Evolutionary diversification of CYC/TB1-like TCP homologs and their recruitment for the control of branching and floral morphology in Papaveraceae (basal eudicots)
Summary
- Angiosperms possess enormous morphological variation in plant architectures and floral forms. Previous studies in Pentapetalae and monocots have demonstrated the involvement of TCP domain CYCLOIDEA/TEOSINTE BRANCHED1-like (CYC/TB1) genes in the control of floral symmetry and shoot branching. However, how TCP/CYC-like (CYL) genes originated, evolved and functionally diversified remain unclear.
- We conducted a comparative functional study in Ranunculales, the sister lineage to all other eudicots, between Eschscholzia californica and Cysticapnos vesicaria, two species of Papaveraceae with actinomorphic and zygomorphic flowers, respectively.
- Phylogenetic analysis indicates that CYL genes in Papaveraceae form two paralogous lineages, PapaCYL1 and PapaCYL2. Papaveraceae CYL genes show highly diversified expression patterns as well as functions. Enhanced branching by silencing of EscaCYL1 suggests that the role of CYC/TB1-like genes in branching control is conserved in Papaveraceae. In contrast to the arrest of stamen development in Pentapetalae, PapaCYL genes promote stamen initiation and growth. In addition, we demonstrate that CyveCYLs are involved in perianth development, specifying sepal and petal identity in Cysticapnos by regulating the B-class floral organ identity genes. Our data also suggest the involvement of CyveCYL genes in the regulation of flower symmetry in Cysticapnos.
- Our work provides evidence of the importance of TCP/CYC-like genes in the promotion of morphological diversity across angiosperms.
Introduction
TCP domain transcription factors form an ancient, plant-specific protein family known to regulate key morphological innovations in plants, thereby contributing to the diversification and adaptation of angiosperm species (Nicolas & Cubas, 2015). The TCP genes have a complex evolutionary history involving lineage-specific duplications associated with functional diversification. They can be divided into three major subfamilies: CYCLOIDEA/TEOSINTE BRANCHED1 (CYC/TB1 or CYC), CINCINNATA (CIN) and PROLIFERATING CELL FACTOR (PCF)-like (Cubas et al., 1999a; Cubas, 2004). CYC-like genes have experienced a number of independent and lineage-specific duplications throughout angiosperms (Kölsch & Gleissberg, 2006; Damerval et al., 2007; Citerne et al., 2013; Horn et al., 2014). In Pentapetalae (all core eudicots excluding Gunnerales; Cantino et al., 2007), the CYC-like genes are divided into CYC1, CYC2 and CYC3 subclades (Howarth & Donoghue, 2006; Citerne et al., 2013; Horn et al., 2014). So far, the phylogenetic relationships amongst basal eudicot (comprising the orders Ranunculales, Proteales, Trochodendrales, Buxales and Gunnerales) CYC-like genes (called TCP/CYL genes hereafter) and their affinities with Pentapetalae homologs are both poorly understood (Citerne et al., 2013; Horn et al., 2014). Although functional studies in Pentapetalae and monocots have revealed highly conserved developmental roles for CYC/TB1-like regulators in the control of shoot branching as well as flower symmetry (Nicolas & Cubas, 2015), their roles in basal eudicots are largely unknown.
Functional studies have revealed that CYC-like genes play major roles in the regulation of organ growth, both as positive and negative regulators of cell proliferation and/or expansion. The highly branched shoot morphology of the maize tb1 mutants indicates that TB1 acts to repress the outgrowth of axillary buds (Doebley et al., 1997). The involvement of TB1 orthologs in branching control has been discovered across flowering plants, for example OsTB1 in rice (Takeda et al., 2003), SbTB1 in sorghum (Kebrom et al., 2006), TaTB1 in wheat (Kebrom et al., 2012), AtTCP18/BRANCHED1 (BRC1) in Arabidopsis (Aguilar-Martínez et al., 2007), PsBRC1 in pea (Braun et al., 2012) and SlBRC1a/b in tomato (Martín-Trillo et al., 2011). In Pentapetalae, the CYC2 genes show highly conserved function in the control of flower symmetry. The molecular basis of floral symmetry control was originally characterized in Antirrhinum majus, where two partially redundant CYC2 paralogs, CYCLOIDEA (CYC) and DICHOTOMA (DICH), define the identity of the dorsal domain of the flowers and establish bilateral symmetry (zygomorphy; monosymmetry) in petal and stamen whorls (Luo et al., 1996, 1999). Loss of CYC and DICH activity results in radially symmetrical flowers (actinomorphy; polysymmetry), in which all petals acquire ventral identity. Moreover, the mutants develop six petals instead of five, and two fertile stamens instead of a dorsal staminode. A similar gene function has also been detected in the closely related Linaria vulgaris (Cubas et al., 1999b). CYC2 genes show broader expression domains in Fabales, in which duplicated CYC2 genes regulate the identity of both dorsal and lateral petals (Feng et al., 2006; Wang et al., 2008). In addition, a ventral expression domain was identified in Asteraceae, in connection with the development of zygomorphic ray flowers in a capitulum inflorescence (Broholm et al., 2008; Juntheikki-Palovaara et al., 2014). CYC-like genes also regulate zygomorphy in monocots. The RETARDED PALEA1 (REP1) gene in rice (Poaceae) is asymmetrically expressed, and specifies the development of the dorsal palea which, together with the ventral lemma, protects the florets (Yuan et al., 2009). In the petaloid monocot order, Zingiberales, shifts in CYC-like gene expression patterns have been shown to be associated with complex floral symmetry patterns, indicating that these genes have repeatedly been recruited to generate zygomorphy (Bartlett & Specht, 2011).
Ranunculales, the sister group to all other eudicots, has been established as a new model order to understand flower morphological diversification (Damerval & Becker, 2017). Its phylogenetic position in angiosperms makes it an ideal target for studies aiming at the elucidation of the origin, evolution and diversification of gene functions in flowers (Becker, 2016; Chanderbali et al., 2016). Ranunculales include several model species that are susceptible to functional genetics by virus-induced gene silencing (VIGS). Among these are the Papaveraceae representatives Papaver somniferum (Hileman et al., 2005), Eschscholzia californica (Wege et al., 2007) and Cysticapnos vesicaria (Hidalgo et al., 2012), and, in Ranunculaceae, Aquilegia coerulea (Gould & Kramer, 2007), Thalictrum ssp. (Di Stilio et al., 2010) and Nigella damascena (Gonçalves et al., 2013). We selected E. californica and C. vesicaria, representing two major Papaveraceae subfamilies, for comparative functional studies. Cysticapnos vesicaria, characterized by the presence of derived zygomorphic (monosymmetric) flowers, represents Fumarioideae (formerly Fumariaceae Marquis, fumitory family), whereas Eschscholzia californica, with actinomorphic (polysymmetric) flowers, represents Papaveroideae (formerly Papaveraceae sensu stricto, poppies) (Hidalgo & Gleissberg, 2010). Zygomorphy in Fumarioideae is peculiar as it is developed along the transverse plane, although the resupination of the pedicel presents the flower in a vertical orientation at anthesis (Damerval et al., 2013, and references therein). By the application of VIGS, we discovered diversified TCP/CYL gene functions affecting both vegetative and reproductive development. In E. californica, TCP/CYL genes showed conserved functions affecting shoot branching, but also petal and stamen development. Furthermore, we showed that, in C. vesicaria, TCP/CYL genes are involved in the specification of sepal and petal identity through the regulation of the B-class MADS-box gene expression domain. We also discovered modifications in floral symmetry, albeit at low frequency.
Materials and Methods
Plant material and growth conditions
The seeds of Eschscholzia californica Cham. ‘Aurantiaca Orange King’ (B&T World Seeds, Aigue Vive, France) were stratified in water for 3 d at 4°C and sown on soil (50% peat, 50% vermiculite). Plants were grown in controlled growth rooms, without natural light, at 23°C under long-day (16 h) conditions with a photosynthetic photon flux density of 70 μmol m−2 s−1 (fluorescent tubes, Philips, warm white). The seeds of Cysticapnos vesicaria (L.) Fedde (HB Göttingen, Germany, Index Seminum 981/2007 (BC)) were sown in pots, and stratified at 4°C for 4 d. Plants were grown at 22°C under continuous light at 60–100 μmol m−2 s−1. For fertilization, NPK fertilizer (Kekkilä Kastelulannoite, Kekkilä, Vantaa, Finland) was used.
Gene identification and cloning
Two partial sequences of E. californica, EscaCYL1 (DQ347820) and EscaCYL2 (DQ347824), were published in Kölsch & Gleissberg (2006). The corresponding full-length coding sequences were retrieved from the oneKP database (Matasci et al., 2014) and from RNAseq data generated in the Becker laboratory. Three partial sequences of C. vesicaria, CyveCYL1A (DQ659312), CyveCYL2A (DQ659313) and CyveCYL2B (DQ659314), were published in Damerval et al. (2007). CyveCYL1B was isolated by half-nested PCR with a selective forward primer CyveCYL1 (Damerval et al., 2007) and two reverse primers AB05 and CyveCYL1&2 (Supporting Information Table S1). 3′ coding sequences of the four CyveCYLs were amplified using a SMARTer RACE 5′/3′ Kit (Takara Bio USA Inc., Mountain View, CA, USA) and cloned into pJET vector (CloneJET PCR Cloning Kit; Thermo Fisher Scientific, Vilnius, Lithuana). The B-class MADS-box genes of C. vesicaria were identified from oneKP, including CyvePI (UDHA-2013443), CyveAP3-1 (UDHA-2096224) and CyveAP3-2 (UDHA-2015403). The CYL and B-class genes of Capnoides sempervirens were identified as described in Methods S1. All primer sequences are given in Table S1. Accession numbers for CyveCYLs are MG893934–MG893937. A list of species as well as accession numbers for genes used for phylogenetic analysis (Methods S2) are given in Table S2.
Sampling and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analysis
For gene expression analyses in wild-type (WT) and VIGS-treated plants, a series of floral organs and plant tissues was collected. The samples of WT E. californica included leaves, leaf nodes, sepals, petals, stamens and carpels at anthesis, as well as floral buds (FBs) at different developmental stages (0–1, 1–2, 2–3 and > 3 mm in diameter). The FBs of WT C. vesicaria corresponded to sizes of < 2, 2–3, 3–5, 5–8 and 8–10 mm in diameter. Other samples were similar to E. californica except that dorsal (spurred outer petal, secondarily dorsal after resupination), ventral (nonspurred outer petal, secondarily ventral after resupination) and lateral petals were sampled separately. The floral organ samples were mixtures of late developmental stages. The leaves and leaf nodes (including axillary buds) from the axils of leaves were collected during the vegetative stage. To verify the efficiency of gene silencing, the first emerging FB was collected from E. californica and young leaves from C. vesicaria VIGS plants. The appropriate developmental stage for leaves was estimated by comparison with positive control plants (pTRV2-CyvePDS) showing phytoene desaturase (PDS) gene silencing and photobleaching.
Total RNA from WT E. californica was extracted with a NucleoSpin® RNA Plant kit (Macherey-Nagel, Düren, Germany) in biological triplicates. cDNA synthesis using 1 μg of RNA template was performed according to a RevertAid H Minus First Strand cDNA Synthesis Kit (Fermentas, St Leon-Rot, Germany). From all other samples, total RNA was isolated using a hexadecyltrimethylammonium bromide (CTAB) method (Chang et al., 1993), and DNase treatment and RNA purification with an RNA clean-up kit (Macherey-Nagel). Reverse transcription of 1 μg of total RNA was conducted with a SuperScript III First-Strand Synthesis System (Invitrogen). qRT-PCR was conducted according to the manufacturer's instructions using a LightCycler 480 instrument (Roche Diagnostics Ltd, Rotkreutz, Switzerland) and monitored with SYBR-green I dye (Roche Diagnostics GmbH, Mannheim, Germany). The housekeeping genes EscaGAPDH and CyveGAPDH were used as internal quantitative controls. qRT-PCR was conducted using two to three biological replicates, each including three technical replicates. The relative expression values were calculated using the ΔCt method for WT samples and the ΔΔCt method for VIGS samples (Pfaffl, 2001). The qRT-PCR primers (Table S1) were tested for their efficiency (between 1.8 and 2).
In situ hybridization
Gene-specific probes were synthesized using a PCR-amplified fragment of the given gene with primers containing a few extra nucleotides (uppercase letters) and a T7 overhang (lowercase letters) (CAtaatacgactcactataGGG) at the 5′-end (Table S1), and labeled using a DIG RNA Labeling Kit (Roche). Sample fixation, sectioning and hybridization were performed as described in Orashakova et al. (2009) for E. californica and in Hidalgo et al. (2012) for C. vesicaria. Signals were recorded using a Leitz Laborlux S microscope equipped with a Leica DFC420 C digital camera (Wetzlar, Germany).
VIGS and scoring of the phenotypes
The tobacco rattle virus (TRV)-based system was applied for VIGS (Liu et al., 2002). For E. californica, we made single as well as double VIGS constructs for the two genes, EscaCYL1 and EscaCYL2. For single gene constructs, the fragments between the TCP and R domains (Fig. S1a) were cloned into a pTRV2 vector between the EcoRI and BamHI restriction sites. To target both genes, the EscaCYL2 fragment flanked by BamHI and SacI sites was cloned into the pTRV2-EscaCYL1 construct. The Agrobacterium tumefaciens strain pGV3101 was transformed with each VIGS construct and an empty vector control, respectively. The VIGS inoculation was conducted as described in Wege et al. (2007) using seedlings at the three- to five-rosette leaf stage and application of syringe infiltration to the hypocotyls. Depending on the construct, 20–49 plants were treated (Table S3; see later Fig. 3). The numbers of activated/elongated axillary meristems, leaf number and stem length were recorded when the first FB (2–4 mm) was collected for expression analysis. The reproductive phenotypes were observed from the second to the fourth flower. Petal width and length, as well as the stamen number and length (from four random stamens of the outermost whorls), were determined from fully open flowers. The measurement of cell sizes is described in Methods S3.
For Cysticapnos single VIGS constructs, each of the four CyveCYL gene fragments including partial TCP and R domains (Fig. S1b), were amplified using gene-specific primers flanked by BamHI and XhoI restriction sites (Table S1) and cloned into the pTRV2 vector. We also made double VIGS constructs for silencing of the paralogous pairs. For the pTRV2-CyveCYL1 construct, we combined a XbaI-tailed CyveCYL1B fragment into linearized pTRV2-CyveCYL1A. For the pTRV2-CyveCYL2 construct, a CyveCYL2B fragment with BamHI tails was amplified and cloned into the linearized pTRV2-CyveCYL2A construct. The Agrobacterium tumefaciens strain pGV3101 was used for transformation. Cysticapnos seedlings with three to five leaves were inoculated by syringe infiltration as described in Hidalgo et al. (2012). Ten to 17 plants were treated and, for each plant, 30 flowers were phenotypically analyzed in double VIGS treatments (Table S4).
Statistical analysis
All statistical analyses were performed in Sas (release 9.3; SAS Institute, Cary, NC, USA). The normality of the variables was evaluated with a UNIVARIATE procedure based on the residuals obtained from a model including group as the fixed effect in the MIXED procedure. Analyses of variance for the variables that passed the normality tests were performed with the MIXED procedure with group as the fixed effect. Analyses of variance for the variables that failed in the normality tests were performed in nonparametric tests using RANK and GLM procedures with group as the fixed effect.
Scanning electron microscopy (SEM)
For SEM, FBs were hand dissected under a stereomicroscope. Sample preparation, fixation (FAA; 50% ethanol, 5% formaldehyde, 10% acetic acid) and dehydration steps were performed as described in Uimari et al. (2004). An automated Leica EM CP300 dryer (Leica Mikrosysteme GmbH, Wien, Austria) was used for critical point drying. Samples were coated with platinum using Quorum Q150TS (Quorum Technologies, Laughton, UK) and examined using a Quanta 250 (FEI Corp., Hillsboro, OR, USA) scanning electron microscope at the Electron Microscopy Unit, University of Helsinki, Finland.
Results
Identification and expression analysis of TCP/CYL genes in Papaveraceae
Altogether, two TCP/CYL genes of E. californica (Kölsch & Gleissberg, 2006) and four genes of C. vesicaria (Damerval et al., 2007; this work) were included in this study. We conducted phylogenetic analysis with selected Papaveraceae as well as other Ranunculales and Proteales sequences (Table S2; Fig. S2). The analysis confirmed the previous results by Damerval et al. (2007), showing that the TCP/CYL genes of Papaveraceae grouped into two paralogous lineages (PapaCYL1 and PapaCYL2) that resulted from a duplication predating the divergence of Fumarioideae and Papaveroideae. The newly identified C. vesicaria CyveCYL1B gene turned out to be the paralog of CyveCYL1A within the PapaCYL1 genes, whereas the previously identified paralogs CyveCYL2A and CyveCYL2B grouped within the PapaCYL2 genes. Expression data for several Papaveraceae TCP/CYL genes (including the data produced here for C. vesicaria and E. californica) were plotted onto the phylogeny (Fig. S2). Generally, the Fumarioideae CYL1 and CYL2 genes are expressed in both leaves and flowers (with the exception of Lamprocapnos LaspCYL1), whereas the expression of Papaveroideae CYL1 and CYL2 genes is restricted to flowers.
qRT-PCR was used for the expression analysis of PapaCYL1 and PapaCYL2 genes in E. californica and C. vesicaria. The analysis was carried out on a series of vegetative and floral tissues at fully differentiated stages, as well as in FBs at different developmental stages (Fig. 1). In E. californica, the expression of both EscaCYL1 and EscaCYL2 was concentrated in FBs (Fig. 1a,b). EscaCYL1 was most highly expressed at the early stages of flower bud development (≤ 3 mm), whereas EscaCYL2 showed higher expression during the later stages (≥ 2 mm). Both genes showed low expression levels in leaves, in nodes of rosette leaves as well as in floral organs at anthesis. In contrast with EscaCYL1, EscaCYL2 transcripts accumulated in the stamens at anthesis.

The four C. vesicaria CyveCYL genes showed considerable variation in both their expression levels and tissue specificity (Fig. 1c,d). Of the paralogous pairs, CyveCYL1B and CyveCYL2B showed higher expression levels than their paralogs in most of the tissues. In contrast with EscaCYLs, CyveCYL2B accumulated abundantly in leaves and leaf nodes. All four CyveCYLs were expressed in sepals. Moreover, among the distinct petal types, especially CyveCYL1B and CyveCYL2B showed higher expression in ventral than in dorsal and lateral petals (Fig. 1c,d). In particular, CyveCYL2B expression was almost four-fold higher in ventral relative to dorsal petals.
In situ hybridization was applied to localize the expression domains of these genes at early developmental stages (Fig. 2). During the vegetative stages of E. californica, EscaCYL1 expression was localized to the axillary meristems of rosette leaves, whereas they were devoid of EscaCYL2 expression (Fig. 2a,b). During the reproductive stages, EscaCYL1 was expressed uniformly and widely throughout FB development, from the naked floral meristem (FM) to the buds with differentiating floral organs (Fig. 2a, FB1–3). EscaCYL2, instead, showed highly restricted expression domains localized in floral organ boundaries (Fig. 2b, FB1–3). In C. vesicaria, we detected strong CyveCYL gene expression signals in young leaves and axillary meristems. During the reproductive stages, all CyveCYL genes presented highly overlapping expression domains (Fig. 2c–f, FB1–2). The expression was localized to FMs, but not detected in the inflorescence meristem (FB1). Later, the expression extended to all flower organ primordia (FB2). When the floral organ primordia were further developed, PapaCYL2 (CyveCYL2A and CyveCYL2B) gene expression was restricted to organ boundaries (Fig. 2e,f, FB3). When zygomorphy of the flowers was established, the expression of, in particular, CyveCYL1B and CyveCYL2B was detected in the basal parts of both the dorsal and ventral petals, stamen filaments and the nectary (Fig. 2c–f, FB4).

EscaCYL1 controls axillary shoot branching
Virus-induced gene silencing was used to silence EscaCYL1 and EscaCYL2 individually, and jointly using the double constructs. Specific downregulation of target genes was confirmed (Fig. S3a,b). Silencing of EscaCYL1 alone led to enhanced shoot branching (Fig. 3a,d). On average, 66.5% of the axillary meristems were activated and elongated in the EscaCYL1 VIGS lines, whereas, in the control lines (treated with the empty vector), 44.9% of the meristems developed into shoots (Table S3). The EscaCYL2-silenced lines were similar to the control plants, and no statistically significant differences in the number of activated axillary meristems were detected (Fig. 3a,d). However, the branching phenotype was enhanced in the double VIGS plants and, on average, 95.5% of the axillary meristems developed shoots (Fig. 3d; Table S3). Inspection of pTRV2-EscaCYL1 and the double VIGS plants during late reproductive stages indicated that, in addition to the branching from the cauline leaf axils at the early reproductive stages, branching from the basal rosette leaf axils was enhanced (Fig. 3b,c,e). The control lines only developed branches from cauline leaf axillary meristems (Fig. 3b,c). In summary, our data suggest that EscaCYL1 controls shoot branching by arresting bud development at rosette leaf positions. However, it is also possible that EscaCYL2 may act redundantly, and enhance the effect. None of the VIGS treatments resulted in significant differences in flowering time, total number of leaves or internode length compared with the control plants (Fig. S4a–c).

EscaCYL1 and EscaCYL2 are involved in the regulation of petal size and stamen number
As EscaCYL1 and EscaCYL2 were expressed in FBs, we followed the reproductive phenotypes in VIGS plants (Fig. 4). We measured the width and length of the outer and inner petals of the first three, fully opened flowers (Fig. 4b). In the single VIGS plants, the width of petals was significantly reduced, but the length was not affected. In the double VIGS plants, both the width and length of the petals were significantly reduced. Similar changes were observed in both inner and outer petals. Altogether, the petals were slightly smaller in VIGS lines, indicating that both EscaCYL1 and EscaCYL2 play a minor role in the regulation of petal size. To study whether the reduced petal size was the result of changes in cell division or cell size, we measured the epidermal cells using SEM images of the outer petals. The length and width of the cells were not significantly altered in double VIGS lines compared with the control (Fig. S4d,e), indicating that the reduced petal size is likely to result from reduced cell number owing to suppressed cell proliferation.

In addition to the reduced petal size, EscaCYL2-silenced single VIGS lines resulted in decreased stamen number (Fig. 4c). A similar phenotype was found in the double VIGS flowers, whereas silencing of EscaCYL1 alone did not have an effect. Furthermore, significant reduction in stamen length was only scored in double VIGS plants (Fig. 4d).
Silencing of CyveCYLs causes homeotic changes of sepals into petals
The functional roles of the CyveCYL genes in C. vesicaria were studied using single and double VIGS constructs for the paralogous genes (pTRV2-CyveCYL1, pTRV2-CyveCYL2, Fig. 5). We were unable to unequivocally link the observed phenotypes to the gene-specific silencing of any of the four CYL genes because of the high sequence similarity between the VIGS fragments and the target genes (Fig. S1b). The nucleotide sequence similarity is 80–81% within the paralogs, and varies between 50% and 56% between the paralogous pairs of PapaCYL1 and PapaCYL2 lineages. Moreover, in the double VIGS lines, the expression of both CyveCYL1B and CyveCYL2B was downregulated. Although the sequence similarity between the VIGS fragments used and the target genes is only 55%, and they do not share any exact 21-bp matches (Figs S1b, S3c, 5d,e). Gene silencing did not affect vegetative growth, leaf morphology or branching in any of the treatments (data not shown). However, during the reproductive stage, we observed diverse floral phenotypes (Figs 5, 6). The most prominent phenotypes showed changes in perianth development. Moreover, although with extremely low frequency (Table S4), we observed flowers with altered floral symmetry in association with nectary initiation and spur formation, as well as a reduced number of stamens.


The changes in perianth development were discovered in c. 16% of flowers in double VIGS lines (Table S4). In control flowers, the perianth is formed of two scale-like sepals and two whorls of petals. The petals follow an alternate arrangement in which dorsal and ventral petals correspond to the outer whorl and two lateral petals to the inner whorl (Hidalgo et al., 2012; see later Figs S6a–c, 6a). In the mild VIGS phenotypes, the sepals were elongated, but their shape remained narrow (Fig. 5b, PSe1). The sepals of a single flower were not always equally affected, and showed variable phenotypes (Fig. 6b). In the more pronounced phenotypes, the modified sepals were morphologically similar to the WT dorsal petals that developed a spur (Fig. 5b, PSe2). In these cases, other flower organs did not show any phenotypic changes.
To confirm the homeotic change of sepals into dorsal petals, we analyzed the cellular structures of the corresponding organs in WT and VIGS lines (Fig. 5a–c). In WT, the cellular structures in dorsal and ventral petals, as well as in sepals, are highly similar (Fig. 5a). The central vein region is occupied by rectangle-shaped cells (green asterisks) that are absent from the tips of the petals. The tips are occupied by puzzle-shaped cells (yellow asterisks) that are also detected in sepals, but aside of the central vein. The mid regions of petals form irregular round-shaped cells (red asterisks) with stomata next to the central vein, whereas the mid regions of the sepals still occupy puzzle-shaped cells. At the basal position, the spur is characterized by conical cells (orange asterisks) next to the central vein. These specific conical cells represent the only criteria to distinguish the dorsal and ventral petals. In the CyveCYLs VIGS lines (Fig. 5c), the cell types in the tip regions of both the elongated sepals (PSe1) and petaloid sepals (PSe2) are identical to those of the dorsal and ventral petals. The middle and basal regions of the elongated sepals (PSe1) are similar to WT sepals, but the changes in petaloid sepals (PSe2) are more pronounced, showing round-shaped cells in the middle region and conical cells marking a spur at the base. Our data thus suggest that the modified sepals in the VIGS lines are morphologically similar to dorsal petals.
We also verified the expression patterns of three B-class MADS-box genes, CyvePI, CyveAP3-1 and CyveAP3-2, in WT Cysticapnos, and showed that CyvePI and CyveAP3-1 expression is specifically localized in petals, whereas they show low expression in sepals (Fig. S5a). In VIGS lines, the expression levels of these genes were upregulated in petaloid sepals (PSe) to similar levels as observed in dorsal petals (Fig. 5f, DP). The changes in B-gene expression were associated with the downregulation of CyveCYL1B or CyveCYL2B in the petaloid sepal (PSe) samples of the double VIGS lines (Fig. 5d,e). Together with the cellular structure analysis, our data indicate that silencing of CyveCYLs causes a homeotic change of sepals into dorsal petals. We thereby suggest that CyveCYL genes in Cysticapnos are involved in the regulation of perianth development and, especially, sepal identity by the suppression of B-gene activity.
In order to test whether the observed association between B and CYL genes holds true for other species in Fumarioideae, we conducted expression analysis in zygomorphic Capnoides sempervirens (Fig. S5b,c). Our data indicate that the expression of B and CYL genes in C. sempervirens follows the same pattern as in C. vesicaria, that is, B genes are highly expressed in petals and CYL gene expression localizes in sepals. However, in contrast with the dominant expression of CyveCYL1B and CyveCYL2B in ventral petals, CaseCYL genes are weakly expressed in all petal types (Fig. S5c). Nevertheless, sepal development of C. sempervirens is likely to be regulated by CYL and B genes.
CyveCYL genes may affect floral symmetry
In addition to the modified sepal phenotypes, we observed a low frequency of disymmetric and radially symmetric flowers (Table S4). In the WT flowers of Cysticapnos, six stamens are arranged in two bundles (Fig. S6g), and the nectary (Ne) develops at the base of one of the bundles in the dorsal domain of the flower (Figs S6d, 6a). Moreover, the base of the dorsal petal develops a spur (Fig. 6a, Sp). In the modified flower phenotypes, the transition from zygomorphy to disymmetry is associated with the development of an extra nectary at the base of the other stamen filament (Fig. S6e), as well as with the subsequent conversion of the ventral petal into a dorsal-like petal with a spur (Fig. 6b). In the radially symmetric flowers, the stamen number is reduced from six to four (Fig. S6h), and four nectaries are formed in a symmetric manner at the bases of the staminal bundles (Figs 6c, S6f). Moreover, the flowers do not develop any sepals and all four petals gain dorsal identity with visible spurs (Fig. 6c).
Discussion
As a result of their phylogenetic position as successive sister lineages to the rest of the eudicots, basal eudicots are useful models to complement our understanding of genetic networks involved in floral developmental diversification which, to date, have been most extensively studied in monocots and Pentapetalae (Soltis et al., 2002; Chanderbali et al., 2016). We conducted the first studies of TCP/CYL gene functions through comparative analysis of Eschscholzia californica and Cysticapnos vesicaria from the Papaveraceae. These species belong to Ranunculales, the sister order to all other eudicots, with wide diversity in developmental traits, including shoot architecture, leaf shape and, particularly, flower forms. Our data indicate that TCP/CYL gene functions affecting shoot branching as well as cellular growth are highly conserved across angiosperms. Moreover, we show species-specific diversification of TCP/CYL gene functions in the regulation of floral traits. Our data call for further comparative studies with a broader selection of model species, both within and outside basal eudicots, to discover the ancestral roles of these key developmental regulators.
Conserved roles of TCP/CYL genes in branching control among angiosperms
Shoot branching control involving CYC/TB1 homologs is highly conserved in both monocots and Pentapetalae (Aguilar-Martínez et al., 2007; Finlayson, 2007), and our functional data in E. californica indicate that this conservation extends to Ranunculales. The role of EscaCYL1 in the repression of rosette branching is associated with its spatial expression in the axillary meristems of rosette leaves (Fig. 2a). As their counterparts in eudicots and monocots, TCP/CYL genes were not expressed in the shoot apical meristem. The phenotype was enhanced in the double VIGS lines, indicating that EscaCYL2 may also contribute to branching control, redundantly with EscaCYL1 (Fig. 3). In C. vesicaria, all four CyveCYL genes were expressed in leaf axillary buds (Fig. 2c–f). However, in contrast with E. californica, the axillary buds in C. vesicaria develop into branches, even in the wild-type, and the VIGS lines did not show any differences in this respect (data not shown). It is plausible that branching is regulated by other mechanisms in C. vesicaria, or involves gene functions downstream of TCP/CYL.
The TB1 homologs repress axillary branching, especially in the basal positions of the plant, suggesting a similar control mechanism along the shoot axis. The loss-of-function mutant brc1 in Arabidopsis shows increased rosette leaf rather than cauline leaf branching (Aguilar-Martínez et al., 2007). Pea PsBRC1 affects branching in cotyledonary axils, as well as in the first and second nodes (Braun et al., 2012), and tomato SlBRC1b functions at lower node positions (Martín-Trillo et al., 2011). A similar effect is observed in tb1 in maize and Ostb1 in rice (Doebley et al., 1997; Takeda et al., 2003). In both eudicots and monocots, BRC1 and TB1 have been shown to mediate strigolactone activity, which is required to repress branching; however, the complex hormonal, environmental and nutritional signaling integrated by TB1/BRC1 genes is not yet fully understood (Rameau et al., 2015).
Parallel recruitment of PapaCYL genes in petal development
PapaCYL genes affected petal development in both E. californica and C. vesicaria, but in a species-specific manner. In the actinomorphic E. californica, silencing of either EscaCYL1 or EscaCYL2 resulted in minor modifications in both inner and outer petal size as a result of suppression of cell proliferation (Figs 4, S4d,e). We anticipate that the reduced size is not associated with a possible trade-off effect of enhanced branching as the EscaCYL2 VIGS lines did not show any alterations in branching. Rather, the observed change is consistent with the widely recognized functions of TCP transcription factors in the promotion and/or repression of tissue growth, often depending on the developmental context (Martín-Trillo & Cubas, 2009). For example, the Antirrhinum majus (Lamiales) CYC acts as both a repressor and an activator of growth. At the early developmental stage, CYC establishes zygomorphy by reducing the growth of the dorsal domain of the flower, whereas, later, it increases the growth rate of the dorsal petals (Luo et al., 1996).
In transverse zygomorphic flowers of C. vesicaria, CyveCYL1B and CyveCYL2B showed stronger expression in the nonspurred outer (and secondarily ventral) petals during the late developmental stages (Fig. 1c,d). In Pentapetalae, ventral expression of CYC-like genes has, so far, only been found in Asteraceae (Broholm et al., 2008; Juntheikki-Palovaara et al., 2014), but is commonly observed in the monocot Zingiberales CsTB1a (Bartlett & Specht, 2011), Commelinaceae TB1a (Preston & Hileman, 2012), magnoliids Aristolochia CYCL1/2 (Horn et al., 2014) and basal eudicots Proteaceae ProtCYC (Citerne et al., 2017). Phylogenetic analyses indicate that CYC2 orthologs do not exist outside Pentapetalae, and that the dorsal-specific expression has evolved at the base of the CYC2 lineage, suggesting a different genetic mechanism underlying bilateral symmetry in noncore eudicots and monocots (Preston & Hileman, 2009). In Pentapetalae, the differential expression domains in the dorsal–ventral axis of the flowers are established early in development. However, in both E. californica and C. vesicaria, the expression of CYL genes was uniform throughout the early developing FBs (Fig. 2). In C. vesicaria, uniform CYL gene expression was observed even at stage FB4, when morphological differentiation of the spur and distinct petal types are already visible (Fig. 2), and the asymmetric expression pattern only appeared during the later stages of organ development (Fig. 1c,d). A similar late asymmetric pattern was also detected in Aristolochia (Magnoliidae; Horn et al., 2014) and Capnoides sempervirens (Fumarioideae; Damerval et al., 2013). We thereby propose that CYL genes are not involved in the initiation of zygomorphy, but are likely to regulate late organ growth in response to as yet unknown factors that establish asymmetry.
Evolution of zygomorphy in Fumarioideae
Zygomorphic flowers are abundant in Pentapetalae or monocots, but almost absent from basal angiosperms, and they are rare among basal eudicots (Reyes et al., 2016). In Pentapetalae, duplications within the CYC2 clade are associated with the evolution of zygomorphy (Hileman, 2014). Three independent transitions from polysymmetry to monosymmetry have occurred in Ranunculales: one each in the families Ranunculaceae, Papaveraceae and Menispermaceae (Damerval & Nadot, 2007). In Delphinieae, characterized by zygomorphic flowers, there is an additional duplication of CYL genes in each of two Ranunculaceae paralog lineages (Jabbour et al., 2014). Jabbour et al. (2014) suggested that these additional duplications coincide with the evolution of zygomorphy; however, the origin of these duplications, or alternatively gene losses, in the ancestors of other tribes remains unclear. In Fumarioideae, there is no evidence of a CYL duplication within Papaveraceae paralog lineages associated with the emergence of zygomorphy in the subfamily, but instead a species-specific additional duplication in our zygomorphic model C. vesicaria.
In Papaveraceae, Fumarioideae shows a transition to monosymmetry that involves an intermediate disymmetric state (Damerval & Nadot, 2007; Sauquet et al., 2015). In disymmetric flowers (e.g. in Lamprocapnos), spurs are formed in both outer petals in association with nectaries at the bases of two stamen bundles (Damerval & Nadot, 2007; Hidalgo & Gleissberg, 2010). In C. vesicaria VIGS lines, we observed both disymmetric and actinomorphic flowers, suggesting that CYL genes may be involved in the proposed scenario of symmetry evolution in Papaveraceae (Damerval & Nadot, 2007). However, the phenotypes occurred with extremely low frequency, and they were also observed in a few wild-type plants (although not in empty vector controls) (Fig. 6; Table S3). Therefore, we cannot rule out the possibility that environmental conditions affect the fluctuation in the phenotypes, as found in Corydalis cheilanthifolia (Tebbitt et al., 2008). The low frequency of the phenotypes may also result from only partial silencing of CyveCYL genes in ventral petals (Fig. 5d,e), or from reduced VIGS efficiency during reproductive stages. This was, however, not true with regard to the frequency of other phenotypes (e.g. presence of petaloid sepals).
Diversified functions of CYC-like genes in stamen development across angiosperms
With the exceptions of Veronica and Gratiola (Plantaginaceae; Preston et al., 2009), CYC activity in Pentapetalae and monocots has been associated with the arrest of stamen development. This occurs in a species-specific manner in the dorsal (Antirrhinum; Luo et al., 1996, 1999), lateral (Mohavea; Hileman et al., 2003) or ventral (Opithandra; Song et al., 2009) domain of the flower. By contrast, our data suggest that TCP/CYL genes promote stamen development in both Papaveroideae and Fumarioideae. In E. californica, silencing of EscaCYL2 decreased stamen number and, in C. vesicaria, stamen number was also reduced in association with the shift to actinomorphy. Stamen expression of CYL paralogs has also been shown, for example, in Nigella damascena, a representative of the Ranunculaceae family (Jabbour et al., 2014). In C. vesicaria, the loss of lateral stamens within the bundles was in concordance with Ronse De Craene & Smets (1992), who stated that the four stamens of the outer whorl are always the last to be affected by reductions in the androecium.
We hypothesize that EscaCYL2 may control stamen number by affecting meristem growth. Eschscholzia develops a ring-like androecial meristem around the gynoecium that remains active even when the central FM ceases its activity (Becker et al., 2005). Previous studies have indicated that mild suppression of E. californica AGAMOUS homologs prolongs floral meristematic activity, leading to increased stamen number (Yellina et al., 2010). However, silencing of SHOOTMERISTEMLESS (STM) orthologs leads to reduced stamen number as well as reduction of the gynoecium (Stammler et al., 2013). It has been proposed that the two STM orthologs in E. californica have evolved specific roles in maintaining the central and ring meristem, respectively. In addition, the initial meristem size was positively correlated with the total organ number per flower in Nigella damascena (Wang et al., 2015). Nigella, however, shows considerable variation in the number of all four types of floral organs. The possible involvement of TCP/CYL proteins in these regulatory networks is still unclear. Nevertheless, it remains to be studied whether the emergence of polyandry (numerous stamens, > 10) across Ranunculales (Damerval & Nadot, 2007) involves the parallel recruitment of TCP/CYL genes.
Perianth evolution in Fumarioideae
A bipartite perianth, that is a perianth consisting of distinct sepals and petals, has evolved several times in angiosperms, resulting in diverse forms of perianth (Ronse De Craene, 2007; Hileman & Irish, 2009). The perianth in early angiosperms, as well as in many Ranunculales genera, is undifferentiated, consisting of tepals (Damerval & Nadot, 2007; Ronse De Craene, 2007). However, in Papaveroideae, the perianth presents distinct sepals and petals, whereas, in Fumarioideae, sepals with petaloid characters are found (Hidalgo & Gleissberg, 2010). Previous studies in two Fumarioideae species, Lamprocapnos spectabilis and Capnoides sempervirens, have shown strong PapaCYL expression in sepals (Damerval et al., 2013; this work Fig. S5c), and we observed a similar pattern in C. vesicaria (Fig. 1c,d). Our functional studies indicated that silencing of PapaCYL genes results in homeotic conversion of sepals into petals in association with the upregulation of B-class MADS-box genes (Fig. 5).
The evolutionary history of the perianth has been debated for a long time. It has been proposed that the original perianth of angiosperms was either sepaloid or petaloid, and involved organ identity transitions in either direction to establish a bipartite identity (Albert et al., 1998; Ronse De Craene, 2007). A recent model-based reconstruction of an ancestral flower of all living angiosperms indicates that the perianth consisted of > 10 undifferentiated tepals arranged in more than two whorls (Sauquet et al., 2017). Albert et al. (1998) suggested that the exclusion of B genes from the outer perianth whorl in early eudicots may be responsible for the sepal–petal distinction. Our data provide evidence that, in C. vesicaria, TCP/CYL expression is necessary to exclude B-gene expression from the outer sepal whorl, identifying the C. vesicaria TCP/CYL genes as cadastral genes, defining the regional specificity in the floral context. Moreover, CYL genes apparently regulate the size of the sepals in C. vesicaria. In the monocot Commelina communis, TB1-like genes establish the early bilateral symmetry and, in addition, DEF-like gene expression has been shown to regulate the differentiation of inner tepals (Preston & Hileman, 2012). Combinatorial interaction of diverse AP3/AGL6 homologs also specifies organ identity in the orchid perianth (Mondragón-Palomino & Theißen, 2008, 2009; Hsu et al., 2015). Furthermore, an interplay between B genes and CYC-like genes has been observed in Pentapetalae (Clark & Coen, 2002). These studies further support the parallel evolution of floral bilateral symmetry in angiosperms, but also the involvement of B-class genes in late developmental shifts within single organ whorls (Preston & Hileman, 2012). Whether B-gene activity is directly regulated by specific TCP/CYL proteins remains to be studied.
Acknowledgements
Evgenia Diel MSc is thanked for her help with VIGS and Andrea Weisert is thanked for carrying out qRT-PCR. Anu Rokkanen is thanked for technical assistance in cloning and expression analyses. O.H. is grateful to Stefan Gleissberg for his advice and support during the preliminary stage of this research carried out at Ohio University. We thank the Electron Microscopy Unit, Institute of Biotechnology, University of Helsinki, Finland for providing excellent facilities. This work was supported by the Academy of Finland grants 1287401 and 1310317 (P.E.), DAAD 57162231 and 57347335 (A.B.), Doctoral Programme in Plant Sciences, and Jenny and Antti Wihuri Foundation (Y.Z.).
Author contributions
A.B., O.H. and P.E. designed the research. Y.Z., K.P., O.H. and A.B.D. performed the research, and collected and analyzed the data. Y.Z., K.P., A.B. and P.E. interpreted the data. Y.Z., K.P., A.B. and P.E. wrote the manuscript with comments from all other authors.