2019 Volume 88 Issue 2 Pages 284-292
The modified ABC model explains the floral morphology of many monocots, such as the lily and tulip, in which the perianth consists of two layers of almost identical petaloid tepals. According to the modified ABC model, B-class genes are expressed in two perianth whorls, inducing the petaloid structure in both whorls 1 and 2. In this study, we analyzed the expression and function of the B-class genes in the grape hyacinth (Muscari armeniacum). We isolated two DEFICIENS (DEF)-like genes (MaDEF1 and MaDEF2) and three GLOBOSA (GLO)-like genes (MaGLOA1, MaGLOA2, and MaGLOB) from M. armeniacum using rapid amplification of cDNA ends (RACE). Expression analysis showed that MaDEF1 and MaDEF2 were expressed in whorls 1, 2, and 3, whereas MaGLOA1, MaGLOA2, and MaGLOB were expressed in all four whorls. These results support the modified ABC model in M. armeniacum. Overexpression of MaGLOA1 and MaGLOB in Arabidopsis thaliana resulted in a morphological change of sepals to petaloid structures in whorl 1, indicating that the function of these genes is similar that of the B-class orthologs PISTILLATA and GLO in A. thaliana and Antirrhinum majus, respectively. In addition, yeast two-hybrid assays revealed strong protein–protein interactions between MaDEF1 and MaGLOA1, suggesting that MaDEF1–MaGLOA1 is likely to have the main B-function in M. armeniacum. These data support the modified ABC model in M. armeniacum.
Floral morphology and color are among the most important agricultural traits affecting the commercial value of ornamental plants. Therefore, understanding the molecular mechanisms that determine the floral structure of ornamental plants is essential. Genetic analysis of floral homeotic mutants of Arabidopsis thaliana and Antirrhinum majus has led to the classical ABC model (Coen and Meyerowitz, 1991; Weigel and Meyerowitz, 1994). This model explains the dicot floral structure comprising four floral organs, sepals, petals, stamens, and carpels, from the outer to inner whorls. According to the ABC model, A-class genes specify sepals in whorl 1; a combination of A- and B-class genes determines petals in whorl 2; a combination of B- and C-class genes determines stamens in whorl 3; and C-class genes specify carpels in whorl 4. The A-class genes include APETALA1 (AP1) and AP2 in A. thaliana and SQUAMOSA in A. majus; B-class genes include AP3 and PISTILLATA (PI) in A. thaliana and DEFICIENS (DEF) and GLOBOSA (GLO) in A. majus; and C-class genes include AGAMOUS (AG) in A. thaliana and PLENA in A. majus. All of these ABC genes, except AP2, encode MADS-box transcription factors (Theissen et al., 2000). Both types of B-class proteins (AP3 and PI in A. thaliana, and DEF and GLO in A. majus) are needed for B-function because they interact to form a heterodimer (Goto and Meyerowitz, 1994; Jack et al., 1992; Schwarz-Sommer et al., 1992; Tröbner et al., 1992). The discovery of E-class genes, whose products interact with the ABC proteins in A. thaliana, has led to the extension of the classical ABC model to the ABCE model (Honma and Goto, 2001; Kanno, 2016; Pelaz et al., 2000, 2001).
In many higher eudicots, the perianth is clearly differentiated into sepals and petals. By contrast, the perianth of monocots, such as the lily (Lilium spp.) and tulip (Tulipa spp.), is composed of two layers of petaloid organs. These indistinguishable sepals and petals of monocots are collectively referred to as tepals. This floral morphology is explained by the modified ABC model, which was originally proposed based on the morphology of wild-type and putative mutant flowers of B- and C-class genes in the tulip (van Tunen et al., 1993). According to this model, both A- and B-class genes are expressed in the two outer whorls, resulting in two layers of petaloid structures. Molecular analysis of tulip B-class genes shows that the DEF-like (TGDEFA and TGDEFB) and GLO-like (TGGLO) genes are expressed in the outer two perianth whorls and stamens, thus supporting the modified ABC model (Kanno et al., 2003). Similar to the tulip, Alstroemeria and the lily (Order Liliales) retain two types of B-class genes (AlsDEFa/b and AlsGLO in Alstroemeria ligtu, LFDEF and LFGLOA/B in Lilium × formolongi); these are expressed in the outer three whorls (Akita et al., 2008; Hirai et al., 2007). Similarly, in Agapanthus praecox, two types of B-class genes (ApDEF and ApGLO) are expressed in the outer three whorls (Nakamura et al., 2005), which is consistent with the modified ABC model. Furthermore, the modified ABC model has been further extended to the modified ABCE model after the discovery of the E-function genes (Kanno, 2016; Soltis et al., 2007).
Recently, more direct evidence supporting the modified ABCE model has been reported in Tricyrtis species (Order Liliales) (Otani et al., 2016). Two DEF-like genes (TrihDEFa and TrihDEFb) and one GLO-like gene (TrihGLO) have been isolated from this plant. Expression analysis shows that these B-class genes are strongly expressed in the outer three whorls. Suppression of TrihDEFa and TrihDEFb gene expression using chimeric repressor gene-silencing technology (CRES-T) results in sepaloid tepals in whorls 1 and 2, and a pistil- or stigma-like structure in the place of stamens (Otani et al., 2016).
Orchids, which are also a member of the Asparagales, exhibit highly differentiated flowers with three outer tepals (sepals) in whorl 1, two lateral tepals (petals) and a single median tepal (lip) in whorl 2, and fused male and female reproductive organs (column) in whorls 3 and 4. Several genera in the orchid family Orchidaceae, including Dendrobium, Phalaenopsis, and Cymbidium, possess two layers of petaloid tepals. Studies in these genera show that both types of B-class genes (DEF-like and GLO-like genes) are expressed in the outer two whorls, again supporting the modified ABC model (Hsu et al., 2015; Tsai et al., 2004; Xu et al., 2006). A notable feature of Orchidaceae is the duplication of the DEF-like genes, resulting in four clades of genes (clades 1–4) (Mondragón-Palomino and Theißen, 2008, 2009; Mondragón-Palomino et al., 2009). Genes belonging to each clade show variations in expression patterns, which are closely linked to the development of sepals, petals, and lips. Based on this differential expression of paralogous DEF-like genes, the orchid code model has been proposed to explain the evolution of the orchid perianth (Mondragón-Palomino and Theißen, 2008, 2009; Mondragón-Palomino et al., 2009). Recently, further study on the orchid code has established the P-code model, showing the involvement of duplicated AGL6-like genes in the morphological diversification of the orchid perianth (Hsu et al., 2015).
Besides orchids, duplicated B-class genes have been isolated and analyzed in only a limited number of Asparagales species. In garden asparagus (Asparagus officinalis), one DEF-like (AODEF) and two GLO-like genes (AOGLOA and AOGLOB) have been isolated. All three of these B-class genes are expressed in stamens and inner petaloid tepals, but not in the outer petaloid tepals, which is inconsistent with the modified ABC model (Park et al., 2003, 2004). On the other hand, all B-class genes isolated from the crocus (Crocus sativus; CsatAPa, CsatAPb, CsatPIA, CsatPIB, and CsatPIC) are expressed in all four whorls (Kalivas et al., 2007; Tsaftaris et al., 2006). In Narcissus tazetta, two DEF-like genes (NtAP3L3 and NtAP3L4) are expressed in the inner petaloid tepals, but not in the outer petaloid tepals. Furthermore, NtAP3L3 is also expressed in stamens, whereas NtAP3L4 is expressed in the ovule (Li et al., 2013). Studies on these Asparagales species complicate the understanding of the conserved role of B-class genes and cast doubt on the generality of the modified ABC model. Clearly, additional studies on Asparagales species are necessary to explain the relationship between gene duplication and functional diversification of B-class genes in an evolutionary context.
In Asparagales, the grape hyacinth (Muscari spp.) is well known as an ornamental garden plant. Muscari armeniacum Leichtlin ex Baker bears two whorls of almost identical petaloid organs. We previously isolated one DEF-like gene (MaDEF) from M. armeniacum. The MaDEF gene is expressed in both the outer and inner petaloid tepals as well as in stamens, supporting the modified ABC model (Nakada et al., 2006). In this study, we report the isolation of two DEF-like and three GLO-like genes from M. armeniacum. In order to characterize these genes, the gene expression was analyzed by real-time PCR (RT-PCR), and protein–protein interactions were examined using yeast two-hybrid assays. In addition, the floral morphology of transgenic Arabidopsis thaliana plants overexpressing these genes was analyzed to infer their potential function in flower development.
M. armeniacum Leichtlin ex Baker cultivar ‘Blue Pearl’ (Sakata seed Co., Yokohama, Japan), which produces standard-type flowers, was used in this study. This plant is the same cultivar we used in a previous study (Nakada et al., 2006). Plants were grown in the experimental field at Tohoku University, Japan without forcing treatments (Fig. 1).
Muscari armeniacum flowers. (A) Mature inflorescence. (B) Mature flower. (C) Transection of a mature flower. (D) Immature inflorescence; white mark indicates the sample used for RNA extraction. (E) Immature flower. (F) Transection of an immature flower. (G) Dissected floral organs from an immature flower. w1, whorl 1 (outer tepals); w2, whorl 2 (inner tepals); w3, whorl 3 (stamens); w4, whorl 4 (pistil). Scale bars: 5 mm (A and D), 1 mm (B, C, E, F and G).
Total RNA was isolated from flower buds (~3 mm in diameter) and poly (A)+ RNA was used as a template for cDNA synthesis. Partial cDNAs were isolated via 3' rapid amplification of the cDNA ends (RACE) PCR, as described previously (Nakada et al., 2006), using four MADS-box degenerate primers (P038: 5'-GATCAAGMGSATCGAGAA-3', P041: 5'-GATGAAGMGSATCGAGAA-3', SP3: 5'-GACARGTCACKTTYTCKAAGC-3', AD: 5'-ARCTCACYGTSCTYTGYGAYGC-3') and a PCR anchor primer (P18M: 5'-GACTCGAGTCGACTACGA-3'). To determine the sequence of full-length cDNAs, 5'-RACE was performed with a 5'/3'-RACE Kit (Roche Diagnostics, Indianapolis, IN, USA) using flower bud cDNA pools as a template. Primers used in 5'-RACE are listed in Table S1. PCR products were cloned into pGEM-T Easy vector (Promega, Fitchburg, WI, USA), and sequenced using an ABI PRISM Dye Terminator Kit (Applied Biosystems, Carlsbad, CA, USA), according to the manufacturer’s instructions.
To construct a phylogenetic tree for B-class MADS-box proteins, amino acid sequences were downloaded from the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/), aligned using Clustal W (Thompson et al., 1994), and manually adjusted using SEAVIEW software (Gouy et al., 2009). Only the conserved MADS domain and K domain of the MIKC type MADS-box proteins were used for phylogenetic analysis. Accession numbers of amino acid sequences used for phylogenetic analysis are listed in Table S2. The phylogenetic tree was constructed using the maximum likelihood method under the LG model with the program PhyML implemented in SEAVIEW. A bootstrap consensus tree was inferred from 500 replicates (Felsenstein, 1985).
Total RNA was isolated from flower buds (~3 mm in diameter) approximately 10 days before flowering (Figs. 1D, E, F). To analyze the expression pattern of B-class genes, approximately 120 flower buds were divided into two biological replicates. Outer tepals, inner tepals, stamens and pistils in whorls 1–4 (w1–w4; Fig. 1G) were dissected, immediately placed in RNAlater solution (Sigma, St. Louis, MO, USA) and stored at 4°C to prevent RNA degradation. Total RNA was extracted from the floral tissues, and first-strand cDNA synthesis was performed using a ReverTra Ace qPCR RT Kit (Toyobo Co., Ltd., Osaka, Japan), according to the manufacturer’s instructions. PCR amplification was performed on the CFX 96 Real-time System (Bio-Rad, Hercules, CA, USA) using a KAPA SYBR FAST Universal qPCR Kit (Kapa Biosystems, Wilmington, MA, USA) and gene-specific primers (Table S3): RT-MaGLOA1Fw and RT-MaGLOA1Rv for MaGLOA1, RT-MaGLOA2Fw and RT-MaGLOA2Rv for MaGLOA2, RT-MaGLOBFw and RT-MaGLOBRv for MaGLOB, RT-MaDEF1Fw and RT-MaDEF1Rv for MaDEF1, RT-MaDEF2Fw and RT-MaDEF2Rv for MaDEF2, and RT-MaActinFw and RT-MaActinRv for ACTIN. PCR conditions were as follows: initial denaturation at 95°C for 3 min, followed by 40 cycles of denaturation at 95°C for 10 s, annealing at 50°C for 15 s and extension at 72°C for 15 s. The expression level of each gene was calculated using the delta Ct method, with ACTIN as the control gene (Livak and Schmittgen, 2001).
A yeast two-hybrid assay was performed using the GAL4 system (de Folter and Immink, 2011). To generate yeast two-hybrid constructs, full-length M. armeniacum cDNAs encoding the five B-class genes were amplified using gene-specific primers (Table S3) and cloned into the pDONR/Zeo vector. The cDNAs were then cloned into the prey vector (pDEST22) harboring the GAL4 transcription activation domain (AD) and tryptophan (Trp) selection marker gene using attL × attR recombination with the ProQuest Two-Hybrid System (Invitrogen, Waltham, MA, USA). All B-class genes were also cloned into the bait vector (pDEST32) harboring the GAL4 DNA-binding domain (BD) and leucine (Leu) selection marker gene. Genes in the pDEST22 and pDEST32 vectors were transformed into yeast (Saccharomyces cerevisiae) strains PJ69-4a and PJ69-4α, respectively, using the lithium-acetate method (de Folter and Immink, 2011). Haploid yeast strains PJ69-4a and PJ69-4α carrying plasmids were mated on a plate with SD medium (0.67% yeast nitrogen base without amino acids [Difco Laboratories, Detroit, MI, USA], 2% dextrose, and 1.7% agar) containing 13 chemical components (0.02% l-adenine [Ade] hemisulfate salt, 0.1% l-Leu, 0.02% l-Trp, 0.02% l-histidine [His] HCl monohydrate, 0.03% l-isoleucine [Ile], 0.15% l-valine [Val], 0.02% l-arginine [Arg] HCl, 0.03% l-lysine [Lys] HCl, 0.02% l-methionine [Met], 0.05% l-phenylalanine [Phe], 0.2% l-threonine [Thr], 0.03% l-tyrosine [Tyr], and 0.02% l-uracil [Ura]). After overnight incubation, spots were transferred to a plate of SD medium containing 11 chemical components (as described above, without l-leucine and l-tryptophan [–Leu/–Trp]) to select diploid yeast cells carrying pDEST22 and pDEST32 plasmids. Plates were incubated at 30°C for 2 days, and diploid yeast cells were transferred to two types of selection plates: selection plate type 1 (SD medium containing 10 chemical components as described above [–Leu/–Trp/–His] and 1 mM 3-amino-1,2,4-triazole [3-AT], which is a competitive inhibitor of the HIS3 gene product and is used to repress HIS3 gene autoactivation) and selection plate type 2 (SD medium containing 10 chemical components as described above [–Leu/–Trp/–Ade]), and the LacZ assay test plate (SD medium containing 11 chemical components [–Leu/–Trp]). These plates were incubated at 20°C for 5 days. LacZ activity was assayed using chloroform-treated permeabilized yeast with X-gal as a substrate (de Folter and Immink, 2011).
Full-length cDNAs of the five B-class genes of M. armeniacum were subcloned into the binary vector pBI-OX-GW (Inplanta Innovations, Yokohama, Japan) under the control of the cauliflower mosaic virus (CaMV) 35S promoter. The construct was introduced into wild-type A. thaliana Columbia (Col-0) via Agrobacterium-mediated transformation using the floral dip method (Clough and Bent, 1998). Seeds of T1 plants were sown on agar plates containing MS salt mixture and 50 μg·mL−1 kanamycin. Plates were transferred to a growth incubator maintained at 22°C, 14 h light/10 h dark photoperiod, and 100 μmol·m−2·s−1 white light for 20 days. Green plants with roots were transferred to soil for seed production. Subsequently, T2 seeds were selected for kanamycin resistance and tested for segregation. Kanamycin-resistant seeds of T2 plants were used for further experiments. A total of 5, 7, 7, 5, and 5 independent transgenic lines were generated for the 35S::MaGLOA1, 35S::MaGLOA2, 35S::MaGLOB, 35S::MaDEF1, and 35S::MaDEF2 constructs, respectively.
Previously, we isolated a single copy of a DEF-like gene, MaDEF, from M. armeniacum (Nakada et al., 2006). Two different B-class genes, DEFICIENS (DEF) and GLOBOSA (GLO), are necessary for the B-function in Antirrhinum majus because the encoded proteins function as a heterodimer (Saedler and Huijser, 1993). Here, we isolated additional B-class MADS-box genes from M. armeniacum using 3'-RACE PCR with MADS-box-specific degenerate primers. BLAST searches of the deduced amino acid sequences of these clones in public databases led to the identification of three GLO-like genes and two DEF-like genes. On the basis of sequence homology and phylogenetic analysis, one of the DEF-like genes was identical to MaDEF isolated previously (Nakada et al., 2006); we changed the name of MaDEF to MaDEF1, and named the other DEF-like gene MaDEF2. The three GLO-like genes identified in this study were designated as MaGLOA1, MaGLOA2, and MaGLOB (Figs. 2 and 3).
Alignment of deduced amino acid sequences of MaGLOs, MaDEFs and related B-class MADS-box proteins. (A) GLO-like MADS-box proteins of Muscari armeniacum (MaGLOA1, MaGLOA2 and MaGLOB), rice (Oryza sativa; OsMADS2 and OsMADS4) and Arabidopsis thaliana (PI). (B) DEF-like MADS-box proteins of M. armeniacum (MaDEF1 and MaDEF2), rice (OsMADS16), and Arabidopsis (AP3). Amino acid residues identical to MaGLOA1 (A) or MaDEF1 (B) are indicated as dots. The MADS domain, K domain, PI motif, PI-derived motif, and paleo AP3 motif were boxed.
Phylogenetic analysis of MaGLOs, MaDEFs and related B-class MADS-box proteins. The phylogenetic tree was constructed from the deduced amino acid sequence of the MADS domain and K domain using the maximum likelihood method. The MaGLOs and MaDEFs proteins are underlined. Groups are indicated to the right. The numbers at each node indicate the percentage of bootstrap values from 500 replications; bootstrap values < 50% are not shown.
Sequence analysis showed that the open reading frames (ORFs) of MaGLOA1, MaGLOA2, and MaGLOB cDNAs were 618, 621, and 630 bp, respectively, encoding deduced proteins of 206, 207, and 210 amino acids, respectively (Fig. 2A). The ORF of MaDEF2 cDNA was 663 bp, encoding a deduced protein of 221 amino acids (Fig. 2B). All genes showed a typical MIKC structure, comprising the MADS domain, an Intervening, a Keratin-like (K) domain, and a C-terminal domain. Amino acid sequences of the three GLO-like genes harbored a conserved PI motif in the C-terminal region (Fig. 2A). The two DEF-like genes harbored a PI-derived motif and a paleoAP3 motif in the C-terminal region (Fig. 2B).
To understand the evolution of B-class genes, amino acid sequences of the MADS domain and K domain were used to construct a phylogenetic tree with the maximum likelihood method (Fig. 3). Phylogenetic analysis showed that MaGLOA1 and MaGLOA2 were closely related to HPI1 and HPI2, the GLO-like genes of Hyacinthus, one of the closest genera within Asparagales (Angiosperm Phylogeny Group, 2016). Thus, it is likely that the duplication of these genes occurred relatively recently within the Asparagales lineage. However, it remains unclear whether the duplication of MaGLOA1 and MaGLOA2 occurred before or after the split between Hyacinthus and Muscari (Figs. 3 and S1). In contrast, MaGLOB grouped in a different clade from MaGLOA1 and MaGLOA2. Because genes from Asparagus and Crocus were found in both clades, the duplication event resulting in the split between MaGLOA1/MaGLOA2 and MaGLOB possibly occurred before the establishment of these three genera (Figs. 3 and S1). Similarly, the two DEF-like genes isolated from M. armeniacum, MaDEF1 and MaDEF2 also grouped in two distinct clades (Figs. 3 and S2). Because genes from Narcissus were found in both clades, duplication of the Muscari DEF-like genes probably occurred before the split between the two genera (Fig. 3).
We carried out northern blot analysis to examine the expression of B-class genes in the flower buds of M. armeniacum. Results showed that all B-class genes were specifically expressed in flower buds; no transcripts were detected in vegetative tissues such as roots, stems, bulbs, or leaves (data not shown). To analyze the expression patterns of isolated genes, real-time RT-PCR was performed using gene-specific primers (Table S3). Transcript levels of the DEF-like and GLO-like genes were normalized relative to ACTIN (Fig. 4). All three GLO-like genes were expressed throughout the four whorls, albeit to different levels (Fig. 4). MaGLOA1 was strongly expressed in the outer three whorls, whereas MaGLOA2 was weakly expressed in all four whorls. MaGLOB was mainly expressed in the stamens (whorl 3). Both MaDEF1 and MaDEF2 were expressed in the three outer whorls (Fig. 4), which was consistent with our RT-PCR results on MaDEF1 (Nakada et al., 2006). Although preliminary RT-PCR analysis showed MaDEF2 expression in all whorls (referred as an unpublished data in Kanno, 2016; Kanno et al., 2007), this was likely because of an excess of PCR cycles in the RT-PCR.
Quantitative expression analysis of MaGLOs and MaDEFs genes in the floral organs of M. armeniacum. The expression of MaGLOs and MaDEFs genes is normalized relative to that of the ACTIN gene. Data represent mean of two biological replicates, each comprising floral organs dissected from approximately 60 flowers. w1, whorl 1 (outer tepals); w2, whorl 2 (inner tepals); w3, whorl 3 (stamens); w4, whorl 4 (pistil); N.D., not detected.
Overall, expression profiles of MaGLOA1 and MaDEF1 were consistent with those of the typical B-class genes proposed in the modified ABC model (Kanno, 2016; Kanno et al., 2003; van Tunen et al., 1993). Among the remaining genes, MaGLOA2 and MaDEF2 seem to have supplemental or perhaps redundant functions, as the expression levels of these genes were low. On the other hand, MaGLOB was specifically expressed in the stamens. Together, these results imply a functional divergence through gene duplication, which has been repeatedly reported among MADS-box genes (Kramer et al., 1998, 2004; Zahn et al., 2005). For example, three paralogous AP3 lineages, termed AP3-I, AP3-II and AP3-III (Kramer et al., 2003), were found in Ranuncilaceae species and AP3-III showed petal-specific expression as well as having a specific role in petal identity (Sharma et al., 2011).
In A. thaliana and A. majus, the B-function is provided by the DEF-like–GLO-like heterodimer; this protein–protein interaction is one of the most common among MADS-box proteins across the angiosperms (Whipple et al., 2004), and constitutes the basis of both the ABC and modified ABC models. In orchids, a specific combination of the duplicated DEF-like and GLO-like proteins determines organ identity (Hsu et al., 2015). Moreover, homodimerization of DEF-like and GLO-like proteins has also been reported in some monocots. In Tulipa gesneriana, the GLO-like protein TGGLO alone binds to the CArG-box (consensus CC[A/T]6GG) as a homodimer (Kanno et al., 2003). Similarly, each of the two GLO-like proteins of L. regale, LRGLOA and LRGLOB, also form homodimers (Winter et al., 2002). In addition, the homodimer of LMADS1, a DEF-like protein from L. longiflorum, also binds to the CArG-box (Tzeng and Yang, 2001; Tzeng et al., 2004). Homodimerization of B-class proteins has also been reported in Phalaenopsis equestris (Order Asparagales) for both DEF-like PeMADS4 and GLO-like PeMADS6 (Tsai et al., 2008). Although the role of these B-class homodimers remains unknown, it is clear that protein–protein interactions underlie the fundamental properties of MADS-box genes.
Here, we performed yeast two-hybrid assays to test protein–protein interactions among the isolated B-class gene products. All interactions detected in the present study involved MaGLOA1 (Fig. 5). MaGLOA1 fused to a prey protein showed interactions with MaDEF1 or MaDEF2, whereas MaGLOA1 fused to a bait protein showed interactions with MaGLOA2 or MaDEF1. Thus, some interactions were not observed when bait and prey constructs were switched. Stable interactions were detected only between MaGLOA1 and MaDEF1. Unlike the tulip and Phalaenopsis orchid, as mentioned above (Kanno et al., 2003; Tsai et al., 2008), no homodimerization was detected in M. armeniacum. This is consistent with the result from Narcissus tazetta, in which the two DEF-like proteins (NAP3L3 and NAP3L4) did not form a homodimer (Li et al., 2013).
A yeast two-hybrid assay showing protein–protein interactions among the MaGLOs and MaDEFs proteins of M. armeniacum. Interactions were measured with the yeast two-hybrid assay by using survival on selection plates (A and B) and LacZ reporter gene activation (C). (A) –Leu, –Trp, and –His supplemented with 3-AT; (B)–Leu, –Trp, and –Ade.
To further understand the function of M. armeniacum B-class genes, 5, 7, 7, 5, and 5 independent transgenic Arabidopsis lines overexpressing 35S::MaGLOA1, 35S::MaGLOA2, 35S::MaGLOB, 35S::MaDEF1, and 35S::MaDEF2, respectively, were generated (Fig. 6). Compared with wild-type Arabidopsis (Fig. 6A), flowers of transgenic 35S::MaGLOA1 and 35S::MaGLOB plants showed partial conversion of sepals into petal-like structures (Fig. 6B, D). This sepal to petal conversion was found in two and four transgenic lines of 35S::MaGLOA1 and 35S::MaGLOB, respectively. This phenotype is similar to the flower phenotype of transgenic Arabidopsis plants overexpressing a GLO-type B-class gene, such as PISTILLATA (PI), ApGLO and MtPI from Arabidopsis, Agapanthus and Medicago, respectively (Krizek and Meyerowitz, 1996; Nakamura et al., 2005; Roque et al., 2016). Therefore, both MaGLOA1 and MaGLOB seem to retain, at least in part, the conserved function of the PI/GLO-like lineage.
Flowers of wild-type and transgenic Arabidopsis thaliana constitutively overexpressing MaGLOs and MaDEFs genes. (A) Wild-type flower. (B–F) Flowers of 35S::MaGLOA1 (B), 35S::MaGLOA2 (C), 35S::MaGLOB (D), 35S::MaDEF1 (E) and 35S::MaDEF2 (F) transgenic plants. Scale bars: 1 mm.
Of the five 35S::MaDEF1 transgenic lines, two lines showed altered phenotypes with an enlarged pistil (Fig. 6E). In general, wild-type A. thaliana plants overexpressing the intrinsic AP3 gene show a homeotic conversion of carpels into stamens (Jack et al., 1994). However, ap3 mutants, as well as the chimeric AP3 repressor, result in an aberrantly fused and enlarged pistil (Mitsuda et al., 2006). Thus, it is possible that the MaDEF1 protein interacts with PI, the intrinsic partner of AP3, giving rise to an inactive protein complex. The observed ap3 phenotype may have resulted from MaDEF1 acting as a dominant negative factor of AP3. The 35S::MaGLOA2 and 35S::MaDEF2 transgenic lines did not show any altered phenotypes compared with the wild type (Figs. 6C and F).
Taken together, MaDEF1 and MaGLOA1 demonstrated typical expression patterns (Fig. 4), protein–protein interactions (Fig. 5), and phenotypic alteration of A. thaliana flowers (Fig. 6), which is consistent with the functional DEF-like and GLO-like genes in monocots. Therefore, these two genes together seem to function as the major B-class genes in M. armeniacum, thus supporting the modified ABC model. Paralogs of these genes, MaDEF2 and MaGLOA2, showed much weaker expression in all four floral whorls (Fig. 4), suggesting a relatively less important role for these paralogs in flower development. However, MaDEF2 showed a partial protein–protein interaction with MaGLOA1 (Fig. 5). In addition, flowers of transgenic 35S::MaGLOA1 and 35S::MaGLOA2 A. thaliana lines were identical (Fig. 6). Therefore, MaDEF2 and MaGLOA1 may be redundant genes undergoing non- or perhaps neofunctionalization. In contrast, MaGLOB showed high expression specifically in the stamens (Fig. 4). Because the MaGLOB protein did not interact with any of the other DEF-like proteins (Fig. 5), it is possible that the MaGLOB gene has already acquired a novel function, independent of the modified ABC model. As a result, it would be interesting to investigate whether genes of MaGLOA1 and MaGLOB lineages retain a conserved function within each lineage (Fig. 3). Functional differentiation has not been reported among duplicated GLO-like genes of Asparagus (Park et al., 2004) or Crocus (Kalivas et al., 2007); only the expression levels have been investigated in these studies. More detailed comparisons are necessary to understand the functional divergence between the two lineages. Gene duplication and the fate of redundant genes have been central topics of discussion in evolutionary biology (Lynch and Conery, 2000; Lynch and Force, 2000; Ohno, 1970; Prince and Pickett, 2002). In this context, MADS-box genes provide an ideal opportunity to study the association between gene duplication and its biological outcomes (Airoldi and Davies, 2012). Although additional investigations are needed in various species, the results from the present study, as well as previous studies in Asparagales, provide valuable insights into gene evolution and the molecular mechanisms of floral development. This provides useful information that may be used to modify the floral architecture to create new flowers such as double flowers and viridiflora flowers in monocots.
We thank Drs. John C. Cushman and Yuichi Uno for providing the yeast strains, and Drs. Hideyuki Takahashi, Atsushi Higashitani and Jun Hidema for helpful discussions.