Genome-wide identification of GH3 genes in Brassica oleracea and identification of a promoter region for anther-specific expression of a GH3 gene

Background The Gretchen Hagen 3 (GH3) genes encode acyl acid amido synthetases, many of which have been shown to modulate the amount of active plant hormones or their precursors. GH3 genes, especially Group III subgroup 6 GH3 genes, and their expression patterns in economically important B. oleracea var. oleracea have not been systematically identified. Results As a first step to understand regulation and molecular functions of Group III subgroup 6 GH3 genes, 34 GH3 genes including four subgroup 6 genes were identified in B. oleracea var. oleracea. Synteny found around subgroup 6 GH3 genes in B. oleracea var. oleracea and Arabidopsis thaliana indicated that these genes are evolutionarily related. Although expression of four subgroup 6 GH3 genes in B. oleracea var. oleracea is not induced by auxin, gibberellic acid, or jasmonic acid, the genes show different organ-dependent expression patterns. Among subgroup 6 GH3 genes in B. oleracea var. oleracea, only BoGH3.13–1 is expressed in anthers when microspores, polarized microspores, and bicellular pollens are present, similar to two out of four syntenic A. thaliana subgroup 6 GH3 genes. Detailed analyses of promoter activities further showed that BoGH3.13–1 is expressed in tapetal cells and pollens in anther, and also expressed in leaf primordia and floral abscission zones. Conclusions Sixty-two base pairs (bp) region (− 340 ~ − 279 bp upstream from start codon) and about 450 bp region (− 1489 to − 1017 bp) in BoGH3.13–1 promoter are important for expressions in anther and expressions in leaf primordia and floral abscission zones, respectively. The identified anther-specific promoter region can be used to develop male sterile transgenic Brassica plants. Supplementary Information The online version contains supplementary material available at 10.1186/s12864-020-07345-9.


Background
The Gretchen Hagen 3 (GH3) gene was first identified in Glycine max (soybean) as an early response gene, which is transcriptionally induced in less than 30 min by treatment of auxin plant hormone [1]. Later studies have found that GH3 genes are found in diverse plant species including mosses and fern, but not in two model algae, Chlamydomonas reinhardtii or Volvox carteri [2][3][4][5][6][7]. Like acyl CoA synthetases, non-ribosomal peptide synthetases, and luciferases in ANL superfamily proteins, GH3 proteins conjugate combinations of amino acids and acyl acids in two-step reactions [8,9]. In the first half-reaction involving ATP and acyl acid, adenylated acyl acid is produced and pyrophosphate is released. In the second half-reaction, adenylated acyl acid intermediate reacts with amino acids, resulting in the release of acyl acid-amino acid amido conjugate and adenosine monophosphate. For example, Arabidopsis thaliana (Arabidopsis) GH3.11, jasmonate (JA) resistant 1 (JAR1), and Arabidopsis GH3.17, reversal of sav 2 (VAS2), catalyze the production of JA-isoleucine and indole acetic acid (IAA)-glutamate, respectively [10,11].
GH3 proteins are involved in various developmental processes and environmental responses in plants, by modulating the activities or availabilities of plant hormones and related compounds, including precursors of plant hormones [12]. Abnormal expressions caused by null mutation or hyper-and mis-expression lead to various phenotypic defects. In Arabidopsis, atgh3.11 (jar1) mutant does not produce bioactive JA-Isoleucine and defective in JA signaling, while atgh3.17 (vas2) mutant over-accumulates free IAA at the expense of IAAglutamate [11,13]. In addition, atgh3.12 (avrPphB susceptible 3 (pbs3)) mutant was found to be more susceptible to bacterial pathogens because production of isochorismoyl glutamate, the precursor of salicylic acid (SA), catalyzed by PBS3, is compromised [14]. Overexpression of AtGH3.6 (Dwarf in Light 1 (DFL1)) or AtGH3.2 (Yadokari 1 (YDK1)), which are induced by auxin, causes hyper-sensitivity to light treatment leading to dwarfism [15,16]. Over-expression of AtGH3.5 (WES1), which is induced by treatment of abscisic acid and SA, as well as auxin, leads to auxin resistant phenotypes [17]. In various plants, important roles played by plant GH3 enzymes have also been demonstrated: nodule numbers and sizes in soybean [18], resistance to Xanthomonas bacteria in citrus [19], drought and salt tolerance in cotton [20], and fruit softening in kiwi [21], were shown to be affected by GH3 gene expressions.
Phylogenetic analyses show that plant GH3 genes can be clustered into 3 groups (GroupI~III) based on overall amino acid sequences or 8 subgroups (subgroup 1~8) based on acyl acid-binding site sequences of Arabidopsis, rice, soybean, maize, Selaginella, and moss GH3 proteins [7,10,12,22]. However, only Group I and II GH3 genes have been identified in Gramineae genomes [23][24][25]. Using GH3 enzymes in various plant species, preferential substrates of GH3 enzymes in terms of acyl acids and amino acids have been determined [8,14,18,22,[26][27][28][29][30]. In addition, a systematic evaluation of sixty GH3 enzymes from Arabidopsis, grape, rice, Physcomitrella, and Selaginella also revealed that not all the enzymes encoded by Group I GH3 genes are involved in JA signaling and 12 out of 16 enzymes encoded by Group II GH3 genes display clear substrate preferences for IAA among three acyl acid substrates -jasmonate, IAA, and 4 hydroxybenzoate (4-HBA) [31]. In case of Group III GH3 enzymes, which are encoded by the largest GH3 group in the plant genomes, no clear substrate preferences were established, except AtGH3.9 or OsGH3.13 for IAA and Arabidopsis PBS3 for 4-HBA. In case of Group III subgroup 6 GH3s, only AtGH3.15 in Arabidopsis was shown to have substrate preference for indole butyric acid (IBA), the auxin precursor [28]. Although decrease in IBA-mediated root elongation inhibition and lateral root formation were observed in transgenic plants constitutively expressing AtGH3.15, in vivo function(s) of other subgroup 6 GH3 genes have yet to be determined. In rapeseed (Brassica napus) and its diploid ancestors, Chinese cabbage (Brassica rapa) and cabbage (Brassica oleracea var. capitata), up to sixty-six GH3coding genes have been identified [32,33]. However, detailed study of GH3-coding genes in kale-type Brassica species (Brassica oleracea var. oleracea), TO1000, which serves as an excellent model for important vegetable crops in Brassica oleracea with various morphological and phytochemical traits [34], have not been performed yet.
The anther is a part of the stamen, the male reproductive organ in plants, and is connected to the flower receptacle by a filament, which is the other part of the stamen [35,36]. Anther development is divided into two phases, culminating in the release of pollen grains, the male gametophytes in plants. Microsporogenesis, the first phase, includes establishment of anther morphology, cell and tissue differentiation, and meiosis of microspore mother cells. Tetrads of haploid microspores produced by meiotic divisions of diploid pollen mother cells are released as distinct unicellular microspores into locules by a mixture of enzymes produced from tapetum cells, which also provide nutrients and pollen wall materials for developing pollens [37,38]. During microgametogenesis, the second phase, differentiation of microspores into pollen grains and tissue degeneration occur for the release of pollens. Microgametogenesis starts with the expansion of the microspore, which is often found with the formation of one large vacuole [39]. This involves movement of the microspore nucleus from the center of the cell to a position close to the cell wall, where the microspore produces two unequal cells, a large vegetative cell and a small generative cell, in a process called pollen mitosis (PM) I. Then, the generative cell, which is spatially separated from the pollen grain wall and engulfed by the vegetative cell, undergoes another round of cell division, called PM II [37]. Depending on whether PM II happens before or after pollen dispersal from the anther, the pollens are called tricellular or bicellular pollen [40]. Plant hormones -JA, auxin, gibberellic acid (GA), and ethyleneare known to play important roles in stamen maturation, locule opening, anther dehiscence, and pollen viability during stamen and pollen development [35,[41][42][43].
To expand our knowledge on the regulation and molecular functions of Group III GH3 genes in plantsespecially those in subgroup 6 whose functions are still elusive -GH3 genes in kale-type B. oleracea var. oleracea were identified genome-wide, and expression patterns of subgroup 6 GH3 genes were investigated. It was found that subgroup 6 GH3 genes in B. oleracea var. oleracea, composed of four genes showing synteny with closely related Arabidopsis subgroup 6 GH3 genes, are not induced by auxin, GA, and JA treatment, but have different organ expression patterns. BoGH3.13-1, a subgroup 6 GH3 gene, is specifically expressed in tapetal cells in anther and pollens when microspores, polarized microspores, and bicellular pollens are produced, as well as in leaf primordia and floral abscission zones. Promoter bash experiments revealed that a 62 base pairs (bp) DNA sequence, − 340 to − 279 bp upstream of BoGH3.13-1 start codon, is required for anther-specific expression, while a~450 bp region (− 1489 to − 1017) is necessary for expression in leaf primordia and floral abscission zones.

Results
Thirty-four GH3-encoding genes (BoGH3s) are present in B. oleracea var. oleracea In the Ensembl Plants database (http://plants.ensembl. org/index.html), protein sequences of 55 GH3 candidate genes in kale-type B. oleracea showed similarities to the 19 Arabidopsis GH3 proteins [10]. Among these, 34 GH3 proteins were found to have intact GH3 domains (pfam03321) and considered as GH3 proteins (Table S1; Figure S1). Although identical genomic sequences were used for annotation, only 30 B. oleracea GH3 candidate proteins, including two with truncations in GH3 domains, were found to have significant similarities to Arabidopsis GH3s in NCBI database (NCBI, http://ncbi.nlm. nih.gov) [34]. The 34 BoGH3 proteins with the intact GH3 domains in Ensembl Plants database include all 28 putative GH3 proteins with the intact GH3 domains identified in NCBI database (Table S1). For proteins showing different protein sequences between two databases, such as BoGH3.12-2 and BoGH3.17-1, NCBI protein models were adopted in our study because they are supported by RNA-seq data in NCBI. While 34 GH3 protein-coding genes were identified from B. oleracea var. oleracea in our study, 25 and 29 GH3 proteincoding genes were previously reported for cabbage-type B. oleracea var. capitata in the comparison with B. napus genes by two independent studies, respectively [32,33].
Similar to previous phylogenetic analyses of GH3 proteins including cabbage-type B. oleracea var. capitata, phylogenetic clustering of Arabidopsis and BoGH3 proteins demonstrated that BoGH3 proteins can be divided into three groups (Group I, II, and III) (Fig. 1a) [6,10,32,33]. It was found that Group I consists of two Arabidopsis and four BoGH3 proteins, while Group II consists of eight Arabidopsis and 11 BoGH3 proteins. In the case of Group III, nine Arabidopsis GH3s and 19 BoGH3 proteins were clustered together. In general, exon/intron structures of BoGH3 genes were same to closely related counterparts in Arabidopsis with some exceptions (Fig. 1b). For example, four protein-coding exons were detected for BoGH3.1 in Group II, based on the distribution of RNA-seq reads in NCBI database, while three protein-coding exons of AtGH3.1 is reported in TAIR JBrowse (https://jbrowse.arabidopsis.org/). In case of BoGH3.11-2 and BoGH3.11-3, which are closely related to AtGH3.11 (JAR1) with four protein-coding exons, only three exons supported by RNA-seq reads were observed. Structural differences were also observed for five BoGH3 genes (BoGH3.8-2, BoGH3.8-5, BoGH3.13-3, BoGH3.18-1, and BoGH3.18-7) that were identified only in Ensembl Plants.
BoGH3.13-1 is strongly expressed in stamen at a specific stage during flower development For four subgroup 6 and two auxin-inducible GH3 genes in B. oleracea var. oleracea, relative expression patterns in six different organs -root, leaf, stem, floral bud, opened flower, and silique -were determined. Among four subgroup 6 BoGH3 genes, BoGH3.13-1 was found to be most strongly expressed in floral bud, although significant expression was also observed in silique compared to that in leaf (Fig. 4a). Only negligible expressions of BoGH3.13-1 were detected in other organs, including open flowers. For the other three subgroup 6 BoGH3 genes, the strongest expression was commonly found in siliques ( Fig. 4 b-d), while comparable expressions in floral bud and open flower were also observed for BoGH3.13-2 (Fig. 4b). For auxin-inducible BoGH3.2 and BoGH3.5-1, which were included as comparison, distinct relative expression patterns were detected: BoGH3.2 and BoGH3.5-1 were found to be most strongly expressed in root and floral bud, respectively ( Fig. 4 e & f). For three subgroup 4 BoGH3 genes, stronger expressions were commonly observed in roots ( Figure S2).
One hundred eighty-six bp region upstream of BoGH3.13-1 is sufficient for anther-specific expression DNA sequences responsible for tissue-specific expression of BoGH3.13-1 was investigated with different DNA regions upstream of the start codon (Fig. 8a). When P1, in which − 1017~− 1 bp region was fused upstream of GUS reporter gene, was used to generate P1 transgenic plants, GUS expressions in anthers and pollens were still detected ( Fig. 8b-d), but those in floral abscission zones and leaf primordia were lost, except one case showing GUS staining in the floral abscission zone (Fig. 8

Discussion
Thirty-four GH3-coding genes of kale-type B. oleracea var. oleracea, which have intact GH3 domains, were identified from Ensembl plants database (Fig. 1a). Among these, 28 gene models were also found in NCBI database, which had used the identical genomic sequence for annotation [34]. The discrepancy in BoGH3 gene numbers between Ensembl plants and NCBI database may result from the use of different gene prediction algorithms or validations. Recently, twenty-nine GH3 protein-coding gene models related to cabbage-type B. oleracea var. capitata GH3 genes were identified from the investigation of genomic sequence of B. napus [32,33], and twenty-eight genes were found to have intact GH3 domains and meet our criteria, while Bol042635 was found to encode a truncated GH3 domain with only 224 amino acids (Table S2) [33]. Among 34 BoGH3 coding-genes reported in this study, putative orthologs of cabbage-type B. oleracea were identified for 19 genes, but clear orthologous relationship could not be determined for the other 15 B. oleracea var. oleracea GH3 genes, based on amino acid sequence identities of over 95%. Considering 6 B. oleracea var. oleracea GH3 genes, whose expression could not be confirmed in NCBI database, are included in these 15 cabbage-type GH3 genes without putative orthologs, we speculate these 6 genes are pseudogenes and lost in B. oleracea var. capitata. In addition, orthologs of 9 cabbage-type B. oleracea GH3 genes could not be determined in B. oleracea var. oleracea (Table S2).
Group III subgroup 6 GH3 genes in B. oleracea var. oleracea and Arabidopsis seem to have evolved by duplications. Four Arabidopsis GH3 genes in subgroup 6 are located within 15 kbp region on the same Arabidopsis chromosome, while 4 BoGH3 genes in the same subgroup are located on 4 different chromosomes of B. oleracea var. oleracea (Fig. 2). Syntenies found between genomic regions around the subgroup 6 AtGH3 and BoGH3 genes suggest that AtGH3.13~AtGH3.16 cluster in Arabidopsis was generated by tandem duplication after Brassica lineage-specific whole genome triplication and/or other BoGH3 genes around BoGH3.13-1, BoGH3.13-2, and BoGH3.13-4 might have been lost after divergence of Arabidopsis and Brassica lineages [47]. Consistent with this idea, one intact and one truncated form of GH3 genes in subgroup 6, BoGH3.13-1 and B02g011230, were identified within 15 kb region on the chromosome 2 of B. oleracea var. oleracea (Fig. 2b). Members in gene family in plants are known to evolve through both tandem (local) duplication and whole genome duplication, which were followed by gene loss or gene retention leading to functional diversification [48]. Nonetheless, close genomic locations of subgroup 4 and subgroup 6 GH3 genes in Arabidopsis and B. oleracea var. oleracea indicate both AtGH3.12-like and AtGh3.13like GH3 genes were present in proximity before the separation of Arabidopsis and Brassica lineages. For exon/intron structures of BoGH3 and AtGH3 genes, overall similarities were observed for the evolutionarily related genes. However, distributions of RNA-seq reads in NCBI database revealed that protein-coding exons of BoGH3.1, BoGH3.11-2, and BoGH3.11-3 are differently organized compared to those of related Arabidopsis GH3 genes (Fig. 1b). Differences in the structures observed for five BoGH3 genes (BoGH3.8-2, BoGH3.8-5, BoGH3.13-3, BoGH3.18-1, and BoGH3.18-7) and related Arabidopsis genes might result from deletions/insertions and incorrect annotations, considering that these five BoGH3 genes are identified only in Ensembl Plants, not supported by RNA-seq data in NCBI database, and encode predicted GH3 proteins with multiple deletions (Fig. 1b & S1).
Four subgroup 6 BoGH3 genes, which seem to be generated from same ancestor gene(s), show distinct expression patterns. At the organ level, BoGH3.13-1 is almost exclusively detected in floral buds by qRT-PCR, while the strongest expressions of BoGH3.13-3 and BoGH3.13-4 are observed in siliques (Fig. 4 a, c & d). In case of BoGH3.13-2, no significant expression preference is found among different organs and constitutively expressed in all parts of flowers (Figs. 4b, 5 c & f). In developing floral buds, BoGH3.13-1 is strongly expressed in stamen when floral buds are about 2~6 mm long (Fig. 5). However, investigation of BoGH3.13-1 promoter activity using GUS reporter revealed that BoGH3.13-1 is also expressed in abscission zones in siliques and leaf primordia, in addition to tapetal cells in stamen and pollen grains (Fig. 6f-l). Relatively weak detection of BoGH3.13-1 in siliques by qRT-PCR may be related to the facts that the gene is expressed only in a small portion of siliques cells, although we do not exclude the possibility that the expression level is also lower in siliques than in stamen. In 2~6 mm floral buds, in which BoGH3.13-1 is most strongly expressed, microspores, polarized microspores, and bicellular pollens are mainly observed in anthers (Fig. 7). Similar to BoGH3.13-1, two syntenic subgroup 6 Arabidopsis GH3 genes, AtGH3.13 and AtGH3.16, are expressed in flower stage 9~11 floral buds and flower stage 12, respectively [49]. More specifically, AtGH3.16 is expressed in polarized microspore and AtGH3.13 is expressed bicellular pollens. Based on the numbers of pollen nuclei and floral bud phenotypes [50], the flower stages, when BoGH3.13-1 is strongly expressed, roughly correspond to stages 8~12 of Arabidopsis flower and overlap with the periods when AtGH3.13 and AtGH3.16 are expressed (Fig. 7 & S2). Although BoGH3.5-1, a group II BoGH3 gene, is also specifically expressed in stamen like BoGH3.13-1 (Figs. 3a-b & 6u-x), BoGH3.5-1 seems to be expressed in a longer time period compared to BoGH3.13-1 (Figs. 5a -b, 6h & o). Different from BoGH3.13-1, neither in floral abscission zones nor in leaf primordia is expression of BoGH3.5-1 observed (Fig. 6 p & r). It needs to be determined which substrate(s) are preferentially used by BoGH3.13-1 and BoGH3.5-1.
BoGH3.13-1 is not induced by auxin (IAA or 2,4-D), JA, or GA, but expressed in a tissue-specific manner. Different from many GH3 genes in other plants, which have been found to be induced by various plant hormones [1,17,33,51,52], no expression changes for BoGH3.13-1 and 3 other subgroup 6 BoGH3 genes were detected in our experimental conditions (Fig. 3). In contrast, expression levels of BoGH3.2 was found to be elevated upon exposure to auxin in the same condition. Similar to our findings, all subgroup 6 GH3 genes in B. napus, an allotetraploid carrying chromosomes with B. oleracea origin, did not show any significant expression changes in response to IAA treatment in leaves [33]. Although BoGH3.13-1 expression is not induced by auxin in our experimental condition, tissues or cells, in which BoGH3.13-1 promoter activity is detected, largely overlap with the regions where auxinresponsive DR5 promoter is activated in Arabidopsis and rice (Figs. 3 & 6) [53][54][55][56]. We do not exclude the possibility that BoGH3.13-1 promoter is less sensitive to auxin treatment than BoGH3.2, but we prefer the idea that expression of BoGH3.13-1 is induced by a transcription factor that is activated in tissue-specific manners downstream of auxin signaling pathway. When expression patterns of BoGH3 genes were probed at the organ level using EMBL-EBI expression atlas (https://www.ebi.ac.uk/gxa/experiments/E-GEOD-4289 1/Results), BoGH3.13-1 was found to be specifically expressed in floral buds, similar to our qRT-PCR results ( Fig. 4 & Table S3). However, expression levels of BoGH3.5-1 was found to be higher in silique than in floral bud, different from our results. Although differences in growth conditions and sampling times might have affected gene expressions, transcription profiling based on RNA-seq could have been confounded by sequence reads produced from highly homologous BoGH3 gene family members. Given that expressions of BoGH3.13-1 in leaf primordia and floral abscission zone could not be detected by transcription profiling, complete understanding of some BoGH3 expression patterns seem to require both qRT-PCR and investigation of promoter activity using promoter-reporter system.
Anther-specific expression of BoGH3.13-1 is directed by 62 bp DNA sequence, from − 340 to − 279 bp from the start codon. Determination of promoter regions important for tissue-specific expressions revealed that about 180 bp P3 region (− 340~− 155) close to the transcription start site is sufficient for anther-specific expression (Fig. 8). The observation that P4 region (− 278~− 155) does not supports anther-specific expression suggests that cis-acting element necessary for antherspecific expression is included by 62 bp DNA sequence from − 340 to − 279 bp. GUS expression detected in 5 out of 12 P5 transgenic lines containing − 418 to − 279 bp region further supported this idea. We suspect that deletion of promoter sequences (− 278~− 155) close to the transcription start site makes anther-specific expression depend on the genomic positions where transgene is inserted. In Arabidopsis, Male Sterility 1, a plant homeodomain-finger, and MYB99 transcription factors functioning in anther and pollen development pathway are expressed in microspores, polarized microspores, and bicellular pollens [49,57]. The findings (1) that BoGH3.13-1 is strongly expressed when microspores, polarized microspores, and bicellular pollens are produced and (2) that MYB core cis-acting element (CTGT TA) is located at − 293~− 288 raises a possibility that Brassica oleracea var. oleracea ortholog of Arabidopsis MYB99 plays an important role for anther-specific expression of BoGH3.13-1 [58]. Because GUS expressions in leaf primordia and floral abscission zones are lost without any obvious effect on anther-specific expression, cis-acting element important for leaf primordia and floral abscission zone expressions must be located in the − 1489 to − 1017 region in BoGH3.13-1 promoter and independent of cis-acting element for anther-specific expression ( Fig. 8af).

Conclusions
In this study, we identified 34 GH3 genes in Brassica oleracea var. oleracea, including four subgroup 6 GH3 genes, and a critical promoter region for anther-specific expression of a subgroup 6 BoGH3 gene, BoGH3.13-1. The information will broaden our understanding of transcriptional regulations during anther development and can be used to develop transgenic male sterile lines for economically important Brassica plants.

Plant growth
Brassica oleracea var. oleracea (TO1000 seeds, stock number CS29002) were obtained from the Arabidopsis Biological Resource Center. In the Ensembl Plants and NCBI database search, E-value thresholds for candidates were set on 1e − 1 and 0.1, respectively. BoGH3 proteins were further determined by the presence of the intact GH3 domains, and their exon/intron structures were determined based on RNA-seq exon coverage and RNA-seq intron spanning reads from NCBI B. oleracea annotation Release 100. Similarly, GH3 protein sequences in B. oleracea var. capitata were identified using the sequence in Bolbase (http://ocri-genomics.org/bolbase/blast/blast. html) [59].

Multiple sequence alignment and construction of phylogenetic tree
The multiple sequence alignment of GH3 proteins was performed using Clustal Omega and visualized using Jalview [60,61]. Phylogenetic analysis was performed using the molecular evolutionary genetics analysis (MEGA) software [62]. The evolutionary history was inferred by using maximum likelihood method based on the JTT matrix-based model [63]. All positions with less than 90% site coverage were eliminated. There were a total of 549 positions in the final dataset. The bootstrap test was repeated 1000 times. An orthologous relationship for synteny between B. oleracea var. oleracea and Arabidopsis was determined using gene information in the Ensembl database (https://plants.ensembl.org/Brassica_ oleracea/Info/Index).

Hormone treatment
For hormone treatment, surface sterilized Brassica oleracea var. oleracea seeds were germinated and grown in 24 well plates containing 1 ml half-strength liquid MS media for 5 days. 2,4-D (D0901), IAA (I0901), GA (G0907) and JA (J0936) from Duchefa (Haarlem, Netherlands) were treated to whole seedlings, after the seedlings were further grown in 2 ml fresh liquid media for 6 h.

Sample collection
Five-day-old seedlings were used to determine whether the BoGH3 gene of interest is induced by hormone treatment. For gene expression analysis by qRT-PCR, root, leaf, stem, floral bud, open flower, and silique were obtained from 3 individual plants: more specifically, 11th to 13th leaves, fifth to seventh node for stems, a mix of unopened floral buds without white petals exposed (bud length less than about 8 mm), a mix of open flowers (bud length larger than 8 mm) with white petals exposed, and siliques with various sizes were collected. Samples for floral buds were further divided into 5 categories by lengths: 0~2, 2~4, 4~6, 6~8, and 8~10 mm sizes ( Figure S3). Sepals, petals, anthers, and pistils were collected from 4~6 mm -long unopened floral buds. After collection, samples were frozen in liquid nitrogen and stored at − 80 C°until RNA isolation. Samples for GUS staining were collected when transgenic Arabidopsis seedlings were 8 days old, or later when inflorescence and siliques were mature enough.

RNA isolation, reverse transcription, and qRT-PCR analysis
Total RNA was extracted using PhileKorea E-Zol RNA Reagent (Seoul, Korea) or Ambion TRIzol® Reagent (Austin, USA) following the manufacturer's instructions. For silique samples, Invitrogen Plant RNA Purification Reagent (Carlsbad, USA) was used to. cDNA was synthesized from RNA with 260/280 ratios between 1.8 and 2.1. First stand cDNA was synthesized with Toyobo ReverTra Ace -α (Osaka, Japan) and 1.0 μg of total RNA, according to the manufacturer's instructions. In case of hormone-treated seedlings, 0.5 μg of total RNA was used. As described in Nam et al. (2019), qRT-PCR was performed with a two-step reaction: 3 min (min) at 95°C, followed by 50 cycles of 10 s at 95°C and 30 s at 60°C. Primer sequences used are listed in Table S4. For each analysis, three technical replicates of at least two independent biological replicates were used.  (Table S5). After SalI and BamHI digestion, the PCR fragments were cloned into pBI101.1 vector between SalI and BamHI sites. The construct was transformed into Arabidopsis by the floral dip method [64].
Histochemical GUS staining and paraffin section of GUSstained samples Histochemical GUS staining was performed with 0.5 mg/ ml MBcell 5-bromo-4-chloro-3-indolyl-beta-D-glucuronic acid-cyclohexylammonium salt (Seoul, Republic of Korea), as previously described [65]. The floral buds of T1 or T2 transgenic plants carrying a GH3 promoter:: GUS fusion transgene were immersed in GUS reaction buffer in the dark condition for 1 day at 37°C, after which samples were washed in 95% ethanol for 1~2 h. At least 7 individual transgenic lines were used to analyze GUS expression patterns.
To perform the paraffin section, GUS-stained samples were fixed in FAA solution (Formaline: ethanol: glacial acetic acid: water = 10: 50: 5: 35) for at least 24 h and washed in water for 24~48 h. Then the samples were dehydrated in 50, 60, 70, 80, 90% ethanol series for 20 min once, and 100% ethanol for 20 min twice. The samples were incubated in a series of ethanol:xylene mix (75: 25, 50:50, and 25:75) for 30 min in each mix, and to a series of xylene:paraffin mix (2:1, 1:1, and 1:2) for 1 h twice in each mix. The samples were incubated in molten paraffin for 24 h and poured into blocks on a slide warmer at 70°C and cooled down to 25°C. Eight μmthick transverse sections of paraffin-embedded samples were made with a microtome. Ribbons of serial sections floated on warm water (50°C) were transferred to slide glasses on the slide warmer at 70°C and cooled down to 25°C. Paraffin in the sections was removed with xylene.