Comprehensive characterization of a floral mutant reveals the mechanism of hooked petal morphogenesis in Chrysanthemum morifolium.

The diversity of form of the chrysanthemum flower makes this species an ideal model for studying petal morphogenesis, but as yet, the molecular mechanisms underlying petal shape development remain largely unexplored. Here, a floral mutant, which arose as a bud sport in a plant of the variety 'Anastasia Dark Green', and formed straight, rather than hooked petals, was subjected to both comparative morphological analysis and transcriptome profiling. The hooked petals only became discernible during a late stage of flower development. At the late stage of 'Anastasia Dark Green', genes related to chloroplast, hormone metabolism, cell wall and microtubules were active, as were cell division-promoting factors. Auxin concentration was significantly reduced, and a positive regulator of cell expansion was down-regulated. Two types of critical candidates, boundary genes and adaxial-abaxial regulators, were identified from 7937 differentially expressed genes in pairwise comparisons, which were up-regulated at the late stage in 'Anastasia Dark Green' and another two hooked varieties. Ectopic expression of a candidate abaxial gene, CmYAB1, in chrysanthemum led to changes in petal curvature and inflorescence morphology. Our findings provide new insights into the regulatory networks underlying chrysanthemum petal morphogenesis.

Introduction

The lateral organs, including leaves and flowers, originate from the peripheral zone of shoot apical meristem (SAM) (Brand et al., 2001). Boundaries are established to separate these organs from the meristem and adjacent organs. A class of transcription factors specifically expressed at the boundaries play critical roles in organ separation, leaf development and floral organ patterning (Townsley and Sinha, 2012). In Arabidopsis thaliana, PETALLOSS (PTL) is expressed in boundaries between sepal primordia and indirectly promotes petal initiation (Brewer et al., 2004; Lampugnani et al., 2012). The expression of the boundary genes CUP‐SHAPED COTYLEDON1/2/3 (CUC1/2/3), which encode NAC family transcription factors, defines the boundaries of the petal domain (Aida et al., 1999; Takada et al., 2001). Similarly, the tomato (Lycopersicon esculentum) CUC homolog GOBLET (GOB) specifies the leaflets’ boundary and influences both leaf and petal development (Berger et al., 2009). SUPERMAN (SUP) maintains floral whorl boundaries by regulating auxin biosynthesis, and the loss‐of‐function mutation of SUP leads to extra stamens in the position of carpel (Sakai et al., 1995; Xu et al., 2018).

Lateral organs grow along three axes, including the proximal–distal axis, adaxial–abaxial axis and medial–lateral axis. The adaxial–abaxial polarity is established when the lateral organ initiates at the flank of SAM, affecting the growth of lateral organs. Several families of transcription factors and small RNAs are involved in the specification of adaxial–abaxial polarity. The class III HD‐ZIP family genes (REVOLUTA (REV), PHABULOSA (PHB) and PHAVOLUTA (PHV)) and ASYMMETRIC LEAVES genes (AS1 and AS2) determine the adaxial cell fate (Iwakawa et al., 2007; Kidner and Martienssen, 2004; Machida et al., 2015; McConnell et al., 2001). The transcripts of REV, PHB and PHV are cleaved by miR165/166 in the abaxial domain and restricted to the adaxial domain (Kidner and Martienssen, 2004). The as1 and as2 mutants exhibit lobed and downwardly curled leaves, and adaxial development of leaves is slightly diminished (Byrne et al., 2002; Semiarti et al., 2001; Serrano‐Cartagena et al., 1999). KANADI 1‐4 (KAN1‐4) expressed in the abaxial domain of lateral organs promotes abaxial fate specification (Eshed et al., 2004; Izhaki and Bowman, 2007). The auxin response factors AUXIN RESPONSE FACTOR 3/4 (ARF3/4) and YABBY (YAB) family are also involved in abaxial domain determination (Pekker et al., 2005; Sawa et al., 1999; Siegfried et al., 1999). Loss‐of‐function mutation in FILAMENTOUS FLOWER (FIL/YAB1) results in filamentous leaves and flowers, and the fil yab3 double mutant shows narrow leaves with defects of polar differentiation (Siegfried et al., 1999). ARF3, ARF4 and YABs are all expressed in the abaxial domain of lateral organs, in a pattern complementary to that of the REV, PHB and PHV genes (Bowman, 2000; Eshed et al., 2001; Pekker et al., 2005). The establishment and maintenance of adaxial–abaxial polarity are based on the mutual antagonistic interactions between adaxial and abaxial genes. KAN genes negatively regulate HD‐ZIP III genes and AS2 expression, while AS2 and HD‐ZIP III genes negatively regulate KANs (Eshed et al., 2004; Wu et al., 2008). Most studies of polarity genes focus on leaf development, but the function in petal morphogenesis is largely unknown.

Vasculature patterning is known to influence the morphogenesis of lateral organs (Dengler and Kang, 2001). The local concentration of a variety of phytohormones (primarily auxin) and the activity of various transcription factors act together to determine vascular development (Lucas et al., 2013). The abaxial–adaxial genes are also involved in vasculature patterning. For example, KANs are expressed in the phloem of the vasculature, while REV, PHB and PHV genes are transcribed in the xylem of the vasculature, regulating vascular development (Emery et al., 2003; Izhaki and Bowman, 2007). Another two members of the HD‐ZIP III family, AtHB8 and ATHB15 exclusively expressed in vascular tissues, play critical roles in vasculature patterning (Baima et al., 1995; Ohashiito and Fukuda, 2003).

Chrysanthemum (Chrysanthemum morifolium) is an economically valuable ornamental plant of Asteraceae (Funk, 2009; Teixeira da Silva, 2003). Its inflorescence is formed by a set of marginal ray florets and a set of central disc florets. Taking advantage of next‐generation sequencing, a number of differentially expressed genes were identified from a pairwise comparison of ray florets and disc florets of chrysanthemum (Liu et al., 2016). Our previous study showed that auxin and some key transcription factors were involved in the growth of chrysanthemum petals (Wang et al., 2017). In addition, CYCLOIDEA (CYC)‐like genes have been reported as being involved in the regulation of flower symmetry in Asteraceae species (Broholm et al., 2008; Huang et al., 2016; Juntheikki‐Palovaara et al., 2014; Kim et al., 2008). However, the molecular basis of petal morphogenesis in chrysanthemum remains largely unknown. Here, morphological analysis and transcriptome profiling were combined to characterize a petal‐shaped mutant in which the petals were straight and completely fused rather than hooked and nonfused in the distal domain. The differences in petal forms can be discerned at the late stage of flower development. Auxin and cell division and expansion were implicated in hooked petal morphogenesis. Key candidate regulator boundary genes and adaxial–abaxial genes have been identified and differently expressed in another hooked petal varieties. Ectopic expression of CmYAB1, a candidate abaxial gene, in chrysanthemum resulted in flat petals and spherical inflorescence. This study provides an overview of the morphogenesis of the hooked petal and identifies candidate genes of relevance to the breeding of varieties producing inflorescences of novel form.

Results

Morphological and cytological characterization of bud sport mutant

Bud sports differing in colour do arise occasionally in chrysanthemum, but those exhibiting variation in floral form are rare. Such a bud sport mutant was identified in the variety ‘Anastasia Dark Green’ (‘ADG’) (Figure S1). The mutation was fixed by regenerating plants from mutant petal explants: this variety is hereafter referred to as ‘MADG’. The ray floret petals of ‘ADG’ are tube‐like with hooked ends, while those of ‘MADG’ are tubular with straight ends. The ray petals are green in ‘ADG’, while they are white in ‘MADG’ (sometimes they are light pink). There are no other visible differences between ‘ADG’ and ‘MADG’. When inflorescences were sampled at the point where their diameter was ~2 mm (S2 stage), it was observed that ring‐shaped petal primordia had begun to form in the outermost whorls in both ‘ADG’ and ‘MADG’ (Figure 1a1–d2). Petals sampled from an S5 inflorescence (~5 mm in diameter) formed a Y shape at their ends, and once again, there was no discernible difference between ‘ADG’ and ‘MADG’ petals (Figure 1e1–g2). However, by S10 (~10 mm in diameter), whereas the dorsal domain of the distal petal remained nonfused in ‘ADG’, it was fused in ‘MADG’ (Figure 1h1–k2). By the early blooming stage (SEB), the petals had begun to curve inward and take on the hooked form in ‘ADG’, while they remained tube‐like in ‘MADG’ (Figure 2a,b). At the full blooming stage (SFB), the ‘ADG’ petals were tubular with nonfused, hooked ends, while in ‘MADG’, they were tubular and straight with fused ends (Figure 2c–f). These observations showed that a distal separation of the petal appeared before the formation of hooked petals and the distal domain of petal curved to form a hook at the SEB stage.

Figure 1

Morphological analysis of the ‘ADG’ and ‘MADG’ inflorescence at the stages of S2, S5 and S10. (a1‐d1, a2‐d2) The inflorescence and ray florets at S2 of (a1‐d1) ‘ADG’ and (a2‐d2) ‘MADG’; (a1, a2) top elevation of the inflorescence, (b1, b2) side elevation of the inflorescence, (c1, c2) SEM view of the inflorescence, (d1, d2) SEM view of the ray florets. The red arrows indicate ring‐shaped petal primordia. Bar: 0.5 mm in a1, a2, b1, b2; 0.1 mm in c1, c2, d1, d2. (e1‐g1, e2‐g2) The inflorescence and ray florets at S5 of (e1‐g1) ‘ADG’ and (e2‐g2) ‘MADG’; (e1, e2) side view of the inflorescence. The insets illustrate the outermost ray florets. (f1, f2, g1, g2) SEM view of the inflorescence. Bar: 1 mm in e1, e2, f1, f2, and 0.1 mm in g1, g2. (h1‐k1, h2‐k2) The inflorescence and ray florets at S10; (i1, k1, i2, k2) the outermost ray florets of (i1, k1) ‘ADG’ and (i2, k2) ‘MADG’. The red arrow in i1 and k1 shows a nonfused petal and in i2 and k2 a fused petal; (j1, j2, k1, k2) SEM images of the inflorescence and ray florets. Bar: 1 mm.

Figure 2

Morphological analysis of the ‘ADG’ and ‘MADG’ inflorescence at the SEB and SFB stages. (a) The inflorescences of ‘ADG’ (on the left) and ‘MADG’ (on the right) at SEB. Bar: 1 cm. (b) The outermost ray florets of ‘ADG’ (on the left) and ‘MADG’ (on the right) at SEB. Bar: 1 mm. The arrow shows the rudimentary hooked petal, and the arrowhead indicates the fused petal. (c and d) The inflorescences of (c) ‘ADG’ and (d) ‘MADG’ at SFB. Bar: 1 cm. (e, f) Ray florets in different whorls of (e) ‘ADG’ and (f) ‘MADG’ at SFB. Bar: 1 cm.

To further explore the differences in petals between ‘ADG’ and ‘MADG’, we observed the vein patterning of petals. The veins were arranged in parallel in the petals (Figure 3a–d). The secondary veins and tertiary veins initiated at the base of the hook domain in ‘ADG’ and distributed throughout the malformed hook (red box in Figure 3a), while branched veins appeared at the distal petal margin in ‘MADG’ (red box in Figure 3b). Phloroglucinol staining showed that the lignified veins in ‘MADG’ were thicker than those in ‘ADG’ (Figure 3c,d). Transverse sectioning of petals sampled at the SEB stage further proved this result. Although the number of vascular bundles present in ‘ADG’ petals was not significantly different from that in ‘MADG’, the xylem did appear larger in ‘MADG’ (Figure 3e,f); an average of 3.24 vessel elements was identified in the ‘ADG’ petals, while an average of 5.04 vessel elements was identified in the ‘MADG’ petals (Figure 3i). The ‘MADG’ adaxial epidermis consisted of a set of rounded rectangle cells, while ‘ADG’ adaxial epidermis cells were much smaller, and the cell length and width were both significantly reduced (t‐test, n ≥ 40, P < 0.01) (Figure 3e–h,j). The results suggested that vein patterning and cell size are different between the hooked petals and straight petals.

Figure 3

Morphological observation of ray petal venation and histological analysis of ray petal. (a–d) The distal petals of (a, c) ‘ADG’ and (b, d) ‘MADG’ ray florets show different venation patterns. The red boxes in (a, b) indicate the position where secondary veins and tertiary veins initiate. (c, d) Petals stained by phloroglucinol–HCL indicate lignified venation. Bar: 2 mm. (e, f) Transverse sections of the petals of (e) ‘ADG’ and (f) ‘MADG’. The arrows indicate the upper epidermis (adaxial epidermis). The arrowheads indicate the vessel elements. ad, adaxial; ab, abaxial. Bar: 100 μm. (g, h) SEM view of adaxial epidermis cells of (g) ‘ADG’ and (h) ‘MADG’ showed larger rounded rectangle cells in ‘MADG’. Bar: 100 μm. (i) Bar chart of the number of veins and vessel elements; (j) Bar chart of length and width of adaxial epidermis cells. ** Significant difference (t‐test, n ≥ 40, P < 0.01).

RNA‐Seq analysis of the ‘ADG’ and ‘MADG’ ray floret transcriptome

RNA prepared from the outermost ray florets of ‘ADG’ and ‘MADG’ sampled at the S2, S5, S10 and SEB stages provided the template for RNA‐Seq analysis. A total of 1.33 × 109 clean reads were generated from 24 samples (three replicates of both varieties sampled at each of the four stages), with each sample producing a minimum of 43 932 812 clean reads (Table S1). The outcome of the assembly procedure was a set of 175 763 unigene sequences of mean length 668 nucleotides (nt); the N50 was 963 nt. Transcript abundances (as estimated from the number of fragments per kilobase of transcript per million mapped reads, or FPKM) were highly correlated (Pearson's correlation coefficient between 0.82 and 0.89) between replicate samples (Figure 4a). A principal component analysis (PCA) of all the expressed unigenes (FPKM > 0.3) showed that 24 samples formed four groups; samples at the same stage were grouped together; stage S5 and S10 were closer than other adjacent stages; and the area occupied by stage SEB was largest (Figure 4b). These results indicated that the gene expression profiles at S5 and S10 were more similar to one another than those in any other pair of stages, and the SEB is the stage in which the gene expression profiles of ‘ADG’ and ‘MADG’ differ the most.

Figure 4

Global analysis of the transcriptome data in the ray florets sampled from ‘ADG’ and ‘MADG’. (a) Pearson's correlation between 24 samples. (b) Principal component analysis of gene expressions in ray florets from ‘ADG’ and ‘MADG’. FPKM values derived from samples taken at S2, S5, S10 and SEB are shown as dots where they were recovered from ‘ADG’ and as triangles from ‘MADG’.

Transcriptomic differentiation during flower development of ‘ADG’ and ‘MADG’

Differentially expressed genes (DEGs) were identified from pairwise comparisons of adjacent or the same development stage in ‘ADG’ and ‘MADG’ through the DESeq R package with a false discovery rate (FDR) value below 0.05 (Table S2) (Anders and Huber, 2010). The accuracy of the DEGs identified by RNA‐Seq was verified using the RT‐PCR analysis (Figure S2). In ‘ADG’, 7704 genes were up‐regulated and 6946 genes were down‐regulated in G_S2 vs G_S5 (Figure 5a). The number of DEGs was decreased in G_S5 vs G_S10 and then increased in G_S10 vs G_SEB: 2468 and 9421 DEGs were up‐regulated and 3874 and 8852 genes were down‐regulated in G_S5 vs G_S10 and G_S10 vs G_SEB, respectively (Figure 5a). In ‘MADG’, the trend of the number of DEGs was similar to that of ‘ADG’: lowest in M_S5 vs M_S10 and highest in M_S10 vs M_SEB.

Figure 5

DEGs analysis of adjacent development stages in ‘ADG’ and ‘MADG’. (a) Histogram illustrating number of DEGs up‐regulated and down‐regulated in pair comparisons of ‘ADG’ and ‘MADG’. (b, c) GO term enrichment of DEGs that are up‐regulated (b) and down‐regulated (c) in pair comparisons. The heat maps show the enrichment in each GO term, which is indicated by FDR value. Triangles indicate microtubule‐related terms, and stars indicate the different GO terms between ‘ADG’ and ‘MADG’.

To explore the molecular differentiation during flower development, these DEGs were further functionally classified through Gene Ontology (GO) analysis. Genes related to oxidoreductase activity and catalytic activity were highly overrepresented among the genes that were up‐regulated in G_S2 vs G_S5 (Figure 5b). Genes related to DNA binding, transcription factor complex and macromolecular complex were up‐regulated in both the G_S5 vs G_S10 and the G_S10 vs G_SEB pairwise comparisons (Figure 5b). Genes related to microtubule were specifically up‐regulated in G_S10 vs G_SEB (Figure 5b). Functional classification of up‐regulated DEGs in ‘MADG’ was similar to that in ‘ADG’ (Figure 5b). Genes related to DNA binding, microtubule and tubulin are down‐regulated in G_S2 vs G_S5 (Figure 5c). Genes related to oxidoreductase activity, catalytic activity and hormone metabolic process were down‐regulated in both G_S5 vs G_S10 and G_S10 vs G_SEB (Figure 5c). Genes related to chloroplast, plastid and cell wall were down‐regulated in G_S10 vs G_SEB (Figure 5c). Down‐regulated DEGs in ‘MADG’ were generally similar to those in ‘ADG’ except three types of genes: (1) genes related to chloroplast, thylakoid and plastid; (2) genes related to cell wall; and (3) genes related to regulation of hormone levels and hormone metabolic process (Figure 5c). The enrichment of genes related to plastid and chloroplast in ‘ADG’ was in accordance with the colour of ray florets, which were green in ‘ADG’ petals, while they were white in ‘MADG’ petals. In addition, hormone‐involved processes may contribute to the formation of hooked petals.

Transcriptomic differentiation between ‘ADG’ and ‘MADG’ at the same development stage

DEGs were identified from comparison of transcriptomes between ‘ADG’ and ‘MADG’ at the same development stage. A total of 7937 genes were differentially transcribed in pairwise comparison between ‘ADG’ and ‘MADG’. A total of 307 DEGs were more abundant in ‘MADG’, and 125 were more abundant in ‘ADG’ at S2. The number of DEGs was gradually increased at S5, S10 and SEB (Figure 6a). The more DEGs at late stage were consistent with the observation that differences in petal shape only developed later during flower development.

Figure 6

DEG analysis of the same development stages between ‘ADG’ and ‘MADG’. (a) Bar plot shows the number of DEGs up‐regulated and down‐regulated between ‘MADG’ and ‘ADG’. (b) Cluster heat map of DEGs between ‘MADG’ and ‘ADG’. Six groups are clustered according to the gene expression pattern.

Cluster analysis was performed using pheatmap R package (Kolde, 2015). The 7937 DEGs were clustered into six coexpression modules (Figure 6b, Table S3). GO enrichment analysis of six modules showed that genes related to cell wall organization and biofilm formation were enriched in C1 modules in which genes were highly expressed at all four stages, especially at S5 and S10 in ‘MADG’ and down‐regulated in ‘ADG’ (Figures 6b and S3). Higher expression of cell wall organization‐ and biofilm‐related genes was consistent with larger epidermal cells in ‘MADG’ as shown by cytological analysis (Figure 3). Genes related to the regulation of transcription and microtubule were enriched in C4 modules in which genes were highly expressed in S5 and S10 and down‐regulated at S10 and SEB in ‘MADG’ (Figures 6b and S3). “Regulation of transcription” and “photosynthesis” were the most enriched GO terms in the C6 module in which genes were up‐regulated at SEB in ‘ADG’ (Figures 6B and S3), suggesting that photosynthesis and transcription were more active at the SEB stage of ‘ADG’ correlating with green petals.

Microtubule‐related genes were enriched in the C4 module, and the two MAP65‐3 homologs (c76593_g1 and c57118_g1) were up‐regulated at the SEB stage in ‘ADG’ (Figure 7a). We also identified development‐related genes in the C4 module and other modules differently expressed between ‘ADG’ and ‘MADG’ (Figure 7a,b, Table S4). Homologs of GROWTH REGULATING FACTOR5 (GRF5) (c55992_g1) and GRF1‐interacting factor 1 (GIF1) (c70468_g1) were more abundant at the SEB stage in ‘ADG’; homologs of GRF2 (c69176_g2) were more abundant at both the S10 and SEB stages in ‘ADG’ (Figure 7a). A gene encoding AINTEGUMENTA (ANT) (c52275_g1) homolog was down‐regulated at SEB in ‘MADG’, while a gene encoding ATHB12 (c50839_g1) was up‐regulated at the S10 and SEB stages in ‘MADG’ (Figure 7a). MADS‐box transcription factors are essential for specifying floral organ identity and subsequent development, and we found that MADS‐box genes were differently expressed at all four stages. Homologs (c52986_g2, c52986_g1, c52253_g2, c56196_g1) of PISTILLATA (PI) and APETALA3 (AP3) were up‐regulated at the S2 and SEB stages (Figure 7b). One SEPALLATA3 (SEP3) homolog (c68326_g3) was up‐regulated at the S2, S5 and S10 stages; another two homologs (c68326_g4, c67344_g2) of SEP3 were down‐regulated at the S5, S10 and SEB stages (Figure 7b).

Figure 7

The transcription intensity of key DEGs across developmental stages of the ‘ADG’ and ‘MADG’ inflorescence, derived from the RNA‐Seq analysis. (a–d) Heat maps illustrating the transcription intensity of (a) genes important for floral development, (b) genes encoding MADS‐box transcription factors, (c) genes encoding boundary‐ and adaxial‐/abaxial‐related transcription factors and (d) genes encoding products involved in auxin pathway at different stages of flower development

The unfused petals suggest that boundary genes promoting organ separation may be involved in hooked petal formation. To investigate whether boundary genes participate in petal morphogenesis in chrysanthemum, homologs of boundary genes were identified from the C1‐C6 modules (Table S4). Genes encoding homologs of CUC2 and CUC3 were differently transcribed in the ‘MADG’ vs ‘ADG’ pairwise comparison. CUC2 homologs (c65461_g2) were up‐regulated at S5 in ‘ADG’; CUC3 homologs (c68632_g1) were up‐regulated at S10 and SEB in ‘ADG’; additional CUC2 and CUC3 homologs (c74119_g5, c73109_g3) were both up‐regulated at SEB in ‘ADG’ (Figure 7c). Genes encoding homologs of PTL (c74128_g1, c58673_g4) were up‐regulated at S10 or SEB in ‘ADG’ (Figure 7c). SUP homologs (c49259_g2, c49259_g4) were up‐regulated at S10 and SEB stages in ‘ADG’ (Figure 7c). The up‐regulation of boundary genes from S5 to SEB was consistent with the appearance of unfused petals at S10, indicating that boundary genes were involved in hooked petal formation.

Given that the distal ends of the petals were curved inward in ‘ADG’, adaxial–abaxial patterning genes were identified from C1 to C6 modules (Table S4). The expression levels of abaxial determinants YAB1, YAB4, YAB5, ARF3 and ARF4 homologs (c47017_g1, c64740_g1, c67661_g1, c70055_g1, c68774_g1) were higher at S10 stage in ‘ADG’ than in ‘MADG’ (Figure 7c). Genes encoding homologs (c65487_g1, c62902_g1) of abaxial determinants KAN2 and KAN4 together with YAB5 and ARF3 homologs (c67661_g1, c70055_g1) displayed higher expression at the SEB stage in ‘ADG’ (Figure 7c). Gene encoding homolog of adaxial determinant AS1 (c66895_g1) was up‐regulated at the SEB stage in ‘ADG’ (Figure 7c). A homolog of WOX1 (c70195_g1), which was expressed in the middle domain in cooperation with adaxial‐ and abaxial‐specific regulators to regulate lamina outgrowth, was up‐regulated at the S10 stage in ‘ADG’ (Figure 7c) (Nakata et al., 2012). Gene encoding homolog of ZPR1, forming a feedback regulatory loop with adaxial gene REV, was up‐regulated at S5, S10 and SEB stages in ‘ADG’ (Figure 7c) (Wenkel et al., 2007). These results indicate that the curvature of distal petals in ‘ADG’ might be caused by polarity genes differently expressed between ‘ADG’ and ‘MADG’.

We also found that genes encoding homologs of ATHB15 (c66734_g1, c67902_g1, c44189_g1, c11680_g1, c58457_g1 and c67902_g1), a negative regulator of vascular cell differentiation, were down‐regulated at SEB or both S10 and SEB stages in ‘MADG’ (Figure 7a). A large number of homologs probably acting downstream of ATHB15 were more strongly expressed in ‘MADG’ than in ‘ADG’ at SEB (Table S5) (Avci et al., 2008; Ruiqin and Zheng‐Hua, 2015). The transcriptional behaviour of all of these putative downstream genes contrasted with that of the ATHB15 homologs, which was consistent with the idea that the products of ATHB15 homologs act to repress vascular development. These data were consistent with the more abundant vessel elements in ‘MADG’ and indicated that the pathway mediated by ATHB15 may be involved in vasculature development.

In addition, our data showed that genes related to hormone metabolism and regulation of hormone levels exhibited different expression patterns in ‘ADG’ and ‘MADG’ (Figure 5c). Thus, we identified auxin‐related genes from DEGs in the pairwise comparison between ‘ADG’ and ‘MADG’. Among these genes, ARF5 homologs (c53241_g1, c50683_g1, c63258_g1) were up‐regulated at the SEB stage in ‘MADG’ (Figure 7d). ARF5 promotes vascular strand formation by regulating PIN1 and ATHB8 (Donner et al., 2009). Accordingly, the expression level of the three ARF5 homologs was lower in ‘ADG’, and vascular development was limited. Furthermore, the concentration of indoleacetic acid (IAA) was reduced at the SEB stage in ‘ADG’ (Figure S4).

A part of the DEGs that showed repeated differential expression cannot be annotated by BLAST using the NR or Swiss‐Prot database. To obtain annotations, we performed a BLAST analysis of the unigenes to the Helianthus annuus genome which is a member of Asteraceae with annotations. Some unigenes can be successfully annotated (Table S2). Among these DEGs, c33869_g1, a GRF zinc finger protein, was up‐regulated in the four stages of ‘MADG’ compared to ‘ADG’. c58673_g2, a probable duplicated homeodomain‐like superfamily protein, was down‐regulated in G_S10 vs G_SEB and up‐regulated in M_S5 vs G_S5 and M_S10 vs G_S10. These DEGs may function in petal morphogenesis and are specific for Asteraceae.

Validation of the expression profiles of key candidates in other varieties

To further investigate the expression profiles of these key candidates, we performed qRT‐PCR in other varieties with similar phenotypes to ‘ADG’ and ‘MADG’. ‘Jierilihua’ and ‘Quanxiangjiliu’ are traditional cultivars with typical hooked petals (Figure 8a–d); ‘Ziyunfeiyue’ and ‘Jinsongyue’ are famous for their straight tubular petals (Figure 8e–h). The abundance of YAB1 and ATHB15 homolog (c47071_g1, c68632_g1) transcripts was higher in ‘Jierilihua’ and ‘Quanxiangjiliu’ than in ‘Ziyunfeiyue’ and ‘Jinsongyue’ at S10 and SEB stages (Figure 8i). Genes encoding CUC3 and ANT homologs (c67902_g1, c52275_g1) were down‐regulated at the SEB stage in ‘Ziyunfeiyue’ and ‘Jinsongyue’ (Figure 8i). Higher expression of these candidates at S10 or SEB in the varieties with hooked petals was similar to that in ‘ADG’. These results further suggested that these genes were associated with hooked petal morphogenesis.

Figure 8

The expression profiles of key candidate genes in other varieties. (a–d) The inflorescence and ray florets (upper part) of ‘Jierilihua’ (a, b) and ‘Quanxiangjiliu’ (c, d), showing the hooked petals. (e–h) The inflorescence and ray florets (upper part) of ‘Ziyunfeiyue’ (e, f) and ‘Jinsongyue’ (g, h), illustrating the straight tubular petals. (i) The heat map based on qRT‐PCR data shows expression level of key candidate genes in the above four varieties.

Ectopic expression of a polarity gene CmYAB1 affected the petal curvature

To further examine the function of the polarity gene YAB1 homolog (CmYAB1) in chrysanthemum, CmYAB1 driven by the cauliflower mosaic virus (CaMV) 35S promoter was introduced to ‘Jinba’, a cultivar of chrysanthemum with spoon‐like ray flowers and decorative inflorescence by Agrobacterium‐mediated leaf disc transformation (Wang et al., 2019). We obtained thirteen transgenic lines, and three of nine lines with similar phenotype were chosen for further analysis. CmYAB1 was up‐regulated by 1.6 to 2.0 fold in the transgenic lines (Figure 9k). The inflorescence morphology of transgenic lines was spheroidal and similar to pompon inflorescence (Figure 9a–h). The petals of ray florets were flat and even epinastic (Figure 9i,j), and the distance between petal margins was largerthan that in wild type (Figure 9l). These results showed that CmYAB1 did control the curvature of petals of ray florets and inflorescence morphology.

Figure 9

Ectopic expression of Cm affected the petal curvature and inflorescence shape. (a–d) Lateral view of wild type (WT) and transgenic lines 1 to 3 show different inflorescence shapes. (e–h) Top view of wild type (WT) and transgenic lines 1 to 3, the abaxial side of ray petal is visible from the top view in WT (arrowhead), but not visible in transgenic plants. (i) The top elevation of outmost ray florets of WT and transgenic lines. Left pane is WT, and right pane is transgenic plants. The petals of Cm overexpression plants are flat. Red range line indicates the distance between petal margins measured in (l). (j) The side elevation of outmost ray florets of WT (left) and transgenic lines (right) indicates reduced degree of curvature. (k) Expression level of Cm in WT and transgenic plants. (l) The histogram of distance between petal margins. ** Significant difference (t‐test, n > 100, P < 0.01). Bar: 2 cm.

Discussion

The hooked petal arises during the late stage of flower development

The morphological analysis failed to discern any difference between the ray florets formed by ‘ADG’ and ‘MADG’ during early flower development (S2 and S5) (Figure 1), a finding supported by the transcriptomic analysis, which revealed relatively low numbers of stage‐specific DEGs (432 in S2 and 995 in S5, compared to >6000 in SEB) (Figure 6). Rudimentary hooked petals did not become visible until the SEB stage (Figure 2). The implication is that the phenotype is not generated until the later stage of floral development, at a time when cell expansion predominates over cell division (Hill and Lord, 1989; Takeda et al., 2013; Yamada et al., 2009). Consistently, cell size in the ‘ADG’ petals was smaller than in those of ‘MADG’ (Figure 3g–j), indicating the defects in cell expansion.

Microtubules are thought to be closely related to cell division and expansion (Ambrose et al., 2007). MAP65‐3 is predominantly transcribed during cell division, and loss of function of MAP65‐3 frequently results in failures of cytokinesis in Arabidopsis (Caillaud et al., 2008; Ho et al., 2012; Li et al., 2017). In Gerbera, the organization of the cortical microtubules is important for the regulation of cell expansion and, hence, for petal growth (Zhang et al., 2012). Changes in the local concentration of phytohormones have been proposed to influence organ growth and cell expansion by orienting the cortical microtubule arrays (Foster et al., 2003; Kim et al., 2002; Shibaoka, 1994). The GO analysis in adjacent stages in ‘ADG’ and ‘MADG’ showed that microtubule‐related genes were up‐regulated at S5 and S10 stages (Figure 5b,c). Microtubule‐related terms were also enriched in the C4 module, in which genes were up‐regulated at the S10 and SEB stages in ‘ADG’ compared to ‘MADG’ (Figures 6b and S3). Genes related to cell wall were up‐regulated at the SEB stage in ‘ADG’ (Figure 5c). These results suggest that cell division or expansion is active at the S5 and S10 stages of flower development, and compared to ‘MADG’, cell division or expansion is more active at S10 and SEB in ‘ADG’. Arabidopsis transcription factor GRF5, which functions partially redundantly with other members of the GRF family, prolongs the duration of cell proliferation phase and increased leaf size (Kim and Lee, 2006; Vercruyssen et al., 2015). ANT gene delays the duration of cell division by sustaining CycD3 expression and promotes cell proliferation and organ growth (Mizukami and Fischer, 2000). However, ATHB12 promotes leaf cell expansion and endoreduplication, and its overexpression increases cell size (Hur et al., 2015). Our transcriptome data show that cell division‐promoting factors homologous to MAP65‐3, GRF2/5 and ANT were all up‐regulated at the S10 and SEB stages in ‘ADG’, and the cell expansion‐positive regulator ATHB12 homolog was down‐regulated at S10 and SEB in ‘ADG’ (Figure 7a). These findings indicate that proliferation of the petal cells rather than cell expansion is active at the late stages of flower development in ‘ADG’, causing smaller petal cells and thus promoting the formation of hooked petals. Taken together, these findings show that at least some of the DEGs involved in microtubule‐based process, along with some transcription factors, are associated with petal morphogenesis, acting by regulating cell proliferation and expansion at the late stages of flower development.

Unfused distal petal regulated by boundary genes is a prerequisite for hooked petal morphogenesis

Prior to the formation of the rudimentary petal hook, the distal petals are not fused. Moreover, almost all the varieties with hooked petals are accompanied by unfused distal petals, but not all varieties with unfused distal petals have hooked petals. Therefore, the unfused distal petals may be a precondition for hooked petal formation. Among the DEGs, eight genes were identified as homologs of the four boundary genes PTL, CUC2, CUC3 and SUP. In A. thaliana, PTL is expressed at the sepal–sepal boundary, and loss of PTL function results in sepal fusion and a reduction—even to zero—in the number of petals formed (Brewer et al., 2004; Quon et al., 2017). The ‘MADG’ petals remained fused, and the abundance of the PTL homolog transcript at S10 and SEB remained lower than in ‘ADG’ (Figure 7c). A similar phenotype has been noted to be associated with the loss‐of‐function ptl mutant, which suggests that the product of the PTL homolog may contribute to petal fusion. The CUCs have been well conserved over evolution; their homologs in petunia (Petunia hybrida), snapdragon (Antirrhinum majus) and tomato (Lycopersicon esculentum) have been shown to have similar functions during meristem development, organ separation and leaf development (Berger et al., 2009; Blein and Laufs, 2008; Hasson et al., 2011; Souer et al., 1996; Weir et al., 2004). The overexpression of the tomato GOBLET gene (a CUC homolog) results in an increase in the number of petals formed due to petal separation. In ‘ADG’ from the S5 to the SEB stage, the transcriptions of CUC homologs were more abundant than in ‘MADG’ at the same stage (Figure 7c). Furthermore, the expression levels of CUC3 homologs were also higher in another two hooked petal cultivars than in another two fused straight petal cultivars at SEB stage (Figure 8). The observations suggest that the products of CUC homologs regulate petal fusion in chrysanthemum as they do in other plant species. SUPERMAN (SUP) maintains floral whorl boundaries by regulating auxin biosynthesis, and mutation of SUP leads to extra stamens in the position of the carpel (Sakai et al., 1995; Xu et al., 2018). Genes encoding homologs of SUP, involved in stamen and carpel development, were up‐regulated at S10 and SEB stages in ‘ADG’ (Figure 7c), suggesting that SUP may evolve a new function in ray petal development of chrysanthemum. In all, the boundary genes are required for hooked petal morphogenesis by promoting the separation of distal petal.

Vascular patterning mediated by ATHB15 homologs and auxin influences petal morphogenesis

Lateral organ morphogenesis depends strongly on the development of the vasculature (Dengler and Kang, 2001). There were distinct differences between the ray floret petal vasculature formed within the ‘ADG’ and the ‘MADG’ inflorescences, largely reflecting an increase of lignified vessels in the latter and different vein patterning (Figure 3). The A. thaliana gene ATHB15 is known to repress vascular development (Du et al., 2015; Kim et al., 2005). A comparison between the transcriptomes of inflorescences sampled at SEB showed that the abundance of the transcript produced by all six ATHB15 homologs was lower in ‘MADG’ than in ‘ADG’ (Figure 7a). The implication is that the abundance of ATHB15 homologs transcripts acts to inhibit vascular development. Reinforcing this conclusion, a number of genes, the products of which are predicted to act downstream of the ATHB15 homologs, were differentially transcribed at the SEB stage between ‘MADG’ and ‘ADG’ in the opposite direction shown by the ATHB15 homologs themselves (Table S5). Furthermore, the transcripts of ATHB15 homologs were more abundant at S10 and SEB stages in another two hooked petal cultivars than that in another two fused straight petal cultivars (Figure 8), suggesting that ATHB15 homologs were involved in the formation of hooked petals. Overall, it would appear that the products of the ATHB15 homologs likely function as repressors of vascularization, thereby affecting petal morphogenesis. Furthermore, auxin plays a critical role in vascular patterning (Aloni et al., 2003; Berleth et al., 2000), and auxin concentration was reduced at the SEB stage in ‘ADG’ (Figure S4). In addition, three homologs of ARF5, which is thought to promote the formation of vascular strands, were more abundantly transcribed at the SEB stage in ‘MADG’ than in ‘ADG’ (Donner et al., 2009) (Figure 7d). Thus, it seems likely that the auxin pathway also exerts some control over petal morphogenesis by regulating vascular development.

Adaxial–abaxial genes are involved in hooked petal morphogenesis

Genes involved in regulating leaf curvature have been found not only in the model plant Arabidopsis thaliana but also in the horticultural crop Brassica rapa. Previous comparative genomic analysis showed that abaxial genes BrARF3.1 and BrKAN2.1 play essential roles in leafy head formation in Brassica rapa (Cheng et al., 2016; Liang et al., 2016). In addition to the unfused distal petals, the petal curved inward and even formed a loop in ‘ADG’, suggesting that it is a polarity defect, at least in part, that causes the hooked petals. We identified several polarity genes that were differently expressed in ‘MADG’ and ‘ADG’. Genes encoding homologs of YABs, ARF3 and ARF4, specifying abaxial cell fate, were strongly transcribed at S10 in ‘ADG’ compared to ‘MADG’ (Figure 7c). The transcripts of WOX1 homologs, especially expressed in the middle domain, were more abundant at the S10 stage in ‘ADG’. However, no adaxial gene was identified at S10 in DEGs (Figure 7c). Until the SEB stage, only the homolog of the adaxial gene AS1, determining the adaxial cell fate, was up‐regulated in ‘ADG’ (Figure 7c). Nevertheless, the transcripts of other abaxial genes KAN2 and KAN4, which repressed the expression of AS2 (Wu et al., 2008), were more abundant in ‘ADG’ at SEB. AS1 together with AS2 directly binds to the promoter of ARF3 to repress the transcription of ARF3 (Iwasaki et al., 2013). Consistently, one homolog of ARF3 (c74255_g1) was down‐regulated in ‘ADG’ at SEB, although another homolog (c70055_g1) was still up‐regulated (Figure 7c). The AS1 homologs also repressed other abaxial genes (Iwasaki et al., 2013; Machida et al., 2015), consistent with the finding that YAB1, YAB4 and ARF4 were not differently expressed between ‘MADG’ and ‘ADG’ at SEB (Figure 7c). Furthermore, a YAB1 homolog was also up‐regulated at the SEB stage in another two hooked petal varieties (Figure 8). Ectopic expression of CmYAB1 influenced the petal curvature and inflorescence morphology (Figure 9). Thus, the polarity gene YAB1 regulates petal curvature, thereby playing a critical role in hooked petal morphogenesis. In all, abaxial genes in ‘ADG’ that interact antagonistically with adaxial genes contribute to the hooked petal formation by affecting the petal curvature during the late stage of flower development.

Petal development relies on the establishment of the petal domain, the specification of petal identity and cell differentiation (Alvarezbuylla et al., 2010). The participation of the products of AP1–3, PI and SEP1–4 in the determination of floral organ identity has long been recognized (Coen and Meyerowitz, 1991; Jack, 2004). Recent studies showed that these MADS‐box genes directly regulate abaxial–adaxial polarity genes, organ boundary genes and development‐related genes (Sablowski, 2015). AP3 and PI promote the expression of SUP and CUC (Ds et al., 2013). YAB1, AS1 and AS2 are the direct target genes of AP1 and SEP3. AP1 and SEP3 bind to GRF genes (Pajoro et al., 2014). In our study, the homologs of AP1, AP3/PI and SEPs were differentially expressed between ‘ADG’ and ‘MADG’ at the early stage S2, at which there was no difference between ‘ADG’ and ‘MADG’ (Figure 7b). Overexpression of the kiwifruit MADS‐box gene SVP results in increased chlorophyll content in petals (Wu et al., 2014). Loss of function of SEP3 homologs in Petunia hybrid and Phalaenopsis orchid leads to green tissues appearing in petals (Matsubara et al., 2008; Pan et al., 2014). Therefore, we speculate that mutations in MADS‐box genes lead to phenotypes of bud sport mutant through the regulation of certain downstream genes, including organ boundary genes and polarity genes.

Overall, the formation of hooked petals involves multiple developmental defects, and multiple genes collectively regulate the hooked petal morphogenesis, including auxin pathway genes, boundary genes and adaxial–abaxial genes. Based on the above results, we put forward a model of the morphogenesis of the hooked petal (Figure 10). During the earliest stage of floral development (S2), the petal primordia are initiated, forming a ring‐shaped structure; the petals continue to elongate up to and beyond S5. By S10, only if the distal petals are unfused can the flower develop hooked petals. By the time SEB is reached, the ‘MADG’ flower has developed straight tubular petals, but in ‘ADG’, the distal petals curve inward and develop rudimentary hooks. By SFB, the hooked petals fully develop. During the development of the hooked petal of ‘ADG’, genes related to hormone metabolism are up‐regulated at SEB, and genes related to the cell wall and chloroplast are active at SEB. Boundary genes and adaxial–abaxial genes are up‐regulated at S10 and SEB stages in ‘ADG’ and are crucial candidates in hooked petal morphogenesis.

Figure 10

Working model for the development of the hooked petal. The early stages of development (S2 and S5) of the hooked and nonhooked petal are similar, diverging at S10. At this point, the distal domain of the ray petal is separated in ‘ADG’. Once the SEB stage has been reached, the ‘ADG’ inflorescence has formed rudimentary hooked petals, while in ‘MADG’, the tubular form has been retained. At SFB stage, the hooked petal is fully developed. Genes related to hormone metabolic, cell wall and chloroplast are up‐regulated from stages S10 to SEB in ‘ADG’, but not found in ‘MADG’. Boundary genes and polarity genes differently expressed at S10 and SEB stages may be involved in hooked petal morphogenesis.

Materials and methods

Plant material and growing conditions

Plants of ‘ADG’, ‘MADG’, ‘Jierilihua’, ‘Quanxiangjiliu’, ‘Ziyunfeiyue’ and ‘Jinsongyue’ were grown in a greenhouse at the Nanjing Agricultural University Chrysanthemum Germplasm Resource Preservation Center (Jiangsu, China). To regenerate full plants from the ray petals of the ‘ADG’ bud sport, surface‐sterilized (0.1%(v/v) HgCl2) explants were cultured in Murashige and Skoog medium containing 1 mg/L 6‐benzylaminopurine and 0.5 mg/L 1‐naphthaleneacetic acid and were rooted by transferring to the same medium lacking both phytohormones. Petals sampled from the outermost whorl of the inflorescence were used both for morphological analysis and as a source of RNA.

Morphological and histological analysis

Whole inflorescences, or only the outermost ray florets, were sampled when inflorescences were ~2 mm, 5 mm and 10 mm in diameter (S2, S5, S10 stage) and at early blooming stage (SEB), and fixed in 2.5% v/v glutaraldehyde after removal of the bracts. After critical point drying and coating with gold, the samples were subjected to scanning electron microscopy using an SU8010 device (Hitachi, Japan). Petals of outermost ray florets were dissected and cleared in a solution containing 70% ethanol and 30% acetic acid for 12 h and then incubated twice in 70% ethanol for 4 h each time and dried through sandwiching in thick filter paper (Szecsi et al., 2006). Images of fresh samples and cleared petals were photographed using a S8AP0 light microscope (Leica Camera AG, Germany). Petals sampled at the SEB stage were fixed, embedded, sectioned and de‐waxed as described elsewhere (Ding et al., 2015). Sections were imaged using a DM1000 microscope (Leica Camera AG, Germany). For the lignified vessel element observation, cleared petals and sections were stained with phloroglucinol and HCL (2:1) as previously described (Du et al., 2015).

RNA extraction, transcriptome sequencing and bioinformatic analysis

The outermost ray florets of ‘ADG’ and ‘MADG’ were sampled at S5, S10 and SEB, while, because of their small size, S2 inflorescences were used in toto. Each stage and each variety were sampled in triplicate. RNA was extracted using an RNA Isolation Kit (Waryong, Beijing, China) and subjected to Illumina sequencing at Beijing Novogene (Tianjin, China) following the manufacturer's protocol. Each sample generated approximately 45‐69 million 150 bp pair‐end reads. The reads were filtered in silico to remove adaptor sequences, sequences harbouring runs of poly N and poor‐quality sequences. The transcriptome was assembled using Trinity (v2.4.0) software (Grabherr et al., 2011). Annotation was based on homology searches against the NCBI, NR, Nt and Swiss‐Prot databases using the BLASTx algorithm. Genes with an a FDR below 0.05, as identified by DESeq software (Anders and Huber, 2010), were assigned as DEGs. A PCA was performed using the prcomp function implemented in R software for all unigenes (FPKM > 0.3) (Rteam et al., 2014). All the DEGs of pairwise comparisons between ‘ADG’ and ‘MADG’ were put into the cluster analysis. The cluster heat map was generated by online OmicShare tool (http://www.omicshare.com/tools/Home/Soft/heatmap). DEG sequences were compared with the GO databases to search for functional enrichment. GO term enrichment analysis was performed using the online OmicShare tool (http://www.omicshare.com/tools/Home/Report/goenrich) with our own background GO documents. GO and expression heatmaps were created using the TBtools software with the ‘Log Scale’ and ‘Row Scale’. The outermost ray florets of other five varieties were also sampled at S5, S10 and SEB in triplicate, and RNA was extracted in the same way.

Quantitative RT‐PCR and IAA measurement

The expression analysis of several key candidate genes in another four varieties was performed using qRT‐PCR. A chrysanthemum gene encoding elongation factor 1α (EF1α) (KF305681) was used as the reference gene. A number of representative DEGs were selected to validate the RNA‐Seq analysis. The primers used for the qRT‐PCRs are given in Table S6. Each qRT‐PCR was represented by three biological and three technical replicates. Relative transcription abundance was derived using the 2−ΔΔCT method (Livak and Schmittgen, 2001). IAA content of ray florets at the S10 and SEB stages in ‘ADG’ and ‘MADG’ was measured as described previously (Wang et al., 2015).

CmYAB1 genetic transformation

The CmYAB1 coding sequence was first cloned into a pENTR 1A vector and then introduced into the pMDC43 vector by LR recombination. The pMDC43‐CmYAB1 vector was introduced to the chrysanthemum ‘Jinba’ by Agrobacterium‐mediated genetic transformation as described previously (Wang et al., 2019). After further screening by PCR at the DNA level, the positive transgenic seedlings were grown in a growth chamber for a week under 16 h/8 h and 25 °C/18 °C day/night and were then transferred to greenhouse under standard management. RNA was extracted from ray petals at the SFB stage, and qRT‐PCR was performed to examine the expression level of CmYAB1 with gene‐specific primers. The outermost ray florets were used to measure the distance between petal margins.

Accession numbers

Sequencing data of this study can be found at the National Center for Biotechnology Information (https://www.ncbi.nlm.nih.gov/bioproject/) with the BioProject ID PRJNA505717.

Funding information

This research was funded by the National Science Fund for Distinguished Young Scholars (31425022), the National Natural Science Foundation of China (31701959), the Natural Science Fund of Jiangsu Province (BK20170717), the China Postdoctoral Science Foundation (2017M611843) and the Fundamental Research Funds for the Central Universities (KJSY201705, KJQN201815).

Conflict of interest

The authors declare no conflict of interest.

Figure S1 The bud sport mutant of the variety ‘Anastasia Dark Green’ (‘ADG’).

Figure S2 Validation of the RNA‐Seq classification of differential transcription via qRT‐PCR.

Figure S3 Functional category of six clusters derived from DEGs between ‘MADG’ and ‘ADG’.

Figure S4 Auxin measurements in ray petals. Auxin concentration of ‘MADG’ and ‘ADG’ at S10 and SEB.

Click here for additional data file.

Table S1 Summary of the RNA‐Seq output and the subsequent transcriptome assembly.

Click here for additional data file.

Table S2 DEGs of different pairwise comparisons.

Click here for additional data file.

Table S3 DEGs of pairwise comparisons in Figure 6.

Click here for additional data file.

Table S4 The expression levels of DEGs for Figure 7.

Click here for additional data file.

Table S5 Homologs of downstream genes of ATHB15.

Click here for additional data file.

Table S6 Primer sequences used in the study.

Click here for additional data file.