A homolog of the ALOG family controls corolla tube differentiation in Torenia fournieri

Many nectar-feeding animals prefer high-quality nectar, and plants have evolved multiple ways to protect their nectar reward (Heyneman, 1983; Nachev et al., 2017; Sprengel, 1793). One major threat to nectar is rain, as raindrops can physically injure the corolla, wash away pollen grains, and spoil and dilute nectar (Darwin, 1876; Hagerup, 1950; Sprengel, 1793). Plants that can protect against the damage caused by rain are favored. In 1793, Sprengel described several structures that protect flowers from rain (Sprengel, 1793). Since then, many biologists have recognized floral features that can adapt to bad weather, including an ability to decrease the openness of the flower (Darwin, 1876), a hydrophobic corolla that gathers water at the base (Hagerup, 1950), changes in the flower orientation (Bynum and Smith, 2001; Huang et al., 2002; Tadey and Aizen, 2001) and production of water-resistant pollen (Mao and Huang, 2009).

Many Torenia flowers have a corolla neck, which is an enlarged cylindrical structure surrounding the nectary located at the floral base. The corollas of Torenia species have differentiated into three regions: a petal lobe, in which petals are separated and exhibit strong zygomorphy; a conical tube, in which different petals fuse together; and a cylindrical corolla neck connected to a conical tube through a constricted region (Fig. S1). The neighboring corolla tissues showed distinct cellular activities according to comparisons of RNA-sequencing data from the conical tubes and corolla necks of T. fournieri. We focused on an ALOG (Arabidopsis LSH1 and Oryza G1) gene, TfALOG3, that is specifically expressed in the corolla neck, indicating its possible role in the development of the corolla neck. We further analyzed the function of TfALOG3 by generating TfALOG3 loss-of-function mutants. The corolla neck structure was no longer apparent in the mutant compared with the wild-type flowers. We also observed that water can easily contact the nectary of the mutant flowers, supporting this hypothetical property of the corolla neck in T. fournieri.

We investigated the flowers of four Torenia species: T. fournieri, T. concolor, T. violacea and T. hybrida. Except for T. hybrida, which is an interspecific hybrid of T. fournieri and T. concolor, the remaining three species are distributed mostly in tropical and/or subtropical areas, such as East and Southeast Asia, where rainfall is always abundant during the flowering season (eFloras, www.efloras.org/). We observed that all of these species possess a corolla neck at the base of the corolla, where the nectary and nectar are protected (Fig. S1).

To validate the water-repellent property of the corolla neck, we mimicked raindrops by applying different amounts of water containing 0.1% Safranin O dye as an indicator into the corolla tubes of T. fournieri. The water ran down to the center of the corolla and stopped once reaching the constricted region (Fig. 1A). Even after increasing the amount of water, water still could not access the nectary (Fig. 1A). Similar results were obtained when we applied 200 μl colored water to the flowers of either T. concolor or T. hybrida (Fig. S2A). When the water volume reached 400 μl, it forced the flowers to change direction, making the water itself flow out from the drooping flowers (Fig. 1A). These results support the property of the corolla neck hypothesized by Sprengel, at least among Torenia.

As biological function relies largely on structural properties, we carefully observed the development of the corolla neck. In a previous study, we divided the process of flower development in T. fournieri into 13 stages (Su et al., 2017). The corolla neck became distinguishable at stage 10 and grew rapidly until stage 13 (Fig. 1). Cell expansion was evident when we examined the outer epidermal cells of the corolla neck (Fig. 1). A large amount of yellow pigment accumulated in the inner epidermal cells of the corolla bottom beginning at stage 10 (Fig. 1B). Cells from the constricted region began to differentiate (Fig. 1C) and further developed into a row of hair cells (Fig. 1D,E). We also examined the corresponding regions in T. concolor or T. hybrida. To our surprise, no hair cells existed in the constricted region of T. concolor, and short hair cells developed in the interspecific hybrid T. hybrida (Fig. S2C).

To test whether the hair cells on the corolla neck also functioned to prevent rain from entering, we removed the reproductive organs and dissected the corolla tubes. By directly applying 20 μl colored water into the cut corolla bottoms, the water traveled to the bottom of the T. concolor corolla but was blocked in the constricted region of T. hybrida (Fig. S2C). These results indicate that, although the constricted region together with the pistil is sufficient to prevent water access, hairs on the constricted region may decrease the ability of water to penetrate. Because under natural conditions, rain is associated mostly with wind (which makes the rain easier to enter the flower), the presentation of hairs functions as another mechanism with which to restrict water access. Compared with T. fournieri, which grows erect, T. concolor has a decumbent growth habit, making its flowers approach the ground; in this case, the wind may not be a problem.

We performed transcriptome sequencing to identify the potential factors involved in regulation of the corolla neck (Fig. 2). Hormone signal transduction and DNA replication were the most enriched pathways among the genes expressed in the corolla neck; however, the biosynthesis of secondary metabolites and flavonoid biosynthesis were the most enriched pathways among the genes expressed in the conical tube, supporting their distinct cellular activities (Fig. 2A,B). These two tissues displayed different expression profiles: 2550 and 1806 genes were more highly expressed in the corolla neck and conical tube, respectively (Fig. 2C). Next, we focused on the genes encoding transcription factors that were more highly expressed in the corolla neck.

The top 10 differentially expressed transcription factor genes included one ALOG, one KNAT6-like KNOX, one AP2, one MYB, several AGAMOUS-like MADS and YABBY genes (Fig. 2D). Phylogenetic analyses of these 10 candidate genes were carried out (Fig. S3). Four MADS genes belonged to AGAMOUS-clade factors, and previous studies have reported that these genes are highly expressed in stamens as well as in pistils (Fig. S3B) and may play a role in the development of plant reproductive organs (Nishijima et al., 2016; Su et al., 2018,b). Two YABBY-like genes are the homologs of Arabidopsis CRC (Fig. S3C), which is required for floral stem cell maintenance and carpel development (Bowman and Smyth, 1999; Yamaguchi et al., 2017). The KNOX-like candidate was homologous to Arabidopsis KNAT2 and KNAT6 (Fig. S3D), two genes enriched in the shoot apical meristem (SAM) and organ boundaries, which are important for the maintenance of SAM and organ separation (Belles-Boix et al., 2006). The MYB-like gene shows homology to Arabidopsis ATMYB16 and ATMYB106 (Fig. S3E), which are involved in the differentiation of the petal epidermis and cutin formation (Baumann et al., 2007; Kannangara et al., 2007; Oshima et al., 2013). The AP2-like gene shares homology to Arabidopsis APETALA2 (Fig. S3F), which is expressed in the floral meristem and various floral organs, and regulates floral stem cells and organ differentiation (Bowman et al., 1991; Drews et al., 1991; Krogan et al., 2012; Wollmann et al., 2010). Among the 10 genes, we selected the ALOG gene (Unigene0004661) as our candidate, due to a lack of sufficient functional information about this gene family. To date, several members from the ALOG family have been functionally elucidated, suggesting their various roles in plant development (Cho and Zambryski, 2011; Lei et al., 2019; MacAlister et al., 2012; Takeda et al., 2011; Yoshida et al., 2013, 2009).

Seven ALOG genes have been isolated previously from T. fournieri, and we confirmed the identification of Unigene0004661 as TfALOG3 after comparison with published sequences (Fig. S3A). The gene is expressed mainly in flower buds, and it encodes a nuclear localized protein belonging to the euALOG1 subclade of ALOG proteins (Xiao et al., 2018). We examined the expression of TfALOG3, together with that of paralog TfALOG1, by reverse-transcription (RT)-PCR (Fig. 3A,B). TfALOG1 was expressed in both the conical tube and corolla neck, whereas TfALOG3 was expressed specifically in the corolla neck, suggesting functional diversification of these two paralogs (Fig. 3B). As mentioned before, T. concolor is devoid of hairs in the constricted region; thus, we measured TfALOG3 homolog expression in dissected T. concolor corollas (Fig. S4). Similar to T. fournieri, TcALOG3 was also expressed specifically in the corolla neck, indicating that ALOG3 is likely to be involved in the 3D shaping of the neck, rather than in the specification of the hair cell fate.

We further determined the spatiotemporal expression pattern of TfALOG3 by RNA in situ hybridization. During the early phase (stage 6), TfALOG3 was expressed at the bottom of the corolla, where petal and stamen whorls fuse together (Fig. 3C). No specific signal was detected from the negative control sense probe (Fig. 3E). During the late phase (stages 9 and 10), TfALOG3 was expressed specifically in the entire corolla neck, with the strongest expression in the inner and outer epidermis (Fig. 3D,F,G), and no expression in the nectar region (Fig. 3D). The specific expression of TfALOG3 strongly supports its potential function in differentiation of the corolla neck.

To elucidate the function of TfALOG3, we generated loss-of-function mutants using the CRISPR/Cas9 genome editing method. Two different plasmids coding for two and three guide RNAs, respectively, were constructed (Fig. S5A). We performed transformation using two T. fournieri cultivars with different floral colors. Mutations were detected in three independent transgenic lines, leading to frame shifts and premature termination (Fig. S5B). In the first mutant, single nucleotide insertions were observed in both targeted sites; in the second mutant, only one nucleotide insertion was detected in the first targeted site (Fig. S5B). In the third mutant, single nucleotide insertions were found in the first and third targeted sites, together with a single nucleotide substitution in the second targeted site (Fig. S5B).

Comparing the floral morphologies between the wild-type and mutant plants, we found that the corolla neck had disappeared in the mutant flowers (Fig. 4A-C, Fig. S6A,B, Fig. S7). Yellow pigmentation was observed at the base of the conical tube in the wild-type flowers, whereas the enlarged corolla neck was whiter (Fig. 4A, Fig. S6A). However, the bottom of the corolla was pigmented in the mutant flowers, and the enlarged corolla neck failed to differentiate (Fig. 4A-C, Fig. S6A,B). Both the inner and outer epidermal cells were smaller in the mutant compared with wild-type plants (Fig. 4D). A row of hairs was present in the boundary between the conical tube and corolla neck in the wild-type flowers, but was no longer apparent in the mutant flowers (Fig. S8). Thus, we concluded that TfALOG3 is involved in regulation of corolla neck development.

During the late stages of petal developmental, cells increase in volume via vacuolation (Huang and Irish, 2016; Sugimoto-Shirasu and Roberts, 2003). In Arabidopsis, the gene encoding the basic helix-loop-helix transcription factor BIGPETALp (BPEp) is regulated by jasmonate signaling (Brioudes et al., 2009). By interacting with AUXIN RESPONSE FACTOR 8, BPEp limits cell expansion and thus regulates petal growth (Szécsi et al., 2006; Varaud et al., 2011). As ARFs function redundantly during multiple aspects of organ development, other ARFs may also be involved in petal morphogenesis. Our transcriptomic analyses indicated high enrichment of factors involved in hormone signal transduction as well as seven ARF unigenes in the corolla neck (Fig. 2A, Fig. S9A). After examining the expression of these genes in the wild-type and mutant plants, significant downregulation in the mutant corolla of an ARF4-like gene was detected, suggesting that this gene is a potential downstream target of TfALOG3 in the regulation of corolla neck development (Fig. S9B-E).

As the cylindrical corolla neck prevents water from entering the nectary, we performed similar assays using the wild-type and mutant flowers. When applying 200 μl colored water to the flowers, we found that the water easily entered the nectary of the mutant flower compared with the wild-type flower (Fig. 4E, Fig. S6C,D). To exclude the effects of male or female reproductive organs, we dissected the corolla bottoms from the flowers and positioned them on slides. After pipetting 20 μl colored water into the cut corollas, we observed that the water was blocked in the neck region of the wild-type corolla, whereas it ran to the bottom of the mutant corolla (Fig. 4F,G). These results indicate the indispensability of the corolla neck. Based on these observations, we propose a potential model to explain how the corolla neck prevents raindrops from entering the flower (Fig. S10). In T. fournieri, the pistil and constriction at the top of the corolla neck, together with hairs, form an intact ‘cover’. When raindrops enter the corolla, they are restricted to this region, thus preventing dilution of the nectar in the nectar container (Fig. S10). However, in the mutant plants, the corolla and pistil fail to form a ‘cover’, allowing the raindrops to enter the nectar container and spoil the nectar (Fig. S10).

The flowers of T. fournieri are upward facing and have relatively wide openings. Raindrops readily enter the flower and gravitate towards its nectar. Thus, the nectar must be secured by measures such as development of a corolla neck. In the future, side-by-side comparisons of pollinator visits between wild-type and mutant flowers would provide a clearer demonstration of the ecological function. Moreover, the corolla neck may also be involved in other biological processes such as maintenance of the nectar microclimate or protection from nectar robbers. For example, many downward-facing flowers of the Penstemon and Salvia genera also have corolla neck-like structures. In these species, the corolla neck may serve as mechanical protection from nectar robbers. These functions need further exploration. As the origin of this corolla structure may be a key innovation during evolutionary history, it will be an interesting issue in the future to explore how the corolla structure originated.

The ABC model explains how the concentric whorls of floral organs are genetically specified, and numerous studies support that B class genes play important roles in the development of petals and stamens (Coen and Meyerowitz, 1991; Irish, 2017). In T. fournieri, two B class factors, TfDEF and TfGLO, have been characterized (Sasaki et al., 2014). In TfDEF/TfGLO-ox plants, sepals change to petal-like structures; however, the petaloid sepals lack a corolla neck at the bottom (Sasaki et al., 2014). When TfDEF and TfGLO are co-suppressed, the petals exhibit sepaloid organs without affecting formation of the corolla neck, indicating that the corolla neck is a B-gene-independent floral structure (Sasaki et al., 2010, 2014).

To summarize, our study revealed two distinct regions in the corolla bottom and characterized an ALOG gene involved in genetic regulation in the corolla neck. In the future, it would be worthwhile to explore the underlying molecular mechanisms controlling differentiation of the corolla tube using T. fournieri as a new eco-evo-devo model.

Torenia spp. plants were cultured at 25°C under a 16 h:8 h light: dark cycle. To produce sterile plants, T. fournieri seeds were sterilized by a series of washes with ethanol, sterile water and then 10% sodium hypochlorite, with four rinses in sterile water. The sterile seeds were germinated on half-strength MS culture medium and refreshed with new medium every 4 weeks.

To observe epidermal cells, floral buds of different stages were collected and dissected under a stereomicroscope. Three types of impression materials were used to produce the epoxy replica, as described previously (Su et al., 2018a). The replicas were sputtered with gold and examined under a JEOL JSM 6360LV scanning electron microscope. To calculate the epidermal cell width, 20 randomly selected cells were measured using Adobe Photoshop CS6 and statistical calculations were performed using multiple t-tests. To observe hair cell development, corollas of different stages were dissected on slides and mounted with mounting medium. To avoid possible dehydration, the slides were immediately examined under a Zeiss Axio Imager light microscope.

We sequenced the transcriptome of the conical tube and corolla neck in two biological replicates. Stage 11 corolla tubes were collected and dissected into two regions, including the conical petal tube and corolla neck. Total RNA was extracted using the Plant RNA Kit (Omega Bio-Tek). Poly(A) mRNA was purified and fragmented prior to sequencing. First-strand cDNA was synthesized using a random hexamer primer, followed by second-strand cDNA synthesis. The sequencing adaptors were ligated to the fragments, which were purified by electrophoresis and amplified by PCR as described previously (Su et al., 2018a). The libraries were paired-end sequenced on the Illumina HiSeq 2000 platform and ∼6 Gb (35.5-43.5 million sequencing reads) were obtained. After obtaining the clean reads, two rounds of de novo assembly were carried out using the Trinity (containing Inchworm, Chrysalis and Butterfly) and CAP3 programs (Gene Denovo Biotechnology) (Haas et al., 2013; Huang and Madan, 1999). After assembly, 42,756 unigenes were obtained, with N50=1901 bp. The reads per kilobase per million mapped reads values were calculated to determine the relative expression of each unigene, and gene ontology annotations were determined by the Blast2GO program (Conesa et al., 2005; Mortazavi et al., 2008). Differentially expressed genes were identified as those with a fold change ≥2 and a false discovery rate <0.05 using the edgeR package (www.r-project.org/). To determine differentially expressed transcription factor genes, we applied ‘log2 Ratio (WT-12-down/WT-12-up) ≥3’ and ‘WT-12-down_count for descending order’ rules. Differentially expressed unigenes and top 10 differentially expressed transcription factor genes are listed in Table S2.

Total RNA was extracted from different tissues using the Plant RNA Kit (Omega Bio-Tek). For RT-PCR analysis, the cDNA templates were prepared using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). For quantitative RT-PCR, the cDNA templates were prepared using the PrimeScript RT reagent Kit with the gDNA Eraser (Takara). PCR was performed using the LightCycler 480 Real-Time PCR System (Roche) according to the manual. β-Actin (TfACT3) expression was measured as an internal control. The primers used in this study are listed in Table S1.

Inflorescences or floral buds were fixed in FAA (formalin-acetic acid-alcohol) fixative containing 50% (v/v) ethanol, 3.7% (v/v) formaldehyde and 5% (v/v) acetic acid for RNA in situ hybridization. The tissues were embedded with Paraplast (Sigma-Aldrich) after treatment with an ethanol series and xylene solution. The probes were labeled with digoxigenin-UTP. Hybridization was performed as described previously (Coen et al., 1990). The primers used in this study are listed in Table S1.

The CRISPR/Cas9 binary vector (provided by Prof. Yaoguang Liu, South China Agricultural University, China) was constructed according to Ma et al. (2015). Target sequences with a protospacer adjacent motif (NGG) are shown in Fig. S5A. Transformation of the vector into T. fournieri was performed as reported previously (Aida, 2012; Su et al., 2017). The edited TfALOG3 sequences are shown in Fig. S5B.

The raw data from the RNA-sequencing analyses have been deposited into SRA under accession numbers SRR8449829, SRR8449830, SRR8449831 and SRR8449832.

We thank Prof. Norihiro Ohtsubo from Kyoto Prefectural University (Japan), Prof. Katsutomo Sasaki from National Agriculture and Food Research Organization (Japan) and Miss Naoe Ando from Nagoya University (Japan) for their patient instructions on plant transformation, and Dr Naoki Yanagisawa from Nagoya University (Japan) for his help on scanning electronic microscopy. We appreciate the editor and three anonymous reviewers for their valuable comments on the manuscript. We are also grateful to Dr Wuxiu Guo from Sun Yat-sen University (China) for helpful discussions and to Mr Chao Huang for his help on collecting T. voilacea in the field.