiting a small, round and tightly packed morphology (Fig. 3e,p). As the bracts in both species begin to enlarge (Fig. 3b,m), the epidermal cells of the adaxial bract surface increase in both length and width, but remain un-domed in both species (Fig. 3f,q). When bracts are fully white and mature (Fig. 3d,o), the adaxial epidermal cells in both species are papillate or conical, and display a surface morphology characterized by a set of epicuticular ridges arranged in a radial pattern (Fig. 3g,h,r,s), as is typically observed in angiosperm petals. The abaxial epidermal cells in both species display an interlocking pavement cell-like morphology, similar to those typically observed in leaves (Fig. 3k,v). Stomatal cells are occasionally observed on the abaxial surface (Fig. 3k,t–v), but we did not observe stomatal cells on the adaxial surfaces in either species. In general, the morphological characteristics of the bract epidermal cells are relatively similar in C. florida and C. canadensis.

Identification of B-class genes from Cornus species

The B-class petal identity genes are defined by the founding members: GLOBOSA and DEFICIENS from Antirrhinum majus and PISTILLATA (PI) and APETALA3 (AP3) from Arabidopsis thaliana. The Cornus orthologs of PISTILLATA (PI) and APETALA (AP3; named CorPI and CorAP3, respectively) were isolated from six Cornus species: C. florida, C. officinalis, C. macrophylla, C. canadensis, C. sanguinea and C. controversa. The species C. sanguinea and C. controversa (from the BW group) were added to the sampling to represent additional subclades within this more speciose group. We named the B-class genes from the Cornus species that we assayed using the first three letters of Cornus following the first three letters of the species specific-nomial and then the gene name. For example, CorcanPI-A is to indicate the A copy of PI ortholog in C. canadensis. We examined the Cornus B-class cDNA sequences by phylogenetic analysis to identify genes most likely to be orthologous to the Arabidopsis PI and AP3 genes using other PI- and AP3-like sequences from Asterids, Rosids and basal eudicots (data not shown). Two PI orthologs (named as CorPI-A and CorPI-B) and a single AP3 ortholog (CorAP3) were identified in C. florida (BB group), C. canadensis (DW group) and C. officinalis (CC group). Only one PI ortholog and one AP3 ortholog were identified in the three species representing the BW group (C. macrophylla, C. sanguinea and C. controversa), which is consistent with the finding of Zhang et al. (2008; Fig. S2 based on a 5′ portion of the genomic and cDNA sequences of PI homologs in Cornus.

Expression pattern of CorPI-A, CorPI-B and CorAP3 during flower development

For in situ hybridization, inflorescences were collected in the summer and fall to assay the earliest stages of bract primordia and floral organ initiation and development. Results of in situ hybridization showed that PI and AP3 orthologs were expressed in petal and stamen primordia during early floral organogenesis in all Cornus species studied (Fig. 4). Expression of CorPI and CorAP3 were strongly detected during the initiation and early development of petals and stamens. The expression levels of these genes in petals and stamens appeared somewhat reduced at later developmental stages (Fig. 4). The two paralogs of CorfloPI exhibited consistent differences in signal intensity in these in situ hybridization experiments. In C. florida, the CorfloPI-B probe generated a strong signal while the CorfloPI-A probe generated a signal that was difficult to detect above background (compare Fig. 4b,c to d). The two paralogs of CorcanPI were both highly expressed in young petal and stamens although the expression level of CorcanPI-A seems slightly higher (cf. Fig. 4i,j to k).

Figure 4.

Expression patterns of CorPI and CorAP3 in early bract and flower development detected using in situ hybridization. (a–c) Expression of CorfloPI-B in Cornus florida developing bracts and flowers; (d) expression of CorfloPI-A in C. florida, expression of Corflo PI-A copy is difficult to detect above background; (e–g) expression of CorfloAP3 in C. florida; (h–j) expression of CorcanPI-A in Cornus canadensis; (k) expression of CorcanPI-B in C. canadensis; (l–n) expression of CorcanAP3 in C. canadensis; (o–q) expression of CoroffPI-A in Cornus officinalis; (r–t) expression of CoroffAP3 in C. officinalis (expression of CoroffPI-B was not detected; images not shown). br, bract; s, sepal; p, petal; st, stamen, IM, inflorescence meristem; sc, scale; Bars, 100 μm.

In order to better estimate the expression levels of the CorPI and CorAP3 transcripts in the flowers and bracts of the different Cornus species, we employed a semi-quantitative reverse transcription PCR (sqRTPCR). Results from sqRT-PCR indicated that both CorPI and CorAP3 transcripts were expressed during the spring growth period as the inflorescence bud opens and reinitiates development (Fig. 5). In order to compare gene expression during the spring growth period we divided that period into three phases according to the morphological development of the inflorescences. Phase I encompasses the period after the inflorescence bud opens, but before bract expansion occurs. Phase II is a mid bract expansion time point and Phase III reflects the fully expanded, terminal bract morphology. In the species in which there is no bract expansion, a comparable developmental stage was estimated using the overall morphological development of the inflorescence. For C. florida, C. canadensis and C. officinalis bracts were separated from the inflorescence during the collection of the tissue (see the Materials and Methods section) thus yielding the tissue samples F1 (floral tissue from phase 1) and B1 (bract tissue from phase1), etc. as reported in Fig. 5. Both the A and B copies of CorPI are highly expressed in floral tissue during phase 1 (lanes marked F1) in C. florida, C. canadensis and C. officinalis. In C. florida the B copy of PI (i.e. CorfloPI-B) was expressed two- to five-fold higher than the A copy (i.e. CorfloPI-A) depending on which stage or tissue is compared (Fig. 5a,b). This higher expression of the B copy of CorPI was not observed in C. canadensis.

Figure 5.

Expression levels of CorPI-A, CorPI-B, and CorAP3 in Cornus bracts and flowers during spring growth period estimated using semi quantitative RT-PCR. (a) Expression of CorfloPI-A, CorfloPI-B and CorfloAP3 in Cornus florida; (b) semi-quantitative estimates of the expression levels shown in (a) (see the Materials and Methods section). (c) Expression of CorcanPI-A, CorcanPI-B and CorcanAP3 in Cornus canadensis; (d) semi-quantitative estimates of the expression levels shown in (c). (e) Expression of CoroffPI-A, CoroffPI-B, and CoroffAP3 in Cornus officinalis; (f) semi-quantitative estimates of the expression levels shown in (e). L, leaf; B1–B3; bract samples from three phases of development during the spring growth period as described in the Results section; F1–F3, floral/inflorescence samples from three phases of development during the spring growth period; 26S, 26S ribosomal rDNA used as an internal control.

Expression pattern of CorPI-A, CorPI-B and CorAP3 during bract development

We employed an in situ hybridization analysis to examine expression of CorPI and CorAP3 during early stages of inflorescence development, but we failed to detect expression of CorPI or CorAP3 in the very young bract primordia in the three species we examined, (i.e. C. florida, C. canadensis and C. officinalis; Fig. 4a,e,h,l,o,r). Even in sections in which strong expression of the B-class genes was observed in the developing floral organs (Fig. 4i), we did not detect expression of Cornus B-class genes in the very young bracts. These results suggest that the petaloid nature of the bracts in C. canadensis and C. florida does not require expression of B-class genes early in the development of the bract (during what might be considered the ‘organ identity specification stage’ based on work in Arabidopsis). We also examined B-class gene expression in the maturing bracts (collected in the spring stages B1, B2 and B3) by sqRT-PCR. Our analysis of C. florida bract samples from the spring growth phase B1 revealed expression of CorfloPI-B and CorfloAP3 and a somewhat lower level of expression of CorfloPI-A (Fig. 5a,b). By contrast, we did not detect expression of CorPI-A or CorPI-B in the bracts of C. canadensis, even at late stages when the C. canadensis bracts display petaloid characteristics (Fig. 5c,d). CorAP3 in C. canadensis bracts was weakly, but consistently, detected. CorPI-A, CorPI-B, and CorAP3 were all detected in the C. officinalis bracts, but at lower levels than were detected in the C. florida samples. Thus, there appears to be higher levels of B-class gene expression in the bracts of C. florida when compared to those of C. canadensis and C. officinalis.

Evolutionary changes in bract development and B-class gene expression along the phylogenetic pathways to C. canadensis and C. florida.

In order to test the hypothesis that the petaloid bracts of C. florida and C. canadensis share a common evolutionary origin and thus petaloidy arose only once in this lineage as has been previously suggested based on outer morphology (Xiang & Thomas, 2008; Xiang et al., 2011), we carried out character-mapping analyses to trace the evolutionary trends of characters with divergent character states onto the existing Cornus phylogenetic tree. To do this, we treated each gene copy as an independent character (e.g. AP3, CorPI-A, CorPI-B). Our comparative study of bract development and B-class gene expression patterns revealed at least four character states that varied among the species we examined (Table 1). These include the pattern of bract initiation (character 1), and the expression patterns of CorAP3, CorPI-A and CorPI-B in B1 sample bracts (characters 2–4). We assigned transcript expression levels to either two (expressed vs not expressed) or three (not expressed, weakly expressed, or strongly expressed) character states. These four characters were traced over the phylogeny based on an existing phylogenetic tree of Cornus species (Xiang & Thomas, 2008; Xiang et al., 2011). Results of likelihood mapping of these four characters indicated that seven evolutionary changes occurred on branches leading to C. florida and five changes occurred on branches leading to C. canadensis. Three of these changes are shared by both species and occurred in the common ancestor of C. canadensis, C. florida, and C. officinalis. These three changes are a change in the location and timing of bract initiation (Character 1), a change in the level of CorAP3 expression in B1 bracts from no expression to weak expression (Character 2), and a change in the level of CorPI-B expression from no expression to weak expression (Character 4; Fig. 6). Three evolutionary changes that are correlated with the presence of bract petaloidy occurred on the terminal branch of C. florida. These are a change in the expression of CorAP3 in B1 bracts from weak to strong expression (Character 2), expression of CorPI-A in B1 bracts from not expressed to weak expression (Character 3), and a change in the expression level of CorPI-B in B1 bracts from weak to strong (Character 4). One evolutionary change occurred in the terminal branch of C. canadensis, a reversal of the bract initiation pattern to the ancestral state (i.e. initiated from the inflorescence branch meristem; Character 1).

Figure 6.

Evolutionary changes in B-class gene expression patterns and bract initiation patterns that are associated with Cornus cladogenesis as inferred from maximum likelihood analyses in Mesquite v2.74. (a–d) Evolutionary trends for four characters: character 1 (ch1) through character 4 (ch4); (e) summary of evolutionary changes for the co-related characters in Cornus florida and C. canadensis. Tick marks on branches represent the evolutionary events. Detailed information on characters and character states is found in Table 1. The pie charts indicate the relative probability of each different character state at a given ancestral node. Tree topology used was simplified from the phylogeny of Cornus derived from DNA sequences of multiple genes (Xiang & Thomas, 2008; Xiang et al., 2011).


Differences in the pattern of bract initiation suggest that the petaloid bract organs in C. florida and C. canadensis are not homologous organs

Here we suggest that the petaloid bracts observed in two Cornus species, C. florida and C. canadensis, originated independently in parallel. In support of this hypothesis we have identified a significant difference in the pattern and timing of bract initiation between these two species. Bracts in C. florida initiate from the inflorescence meristem before that meristem generates inflorescence branch meristems, while in C. canadensis bracts initiate from the abaxial portions of the individual inflorescence branch meristems. In this sense the C. florida bracts subtend the rest of the inflorescence and are true involucral bracts, while C. canadensis bracts subtend the inflorescence branches rather than subtending the entire inflorescence. Because the inflorescence branches do not elongate, the bracts appear to be involucral, although they might better be classified as pseudo-involucral. This difference in the pattern of initiation suggests that petaloid bracts in C. florida are not homologous with the petaloid bracts seen in C. canadensis. The pattern of bract initiation in C. florida (generating true involucral bracts) was also observed in C. officinalis. In addition to the involucral bracts observed in C. florida and C. officinalis, bracts that subtend the inflorescence branch meristems also initiate (Fig. 2f). However, the development of these bracts is suppressed at later developmental stages. We suggest that these cryptic bracts in C. florida and C. officinalis may be considered homologous to the petaloid bracts seen in C. canadensis. Our character mapping analysis suggests that the pattern of bract initiation in C. canadensis (from the inflorescence branch meristem) represents an evolutionary reversal to the ancestral state. This analysis also suggests that the switch to the derived state (initiation from the inflorescence meristem) likely occurred in the common ancestor of C. canadensis, C. florida and C. officinalis.

Differences in the patterns of B-class gene expression in the petaloid bracts of C. florida and C. canadensis are consistent with the proposed independent origins

The results of our analysis of B-class gene expression in C. florida and C. canadensis are also consistent with the hypothesis that petaloidy of the bract organ may have arisen independently in these two Cornus species. During bract maturation the expression of the CorPI-B, CorPI-A and CorAP3 genes can be detected in C. florida (particularly in the B1 bract samples) but they are only very weakly detected (CorAP3) or undetectable (CorPI-A and CorPI-B) in C. canadensis bracts. This is in contrast to the expression of the B-class genes in the developing petals and stamens in both C. florida and C. canadensis, that are easily detected in both species. The expression patterns of Cornus B-class genes we report here are consistent with those reported by Maturen (2008), who assayed the expression levels of A-, B- and E-class genes in C. florida and C. canadensis using quantitative PCR. In that study, the A-, B- and -E class genes were found to be expressed in bracts of C. florida, but not in comparably staged bracts of C. canadensis. Although our data are consistent with independent origins of bract petaloidy in C. florida and C. canadensis, we cannot rule out the possibility that petaloidy of bracts in these species had a single evolutionary origin in their common ancestor and that the differences in patterns of gene expression between these two species evolved at a later evolutionary time after the two lineages diverged. Additionally, although our data indicate a correlation between the expression of the Cornus B-class genes and petaloid bract development, we cannot at this time conclusively assign a functional role for the Cornus B-class genes during petal or petaloid bract development. The future analysis of RNAi knockdown lines in C. canadensis using an available regeneration/transformation system (Feng et al., 2009; X. Liu et al., unpublished) to knockdown the functions of the B-class genes in C. canadensis may illuminate the functional roles of the Cornus B-class genes during both petal and petaloid bract development.

It is interesting that we did not detect expression of Cornus B-class genes in the early initiating bract primordia, but rather B-class gene expression was only detected later during phase B1 (Fig. 5). This indicates that bract petaloidy in C. florida may not require the early expression of B-class genes in the developing bract primordia. It further suggests that the bract organ identity in the species is probably not specified by B-class gene orthologs, but rather that petaloidy of the bracts during their expansion in the species may require AP3 and PI orthologs. In Arabidopsis thaliana early expression of the B-class genes AP3 and PI is thought to coordinate petal identity specification from a very early developmental stage. Thus, the mechanism of ‘petal identity specification’ in the C. florida bract may be quite different from that in the Arabidopsis petal or in the Cornus petal for that matter. Furthermore, the very reduced levels of B-class gene expression detected in the C. canadensis bracts suggest that B-class genes may not be required for petaloidy in C. canadensis. However, we cannot at this point rule out an important function for the residual B-class gene expression that we detect in C. canadensis bracts. With respect to the development of stamens and petals within the Cornus flower, the expression of CorPI and CorAP3 in petals and stamens in all of the Cornus species we examined suggests that the B-class gene-dependent mechanism of petal and stamen identity specification that has been observed in a diversity of Eudicots lineages may have been conserved in dogwood flowers (Coen & Meyerowitz, 1991; Vandenbussche et al., 2004; De Martino et al., 2006; Drea et al., 2007; Irish, 2009; Litt & Kramer, 2010).

B-class gene function and heterotopic petaloidy

Investigations in a variety of plant species suggest that the requirement for B-class gene function during heterotopic petaloidy varies dramatically, and may function early, late, or not at all during the development of petaloid organs (Park et al., 2003, 2004; Jaramillo & Kramer, 2004; Vekemans et al., 2011; Brockington et al., 2012; Landis et al., 2012). In Aristolochia (Aristolochiaceae, basal angiosperm), expression of an AP3-like gene was detected only during the late stages of petaloid perianth development and expression of B-class paralogs was detected in late development of Aquilegia petaloid sepals, suggesting that B-class homologs may only be required relatively late in the development of petaloid organs in these species (Jaramillo & Kramer, 2004; Kramer & Hodges, 2010). Jaramillo et al. suggest that the Aristolochia B-class genes may be required for the specification of the papillate or conical cell morphology observed in the epidermal cells of the adaxial surface of the perianth tube and thus may be a conserved function of the B-class orthologs. The detection of B-class genes only in maturing bracts in Cornus, as we report, suggests that a similar mechanism for heterotopic petaloidy may exist in the Cornus lineage. We also observed epidermal cells with papillate or conical morphologies in the Cornus petaloid bracts (Fig. 3). By contrast, in Davidia the expression of two B-class genes (DiTM6 and DiGLO1/2), as well as a C-class gene (DiAG), was detected early during the development of the petaloid involucral bracts (Vekemans et al., 2011). Interestingly, strong expression of the orthologs of Arabidopsis B- and C-class genes in the Davidia petaloid bracts is not maintained into later developmental stages, suggesting that the development of petaloid bracts in Davidia does not require expression of B-class genes during late stages of bract development. Of note, the adaxial epidermal cells of the Davidia petaloid bracts do not exhibit the conical cellular morphology that we have detected in C. florida and C. canadensis and that is commonly associated with Angiosperm petals. Finally, studies in Asparagus officinalis (Asparagaceae), Rhodochiton atrosanguineum (Plantaginaceae), Sesuvium portulacastrum (Aizoaceae) and Delosperma napiforme (Aizoaceae) conclude that the patterns of B-class gene expression in these species are not consistent with a role for B-class genes in the heterotopic petaloidy observed in these species (Park et al., 2003, 2004; Brockington et al., 2012; Landis et al., 2012). These studies indicate that the sliding boundary modification of the ABCE model is insufficient to explain the development of the many heterotopic petaloid organs observed in flowering plants. They also suggest that several different molecular mechanisms that support heterotopic petaloidy have evolved and that either early or late petal development programs may be independently recruited.


The authors would like to give special thanks to S. Azhakanandam for technical assistance with in situ hybridization and Dr D. E. Boufford and Dr L. J. Mehrhoff for assistance in collecting C. canadensis in New Hampshire. We thank Dr D. E. Soltis, Dr P. E. Soltis and Dr R. E. Koes for helpful discussion, and Wenheng Zhang, Jian Zhang, Juan Liu and anonymous reviewers for providing extensive and helpful comments for this manuscript. We acknowledge generous support from the JC Raulston Arboretum, the phytotron of College of Agriculture and Life Sciences at North Carolina State University and Center for Electron Microscopy at North Carolina State University. 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) and benefited from a CAS/SAFEA International Partnership Program for Creative Research Teams.