Floral organ MADS-box genes in Cercidiphyllum japonicum (Cercidiphyllaceae): Implications for systematic evolution and bracts definition

The dioecious relic Cercidiphyllum japonicum is one of two species of the sole genus Cercidiphyllum, with a tight inflorescence lacking an apparent perianth structure. In addition, its systematic place has been much debated and, so far researches have mainly focused on its morphology and chloroplast genes. In our investigation, we identified 10 floral organ identity genes, including four A-class, three B-class, two C-class and one D-class. Phylogenetic analyses showed that all ten genes are grouped with Saxifragales plants, which confirmed the phylogenetic place of C. japonicum. Expression patterns of those genes were examined by quantitative reverse transcriptase PCR, with some variations that did not completely coincide with the ABCDE model, suggesting some subfunctionalization. As well, our research supported the idea that thebract actually is perianth according to our morphological and molecular analyses in Cercidiphyllum japonicum.


Introduction
Cercidiphyllum japonicum Sieb. Et Zucc. is a tertiary relic plant and only occurs as a species of east Asian flora. Paleontology research shows that it was once widely distributed in the northern hemisphere. Due to quaternary glaciations, it is now only sporadically found in China and Japan [1,2]. As a cretaceous relic, C. japonicum has considerable presence as a tree with colorful leaves. The tree displays typically colored leaves showing amaranthine in the spring, emerald in the summer, golden in the fall and carmine in the winter. As well, it has great economic value given that its fruits and leaves can be used as medicines and the bark is used for tannic extracts. Furthermore, its dioecious, achlamydeous and extreme simplification inflorescence makes it an ideal material for the study of sexual differentiation and regulation of floral development.
Since it was established by Siebold and Zuccarini in 1846 [3], the systematic position of C. japonicum has always been in dispute. In the early years, researchers classified it according to its morphology and it was once placed in the Magnoliaceae [4]. Baillon [5] proposed that a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Plant materials
Flower buds were collected from C. japonicum growing under natural conditions in Beijing with the cooperation of Dr. Guoke Chen from Institute of Botany, the Chinese Academy of Sciences. One part of the buds were immersed in glutaraldehyde. The others buds for cloning were separated into seven parts-outer scale (OS), middle scale (MS), inner scale (IS), stamens (ST) or carpels (CA), juvenile leaves (LE), stipule (STI) and bracts (BR) and immediately frozen in liquid nitrogen and stored at -80˚C until used.

Isolation and identification of genes
Total RNA was extracted from floral buds using the EASYspin plant RNA Extraction Kit (Aidlab, China) following instructions from the manufacturer. First-strand cDNA was synthesized from 1 μg of the DNase I-treated RNA, using adaptor primers and M-MLV Reverse Transcriptase (TaKaRa, Japan). Initial amplification for core sequences were based on homologous cloning. The PCR reagents were composed of 1 μL cDNA, 0.5 μL of each primer (10 mM each), 2.5 μL Ex Taq buffer, 2 μL dNTP (2.5 mM each), 0.3 μL Ex Taq plymerase (TaKaRa, Japan) and adjusted with water to a final volume of 25 μL. PCR was performed with a 3 min 95˚C denaturation step, followed by 35 cycles of 30 s at 95˚C, 30 s annealing at 52-57˚C, a 30-60 s extension at 72˚C and a final extension period of 10 min. The PCR products were purified with the gel extraction kit (TaKaRa) and cloned into pMD18 1 -T vector (TaKaRa). Ligation products were transformed into Escherichia coli Top10 cells (Aidlab China) following instructions by the manufacturer. Then we used 3' RACE and 5' RACE system kits (TaKaRa) to obtain the 3'-and 5'-end sequences of each gene. Full-length cDNA of each gene was obtained by PCR-based cloning with gene-specific forward and reverse primers designed according to the corresponding 3'-and 5'-end sequences. Names and sequences of the primers used in this study are presented in Tables 1 and 2.

Sequence alignments and phylogenetic analysis
Selected sequences were downloaded from the National Center for Biotechnology Information GenBank. The taxa were selected on the basis of aligning results and the representative angiosperm classification according to the APGIII system (APGIII, 2009). Only one taxon provided relatively complete cds and was chosen per order. Alignments were conducted by Clustal X 2.0 using protein sequences and phylogenetic trees were formed by software MEGA7.0 using the Neighbor-Joining (NJ) and Maximum Likelihood (ML) Method. Gnetum gnemon and Picea abies were chosen as outgroups. Relative species and accession numbers are shown in Table 3. Support for the branches was assessed using bootstrap analysis with 1000 replicates.

Gene expression analysis
For our semi-quantitative RT-PCR analysis, total RNA was extracted from seven parts described earlier. Each first-strand cDNA was synthesized using an oligo (dT)15 primer and the M-MLV reverse transcriptase kit. To precisely analyze the tissue-specific expression patterns of each lineage genes, real-time quantitative PCRs are conducted. The experiment was accomplished with SYBR premix Ex Taq (Takara, Japan) using the following program: 95˚C for 30 s; 40 cycles of 95˚C 5 s, and 60˚C for 30 s. The beta-actin gene of C. japonicum Cejaactin is referred as internal reference.

Morphological observations
Mature floral buds from pistillate and staminate flower of C. japonicum were dissected with a needle and photographed under a stereoscopic microscope. All parts were separately fixed overnight in glutaraldehyde (2.5% glutaraldehyde in a 25 mM sodium phosphate buffer, pH 6.8) at 4˚C. After dehydration in a graded ethanol series, the specimens were introduced at a critical point into liquid CO 2 . The dried material was mounted and coated with gold-palladium using a Hitachi E-1010 sputter Coater. Specimens were examined using a FEI-Quanta 200F scanning electron microscope with an accelerating voltage of 15 kV.

Morphological observations
The flowers of C. japonicum are small and inconspicuous, with similar flowering buds and leaf buds. The inflorescence has a juvenile leaf and a stipule which are embedded in three scales. The outer scales are russety, thick and sclerotic. The middle and inner scales are membranous, stretching out from the outer ones as they develop. When young, the middle and inner scales are peak green with a rose-red margin and turn yellowish with a red margin when mature. Juvenile leaves and stipules are found at the bottom of the pedicel. Juvenile leaves with transparent scrotiform glands in the margin are involute when they are wrapped in scales. The stipules are lanceolate, subtranslucent and membranous. The inflorescence of C. japonicum is highly simplified, with their pistillate inflorescence formed by four subtranslucent peak green bracts and 2-6 carpels, whose flat and upturned stigma is yellowish-green when young and turn scarlet when mature (Fig 1A). From our observations, we conclude that there are only two membranous bracts and several stamens whose heads are a bit sharp. The anthers are greenish when young and turn crimson when mature, with filaments almost did not elongate until when they are nearly mature (Fig 1B).
For an individual flower, the morphology of epidermal cells among the various parts-three scales, juvenile leave, stipule, stamen or carpel and bract-are clearly distinct. When comparing the male and female flowers, except for the carpels and stamens, the other corresponding parts of flowers do not show clear differences on epidermal cells. The abaxial epidermal cells on the outer scales are long, fibrous and relatively smooth except for a few short horns (Fig 2A). While the adaxial epidermis can be clearly distinguished, the cells are short, irregular and rough with a raised edge in the middle (Fig 2B). Most epidermal cells on both adaxial and  abaxial sides of the middle scales are short and square, while cells on the edge are longer and with irregular prismatic protuberances (Fig 2C). The inside and outside epidermal cells on the inner scales are basically the same, regular and square in the middle, longer in the margin and straddle parallel grooves (Fig 2D). Epidermal cells on stigma are sunken and irregular in shape; it is hard to distinguish between individual cells. Cells on ventral sutures are square and arranged densely, while the peripheral cells are relative long and smooth (Fig 2F). The epidermal cells on the head of stamens and cells at the stomium of anther are spheroidal or square, but other places of the anthers are irregular, distorted strips, difficult to affirm as single cells (Fig 2G). Elsewhere, the filament cells are smooth and regular and elongated (Fig 2H). Cells of veins are larger and protuberant, while the mesophyll cells are smaller, round or square protuberances (Fig 2I). Epidermal cells of glands on the edge of juvenile leaves are nearly square and smooth (Fig 2J). The epidermal cells on the cusp of stipules are short and round and the margin consists of monolayer cells, while the lower cells are regular strip foundations with parallel contorted folds with spiny protuberances in the margin (Fig 2K). The epidermal cells on bracts are distinct ellipsoid with regular horizontal slender striate bulges and most of them are slotted in the middle or have tee or cross grooves (Fig 2L).

Screenening and phylogenetic analysis of homeotic genes
Ten floral organ identity genes were obtained by homologous cloning and RACE methods. Among these, four clones were identical to AP1, FUL, FUL-like and AGL6 genes. These genes were respectively referred as CejaAP1, CejaFUL, CejaFUL-like and CejaAGL6. Three B-class transcripts were identified and referred as CejaPI, CejaAP3_1 and CejaAP3_2. Two C-classgene were called CejaAG1, CejaAG2 and the only D-class homologous gene was named CejaAGL11. We performed phylogenetic analyses and constructed trees of each gene and classified them into four trees.
According to the phylogenetic analysis of A-class genes, CejaAP1, CejaFUL and CejaFULlike genes are respectively classified with euAP1, euFUL and FUL-like lineages in the basal core eudicots. CejaAP1 and CsAP1 of Corylopsis sinensis (Saxifragales) are sister groups, given bootstrap support under ML (94%) and form a clade with other euAP1 homologues of Saxibragales. CejaFUL and CsFUL of Corylopsis sinensis (Saxifragales) are sister groups and form a clade with HeaFUL of Heuchera americana (Saxifragales) with bootstrap support under ML (95%). CejaFUL-like also forms sister groups with HeaFUL-like of Heuchera americana (Saxifragales) (Fig 3). Since AGL6 lineage was not a typical A-class gene, the phylogenetic tree of CejaAGL6 was constructed only with its own lineage genes. The analysis shows that CejaAGL6 groups with RsAGL6 of Ribes sanguineum (Saxifragales) in the basal core eudicots (bootstrap 82%) (Fig 4).
Two C-class genes, CejaAG1 and CejaAG2, were isolated; phylogenetic analysis showed that CejaAG1 belongs to the euAG lineages and CejaAG2 to the PLE lineages. CejaAG1, PasuAG of Paeonia suffruticosa (Saxifragales) and SxcAG1 of Saxifraga careyana (Saxifragales) gather in a group with bootstrap support under ML (61%). CejaAG2 and LAG of Liquidambar styraciflua (Saxifragales) is a sister group in the ML analysis (bootstrap 61% support). The only D-class gene CejaAGL11 forms a clade with SxcAG2 of Saxifraga careyana (Saxifragales) in the ML analysis (bootstrap 71% support) (Fig 6).

Expression of ABCD Homologs in C. japonicum
The expression patterns of the ABCD Homologs were analyzed by qRT-PCR. The expression patterns of these genes were shown in Fig 7. Except for CejaPI which is expressed strongly in male ones and weakly in female ones, the remaining target genes are barely expressed in juvenile leaves.
For A-class genes, cejaAP1 has similar expression patterns between male and female buds, expressed in inner scales, stipules and bracts. CejaFUL is expressed in all scales, stipules and bracts of male and female buds as well as in carpels. CejaFUL-like is almost only expressed in bracts. CejaAGL6 shows different expression patterns between male and female flowers, with relatively strong expressions in the outer scales of males while weakly in those of females, but expressed relatively week in carpels and stipules. Elsewhere, CejaAGL6 is detected in female bracts but not in male ones. B-class genes are expressed in almost all male floral organs, especially CejaPI which is barely expressed in female buds. CejaAP3_1 is expressed most often in both male and female bracts. This CejaAP3_1 is expressed most in both male and female bracts, where the expression level of CejaAP3_1 is 3-4 times compared with CejaActin.

Systematic evolution of Cercidiphyllum japonicum
CejaAP3_2 is expressed higher than CejaAP3_1 in stamens and carpels, but in both male and female bracts, expression level of CejaAP3_2 is much less than CejaAP3_1. Apart from this observation, we found that, CejaAP3_1 displays a similar expression pattern with CejaAP3_2 between other male and female floral parts (low level). For C-class genes, CejaAG1 is mainly expressed in carpels, stamens and both bracts. CejaAG2 is expressed in carpels and both bracts (low level), but less than CejaAG1. The D-class gene CejaAGL11 is expressed quite strongly in carpels.

Discussion
Since species identification and classification are based on morphology, an increasing number of studies suggested that sole reliance on this approach may lead to the neglect of a significant number of relevant species [25]. As the development of molecular phylogenetics, DNA and amino acid sequence analyses have been an important method to study systematic evolution and development. As Woese [26] argues, sequencial information contains the promise that we will have potentially more evolutionary information than we now possess and allows us to infer a great deal of assurance than we can now.

MADS-box homologs and systematic place
We obtained three A-class, three B-class, two C-class homologs and one D-class homolog from Cercidiphyllum japonicum, which has never been reported before. Phylogenetic analyses show that these floral organ identity genes group with the respective classes of the MADSbox genes from other Saxibragales plants, indicating that placing Cercidiphyllum japonicum in Saxibragales in the basal core eudicots is suitable. The C-terminal regions of C. japonicum genes contained conserved characteristic motifs, typical of the genes of each class (Fig 8), therefore indicating their functional similarities with other homologs regulating flower formations in other plants [27,28].
Only the C terminal is shown. Conserved motifs are boxed, as defined by previous studies for the AP1 motif, the PI and AP3 motifs, and the AG motif.
Recent studies suggested that the major duplication events for floral ABC-class genes occurred at the base of core eudicots [29][30][31][32]. For A-class genes, it has been proposed that a major duplication event occurred near the base of their core eudicots, giving rise to euAP1, euFUL and FUL-like lineages [31,33,34]. All the three A-class lineages we obtained from C. japonicum, thus suggesting that it could have originated after this duplication period. For the AP3/PI subfamily, one duplication formed DEF/AP3 (paleoAP3) and GLO/PI lineages. Subsequently, following the duplication in the base of core eudicots, a frame shift mutation occurred in DEF/AP3 copies and formed TM6 and euAP3 lineages [29,35]. Predicted amino acid sequense of CejaAP3_1 contains a paleoAP3 motif, suggesting that C. japonicum may not originate may not have originated later than the base of the core eudicots. In addition, euAG-and PLE-lineage originated on account of a major duplication in the early period of core eudicots and undergone the functional switch between them after rosid and asterid differentiations [30,36,37]. Since both euAG and PLE homologs were found in C. japonicum, it is further demonstrated that C. japonicum may not have originated earlier than the rosid and asterid divergent period. Hence, the summation of molecular evidence limited the systematic place of C. japonicum to the base of core eudicots.
Studies of earlier ABCDE-models were based on the Arabidopsis and Antirrhinum model systems [16,17]. Based on ABCDE-model, we speculated that the sexual differentiation of C. japonicum may be related to the B-/C-class homologs. In the most recent common ancestor of gymnosperms and angiosperms, the primitive function of AG lineage was to differentiate the reproductive organs from nutritional organs [38,39]. The function of DEF/GLO lineage is to differentiate male and female [40]. The B-class gene SlAP3Y in Silene latifolia is located in the Y chromosome and related to gender decision [41]. The qRT-PCR results show that CejaAG1 is highly expressed in stamens and carpels, while the CejaAG2 is almost only expressed in carpels strongly. Previous studies have indicated that the B-class genes of core eudicots are stably Systematic evolution of Cercidiphyllum japonicum expressed in petals and stamens, but this is not always coincident with the B-class genes of basal eudicots and basal angiosperms [42]. For instance, the CejaPI is almost male specific, since it is strongly expressed in all male organs and barely examined in female ones. These results may indicate that CejaAG1 plays an important role in reproductive organ formation. As well, CejaAG2 and CejaPI are crucial to carpels and stamens in floral development of C. japonicum respectively. Since functional verification is difficult to conduct in woody material, evidence for functions of identified ABCDE genes of Cercidiphyllum japonicum should use the corresponding mutant Arabidopsis as medium in future studies.
Confusing structure of C. japonicum In general, C. japonicum is thought to be missing the perianth. When we observed the male and female inflorescences, we encountered that there were two lamelliform and membranous bracts in male while there were four in female ones. Ding [43] described the 'bracts' as four sepals in the Flora of Henan. In another point of view, Yan et al. [15] observed morphogenesis of C. japonicum and considered that bracts should be closer to phyllome, but the so called bracts in C. japonicum developed with their basal stamens or pistils correlatively; hence they proposed that the so called bracts are more closely related to tepals. We found that leaf buds and flower buds are much the same except for their reproductive parts. According to the Agricultural Dictionary, a bract is actually a phyllome. Based on the model, the absence of petals in C. japonicum might be due to the null function of A-class and B-class homologs. The APl/ SQUA family, such as AP1 mutant of Arabdopsis and SQUA mutant of Antirrhinum majus, may cause changes of petals and sepals [33,44,45]. Moreover, the petals were converted to sepals and stamens to carpels in the ap3 and def mutants [40]. Unfortunately, definite evidence of A-class homologs has never been demonstrated in woody plants and the expression patterns are not strictly conserved. In most primitive angiosperms, it is the petals not bracts or sepals having high expression levels of both A-and B-class genes, such as Orchid [46,47], Trochodendron [48] and Eucalyptus of Saxifragales [49]. Wróblewska et al. [50] analyzed expression patterns of key flower genes of several Magnoliaceae and found that the B-class genes, AP3 and PI, were restricted to the second and third whorl. In our research, the qRT-PCR results show that both A-and B-class genes, especially CejaAP1 and CejaAP3_1/_2 whose homologous genes are petal decisive in Arabidopsis, had significant expressions in the bracts that are different from other organs of C. japonicum. Recent studies in Arabidopsis and Antirrhinum, as well as several other species, indicate that the function of floral MADS-box genes is largely associated with the expression patterns of these genes, particularly when expression levels are high [51]. What is more, the epidermal cells of the bracts show considerable differences from other phyllomes. In view of this inference, we recommended that the so-called bracts actually should be considered as perianth.

Exon skipping of CejaAP3
Alternative splicing has been found in several MADS-box genes which, to some extent, might have either an important positive or negative impact, typical in Magnolia stellata [52]. During the screening, two CejaAP3_1/_2 spliceosomes were found. After examining the genomic sequence, we found that the two clones may be formed by alternative splicing. In addition, the shorter spliceosome, CejaAP3_2, was confirmed to be missing an exon 4 (Fig 9). What is more, the results of qRT-PCR shows that CejaAP3_2 displays a high expression in stamens and moderate expression in other floral parts, indicating that this abnormal splicing may have a significant impact on the floral development of C. japonicum, especially the perianth. However, the exact nature of this product and its interactions need further study.
We conclude that all floral homeotic gene phylogenies show that C. japonicum is closely related to the plants of Saxifragales, suggesting that our species should be placed in Saxifragales at the base of core eudicots. This result confirms the APGIII system and supports a new train of thought when investigating systematic evolution based on floral organ identity genes. As well, our research supports the conjecture that the so called bracts of C. japonicum actually are perianth, a conclusion based on morphology and expression patterns.