Molecular basis of floral petaloidy: insights from androecia of Canna indica

Floral organs usually take on the characteristics of petals in ginger plants. In Canna indica, the most ornamental parts of the flowers are considered to be staminode, which means rudimentary and sterile stamen. However, the precise nature of these petaloid organs is yet to be determined. Two floral organ identity genes GLOBOSA (GLO) and AGAMOUS (AG) are isolated from Canna indica. Their expression patterns suggest that petaloid staminodes and labellum are of androecium identity, in agreement with their position within the flower. But the current molecular, morphological and anatomical data, are still not sufficient to explain the distinct morphology observed in staminodes and fertile stamen in Canna indica.


Introduction
A landmark accomplishment in plant developmental biology was the proposition of the ABC model of flower organ identity. This model provides a framework for describing a conserved pattern of gene expressions associated with the specification of floral organs in model species (Bowman et al. 1991;Coen and Meyerowitz 1991). Soon after its proposition, other floral identity genes including D-and E-class genes were identified and the ABC model was then extended to an ABCDE model (Angenent et al. 1995;Pelaz et al. 2000;Ditta et al. 2004). ABC model genes, with the exception of APETALA2 (AP2) (Jofuku et al. 1994), belong to the MADS-box family of transcription factors. In angiosperms, MADS-box transcription factors are important components of various developmental processes, although the most well-characterized MADS-box genes to date are involved in floral organ identity and initiation (Theißen and Saedler 2001).
The ABC model has been extensively characterized due to molecular studies in two model plants, Arabidopsis thaliana (Yanofsky et al. 1990;Jack et al. 1992;Mandel et al. 1992;Goto and Meyerowitz 1994;Gustafsonbrown et al. 1994) and Antirrhinum majus (Sommer et al. 1990;Huijser et al. 1992;Trobner et al. 1992;Bradley et al. 1993). Studies in other angiosperm lineages generally support the idea that the expression patterns described by the ABC model are widely conserved, at least among eudicots. However, the extent to which the eudicot-based ABC model of flower organ identity describes floral morphologies in monocots is still not well understood (Ambrose et al. 2000).
The role of A-class genes appears to be conserved only in the lineage containing Arabidopsis (Causier et al. 2010), and to date the A-function of the ABC model has not been shown to play a role in organ identity in monocots. Studies carried out in maize, rice and wheat (Mena et al. 1996;Kang et al. 1998;Ambrose et al. 2000;Munster et al. 2001;Meguro et al. 2003), however, do indicate that B-and C-class genes and gene activities are somewhat conserved between grasses and eudicots, despite evidence for gene duplication in various lineages. However, the derived morphology of grass flowers results in inconclusive debates regarding the interpretation of the lemma and palea as equivalent to sepals and petals of eudicots (Bowman 1997;Ambrose et al. 2000). Also, studies in other monocot lineages such as the Liliaceae show divergent results from those predicted by the canonical BC model. In this family, flowers have two outer whorls of almost identical petaloid organs called tepals. In order to explain such a derived pattern, a modified BC model has been proposed (Van Tunen et al. 1993) and ascertained (Kanno et al. 2003). In this modified model, B-class genes are expressed in whorl 1 as well as whorls 2 and 3, resulting in similar expression patterns in whorls 1 and 2 and, therefore, similar petaloid structure (Kanno et al. 2007). However, the expression patterns of B-class genes in asparagus, for instance, do not fit this model (Kanno et al. 2004).
Furthermore, the intricate morphology and unprecedented diversity of orchid floral morphology has resulted in a proposed 'orchid code' that builds on the BC model. This idea assumes that the identity of the different perianth organs is specified by the combinatorial interaction of four DEF-like MADS-box genes with other floral homeotic genes (Mondragón-Palomino and Theißen 2008Theißen , 2009).
Owing to the lack of consistency regarding the conservation of the BC model in monocots, it is likely that further studies on other non-grass, petaloid monocots will provide a novel understanding of the evolution of floral organ identity specification outside the eudicots.
The Zingiberales is an order of tropical monocots and consists of eight families (Tomlinson 1962;Dahlgren and Rasmussen 1983). The families are often divided into two groups: the paraphyletic 'banana group' (including Musaceae, Streliziaceae, Heliconiaceae and Lowiaeae) with five (occasionally six) fertile stamens, and the monophyletic 'ginger group' (including Zingiberaceae, Costaceae, Cannaceae and Marantaceae) with one or a half fertile stamen (Kress 1990;Rudall and Bateman 2004). Thus, an impressive reduction in the number of fertile stamens occurred in the ancestor of the ginger clade.
The Cannaceae family is derived within the Zingiberales with only a single fertile stamen, further reduced to a single theca (half of a fertile stamen) with a prominent, expanded petaloid appendage (Kirchoff 1983). For many years, the interpretation of the androecium members of the Cannaceae has sparked much debate, especially regarding the developmental origins of the stamen and the labellum in Canna indica (see the brief review by Miao et al. 2014). In recent years, this scientific problem has been the subject of both morphological observations and molecular studies. Almeida et al. (2013) combined a developmental study of the petaloid fertile stamen with data on the expression of three B-class and two C-class floral organ identity genes to elucidate the organogenesis of the petaloid stamen and staminodes. In that study, the authors (Almeida et al. 2013) proposed that the canonical BC model is not sufficient to explain petaloidy in the androecial whorl in Canna sp. In addition, based on floral vasculature and ontogeny, Miao et al. (2014) suggested that the labellum incorporates two androecial members from both the outer and inner whorls, one diverging from the carpellary dorsal bundle and the other from the parietal bundle. The functional stamen also incorporates two androecial members from both the outer and inner whorls. In order to obtain a better understanding of the identity of the androecium members, we carried out a molecular investigation of C. indica B-and C-class genes. Based on our molecular results, we discuss the identity of the androecium members in C. indica in the context of the current molecular, morphological and anatomical data.

RNA extraction and cDNA synthesis
Fresh floral materials were wrapped in aluminium foil, flash frozen in liquid nitrogen and preserved at 280 8C to avoid RNA degradation. To ensure RNase-free materials, the plastic experimental apparatus were dipped in DEPC dH 2 O overnight, whereas all the glass apparatus and mortar are sterilized at 180 8C in the pressure vapour sterilizer overnight before the experiment. Total RNA was extracted from at least 100 mg of floral materials using the Plant RNA Extraction Reagent kit according to the manufacturer's instructions (Bioteke, China). The RNA pellet was re-eluted in 40 mL of RNase-free ddH 2 O and stored at 280 8C. RNA quality was characterized by running 4 mL of extracted RNA through 1.0 % agarose gel electrophoresis. Four microlitres ( 10 mg) of RNA was used in a reverse transcription into cDNA using a poly-T primer designed with software Primer Premier 5 (3 ′ -RACE primer: AAG CAG TGG TAT CAA CGC AGA GTA CTT TTT TTT TTT TTT TTT TTT TTT TTT TTT T) and M-MLV (Invitrogen SuperScriptIII First-Strand Synthesis System for RT-PCR, USA) following the manufacturer's protocol. The success of the reverse transcription reaction was assessed by amplifying 18sRNA from the cDNA, using the following primer pair: 1265S: GGC TAC CAC ATC CAA GGA AG; 1265AS: CCA TTA AAT AGG TAG GAG C.

Sequence analyses
cDNA was diluted 10 times in dH 2 O, and 4 mL of the dilution was used in 20 mL polymerase chain reaction (PCR) reactions containing 1 mL of 20 mM of each of the forward and reverse primers, 1 U of Taq DNA polymerase (TaKaRa, Japan), 2 mL of dNTP (2 mM) and 2 mL of 10× PCR buffer. The forward primers were designed according to the MADS domain and a RACE-AP Reverse. Two degenerate primer pair combinations were designed according to the conserved region in some monocot plants and were used for each cDNA sample: RACE-AP(R): AAG CAGTGGTATCAACGCAGAGTACGCGGG; GLO-F: ATGGGGCG MGGRAAGATCGAGATCAAG; AG-F: ATGGGSMGRGGRAAG ATCGAGATC.
Cycling conditions were set as follows: initial denaturation at 94 8C for 3 min, touchdown procedure from 62 to 48 8C, 15 amplification cycles (94 8C for 30 s, 62 8C for 30 s and 72 8C for 1 min), the cycle of 48 8C followed by another 15 amplification cycles (94 8C for 30 s, 54 8C for 30 s and 72 8C for 1 min) and a final extension step carried out at 72 8C for 10 min.
The full volumes of the PCR reactions were run on 1.0 % agarose gels and bands of the appropriate size (800 bp -1 kb) were excised and cloned on PMD18-T vectors (TaKaRa). Positive colonies were identified by bacterial colony PCR. Then, all the positive clones (six to eight) were picked from each cloning reaction and sequenced using T-vector-specific primers M13-47 and M13-48. All colonies were sequenced to verify identity, and two sequences were submitted to GenBank.
Analyses of the obtained sequences as well as their inferred amino acid translations were performed using the NCBI Blast program (http://www.ncbi.nlm.nih.gov). Alignments of the deduced amino acid sequences were carried out by ClustalX v2.0 (Thompson et al. 1997).
Forty MADS-box B-, C-and D-class genes including the two genes of this study were downloaded from NCBI (http://www.ncbi.nlm.nih.gov/nuccore) and a multiple sequence alignment was performed. Gene and species names as well as accession numbers are as follows: and CiGLO (JQ180192) of C. indica. The evolutionary history of the genes in the multiple sequence alignment was inferred by using maximum likelihood and the Tamura -Nei model (Tamura and Nei 1993). The Kimura two-parameter model (Kimura 1980) was also tested and resulted in the same tree topology. Initial trees for the heuristic search were obtained automatically by applying neighbour-joining and BioNJ algorithms to a matrix of pairwise distances estimated using the maximum composite likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. Evolutionary analyses were conducted in MEGA6. Bootstrap values on the consensus trees were derived from 1000 bootstrap replicates.
Because the gene sequences (accession numbers GU594899-GU594995) submitted by Bartlett and Specht (2010) and Almeida et al. (2013) are only partial CDS, they were not included in this phylogenetic construction.

Real-time reverse transcription PCR
Four whorls of floral organs were dissected from C. indica flowers just before anthesis. The dissected whorls are as follows: sepals in whorl 1, petals in whorl 2, petaloid staminodes and the labellum in whorl 3o (outer androecial whorl), petaloid staminodes and a single fertile stamen in whorl 3i (inner androecial whorl) and carpels in whorl 4. These materials were then separately put into liquid nitrogen to avoid RNA degradation. Meanwhile, a fresh leaf was also collected, flash frozen and used as a reference material. RNA for real-time reverse transcription (RT)-PCR was extracted separately from the four whorls of the floral organs and leaf. RNA extraction was carried out using the Plant RNA Extraction Reagent Kit according to the manufacturer's instructions as described above (Bioteke). The RNA pellet was re-eluted in 40 mL of RNasefree ddH 2 O and stored at 280 8C. RNA quality was tested by running 2 mL of the extracted RNA through a 1.0 % agarose gel electrophoresis (AGE). RNA was quantified using a NanoDrop 1000 spectrophotometer (Thermo Scientific, USA). One microgram of RNA from each sample was reverse transcribed into cDNA for real-time RT-PCR following the manufacturer's protocol (TaKaRa, Prime-Script RT Master Mix Perfect Real Time).
Primer pairs of the reference gene were designed according to the b-actin sequence of Musa acuminata, a member of the banana family. The primer pairs used for real-time RT-PCR are actin-F: TCCTTTCCCTCTATGC TAGTGGC; actin-R: CCTCCAATCCAGACACTGTACTTA; AG-F: TAGAGATTATGGGGCGAGGAA; AG-R: ATCCGAGTTGGAGGG CAGT; GLO-F: GAGGCACTCCAGATTGGGCTTAC; and GLO-R: TGGGAAAAGGATGACGGGAAG. Reagent concentrations were used as follows: 2× SYBR buffer solution (TaKaRa) 10 mL, RoxII 0.4 mL, dH 2 O 2 mL, reverse primer 0.8 mL (10 mM, 0.052 mg) and forward primer 0.8 mL (10 mM, 0.065 mg) and diluted cDNA 3 mL for the GLO reaction system and 6 mL for the AG reaction system. Owing to the already described low expression of AG, the volume of diluted cDNA was doubled for the AG reaction. Each reaction was amplified by two pairs of primers and processed with three biological replicates of the six floral organs extracted from multiple flowers: sepals, petals, petaloid staminodes, labellum, fertile stamens and carpels. b-Actin was amplified from the six floral organs as a reference. The real-time RT-PCR was run on an ABI7500 (USA) for 45 cycles, and cycling conditions were set as follows: pre-heating at 95 8C for 30 s, PCR reaction at 95 8C for 3 s followed by 60 8C for 30 s, and melt curve stage at 95 8C for 15 s, 60 8C for 60 s and 95 8C for 15 s. The relative quantization approach through a DeltaDelta CT method was used to analyse the results of real-time PCR experiments. ABI 7500 software v2.0.1 was used to collect data and carry out the statistical analysis.

Phylogenetic analyses
Phylogenetic analyses (Fig. 3) indicate that our recovered sequences fall into two distinct clades. Three groups can be distinguished by representative genes of the two clades, which are the AGAMOUS (AG), GLOBOSA (GLO) and DEFICIENS (DEF) subfamilies, respectively. Among the three groups, the two B-class subfamilies, DEF and GLO groups, were resolved as sister lineages and together form a sister clade to the C-class subfamily, the AG group. CiAG and CiGLO reported in this study are present in the AG and GLO clades, respectively. CiAG clustered together with Ah-MADS6, AoAG and other monocot C-class members and formed a monocot C sub-group with a 100 % bootstrap value. CiGLO clustered together with AhMADS8, AoPI and other monocot B-class members and formed a monocot B sub-group on a 97 % bootstrap value. D-class genes SHP1, SHP2, FBP11, STK and MADS13 are included in the AG group and show close relationships with C-class members.

Expression patterns of two MADS-box genes
The expression patterns of CiAG and CiGLO in different floral organs were analysed using real-time RT-PCR, as shown in Fig. 4.
CiAG expression signals were observed in both the outer and inner whorls of staminodes (including petaloid staminodes and the labellum), the fertile stamen and carpels. However, CiAG expression was not observed in sepals and petals as previously reported (Almeida et al. 2013). In other words, CiAG was expressed in floral organs of whorls 3 and 4, and not expressed in whorls 1 and 2, similar to the expression pattern of other AG-like homologues' based on the BC model. The strongest expression signal of CiAG was detected in the labellum of whorl 3o.
CiGLO expression was observed in petals, the outer whorl of petaloid staminodes, the labellum and the fertile stamen as well as carpels, while it was not observed in sepals. In other words, CiGLO was expressed in the floral organs of whorls 2, 3 and 4, but was not expressed in whorl 1. The strongest expression signal of CiGLO was detected in the inner whorl where the single fertile stamen is initiated.
In addition, the CiAG expression signal was also detected in the leaf, whereas no CiGLO expression signal was detected in the leaf.

Discussion
Morphological and anatomical studies of C. indica (Kirchoff 1983) have shown that the sequence of floral organ initiation is as follows: calyx (sepals), corolla (petals), inner androecial whorl, outer androecial whorl and gynoecium. The developmental work of Kirchoff (1983) supports the hypothesis proposed by Eichler (1875), that the stamen, the labellum and one of the staminodes (the inner staminode) represent the inner androecial whorl while the remaining one or two petaloid staminodes represent the outer androecial whorl. The nature of the labellum was not formulated by Kirchoff (1983). The most extraordinary and ornamental parts in the flower of Canna L. are staminodes, which are usually taxonomically interpreted as petals by people Chen 1981, 2000). In the flora of China, Canna flowers are interpreted as having two whorls of petaloid staminodes, with the reflexed staminode in the inner whorl referred to as the 'labellum'. The labellum is narrower than the outer whorl of staminodes and is useful in distinguishing among species of Canna. In the light of their statements Chen 1981, 2000), the androecium members, except for the fertile stamen, were replaced by petaloid staminodes and the labellum. However, no detailed evidence has been put forward.
In the present study, two genes, CiGLO and CiAG, have been isolated from C. indica. Sequence characterization and phylogenetic relationships indicate that they belong to B-and C-class gene families, respectively. The described expression patterns in eudicot model species indicate that stamens are determined by B-and C-class genes in whorl 3, whereas C-class genes are essential for carpels in whorl 4 (Theißen 2001). The expression pattern in different floral organs analysed by qPCR shows that CiAG was expressed in whorl 3o of petaloid staminodes and the labellum, whorl 3i of petaloid staminodes plus the fertile stamen, and whorl 4 of carpels, while it was not expressed in whorl 1 of sepals and whorl 2 of petals. The qPCR results of CiAG are largely consistent with the expression patterns of C-class genes, as described in eudicot model species. In the study of Almeida et al. (2013), two copies of AGAMOUS genes were isolated from C. indica, and RT-PCR was used to assess the expression pattern. The genes AG-1 and AG-2 show an expanded pattern of expression when compared with the canonical BC model. AG-1 is expressed in a gradient increasing from sepals to gynoecium while AG-2 is evenly expressed in all floral parts studied except in petals. At that time, Almeida et al. (2013) did not amplify the full-length sequences for the two genes. According to the accession number, AG-1 may be a part of CiAG of our study. The expression patterns of AG-1 and CiAG are slighly different, which may result from the different reseach methods used. Moreover, the expression signal of CiAG has been detected in the leaf. CiAG expressed in non-floral tissue as well as in floral organs is in accordance with the research of Kim et al. (2005) in some basal angiosperms and of Chanderbali et al. (2006) in Pe.am.AG of Persea from the Lauraceae family. In this study, the authors proposed that the expression of AG in vegetative tissue is an evolutionary novelty (Chanderbali et al. 2006). Chanderbali et al. (2006) also hypothesized that the tepals of Persea and perhaps in other Lauraceae are derived from stamen primordia on the basis of gene expression and the occasional presence of tepaloid organs in stamen whorls.
CiGLO was expressed in whorl 2 of petals, whorl 3o of petaloid staminodes and the labellum, whorl 3i of petaloid staminodes and the single fertile stamen, and whorl 4 of carpels, while it was not expressed in whorl 1 of sepals. The expression of CiGLO in whorls 2, 3 and 4 seems consistent neither with the aforementioned modified monocot BC model (Van Tunen AJ 1993;Kanno et al. 2003;Kanno et al. 2007) nor with the shifting/sliding boundary model (Bowman 1997;Kramer et al. 2003), although it is still consistent with the canonical BC model of eudicots. The 'shifting boundary ' (Bowman et al. 1991) and 'sliding boundary' (Kramer et al. 2003) models allow the boundary of expression of the B-class genes to 'slide' from that observed in Arabidopsis and Antirrhinum to one that includes the outer perianth whorl (outer tepals) of Ranunculus, Tulipa and other species with an entirely petaloid perianth. The extension of B-class gene expression into whorl 4 of carpels might explain the petaloidy observed in the style of C. indica (Glinos and Cocucci 2011).
Two copies of GLOBOSA were studied by Almeida et al. (2013), and both GLO-1 and GLO-2 show an expansion in their expression domains. GLO-1 and GLO-2 show great similarity to CiGLO of our study, but they are only partial CDS, which makes it difficult to address the relationship between them and CiGLO. Further data are required to see whether CiGLO is a new copy due to its divergent expression patterns. Broad expression patterns observed for ZinGLO1, ZinGLO2, ZinGLO3 and ZinGLO4 support a combination hypothesis which supports the idea that labellum identity is a result of the combination of multiple GLO homologues rather than expression of a single orthologue (Bartlett and Specht 2010). Whether the combination hypothesis is applicable in Canna needs broader investigation.
In the canonical BC model, simultaneous expression of B-and C-class MADS-box genes would result in the specification of stamen identity. CiAG and CiGLO are expressed simultaneously in the labellum, petaloid staminodes and the fertile stamen in the present study. If molecular evidence is sufficient to determine organ identity, one could claim that these results suggest that the three floral organs are of the same identity. However, despite their similar expression patterns, these organs present very different morphologies. The role of B-and C-class genes Other studies, such as stage-specific in situ hybridization, would be necessary for a more complete picture of the expression profiles during development.
In situ hybridization experiments carried out on a C-class gene AhMADS6 and two B-class genes AhMADS5 and AhMADS8 of A. hainanensis (Song et al. 2010) have shown a case similar to the present study, which has corroborated the hypothesis that the labellum of A. hainanensis originated from stamens. Recently, Miao et al. (2014) showed that the vascular bundle of the labellum of C. indica originates from two stamen primordia at an early developmental stage, providing anatomical evidence for the interpretation that the labellum should be regarded as a derivative of staminodes in the inner whorl (Miao et al. 2014). Although the molecular data presented in this study do not provide sufficient evidence to support that anatomical result, they are also not inconsistent with this interpretation. In accordance with the expression pattern of B-and C-class genes described by Almeida et al. (2013), our present data do not show differential expression patterns between fertile and sterile elements within the androecial whorl, and neither set of results can explain the observed petaloidy or how, mechanistically, the staminodes and the fertile stamen show distinct morphologies.

Conclusions
Based on the expression patterns of B-and C-class genes, the molecular data presented in this study suggest that petaloid staminodes and the labellum are of androecial identity, in concordance with their position in the floral meristem and in agreement with largely described expression patterns in model species. However, these expression patterns are not sufficient to explain the distinct morphologies observed in staminodes and the fertile stamen. Broader and in-depth exploration of other copies of GLO and AG genes and sequence evolution, as well as a broader understanding of the involvement of B-and C-class genes in floral organ identity regulatory networks are required for a better understanding of the relationship between B-and C-class gene expression patterns and the floral organ morphology in Canna.

Contributions by the Authors
Isolation of the sequences and real-time PCR analyses and data collection were accomplished by Q.F. following supervision by J.P.L. and the laboratory assistance of H.F.L. and P.Z. Phylogenetic analysis was performed by Y.F.K. Writing and editing and comprehensive revision of the manuscript was performed by Y.F.K. and A.M.R.A.

Conflicts of Interest Statement
None declared.