Background

The ABC model of floral development, which describes the role of three classes of organ identity genes in the spatial arrangement of floral organs, was proposed in accordance with studies in two model core eudicots, Arabidopsis thaliana and Antirrhinum majus[1, 2]. In the determination of reproductive organs, expression of C-function genes alone specifies carpels and expression of both C-function and B-function genes gives rise to stamens. In the outer perianth whorl, sepal identity is determined by the lack of B-function and C-function genes. In the inner perianth whorl, the lack of C-function genes and the activity of B-function genes are required for determination and maintenance of petal identity [3]. All previously described B-function genes belong to the family of MADS-box containing transcription factors.

While the ABC model is well known, little is known about the evolutionary events that have led to the striking structural diversity among petaloid organs, including sepal or petal loss and transfer of petaloid characters between floral whorls. The presence of homeotic petaloid organs in other plant groups, such as the Ranunculales, monocots, magnolids and basal angiosperms, have been shown to be correlated with ectopic expression of B-function MADS box genes [4–7]. It has been suggested that similar core eudicot examples of apparent divergence from the standard condition, in which petaloid organs are found only in the second floral whorl, may be associated with heterotopic expression of B-function genes. However, these systems have remained inadequately studied.

Within the core eudicots, perianth organization rarely deviates from the standard sepal-petal format. In cases where ectopic petaloid organs do occur in core eudicots, the organs remain morphologically differentiated from petals, as compared with the identical perianth whorls that can often be observed in the monocots. Furthermore, petaloid organs which occur outside of the inner perianth whorl, such as petaloid bracts in dogwood (Cornus), the petaloid sepal in Impatiens, and sepal-derived and stamen-derived petaloid organs in the Caryophyllales, have revealed employment of alternate petal identity programs in the formation of petaloid organs, discounting simple extension of B-class gene expression in these systems [8–11]. Similarly, some monocots have been shown to lack expression of B-function genes in the outer tepals [12].

The Hawaiian endemic lobelioids, Campanulaceae, comprise 6 genera and 126 species and represents the largest plant radiation on any ocean island archipelago. Clermontia, a genus within the lobelioids, includes 22 species of bird-pollinated trees and shrubs possessing morphologically diverse flowers [13]. Clermontia flowers exhibit inferior ovaries and produce fleshly fruits. A drastic homeotic phenotype, in which sepals are replaced by an extra whorl of organs that appear to be true petals, characterizes 15 of the 22 species: the double-corolla species. This petal-petal format is a unique trait within the standard groundplan, consisting of a sepal-petal perianth, of the core eudicots. For most double-corolla Clermontia species, the sepal-to-petal transformation is a complete one-for-one conversion. C. arborescens demonstrates the ancestral condition, displaying an outer perianth whorl of small green sepals and larger, showier petal organs in the inner whorl (Figure 1A). The identical first and second whorl organs present in C. parviflora, both of which are petaloid, reveal the double-corolla phenotype (Figure 1C). Two other Clermontia species are also figured to make the point about the double-corolla phenotype: C. fauriei (Figure 1B), with tiny sepals and one petal whorl, is otherwise very similar to C. oblongifolia (Figure 1D), which bears only petal-like organs.

Figure 1
figure 1

Groundplan diversity in representative Clermontia species. Floral morphology of the standard sepal-petal species (A) Clermontia arborescens and (B) C. fauriei and double-corolla species (C) C. parviflora and (D) C. oblongifolia are shown. C. arborescens and C. fauriei exhibit small green sepals (s) in the outer perianth whorl and five large showy petals (p5) in the inner perianth whorl. C. parviflora and C. oblongifolia bear 10 petal-like organs (p10) in both perianth whorls. Photos courtesy of Anne-Cathrine Scheen (A), John Game (B) and Jupiter Nielsen (C,D).

Full size image

The genus Clermontia is endemic to the Hawaiian Islands, which allows for a system that can be studied on a recent and established evolutionary timescale. The Hawaiian Islands were formed by movement of the Pacific plate over a mantle plume; this resulted in islands that emerged from volcanic eruptions and erode in a southeast to northwest geographic progression. Further, unstable environments may allow for increased ecological viability of drastic groundplan differences that may not necessarily create any immediate increase in fitness. This may account for the success of a morphological radiation where a drastic groundplan difference is caused by small regulatory changes. The naturally occurring homeotic mutation in Clermontia offers an appropriate model for studying the potential role of altered B-function gene expression contributing to heterotopic petaloid organs within the core eudicots [14], and provides an example that may have general implications for the study of island radiations.

Methods

Scanning electron microscopy

RNAlater-preserved samples (Applied Biosystems, Carlsbad, CA, USA) were dissected to separate inner and outer perianth organs. Tissues were dehydrated in a graded ethanol series and chemically dried with hexamethyldisilazane as described in Costea et al. [15]. Samples were mounted on stubs, sputter coated with evaporated carbon, and micrographs were taken using a Hitachi-S-4000 scanning electron microscope (Hitachi, Ltd., Tokyo, Japan). Both adaxial and abaxial surfaces were analyzed.

Sequence generation and phylogenetic analysis

Genomic DNA was extracted from herbarium- or silica-preserved vegetative material using Qiagen DNeasy Plant Kits (Qiagen, Valencia, CA, USA). Isolation was performed with the use of universal 5S ribosomal DNA non-transcribed spacer (5S-NTS) primers previously described in Cox et al. [16]. Amplicons were cloned into the PDrive cloning vector (Qiagen) and five to eight colonies per sample were sequenced. 5S-NTS sequences were submitted to GenBank and have accession numbers [GenBank:JQ734775] to [GenBank:JQ734917]. B-function MADS box genes that were not obtained from 454 transcriptome sequencing (KAH and VAA, unpublished results) were isolated using RT-PCR with previously reported primers [17]. To gain insight into the relationships among the 5S-NTS sequences, a maximum likelihood phylogenetic tree was constructed. Similarly, a maximum likelihood tree was generated comprising the two Clermontia PI homolog duplicates as well as PI homologs from selected other taxa, including Campanulaceae, Asteraceae and outgroups. Sequences were aligned using MUSCLE with default settings [18]. The sequence alignments, with complete taxon names, voucher information (where available) and GenBank accession numbers are available in Additional files 1 and 2. A single most optimal tree for each data set was computed using the RaxML BlackBox web server (http://phylobench.vital-it.ch/raxml-bb/) running RaxML version 7.2.8 [19]. Default settings were used with the general time reversible plus gamma model of molecular evolution. One hundred bootstrap samples were generated to assess support for the inferred relationships. Local bootstrap values (in percentages) are indicated for branches with >50% support.

Quantitative polymerase chain reaction

RNAlater-preserved tissues (except for young leaves, these were floral organs from buds in various pre-anthesis stages) were dissected into organ groups as needed, total RNA was isolated using a Qiagen RNeasy Plant Kit, and mRNA was subsequently isolated using a Qiagen Oligotex mRNA Kit. Reverse transcription of mRNA with a Bio-Rad iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA) was performed to synthesize cDNA. Each reaction comprised 20 ng of cDNA. Primers were designed from sequences of B-function and other MADS box genes discovered either by transcriptome sequencing (KAH and VAA, unpublished data) or isolated by RT-PCR and are listed in Additional file 3. Primers for each PI-like homolog yielded unique sequences when checked for specificity. Amplification and real-time detection was performed using Bio-Rad iQ SYBR Green Supermix on an MyiQ2 Real-Time Detection Thermal Cycler (Bio-Rad). Standardization was performed using β-actin to determine delta-delta critical threshold values. Bars represent standard deviations based on means of three independent experiments, each of which included two replicates. Gene expression is presented in as fold differences relative to expression in the highest expressing organ in each gene category, which was set to 10.

In situ hybridization

Clermontia flower buds were fixed with formalin-acetic acid-alcohol (FAA), embedded in paraffin and sectioned using standard methods. In situ hybridizations were performed essentially as described in Ruonala et al. [20]. For the post-hybridization washes, an InsituPro VsiVSi 3.0 (Intavis AG, Koeln, Germany) device was used with the protocol shown in Additional file 4. The color reaction to visualize hybridized probe was done for approximately 24 hours at room temperature. Sections were photographed using an AxioImagerZ1 (Zeiss, Munich, Germany) equipped with AxioCam MRc (Zeiss). The digoxigenin-labeled probe covered 430 bp of the Clermontia PI coding region, excluding the MADS domain. Primers used to amplify the Clermontia PI probe fragment were: forward, 5′ GCATGAGTACTGTAGCCCTTCC 3′; reverse 5′ GAAGAGTGGGGAAGGAAGATTT 3′.

Results

Perianth epidermal morphology

Scanning electron microscopy experiments were performed on the adaxial and abaxial surfaces of organs located in the outer two floral whorls for standard sepal-petal and double-corolla flowers. Cell surface morphologies of first and second whorl organs of the double-corolla species Clermontia parviflora revealed conical epidermal cells on the adaxial surfaces of both first and second whorl petaloid organs, strongly suggesting a homeotic conversion in the former (Figure 2A,C). While the inner whorl of the standard bipartite perianth species, C. arborescens, revealed conical epidermal cells on the adaxial surface, the outer whorl epidermal cells lacked the well-defined conical shape; such flattened, irregular cell morphologies are typical for sepal organs (Figure 2E,G).

Figure 2
figure 2

Epidermal micromorphology of perianth organs. Scanning electron microscope images from (A-D) C. parviflora and (E-H) C. arborescens on the adaxial (left panels) and abaxial (right panels) epidermal surfaces. Canonical Conical epidermal cell morphology, an indicator of petal identity, is discovered are apparent on the adaxial surface of the outer perianth organs of C. parviflora (A). Similar cone-shaped cells are observed on the adaxial surface of inner perianth organs in both C. parviflora (C) and C. arborescens (G). However, the outer perianth organs of C. arborescens display a flattened morphology more typical of sepal organs (E). Scale bar denotes 100 μm.

Full size image

Phylogenetic analysis

Phylogenetic analysis of 5S-NTS sequences shows clear divergence of Clermontia from outgroup genera Brighamia and Delissea, and Clermontia’s sister genus Cyanea, and demonstrates extremely low sequence divergence among Clermontia species (Figure 3). The single-corolla species C. fauriei, the only species found on the oldest island, Kauai, forms a significantly distinct group from other Clermontia, which include both single- and double-corolla species (Figure 3 and Additional file 5). This indicates a probable single and geologically recent origin of the double-corolla trait within the genus, likely by the time of the emergence of Oahu (circa 3.5 million years ago), with numerous potential reversals to the standard sepal-petal format. Oahu holds two species with the double-corolla phenotype, C. persicifolia and C. oblongifolia, the former of which is endemic to that island. 5S-NTS sequence details and an alignment are presented in Additional files 1 and 6.

Figure 3
figure 3

Schematic representation of phylogenetic analysis of 5S-NTS ribosomal DNA sequences. Outgroups Brighamia and Delissea are displayed along with Clermontia’s sister genus Cyanea. C. fauriei (f) exhibits the standard sepal-petal groundplan and forms a distinct group separate from all other Clermontia species, including both sepal-petal and double-corolla species. Red indicates sequences derived from double-corolla DNA samples.

Full size image

Expression analysis

qPCR analysis performed on homologs of the B-function genes PISTILLATA (PI), APETALA3 (AP3), and TOMATO MADS 6 (TM6) was used to elucidate expression patterns in outer perianth, inner perianth, stamen, carpal, vegetative apices and hypanthium (inferior ovary) tissues of the double-corolla species Clermontia parviflora and the standard sepal-petal species C. arborescens. This analysis indicated ectopic expression of the two PI paralogs in the first whorl of C. parviflora; no such drastic homeotic expression was observed for the other two genes, although AP3 expression appears to be slightly elevated relative to C. arborescens (Figure 4). Transcriptional analysis of other Clermontia MADS box gene homologs, SHORT VEGETATIVE PHASE (SVP), SUPPRESSOR OF OVEREXPRESSION OF CONSTANS (SOC1), AGAMOUS-LIKE 6 (AGL6) and SEPALLATA3 (SEP3), which are involved in various floral and non-floral functions [21–24], was also performed. No remarkable differences were found between expression patterns in these genes between sepal-petal and double-corolla floral organs (Additional file 7). In C. arborescens, detection of PI transcripts was restricted to the inner perianth organs and stamens, with no ectopic expression of PI homologs being observed in whorl 1, which is consistent with eudicot-standard lack of B-function genes in sepals (Figure 4). No expression was detected for either of the PI homologs in the carpel organs of C. parviflora (Figure 4), indicating that the up- versus downregulation of the PI homologs is precisely constrained to the perianth and stamen whorls, excluding a simple overexpression phenotype that would be expected to engender a pleiotropic phenotype.

Figure 4
figure 4

Quantitative gene expression of B-class genes in dissected floral organs. Expression of both PISTILLATA (PI) homologs is constrained to whorls 2 and 3 of Clermontia arborescens (top two right panels) but is extended to whorls 1, 2 and 3 in C. parviflora (top two left panels). No substantial difference in expression pattern is observed in other B-function genes, AP3 and TM6 homologs (bottom two panels). β-actin was used as an internal control. Bars indicate mean and standard deviation for three independent experiments each including two replicates. 1, outer perianth; 2, inner perianth; 3, stamen; 4, carpel; L, leaf; H, hypanthium.

Full size image

In situ hybridization was used to study expression of both PI homologs in primordial stages of floral organ development of the double-corolla species Clermontia parviflora. Consistent with qPCR data, this analysis showed ectopic expression of PI homologs in the outer perianth whorl of developing double-corolla flower buds, as well as in the petal and stamen whorls, as expected (Figure 5).

Figure 5
figure 5

In situ analysis of PISTILLATA expression in flower primordia of Clermontia parviflora. In situ hybridization on flower primordia of C. parviflora displays expression of PI homologs in (A) early stage outer perianth organs and (B and C) in both perianth organs and whorl 3 in a late-stage bud. (D) Sense PI homolog controls show low background staining. Images are displayed at 20× (A), 10× (B and D) and 40× (C). Organs are numbered by whorl; b = bract.

Full size image

Discussion

The petals of angiosperm flowers, laminar organs occupying the second floral whorl, often share characteristics such as expression of B-function MADS genes, showy morphology, pigmentation and production of aromatic compounds. However, petals exhibit substantial diversity and organs with petaloid characteristics can be found outside of the second floral whorl, making the definition and identification of these organs rather challenging. The adaxial epidermal surfaces of petals are frequently composed of conical cells whereas those of other floral and vegetative cells are typically not [25]. Identification of petaloid organs has been determined by observation of conical epidermal cells in cases of homeotic conversions involving petals [26, 27]. Analysis of micromorphological characteristics on the adaxial epidermal surfaces of first-whorl organs in the double-corolla species Clermontia parviflora reveals conical cells, a marker of petal identity. This supports the hypothesis of homeotic transformation of sepals to petals.

Based on the phylogenetic analysis presented here and the geographic distribution of Clermontia, the origin of this homeotic transformation was likely established via a single and geologically recent occurrence. The Hawaiian Islands were formed by the northwestward movement of the Pacific tectonic plate over a fixed volcanic plume, resulting in a layout where the islands progress from oldest to youngest in a northwest to southeast manner [28]. C. fauriei is the only species found on Kauai, the oldest of the Hawaii Islands, and it displays the ancestral sepal-petal format. The neighboring island Oahu is home to five species, including C. fauriei and four double-corolla species [29]. 5S-NTS data support previous findings suggesting Clermontia’s substantial sequence divergence from its sister genus Cyanea but low sequence divergence within the genus [30]. Our data do indicate sequence divergence between C. fauriei and all remaining Clermontia species, including both standard sepal-petal and double-corolla species, perhaps reflective of a split age at least by the time of the origin of Kauai. The phylogenetic clustering of ancestral state and petal-petal format species observed suggests multiple potential reversals occurred during the radiation of Clermontia. Similar findings on Hawaiian lobelioid inter-relationships, including evolution of reversals in the Clermontia clade, have been reported by Givnish and collaborators [13, 31, 32].

The transformation of first-whorl sepal organs into organs bearing petal identity in double-corolla Clermontia species, which is often complete, led us to question whether this phenomenon may be regulated by ectopic expression of B-class genes. Our detection of ectopic expression of PI homologs in floral primordia indicates a likely role of PI in the differentiation and determination of outer whorl floral organs, resulting in the development of petal identity. Lacandonia schismatica (Triuridaceae) exhibits a homeotic transformation in which central stamens are surrounded by carpels. The simple displacement of the B-function has been shown to play a role in this morphological shift [33]. In dove tree (Davidia), early B-class gene expression in petaloid bracts has been suggested to lead to a partial petaloid phenotype; however, the lack of late-stage expression makes the mechanism unclear [34]. In Clermontia, not only is expression of PI homologs detected in floral primordia, but we have also shown continued expression in late-stage outer whorl floral organs, supporting the function of B-class genes in the maintenance of homeotic petal identity. Furthermore, we show that expression of a variety of other MADS-box gene homologs involved in various floral and non-floral functions, specifically AP3, TM6, SEP3, AGL6, SVP and SOC1, present expression patterns that are largely or completely consistent between standard sepal-petal and double-corolla species (Additional file 7).

Expression of a Clermontia AP3 homolog is detected in the outer whorl of both standard groundplan species and double-corolla species. Therefore, we hypothesize that ectopic expression of PI homologs would be sufficient to induce the obligate AP3-PI heterodimer autoregulatory feedback loop. In Arabidopsis, expression of PI alone is not able to induce petaloidy in vegetative organs; however, it is when co-expressed with AP3 in vegetative organs or expressed alone in outer whorl perianth [3]. The expression of an AP3 homolog in the outer whorl perianth in Clermontia indicates that the genetic background condition required for ectopic PI expression to induce petaloidy is present. SEP3 is not able to induce petaloidy on its own, but has been shown to increase petaloid characteristics of ectopic petaloid organs, and the heterodimer AP3-PI forms a ternary complex with SEP3 [35, 36]. We hypothesize that in Clermontia, low-level expression of SEP3 and AP3 homologs in the ancestral condition would set up a context in which ectopic expression of PI would be enough to induce petal identity in the outer whorl perianth.

Gene duplications among B-class genes, resulting in the euAP3, TM6 and PI groups, have been shown to be of evolutionary significance in the production of floral diversity. It has been suggested that extension of the B-class gene model has played a diversity-generating role in cases with a history of gene duplication among PI homologs or AP3 homologs in taxonomic groups with otherwise undifferentiated first and second whorls [37]. Here, we demonstrate ectopic and sustained expression of both Clermontia PI homolog duplicates, which clearly derive from a lineage-specific duplication within Campanulaceae (Additional file 8). This apparent subfunctionalization event following gene duplication may have played a key role in the events following PI duplication in Clermontia. The two homologs may be largely functionally redundant, or one may be expressed first and induce the expression of the other, consistent with an autoregulatory feedback loop when obligatory dimerization with AP3 occurs.

The precise regulation that restricts ectopic expression of PI homologs to the first-whorl primordia in early and late-stage organs, as seen in Clermontia parviflora, indicates that small regulatory changes may be responsible for the dramatic change in the floral groundplan established in eudicots. This tight spatiotemporal regulation allows for the deployment of a drastic homeotic mutation without causing potentially deleterious pleiotropic effects, such as transformation of the carpel whorl into stamens. In Arabidopsis, ectopic expression of B-function genes has been shown to be sufficient to transform the carpel whorl into staminoid organs [3]. The specific regulation of this heritable mutation is necessary to produce viable organisms, since simple overexpression would be likely to cause infertility through the production of staminoid carpels. Other viable homeotic transformations within the angiosperm groundplan have occurred but have not led to radiations as seen in Clermontia.

The double-corolla phenotype may not have initially acted with adaptive advantage, its success perhaps relying instead on passive expansion of the mutation by random genetic drift in small populations in the unstable environments of the Hawaiian Islands. The sepal-petal ancestral status of Clermontia is supported by our phylogenetic analysis and the uniformly eudicot-standard floral morphology of its sister genus Cyanea and further closely related outgroup lobelioid species. While phylogenetic analysis demonstrates multiple possible reversals to the ancestral condition, the radiation of the double-corolla mutation indicates the long-term viability of the mutation. This gives insight into possible mechanisms that may have shaped diversity among floral groundplans in past unstable environments; small regulatory changes, as opposed to large coding-sequence differences, could account for morphological differentiation in other island groups such as the Hawaiian silverswords, Hawaiian mints and Canarian Crassulaceae.

Conclusions

Our morphological and gene expression data strongly suggest that a drastic and heritable phenotypic change, at the level of the floral groundplan, can originate from a homeotic mutation that is likely regulatory, being under precise spatiotemporal control as opposed to having pleiotropic characteristics. The uniqueness of this trait among core eudicots could be linked to increased ecological viability in an unstable island environment, a chance event that need not have posed any immediate adaptive benefit. We argue that the evolutionarily young morphological radiation of Clermontia may form a model system for general understanding of mechanisms of larger-scale angiosperm diversification in past, similarly unstable environments, in which small regulatory changes may have been responsible for modern-day groundplan differences.