Functional diversification of duplicated CYC2 clade genes in regulation of inflorescence development in Gerbera hybrida (Asteraceae)



The complex inflorescences (capitula) of Asteraceae consist of different types of flowers. In Gerbera hybrida (gerbera), the peripheral ray flowers are bilaterally symmetrical and lack functional stamens while the central disc flowers are more radially symmetrical and hermaphroditic. Proteins of the CYC2 subclade of the CYC/TB1-like TCP domain transcription factors have been recruited several times independently for parallel evolution of bilaterally symmetrical flowers in various angiosperm plant lineages, and have also been shown to regulate flower-type identity in Asteraceae. The CYC2 subclade genes in gerbera show largely overlapping gene expression patterns. At the level of single flowers, their expression domain in petals shows a spatial shift from the dorsal pattern known so far in species with bilaterally symmetrical flowers, suggesting that this change in expression may have evolved after the origin of Asteraceae. Functional analysis indicates that GhCYC2, GhCYC3 and GhCYC4 mediate positional information at the proximal–distal axis of the inflorescence, leading to differentiation of ray flowers, but that they also regulate ray flower petal growth by affecting cell proliferation until the final size and shape of the petals is reached. Moreover, our data show functional diversification for the GhCYC5 gene. Ectopic activation of GhCYC5 increases flower density in the inflorescence, suggesting that GhCYC5 may promote the flower initiation rate during expansion of the capitulum. Our data thus indicate that modification of the ancestral network of TCP factors has, through gene duplications, led to the establishment of new expression domains and to functional diversification.


Asteraceae, the sunflower family, is among the largest families of plants including approximately 10% of all angiosperm species. Typical for the Asteraceae is a highly compressed inflorescence, the capitulum, which superficially resembles a solitary flower but is composed of tens to hundreds of flowers. In homogamous (unisexual) capitula all flowers are similar whereas heterogamous (hermaphroditic) capitula harbour showy, marginal ray flowers that differ morphologically and functionally from the centrally located, less conspicuous disc flowers. The ray flowers with their large corolla ligules are bilaterally symmetrical (zygomorphic/monosymmetric) and sterile or female, while the disc flowers are radially symmetrical (actinomorphic/polysymmetric) and develop pollen-producing stamens in addition to carpels. The presence of the showy marginal flowers in heterogamous heads has been shown to be associated with pollination specialization, outcrossing rate, genetic diversity and fitness, and is the likely reason for the evolutionary success and rapid tribal radiation of this large plant family (Marshall and Abbott, 1984; Stuessy et al., 1986; Sun and Ganders, 1990; Endress, 1999; Cubas, 2004; Sargent, 2004; Andersson, 2008). Classical genetic studies have suggested that the presence or absence of ray flowers is under simple genetic control involving one or two major genes and an unidentified number of modifier genes (reviewed in Gillies et al., 2002).

Studies in the model species gerbera (Gerbera hybrida), groundsel (Senecio vulgaris) and sunflower (Helianthus annuus) have brought insight into the molecular control of flower type identity in Asteraceae by implicating the involvement of CYCLOIDEA/TEOSINTE BRANCHED1-like (CYC/TB1) TCP domain transcription factors in the establishment and generation of flower-type heterogeneity (Broholm et al., 2008; Kim et al., 2008; Fambrini et al., 2011; Chapman et al., 2012; Tähtiharju et al., 2012). Members of the CYC/TB1 subfamily of TCP proteins are considered as general developmental regulators of reproductive and vegetative axillary structures, i.e. flowers and lateral shoots (Martín-Trillo and Cubas, 2010). Phylogenetic analysis has revealed that in core eudicots the CYC/TB1 (or ECE) clade has experienced gene duplications and can be divided into three subclades, CYC1, CYC2 and CYC3 (Howarth and Donoghue, 2006). Early studies in Antirrhinum majus showed that the CYC2 clade genes CYCLOIDEA (CYC) and its paralogue DICHOTOMA (DICH) function partially redundantly in establishing the identity of the dorsal domain of the flower and to generate asymmetry (Luo et al., 1996, 1999). Later, numerous studies across rosids and asterids indicated that CYC2 clade genes have been recruited several times independently for the common genetic function of regulating bilateral symmetry of single flowers (reviewed by Busch and Zachgo, 2009; Rosin and Kramer, 2009; Martín-Trillo and Cubas, 2010; Preston et al., 2011). In Asteraceae, the CYC/TB1-like gene family has expanded, and in both gerbera and sunflower 10 gene family members have been identified (Chapman et al., 2008; Tähtiharju et al., 2012). In particular, the duplicated CYC2 clade genes have been shown to be associated with complex capitulum structure and to regulate differentiation of flower type.

Our previous studies in gerbera have focused on characterization of a single CYC2 clade gene, GhCYC2 (Broholm et al., 2008). GhCYC2 expression followed the radial organization of the inflorescence, being upregulated in the marginal ray flowers while no expression was detected in the centremost disc flowers. Broholm et al. (2008) demonstrated that ectopic expression of GhCYC2 in transgenic gerbera converted disc flowers into ray-like ones with elongated petals and disrupted stamen development. Kim et al. (2008) showed that the RAY locus, which controls the radiate condition of Senecio vulgaris flower heads, corresponds to two closely linked CYC2 clade genes that also show specific upregulation in the marginal ray flower primordia. In sunflower, chrysanthemoides (chry) or double (dbl) mutants bear flowers that are solely of ray identity (Fambrini et al., 2003; Berti et al., 2005). In contrast, in the tubular ray flower (turf) and tubular-rayed (tub) mutants, marginal flowers are almost radially symmetrical and develop male and female reproductive organs, as do wild-type disc flowers (Berti et al., 2005; Chapman et al., 2012). These mutants show an altered expression pattern of the CYC2 clade gene HaCYC2c (Fambrini et al., 2011; Chapman et al., 2012). The turf mutation is caused by an insertion of a transposable element in the HaCYC2c TCP motif that leads to a premature stop codon (Fambrini et al., 2011; Chapman et al., 2012), causing reduced expression and radialization of the normally zygomorphic ray flowers (Chapman et al., 2012). In the dbl mutant, due to an insertion in the promoter region, HaCYC2c is expressed throughout the capitulum, converting disc flowers into ray-like flowers (Chapman et al., 2012).

Our previous study (Broholm et al., 2008) demonstrated that although overexpression of GhCYC2 converted disc flowers into ray-like ones, suppression of GhCYC2 expression was not sufficient to cause loss of ray flower identity. This indicates that there must be additional genes involved in the regulation of flower type differentiation. Comparative analysis of the gene family in gerbera and sunflower showed that all six CYC2 clade genes in gerbera and five in sunflower are upregulated during early stages of ray flower primordia (Tähtiharju et al., 2012). The expression analysis also suggested that CYC2 clade genes may have more specialized functions at the level of single flowers, including late functions in reproductive organs (Tähtiharju et al., 2012). Pairwise protein–protein interaction assays further indicated that most CYC/TB1-like proteins in gerbera and sunflower show competence to form higher-order protein complexes (Tähtiharju et al., 2012). In this study we extended the analysis of the gerbera CYC2 clade genes to localize their expression at the level of single flowers. Functional studies in an Arabidopsis background indicate that the CYC2 clade proteins have highly conserved biochemical functions. Still, ectopic expression in gerbera shows evidence for both functional redundancy and diversification among the gene family members, suggesting that functional specificity for CYC2 clade proteins is obtained by the formation of context-specific protein complexes involving CYC2 proteins and their co-regulators that may target different downstream genes.


Gerbera CYC2 clade genes show highly overlapping expression patterns

Our earlier results indicated that the CYC2 clade genes share strikingly similar expression patterns during the development of ray and disc flower primordia. Quantitative real-time PCR (qPCR) for excised floral primordia at early developmental stages indicated that all CYC2 clade genes, with the exception of GhCYC7, are predominantly expressed in developing ray flower primordia and are lacking from the centremost, most strongly actinomorphic disc flowers (Tähtiharju et al., 2012). In order to localize the expression domains in more detail at the organ and tissue level, we performed in situ hybridization analysis.

Our previous qPCR data indicated that GhCYC3 is the only CYC2 clade gene that is upregulated in ray flower primordia and is absent from both trans and disc flower primordia (Tähtiharju et al., 2012). In situ hybridization confirmed the abundant expression in ray flowers but we also detected expression in petals and carpels of outermost disc flowers during later developmental stages (Figure 1). In contrast to the previously characterized GhCYC2 gene, GhCYC3 expression was absent from the rudimentary stamens in ray flowers but showed expression in both the ventral ligule (formed of three fused petals) and in the two dorsal petals (Figure 1). GhCYC4 showed an identical expression pattern to GhCYC3 in ray flowers (Figure 1). The expression of GhCYC9, which is the recently duplicated paralogue of GhCYC4 (Tähtiharju et al., 2012), was below the level that could be detected by in situ analysis (Figure S1a,b in Supporting Information). Expression of GhCYC5 and GhCYC7 was detected in all whorls of organs, but the latter showed relatively weak expression in 12-mm diameter inflorescences (Figure 1). The expression of GhCYC7 was more prominent at an earlier developmental stage, in inflorescences of 3–4 mm in diameter, where its expression was ubiquitously localized in the undifferentiated inflorescence meristem and young developing flower primordia (Figure S1c,d). In the outermost disc flowers, all CYC2 clade genes were expressed in petals and carpels, and GhCYC5 was additionally expressed in pappus bristles (whorl 1) (Figure 1). In conclusion, with the exception of GhCYC2 being absent from dorsal petals (Broholm et al., 2008), the CYC2 clade genes were redundantly expressed in all five petals as well as in carpels in both ray and disc flowers. The ray flower stamen primordia lacked GhCYC3 and GhCYC4 expression, while GhCYC2, GhCYC5 and GhCYC7 expression was present.

Figure 1.

In situ expression analysis of CYC2 clade gene family members. The panels on the left show the expression patterns in the marginal ray flowers (RF) and those on the right in the outermost disc flowers (DF) of gerbera inflorescences (diameter 12 mm). dPe, dorsal petal; vLi, ventral ligule; ruSt, rudimentary stamen; Ca, carpel; St, stamen; Pe, petal; Pa, pappus bristles. Scale bars: 100 μm. (a, b) GhCYC3, (c, d) GhCYC4, (e, f) GhCYC5, (g, h) GhCYC7 and (i, j) sense control gene expression patterns in ray flowers and disc flowers, respectively.

We also investigated temporal expression of the CYC2 clade genes during later stages of petal development in ray flowers. Analysis of pooled ray flower petal samples from Stages 2, 4, 6 and 8 (stages described in Helariutta et al., 1993; Kotilainen et al., 1999; Laitinen et al., 2007) revealed that the CYC2 clade genes GhCYC2, -3, -4 and -5 showed the highest expression levels; GhCYC9 showed only a very low signal and GhCYC7 was not expressed (Figure 2a) and were excluded from further analysis. Analysis of petal samples from ray flowers at developmental Stages 1–11 showed that the expression of GhCYC3 peaks at Stage 1 and then gradually decreases until Stage 9 when the inflorescence is already fully open and the petals have achieved their final size and shape (Figure 2b). The other CYC2 clade genes, GhCYC2, GhCYC4 and GhCYC5, are expressed at lower levels in petals. GhCYC2 and GhCYC4 are expressed quite uniformly throughout development, while the expression of GhCYC5 peaks at late Stage 9.

Figure 2.

Gene expression of gerbera CYC2 clade genes in ray flower petals. (a) Relative gene expression of gerbera CYC2 clade genes in pooled ray flower petal samples. The expression levels were calculated relative to their expression in a root sample. (b) Expression of GhCYC2, GhCYC3, GhCYC4 and GhCYC5 during ray flower petal development (Stages 1–11). Bars show the mean expression of two biological replicates. Expression levels of GhACTIN were used for normalization.

Heterologous expression in Arabidopsis indicates a conserved biochemical function for the gerbera CYC2 clade proteins

For functional analysis, all six gerbera CYC2 clade genes were ectopically expressed in Arabidopsis thaliana. For comparison, we also generated plants that ectopically expressed the orthologous Arabidopsis gene TCP1. At the seedling stage (7 days after germination, DAG) many of the transgenic 35S::TCP1 lines were clearly smaller in size than the wild-type seedlings (Figure S2). The 35S::TCP1 lines did not show root phenotypes. In contrast, constitutive expression of the gerbera CYC2 clade genes did not affect seedling size but did result in shorter root length for all genes except GhCYC2 (Figure S2).

The T1 lines could be grouped based on the severity of the phenotype into strong and intermediate lines, as well as lines with no clear phenotype (similar phenotypic categories were described for 35S::TCP1 by Busch and Zachgo, 2007). Constitutive expression of all the gerbera CYC2 clade genes, except GhCYC2, led to phenotypes similar to the intermediate lines of 35S::TCP1 (Figure 3). The overall plant size was smaller than in the wild type and they produced flowers that had a smaller petal size. Furthermore, they produced short siliques with fewer seeds than the wild type. Constitutive expression of GhCYC4 and GhCYC7 also led to phenotypes that were very similar to those of the strongest 35S::TCP1 lines (Figure S3). These plants were dwarfed and their first flowers failed to open, lacked mature floral organs and produced only a few if any seeds. In the 50 35S::GhCYC2 T1 lines studied, and the four T2 lines selected for phenotypic verification, we did not see additional phenotypic changes other than that some lines produced short siliques and thereby fewer seeds. Thus, reduced reproductive success was common to all the transgenic Arabidopsis lines showing constitutive expression of the gerbera CYC2 clade genes. We can conclude that, when tested in a heterologous system, the gerbera CYC2 clade genes (excluding GhCYC2) are capable of causing similar phenotypes as the orthologous Arabidopsis TCP1 gene, and thus their biochemical function is conserved in this context.

Figure 3.

Ectopic expression of the gerbera CYC2 clade genes in Arabidopsis affects vegetative growth and petal size. In comparison to the wild type Columbia (a), ectopic expression of GhCYC3 (d), GhCYC4 (e), GhCYC5 (f), GhCYC7 (g) and GhCYC9 (h) but not GhCYC2 (c), led to reduced vegetative growth and smaller petal size, similarly to the control 35S::TCP1 lines (b). Scale bars: 1 cm.

Ectopic activation of GhCYC3 and GhCYC4 functions indicate redundancy in regulation of flower type identity in gerbera

We observed that the constitutive expression of CYC genes in gerbera caused severe growth defects (e.g. Broholm et al., 2008) and difficulties in regeneration of transgenic shoots. Therefore we performed functional analyses in gerbera by using an inducible system to transiently activate GhCYC function. The dexamethasone (DEX)-inducible rat glucocorticoid receptor (GR) fusion system is suitable for nuclear transcription factors. All gerbera CYC2 clade proteins have a nuclear localization signal in their amino acid sequences and are predicted to be targeted to the nucleus, except for GhCYC5 which TargetP analysis suggested chloroplast localization. We tested the subcellular localization of the gerbera proteins using transient expression of GhCYC:GFP fusions in onion epidermal cells and confirmed that all of them (including GhCYC5) are localized in the nucleus (Experimental procedures S1, Figure S4).

For functional analyses in gerbera, we focused on GhCYC3, GhCYC4 and GhCYC5 which showed the highest expression levels in young inflorescences as well as specific expression patterns during ligule development in ray flowers. We produced 6 to 16 transgenic lines constitutively expressing GhCYC3:GR, GhCYC4:GR or GhCYC5:GR fusion constructs (Figure S5). Induction of ectopic activity of GhCYC3 and GhCYC4 in gerbera inflorescences altered the morphology of disc flowers to resemble ray flowers. The ligule length of disc flowers increased significantly (Tables 1 and 2), leading to pronounced bilateral symmetry, and stamen development was arrested (Figure 4). In 35S::GhCYC3:GR lines, the ligule lengths of ray and trans flowers were also significantly altered, being shorter in rays and longer in trans flowers when compared with the DEX-treated wild-type Terra Regina (Table 1). In 35S::GhCYC4:GR lines, the ligule length of ray flowers was unaltered but those of trans flowers were longer (Table 2). Furthermore, ectopic activation of GhCYC3 and GhCYC4 significantly promoted the growth of the two dorsal petals (marked with arrows in Figure 4). In conclusion, the ectopic activation of GhCYC3 and GhCYC4 function led to similar but more extreme phenotypes, as previously reported for lines with constitutive GhCYC2 expression (Broholm et al., 2008). To better compare the phenotypes, we generated inducible 35S::GhCYC2:GR lines. The induced ectopic GhCYC2 function affected disc flower petals and stamens in a similar manner to that described for the overexpression lines in Broholm et al. (2008) (Figure S6). Thus, the more complete conversion, compared with GhCYC2, of disc flower morphology upon activation of GhCYC3 and GhCYC4 was not due to the use of different methods for ectopic activation. Despite numerous attempts, we were not able to produce RNA interference (RNAi) silencing lines for the CYC2 clade genes.

Table 1. Length of the ventral ligules (mm) of different flower types in wild-type and transgenic gerbera lines with activated GhCYC3 function
LineRay flowerTrans flowerDisc flower
  1. The lengths of the ventral ligules were measured from two independent transgenic lines in comparison with the wild type Regina after dexamethasone treatment. Differences were tested by pairwise comparison with the Dunnett test using five biological replicates.

  2. Statistical significance is indicated with * (< 0.001).

Wild type47.3 ± 1.717.8 ± 2.011.3 ± 0.6
35S::GhCYC3:GR TR1137.8 ± 2.8*26.4 ± 1.7*16.5 ± 1.2*
35S::GhCYC3:GR TR1238.8 ± 3.0*27.1 ± 1.6*15.7 ± 2.4*
Table 2. Length of the ventral ligules (mm) of different flower types in wild-type and transgenic gerbera lines with activated GhCYC4 function
LineRay flowerTrans flowerDisc flower
  1. The lengths of the ventral ligules were measured from two independent transgenic lines in comparison with the wild type Regina after dexamethasone treatment. Differences were tested by pairwise comparison with one-way anova using five biological replicates.

  2. Statistically significant differences are indicated with * (for TR7 < 0.002 and TR5 < 0.001).

Wild type 145.4 ± 2.517.9 ± 0.911.7 ± 0.8
35S::GhCYC4:GR TR742.2 ± 2.329.3 ± 1,7*19.4 ± 3.6*
Wild type 247.7 ± 1.516.4 ± 1.811.3 ± 0.3
35S::GhCYC4:GR TR544.2 ± 4.335.0 ± 3.0*20.4 ± 1.6*
Figure 4.

The effect of activated GhCYC3 and GhCYC4 function on inflorescences, flower types and stamen development in transgenic gerbera. (a) Wild-type Terra Regina inflorescence. (b) Inflorescence of a transgenic 35S::GhCYC3:GR gerbera line. (c) Inflorescence of a transgenic 35S::GhCYC4:GR gerbera line. (d) Ray, trans and disc flowers of the wild type. (e) Ray, trans and disc flowers of the transgenic 35S::GhCYC3:GR gerbera line. (f) Ray, trans and disc flowers of the transgenic 35S::GhCYC4:GR gerbera line. (g) Stamen of the wild type. (h) Disrupted stamen of the transgenic 35S::GhCYC3:GR gerbera line. (i) Disrupted stamen of the transgenic 35S::GhCYC4:GR gerbera lines. Scale bars: 1 cm (a–f), 1 mm (g–i). The arrows in (e) and (f) point to the enlarged dorsal petals that are rudimentary in the wild type (d).

To study whether the larger ligule size of the 35S::GhCYC3:GR and 35S::GhCYC4:GR lines was achieved by affecting cell proliferation or cell size, we measured the size of epidermal cells in SEM images of the ventral ligules of disc flowers from the DEX-treated lines. The length of ligule cells was not significantly altered in the 35S::GhCYC3:GR or 35S::GhCYC4:GR plants in comparison with the wild type (Figure 5). Thus, the larger ligule size is likely to result from an increased cell number.

Figure 5.

Epidermal cell lengths of disc flower ligules in wild-type and transgenic gerbera lines. (a) 35S::GhCYC3:GR line and (b) 35S::GhCYC4:GR line in comparison with the wild type Terra Regina after dexamethasone DEX-treatment. The error bars represent the standard error of mean. The Student's t-test showed that differences in cell numbers are not statistically significant (P-values for 35S::GhCYC3:GR- and 35S::GhCYC4:GR-lines were 0.119 and 0.120, respectively).

The crested gerbera cultivar shows upregulation of GhCYC3 expression

Due to its resemblance to the transgenic phenotypes, a crested gerbera cultivar obtained from a breeder's collection was further investigated for CYC2 clade gene expression (Figure 6). This cultivar only produces flowers with a ray identity. All CYC2 clade genes were expressed in the outermost ray flower primordia of the crested cultivar similar to the wild-type variety Regina. However, unlike in the wild type, only GhCYC3 was highly upregulated in the centremost flower primordia of the crested cultivar.

Figure 6.

Gene expression of CYC2 clade genes in a crested gerbera cultivar. (a, b) Expression of GhCYC genes in marginal and central flower primordia of the crested cultivar CH02.663 (b) in comparison to the radiate cultivar Terra Regina (a). (c) Quantitative PCR indicates that GhCYC3 is upregulated in centremost flower primordia in the crested (Cre) cultivar. Two different developmental stages (4, 6) were included in the analysis. RF, ray flower primordia; CF, central flower primordia. Mean fold changes were calculated relative to the expression in Terra Regina separately at each developmental stage, for each gene and flower primordia type. Error bars represent standard deviations of three technical replicates. Fold changes were not analysed for GhCYC2, GhCYC4 and GhCYC5 in the central flowers since the expression levels were very low (Cp values >29).

GhCYC5 shows divergent function in regulating the flower density of the inflorescence

The induced activation of GhCYC5 function did not cause any striking visual changes at the inflorescence level (Figure 7a,b). However, closer analysis of the transgenic lines indicated changes in the number of flowers in the capitula (Table S1). As the capitulum area may vary according to the season or growth conditions and in individual inflorescences of the same plant, we calculated flower density, i.e. the number of flowers per unit area of the capitulum (mm2). By analysing five independent transgenic lines in comparison with the DEX-treated wild type Terra Regina we observed a statistically significant increase in flower density (Figure 7, Table S1). We did not detect any morphological alterations, including organ identity changes, in the phenotypes of individual flower types. We also followed ligule development and found that, in contrast to activation of GhCYC2, GhCYC3 or GhCYC4 functions, the lengths of the ligules were not changed in any of the flower types (Table S2).

Figure 7.

The effects of activated GhCYC5 function. (a) Inflorescence of the wild type Terra Regina. (b) Inflorescence of a transgenic 35S::GhCYC5:GR gerbera line. (c) The density of flowers in an inflorescence indicated as a number of flowers per square millimetre (n per mm2). The density of flowers was measured from five independent transgenic lines. Differences between the dexamethasone-treated wild type Terra Regina and transgenic lines were tested by pairwise comparison with a Dunnett test using 10 biological replicates. Statistically significant differences are marked with * (< 0.001). Scale bars: 1 cm.


The expanded CYC2 clade gene family in Asteraceae has been shown to be associated with the development of the complex inflorescence architecture (Broholm et al., 2008; Kim et al., 2008; Chapman et al., 2012; Tähtiharju et al., 2012). In this paper, we extended characterization of the individual CYC2 clade gene family members in gerbera. Our data indicate a ventralized pattern for CYC2 clade gene expression in the petal whorl that may have evolved after the origin of the Asteraceae. The largely overlapping expression domains at the level of individual flowers suggest that functional specificity for the CYC2 clade genes may be obtained through context-specific protein complexes that activate different downstream targets. Functional analysis in transgenic gerbera further indicated that in addition to GhCYC2 (Broholm et al., 2008), GhCYC3 and GhCYC4 show redundant functions in the regulation of ray flower identity and promoting petal development in ray flowers. GhCYC5 instead shows a diverged function in regulating the rate of flower initiation in the inflorescence.

Gerbera CYC2 clade expression is shifted from the ancient dorsal domain

Typically, the CYC2 clade genes involved in regulating floral zygomorphy in core eudicots show asymmetric expression patterns localizing to the dorsal domain of the flower, as shown, for example, in Antirrhinum majus (Luo et al., 1996, 1999) and Iberis amara (Busch and Zachgo, 2007). Broader functional domains of duplicated CYC gene paralogues in dorsal and lateral petals have been observed in Fabales, in Lotus japonicus and Pisum sativum (Feng et al., 2006; Wang et al., 2008) and in Primulina heterotricha, Gesneriaceae (Gao et al., 2008; Yang et al., 2012). The single CYC2 clade gene of Arabidopsis, TCP1, also shows expression in the dorsal side of the meristem of actinomorphic flowers, but only transiently in the earliest stages of flower meristem development (Cubas et al., 2001).

Dorsal-specific gene expression is likely to have evolved after the duplication event within the CYC/TB1 clade that gave rise to the CYC2 subclade (Preston and Hileman, 2009), but at a time pre-dating the divergence of Antirrhinum (asterids) and Arabidopsis (rosids) (Cubas et al., 2001). Our data, however, indicate that none of the gerbera CYC2 clade genes behave in this dorsal-specific manner. In contrast, GhCYC2 expression is excluded from the dorsal petals and has shifted to the ventral domain (Broholm et al., 2008), while the expression of GhCYC3, GhCYC4, GhCYC5 and GhCYC7 is detected both in the two rudimentary dorsal petals and the fused ventral ligule, i.e. in all five members of the petal whorl. So far, a ventralized pattern of CYC gene expression in the petal whorl has not been reported in any other plant lineages, indicating that it may have evolved specifically after the origin of Asteraceae. However, CYC genes also control stamen number, and Song et al. (2009) showed evidence that in Opithandra (Gesneriaceae) expression of a CYC gene is associated with the abortion of the ventral stamen. However, in Veronica and Gratiola (Plantaginaceae) CYC gene expression and stamen reduction in the ventral domain of the flower are not correlated, suggesting a different genetic mechanism (Preston et al., 2009).

Ectopic expression of GhCYC genes in Arabidopsis causes reduced growth during both vegetative and reproductive development

Ectopic expression of the gerbera CYC2 clade genes in A. thaliana indicated that GhCYC3, GhCYC4, GhCYC5, GhCYC7 and GhCYC9 all affect growth in a similar way as the endogenous Arabidopsis homologue TCP1. Intriguingly, although the gerbera CYC2 clade genes show subfunctionalization in gerbera and generally promote growth in reproductive tissues (see below), their effect on growth in the heterologous Arabidopsis background is repressive. We observed defects in vegetative growth and reproductive success. In addition, we detected reduction in petal growth, similarly to the 35S::TCP1 lines, as previously reported by Busch and Zachgo (2007). Interestingly, the ectopic activation of the Antirrhinum majus CYC in Arabidopsis led to an opposite effect on petal growth, i.e. enlarged petals due to enhanced cell expansion (Costa et al., 2005). This suggests that the growth effect caused by the rosid genes (A. thaliana and Iberis amara; Busch and Zachgo, 2007) and asterid CYC2 clade genes (from gerbera) represent the more ancestral functions of the CYC2 clade genes, while the growth effect caused by the asterid Antirrhinum CYC represents a more recent innovation. However, the ectopic expression of all of these genes causes reduced vegetative growth, indicating that, at least in vegetative organs, their function is conserved. Notably, also in gerbera, the ectopic expression of GhCYC2 led to strongly reduced vegetative growth (Broholm et al., 2008), an effect that we deliberately avoided in this study by only performing the DEX treatments on early inflorescence primordia, not on vegetative tissues.

Although the 35S::GhCYC2 lines did not produce phenotypes similar to the other transgenic Arabidopsis lines, its ectopic expression in gerbera did produce an inflorescence phenotype similar to 35S::GhCYC3:GR and 35S::GhCYC3:GR lines (Broholm et al., 2008). It is possible that GhCYC2 function is truly different in a heterologous Arabidopsis background but we cannot exclude the possibility that we would detect a similar phenotype with other CYC2 clade genes by generating more lines or that the phenotype would be visible in some other growth conditions. Markedly, the 35S::GhCYC5 phenotypes in Arabidopsis are similar to the phenotypes caused by the other CYC2 clade genes, but in gerbera show clear functional differentiation. One possible explanation is that GhCYC5 has diverged protein interaction capacity (see below) and/or target genes that are not present in Arabidopsis.

Gerbera CYC2 clade genes show redundant functions in regulating ray flower identity

The gerbera CYC2 clade genes showed partially overlapping expression domains during the early stages of development of flower primordia. Ectopic activation of GhCYC2, GhCYC3 and GhCYC4 caused very similar phenotypes in transgenic gerbera (Broholm et al., 2008; this study). All gene activities converted disc flowers into ray-like ones by promoting ligule growth through enhanced cell proliferation and suppressed stamen development. However, the phenotypic changes caused by GhCYC4 activation were more pronounced. In trans flowers, both GhCYC3 and GhCYC4 promoted ligule growth whereas GhCYC2 did not (Broholm et al., 2008). In ray flowers instead, both GhCYC2 and GhCYC3 activation led to reduced petal length. Ectopic activation of GhCYC3 and GhCYC4 also promoted the growth of the dorsal petals. Interestingly, GhCYC3 and GhCYC4 are expressed in the dorsal petals of wild-type ray flowers which, however, remain rudimentary.

Together, these observations indicate that the growth effects of these genes may vary based on the site and timing of their expression. Opposite growth effects have also previously been reported for CYC/TB1-like transcription factors. For example, the Antirrhinum CYC suppresses the growth of the dorsal-most stamen, and while initially suppressing the growth of the dorsal petals it later on promotes their growth (Luo et al., 1996). Tähtiharju et al. (2012) showed that the gerbera CYC2 clade proteins have the ability to interact in yeast two-hybrid assays. Therefore, we postulate that the functional specificity in given tissues (or in various flower types) is connected with formation of context-specific protein complexes involving CYC2 proteins and their co-regulators that may target different downstream genes. CYC2 clade proteins have also been shown to autoregulate themselves and to cross-regulate each other, leading to the formation of autoregulatory loops that may trigger threshold-dependent genetic switches (Yang et al., 2012). The presence of autoregulatory connections between the gerbera CYC2 clade genes may also affect the differential growth effects and will form an interesting target for future studies.

CYC2 clade genes show functional diversification in regulation of stamen development and late petal development in ray flowers

In both gerbera and sunflower, CYC2 clade genes are predominantly expressed in ray flowers, in which the stamen development is disrupted and their expression is absent from the actinomorphic disc flowers with functional stamens (Tähtiharju et al., 2012). In situ hybridization analysis indicated that GhCYC3 and GhCYC4 expression is absent from the rudimentary stamens of wild-type ray flowers while GhCYC5, GhCYC7 and GhCYC2 (Broholm et al., 2008) are present, suggesting functional diversification among the gene family members. Still, ectopic activation of GhCYC2, GhCYC3 and GhCYC4 led to suppression of stamen development in the modified disc flowers, suggesting that these proteins may share common downstream targets. However, we cannot exclude the possibility that the ectopic expression of GhCYC3 and GhCYC4 might activate GhCYC2 through an autoregulatory mechanism.

Temporal expression analysis during ligule development in ray flowers indicated that CYC2 clade genes regulate the late stages of ligule growth. GhCYC3 showed highest expression levels in petals at Stages 1–4 but was maintained throughout ligule development until Stages 8–9. The expression of other CYC2 clade genes was lower and relatively uniform throughout development. The GhCYC3 expression level correlated with our previous biometric studies showing that petal growth is tightly controlled and invariable; the ligule expands at a constant rate both longitudinally and laterally until Stage 9 when its growth ceases and the petals have reached their final size and shape (Kotilainen et al., 1999). Based on the expression dynamics of genes annotated in the functional class ‘cell growth and structure’, microarray analysis of gerbera petal organogenesis indicated that Stage 4 can be considered as a transition stage between cell division and elongation, while from Stage 6 organ growth is caused by cell expansion (Laitinen et al., 2007). The highest expression of GhCYC3 thus correlates with the cell division phase. This also fits well with our conclusion that larger petal size after ectopic activation of GhCYC3 and GhCYC4 is probably due to increased cell proliferation and not to enlarged cell size. A recent study exploring the natural variation of petal size and shape in three A. thaliana ecotypes identified ERECTA as a major locus determining petal shape. The allelic variation in this locus was associated especially with petal cell proliferation and not cell size (Abraham et al., 2013). Further studies are required to assess the possible regulatory links between TCP factors and, for example, ERECTA-like receptor kinases, as well as to identify the direct target genes required to define the final size and shape of petals. We also showed that GhCYC3 expression was upregulated in the centremost flowers of gerbera crested cultivars, further supporting the major role of this gene in promoting petal growth and flower-type identity.

GhCYC5 affects the rate of flower initiation

Ectopic activation of GhCYC5 uncovered a diverged function for a CYC2 clade gene. We observed a statistically significant increase in the number of flowers per unit area of the capitulum. In Asteraceae, the number of seed-producing flowers is an important component of final yield (López Pereira et al., 1999; Cantagallo and Hall, 2002). Studies in sunflower have indicated that flower initiation and meristem tissue expansion are strongly coordinated and responsive to various environmental signals (Dosio et al., 2006, 2011). Our previous studies in gerbera give similar indications. Transgenic lines with suppressed GRCD2 (a SEPALLATA-like gene) expression retain an undifferentiated, expanding region in the centre of the inflorescence that permits continuous initiation of new disc flower primordia (Uimari et al., 2004). In sunflower, the final number of flowers in a capitulum is determined by the initial meristem area before flower initiation and the rate of tissue expansion in the meristem during flower initiation, while the duration of the expansion phase remain stable under different environmental conditions (Dosio et al., 2006). Our data suggest that in gerbera, GhCYC5 affects the rate of flower initiation in a given time frame leading to increased flower density. Interestingly, the organization of the capitulum was not affected, and its appearance did not differ from the wild type.

The phenotypic changes caused by ectopic expression of GhCYC5 in Arabidopsis were similar to the other gerbera CYC2 clade genes, indicating differentiation of the regulatory network in Asteraceae. The yeast two-hybrid assays by Tähtiharju et al. (2012) indicated that GhCYC5 is exceptional compared with the other CYC2 clade proteins as it did not show the ability to interact with the other CYC-like proteins nor to homodimerize. The early stages of meristem patterning and flower initiation in Asteraceae are poorly understood and, in fact, functional analyses of putative meristem identity determinants are entirely lacking. Thus, more detailed analyses using lines in which GhCYC5 is silenced as well as analyses of GhCYC5 target genes and its interactome are needed to discover the mechanisms of GhCYC5 function and possible connections with other key regulators involved in flower initiation.

Experimental Procedures

In situ expression analysis

For in situ analysis, the treatment of the plant material and sectioning was done as in Elomaa et al. (2003). Gene-specific probes for the CYC2 clade genes GhCYC3, GhCYC4, GhCYC5, GhCYC7 and GhCYC9 corresponding to their 3′-untranslated region sequences were amplified with the primers shown in Table S3. All probes were amplified with primers containing a few extra nucleotides and a T7 overhang (CAtaatacgactcactataggg) in the 5′ end. When included in the forward primer, the T7 recognition site gave the sense probe, whereas the in the reverse primer an antisense probe was obtained. The gel-purified PCR products (200–300 bp) were used as templates for in vitro transcription with T7 polymerase and labelled using the DIG RNA Labelling Kit (Roche, according to the manufacturer's instructions. Hybridizations were performed essentially as described in Ruonala et al. (2008). For the post-hybridization washes, an InsituPro Vsi 3.0 (Intavis AG, device was used as in Hofer et al. (2012). The colour reaction to visualize the hybridized probe was done for approximately 24 h at room temperature (20–22°C). Sections were photographed using an AxioImagerZ1 (Zeiss, equipped with AxioCam MRc (Zeiss).

Quantitative real-time PCR analyses

Petal samples from ray flowers corresponding to inflorescence developmental Stages 1–11 (Helariutta et al., 1993; Laitinen et al., 2007) were collected from the variety Terra Regina (Terra Nigra BV, During Stages 1–3 the petals are covered by involucral bracts. At Stage 5, anthocyanin pigmentation becomes visible. The opening of the inflorescence occurs between Stages 7 and 8 and at Stage 9, the petals have reached their final length and width. After Stage 11 the petals start to wilt (Kotilainen et al., 1999). For the crested cultivar CH02.663 (Terra Nigra BV) and the radiate cultivar Terra Regina, marginal and central flower primordia samples corresponded to primordia Stages 4 and 6 (Laitinen et al., 2006), and samples were pooled from four to six inflorescences. Total RNA was isolated using the TRIzol reagent (Life Technologies/Gibco-BRL, following the manufacturer's instructions. The RNA was treated with RNase-free DNase and cleaned up with the RNeasy Mini Kit according to the manufacturer's instructions (Qiagen, The RNA concentrations were equalized within a sample set and the RNA integrity was analysed by gel-electrophoresis. The cDNA synthesis, real-time RT-PCR and evaluation of product specificity were carried out as previously described (Tähtiharju et al., 2012). The qPCR primers were the same as used previously (Tähtiharju et al., 2012, Table S3). The PCR efficiencies were analysed for all primer combinations to make sure that they were close to 100% (Table S3). Relative expression levels were calculated using the inline image method (Pfaffl, 2001). The expression levels were normalized to GhACTIN expression levels. Two biological replicate samples each with three technical replicates were used for the real-time RT-PCR.

Inducible transgene constructs

The GhCYC open reading frames (ORFs) were fused to the rat glucocorticoid receptor (GR) encoding sequence (Lloyd et al., 1994) and expressed under the CaMV35S promoter. The forward primers used to amplify GhCYC ORFs introduced a restriction site immediately before the respective start codons and the reverse primers removed the GhCYC stop codons and created a restriction site at the C-terminus. The PCR fragments were digested with and cloned into the BamH1 restriction site of the plasmid GR-pBluescript (pRS020, obtained from the Arabidopsis Biological Resource Center, (Lloyd et al., 1994; Sablowski and Meyerowitz, 1998) to create in-frame translational fusions at the C-terminus of GhCYCs with the rat glucocorticoid hormone-binding domain. The fusion fragments (GhCYC:GR) were further cloned into the plasmid vector pHTT603 Sma1 site. The vector pHTT603 is similar to pHTT602 (Elomaa and Teeri, 2001) but contains the multiple cloning site in reverse orientation. The 35S::GhCYC:GR constructs in pHTT603 were conjugated into Agrobacterium tumefaciens strain C58C1 (van Larebeke et al., 1974) with the disarmed Ti plasmid pGV2260 (Deblaere et al., 1985) using triparental mating (van Haute et al., 1983).

Plant transformation

The ORFs for gerbera CYC2-clade genes and the Arabidopsis orthologue TCP1 were cloned into the Gateway® entry vector pDONR221 (Tähtiharju et al., 2012) and then under the CaMV35S promoter in pK7WG2D (Karimi et al., 2002), following the manufacturer's instructions (Invitrogen, The Arabidopsis ecotype Columbia was transformed by floral dipping method (Clough and Bent, 1998). Kanamycin-resistant transformants (T1 generation) were first selected on MS plates (50 μg ml−1 kanamycin) and then transplanted on soil and grown in long-day conditions. We screened for 27–63 T1 plants per construct (52 T1 plants for 35S::TCP1, 50 for 35S::GhCYC2, 60 for 35S::GhCYC3, 27 for 35S::GhCYC4, 62 for 35S::GhCYC5, 38 for 35S::GhCYC7 and 43 for 35S::GhCYC9). Phenotypes were first screened in the T1 generation, and four to six representative lines were chosen for characterization in T2 generation to confirm the heritability of the phenotypes. All the figures shown are taken from representative T2 plants. Transgene expression was verified with RT-PCR from young leaf tissue of representative T2 lines.

Gerbera transformation was conducted as described in Elomaa and Teeri (2001). Transgene expression was verified with RT-PCR from capitula of diameter 10–16 mm. For RT-PCR total RNA was isolated using the TRIzol reagent (Life Technologies/Gibco-BRL) following the manufacturer's instructions. Five micrograms of total RNA was used for the first-strand cDNA synthesis according to Invitrogen's Superscript III RT protocol. Two microlitres of cDNA was used as template for RT-PCR that was conducted with a gene-specific forward primer for GhCYC3, GhCYC4 or GhCYC5 and a reverse primer GER297 (CGAAGTGTCTTGTGAGACTCC) that anneals to the GR sequence. The PCR conditions were as follows: 94°C for 30 sec, 60°C for 60 sec and 72°C for 45 sec for 35 cycles.

Dexamethasone inductions and analysis of the induced plants

To activate the CYC protein function in transgenic gerbera lines, we used 10 μm DEX solution with 0.015% Silwet. The induction was done by pipetting the DEX solution on top of very young inflorescences located in the middle of the rosette between the emerging young leaves (3 ml per plant). The DEX treatment was done five times, once a day, for all the other lines except for 35S::GhCYC4:GR, in which the 5-day treatment stopped all growth and led to senescence of the plants. Thus, we used a 3-day DEX treatment for 35S::GhCYC4:GR lines.

From the DEX-induced plants (GhCYC2, -3, -4 and -5) and DEX-treated wild-type control, the lengths of the ligules (excluding the basal tube) were measured using five independent inflorescences per line. From a fully open inflorescence (Stage 9; Helariutta et al., 1993) all ray flowers were measured as well as more than 50 trans and disc flowers from the same inflorescence section. Stamen phenotypes were documented using light microscopy (Zeiss SteREO Discovery V20) equipped with AxioCam ICc3 (Zeiss). Statistical pairwise testing was done with spss (IBM SPSS statistics, using Dunnett's test or one-way anova.

For the DEX-treated 35S::GhCYC5:GR we also calculated the number of flowers in the inflorescences, capitulum area (mm2) and density of flowers (number per mm2) from 10 biological replicates per line (Stage 9). The diameter of the capitulum was measured as an average of five measurements when all flowers and bracts were removed. The surface area was calculated with a formula of a circle or, when perpendicular diameter measurements differed more than 1.5 mm, with a formula of an ellipse. Statistical pairwise testing was done using Dunnett's test.

Scanning electron microscopy and cell measurements

Scanning electron microscopy (SEM) was used to measure cell sizes in disc flower ligules of the DEX-treated 35S::GhCYC3:GR and 35S::GhCYC4:GR lines in comparison to DEX-treated wild type. We analysed cell sizes from seven individual disc flower ligules from two inflorescences of a 35S::GhCYC3:GR line (TR12), two inflorescences of a 35S::GhCYC4:GR line (TR7) and three wild-type inflorescences. SEM samples were prepared as described in Uimari et al. (2004). Samples were examined using the FEI Quanta 250 field emission gun scanning electron microscope ( at the Electron Microscopy Unit of the Institute of Biotechnology at the University of Helsinki. Images were taken from adaxial side of disc flower ligules, at the same distance from the tip of the ligule in relation to the ligule length (for wild-type ligules, at a distance of 1.6 mm from the tip, and for the 35S::GhCYC3:GR and 35S::GhCYC4:GR lines, 2.8 mm from the tip of the ligule). Two images were taken from each sample, and the lengths of 20 cells per image were measured using imagej 1.47 ( index.html).


We thank Eija Takala and Anu Rokkanen for excellent technical assistance and Sanna Peltola and Johanna Boberg for taking care of the plants in the greenhouse. We also thank the Electron Microscopy Unit of the Institute of Biotechnology, University of Helsinki, for providing laboratory facilities. This work was supported by the Academy of Finland (139092 to PE), the Netherlands Organization for Scientific Research (825.08.037 to ASR) and the Finnish Doctoral Programme in Plant Science (IJ-P).