(This chapter has been published in Botanical Studies 59:24. 2018)
Kuan-Ting Hsin and Chun-neng Wang
Authors’s contribution
CNW conceived the context of the manuscript. CNW and KTH designed the experiment. KTH performed the experiment. Both authors read and approved the final manuscript.
Abstract
Bilateral symmetry flower (zygomorphy) is the ancestral state for Gesneriaceae species. Yet
independent reversions to actinomorphy have been parallelly evolved in several lineages. Conandron ramondioides is a natural radially symmetrical species survived in dense shade mountainous habitats where specialist pollinators are scarce. Whether the mutations in floral symmetry genes such as CYC, RAD and DIV genes, or their expression pattern shifts contribute to the reversion to actinomorphy in C. ramondioides thus facilitating shifts to generalist pollinators remain to be investigated. To address this, we isolated putative orthologues of these genes and relate their expressions to developmental stages of flower actinomorphy. Tissue specific RT‑PCR found no dorsal identity genes CrCYCs and CrRADs expression in petal and stamen whorls, while the ventral identity gene CrDIV was expressed in all petals. Thus, ventralized actinomorphy is evolved in C. ramondioides. However, CrCYCs still persists their expression in sepal whorl. This is congruent with previous findings that CYC
expression in sepals is an ancestral state common to both actinomorphic and zygomorphic core Eudicot species. The loss of dorsal identity genes CrCYCs and CrRADs expression in petal and stamen whorl without mutating these genes specifies that a novel regulation change, possibly on cis
‑elements of these genes, has evolved to switch zygomorphy to actinomorphy.
Keywords: Peloria, Reversal, Zygomorphy, CYCLOIDEA, DIVARICATA, RADIALIS, Bilateral symmetry
Introduction
Evolutionary reversal to actinomorphy (flower radial symmetry) from zygomorphy (flower bilateral symmetry) have occurred multiple times independently across flowering plant
diversification (Hileman, 2014). Although zygormorphy enhances pollination specialization, these reversals may have evolved in a benefit of increased pollinator generalization when pollinators are scarce or in harsh conditions (Cronk & Moeller, 1997; Donoghue et al., 1998). The frequent
transitions of floral symmetry reversals along many flowering plant lineages imply that a similar or modified developmental program has been independently recruited for floral symmetry transition (Zhang et al. 2013; Hileman 2014).
Gesneriaceae species are predominantly bilateral flower symmetry (zygomorphy) and exhibit a great diversity of floral forms (Endress, 1999, 2001; Weber, 2004). Their flowers have been
adaptively evolved with bee, fly, moth, birds and even bat into a variety of pollination syndromes (Harrison et al. 1999; Perret et al. 2007). However, there are also certain lineages evolved with flower reversion to actinomorphy. There are at least five and four independent reversals to actinomorphy events occurred in old world and new world Gesneriaceae lineages, respectively (Wang et al., 2010; Smith et al., 2004; Clark et al. 2011). This frequency of reversals to
actinomorphy in Gesneriaceae is the highest among Lamiales, probably because it is the basal most lineage of Lamiales which is just derived from the ourgroup actinomorphic Oleaceae species (Endress, 1999). The repeated reversals to actinomorphy in Gesneriaceae species also implied there are perhaps similar yet modified developmental programs repeatedly recruited among these
reversals.
It has been argued that the reversals to actinomorphy may contain selection disadvantage by losing its specific pollinators (Cronk & Moeller, 1997). However, in the aforementioned examples, the actinomorphic floral forms caused by reversals in Gesneriaceae species are usually compromised by pollinator shifts and pollination strategies switched from nectar to pollen rewards (Weber, 2004).
The reversal to actinomorphy with corolla tube fully opened to attract every kinds of general
pollinators may be selected for when only few pollinators are available (Cronk and Moeller, 1997).
For example, the European relict actinomorphic species in Pyrenees, Ramonda myconi, has been inferred that reversion to actinomorphy by opening of the corolla tube provides adaptive advantage to attract more available general pollinators in harsh alpine habitat (Cronk & Moeller, 1997; Pico et al. 2002). Additionally, its pollination syndrome has shifted from nectar reward which is usually favored by specific pollinators to pollen reward allowing general pollinators to visit (Weber, 2004;
Wang et al., 2010).
Conandron ramondioides is another relic and paleoendemic genus in East Asia which is also apparently evolved as an actinomorphic reversal from zygomorphic ancestor (Kokubugata and Peng 2004; Wang et al. 2010; Xiao et al. 2011). Like R. myconi, its corolla tube is lost and five petals are equally large with five stamens fully developed. These indicate they are natural peloria with
complete actinomorphy. Nectary glands of C. ramondioides are lost. Also, its stamens are dehisced by apical pores and pollens are powdery suggest they are pollinated by pollen-collecting bees (Wang, pers. obs.). C. ramondioides can only survive in limestone cliff, often in deep shade forest, where pollinators are scarce. Taken together, the reversal to actinomorphy in C. ramondioides could prevent it from relying only on specific pollinators because insect activity is relative low under this habitat.
The establishment of flower zygomorphy requires CYCLOIDEA (CYC) gene specifically
expressed in dorsal side of the flower to promote growth difference between dorsal and ventral petals and retarding dorsal stamen (Luo et al., 1996; Luo et al., 1999; Corley et al., 2005). Thus the reversal to actinomorphy in C. ramondioides could probably results from loss of CYC function or a CYC expression shifts in flower bud. In Antirrhinum majus, mutation of CYC and its paralog DICH can result in complete actinomorphy with all petals equal in size resembling the ventral one and no retardation on stamens. CYC in C. ramondioides may also be mutated thus becoming actinomorphy.
However, CYC and DICH‘s effect is through activating a downstream MYB family gene RADIALIS (RAD) at dorsal side, whose encoded protein restricts another MYB-like protein DIVARICATA (DIV)
to the ventral region (Corley et al., 2005; Costa et al., 2005; Raimundo et al. 2013). Thus in addition to CYC, fully functional RAD is also needed to develop flower zygomorphy. In cyc dich double mutant, RAD could not be activated to restrict ventral identity gene DIV to ventral region resulting in all petals resembling to ventral petal of wild type. This type of actinomorphic reversal due to
mutation of CYC and/or its downstream RAD therefore is often called abaxialized (ventralized) effect (Cronk, 2006; Zhang et al. 2013; Hileman 2014; Spencer & Kim 2017).
In contrast to ventralization, actinomorphy can be established through expanded expression of CYC and its homologues in petal whorl, an adaxialized (dorsalized) effect. In actinomorphic species such as Cadia of legumes and certain Malpighiaceae species, the expression of CYC extended from dorsal regions to lateral and ventral regions of the corolla (Citerne et al. 2006; Zhang et al. 2013). It thus appears that both the loss of CYC-like gene expression (ventralization) and the expansion (dorsalization) of CYC-like gene expression are two major mechanisms in creating flower actinomorphic reversal in angiosperm.
In Gesneriaceae, reversal to actinomorphy through both ventralization or dorsalization were reported (e.g. Bournea leiophylla and Tengia scopulorum) (Zhou et al., 2008; Pang et al., 2010). In B. leiophylla, the BlCYC1 and BlRAD genes were transiently expressed in floral meristem initiation stage and then quickly vanished at latter developmental stages. The loss of CYC expression at later stages correlates with the fact that all petals resembling ventral ones, demonstrating a ventralization form of reversal (Zhou et al. 2008). Unlike B. leiophylla, CYC-like expression in T. scopulorum has ubiquitously expressed in all petals across dorsoventral axis, a dorsalized form of actinomorphic reversal (Pang et al., 2010). Partial CYC sequences have been isolated from both C. ramondioides and R. myconi but no apparent SNP mutations result in premature stop codon in their coding regions (Xiao et al. 2007; Pico et al. 2002). These imply CYC in C. ramondioides may still function. It would therefore essential to investigate whether expression patterns of CYC together with downstream RAD and DIV shifts, which may correlate and explain the developmental switch of actinomorphic reversal in C. ramondioides.
To examine possible roles of floral symmetry genes (homologues of CYC, RAD and DIV) involving in establishment of actinomorphiy in C. ramondioides, we examined their expression patterns along floral developmental stages and separated floral organs. In order to ascertain the developmental process of actinomorphy in C. ramondioides, we also observed the bud development using scanning electron microscope. From these results, we hope to find whether there is a correlation between shifts of expression patterns among these floral symmetry genes and corresponded floral symmetry
transition.
Material and methods
Floral development in C. ramondioides
Floral buds of Conandron ramondioides from stage 3 to 7 stage were collected for SEM examination. The materials were fixed in FAA overnight then transfer to 70％ EtOH for preservation. Fixed materials were pre-dissected under stereo microscope S8APO (Leica) then dehydrated through an ethanol series (85％, 95％, and 100％twice) with each step for 20 minutes.
After dehydration in 100％ ethanol, materials were dehydrated through ascending gradients of acetone, dried with molecular sieve, and finally dried in a critical point dryer (Hitachi E101). Dried samples were mounted on aluminum stubs and then coated with gold-palladium using Hitachi E1011I sputter. Specimens were viewed using FEI SEM at working distance at 10 mm, and
operating at 15kV. Stages of flower development were summarized in the supporting file 1 table 1.
Isolation and characterization of CYC, RAD and DIV homologues
CYC, RAD and DIV homologues were isolated from C. ramondioides total genomic DNA by using degenerate primer pairs. To isolate CYC homologues from C. ramondioides, a pair of well-known primer pair GcycFS and GcycR, for amplifying CYC homologues in Gesneriaceae was used for amplification (Moeller et al., 1999). The PCR products were then cloned into pGEM-T easy vector system (Promega, USA) and at least 7 clones were sequenced for checking numbers of CYC
homologues of C. ramondioides. Then, to isolate RAD homologues of C. ramondioides, degenerate primer pair QAL-F (5’-RTTRGCRGTKTAYGACA-3’) and FPN-R (5’-
TTYCCYAAYTACWGGACCA-3’) locating at conserve MYB domain and conserve 3’ end were designed according to available RAD homologues of other Gesneriaceae species. Last, to isolate DIV homologues of C. ramondioides, degenerate primer pair DIV-F MEI
(5’-ATGGAGATTTTRDCMCCAAGTT-3’) and DIV-YGK-R1 (5’-CTCCARTCYCCYTTYCCATA-3’) locating at conserve 5’ end and R3 domain were designed based on available Gesneriaceae sequences and DIV form A. majus respectively. Both RAD and DIV homologue PCR products were
cloned and sequenced following the same process as previously mentioned above. To extend into the 5’ and 3’ unknown sequence region of amplified RAD and DIV partial sequence above, rapid
amplification of cDNA ends (5’- and 3’-RACE, SMART RACE cDNA amplification kit, Clontech) technique is applied for obtaining full length cDNA according to manufacturer suggestion. To investigate the homology of isolated CYC, RAD and DIV of C. ramondioides, we aligned full length sequences of them with their genbank available homologs from subfamily Didymocarpoideae where it belongs to and those from closely related model species Antirrihum majus (Scrophulariaceae) (Luo et al., 1996; Almeida et al., 1997; Zhou et al., 2008; Yang et al., 2012).
Phylogenetic analysis of isolated CYC, RAD and DIV homologues in C. ramondioides
To check the homology of these isolated CYC, RAD and DIV genes from C. ramondioides, available NCBI homologues of Gesneriaceae species, A. majus and Arabidopsis were downloaded and used for reconstructing their phylogenies respectively. Sequences used to reconstruct phylogeny were listed in supporting table 2. Nucleotide sequences were first translated into amino acid and aligned using default settings in CLUSTALX (Thompson et al. 1997) with major domains specified, then manually aligned afterward. We apply both neighbor joining (NJ) and maximum likelihood (ML) algorithm for testing the robustness of reconstructed CYC, RAD and DIV phylogeny. The NJ tree of each gene dataset was reconstructed using MEGA 6 (Tamura et al., 2013). For ML tree, the web interface PhyML 3.0 was applied (Guidon et al., 2010). Best-fit nucleotide substitution model of each dataset was evaluated by smart model selection (SMS) which is implementing in PhyML 3.0 (Lefort et al., 2016). For CYC-like gene dataset, the best-fit model is HKY + G model. For RAD, TN93 + G model is suggested and GTR + I + G model for DIV.
Locus-specific RT-PCR
Flower buds were categorized into three stages (Fig. 3-1): early stage (Stage 10, 2-4mm in diameter, sepal longer than petal), middle stage (Stage 13, 5-7mm in diameter, petal longer than sepal), and
late stage (Stage 15, anthesis) (see supporting file 1 table 1). They were collected, in the field
through fixing in RNAlater (Ambion, Life technologies, USA), or freshly collected from individuals grown in the greenhouse. Next, to detect expression locations of CYC, RAD and DIV homologues on petals, a single petal was dissected from flower buds at early developmental stage. Because RNA yield may be low in single petal, pooled sepals, petals, stamen and gynoecium were dissected from flower buds at early stage to confirm the expression locations of CYC, RAD and DIV homologues in C. ramondioides. Total RNA of floral buds and dissected floral organs were extracted following TRIzol® Reagent (Invitrogen, USA) protocol. Single-strand cDNA (20ng/ul) were reverse
transcribed from total RNA of these samples by SuperscriptIV Reverse Transcriptase (Invitrogen).
Gene specific primer pairs were used to examine each candidate gene’s expression level: CrCYC1C (forward: 5’-AGACATGCTTTCTGGCCACT-3’, Reverse: 5’-CTTCTTCGCCTTCTGAATGC-3’), RT-PCR mixture contains 12ul of Ampliqon master mix Red III (Denmark), 6.25mM of each primer, 9.5ul ddH2O and 1ul first-strand cDNA. The PCR condition used for amplifying CrCYC1C,
CrCYC1D and CrCYC2 genes were 94℃ for 3 minutes followed by 35 cycles of 94℃ for 30s, 53℃
for 40s, 72℃ for 20s and a 2 minutes final extension at 72℃. To amplify higher Tm CrRAD1 and CrRAD2 genes, the program is 94℃ for 3 minutes followed by 35 cycles of 94℃ for 30s, 57℃ for 40s, 72℃ for 20s and a 2 minutes final extension at 72℃. For CrDIV gene, the program is 94℃ for 3 minutes followed by 35 cycles of 94℃ for 30s, 53℃ for 40s, 72℃ for 20s and a 2 minutes final
extension at 72℃. Cr18S was used as positive control. Two biological replicates were carried out to validate the reproducibility of the results (see supporting file 3: Figure S1).
Results
Floral development
Development of flower in C. ramondioides can be divided into 15 milestone stages. From stage 3 SEM picture (Fig. 3-2a), all five sepals were already initiated but dorsal and lateral sepals were slightly smaller than ventral ones (a residual zygomorphy). During stage 7 (sepal removed), petals and stamens appeared to be equaled in size when initiated (Fig. 3-2c). This is more evident during stage 7A when gynoecium started to emerged, in which all petals and stamens were grown in equal size (Fig. 3-2d). Petals and stamens continued to grow in equal rate thus all 5 petals and stamens are the same size toward anthesis (Fig. 3-2e). The floral diagram of fully developed C. ramondioides thus can be drawn as Fig. 3-2f, showing complete actinomorphy of C. ramondioides flower at anthesis.
Characterization and phylogeny of CrCYC, CrRAD and CrDIV genes
Among CYC clones, three out of 8 belonged to CrCYC1C, two belongs to CrCYC1D and three belongs to CrCYC2. For CrRADs, four clones belonged to CrRAD1, while the other four were CrRAD2. For CrDIV, all seven clones belonged to CrDIV. We believed our approach can effectively isolate all possible copies of each gene because we designed the primers in most conserved domain of each gene (see “Material and methods” section). Full-length cDNA of CrCYC genes, CrRAD genes and CrDIV were isolated from developing floral tissues and dissected tissues. The CrCYC, CrRAD and CrDIV sequences we isolated from C. ramondioides have been deposited in NCBI database (Accession number MH366524 to MH366529, detailed information see supporting file 2:
Table S2). There are three CYC homologs, CrCYC1C, CrCYC1D, CrCYC2 identified in C.
ramondioides (Fig. 3-3a). Their full length amino acid sequences are 339, 338, and 335, respectively.
According to phylogeny (Fig. 3-4a), we designated them as CrCYC1C, CrCYC1D and CrCYC2.
Sequence analysis shows that CrCYC1C, CrCYC1D and CrCYC2 are 26.1%, 17% and 23% identical to Antirrhinum CYC, respectively. When comparing the TCP domain, R domain, and ECE domains,
CrCYC1C, CrCYC1D, and CrCYC2 shared 92.6%, 90.2% and 92.6% amino acid sequence identity with Antirrhinum CYC, suggesting these genes are functionally related. When compared with CYC-like genes from available closely related Didymocarpoid Gesneriaceae species, CrCYC1C and
CrCYC1D are 86% and 78.4% identical to BlCYC1 of Bournea leiophylla, respectively, and CrCYC2 is 88.1% identical to BlCYC2.
The sequence difference between these CYC-like genes are mainly located in the intervening regions of CYC domains mentioned above. Phylogenetic analysis shows that CrCYC1C, CrCYC1D (first isolated in this study) and CrCYC2 formed three monophyletic clades (GCYC1C, GCYC1D and GCYC2) with other Gesneriaceae CYC homologs at amino acid level, confirming to previous
phylogenetic trees (Fig. 3-4a)
Next, two RAD homologues isolated from C. ramondioides were CrRAD1 and CrDAD2. They shared 76% and 68% amino acid identity with RAD from A. majus, respectively. Both CrRAD1 and CrDAD2 have one conserved 55aa-MYB-domain as RAD does (Fig. 3-3b). Phylogenetic analyses based on neighbor-joining method showed that CrRAD1 and CrRAD2 formed two distinct clades with high support (bootstrap/ML: 96/99 for RAD1 clade; 100/100 for RAD2 clade) (Fig. 3-4b). With A. majus RAD as outgroup, CrRAD1 formed one monophyletic clade with Didymocarpoid
Generiaceae species such as RAD1 of Saintpaulia ionantha with high support, while CrRAD2 formed another monophyletic clade with Saintpaulia RAD2 and RAD-like gene of B. leiophylla according to nucleotide and amino acid NJ tree (Fig. 3-4b).
We only isolated one DIV homolog (CrDIV) from C. ramondioides which encode a protein of 296 amino acids. CrDIV is 54% identical to DIV from A. majus at amino acid level, and 90% and 89%
identical to BlDIV1 and BlDIV2 from B. leiophylla. Conserved R2 and R3 domain and MYB-specific motif “SHAQKY” were found in CrDIV (Fig. 3-3c). Phylogenetic trees reconstructed from
nucleotide and amino acid showed that CrDIV forms a monophyletic clade with BlDIV1 (Fig. 3-4c).