RNA-Binding Protein MAC5A Is Required for Gibberellin-Regulated Stamen Development

The development of floral organs is coordinated by an elaborate network of homeotic genes, and gibberellin (GA) signaling is involved in floral organ development; however, the underlying molecular mechanisms remain elusive. In the present study, we found that MOS4-ASSOCIATED COMPLEX 5A (MAC5A), which is a protein containing an RNA-binding motif, was involved in the development of sepals, petals, and stamens; either the loss or gain of MAC5A function resulted in stamen malformation and a reduced seed set. The exogenous application of GA considerably exacerbated the defects in mac5a null mutants, including fewer stamens and male sterility. MAC5A was predominantly expressed in pollen grains and stamens, and overexpression of MAC5A affected the expression of homeotic genes such as APETALA1 (AP1), AP2, and AGAMOUS (AG). MAC5A may interact with RABBIT EARS (RBE), a repressor of AG expression in Arabidopsis flowers. The petal defect in rbe null mutants was at least partly rescued in mac5a rbe double mutants. These findings suggest that MAC5A is a novel factor that is required for the normal development of stamens and depends on the GA signaling pathway.


Introduction
Gibberellins (GAs) are a family of tetracyclic diterpenoid phytohormones that stimulate plant growth and development, including flower development and inflorescence meristem size [1][2][3][4][5][6]. Flowers are frequently composed of four different floral organs: sepals, petals, stamens, and carpels; and the A, B, and C class genes of the ABC model are critical for floral organ identity [7,8]. The identity of sepals in the outer whorl is determined by class A genes APETALA1 (AP1) and AP2, while that of petals in the second whorl is determined by the combined activity of class A and B genes AP3 and PISTILLATA (PI). Stamens in the third whorl are determined by the combined effects of class B genes and the class C gene AGAMOUS (AG), which is also solely responsible for the determination of carpels in the innermost whorl. The anthers, which develop in the third whorl, are suggested to be a major site of bioactive GA synthesis during flower development in several plant species [9][10][11]. GA-deficient ga1-3 mutants develop flowers with retarded growth of all floral organs despite their normal identities, especially impaired another development, which results in male sterility owing to a lack of mature pollen [2,[12][13][14]. Short anthers Int. J. Mol. Sci. 2022, 23,2009 2 of 13 and complete male sterility in ga1-3 mutants can be restored by exogenous GA application, which upregulates the expression of AP3, PI, and AG [2,15,16]. The overexpression of AG partially rescues floral defects in young flowers of ga1-3 mutants [3].
GA promotes floral organ development in Arabidopsis thaliana by blocking the functions of DELLA proteins [2]. DELLA proteins have an N-terminal with two conserved domains: the DELLA and TVHYNP domains. After the binding of bioactive GA, the C3-hydroxyl group of the GA molecule becomes hydrogen-bound to the GA receptor GIBBERELLIN INSENSITIVE DWARF1 (GID1), inducing a conformational change of GID1 in the Nterminal extension to cover the GA pocket [17,18]. Once the pocket is closed, the upper surface of the lid binds to the DELLA and TVHYNP domains of the DELLA protein to form a GA-GID1-DELLA complex [12,19,20]. Subsequently, the GA-GID1-DELLA complex is recognized by the SCFSLY1/GID2 ubiquitin E3 ligase complex for polyubiquitination, and the DELLA protein is then degraded by the 26S proteasome [21,22]. A reduction in GA concentration promotes the accumulation of DELLA proteins that repress GA responses, whereas an increase in GA results in the degradation of DELLA proteins and the activation of GA responses. Five DELLA genes (GAI, RGA, RGL1, RGL2, and RGL3) are present in the Arabidopsis genome. The loss of function of RGA and RGL2 is almost sufficient to compensate for floral organ deficiency in ga1-3 mutants, which indicates that both RGA and RGL2 are major repressors of GA response in this process [21].
MOS4-ASSOCIATED COMPLEX 5A (MAC5A) is involved in the stress response, cell division, and development of Arabidopsis. In yeast and humans, MAC5A homologs are components of the NINETEEN COMPLEX (NTC) or Prp19 Complex (Prp19C) [23]. MAC/NTC/Prp19 is a large flexible RNA-protein complex associated with the spliceosome that permits alternative splicing and generates multiple transcript variants from a single gene, and it is required for development and immunity in all three organisms. The core components of MAC include CELL DIVISION CYCLE5 (CDC5/MAC1), WD-40 protein PLEIOTROPIC REGULATORYLOCUS1 (PRL1/MAC2), MAC3A and MAC3B, and MOS4 [23]. MOS4 encodes a nuclear protein that binds directly to CDC5, an atypical R2R3 Myb transcription factor required for optimized Pol II activity at the promoters of miRNA genes [24]. During transcription, PRL1 binds to pri-miRNAs to prevent their degradation [25]. MAC3A and MAC3B interact physically and genetically with the Skiinteracting protein and mediate the alternative splicing of 50% of the expressed genes in Arabidopsis [26]. mac5a null mutants exhibit early flowering and have serrated leaves, short roots, and reduced seed sets [27]. The loss of function of MAC5A and its close homolog MAC5B is lethal, and mac5a homozygous plants with a mac5b heterozygous background exhibit defects in reproductive organogenesis [27]. In this study, we found that the overexpression of MAC5A and loss of MAC5A function affected stamen development. Furthermore, MAC5A predominantly expressed in anthers and regulated anther development via a GA-dependent pathway.

Loss of MAC5A Function Affects Floral Organ Development in Arabidopsis
Salk_132281 (mac5a) carries a T-DNA insertion in the second exon of MAC5A and is a genuine mac5a null mutant, as previously reported [27]. In the present study, we showed that the petals of mac5a single mutants were narrower than those of the Columbia ecotype (Col-0) of Arabidopsis ( Figure 1A,B,D). A normal Arabidopsis flower has two short and four long stamens, all of which develop in the third whorl ( Figure 1B). mac5a had a reduced number of stamens ( Figure 1B-D); 42% of mac5a flowers did not have any short stamens and produced four long stamens only, whereas 36% had five long stamens ( Figure 1B,C). The outer epidermal cells of sepals exhibit a characteristic pattern with diverse sizes, ranging from giant cells of an average length of 360 µm to smallest cells only approximately 10 µm in length [28]. In this study, the Col-0 sepals developed a group of tightly arranged small cells to form a smooth edge ( Figure 1E). In mac5a, the number of small cells was reduced, and giant cells frequently developed on the edge of the sepals, forming a jagged tip ( Figure 1E).

GA Application Enhances the Deficient of Stamen Development in mac5a Mutant
GA treatment exacerbated the abnormalities in the floral organs of mac5a null mutants, whereas PAC treatment partially alleviated them ( Figure 2). Jagging of the sepal tip in mac5a mutants seemed to be aggravated by GA but was not recovered by PAC ( Figure  2A). Following GA application, the number of stamens decreased ( Figure 2B), and anther development was retarded in mac5a, resulting in a lack of mature pollen ( Figure 2C). Pollen viability was assessed by Alexander cytoplasmic staining. In the presence of 100 μM GA3, around 70% of anthers in Col-0 developed viable pollen ( Figure 2D,E). In contrast, pollen development was consistently aborted or invisible in mac5a anthers in response to GA treatment ( Figure 2D,E), leading to sterility ( Figure 2F). After PAC treatment, the number of stamens in mac5a increased and was comparable to that in Col-0 ( Figure 2B).

GA Application Enhances the Deficient of Stamen Development in mac5a Mutant
GA treatment exacerbated the abnormalities in the floral organs of mac5a null mutants, whereas PAC treatment partially alleviated them ( Figure 2). Jagging of the sepal tip in mac5a mutants seemed to be aggravated by GA but was not recovered by PAC (Figure 2A). Following GA application, the number of stamens decreased ( Figure 2B), and anther development was retarded in mac5a, resulting in a lack of mature pollen ( Figure 2C). Pollen viability was assessed by Alexander cytoplasmic staining. In the presence of 100 µM GA 3 , around 70% of anthers in Col-0 developed viable pollen ( Figure 2D,E). In contrast, pollen development was consistently aborted or invisible in mac5a anthers in response to GA treatment ( Figure 2D,E), leading to sterility ( Figure 2F). After PAC treatment, the number of stamens in mac5a increased and was comparable to that in Col-0 ( Figure 2B).

Overexpression of MAC5A Affects Stamen Development in Arabidopsis
To gain further insights into the physiological role of MAC5A, a full-length MAC5A cDNA was fused in-frame with the GFP gene at the 3′-terminus and driven by the constitutive CaMV35S to overexpress the fusion gene MAC5A-GFP in Arabidopsis (ox lines). Eleven independent MAC5A-overexpressing lines were identified. The T2 populations of White arrows indicate shrunken underdeveloped seeds. Scale bar = 500 µm.

Overexpression of MAC5A Affects Stamen Development in Arabidopsis
To gain further insights into the physiological role of MAC5A, a full-length MAC5A cDNA was fused in-frame with the GFP gene at the 3 -terminus and driven by the constitutive CaMV35S to overexpress the fusion gene MAC5A-GFP in Arabidopsis (ox lines). Eleven independent MAC5A-overexpressing lines were identified. The T 2 populations of ox2, ox4, ox8, and ox11 segregated abnormal phenotypes, such as compacted inflorescence and abnormal floral organs, during the reproductive stage ( Figure 3A-E). The ox2 plants showed the most severe developmental defects and were used for further analysis. In ox2 plants, 52% of flowers (n = 180) developed six normal stamens at the correct position ( Figure 3G), but the rest exhibited various defects in stamen development ( Figure 3H-M). These defects included flowers with five stamens (18.9%, Figure 3H), more than four short stamens (11.1%, Figure 3I), and undeveloped or malformed anthers such as a part of the anther transformed into a petal-like structure (3.88%, Figure 3J-L). ox2 plants also developed flowers with retarded stamens and petals (13.33%, Figure 3M) and consistently produced significantly swollen or shrunken pollen (arrows in Figure 3 N,O). ox2 plants rarely produced seeds after self-pollination ( Figure 3P). ox2, ox4, ox8, and ox11 segregated abnormal phenotypes, such as compacted inflorescence and abnormal floral organs, during the reproductive stage ( Figure 3A-E). The ox2 plants showed the most severe developmental defects and were used for further analysis. In ox2 plants, 52% of flowers (n = 180) developed six normal stamens at the correct position ( Figure 3G), but the rest exhibited various defects in stamen development ( Figure  3H-M). These defects included flowers with five stamens (18.9%, Figure 3H), more than four short stamens (11.1%, Figure 3I), and undeveloped or malformed anthers such as a part of the anther transformed into a petal-like structure (3.88%, Figure 3J-L). ox2 plants also developed flowers with retarded stamens and petals (13.33%, Figure 3M) and consistently produced significantly swollen or shrunken pollen (arrows in Figure 3 N,O). ox2 plants rarely produced seeds after self-pollination ( Figure 3P).  Floral organ identity is mainly defined by ABC class genes, such as AP1, AP2, AP3, PI, and AG. In the inflorescences of mac5a and MAC5A-overexpressing plants, the expression of AP1, AP2, and AG changed at different levels ( Figure 4A). PI, and AG. In the inflorescences of mac5a and MAC5A-overexpressing plants, the expression of AP1, AP2, and AG changed at different levels ( Figure 4A).
To gain insight the role of MAC5A in the GA pathway, we examined the expression levels of genes involved in GA biosynthesis and signaling pathways in inflorescences of mac5a mutants and MAC5A-overexpressing plants. The expression levels of ENT-KAU-RENOIC ACID HYDROXYLASE 2 (KAO2), ENT-KAURENE OXIDASE (GA3) and GID1B, RGA, and RGL2 were higher in mac5a mutants and MAC5A-overexpressing plants ( Figure  4B). In response to GA application, the dwarf phenotype of plants overexpressing MAC5A was partially rescued; however, their fertility was not elevated by GA inhibitor PAC treatment (Supplementary Figure S1).

MAC5A Is Highly Expressed in the Stamen and Cooperates with RABBIT EARS (RBE) to Regulated Floral Organ Development
To examine the spatiotemporal expression pattern of MAC5A, a 2-kb genomic sequence of MAC5A native promoter was fused with the β-glucuronidase (GUS) reporter gene (PMAC5A:GUS). In the reproductive phase, the PMAC5A:GUS signal was low in the cauline leaf ( Figure 5A), but high in the anther ( Figure 5B,C,F) and the carpel just after pollination ( Figure 5E,G,H). In siliques, the PMAC5A:GUS signal was highly detected in developing seeds ( Figure 5I), whereas it disappeared in mature seeds ( Figure 5J). To gain insight the role of MAC5A in the GA pathway, we examined the expression levels of genes involved in GA biosynthesis and signaling pathways in inflorescences of mac5a mutants and MAC5A-overexpressing plants. The expression levels of ENT-KAURENOIC ACID HYDROXYLASE 2 (KAO2), ENT-KAURENE OXIDASE (GA3) and GID1B, RGA, and RGL2 were higher in mac5a mutants and MAC5A-overexpressing plants ( Figure 4B). In response to GA application, the dwarf phenotype of plants overexpressing MAC5A was partially rescued; however, their fertility was not elevated by GA inhibitor PAC treatment (Supplementary Figure S1).

MAC5A Is Highly Expressed in the Stamen and Cooperates with RABBIT EARS (RBE) to Regulated Floral Organ Development
To examine the spatiotemporal expression pattern of MAC5A, a 2-kb genomic sequence of MAC5A native promoter was fused with the β-glucuronidase (GUS) reporter gene (P MAC5A :GUS). In the reproductive phase, the P MAC5A :GUS signal was low in the cauline leaf ( Figure 5A), but high in the anther ( Figure 5B,C,F) and the carpel just after pollination ( Figure 5E,G,H). In siliques, the P MAC5A :GUS signal was highly detected in developing seeds ( Figure 5I), whereas it disappeared in mature seeds ( Figure 5J). Using yeast two-hybrid (Y2H) screening, we found that MAC5A may be a potential interacting protein of RBE, which is a repressor of AG expression in Arabidopsis flowers [29][30][31]. Subsequently, their interaction was validated using a Y2H assay ( Figure 6A). A bimolecular fluorescence complementation (BiFC) experiment was performed to further confirm their interaction ( Figure 6B). The second whorl petals in rbe mutants are missing or replaced with filaments ( Figure 6C) [29,30]. We constructed a mac5a rbe double mutants to detect the genetic interaction of MAC5A and RBE. The petal defect in rbe was at least partly rescued in the mac5a rbe double mutants ( Figure 6C), suggesting that MAC5A is responsible for the second-whorl defect in rbe mutants. We found that the pollen grains in both rbe and mac5a rbe displayed an irregular and collapsed shape ( Figure 6D), similar to those in the mac5a single mutant ( Figure 1F). Using yeast two-hybrid (Y2H) screening, we found that MAC5A may be a potential interacting protein of RBE, which is a repressor of AG expression in Arabidopsis flowers [29][30][31]. Subsequently, their interaction was validated using a Y2H assay ( Figure 6A). A bimolecular fluorescence complementation (BiFC) experiment was performed to further confirm their interaction ( Figure 6B). The second whorl petals in rbe mutants are missing or replaced with filaments ( Figure 6C) [29,30]. We constructed a mac5a rbe double mutants to detect the genetic interaction of MAC5A and RBE. The petal defect in rbe was at least partly rescued in the mac5a rbe double mutants ( Figure 6C), suggesting that MAC5A is responsible for the second-whorl defect in rbe mutants. We found that the pollen grains in both rbe and mac5a rbe displayed an irregular and collapsed shape ( Figure 6D), similar to those in the mac5a single mutant ( Figure 1F).

Discussion
In this study, we unveiled the molecular role of MAC5A in floral organ development in Arabidopsis. Both the loss and gain of the MAC5A function frequently resulted in the malformation of floral organs, especially the stamens (Figures 1 and 3), which indicates that an optimum expression of MAC5A is critical for the normal development of floral organs. Arabidopsis flowers consist of four whorls of organs: sepals, petals, stamens, and carpels; and their identity is defined by three classes of homeotic genes: A, B, and C. The class A genes AP1 and AP2 specify the identity of sepals; class A and B genes AP3 and PI specify the identity of petals; class B and C genes specify the identity of stamens; and the class C gene AG specifies the identity of carpels. In the absence of the class C gene, the class A genes are active throughout the floral meristem. AG is required to repress the activity of class A genes in the third and fourth whorls [8,32]. The loss and gain of the MAC5A function affect the expression of the class A and C genes ( Figure 4A), and MAC5A promoter activity was identified in the stamens ( Figure 5). These results suggest that MAC5A may be involved in floral organ development by regulating class A and C genes.
RBE, which encodes a C2H2 zinc finger transcriptional factor, is specifically expressed in the petal primordia and acts as a second-whorl repressor of AG to limit its expression in flow-

Discussion
In this study, we unveiled the molecular role of MAC5A in floral organ development in Arabidopsis. Both the loss and gain of the MAC5A function frequently resulted in the malformation of floral organs, especially the stamens (Figures 1 and 3), which indicates that an optimum expression of MAC5A is critical for the normal development of floral organs. Arabidopsis flowers consist of four whorls of organs: sepals, petals, stamens, and carpels; and their identity is defined by three classes of homeotic genes: A, B, and C. The class A genes AP1 and AP2 specify the identity of sepals; class A and B genes AP3 and PI specify the identity of petals; class B and C genes specify the identity of stamens; and the class C gene AG specifies the identity of carpels. In the absence of the class C gene, the class A genes are active throughout the floral meristem. AG is required to repress the activity of class A genes in the third and fourth whorls [8,32]. The loss and gain of the MAC5A function affect the expression of the class A and C genes ( Figure 4A), and MAC5A promoter activity was identified in the stamens ( Figure 5). These results suggest that MAC5A may be involved in floral organ development by regulating class A and C genes.
RBE, which encodes a C2H2 zinc finger transcriptional factor, is specifically expressed in the petal primordia and acts as a second-whorl repressor of AG to limit its expression in flowers. The second whorl petals in rbe mutants are missing or replaced with filaments due to misexpression of AG in the second whorl [29,30]. RBE directly interacts with the promoter of microRNA164 (miRNA164) and negatively regulates its expression [31]. MAC5A is an RNA-binding protein required for microRNA (miRNA) biogenesis [33]. Indeed, miRNA164 expression is decreased in the mac5a mutants [33]. We found that MAC5A may be a potential interacting protein of RBE, and the petal defect of rbe was at least partly rescued in the mac5a rbe double mutants (Figure 6), suggesting that MAC5A is responsible for the rbe second-whorl defects. In Arabidopsis flowers, miRNA164 genes regulate the expression of CUP-SHAPED COTYLEDON1 (CUC1) and CUC2, which encode critical transcriptional regulators involved in organ boundary specification [34,35]. RBE coordinately regulates the expression of miRNA164 genes to impact the expression of CUC1 and CUC2, which in turn regulate the events required for sepal and petal development [31]. On the other hand, the abundance of miRNA172 is also reduced in mac5a mutants [33]. miRNA172 mediates the repression of AP2, which is critical for stamen and carpel development [36,37]. We propose that MAC5A and RBE work together to regulate the class A and C genes through an miRNA-dependent pathway.
GAs play an essential role in plant growth, floral development, and late embryogenesis [5,38,39]. GA promotes floral organ development in Arabidopsis by blocking the functions of DELLA proteins [2]. After the binding of bioactive GA, the GA receptor GID1 binds to the DELLA protein to form a GA-GID1-DELLA complex [12,19,20]. Subsequently, the GA-GID1-DELLA complex is recognized by the SCFSLY1/GID2 ubiquitin E3 ligase complex for polyubiquitination, and the DELLA protein is then degraded [21,22]. Reduction in GA concentration promotes the accumulation of DELLA proteins that repress GA responses, whereas an increase in GA results in the degradation of DELLA proteins and activation of GA responses. In the inflorescence of mac5a mutants and MAC5A-overexpressing plants, the upregulated expression of GA biosynthesis and signaling genes ( Figure 4B) may indicate that excessive levels of GA or exaggerated GA signaling lead to stamen deficiency in mac5a and MAC5A-overexpressing plants (Figures 1 and 3), which also supported by the GA application exacerbating the deformation of the stamen in mac5a null mutants ( Figure 2). The dwarf phenotype of ox2 plants was partially rescued by GA application; however, the fertility of plants overexpressing MAC5A was not significantly elevated by GA inhibitor PAC treatment ( Figure S1), indicating that MAC5A also regulate stamen development in a GA-independent pathway. Five DELLA genes (GAI, RGA, RGL1, RGL2, and RGL3) are present in the Arabidopsis genome. The loss of function of RGA and RGL2 is almost sufficient to compensate for floral organ deficiency in ga1-3 mutants, which indicates that both RGA and RGL2 are major repressors of GA response in this process [21]. We found that the expression levels of RGA and RGL2 were significantly increased in mac5a and ox2 plants ( Figure 4B), which suggested that MAC5A may play an important role in regulating the expression of RGA and RGL2 during floral organ development. Taken together, these results indicate that optimal levels of MAC5A are necessary for the development of stamen. MAC5A gene may be involved in the GA pathway to regulate stamen development. Further analysis of the relationship between MAC5A and factors related to GA biosynthesis and signaling will provide insights into GA-mediated stamen development in plants.

Plants Materials and Growth Conditions
Col-0 plants were used as the wild type. mac5a (Salk_132881, mac5a-1) and rbe (CS6396, rbe-2) with a Col-0 genetic background were obtained from the Arabidopsis Information Source (TAIR, http://www.arabidopsis.org, 11 February 2022). Seeds were surface-sterilized and plated on a 0.5 × Murashige und Skoog basal salt (Wako, Osaka, Japan) medium containing 0.8% agar, 1% w/v sucrose, and 0.5 g/L MES (pH 5.8). After sowing, the plates were chilled at 4 • C for 72 h in dark and subsequently incubated at 22 • C under long-day photoperiods (LDs, 16/8 h light/dark). Light intensity was 120 µmol m −2 s −1 . After ten days of growth, the plantlets were planted on rock wool cubes and grown with regularly supplied 0.2 × MS solution. To determine the effect of exogenous GA and PAC treatments on floral organs, GA 3 and PAC were weekly applied to 20-day-old plants by spraying them with 50 µM or 100 µM GA 3 (Sigma-Aldrich, St. Louis, MO, USA) or 20 µM PAC (Sigma-Aldrich) in 0.001% Triton X-100 or Triton X-100 alone.

Microscopy
Pollen viability was assessed using Alexander cytoplasmic staining [40] and observed using a stereomicroscope (SZX16; Olympus, Tokyo, Japan). Here, the cytoplasm of pollen was colored red, which indicated viable pollen grain. When the cytoplasm was absent, the cell wall of pollen grain can be stained, which indicated aborted pollen grain. The anther cannot be stained, which indicated invisible pollen grain. Progress in pollen development was determined using 4 ,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich), and the DAPI signal was examined using a confocal microscope (LSM880; Zeiss, Oberkochen, Germany). Floral organs in fresh samples were observed using a benchtop scanning electron microscope (TM4000; Hitachi, Tokyo, Japan).

Preparation of Chimeric Constructs and Plant Transformation
To create the pBI121:P MAC5A :GUS construct, an approximately 2-kb long promoter sequence upstream of MAC5A was cloned from Arabidopsis genomic DNA using primer pairs with HindIII and SmaI sites. The CaMV 35S promoter of the pBI121 vector (Clontech, Mountain View, CA, USA) [41] was replaced with a MAC5A native promoter that was doubledigested with HindIII/SmaI. MAC5A cDNA was obtained from ABRC and sequenced.  Supplementary Table S1.

Expression Analysis
Total RNA was isolated from inflorescences using an RNeasy Plant Mini Kit (QIAGEN, Dusseldorf, Germany), and cDNA was prepared using PrimeScript Reverse Transcriptase (TAKARA Bio, Shiga, Japan). Quantitative reverse-transcription PCR (qRT-PCR) was performed using a StepOne Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) and SYBR Premix Ex Taq (TAKARA Bio). Actin (NM_112764) levels were used to normalize the expression patterns between samples. qRT-PCR primers used in the study are shown in Supplementary Table S1.

Y2H Assay
The ORFs of MAC5A and RBE were cloned into the EcoRI/BamHI I sites of pGADT7 and pGBKT7 vectors (Clontech, CA, USA), respectively. Bait and prey plasmids were cotransformed into yeast stain AH109 [45]. Empty vectors were used as negative controls. SD/-Leu/-Trp (DDO) was used to select for the bait and prey plasmids. SD/-Ade/-His/-Leu/-Trp (QDO) dropout supplement was used to confirm interactions. The yeast transformants were cultured with DDO at 30 • C for 24 h and subsequently harvested and diluted with distilled water. Then, 5 µL of the indicated dilution was dropped on DDO and QDO medium. Images were taken after the yeast cells were cultured for three to five days at 30 • C. The primers used for plasmid construction are shown in Supplementary Table S1.

Subcellular Localization and BiFC Assay
Full-length cDNAs of MAC5A and RBE without the stop codon were cloned into the KpnI/SpeI site of the vector pBluescript II SK:35S:GFP (pBS:35S:GFP) [9], generating pBS:35S:MAC5A-GFP and pBS:35S:RBE-GFP. Vectors for the BiFC assay were constructed by replacing GFP in the vector pBS:35S:GFP with the N-terminus (154 amino acids) or C-terminus (80 amino acids) of YFP, generating pBS:35S:nYFP and pBS:35S:cYFP, respectively [42]. The ORFs of MAC5A and RBE without stop codons were amplified using a primer pair with the KpnI and SpeI sites and cloned into pBS:35S:cYFP and pBS:35S:nYFP, respectively. The primers are shown in Table S1. Plasmid DNA was introduced into onion epidermal cells using a bombardment system (Bio-Rad, Hercules, CA, USA, PDS-1000), and images were processed using Canvas X software (ACD Systems, Victoria, Canada).