Author for correspondence: Yin-Zheng Wang Tel:+86 10 62836474 Fax:+86 10 62590843 Email: firstname.lastname@example.org
• The shift from zygomorphy to actinomorphy has been intensively studied in molecular genetic model organisms. However, it is still a key challenge to explain the great morphological diversity of derived actinomorphy in angiosperms, since different underlying mechanisms may be responsible for similar external morphologies. Bournea (Gesneriaceae) is of particular interest in addressing this question, as it is a representative of primarily derived actinomorphy characteristic of a unique developmental transition from zygomorphy to actinomorphic flowers at anthesis.
• Using RNA in situ hybridization, the expression patterns were investigated of three different Bournea orthologues of TCP and MYB genes that have been shown to control floral symmetry in model species.
• Here, it is shown that the initial zygomorphic pattern in Bournea is likely a residual zygomorphy resulting from conserved expression of the adaxial (dorsal) identity gene BlCYC1. As a key novel event, the late downregulation of BlCYC1 and BlRAD and the correlative changes in the late specific expression of the abaxial (ventral) identity gene BlDIV should be responsible for the origin of the derived actinomorphy in Bournea.
• These results further indicate that there might be diverse pathways in the origin and evolution of derived actinomorphy through modifications of pre-existing zygomorphic developmental programs under dynamics of regulatory networks.
One important event during the evolution of angiosperms is the secondary shift from zygomorphy (bilateral symmetry along the dorsoventral axis) to actinomorphy (radial symmetry) of their flowers. The shift in floral symmetry has attracted considerable interest for over 200 yr since Carl Linnaeus first described a natural peloric mutant of toadflax (Linaria vulgaris) (Gustafsson, 1979). In the past decade, significant progress has been made in the area of flower symmetry regulation especially in model genetic systems including Antirrhinum majus (Luo et al., 1996, 1999; Cubas et al., 1999a; Endress, 1999; Citerne et al., 2006; Feng et al., 2006; Howarth & Donoghue, 2006; Damerval et al., 2007). However, the evolutionary shift from zygomorphy to actinomorphy is still a vast unexplored field at the molecular developmental level in naturally occurring nonmodel organisms. In Lamiales sensu lato, a major clade of predominantly zygomorphically flowered angiosperms, derived actinomorphy frequently occurs (Donoghue et al., 1998; Endress, 1998).
Note that actinomorphy is considered the ancestral state for angiosperms, and zygomorphy has evolved several times independently from actinomorphic ancestors (Crepet, 1996; Donoghue et al., 1998). However, zygomorphy is frequently lost in the Asteridae, especially in Lamiales sensu lato (Donoghue et al., 1998; Endress, 1998). The secondary shift from zygomorphy to actinomorphy may be a reversal to an earlier ancestral form, but may also be an innovative process or innovative homoeotic transformation (Citerne et al., 2006). Since it is not exactly known to date which genes or gene interactions are involved in controlling the ancestral actinomorphy, the term ‘derived actinomorphy’ as used here is not concerned whether it is a reversal or innovative process, but only indicates an actinomorphy derived from an otherwise zygomorphic clade.
In A. majus with zygomorphic flowers, the floral dorsoventral (adaxial/abaxial) asymmetry is established through transcriptional and post-transcriptional interactions among three key genes encoding transcription factors of the TCP and MYB families. CYCLOIDEA (CYC) and DICHOTOMA (DICH), encoding two related transcription factors of the TCP (Teosinte Branched 1, CYCLOIDEA and PCF) family (Cubas et al., 1999b), function redundantly in specifying the adaxial (dorsal) identity (Luo et al., 1996, 1999). Conversely, DIVARICATA (DIV) encodes a MYB transcription factor and determines the abaxial (ventral) petal identity (Galego & Almeida, 2002; Perez-Rodriguez et al., 2005). The function of CYC and DICH is mediated through activating the downstream target RADIALIS (RAD), another MYB gene, whose encoded protein antagonizes DIV protein function in the adaxial region and therefore limits DIV to the abaxial domain (Corley et al., 2005; Costa et al., 2005). In Antirrhinum, the cyc/dich double mutant has a fully abaxialized (ventralized) peloric flower because RAD is not activated (Corley et al., 2005) to exert an antagonistic effect on DIV in the adaxial region. The abaxialized peloric mutants of both A. majus and L. vulgaris are caused by complete silencing of CYC and LCYC (Linaria CYC) through transposon insertion and extensive DNA methylation, respectively (Luo et al., 1996; Cubas et al., 1999a). In legumes, distantly related to Antirrhinum, CYC homologues legCYC genes are involved in the establishment of zygomorphy (Feng et al., 2006). Actinomorphic flowers in legumes come from legCYC expression in all five petals, similar to the adaxialized (dorsalized) mutant in Antirrhinum as a result of ectopic expression of CYC (Citerne et al., 2006). However, the formation of the great diversity of actinomorphy derived from different clades of zygomorphy in nature seems not to involve simple loss-of-function or gain-of-function mutations and the same mechanism that underlies the genetic control of floral symmetry in the model genetic system (Donoghue et al., 1998; Reeves & Olmstead, 2003; Smith et al., 2004). Using the mechanisms that underlie mutant phenotypes in model species as a starting point, detailed explorations of natural types of derived actinomorphy may reveal new evolutionary pathways from zygomorphy to actinomorphy, involving different suites of genes and acting at different times during development.
The family Gesneriaceae, sister to the remainder of Lamiales s. l. ( Wortley et al., 2005), with weak floral zygomorphy (especially in subfamily Cyrtandroideae), has the most diverse forms of derived actinomorphy and possesses the largest proportion of genera with actinomorphic flowers in Lamiales s. l. (Endress, 1998, 1999). In Gesneriaceae, Bournea leiophylla exhibits actinomorphic flowers at anthesis (Fig. 1). However, its very short corolla tube with five filaments adnate to the corolla and the adaxial petals somewhat smaller than others hints at their evolution from a zygomorphic ancestor (Fig. 1). Phylogenetic analyses confirm that the ancestor to Bournea was likely a species with zygomorphic flowers (Li & Wang, 2004; Du & Wang, 2008). The floral development further shows that the flowers in B. leiophylla undergo a morphological transition from a zygomorphic pattern during organ initiation and early development to actinomorphy at anthesis (Figs 1 and 2). This developmental pattern implies that CYC-like genes should be functional in controlling the floral dorsoventral asymmetry in B. leiophylla, at least in early floral development. Given that Gesneriaceae and Veronicaceae (family of Antirrhinum) are relatively closely related ( Wortley et al., 2005) and CYC-like genes function in arresting stamen development (Luo et al., 1996, 1999; Hileman et al., 2003), it is unlikely that the actinomorphic flower in B. leiophylla results from either loss-of-function or gain-of-function of CYC-like genes. Instead, the developmental transition from the initial zygomorphy to actinomorphy at anthesis may involve an alteration in the regulatory interaction among TCP and MYB family genes during floral development.
Recent molecular genetic studies have begun to reveal the dynamics of gene-regulatory interaction, and the complex combinatorial mechanisms that create a distinct final floral architecture and form (Krizek & Fletcher, 2005). Bournea represents an ideal candidate for exploring a potentially novel evolutionary pathway for derived actinomorphy in angio-sperms, especially in Lamiales s. l. To address this, we carried out a comprehensive investigation on the expression patterns of all three kinds of floral symmetry genes known in Antirrhinum (i.e. CYC, RAD and DIV homologues; DICH only exists in Antirrhineae) (Hileman & Baum, 2003; Cubas, 2004). The first goal of this study was: to identify whether the early expression patterns of the three kinds of floral symmetry genes are correlated with the initial floral zygomorphy in B. leiophylla as in those in A. majus; to identify whether their expression is altered when the floral development shifts to an actinomorphic pattern; to understand the functional and developmental homology of the floral symmetry genes between B. leiophylla and A. majus. Second, we focused on how the expressional alteration and evolution of floral symmetry genes relate to the developmental transition from zygomorphy to actinomorphy in B. leiophylla, which would provide novel evolutionary pathway for derived actinomorphy in angiosperms.
Materials and Methods
Molecular cloning and sequence analyses
We isolated CYC, RAD and DIV homologues from Bournea leiophylla using degenerate oligonucleotide primers in 3′- and/or 5′-RACE (rapid amplification of cDNA ends) according to described methods (Sambrook & Russell, 2001) and the manufacturer's protocol (Invitrogen, Carlsbad, CA, USA), respectively. Total RNA was extracted from the young floral buds of B. leiophylla using the Plant RNA Purification Reagent (Invitrogen) according to the manufacturer's instructions. First-strand cDNAs were synthesized from total RNA for 3′-RACE with the Supertranscript III RNase H− Reverse Transcriptase (Invitrogen). To examine the intron/exon structures we isolated and sequenced the corresponding genomic DNA of BlCYC, BlRAD and BlDIV from leaves. The oligonucleotide sequences for primers are included in the Supplementary Material Text S1.
According to the known and putative amino acid information, phylogenetic analyses of DIV-like and RAD-like proteins were conducted, respectively (Rose et al., 1999; Galego & Almeida, 2002; Barg et al., 2004; Guo et al., 2005; Baxter et al., 2007). Protein sequences to be analysed were aligned using clustalx 1.81 (Thompson et al., 1994) or dnaman version 5.2.2 (Lynnon BioSoft, Pointe-Claire, Quebec, Canada) and manually refined. Amino acid sequences for phylogenetic analysis were aligned and all gaps were removed manually. The neighbor-joining method with p-distance was carried out using mega 3.1 (Kumar et al., 2004). Bootstraps (1000 replicates) were conducted to assess the statistical reliability of the inferred topology with mega 3.1 (Kumar et al., 2004). The sequences reported in the paper for B. leiophylla BlCYC1, BlCYC2, BlRAD, BlDIV1 and BlDIV2 have been deposited in the GenBank database (Accession Nos. EF211118, EF211119, EF211120, EF211121, EF207557, EF211122, EF486282, EF486283 and EF486284).
RNA in situ hybridizations
Materials for in situ hybridization were fixed, sectioned and hybridized to digoxygenin-labeled probes of BlCYC1, BlRAD, BlDIV1 and BlDIV2 with reference to described methods (Bradley et al., 1993). The templates for transcriptions of probes specific to BlCYC1, BlRAD, BlDIV1 and BlDIV2 include translated and nontranslated regions. The fragments were amplified from single-strand cDNA of young floral buds from Bournea by using primer pairs specific for BlDIV1, BlDIV2, BlRAD, and BlCYC1, and then purified and cloned into pGEM-T easy vectors. Linearized templates were amplified by using primers YT7 and YSP6, and then single-strand digoxygenin-labeled antisense and sense probes were prepared from the linearized templates (Divjak et al., 2002). The oligonucleotide sequences for primers are included in the Supplementary Material Text S1.
Tissues used for RT-PCR including petals and sepals were dissected from middle-late B. leiophylla flowers whose organs had undergone a degree of differentiation. Sepals were dissected from the flowers. The petals were partitioned into abaxial, lateral and adaxial regions. Total RNA was extracted from the three types of petal regions and from the sepal tissues respectively. cDNA was prepared from 4 µg of different total RNA with the Supertranscript‘ III RNase H− Reverse Transcriptase (Invitrogen). To make sure each pair of primers was suitable, we first used them to amplify genomic DNA of B. leiophylla. The PCR products were then cloned. At least 20 clones of each PCR product were sequenced and only the primers that could specifically amplify BlDIV1, BlDIV2, BlRAD, and BlCYC1, BlCYC2 genes were used further. The oligonucleotide sequences for primers are in the Supplementary Material Text S1. The PCR reaction mixes included 0.75 units of Ex-taq (Takara, Tokyo, Japan), 2 mm MgCl, 200 µm of each dNTP, 40 mm of corresponding primers and 1.5 µl corresponding first-strand cDNA. The amplification program was: 94°C for 3 min, followed by 35 cycles of 94°C for 30 s, 52°C or 54°C for 30 s, and 72°C for 40 s and a final 8-min extension at 72°C. Amplification of ACTIN was used as a positive control (Jin et al., 2000). The primers for ACTIN are actinF and actinR. All RT-PCR products were cloned into pGEM-T easy vectors, and 20 clones from each RT-PCR were sequenced to confirm locus-specificity.
Scanning electron microscopy
Floral buds were collected at different developmental stages and were examined under a Hitachi S-800 scanning electron microscope in Institute of Botany, CAS. The electron microscopy was performed according to previously described protocol (Matthews & Endress, 2004).
Floral development in Bournea leiophylla
We first examined the floral developmental process in Bournea leiophylla, using scanning electron microscopy, and compared it with the previously described floral developmental stages in A. majus (Galego & Almeida, 2002; Vincent & Coen, 2004) (also see ‘Definition and additional description of the floral developmental stages in Bournea leiophylla’ in the Supplementary material Text S2). It was found that the petals and stamens were unidirectionally initiated from the abaxial side and subsequently extended to the adaxial side of developing flowers, and the adaxial stamen was much delayed in initiation (Fig. 2a–c), exactly matching stages 4 and 5 of floral development in A. majus and zygomorphic taxa in Gesneriaceae (Wang et al., 1997, 2002; Vincent & Coen, 2004). However, rather than stop growth after carpel initiation (stage 6) as in Antirrhinum and zygomorphic taxa in Gesneriaceae, the adaxial stamen of B. leiophylla continued to elongate and expand in size (Fig. 2d). In A. majus and zygomorphic taxa in Gesneriaceae (Wang et al., 1997, 2002; Vincent & Coen, 2004), Stage 7 is characteristic of the abaxial (ventral) and lateral stamens beginning to widen (or swell) (microsporangia) with corolla almost covering the stamens and Stage 8 shows an important feature of basic external structure of the anther becoming visible and the rudimentary filament seen. During stages 7 and 8 in B. leiophylla, the development of the corolla and stamens shows a common feature with those of A. majus and zygomorphic taxa in Gesneriaceae except for the adaxial corolla lobes (data not shown) and the adaxial stamen gradually catching up with the abaxial and lateral ones in size (Fig. 2e–g). When anther sacs were clearly defined and two carpels were closed into the ovary, which is defined as stage 9 and 10 in A. majus (Vincent & Coen, 2004), the adaxial stamen became almost equal in size with other stamens, and two adaxial petals were only slightly smaller than others but similar in shape (Fig. 2 h–i). Lateral growth between adjacent petals and stamens only lasted a short duration, forming a short corolla tube (Fig. 2b–f). At anthesis, all five petals were equally expanded and similar in shape albeit the two adaxial petals were slightly smaller than others. All five stamens were almost equal in length and were extending radially and horizontally (Fig. 1).
Sequence analyses of BlCYC, BlRAD and BlDIV
There are two CYC homologous genes in B. leiophylla (see the Materials and Methods section). They encode two related proteins of 333 and 331 amino acids, respectively. We designated them as BlCYC1 and BlCYC2. Sequence analysis shows that BlCYC1 and BlCYC2 are, respectively, 48% and 45% identical to CYC over the entire amino acid sequence. When comparing the TCP and R functional domains, BlCYC1 and BlCYC2 share 94.8% and 90.9% amino acid sequence identity with CYC, respectively, suggesting they are functionally related (Supplementary Material Fig. S1a). When compared with CYC-like genes from other species in Gesneriaceae, BlCYC1 and BlCYC2 are 91% and 86% identical to ObCYC1 and ObCYC2 from Oreocharis, respectively (Supplementary material Fig. S1a). Oreocharis, a genus with weak zygomorphy, is closely allied with Bournea both in morphology and GCYC phylogeny in Gesneriaceae (Li & Wang, 2004; Du & Wang, 2008). Proteins encoded by these CYC-like genes are very similar in the amino acid sequences in the three motifs, TCP domain, R domain and end box (see the Supplementary Material Fig. S1a). The sequence differentiation among these CYC-like genes are mainly located within the nonconserved regions (Fig. S1a). The close phylogenetic relationship between CYC-like genes in Gesneriaceae and CYC in Antirrhinum was shown in previous phylogenetic analyses (Möller et al., 1999; Citerne et al., 2000; Smith et al., 2004; Du & Wang, 2008) (The sequences of BlCYC1 and BlCYC2 are included in the phylogenetic tree in Du & Wang (2008) and alignment in this paper).
Two DIV homologues were isolated from B. leiophylla, which encode two proteins of 295 and 291 amino acids, respectively. We designated these two genes as BlDIV1 and BlDIV2. They are 91% identical at the nucleotide sequence level. At the amino acid level, BlDIV1 and BlDIV2 are 71% and 73% identical to DIV, respectively. All three proteins share a conserved R3 domain (Fig. S1b).
DIV, BlDIV1 and BlDIV2 belong to the atypical plant R2R3-MYB class of transcription factors (i.e. the DIV-like class) (Galego & Almeida, 2002). Distinctively different from the typical plant R2R3MYB protein, DIV-like proteins have different substitutions for the conserved tryptophan residue, longer linkers between R2 and R3 MYB-domains and a specific ‘VASHAQKYF’ motif (Fig. S1b). Sequence alignments revealed that the Arabidopsis At5g58900 has a higher sequence similarity to BlDIV1, BlDIV2, DIV and DVL1 than to any other known R2R3-MYB protein from A. thaliana. These five genes also share the same intron insertion positions and phases (Fig. S1b). In their C-terminal regions, there were some conserved motifs (Fig. S1c). Phylogenetic analyses based on neighbor-joining showed that BlDIV1, BlDIV2, DIV, DVL1 and At5g58900 formed a monophyletic clade with a high bootstrap support (Fig. 3a). Within the clade, BlDIV1, BlDIV2, DIV and DVL1 were further clustered together, sister to At5g58900 with a bootstrap value of 100% (Fig. 3a).
The RAD homologue was isolated from B. leiophylla and it was designated as BlRAD. It shares 69% amino acid identity with RAD from Antirrhinum (Fig. S1d). BlRAD has a single 84aa-MYB-domain conserved among RAD and DIV-like proteins, including DIV and BlDIV proteins (Fig. S1d). Phylogenetic analyses based on neighbor-joining of proteins encoded by RAD-like genes showed that BlRAD was sister to RAD from Antirrhinum with high support and they were further clustered with AtRL1 and AtRL2 proteins from Arabidopsis in a monophyletic clade (Fig. 3b). Other RAD-like genes from Antirrhinum and Arabidopsis that do not function in controlling floral dorsoventral asymmetry were shown to not be closely related to BlRAD (Fig. 3b).
Tissue-specific expressions of BlCYC, BlRAD and BlDIV
Full-length BlCYC, BlRAD and BlDIV cDNA was isolated from developing floral tissues. To elucidate their role in floral development, their temporal and spatial expression patterns were examined during floral development in B. leiophylla using RNA in situ hybridization. Transcripts of BlCYC1 were first detected at the adaxial side of the floral meristem (Fig. 4a). When petals and stamens were just becoming visible, the mRNA of BlCYC1 was specifically detected at the adaxial side of the floral apex inside the adaxial sepal (Fig. 4a). After initiation of petals and stamens at stage 5 in floral development, BlCYC1 was highly expressed in the two adaxial petals and one adaxial stamen up to stage 6 after carpel initiation (Fig. 4b). The BlCYC1 transcripts were then sharply reduced in the adaxial petals and stamen when stamens began to enlarge laterally and differentiate into anthers and filaments at stage 7 or at early stage 8 in floral development, and eventually became barely detectable in the developing flower from stage 8 onward (Fig. 4 c,d).
Similar to BlCYC1, BlRAD transcripts were first detected at the adaxial side of the floral meristem (Fig. 4e). Then, BlRAD was highly expressed in the two adaxial petals and one adaxial stamen up to stage 6 or early stage 7 after carpel initiation (Fig. 4e–f). Like BlCYC1, BlRAD transcripts were then rapidly reduced and barely detectable from stage 8 onward (Fig. 4g,h).
Our results showed that the expression patterns of BlDIV1 and BlDIV2 were indistinguishable during all stages of flower development (data not shown). During stage 1 to stage 3, BlDIV1 and BlDIV2 mRNA accumulated in both inflorescence and floral meristems (Fig. 5a,b). Between stages 4 and 6, when floral organs were initiated, BlDIV1 and BlDIV2 were highly expressed in the primordia of petals, stamens and carpels (Fig. 5b–f). After stage 6, the level of BlDIV1 and BlDIV2 transcripts were sharply reduced in petals and their mRNA accumulation became restricted to the lateral edges of each petal (Fig. 5g–i). Transcripts of BlDIV1 and BlDIV2 were only detected at the lateral edges of each of the five petals at stage 8 (Fig. 5j). While greatly reduced in petals, the transcripts of BlDIV1 and BlDIV2 were still accumulated to a high level in all five stamens after stage 6 but were absent in small patches on the abaxial side (Fig. 5h–i). From stage 9 onward, transcript signal of BlDIV1 and BlDIV2 gradually became weak in petals and stamens (Fig. 5j,k).
To confirm our RNA in-situ hybridization data, we performed RT-PCR analysis of the expression of BlCYC, BlRAD and BlDIV genes using gene-specific primers and dissected petals from mid-to-late stages of developing flowers. The expression of BlDIV1 and BlDIV2 was detected in all petals (Fig. 6), while the expression of BlCYC1 and BlRAD was only observed in the adaxial petals (Fig. 6). The RT-PCR results confirmed our RNA in situ hybridization data. Interestingly, no BlCYC2 transcripts were detected in sepals, petals and stamens of all stages of flowers that we tested (data not shown), suggesting that BlCYC2 may not play a significant role in floral development in B. leiophylla. This is not surprising because we failed to isolate BlCYC2 from cDNA pools prepared from total RNA isolated from flower buds by RACE-PCR.
Early zygomorphy and expression patterns of BlCYC1, BlRAD and BlDIV
The early pattern of BlCYC1, BlRAD and BlDIV expression in B. leiophylla is quite similar to the pattern of CYC, RAD and DIV expression in Antirrhinum (Luo et al., 1996, 1999; Galego & Almeida, 2002; Corley et al., 2005; Costa et al., 2005) (Fig. 7a,d). Similarly, both plants have flowers showing dorsoventral asymmetry at this early stage. In Antirrhinum, even though DIV is transcribed in all petals during early floral development, its early effect on growth is restricted to the abaxial petal and its adjacent region of the lateral petals. This is because DIV is inhibited post-transcriptionally by CYC/DICH through an antagonistic effect of RAD on DIV in the adaxial petals (Galego & Almeida, 2002; Corley et al., 2005) (Fig. 7a). Because of a high degree of sequence homologies and similar patterns of expression during early flower development, it is likely that a similar antagonistic interaction between BlRAD and BlDIV is involved in the growth control of the abaxial petal at the early developmental stages (Fig. 7d). By contrast, the delay in the development of two adaxial petal and one adaxial stamen primordia correlates well with the expression of BlCYC1 and BlRAD being restricted to these organs. Together, our data suggest that BlCYC1, BlDIV and BlRAD, similar to CYC, DIV and RAD in Antirrhinum, play a role in the establishment of the early zygomorphic floral development in Bournea.
Late actinomorphy and BlDIV late specific expression
The later stages of floral development as well as corresponding gene expression differ between Antirrhinum, which results in a zygomorphic flower at anthesis, and Bournea, which is nearly actinomorphic at anthesis. In Antirrhinum, a cyc/dich double mutation does not express RAD in the adaxial region, thus, the abaxial pattern of the late specific expression of DIV in Antirrhinum spreads to all five petals, resulting in an abaxialized peloric flower (Luo et al., 1996, 1999; Galego & Almeida, 2002) (Fig. 7b). In Bournea, both BlCYC1 and BlRAD expressions in the adaxial petals are downregulated after stage 7. Coincidentally, the pattern of BlDIV expression is changed in late floral development in Bournea. We found that the expression of BlDIV genes is restricted to the lateral edges of all five petals from stage 7 onward in Bournea (Fig. 7d). This pattern of expression is similar to the late specific expression of DIV in cyc/dich double mutants in that DIV is no longer restricted abaxially (Galego & Almeida, 2002), but unique in that it is only found in the lateral edges of the petals (Fig. 7d). DIV is mainly expressed in the inner epidermis of the furrow corresponding to the boundary between corolla tube and lobe. Therefore, there are additional differences in the intrapetal spatial patterns with respect to the late specific expression between BlDIV in Bournea and DIV in snapdragon (Fig. 7b,d). The downregulation of BlCYC1 accompanied by BlRAD in the adaxial petals after early floral development is most likely responsible for the late BlDIV specific expression distributed in all five petals, which are correlated with the developmental change from initial zygomorphy to actinomorphy at anthesis in Bournea. Our RNA in-situ hybridization data further indicate that BlDIV transcripts were accumulated in stamen primordia during early development and were continuously accumulated in all five stamens after stage 6 except in small patches on the abaxial side. This pattern of BlDIV expression has not been observed for the corresponding DIV gene in snapdragon. However, it correlates well with the observation that all five stamens are almost equal in size at anthesis (Fig. 7d), suggesting that BlDIV may also play a role in stamens as in petal development in Bournea (Fig. 7d).
Significance of the changed pattern from zygomorphy to actinomorphy in Bournea
As the sister group to the remainder of Lamiales s. l. ( Wortley et al., 2005), Gesneriaceae is characteristic of predominantly weak zygomorphy with four fertile stamens plus an adaxial staminode (Endress, 1999; Cubas, 2004; Li & Wang, 2004). In the derived actinomorphic taxa in Gesneriaceae, some groups have noticeable zygomorphic vestiges, such as Thamnocharis, Niphaea, Phinaea, and Bournea, while some genera have completely actinomorphic flowers, resembling ‘natural pelories’, such as Tengia, Bellonia and Marssonia (Donoghue et al., 1998; Endress, 1998; Smith et al., 2004; Li & Wang, 2004). For example, the adaxial petals and stamen are usually somewhat smaller or shorter than others in Thamnocharis and Bournea, while no residual zygomorphy can be easily observed in the flowers of Tengia (Li & Wang, 2004). Between them, there are some taxa with almost completely actinomorphic flowers, such as Ramonda and Protocyrtandra (Donoghue et al., 1998; Endress, 1998; Citerne et al., 2000; Cubas, 2004). According to Citerne et al. (2000), Ramonda shows no sign of residual zygomorphy when petals and stamens are initiated. The lack of zygomorphy suggests that the CYC homologues in Ramonda may be expressed before floral organ initiation (Citerne et al., 2000). In Oreocharis, which is closely related to Bournea, both in terms of morphology and CYC sequence homologies in Gesneriaceae (Li & Wang, 2004; Du & Wang, 2008), ObCYC1 is expressed in the adaxial petals and stamen (staminode) throughout floral development (Du & Wang, 2008). This is consistent with mature flowers of Oreocharis being characteristic of a weakly zygomorphic (slightly bilabiate) corolla with four didynamous stamens plus a staminode at the adaxial position (Li & Wang, 2004; Du & Wang, 2008). Several other studies have also implicated CYC homologues functioning in the development of floral symmetry in Gesneriaceae. The genus Chirita develops zygomorphic flowers with bilabiate corolla, two fertile abaxial stamens and three staminodes at the adaxial and lateral positions. In Chirita heterotricha, a CYC homologue ChCYC1D is expressed in the adaxial floral region, while the expression of another copy ChCYC1C is expanded from the adaxial to the lateral region, correlating with the abortion of both the adaxial and lateral stamens (Gao Qiu, unpublished). This expression pattern is similar to McCYC/McDICH expression in the flower of Mohavea, which aborts both the adaxial and lateral stamens (Hileman et al., 2003). In Sinningia speciosa, a zygomorphic species in Gesneriaceae, its actinomorphic cultivars carry a frame-shift mutation in the single Sinningia CYC homologue (GCYC) that yields a truncated protein, suggesting that the actinomorphic flowers may be caused by the mutation of GCYC (Citerne et al., 2000; Cubas, 2004). In Arabidopsis thaliana, a model eudicot species with ancestral actinomorphic flowers, a CYC homologue TCP1 is transiently expressed at the adaxial region of the floral meristem that prepatterns the dorsoventral asymmetry (Cubas et al., 2001). In the basal eudicot family Papaveraceae sensu lato, it has been shown that the duplication and diversification of CYC-like TCP genes is accompanied by alterations in expression patterns, some of which play a role in floral symmetry as the adaxial identity genes (Damerval et al. 2007). Given that CYC-like genes are widely conserved in the adaxial identity function in eudicots and the expression patterns and loss-of-function of CYC-like genes are correlated with both natural and mutant phenotypes in Gesneriaceae, we come to the conclusion that CYC homologues in Gesneriaceae also function as adaxial identity genes in establishing the floral dorsoventral asymmetry. Therefore, the initial zygomorphic pattern of floral development in Bournea may be interpreted as a residual zygomorphy that results from the conserved pattern of early BlCYC1 and BlRAD expression. The key novel event seen in Bournea is the downregulation of BlCYC1 accompanied by BlRAD after early floral development and the correlative changes in the late specific expression of the abaxial identity gene BlDIV that is distributed in lateral edges of all five petals and stamens, which should be responsible for the origin of the derived actinomorphy in Bournea.
The epimutant flowers of L. vulgaris caused by heavy DNA methylation of LCYC provide an example that the abaxialized actinomorphic flower in nature can come about via the same mechanism that underlies the cyc/dich double mutant phenotype in Antirrhinum (i.e. loss-of-function of adaxial identity TCP genes) (Cubas et al., 1999a). The study in Cadia indicates that an adaxialized mutant resulting from ectopic expression of CYC homologues can also occur in natural species (i.e. the adaxial identity gene, legCYC1B expression distributed in all five petals) (Citerne et al., 2006) (Fig. 7c). However, the floral actinomorphy in Cadia is established in consequence of the late downregulation rather than complete loss-of-function of another copy, legCYC1A (Citerne et al., 2006). Our findings provide another example that the downregulation of a CYC homologue triggers the developmental change from initial zygomorphy to the actinomorphic flower in natural species. The difference between Bournea and Cadia lies in the late specific expression of the abaxial identity gene (BlDIV) (Fig. 7d), rather than the expression of the adaxial identity gene (legCYC1B) (Fig. 7c), distributed in all five petals and stamens, which has been previously observed only in the model species snapdragon upon complete loss-of-function of CYC-like genes (Fig. 7b). The expression data in Bournea, along with previous reports, suggest that there might be diverse pathways in the origin of derived actinomorphy through modifications of a pre-existing zygomorphic developmental program. A special organ type usually involves dynamics of the gene-regulatory interaction in elaborate networks of positive and negative factors that intersect at various levels to regulate floral morphogenesis, finally creating a distinct floral architecture and form (Krizek & Fletcher, 2005). Our results further shed light on the evolution of actinomorphy from zygomorphy dependent upon the spatial and temporal changes in gene-regulatory interactions between RAD and DIV homologues promoted by CYC-like genes in floral development. Different shifts from zygomorphy to actinomorphy may depend on the timing of when the expression of CYC-like genes is downregulated in floral development, attributed to the degree of conserved early expression pattern in different lineages switching from zygomorphy to actinomorphy in Lamiales s. l., especially in Gesneriaceae. Further investigation of the expression patterns and functional analyses of the key regulatory genes in additional taxa with derived actinomorphy, as well as identification of upstream primary genes responsible for the spatiotemporal switches of expression of CYC genes promises to shed new light on mechanisms that underlie the vast morphological diversification of derived actinomorphy in Lamiales s. l.
We thank Yu-Xin Hu, Song Ge, Da Luo for their help and critical comments on the manuscript. This work was supported by National Natural Science Foundation of China Grant, nos. 30121003, 30740001, KSCX2-YW-R-135.