Two cultivars of C. morifolium and their F1 progenies (Figure 2A), as well as the wild diploid species C. lavandulifolium were grown in the standard experimental fields at Xiao Tang Shan, affiliated to Beijing Forestry University, Beijing, China. C. morifolium ‘Guoqing xiaoliuhao’ (GQ) and ‘Mao xiangyu’ (MXY) (Figure 2B) are two hexaploid (2n = 54) ground-cover chrysanthemum cultivars with different whorls of ray florets. C. lavandulifolium (2n = 18) used for plant transformation is one of the original species of C. morifolium with relatively simple genetic background. Wild-type (WT) and transgenic plantlets of C. lavandulifolium with six leaves were transplanted into flowerpots containing a mixture of turf and vermiculite substrate (V:V = 1:1). All these plantlets were managed routinely in the standard greenhouse under a 16-h light/8-h dark cycle for 150 days, and then an 8-h light/16-h dark cycle at 25 ± 2°C with ∼55% relative humidity.

Total RNA was extracted from young inflorescences of C. morifolium ‘Guoqing xiaoliuhao,’ ‘Mao xiangyu,’ and C. lavandulifolium using the TRIzol reagent (Invitrogen, USA) according to the manufacturer’s instructions. The RNA was treated with RQ1 RNase-free DNase (Promega, USA) according to the manufacturer’s protocol to remove residual genomic DNA. The RNA integrity was analyzed by gel electrophoresis, and its concentrations were measured and equalized within sample sets. The first strand of cDNA was synthesized based on 2 μg of total RNA using the M-MLV reverse transcription system (Promega, Madison, WI, USA). The TCP and R domain factors of CYC2 subclade genes were amplified using one pair of degenerate primers designed corresponded to the amino acid sequences ASKTLDWL and RARARERT of available GenBank Asteraceae sequences (accession number: FJ356704.1, FJ356700.1, EU429303.1, EU429304.1, and EU088368-EU088372). PCR conditions were the following: 94°C for 30 s, 54°C for 30 s, and 72°C for 1 min for 30 cycles. Multiple clones were sequenced and analyzed in an attempt to find all the CYC-like paralogs. The full length cDNAs were amplified by using the SMART RACE cDNA amplification kit (Clontech, Japan) to carry out 3′ and 5′ RACE (rapid amplification of cDNA ends). All of the amplified products were sub-cloned into the pMD18-T vector (Takara, Japan) and transformed into E. coli DH5α for sequencing. All primers were listed in Supplementary Table S1. The mRNA coding sequences of these genes have been uploaded to the GenBank database (accession numbers KU595426-KU595431 and KX161379-KX161384).

Multiple alignment of CYC/TB1-like protein sequences from C. morifolium and the other selected eudicots species were performed using ClustalX 2.0 with default parameters (Larkin et al., 2007), and the BioEdit software (version 7.1) was used to edit the aligned sequences. Protein sequences between the TCP and R domains were selected to generate maximum likelihood (ML) tree using PhyML (Guindon et al., 2010) under the Jones-Taylor-Thornton (JTT) model with 100 bootstrap replicates. Numbers above branches indicate local bootstrap value (>50% support). Prediction of conserved motifs of CmCYC2 clade genes was performed using the MEME online tool ¹.

The young developing flower heads of MXY at different stages were collected and fixed overnight in FAA, and then dehydrated through a gradient ethanol series into 100% after being dissected under a stereomicroscope ethanol. After being chemically dried by washing in a 2-Methyl-2-propanol series and air-dried under vacuum, the samples were mounted on aluminum stubs and sputter coated with gold. Photographs were taken by a Hitachi S-3400N scanning electron microscope (SEM) in Beijing Forestry University.

The flower buds in 100% ethanol (procedures before this were the same as that mentioned above for SEM) were moved into a gradient xylene series. Then, the samples were embedded in paraffin and cut into 8 μm sections. For histological staining, paraffin was removed with xylene and the sections were stained with safranin (1% in water) and fast green (0.3% in 95% ethanol). The sections were observed and photographed under a light microscope (Zeiss Axio scope. A1) in Beijing Forestry University.

Quantitative RT-PCR (qRT-PCR) assay was used to detect the expression levels of the CmCYC2 genes in inflorescence primordia of MXY at six early developmental stages (stages I-III, Figure 3) and one late stage (stage IV) at which the petal ligules of ray florets began to expand (Supplementary Figure S1). To analyze the specific expression patterns of the CmCYC2 genes in ray and DF of C. morifolium, the double-ray flowered cultivar GQ and the radiated cultivar MXY, as well as two of their F1 progenies (F1a and F1b) with intermediate whorls of ray florets were prepared (Figure 2A). Each sample was a pooling of six to eight inflorescences at later developing stages (pooled from stage 3, 4, 6, 7, 8, and 9) (Supplementary Figure S1). We defined the later inflorescence developmental stages comparable to those defined in Gerbera (Bradley et al., 1993; Laitinen et al., 2007). In addition, the disc floret samples were only excised from their centermost florets while the ray floret samples were a mix of flowers from every whorl. For tissue-specific expression analysis, vegetative and reproductive organs (a mixture of stage 8 and 9) were dissected from another F1 progeny (F1c) which developed some trans-like florets between the normal ray and disc ones (Figure 2A). Additionally, the stamen sample was pooled from DF only, and the ovary plus stigma and style sample was a mixture of all the three types of florets. All plant materials collected were immediately placed in liquid nitrogen and stored at -80°C until extraction. Three biological replicates were collected for each sample. The extraction of total RNA and the synthesis of the cDNA were performed as described above. The cDNA sample was diluted five times with 80 μl of deionized water for all gene expression analyses. The qPCR reactions were performed using a Mini Opticon Real-time PCR System (Bio-Rad, USA) with SYBR Premix Ex TaqII kit (TaKaRa, Japan) following the manual’s recommendations. All gene-specific primer pairs for qRT-PCR were listed in Supplementary Table S1, and their amplification efficiencies were analyzed to make sure that they were closed to 100%. The PCR program was conducted with an initial step of 30 s at 95°C, followed by 40 cycles at 95°C for 5 s, 58°C for 30 s, and 72°C for 30 s. Product specificity for each primer pair was verified by melting-curve analysis. The relative expression levels were calculated using the 2^-ΔΔCt method, with CnActin (GenBank accession number KF305683.1) gene as the endogenous control.

The ORF of CmCYC2c was obtained by PCR amplification using gene-specific primers CmCYC2c-F2 and CmCYC2c-R2 (both containing NcoI enzyme site). Driven under the control of the CaMV 35S promoter, the CmCYC2c was cloned into a modified binary pCOMBIA1304 vector. Then the 35S::CmCYC2c plasmid was transformed into C. lavandulifolium by the Agrobacterium tumefaciens EHA105-mediated freeze-thaw method (Chen et al., 2004). For transformation, small leaf pieces with a diameter less than 8 mm were cut from newly formed expanding leaves of C. lavandulifolium using scalpels. After 48 h pre-culture on MS medium, these leaves were soaked in the Agrobacterium bacteria solution containing 20 μM acetosyringone (final OD₆₀₀ = 0.4–0.6) for 10 min, with excess broth sucked away by sterilized absorbent paper. Then the leaf pieces were cultured on the CI (callus-induction) medium (MS medium +1.0 mg l^-1 6-BA, 0.5 mg l^-1 NAA) for 3 days in darkness. After being washed with sterilized water, the explants were placed on the bacteria elimination medium (CI medium containing 400 mg l^-1 carbenicillin) for 5 days, and then were transferred to the selection medium I (CI medium containing 400 mg l^-1 carbenicillin and 8 mg l^-1 hygromycin) for the selection of putatively transformed calli. After two subcultures, the explants were placed on the selection medium II (containing only 200 mg l^-1 carbenicillin) to promote callus differentiation. The medium was replaced every 2 weeks until the regenerated plantlets grow up with 2–4 leaves. Then the putatively transformed plantlets were transferred to the rooting medium (MS medium + 10 mg l^-1 hygromycin, 200 mg l^-1 carbenicillin). All the materials were cultured at the same conditions with 16 h light/8 h dark at 24 ± 2°C. The transgenic plants selected from hygromycin were verified using a pair of CaMV 35S promoter-specific forward primer and CmCYC2c-specific reverse primer (Supplementary Table S1).

RNA extraction and cDNA synthesis were performed as described above from full opened inflorescence. We used two primer pairs to distinguish the expression levels of the transgene (CmCYC2c) and the endogenous gene (ClCYC2c) in full opened inflorescences of six positive transgenic lines. One specific primer pair CmCYC2c-F1/CmCYC2c-R1 was designed corresponded to the ORF sequence of the transferred gene CmCYC2c, the other pair ClCYC2c-utr-F1/ClCYC2c-utr-R1 was designed according to the 3′ UTR sequence of endogenous ClCYC2c. The ray florets of the three positive lines, showing obvious phenotypic changes, were used to detect the expression levels of all the six ClCYC2 clade genes. The ClActin gene (GenBank accession number JN638568.1) was used as an internal control. All primers were shown in Supplementary Table S1. Both WT plants and transgenic lines expressing the empty vector were used as the negative control plants.

The phenotypes of transgenic lines were observed and recorded during the development process. The numbers of ray and DF were calculated from the same inflorescence using fifteen independent flower heads per line. In addition, the petal ligule lengths of ray florets were measured from 15 inflorescences of each line. The top three fully open inflorescences were sampled from five lateral branches, the second one below each shoot apex, of both positive transgenic lines and control lines. Statistical testing was done with SPSS17.0 (SPSS Inc., Chicago, IL, USA) using one-way ANOVA and Duncan LSD multiple comparisons tests.

The capitulum of MXY contains 8–9 whorls of DF and two whorls of peripheral ray florets (Figure 2A). To better understand the morphologic traits of ray and DF of MXY, we normalized the development process into four phases: (I) initiation of floral primordia; (II) differentiation of floral organs; (III) growth of floral organs; and (IV) maturation of inflorescence (data not shown as the capitulum of this stage is big enough to be clearly observed with naked eye). The sizes of the capitula corresponding to each stage were shown in Supplementary Table S2. At stage I, the differentiation of inflorescence primordia was completed and individual flower primordia (arrowed) arose acropetally through the capitulum (Figures 3A,C, SI). The ray and disc floret primordia were indistinguishable undifferentiated bumps at this stage (Figures 3A,B, SI). At stage IIa, the marginal four whorls of DF began to form petal and stamen primordia (arrowed) while the inner four whorls were still undifferentiated bumps (Figure 3A, SIIa). From stage IIb to IIc, the differentiation of floral organ was continuing, those undifferentiated innermost bumps gradually turned into ring-shaped petal primordia (arrowed) with recognizable raised stamen primordia (Figure 3A, SIIb). Additionally, the petal lobes of peripheral DF continued to elongate and finally covered the developing stamen and carpel primordia (arrowed, Figure 3A, SIIc). The floral organ differentiation of ray florets seemed to lag behind the outermost DF at stage IIa, which became noticeable at stage IIb when the stamen primordia (st, arrowed) appeared in ray florets (Figure 3B, SIIa and SIIb). In ray florets, the zygomorphy of five petal ligules appeared (arrowed) soon after the petal primordia occurred at stage IIa. Further, the asymmetry became prominent with accelerated expanding of three fused ventral ligules (vl) and ceased growth of two dorsal petals (dp) when the stamen primordia kept on developing (Figure 3B, SIIa-SIIc). At stage IId, nearly all ray and DF finished their organ differentiation except for the innermost DF (arrowed, Figure 3A, SIId). The marginal ray florets eventually developed into female with arrested growth of stamens at stage IId, whereas the DF were fully bisexual (Figure 3C, SIId). Furthermore, the hair-like structures (hl) began to emerge on the surface of the peripheral flowers (Figure 3A, SIId). At stage III, the differentiation of floral organs was completed, and all flowers continued to grow and undertook further development to bloom (Figures 3A,B, stage III). In addition, all flowers were attached with those hair-like structures.

To study the role of CYC2 subclade transcription factors in establishment and evolution of the complex flower head of chrysanthemum, six genes encoding these proteins were isolated from young developing inflorescences by an extensive screening of putative homologs from PCR products. These genes were named according to a previously published nomenclature and are as follows: CmCYC2a, CmCYC2b, CmCYC2c, CmCYC2d, CmCYC2e, and CmCYC2f (GenBank accession nos. KU595426-KU595431). Excluding introns, the length of each gene varied from 774 to 969 bp (258 to 323 inferred amino acids). A conserved motif analysis showed that all these genes contain the conserved TCP and R domain typical for the CYC2/ECE subfamily (Figure 4B). The sequence variation among these genes are mainly located within the non-conserved regions (Figure 4C).

We performed phylogenetic analysis on a selected set of TCP class II proteins (between the TCP and R domains) from various plant species to explore the evolutionary relationships and diversification of the CmCYC2 subclade genes. The maximum likelihood analysis placed all the six CmCYC2 genes in CYC2 clade together with all the other listed Asteraceae sequences (Figure 4A). Therefore, our focus here was on the CYC2 subfamily. The most clearly supported orthologous were CmCYC2a and HaCYC2a along with GhCYC7, as well as the recently duplicated gene pair CmCYC2b and HaCYC2b grouping with CmCYC2c. Although the relationship was not well supported, CmCYC2f and CmCYC2e were grouped into one clade, and sister to a Senecio ray floret specific gene SsRAY2. The CmCYC2d gene was clustered with the other Senecio ray floret specific gene SsRAY1. In conclusion, CmCYC2f and CmCYC2e showed a sister relationship whereas the other four CmCYC2 genes were sister to the other Asteraceae sequences, suggesting both lineage-specific and shared duplications in Asteraceae.

According to the earlier results in Senecio, Gerbera and Helianthus (Broholm et al., 2008; Kim et al., 2008; Fambrini et al., 2011; Tahtiharju et al., 2012), CYC2 subclade genes are involved in regulating the complex inflorescence architecture of Asteraceae. To look into the possible role of the CYC-like genes in chrysanthemum, we investigated the transcription levels of the six CmCYC2 genes in different tissues of F1c (Figure 5A). All the six CmCYC2 genes were mainly expressed in floral reproductive organs, although they were weakly expressed in vegetative organs, in accordance with the results in Gerbera and sunflower (Broholm et al., 2008). In addition, all these genes were found to be predominantly expressed in petal of ray florets and barely expressed in that of DF. CmCYC2b and CmCYC2c in contrast were strongly expressed in petal of middle trans-like florets, which was not generally true for other CmCYC2 genes. Contrary to the previous studies in other species (Broholm et al., 2008; Tahtiharju et al., 2012), all the CmCYC2 genes were surprisingly relatively highly expressed in the involucrate bracts while slightly expressed in stamens and pistil (including the ovary, styles, and stigma). qRT-PCR reactions were also performed to compare the expression patterns of these genes in flower buds of MXY at both early and late development stages, as shown in Figure 5B. Interestingly, the expression of these genes all peaked at stage IIa and then decreased gradually and maintained at relative low levels, except for CmCYC2c, which was remarkably up-regulated at stage IV (mature flower buds).

To further investigate any flower-type specific role of the CYC-like genes, we compared the expression patterns of these genes in two chrysanthemum cultivars and two of their F1 progenies (Figures 2A and 5C). Similar to the previous studies (Broholm et al., 2008; Tahtiharju et al., 2012), all genes were found to be expressed at much higher levels in ray florets than in DF, as shown in Figure 5C, except that CmCYC2f was expressed at more or less comparable levels in all florets. CmCYC2d and CmCYC2e were expressed at generally similar levels in all of the tested ray florets, and did not display phenotype difference among four types of flower heads. In contrast, the expression of CmCYC2c was dramatically up-regulated in ray florets of the double-ray flowered heads (GQ and F1a), compared with the expression in the radiate (MXY) and semi-double ray flowered (F1b) heads. Although CmCYC2a and CmCYC2b exhibited an expression pattern similar to the CmCYC2c, they were expressed at much lower levels.

In short, CmCYC2c was abundantly expressed at two key stages of petal development in inflorescence primordia, and in petals of both ray and trans-like florets. Moreover, in a series of flower heads with various whorls of ray florets, the transcript levels of CmCYC2c were significantly up-regulated in ray florets of the double-ray flowered heads. We therefore conclude that CmCYC2c is a strong candidate as a key regulator of ray floret identity in chrysanthemum, and hence need a further verification through transgene.

According to our results, CmCYC2c was predominantly expressed in ray and trans-like florets (if presented) of the double-ray flowered inflorescence. Thus, the ORF of CmCYC2c was transformed under the CaMV 35s promoter into C. lavandulifolium. Finally, three positive lines showing obvious phenotypic changes were chosen for further analysis. Ectopic expression of 35S::CmCYC2c retarded the vegetative growth and outgrowth of lateral branches in young plantlets (Figure 6A). However, the growth of these young seedlings restored gradually later on and they produced inflorescence finally (Figure 6B). Further, the capitula appeared larger in the over-expression lines due to increased flower number and elongated petal ligule length of ray florets, compared with the WT plants and transgenic plants with empty vector (Figure 6C). The average length of petal ligule in ray florets was remarkably increased from 4.72 mm in empty vector to 6.12–7.04 mm in positive transgenic lines (Figure 6D; Table 1). In addition, the average number of ray florets per inflorescence was significantly enhanced from 13.8 in empty vector to 16–18 in positive transgenic lines (Table 1). Furthermore, we found some trans-like florets in one of the transgenic lines, which were morphologically similar to ray florets but with short petals and abnormal stamens (Figure 6E). In some inflorescences, the growth of ray florets was severally inhibited and resulted in malformation (data not shown). Interestingly, no phenotypic difference was observed in DF among different lines, in accordance with the results in Senecio (Kim et al., 2008; Garcês et al., 2016).

Moreover, the six independent positive transgenic lines that expressed CmCYC2c constitutively (Figure 7A) were further confirmed by qRT-PCR assay (Figure 7B). The expression of CmCYC2c was highly up-regulated in positive lines when compared with that of the endogenous ClCYC2c gene. Furthermore, given that the CYC2 clade proteins have been shown to cross-regulate each other to positive regulate gene expression (Tahtiharju et al., 2012; Yang et al., 2012), the effect of CmCYC2c on the other five ClCYC2 genes was also analyzed by qRT-PCR assay (Figure 7C). The results showed that, besides a significant up-regulation for ClCYC2c, the expression of ClCYC2f was also strongly induced in ray florets of the transgenic lines when comparing with that in both control ray florets. No obvious difference was found for the expression levels of the other four ClCYC2 genes among these lines.

CYC-like genes in Asteraceae have been shown to function as key regulators controlling the differentiation of the specific flower type (Marshall and Abbott, 1984; Broholm et al., 2008; Kim et al., 2008; Fambrini et al., 2014b). Much of the work has focused on the three model species: S. vulgaris (Kim et al., 2008; Garcês et al., 2016), H. annuus (Chapman et al., 2008, 2012) and G. hybrid (Broholm et al., 2008; Juntheikki-Palovaara et al., 2014). Surprisingly, few reports have been published on the function of these genes in the chrysanthemum, which is characterized by the extreme diversity of floral morphology. It has been previously reported that the CYC2 subclade has undergone the greatest number of gene duplications within the Asteraceae, such as five CYC-like genes in H. annuus and six in Gerbera (Broholm et al., 2008). No more copies have previously been reported in any other plant lineages studied thus far. Correspondingly, six CmCYC2 genes were obtained in chrysanthemum with highly conserved TCP and R domains. Gene and whole-genome duplications in plants are known to increase genomic complexity and have been suggested to produce new genes that allow the evolution of functional novelty (Jiao et al., 2011). The duplication of CYC2 subclade in Asteraceae seems to support the idea that members of this subclade may have novel and important functions contributing to the complex inflorescence structure in Asteraceae. Intriguingly, the strong gene candidates in specifying flower zygomorphy in Asteraceae were grouped into different clades according to the phylogenetic trees reported previously by Chapman et al. (2012) and Tahtiharju et al. (2012) and the current study. For example the HaCYC2c in sunflower (Chapman et al., 2012), the GhCYC2 in Gerbera (Tahtiharju et al., 2012), the SsRAY1 and SsRAY2 in Senecio (Kim et al., 2008), and the CmCYC2c in chrysanthemum were placed into five different groups, rather than showing sister relationships. These results indicate that paralogs of this gene family have been independently recruited to regulate zygomorphy in different species within the Asteraceae. Moreover, similar to the reports in sunflower and C. lavandulifolium (Tahtiharju et al., 2012; Ren and Guo, 2015), ray florets could be easily distinguished from disc ones in chrysanthemum at early stage of flower primordia development. The reason is that the asymmetric growth of the five petals in ray florets occurred soon after the petal primordia emerged. In Gerbera, however, the ray and disc petal primordia do not show difference in growth until late developing stages (Laitinen et al., 2006; Tahtiharju et al., 2012). Thus, the morphological divergence within Asteraceae in this respect is also consistent with the organismal phylogeny. As G. hybrida (Mutisieae) is a phylogenetically more distant specie when compared with the other three species (Bremer and Anderberg, 1994; Panero and Funk, 2008). This explanation can be further supported by the phylogenetic analysis that all the CmCYC2 genes were sister to the corresponding CYC2 members of sunflower and Senecio, whereas they were not clustered with most Gerbera CYC2 genes.

Typically, the expression of CYC2 subclade genes is restricted to the dorsal regions of the flower, for example, in A. majus (Luo et al., 1999) and Veronica montana (Preston et al., 2009). In Asteraceae, however, the expression of CYC-like genes is excluded from the dorsal petals and has shifted to the ventral part of ray florets (Chapman et al., 2008; Kim et al., 2008; Tahtiharju et al., 2012; Juntheikki-Palovaara et al., 2014). For example, RAY1, RAY2, and RAY3 in Senecio, proved to control ray versus disc floret identity, are expressed only in ray florets (Kim et al., 2008; Garcês et al., 2016). Here in chrysanthemum, the CmCYC2 genes were also predominantly expressed in petals of ray florets, consistent with what has been reported previously in Asteraceae (Chapman et al., 2008; Tahtiharju et al., 2012). The appearance of trans-like florets in the F1 progeny (F1c) suggests it is a transition type between ray and DF. In addition to petals of ray florets, CmCYC2c and CmCYC2b were also abundantly expressed in petals of trans-like florets, suggesting their possible role in promoting the growth of ventral petal ligule. This fits well with our data in transgenic lines that over-expression of CmCYC2c leads to elongated ventral ligule length in ray florets. Similar results have been reported for GhCYC4, which also shows ray/TF specific expression, and function in enhancing the growth of petal ligules (Juntheikki-Palovaara et al., 2014). Detailed expression analysis in Gerbera show that GhCYC2 is exclusively expressed in the ventral domain, and the expression of GhCYC3, GhCYC4, GhCYC5, and GhCYC7 is detected both in the fused ventral ligule and two rudimentary dorsal petals.

The transcript levels of the six CmCYC2 genes were analyzed in two chrysanthemum cultivars and their F1 progenies with intermediate whorls of ray florets. All these genes were predominantly expressed in ray florets, and most of them showed no or some expression variation across genotypes. However, the expression of CmCYC2c was significantly up-regulated in the double-ray flowered heads, indicating it is a strong candidate as a regulator of ray floret identity. It is supported by the fact that the 35s::CmCYC2c transgenic lines generate more ray florets per capitulum in comparison with that in the control lines. Despite well-known for the abundant flower shape, nearly all wild original crossing parents of modern chrysanthemum are characterized by multiple whorls of DF surround by a single whorl of peripheral ray florets. Of these wild original germplasm, C. lavandulifolium is one typical and vital diploid specie. The regulatory mechanism controlling the growth of ray florets in chrysanthemum is still not clear yet. Therefore, it is interesting and meaningful that the number and morphology of the ray florets of C. lavandulifolium are changed by ectopic activation of CmCYC2c. According to the reports by Tahtiharju et al. (2012), HaCYC2c and GhCYC3 are indicated to be strong candidates in controlling of ray floret identity owing to their specific expression in ray flowers. In sunflower, the expression of HaCYC2b and HaCYC2e are generally similar across different genotypes, whereas increased HaCYC2c expression is detected across the inflorescence of double-flowered (dbl) mutant (with only ray florets) (Chapman et al., 2008, 2012). Furthermore, overexpression of HaCYC2c due to an insertion upstream of the start codon generates dbl mutant, providing functional evidence for its role in ray formation (Chapman et al., 2012). In addition, GhCYC3 in Gerbera is highly up-regulated in the centermost ray floret primordia of the crested cultivar, which can convert DF into ray-like ones by promoting petal ligule growth in transgenic lines (Broholm et al., 2008).

The CmCYC2 subclade genes also share strikingly similar expression patterns during the early inflorescence developing stages (from SI to SIII). All these genes were highly expressed at stage IIa and then decreased gradually. However, CmCYC2c was also highly expressed at stage IV when compared with the other five genes. Since the petal of ray florets begin to differentiate at stage IIa and further expand at stage IV, we speculate that CmCYC2c may function in regulating petal development at both early and late stages. Similar functions have been reported previously for CYC and GhCYC3, which are involved in modifying petal growth owing to cell expansion or proliferation at different stages of development (Luo et al., 1996; Juntheikki-Palovaara et al., 2014). However, further studies are required to elucidate the expression patterns of the CmCYC2 genes at the organ and tissue level using, for example, RNA in situ hybridization.

An increasing number of evidence demonstrate that CYC-like genes participate in controlling flower symmetry and typically the growth of the dorsal parts of a flower as shown, for example, in A. majus (Luo et al., 1996, 1999) and I. amara (Busch and Zachgo, 2007). In Antirrhinum, for example, CYC and DICH regulate flower symmetry by promoting dorsal petal growth but repressing dorsalmost stamen development (Luo et al., 1996, 1999). Similarly, in Mohavea confertiflora, the expression of McCYC and McDICH in both dorsal and lateral regions is correlated with the abortion of both dorsal and lateral stamens and the increase of dorsal petal sizes (Hileman and Baum, 2003). In addition to controlling symmetry in a single flower, the TCP domain regulatory proteins in Asteraceae have developed a novel role in regulation of floral identity within the capitulum (Broholm et al., 2008; Fambrini et al., 2014b; Juntheikki-Palovaara et al., 2014; Garcês et al., 2016). Abundant expression data have shown that these members of CYC2 subclade in Asteraceae play a pivotal role in promoting or repressing the growth of ventral petal ligule, and sometimes affecting stamen development, hence change the symmetry of corresponding flowers (Broholm et al., 2008; Fambrini et al., 2014b; Juntheikki-Palovaara et al., 2014; Garcês et al., 2016).

Our transgenic data show that overexpression of CmCYC2c in C. lavandulifolium leads to significant elongation in ventral ligule length of ray florets. Based on the transgenic phenotype, the high expression of CmCYC2c in the petals of ray and trans-like florets appears to be crucial for the ventral petal ligule growth. This is in consistent with the results in Gerbera, in which overexpression of GhCYC3 and GhCYC4 enhances petal ligule growth in trans flowers (Juntheikki-Palovaara et al., 2014). Conversely, ectopic activation of GhCYC2, GhCYC3, and RAY1 can reduce ventral petal growth of corresponding ray florets (Kim et al., 2008; Juntheikki-Palovaara et al., 2014). Moreover, in sunflower and gerbera, overexpression of HaCYC2c, GhCYC2, and GhCYC3 can cause their DF to obtain some characteristics of ray florets, while no effect is observed on the DF of RAY1, RAY2, and RAY3 transgenic Senecio lines (Chapman et al., 2012; Juntheikki-Palovaara et al., 2014; Garcês et al., 2016). In this study, although one transgenic line generated some trans-like flowers, no difference was observed in DF between transgenic lines and control lines. Therefore, it appears that the CYC2 subclade proteins in Asteraceae have experienced functional diversification. Similar functional deviations have also been reported in regulation of stamen development (Juntheikki-Palovaara et al., 2014; Garcês et al., 2016). Besides promoting ventral petal length of ray florets, ectopic activation of CmCYC2c also statistically increased the number of ray florets per capitulum. Several studies have shown that the showy marginal flowers in Asteraceae increases the attractiveness of the inflorescence to pollinators, thereby increasing genetic diversity and fitness (Marshall and Abbott, 1984; Bremer and Anderberg, 1994; Andersson, 2008). Contrary to the single whorl of ray florets with changeless shape in the wild original species, the modern chrysanthemum cultivars are characterized by the various whorls of ray florets and the diversity of petal shape. Taking together the high expression levels of CmCYC2c in ray florets of the double-ray flowered chrysanthemum and the transgenic phenotypes in C. lavandulifolium, we can postulate that the CmCYC2c could be a vital regulator in evolution of modern chrysanthemum.

CYC2 clade proteins have been shown to influence vegetative and reproductive growth to produce morphological novelties in plants (Cubas et al., 1999b; Costa et al., 2005; Zhong and Kellogg, 2015). In this study, over expression of CmCYC2c in C. lavandulifolium resulted in retarded vegetative growth and reduced branching of young plantlets, which were restored gradually later, fortunately. Similar effect has been reported on many other TCP transcription factors. For example in Gerbera, constitutive expression of GhCYC2 causes severe growth defects and difficulties in regeneration of transgenic shoots (Broholm et al., 2008). Similarly, ectopic activation of CYC-like genes in Arabidopsis leads to reduction in vegetative growth (Costa et al., 2005; Yang et al., 2012; Juntheikki-Palovaara et al., 2014). Thus, at least in vegetative organs, their function is conserved. However, negative regulation of bud formation or branching outgrowth is mainly documented for some other genes in CYC/TB1 group, such as TB1 in maize (Doebley et al., 1995; Hubbard et al., 2002), OsTB1 in rice (Takeda et al., 2003) and BRC1 in Arabidopsis (Aguilar-Martínez et al., 2007). Conversely, heterologous expression of CYC1C in Arabidopsis promotes the outgrowth of lateral branches (Yang et al., 2012). Therefore, it seems like that ECE clade genes have undergone functional divergence in this content.

Intriguingly, overexpression of CmCYC2c can strongly induce the expression of CmCYC2f in all transgenic lines. According to the reports in Primulina, CYC1C and CYC1D can positively autoregulate themselves and cross-regulate each other to form auto-regulatory loops, which may trigger threshold-dependent genetic witches (Yang et al., 2012). In addition, Tahtiharju et al. (2012) shows that Gerbera CYC2 clade proteins have the ability of interacting in yeast two-hybrid assays, and the co-regulators may target different downstream genes. Therefore, we postulate that CmCYC2f may work as a putative downstream target of CmCYC2c to induce the differentiation of specific flower in chrysanthemum. Functional analysis of CmCYC2f is currently underway.