ATP-Binding Cassette G Transporters SGE1 and MtABCG13 Control Stigma Exsertion1[OPEN]

Butuo Zhu,a,2 Hui Li,a,b,2 Xiuzhi Xia,a,2 Yingying Meng,a Na Wang,a LuLu Li,a Jianxin Shi,c Yanxi Pei,b Min Lin,a Lifang Niu,a and Hao Lina,3,4 Biotechnology Research Institute, Chinese Academy of Agricultural Sciences, Beijing 100081, China College of Life Science, Shanxi University, Taiyuan, Shanxi 030006, China Joint International Research Laboratory of Metabolic and Developmental Sciences, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai 200240, China

Stigma exsertion is an important agricultural trait that facilitates the application of heterosis in crop breeding. Although several quantitative trait loci associated with stigma exsertion have been fine-mapped or cloned, the underlying genetic basis, particularly in legumes, remains unclear. In this study, we identified and characterized the exserted stigma mutant stigma exsertion1 (sge1) in the model legume Medicago truncatula. The exserted stigma phenotype of sge1 is mainly caused by physical interaction between floral organs, in which normal petal and stamen elongation are inhibited due to flower cuticle defects. SGE1 encodes an ATP-binding cassette G (ABCG) transporter that plays a critical role in regulating floral cutin and wax secretion in M. truncatula. SGE1 physically interacts with another half-size transporter, MtABCG13, to form a functional heterodimer. Mutation of MtABCG13 results in flower cuticle defects similar to those in sge1 as well as stigma exsertion, indicating that SGE1 and MtABCG13 are indispensable for flower cuticle secretion and collaboratively control stigma exsertion in M. truncatula. Our findings reveal novel functions for ABCG transporters in determining stigma exsertion by affecting the physical interactions of floral organs, providing insight into the molecular mechanism underlying stigma exsertion in leguminous plants with complex zygomorphic flowers.
The evolution of mating systems in flowering plants, which exhibit extraordinary diversity, is of central interest in plant biology. Self-fertilization can be adaptive when mates or pollinators are limited in nature, as originally proposed by Darwin (1876), because it confers reproductive assurance. However, this advantage can be offset by inbreeding depression, leading to reduced fitness of the offspring compared with outcrossed progeny (Darwin, 1876;Stebbins, 1974;Charlesworth and Charlesworth, 1987). As a consequence, most plants have evolved a wide variety of changes in floral morphology to facilitate or ensure outcrossing (Darwin, 1876;Stebbins, 1974;Barrett, 2003;Luo and Widmer, 2013).
Stigma exsertion is an efficient floral alteration that improves outcrossing pollination, making it an important agronomic trait that contributes to the application of heterosis to crop breeding (Kato and Namai, 1987;Karron et al., 1997;Richards, 1997;Motten and Stone, 2000). To date, several key regulators and quantitative trait loci (QTLs) that control stigma exsertion have been identified in diverse plant species. In tomato (Solanum lycopersicum), the transcription factor Style2.1 is responsible for long styles in wild tomatoes; a mutation in the Style2.1 promoter causes cultivated tomato to become autogamous and bear flowers with inserted stigmas . Characterization of the tomato procera mutant revealed novel functions for SlDELLA (a repressor in the gibberellin signaling pathway) in controlling stigma exsertion by affecting pistil elongation (Carrera et al., 2012). By contrast, stigma exsertion in rice (Oryza sativa) is a polygenic, complex quantitative trait that is strongly influenced by environmental factors Athwal, 1973, 1974;Chen and Tanksley, 2004;Chen et al., 2007). At least 38 QTLs affecting stigma exsertion have been identified in rice, which are distributed on all 12 chromosomes (Uga et al., 2003;Yamamoto et al., 2003;Miyata et al., 2007;Liu et al., 2015). Genome-wide association studies using 533 diverse accessions of rice identified 23 genomic loci that are significantly associated with stigma exsertion and indicated that GS3, GW5, and GW2 affect stigma exsertion by controlling the size and shape of the spikelet and stigma (Zhou et al., 2017). Despite these advances, the molecular basis of stigma exsertion across diverse species remains largely unexplored.
Legumes form one of the largest monophyletic plant families (Dong et al., 2005) and are second only to grasses in economic and nutritional value (Graham and Vance, 2003). The subfamily Papilionoideae, which is known for its peculiar arrangement of zygomorphic flowers (Tucker, 2003), is the largest of the three subfamilies in Leguminosae. Papilionoideae contains approximately 455 genera and 12,000 species (Tucker, 2003;Dong et al., 2005), including economically important crops such as soybean (Glycine max), pea (Pisum sativum), and the model species Lotus japonica and Medicago truncatula. Papilionoideae flowers are organized into four whorls: five sepals, five petals, 10 stamens, and a central carpel. The outermost whorl contains five sepals fused toward the base, forming a calyx tube with acute free lobes. Whorl 2 is pentamerous, with three different types of petals, including (from the adaxial to abaxial positions) one vexillum (standard petal), two lateral petals (alae, or wings), and two short petals (keel petals). The two keel petals are fused together along their adjacent abaxial edges and enclose the stamens and carpel. Whorl 3 is composed of a staminal tube, which is fused with nine stamen filaments around the carpel, and the tenth vexillary stamen filament at the adaxial position, which is free-standing. The most interior whorl, whorl 4, comprises a closed central carpel (Tucker, 2003).
Heterosis is routinely exploited in plant breeding, leading to improved biomass, yield, or resistance to biotic and abiotic stimuli in the hybrid progeny (Baranwal et al., 2012). However, legume breeders have not been able to take advantage of this genetic phenomenon in selfpollinating legume crops such as soybean. One of the limiting factors is that legumes possess complex fused floral architecture, in which stamens and pistils are covered by petals and the stigma is buried within anthers, greatly restricting cross-pollination and artificial pollination. Thus, isolating germplasm/mutants with stigma exsertion in legumes and uncovering the underlying molecular basis of this trait should facilitate the application of heterosis to leguminous crops, especially for papilionaceous species.
In this study, we identified and characterized the exserted stigma mutant stigma exsertion1 (sge1) in the model legume M. truncatula. The exserted stigma phenotype of the sge1 mutant is primarily caused by physical interaction between floral organs. SGE1 encodes an ABCG transporter, an ortholog of Arabidopsis ABCG11, that physically interacts with MtABCG13, collaboratively regulating floral cutin and wax secretion. Our results unveil novel functions for ABCG transporters in determining stigma exsertion by affecting the straight elongation of petals and stamens via regulating flower cuticle secretion in M. truncatula.

Identification and Characterization of the sge1-1 Mutant
In a screen of a collection of M. truncatula R108 Tnt1 insertional mutants (Tadege et al., 2008), we isolated a mutant with obviously exserted stigmas named sge1-1. In contrast to the wild-type flower, which has normally expanded petals and a stigma and anthers fully enclosed by petals from an early development stage (Fig. 1 Benlloch et al., 2003). We further explored the mutant phenotype by dissecting mature sge1-1 flowers. The petals of sge1-1 are wrinkled, and the compressed staminal tube fails to enclose the carpel, resulting in stigma exsertion that does not occur in the wild type (Fig. 1, I-K and M-O). The fused part of sge1-1 sepals is burst by massed petals inside the sepals at the late flower development stage (Supplemental Fig.  S1B). These observed morphological changes in floral organs resulted in spatial separation of the stigma and stamens in sge1-1 plants, which removed the physical contact between anthers and the stigma. Therefore, although the enclosed anthers bear fertile pollen (Fig. 1, L and P), mutant plants were completely sterile.
Intriguingly, the artificial application of pollen to the exserted stigmas of sge1-1 yielded normal pods with fertile seeds, whereas unpollinated flowers usually dropped after forming two to three pod spirals (Fig. 1, Q-S), indicating that the exserted stigmas of sge1-1 remain active. To identify the cause of stigma exsertion in sge1-1, we conducted histological analysis of flowers from wild-type and sge1-1 plants. Analysis of transverse cross sections showed that in sge1-1, floral organ primordia formed normally during the early stage of development (Benlloch et al., 2003), but as the floral bud grew, the petals and staminal tube exhibited abnormal curvature due to their irregular folding (Fig. 2, A-F). Analysis of longitudinal cross sections of floral buds at different stages (Benlloch et al., 2003) showed that, in contrast to the wild type, whose petals elongated smoothly through the space between the sepals and anthers (Fig. 2, G-K), the elongation of sge1-1 petals was blocked due to contact between the anthers and sepals, causing the petals to become wrinkled and develop into a mass (Fig. 2, L-P). The folded petals occupied most of the space within the sge1-1 flowers and restricted the normal elongation of the staminal tube, preventing it from covering the carpel, thereby resulting in stigma exsertion (Fig. 2,. In agreement with this notion, scanning electron microscopy (SEM) revealed that the epidermal cells of sge1-1 petals showed obvious traces of contact and morphological changes (Fig. 2Q).
To confirm this finding, we first examined whether the wrinkled petals and staminal tubes of sge1-1 flowers would be restored in open flower buds: in these buds, sepals would not form a tight covering, and therefore, the physical contact between petals and sepals or the staminal tube would not be strong. To this aim, we employed two previously reported mutants defective in petal expansion and floral organ fusion, stenofolia (stf) and loose flower (lfl), with mutations in the WOX family genes STF/MtWOX1 and LFL/MtWOX3, respectively, which are required for petal expansion and floral organ fusion in M. truncatula (Tadege et al., 2011;Niu et al., 2015). stf displays narrow petals (Tadege et al., 2011), while lfl shows a loose-flower phenotype due to fusion defects in the sepals and petals (Niu et al., 2015). The flowers of the lfl stf double mutant open completely, resulting in the exposure of the stigma and anthers from early flower development onward (Niu et al., 2015). We crossed sge1-1 with the heterozygous lfl 1/2 stf 1/2 mutant and analyzed the flower morphology of the resulting triple mutant. As expected, the wrinkling of the petals was suppressed and both the petals and the staminal tube elongated normally in the sge1-1 lfl stf triple mutant (Fig. 2R). These results confirmed that the wrinkled petals and compressed staminal tubes in sge1-1 flowers are caused by physical interaction between floral organs, resulting in stigma exsertion. As the cuticle plays an important role in floral organ development, enabling the uninhibited growth of petals as they extend between the sepals and anthers (Takeda et al., 2013), we next examined the cuticle integrity of floral organs using a well-developed Toluidine Blue test (Tanaka et al., 2004). Except for the fusion sites between the alae and keel, petals from wild-type plants exhibited little staining, whereas all three types of petals (vexillum, fused alae, and keel) from sge1-1 plants were deeply stained (Fig. 2S). Taken together, these results suggest that mutation of SGE1 affects the permeability of the cuticle and that the cuticle defects in petals and staminal tubes might lead to physical interaction between floral organs in sge1-1.

Cloning and Characterization of SGE1
We cloned SGE1 by PCR-based genotyping of flanking sequence tags (FSTs) in segregating populations (Tadege et al., 2008). Among the 56 progeny plants in a population produced by heterozygous parents, 15 were mutants, all of which were homozygous for FST 23. The ratio between mutant plants and wild-type-like plants was approximately 1:3 (x 2 5 0.095, P . 0.05), indicating that the sge1-1 mutant phenotype segregates as a single recessive Mendelian trait. We recovered the full-length sequence corresponding to this FST, representing the SGE1 gene (Medtr6g066240). SGE1 contains 10 exons, and the Tnt1 retrotransposon in the sge1-1 mutant was found to be inserted in the first exon ( Fig. 3A). Full-length SGE1 transcript was undetectable in the sge1-1 mutant in a reverse transcription (RT)-PCR assay (Fig. 3B). To confirm that the Tnt1 insertion in SGE1 is responsible for the mutant phenotype, we performed BLAST analysis of the FST database generated from the Tnt1 mutant population using SGE1 as a query (Sun et al., 2019) and identified an additional Tnt1 mutant that we named sge1-2 with a Tnt1 insertion at a different location in exon 1 of SGE1 (Fig. 3A). The sge1-2 homozygous mutant showed similar phenotypes to those of sge1-1 and undetectable transcription of fulllength SGE1 (Fig. 3B). Further genetic complementation confirmed the function of SGE1, in which a vector containing the 2.6-kb SGE1 promoter region and the SGE1 coding sequence (CDS) was introduced into sge1-1 plants by Agrobacterium tumefaciens-mediated transformation, which rescued the abnormal flower phenotype of sge1-1 (Fig. 3C). Collectively, these findings confirmed that disrupting SGE1 function results in stigma exsertion in sge1 mutants.
To further characterize the function of SGE1, RTquantitative PCR (qPCR) was used to examine the temporal and spatial expression patterns of SGE1. As shown in Figure 3D, SGE1 was expressed at high levels in flowers, moderate levels in leaves, stems, and pods, and low levels in roots. The high expression level of SGE1 in flowers supports its function in flower development in M. truncatula. To determine the subcellular localization of SGE1, the C terminus of SGE1 fused with the GFP gene under the control of the cauliflower mosaic virus 35S promoter was constructed, which, together with an mCherry-labeled plasma membrane localization marker pm-rk (CD3-1007; Nelson et al., 2007), was cotransferred into Nicotiana benthamiana leaf epidermal cells via an A. tumefaciens-mediated transfection.
We also fused SGE1 with RFP under the control of the 35S promoter, and the construct together with the plasma membrane localization marker pm-GFP were cotransferred into M. truncatula leaf protoplasts by the polyethylene glycol-mediated method. The signal of the SGE1-GFP or SGE1-RFP fusion protein was mainly detected in the plasma membrane in both N. benthamiana and M. truncatula (Fig. 3E), indicating a plasma membrane localization of SGE1 protein, which was consistent with a function in cuticle integrity.
Mutation of MtABCG13 Leads to a Similar Stigma Exsertion Phenotype to sge1 SGE1 encodes an ABC transporter sharing 82.95% amino acid sequence identity with Arabidopsis ABCG11 Values are means 6 SD of three biological replicates. E, Subcellular localization of the SGE1-GFP or SGE1-RFP protein in N. benthamiana leaf epidermal cells (top row) or M. truncatula leaf protoplasts (bottom row), respectively. pm-rk or pm-GFP was used as a plasma membrane localization marker. Bars 5 50 mm in N. benthamiana and 5 mm in M. truncatula. F, Phylogenetic analysis of SGE1 and its homologs in M. truncatula, Arabidopsis, alfalfa, and soybean. The SGE1 protein is marked with a red dot. The sequences were aligned using ClustalW, and a neighbor-joining phylogenetic tree was constructed using MEGA6 software. Numbers on branches indicate bootstrap percentages for 1,000 replicates. that is involved in the secretion of cuticular lipids (Bird et al., 2007;Luo et al., 2007;Panikashvili et al., 2007Panikashvili et al., , 2010Ukitsu et al., 2007). Protein sequence analysis revealed the presence of several characteristic ABC transporter domains near the N terminus of SGE1, including the Walker A and B motifs and an ABC signature constituting the nucleotide-binding domain, as well as a transmembrane domain containing six transmembrane a-helices near the C terminus (H1-H6; Supplemental Fig. S2). Phylogenetic analysis showed that SGE1 has homologs in several important legume crops, including alfalfa (Medicago sativa) and soybean, and clustered in a separate subclade with Arabidopsis ABCG transporters including AtABCG11, AtABCG12, AtABCG13, AtABCG15, and AtABCG3 (Fig. 3F). In addition to AtABCG11, both AtABCG12 and AtABCG13 are involved in cuticular lipid secretion (Pighin et al., 2004;Panikashvili et al., 2011;Takeda et al., 2014).
Given that the sge1 mutants showed cuticle defects and an exserted stigma phenotype, we wondered whether mutations in M. truncatula homologs of AtABCG12 and AtABCG13 would also result in a similar exserted stigma phenotype. By searching in the Medicago truncatula Genome Database (http://www.medicagogenome.org) for potential homologs of AtABCG12 and AtABCG13 using the AtABCG12 and AtABCG13 proteins as query sequences, we identified two putative ABCG proteins (Medtr4g076900 and Medtr4g076940) that were closely related to both AtABCG12 and AtABCG15 (which we named MtABCG12 and MtABCG15, respectively; of the mtabcg12 and mtabcg15 single mutants, nor the MtABCG15-CR/mtabcg12-1 double mutant, in which MtABCG15 was knocked out by CRISPR/Cas9 in the mtabcg12-1 background, was distinguishable from wildtype flowers (Supplemental Fig. S5, C-E). However, all three mtabcg13 mutant lines showed a stigma exsertion phenotype similar to that of sge1 (Fig. 4A), which was functionally verified in a pollination test (Fig. 4B). Moreover, mutations of both STF and LFL in the mtabcg13 mutant background completely restored the wrinkled petals and compressed staminal tube of the mutant (Fig. 4C). Furthermore, like sge1, the petals and staminal tube of mtabcg13 showed deeper Toluidine Blue staining compared with the wild type (Fig. 4D). Together, these findings indicate that MtABCG13 controls stigma exsertion by affecting the physical interaction of floral organs in a similar manner as SGE1.

SGE1 Physically Interacts with MtABCG13 to Form a Heterodimer
ABCG subfamily members contain both full-length transporters, like the pleiotropic drug resistance proteins, and half-transporters requiring dimerization to form a functional ABC transporter (Verrier et al., 2008;Kang et al., 2011;Do et al., 2018). Given that both SGE1 and MtABCG13 encode half-transporters and that the sge1 and mtabcg13 mutants exhibited similar stigma exsertion phenotypes, it is plausible to assume that SGE1 may form a functional heterodimer with MtABCG13. Like SGE1, MtABCG13 was highly expressed in flowers (Supplemental Fig. S6C). Transfection assays in both N. benthamiana leaves and M. truncatula leaf protoplasts confirmed that the MtABCG13 protein mainly localized to the plasma membrane (Supplemental Fig. S6D), indicating that both the spatial expression pattern and localization of MtABCG13 are associated with that of SGE1.
Further yeast two-hybrid assay showed that SGE1 interacts with MtABCG13 (Fig. 5A), and this interaction was verified in N. benthamiana leaves via bimolecular fluorescence complementation (BiFC) assays using split YFP. Strong yellow fluorescence was observed in cells cotransformed with constructs containing SGE1 fused to the N-terminal half of YFP (nYFP-SGE1) and MtABCG13 fused to the C-terminal half of YFP (MtABCG13-cYFP; Fig. 5B), indicating that SGE1 and MtABCG13 are capable of forming a heterodimer in planta. Notably, SGE1 could not form homodimers while MtABCG13 could, as tested by the yeast twohybrid and BiFC assays (Fig. 5, A and B). The heterodimerization between SGE1 and MtABCG13 and the homodimerization of MtABCG13 were further verified by luciferase complementation imaging (LCI) assays in Arabidopsis leaf protoplasts, in which SGE1 and MtABCG13 were fused to the N-terminal part of LUC (nLUC) to produce 35S:SGE1-nLUC and 35S:MtABCG13-nLUC and were fused to the C-terminal part of LUC (cLUC) to generate 35S:SGE1-cLUC and 35S:MtABCG13-cLUC, respectively. In contrast to the negative controls, where one-half of the split LUC (nLUC or cLUC) was fused to either SGE1 or MtABCG13 and the other half was used alone, significantly increased LUC activity was detected only in samples coexpressing 35S:SGE1-nLUC with 35S:MtABCG13-cLUC, 35S:MtABCG13-nLUC with 35S:SGE1-cLUC, or 35S:MtABCG13-nLUC with 35S:MtABCG13-cLUC (Fig. 5C), confirming that SGE1 and MtABCG13 form a heterodimer and that MtABCG13 forms homodimer.

SGE1 Associates with MtABCG13 to Collaboratively Regulate Flower Cutin and Wax Secretion
To further evaluate the interaction between SGE1 and MtABCG13, we generated the sge1 mtabcg13 double mutant by crossing the sge1-2 1/2 heterozygote with the mtabcg13-1 homozygote. In contrast to the wild type, the sge1-2 mtabcg13-1 double mutant had a similar functional and exserted stigma to that of either the sge1-2 or mtabcg13-1 single mutant ( Fig. 6A; Supplemental Fig. S7). Transmission electron microscopy (TEM) showed that, while wild-type petal cuticle appeared as darkly stained and evenly distributed electron-opaque ridges, that of sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant looked less stained with unevenly distributed or no ridges (Fig. 6B). These results suggest that both SGE1 and MtABCG13 are essential for flower cuticle secretion and that they collaboratively control stigma exsertion in M. truncatula.
To gain insight into the precise roles of SGE1 and MtABCG13 and to explain the observed cuticle integrity changes in mutants, we conducted comprehensive cutin and wax analyses of petals from wild-type, sge1-2, mtabcg13-1, and sge1-2 mtabcg13-1 plants using gas chromatography-mass spectrometry (GC-MS) and gas chromatography-flame ionization detection (GC-FID) analyses. The total wax and cutin loads in wild-type petals were 4.5 and 4.2 mg mg 21 dry mass of tissue, respectively (Fig. 6C). As compared with the wild type, the total wax loads in sge1-2 and mtabcg13-1 were reduced by 29% and 22%, respectively (Fig. 6C), and the total cutin loads in sge1-2 and mtabcg13-1 petals were reduced by 36% and 40%, respectively (Fig. 6C). The decline in the levels of almost all typical flower waxes, including the most abundant flower wax C29 alkanes, contributed to the significant reduction in wax load, while the reduction of all classes of classical cutin monomers, including the dominant C16 dihydroxy fatty acid, accounted for the remarkably reduced cutin loads (Fig. 7). These results are consistent with the observation that their cuticles were less electron opaque than the wild type (Fig. 6B). Notably, both total cutin and wax contents were further reduced in the petals of the sge1-2 mtabcg13-1 double mutant (Fig. 6C), suggesting that SGE1 and MtABCG13 function redundantly and show additive effects on cuticular lipid secretion in M. truncatula flowers.

DISCUSSION
Stigma exsertion is an important agricultural trait that facilitates the use of heterosis in crop breeding (Kato and Namai, 1987;Karron et al., 1997;Richards, 1997;Motten and Stone, 2000). Although significant progress has been made in characterizing stigma exsertion mutants and identifying the corresponding Figure 6. SGE1 associates with MtABCG13 to collaboratively regulate floral cutin and wax secretion. A, Flowers of the wild type (WT), sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant. Bars 5 1 mm. B, TEM comparison of vexilla from the wild type, sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant. Bars 5 1 mm. C, Total wax load and cutin load on the petal surface of the wild type, sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant. Data are mean values 6 SD of three independent experiments. Lowercase letters denote means that are significantly different as determined by one-way ANOVA and posthoc Tukey's test (P 5 0.05).
QTLs and regulatory genes Athwal, 1973, 1974;Uga et al., 2003;Yamamoto et al., 2003;Chen and Tanksley, 2004;Chen et al., 2007;Miyata et al., 2007;Carrera et al., 2012;Liu et al., 2015;Zhou et al., 2017), the molecular mechanisms underlying stigma exsertion, particularly in legumes with complex zygomorphic flowers, were largely unclear. In this study, we cloned genes encoding two ABCG transporters, SGE1 and MtABCG13, which are involved in flower cuticle secretion, and functionally characterized their roles in stigma exsertion in the model legume M. truncatula. Our study reveals that ABCG transporters have novel roles in determining stigma exsertion by influencing the physical contact between floral organs.

SGE1 Forms Heterodimers with MtABCG13 That Regulate Flower Cutin and Wax Transport
ABC transporters exist in a wide range of organisms in all kingdoms of life. These proteins transport a diverse array of molecules, including ions, nutrients, secondary metabolites, and lipids, across various cellular membranes (Kang et al., 2011;Do et al., 2018). SGE1 and MtABCG13, which we identified in M. truncatula based on their relatedness and identical phylogenetic profiles, are half-transporters of the ABCG subfamily that are closely related to Arabidopsis AtABCG11 and AtABCG13, respectively (Fig. 3F). AtABCG11 is involved in the transport of cuticular lipids in both vegetative and reproductive organs (Bird et al., 2007;Panikashvili et al., 2007Panikashvili et al., , 2010, and AtABCG13 is required for floral cutin deposition (Panikashvili et al., 2011). M. truncatula sge1 and mtabcg13 insertional knockout mutants displayed mutant phenotypes specific to floral organs and the flower cuticle, such as altered floral morphology, particularly with respect to petals (Figs. 1, A-K and M-O, 2, A-P, and 4A). Toluidine Blue staining, transmission electron microscopy, and chemical analyses supported the involvement of SGE1 and MtABCG13 in the transport of cuticular lipids, primarily cutin and wax monomers, in flowers (Figs. 2S, 4D, 6, B and C, and 7), suggesting that the role of ABCG11 and ABCG13 homologs in the transport of cuticle lipids is conserved among diverse plant species. In agreement with this notion, disrupting OsABCG26, an ortholog of AtABCG11, led to reduced levels of anther cutin monomers and wax components in rice (Zhao et al., 2015). Nevertheless, unlike those reported for the Arabidopsis abcg11 mutant (Bird et al., 2007;Panikashvili et al., 2007Panikashvili et al., , 2010, no obvious cuticle defects were observed in the vegetative organs of sge1, mtabcg13, or sge1 mtabcg13 mutants of M. truncatula (Supplemental Fig. S8). This result indicates that the contributions of SGE1 and MtABCG13 to the transport of cuticular lipids in vegetative organs are minor and redundant, even though SGE1 is expressed in M. truncatula leaves (Fig. 3D). However, our studies of SGE1 and MtABCG13 provided only phenotypic and chemical evidence for the involvement of these putative transporters in the deposition of cuticular components. Further examination of the transport activity of these proteins would elucidate the process in which the cuticular lipids are transported to the extracellular matrix.
The ABCG transporters dimerize to create a full-size transporter (Kang et al., 2011;Do et al., 2018). Given the characteristics of SGE1 and MtABCG13, we reasoned that they might function together. Consistent with this hypothesis, both SGE1 and MtABCG13 are localized to the plasma membrane, and yeast two-hybrid, BiFC, and LUC complementation assays confirmed that SGE1 is able to heterodimerize with MtABCG13 (Figs. 3E and 5; Supplemental Fig. S6D). Furthermore, consistent with the notion that the loss of either partner impairs heterodimer formation, disrupting either SGE1 or MtABCG13 led to similar exserted stigma phenotypes, along with reduced cutin and wax secretion (Figs. 1,  A-H, 4A, 6, B and C, and 7). A similar observation was previously made for the M. truncatula half-ABC transporters STUNTED ARBUSCULE (STR) and STR2, which interact to form heterodimers: the loss of either half-transporter resulted in stunted arbuscule development (Zhang et al., 2010). Even though MtABCG13 can form homodimers, disrupting SGE1 alone led to cuticle defects, suggesting that both SGE1 and MtABCG13 are indispensable for cuticular lipid transport in M. truncatula flowers. In agreement with this notion, the amounts of most detected cutin and wax monomers Figure 7. Biochemical analysis of surface lipids from petal extracts of the wild type (WT), sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant. A, Wax profile of petals from the wild type, sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant. Alk, Alkane; FA, fatty acid; OL, alcohol; UI, unidentified. B, Cutin profile of petals from the wild type, sge1-2, mtabcg13-1, and the sge1-2 mtabcg13-1 double mutant. DFA, Double fatty acids; DHFA, dihydroxy fatty acids; v-HFA, v-hydroxy fatty acids; 2HFA, 2-hydroxy fatty acids. Data are mean values 6 SD of three independent experiments. Lowercase letters denote means that are significantly different as determined by one-way ANOVA and posthoc Tukey's test (P 5 0.05).
were reduced in sge1 and mtabcg13 petals (Figs. 6C and 7), suggesting that SGE1 and MtABCG13 have broad substrate specificity for a variety of structurally diverse cuticular lipids. Moreover, compared with those of the sge1 and mtabcg13 single mutants, the total cutin and wax contents were further reduced in the petals of the sge1 mtabcg13 double mutant (Fig. 6C). Nevertheless, even though SGE1 and MtABCG13 form heterodimers, and MtABCG13 can homodimerize, the petals of sge1, mtabcg13, and sge1 mtabcg13 flowers showed only partial reductions in cutin and wax levels (Figs. 6, B and C, and 7), suggesting that SGE1 and MtABCG13 play redundant roles in flower cuticle secretion and may form heterodimers with other ABC transporters or proteins. Further identifying the interaction partners of SGE1 and MtABCG13 and demonstrating their substrate specificity, and their modes of action, will increase our understanding of the genetic network underlying this fundamental process.

SGE1 and MtABCG13 Control Stigma Exsertion by Affecting the Physical Interaction of Floral Organs
Stigma exsertion is a major factor increasing the opportunity for outcrossing pollination. Many studies have shown that stigma exsertion and stigma length are highly positively correlated Athwal, 1973, 1974;Chen et al., 2007;Carrera et al., 2012;Liu et al., 2015). In this study, we demonstrated that the ABCG transporters SGE1 and MtABCG13 regulate stigma exsertion in M. truncatula by affecting the physical contact of floral organs. Our results suggest that the defective flower cuticle in sge1 and mtabcg13, with reduced levels of cutin and wax monomers, prevents petals from elongating smoothly through the narrow space in floral buds. This causes the petals to become wrinkled and thus affects the normal elongation of staminal tubes, resulting in an exserted stigma phenotype (Fig. 2, A-P). The finding that ABCG transporters facilitate flower cuticle secretion and stigma exsertion by affecting the physical interaction of floral organs is quite interesting. The cuticle, which is composed of cutin and wax, is essential for proper development of the diverse surface structures on aerial plant organs (Shao et al., 2007;Mintz-Oron et al., 2008). It is thus not surprising that a defective cuticle would lead to physical interaction between floral organs. Indeed, floral organ friction leading to abnormal petal elongation has been reported in an Arabidopsis mutant of FOLDED PETALS1, encoding a member of the wax synthase/diacylglycerol acyltransferase family likely involved in the biosynthesis of wax-related products in the petal epidermis (Takeda et al., 2013). These findings are in agreement with the notion that the cuticle contains lubricating materials (e.g. wax or cutin) that enable the uninhibited growth of petals as they extend between sepals and anthers during flower development.
Notably, even though disrupting either SGE1 or MtABCG13 can lead to obvious stigma exsertion in M. truncatula, to date, no transporters in the ABCG subfamily have been implicated in stigma exsertion in other species, including Arabidopsis and rice, suggesting that this type of stigma exsertion is specific to leguminous plants, which possess specific fused floral architecture. In agreement with this hypothesis, the L. japonica mutant wrinkled petal and stamens1, whose petals show abnormal epidermal cell development, exhibits a similar stigma exsertion phenotype (Chen et al., 2006). Together, these findings suggest that the stigma exsertion mechanism caused by the physical contact of floral organs due to impaired flower cuticle secretion may be conserved among leguminous plants. Identifying and characterizing additional stigma exsertion mutants in other legumes would help confirm this notion.

CONCLUSION
In this study, we identified and characterized the exserted stigma mutant sge1 in the model legume M. truncatula. SGE1 encodes an ABCG transporter protein that plays a critical role in regulating floral cutin and wax secretion in M. truncatula. Our results demonstrate that SGE1 physically interacts with another half-size transporter, MtABCG13, to form a functional heterodimer in flower cuticle secretion and collaboratively control stigma exsertion in M. truncatula. Our findings unveil novel roles for ABCG transporters in determining stigma exsertion by influencing the physical contact between floral organs in the model legume M. truncatula, which may facilitate the use of heterosis for breeding leguminous crops with complex zygomorphic flowers.

Plant Materials and Growth Conditions
Medicago truncatula ecotype R108 was used for all experiments. Seeds of the Tnt1 insertion lines for SGE1 (NF2856 and NF3589), MtABCG13 (NF19675, NF14424, and NF16688), MtABCG12 (NF16587 and NF2014), and MtABCG15 (NF12478 and NF12855) were obtained from the Noble Research Institute. Primers used for genotyping are listed in Supplemental Table S1. Scarified M. truncatula ecotype R108 and mutant seeds were germinated overnight in petri dishes containing moist filter paper, incubated in darkness at 4°C for 7 d, and grown in a greenhouse at 24°C day/22°C night temperatures, a 16-h-day/ 8-h-night photoperiod, 150 to 200 mE m 22 s 21 light intensity (full-spectrum white fluorescent light bulbs), and 60% to 70% relative humidity.

Pollen Staining
Pollen was stained as previously described (Alexander, 1969). Mature wild-type and sge1 flowers were collected, fixed in Carnoy's fixative for 2 h, and stained with Alexander's solution for 2 h at room temperature. The samples were destained in 10% (v/v) glycerol for 45 min prior to observation.

Vector Construction and Plant Transformation
For complementation, the SGE1 promoter containing a 2.6-kb region upstream of the translation start codon was amplified by PCR and cloned into the EcoRI site of the pCAMBIA2300 vector. The SGE1 CDS was cloned into the SmaI site to generate the ProSGE1::SGE1 vector using an In-Fusion HD Cloning Kit (Clontech).
For CRISPR/Cas9 vector constructs, the vector used for the CRISPR/Cas9 assay (pFGC5941-Cas9) was described previously (Meng et al., 2017). The MtU6 promoter was amplified from R108 genomic DNA using the sense primer MtU6-F1 containing the In-Fusion reaction adaptor and the antisense primer MtU6-R1. The single guide RNA and scaffold fused fragment (sgRNA-scaffold) was obtained by amplifying the pU3-gRNA using the sense primer MtABCG15-sgRNA-F1 containing sgRNA sequence and the antisense primer MtABCG15-sgRNA-R1 containing the In-Fusion reaction adaptor. The MtU6p::gRNA-scaffold fragment was prepared by fusing the MtU6 promoter and sgRNA-scaffold fragments using the primer pairs MtU6-F1 and MtABCG15-sgRNA-R1, and then the pMtU6::sgRNA cassette was cloned into the linearized destination vector pFGC5941-Cas9 digested with XbaI using the In-Fusion cloning system to generate pFGC5941-Cas9-MtABCG15 vector (Clontech). The primers used are listed in Supplemental Table S1.
The ProSGE1::SGE1 and pFGC5941-Cas9-MtABCG15 constructs were introduced into Agrobacterium tumefaciens AGL1 by chemical transformation. A. tumefaciens-mediated transformation of M. truncatula was performed as described previously (Cosson et al., 2006). The primers used are listed in Supplemental Table S1.

Sequence Alignment and Phylogenetic Analysis
The full-length amino acid sequences of M. truncatula, Arabidopsis (Arabidopsis thaliana), alfalfa (Medicago sativa), and soybean (Glycine max) proteins were aligned using ClustalW (http://www.genome.jp/tools-bin/clustalw), and a neighbor-joining phylogenetic tree was constructed using MEGA6 software (Tamura et al., 2013). The most parsimonious tree with bootstrap percentages from 1,000 trials was used.

RNA Extraction and RT-qPCR Analysis
Total RNA was extracted from various M. truncatula tissues using TRIzol reagent (Invitrogen). Five micrograms of total RNA was used for cDNA synthesis with TransScript-Uni One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotech). RT-qPCR was performed using the Roche LightCycler 96 detection system with TransStart Tip Green qPCR SuperMix (TransGen Biotech). RT-qPCR data were obtained using three biological replicates, each with four technical replicates. The primers used are listed in Supplemental Table S1.

Subcellular Localization Analysis
The CDSs of SGE1 and MtABCG13 were amplified from R108, cloned into the pENTR/D-TOPO cloning vector (Invitrogen), and transferred into the pMDC83 vector using the Gateway LR reaction (Invitrogen) to generate the destination vectors. The correct constructs and pm-rk were introduced into A. tumefaciens strain GV2260 by chemical transformation, and the resulting agrobacteria were infiltrated into the leaves of 4-week-old Nicotiana benthamiana plants. P19 was used to inhibit transgenic silencing. After culturing the plants for 2 to 3 d, GFP and mCherry signals were visualized with a Zeiss LSM700 confocal laser-scanning microscope.
For M. truncatula protoplast transformation assay, the CDS of RFP was amplified from AHL22 (Wang et al., 2013) and inserted into the BstBI/EcoRI sites of pGreenII-0800-LUC (Liu et al., 2008) to produce the pGreenII-RFP vector, then the CDSs of SGE1 and MtABCG13 were cloned into the NheI site of pGreenII-RFP to generate SGE1-RFP and MtABCG13-RFP, respectively, using the In-Fusion cloning strategy (Clontech). The AtPIP2 gene was cloned from the pm-rk plasmid and inserted into the SalI/HindIII sites of the cauliflower mosaic virus 35S-GFP vector (Lin et al., 2009) to produce the plasma membrane marker pm-GFP. Mesophyll protoplasts were isolated from the young leaves of M. truncatula and transformed as previously described (Xiong et al., 2019). After overnight incubation in the dark, GFP and RFP signals were visualized with a Zeiss LSM700 confocal laser-scanning microscope. All primers used are listed in Supplemental Table S1.

Yeast Two-Hybrid Assay
For the yeast two-hybrid assay, the CDSs of SGE1 and MtABCG13 were cloned into in the pBT3-N vector (bait) and the pPR3-N vector (prey), respectively, using the In-Fusion cloning strategy (Clontech). Interaction was determined by cotransformation of the prey and bait vectors into the yeast strain NMY51, supplied by the DUAL hunter system according to the DUALmembrane starter kits user manual (Dualsystems Biotech). Cotransformed yeast clones were patched onto SD/-Trp/-Leu (double dropout) and SD/-Trp/-Leu/-His/-Ade (quadruple dropout) plates and grown at 28°C for 3 d. The primers used are listed in Supplemental Table S1.

BiFC Analysis
The vectors used for the BiFC assay (p2YN and p2YC) were described previously (Yang et al., 2007). The full-length CDSs of SGE1 and MtABCG13 were inserted into the PacI/SpeI sites of the p2YN and p2YC vectors to generate inframe vectors with the N terminus or C terminus of YFP using the In-Fusion system. The vectors were introduced into A. tumefaciens strain GV2260 by chemical transformation. Various combinations of vectors were coinfiltrated into N. benthamiana leaves. P19 was used to inhibit transgenic silencing. YFP signals were observed 1.5 to 2 d after infiltration with a Zeiss LSM700 confocal laserscanning microscope. The primers used are listed in Supplemental Table S1.

Histological Analysis
Dissected wild-type and mutant M. truncatula flowers were fixed and embedded as previously described (Lin et al., 2009). The tissues were sliced into 8-to 10-mm sections with a Leica RM2265 microtome, affixed to microscope slides, and stained with Toluidine Blue. Images were obtained with a digital camera mounted on an Olympus BX-51 compound microscope.

SEM
SEM analysis was performed as previously described (Zhu et al., 2018). Fresh M. truncatula floral organs from wild-type and mutant plants were fixed in 3% (v/v) glutaraldehyde in 25 mM phosphate buffer (pH 7) for 2 d, followed by 1% (w/v) osmium tetroxide in 25 mM phosphate buffer (pH 7) for 2 h, and dehydrated in a graded ethanol series. The desiccated tissues were critical-point dried in liquid CO 2 , mounted on aluminum stubs, and sputter coated with gold. The specimens were observed with a JSM-8404 microscope (S3400N; Hitachi).

Toluidine Blue Staining
Toluidine Blue staining was performed as previously described (Tanaka et al., 2004). Flowers (with the sepals removed) from wild-type and mutant plants were immersed in TB solution (0.05% [w/v] Toluidine Blue and 0.01% [v/v] Tween 20) for 5 min, washed in water, and photographed. At least 15 mutant and wild-type flowers were examined.

TEM
The top 3 mm of vexilla from wild-type and mutant flowers was fixed in 4% (v/v) paraformaldehyde and 2% (v/v) glutaraldehyde in 0.1 M potassium phosphate buffer (pH 7.4) for 2 h at room temperature and postfixed in 1% (w/v) osmium in 0.1 M potassium phosphate buffer (pH 7.4) for 1 h at 4°C. Following serial dehydration in ethanol at room temperature, the material was embedded in Spurr's standard epoxy resin and polymerized at 62°C overnight. Thin sections (Leica EM UC7) were stained with 2% (w/v) uranyl acetate (in 50% [v/v] acetone) and 80 mM alkaline lead citrate, each for 30 min. The specimens were observed with a microscope (HT7700; Hitachi).

Analysis of Wax and Cutin
Wax and cutin in the samples were analyzed as previously described with minor modifications (Panikashvili et al., 2011;Zhu et al., 2013). Fresh petals (including vexillum, alae, and keel) from mature wild-type and mutant flowers were collected and extracted with 8 mL of chloroform for 1 min at room temperature. Each extraction was repeated three times, and each contained about 100 mg of fresh materials of petals from five to 10 individual plants. The chloroform extract solution was spiked with 10 mg of tetracosane (Fluka, Ronkonkoma) as an internal standard and evaporated under nitrogen gas to a final volume of 100 mL. After the addition of 20 mL of N,N-bis-trimethylsilyltrifluoroacitamide (Macherey-Nagel) and 20 mL of pyridine to each sample, the extracts were incubated for 40 min at 70°C. The derivatized samples were then analyzed by GC-FID (Agilent Technologies) and GC-MS (Agilent gas chromatograph coupled to an Agilent 5973N quadrupole mass selective detector; Zhu et al., 2013). For cutin analysis, petal samples that had been used in the wax extraction were exhaustedly delipidated in a 10-mL mixture of methanol:chloroform (1:1) with gentle shaking for 2 weeks (solvent changed daily), dried, weighed, and used for analysis. The dry residue (4-18 mg) was depolymerized by transesterification in 2 mL of 1 N methanolic HCl (1 M HCl in methanol) for 2 h at 80°C. After stopping the reaction by adding 2 mL of saturated NaCl, 20 mg of dotriacontane (Fluka) was added as an internal standard, and the cutin monomers were extracted with 1 mL of hexane three times. The three hexane phases were combined, evaporated, and derivatized as described above and then analyzed with GC-MS and GS-FID as for the wax analysis .

LCI Assay
For LCI assays in Arabidopsis protoplasts, the full-length SGE1 and MtABCG13 sequences were cloned into the 35S:nLuc and 35S:cLuc vectors, respectively (Chen et al., 2008). A plasmid expressing the Renilla Luciferase (REN) gene was used as the internal control. Mesophyll protoplasts were isolated from 2-week-old plants grown in one-half strength Murashige and Skoog medium (PhytoTech) under long-day conditions (16 h of light/8 h of dark). LUC and REN activities were separately determined 16 h posttransformation using the Dual-Luciferase Reporter Assay System (Promega). The primers used are listed in Supplemental Table S1.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S2. Amino acid sequence alignment of SGE1 and AtABCG11.
Supplemental Figure S4. Amino acid sequence alignment of AtABCG13 and MtABCG13.
Supplemental Figure S6. Mutant identification, expression pattern, and protein subcellular localization of MtABCG13.
Supplemental Table S1. Primers used in this study.