The genomic sequence, annotated file, and protein sequences of the six Brassica species in U’s triangle were downloaded ( Supplementary Table S1 ), and the longest transcript was used for subsequent analysis. In summary, 105 AtMADS protein sequences downloaded from the TAIR database (https://www.arabidopsis.org/) (Parenicova et al., 2003) were used to perform a BLASTP search of the local protein database of the six U’s triangle species with an E-value < e^-10 and identity ≥ 50% (Camacho et al., 2009). The hidden Markov model (HMM) profiles of SRF-TF (PF00319) and K-box (PF01486) were retrieved from the Pfam database (http://pfam-legacy.xfam.org/), and HMMER software (version 3.0) was used to search for homeodomains with an E-value < 1e^-5. Subsequently, all MADS-box protein sequences were submitted to the National Center for Biotechnology Information Conserved Domain Database (NCBI-CDD) (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi) (Lu et al., 2020) and SMART database (http://smart.embl-heidelberg.de/smart/batch.pl) to examine the integrity of the MADS-box or K-box domains (Letunic and Bork, 2018). The theoretical isoelectric point (pI) and molecular weight (MW) of all candidate MADS-box proteins were predicted using ExPASy (https://www.expasy.org/) (Gasteiger et al., 2003).

Multiple sequence alignment of MADS-box protein sequences was performed using MAFFT software with the FFT-NS-I method (Katoh et al., 2019). Unrooted and rooted Maximum Likelihood (ML) trees were constructed using FastTree software (Price et al., 2009) and visualized using EvolView (http://www.evolgenius.info/evolview) (Zhang et al., 2012) and iTOL (http://itol.embl.de/) (Letunic and Bork, 2007). The gene structure files of all the MIKC-type genes were extracted using a genome annotation file. The conserved motifs were predicted using the MEME software, with the number of motif sets at 10 (http://meme-suite.org/tools/meme) (Bailey et al., 2009). TBtools was used to draw the exon-intron structure and conserved motif features (Chen et al., 2020).

The MIKC-type protein sequences of the U’s triangle species were aligned using BLASTP with an E-value < 1e^-10 (Camacho et al., 2009). The tandem and segmental duplication events in each species were analyzed using the MCScanX program (Wang et al., 2012) and visualized using Circos (Krzywinski et al., 2009). Genomic synteny was analyzed using JCVI software and the detailed workflow reference Python version MCscan (https://github.com/tanghaibao/jcvi/wiki/).

RNA-seq data for six tissues (roots, stems, leaves, flowers, seeds, and siliques) and four abiotic stressors (dehydration, salt, ABA, and cold) in B. napus were downloaded from the National Genomics Data Center (NGDC) (project ID: CRA001775) (https://ngdc.cncb.ac.cn/?lang=en). The transcriptome data of embryos and seed coats at different developmental stages and two different white powder-resistant rapeseed lines after inoculation with powdery mildew were downloaded from the NCBI database (project ID: PRJNA641876 and PRJNA778657). All transcriptome data were mapped to the Darmor-bzh genome using HISAT2 (Kim et al., 2015). The Transcripts Per Million (TPM) values were calculated using the FeatureCounts R package, and a heatmap of expression level was plotted using TBtools software (Liao et al., 2014; Chen et al., 2020).

First, four BnaAG genes with the highest homology to A. thaliana (AT4G18960) were selected for designing single guide RNA (sgRNA). The sgRNA1 was designed to target the MADS-box and sgRNA2 to target the K-box, and then a multiplex genome editing tool CRISPR/Cas9 with two sgRNAs was used to mutate BnaAG genes (Xing et al., 2014). Finally, the successfully constructed recombinant plasmid (pHSE401-BnaAG) was transformed into Agrobacterium tumefaciens strain GV3101, followed by genetic transformation of 19YB437 (recipient material) using a detailed transgenic operation method described by Bhalla and Singh (Bhalla and Singh, 2008). Primers used to construct the vectors are listed in Supplementary Table S12 .

All positive tissue culture plantlets were transplanted into humus soil and grown under a 16 h light/8 h dark photoperiod at 22°C until they reached the five-leaf stage. After four weeks of vernalization, the seedlings were transplanted into a normal environment for growth. The variant phenotypes of floral organs were observed and recorded; DNA was extracted from leaf tissues of each type of flower organ mutant, and the fragments near the two target sequences were amplified using PCR with specific primers ( Supplementary Table S12 ). The PCR products and positive colonies were sequenced, and the editing events of each mutant were analyzed according to the results of Sanger sequencing.

In total, 1430 MADS-box genes were identified in the genome of the six U’s triangle species. To uncover the kinship relationships of MADS-box genes among A. thaliana and U’s triangle species, the ML method was used to construct the phylogenetic tree based on the alignment of 105 MADSs in A. thaliana and 1430 MADS-box proteins in U’s triangle species ( Figure 1A ; Supplementary Table S2 ). The 1,535 MADS-box genes were divided into M-type and MIKC-type subfamilies; among them, the MIKC subfamily had the highest number of members (842), followed by the Mα (287), Mβ (143), Mγ (164), and Mδ (99) subfamilies ( Figure 1B ; Supplementary Table S3 ). The three diploid Brassica species B. rapa (AA), B. nigra (BB), and B. oleracea (CC) in the U’s triangle have undergone a whole genome triplication event (WGT) (Wang et al., 2011; Woodhouse et al., 2014) resulting in the expansion of MADS-box genes compared with A. thaliana; however, the gene numbers were significantly different. For example, 159, 184, and 180 MADS-box genes were identified in B. rapa, B. nigra, and B. oleracea genomes, respectively. In addition, although allopolyploidy facilitated the expansion of MADS-box genes, the total number of MADS-box genes identified in allotetraploid species differed from the total number of genes in its two progenitor species. For example, 240 MADS-box genes were identified in the genome of B. juncea, which was significantly less than the sum of genes in its progenitor species B. rapa (159) and B. nigra (184), and the number of MADS-box genes in B. carinata genome (335) was slightly lower than the sum of genes in its progenitor species B. nigra (184) and B. oleracea (180). The 1430 MADS-box genes were renamed based on the gene information about their distribution on chromosomes, and the protein length, molecular weight (MW), and isoelectric points (pI) of each MADS protein were analyzed synthetically to understand the basic information more comprehensively ( Supplementary Table S3 ).

MIKC-type MADS-box genes are numerous, and their classification in plants is complicated. The ML phylogenetic tree was constructed for 39 MIKC-type genes in A. thaliana and 803 in the U’s triangle species ( Figure 2 ). Subsequently, 842 MIKC-type genes were divided into 35 subclades according to classification criteria and nomenclature in A. thaliana. Among them, 32 subclades contained MIKC-type genes from each species, and these subclades were renamed referring to the corresponding name in A. thaliana. The other three subclades (AGL72-like, AGL18-like, and FLC-like) did not contain corresponding homologous genes in A. thaliana; therefore, their renaming was based on the name of the branch closest to their evolutionary relationships ( Figure 2 ). AGL72-like subclades had the highest number of genes, which may be another sister group of AGL72 subclades. In addition, most subclades had a clear evolutionary classification, whereas the SEP subclade had the highest number of genes and a complicated branch topology, implying that SEP subclades may have undergone multiple duplication events in the U’s triangle species.

The exon-intron structural characteristics of MIKC-type genes varied remarkably in different subclades; however, the number and arrangement of exon-introns were similar in the same subclade ( Figure S1 ). Most of the genes in the AG and AP3 subclades comprised seven exons and six introns ( Figures 3A, D ), AP1 and CAL subclades possessed the closest evolutionary relationship, and the majority of MIKC-type genes in AP1 and CAL subgroups generally contained eight exons and seven introns ( Figure 3B ); the PI subclade ( Figure 3E ) had six exons, and the SEP subclade had eight exons ( Figure 3C ).

Ten conserved motifs were observed in 803 MIKC-type protein sequences ( Supplementary Table S4 ). Motif 1 and motif 3 existed in most of the MADS-box domains, and motif 2, comparatively conserved in the K-box domain, may be responsible for the conservation of MIKC-type proteins among different subclades. However, other motifs were specifically distributed in different subclades and determined the specificity of the MIKC-type subclades ( Figure 3 ; Figure S1 ). For example, comparing the conserved motifs of the AG, AP1, and CAL subclades ( Figures 3A, B ), motif 8 was only found in the AP1 and CAL subclades, whereas motifs 6 and 7 were only found in the AG subclade. Furthermore, motif 4 and motif 10 were distinctively distributed in the AP3 and PI subclades ( Figures 3D, E ), respectively, determining the specificity of these two subclades. The gene structure and conserved motifs of the other subclades are shown in Figure S1 .

BLAST and MCScanX software were used to analyze the gene duplication events of MIKC-type genes in the U’s triangle species. A mass of tandem and segmental duplication events were detected in the genomes of B. rapa (3/69) ( Figure 4A ), B. nigra (4/75) ( Figure 4B ), B. oleracea (8/66) ( Figure 4C ), B. juncea (2/137) ( Figure 4D ), B. napus (4/194) ( Figure 4E ), and B. carinata (8/281) ( Figure 4F ). The duplication event numbers of MIKC-type genes in three diploid basic species were nearly the same. However, there was a significant difference among the three allotetraploids, particularly in segmental duplication events ( Supplementary Tables S5 , 6 ), suggesting that MIKC-type genes have different evolutionary processes in the three allotetraploid species. Furthermore, the subgenome distribution of segmental duplication events in the genomes of three tetraploid Brassica species was analyzed. In B. juncea, 88 segmental duplication events occurred across the AA and BB subgenomes, 22 across the AA subgenome, and 27 across the BB subgenome ( Figure 4D ). In B. napus, 125 segmental duplication events occurred across the AA/CC subgenome, 41 across the AA subgenome, and 28 across the CC subgenome ( Figure 4E ). In B. carinata, 159 segmental duplication events occurred across the BB/CC subgenome, 85 across the BB subgenome, and 37 across the CC subgenome ( Figure 4F ). These results indicate that segmental duplication events among the three tetraploid Brassica species mainly occurred across the different subgenomes.

To elucidate the evolutionary process of MIKC-type genes, synteny between three allotetraploid species (B. carinata, B. juncea, and B. napus) and its progenitor Brassica species (B. rapa, B. nigra, and B. oleracea) was performed ( Supplementary Table S7 ). A total of 15 gene pairs were identified between B. rapa and the A subgenome of B. napus and B. juncea ( Figure 4G ). Among them, BnaMADS148 (BnaAnng06990D) may be located on chrA02 of B. napus because its syntenic gene pairs were all located on chrA02. In addition, 32 gene pairs were detected across the B. nigra genome and B subgenomes of B. juncea and B. carinata ( Figure 4H ), of which 28 gene pairs were shared on the same chromosome among the B genome of progenitor species and B subgenome of its hybridized allotetraploid species, whereas the other four were shared on different chromosomes. Additionally, 37 MIKC-type gene pairs were identified across the B. oleracea genome and C subgenomes of B. napus and B. carinata ( Figure 4I ). Overall, the number of syntenic gene pairs in subgenome A was significantly lower than that in subgenome B or C.

B. napus is an important oil crop in the U’s triangle, and its abundant and reliable transcriptome data provide a basis for exploring the expression patterns of MIKC-type genes. Therefore, transcriptome data of various tissues, embryos, and seed coats at different developmental stages were downloaded from the public RNA-seq database to analyze the expression characteristics of MIKC-type genes. A total of 63 MIKC-type BnaMADS-box genes were screened from the expression matrix of different tissues ( Supplementary Table S8 ), which were classified into five subgroups according to their expression patterns ( Figure 5A ). Subgroup I genes were highly expressed in stems and leaves, and some genes were also highly expressed in siliques and other tissues; however, all genes were expressed at low levels in the seed. Subgroup II genes were mainly expressed in flowers, seeds, and siliques, but their expression was low in roots. The subgroup III genes were root-specific. Nine genes, including especially BnaMADS37, BnaMADS111, BnaMADS155, and BnaMADS189, in subgroup IV were highly expressed in seeds. The genes in subgroup V, including BnaMADS12, BnaMADS30, BnaMADS300, and BnaMADS313, were only highly expressed in the flowers. Overall, the expression patterns of the 63 BnMADS-box genes were tissue-specific, suggesting that these genes may have vital biological functions in tissues during specific developmental processes.

To explore the expression patterns of MIKC-type BnaMADS-box genes at different developmental stages of the embryo and seed coat, the expression patterns of 48 BnaMADS-box genes were divided into two major subgroups ( Figure 5B ; Supplementary Table S9 ). Subgroup I genes were mainly expressed during the unfertilized ovule stage and first three periods of embryonic development (zygote, octant, and globular stages), whereas subgroup II genes were generally expressed during the embryo and seed coat development stages, especially in the middle and late stages of embryo development, such as BnaMADS27 was highly expressed in torpedo and bent stage embryos, BnaMADS141 and BnaMADS323 were highly expressed in octant, globular, and heart stage embryos. In addition, BnaMADS98 was highly expressed in globular stage embryo, whereas it was less expressed during seed coat development. Therefore, analyzing the expression pattern of MIKC-type genes in B. napus provided a guide for exploring the biological functions of MIKC-type genes in various tissues during the entire developmental process.

To determine the mechanism of MIKC-type genes responding to biotic and abiotic stress, transcriptome data for dehydration, salt, ABA, and cold stress were downloaded from the public RNA-seq database ( Supplementary Table S10 ). Based on gene expression trends under abiotic stress, the 38 MIKC-type genes were divided into three categories: subgroups I, II, and III ( Figure 6A ). Subgroup I genes were expressed under various types of abiotic stresses. BnaMADS227, BnaMADS286, and BnaMADS297 were significantly upregulated after dehydration, salt, ABA, and cold treatment, whereas BnaMADS104 and BnaMADS311 were more susceptible to dehydration, salt, and ABA stress. BnaMADS222 was downregulated at 1 h and 8 h of dehydration stress, indicating that BnaMADS222 was more sensitive to dehydration stress. A small subset of the genes in subgroup II was sensitive to abiotic stress. BnaMADS2 and BnaMADS164 were induced by dehydration and cold treatment; BnaMADS38 and BnaMADS302 were upregulated under dehydration and salt stress, and BnaMADS56 was significantly downregulated after dehydration stress. Subgroup III genes had low expression under abiotic treatments; however, individual genes were upregulated after stress treatment. For example, BnaMADS123 and BnaMADS63 were significantly highly expressed after 24 h of cold treatment, and BnaMADS35 was upregulated after 24 h of ABA treatment. The rapeseed leaf transcriptome data of resistant and susceptible lines after inoculation with powdery mildew were downloaded for further analysis ( Figure 6B ; Supplementary Table S11 ). A total of 21 MIKC-type genes responded to powdery mildew stress, of which 16 genes, including BnaMADS113, BnaMADS118, and BnaMADS321, showed significantly higher expression in resistant strains than in sensitive strains. Furthermore, BnaMADS2, BnaMADS11, BnaMADS164, BnaMADS190, and BnaMADS215 showed higher expression in susceptible strains than in resistant strains.

To further verify the contribution of BnaAG genes to floral organ development, four highly homologous BnaAGs, BnaA01g09760D, BnaA03g43820D, BnaC01g11460D, BnaC03g62970D (also named BnaAG.A01, BnaAG.A03, BnaAG.C01, and BnaAG.C03, respectively), located on chromosomes A01, A03, C01, and C03, respectively, were isolated by Blastp with AtAG. Phylogenetic tree analysis showed that BnaAG.A01 and BnaAG.C01 had the closest relationships with AtAG ( Figure 7A ). The observation of gene structure features showed that most of the BnaAG genes contained seven exons, whereas BnaAG.C01 was composed of nine exons but had no effect on its conserved domain at the gene structure level ( Figures 7B ). Based on multiple alignments of BnaAG homologous genes, CRISPR/Cas9 vectors were constructed to target the conserved regions of four BnaAG homoeologs. In brief, the consensus sequence on the first exon of BnaAG.A01, BnaAG.A03, BnaAG.C03, and the third exon of BnaAG.C01 were targeted by sgRNA1. SgRNA2 targeted the consensus sequence on the third exon of BnaAG.A01, BnaAG.A03, BnaAG.C03, and the fifth exon of BnaAG.C01 ( Figures 7C ). Based on these results, sgRNA1 was located on the MADS-box domain and sgRNA2 on the K-box domain.

After the genetic transformation of the rapeseed hypocotyl mediated by Agrobacterium tumefaciens, more than 40 T0 generation hygromycin B-positive plants were obtained, of which 20 lines consisted of four concentric wheels and showed no difference from the wild-type (WT), whereas the other lines showed a disorganized floral structure. Therefore, T0 generation lines can be classified into four types based on their floral organ mutation phenotypes. To verify the mutation type, plants of all phenotypes were screened using Sanger sequencing, and the editing efficiency of sgRNA1 was found to be significantly lower than that of sgRNA2 ( Figures S2 , 3 ). Specifically, compared to wild type (WT) ( Figures 8A–C ), there were no phenotypic differences found in wt-like (wtl) ( Figures 8D–F ). Sequencing results showed that no editing events existed in target 1, and all BnaAG homologs had a single base inserted except BnaAG.A01 in target 2 ( Figure 8S ). The female flower-like (feml) mutant ( Figures 8G–I ), which led to the conversion of stamens into petals and expansion of carpels, generated low seed-setting and distorted siliques after cross-pollination ( Figure 8P ). The sequencing results showed that no editing events happened except for single base insertion in BnaAG.C01 of target 1. BnaAG.C01, BnaAG.C03, and BnaAG.A03 had a single base insertion in target 2, whereas BnaAG.A01 had a deletion of three bases (CGA) in target 2 that did not cause frameshift mutation ( Figure S4 ). The next mutant, like ag-2 mutant in A. thaliana, gave rise to the following phenotype: six stamens replaced by six petals, and the carpel replaced by a new flower with an elongated pedicel, in which stamens and carpels were replaced by petals (also called agamous-like flower-2 (aglf-2) (Yanofsky et al., 1990; Tsutsui and Higashiyama, 2017) ( Figures 8J–L, Q, R ). Target 1 of the BnAG homologous gene was not edited; however, insertion or deletion occurred in all BnAG homologous genes in target 2, resulting in a frameshift mutation. The phenotype of the last mutant was similar to the ag-1 mutant in A. thaliana (Yanofsky et al., 1990); its outermost whorl had four normal sepals, while its third whorl stamens and fourth whorl carpel were all replaced by a multitude of petals. The resulting flower mutant, agamous-like flower-1 (aglf-1), contained more than 85 petals ( Figures 8M–O ), in which BnaAG.A01, BnaAG.C01, and BnaAG.A03 with base insertion or deletion in target 2 caused a frameshift mutation, whereas BnaAG.C03 had a single base insertion in both the target squences, resulting in a frameshift mutation.

Previous studies showed that a total of 160, 91, and 307 MADS-box genes have been identified in B. rapa, B. oleracea, and B. napus genomes, respectively. However, there was no related research regarding MADS-box genes in B. nigra, B. juncea, and B. carinata genomes (Duan et al., 2015; Wu et al., 2018; Sheng et al., 2019). With the disclosure of the genome information of the remaining three Brassica species (B. nigra, B. juncea, and B. carinata), this study was the first to perform the systematic identification and classification of MADS-box genes in U’s triangle species. A total of 159 MADS-box genes were identified in the B. rapa genome, which is consistent with previous results (Duan et al., 2015). However, 180 MADS-box genes were investigated in the B. oleracea genome, which was approximately twice the number previously reported (Sheng et al., 2019), and 332 MADS-box genes were identified in the B. napus genome, which was 25 more than originally reported (Wu et al., 2018) ( Figure 1B ). The differences in the number of members may be confined to the reference genome quality of previously assembled Brassica species. The MADS-box genes can be divided into M-type and MIKC-type according to their phylogenetic relationships. The gene number ratio of the two types was 1.69 (66:39) in A. thaliana, 0.8 (71:88) in B rapa, 1.0 (92:92) in B. nigra, 0.87 (84:96) in B. oleracea, 0.87 (112:128) in B. juncea, 0.75 (143:189) in B. napus, and 0.59 (125:210) in B. carinata, indicating that the ratio of M-type and MIKC-type genes decreased gradually compared with that of A. thaliana. Moreover, the genome of the three diploid U’s triangle species has undergone genome-wide triploid events during the evolutionary process (Wang et al., 2011; Woodhouse et al., 2014), leading to a significantly higher ratio of MIKC-type genes in A. thaliana and three diploid species (A. thaliana:B. rapa:B. nigra:B. oleracea = 39:88:92:96) than that of M-type genes (A. thaliana:B. rapa:B. nigra:B. oleracea = 66:71:92:84). The function of M-type genes in plants is poorly understood, whereas MIKC-type genes are implicated in plant growth and development and environmental adaptation (Kofuji et al., 2003; Yang et al., 2016; Guo et al., 2021). Therefore, the significant expansion of MIKC-type genes in the U’s triangle species may be an adaptive evolution.

Public transcriptome data helps effectively identify valuable genes. In this study, a large number of tissue-specific MIKC-type BnaMADS-box genes were discovered in various tissues and during embryo development processes in B napus, of which BnaMADS1 and BnaMADS297 were highly expressed in roots ( Figure 5A ). AtAGL14 (XAL2, AT4G11880), a homologous gene of BnaMADS297 in A. thaliana, which binds to the CArG box (a cis-regulatory element of PIN1 and PIN4), regulates root elongation by upregulating the abundance of PIN protein (Garay-Arroyo et al., 2013). AGL21 (AT4G37940), a homolog of BnaMADS1, mainly regulates lateral root development, especially during lateral root growth under low-nitrogen conditions (Yu et al., 2014). BnaMADS111 is specifically expressed in seeds and highly expressed during the early stage of seed coat development (zygote, octant, and globular stage) ( Figure 5B ); its homolog AGL32 (TT16, AT5G23260) modulates the development of the sub-epidermal integument cell layer in A. thaliana seeds (Coen et al., 2017). BnMADS12 and BnMADS313 were specifically highly expressed only in flowers, and their homologous gene PI in A. thaliana acts as a floral homeotic gene mainly involved in the regulation of petal and stamen development (Goto and Meyerowitz, 1994). BnaMADS28 is highly expressed during the later stages of embryo and seed coat development, and its homologous gene AGL15 (AT5G13790) facilitates somatic embryogenesis in A. thaliana (Paul et al., 2022). BnaMADS161 is highly expressed during the early stage of embryo development and throughout the seed coat development process, and its homologous gene AGL11 (STK, AT4G09960) in Arabidopsis is involved in seed and silique development (Di Marzo et al., 2022).

Plant responses to biotic and abiotic stresses reflect resilience to external challenges. Plants usually respond to drought stress by flowering early to complete their life cycle, a phenomenon called drought escape (Waadt et al., 2022). Transcriptome data analysis showed that some BnaMADS-box genes involved in flowering time are also highly expressed under drought stress, possibly regulating drought escape. For example, AGL20 (SOC1, AT2G45660), a homologous gene of BnaMADS227 in A. thaliana, is upregulated by ABF3 and ABF4 under drought stress, thus promoting flowering (Hwang et al., 2019). BnaMADS227 is not only upregulated under drought stress but also induced by salt, ABA, and cold stress, indicating that BnaMADS227 participates in drought response and other abiotic stresses. In addition, BnaMADS2 and BnaMADS164 also respond to various abiotic stresses, such as dehydration and cold stress, and homologous AGL16 (AT3G57230) negatively regulates drought stress through stomatal density and movement in A. thaliana (Zhao et al., 2020). Thus, we speculated that these two genes may be involved in drought or cold stress. The expression of some genes, such as BnaMADS118, BnaMADS113, and BnaMADS321, was high in resistant strains and low in susceptible strains; however, the homologs of these genes in Arabidopsis are not reportedly involved in biotic stress. Exploring the mechanism of these genes in powdery disease can enrich the function of MIKC-type BnaMADS-box genes.

Polyploid plants enriched the floral organ phenotype due to the complex gene copy number but interfered with the systematic study of floral organ development. AG is responsible for stamen identity, carpel identity, and floral meristem determination (Theissen et al., 2000; Dreni and Kater, 2014). In this study, CRISPR/Cas9 technology was used to knock-out four BnaAG genes, and four lines with significant phenotypic differences in the T0 generation were obtained: wtl, feml, aglf-2, and aglf-1 ( Figure 8 ; Figure S3 ). The BnaAG.A01 was not edited in wtl, suggesting that the presence of only one copy of BnaAG can maintain the normal development of floral organs. However, the amino acid mutation of BnaAG.A01 did not cause a termination mutation, and the frameshift mutation occurred on the other three homologous genes, resulting in the feml phenotype. In addition, termination mutations of four BnaAG homologs in aglf-2 and aglf-1 resulted in phenotypes similar to those in Arabidopsis (Yanofsky et al., 1990). These results confirmed that BnaAG genes were involved in the development of third and fourth whorls in flowers, suggesting that there may be functional redundancy among the four BnaAG genes.

AG is responsible for the floral organ sex determination of plants (Zhang et al., 2020; Wang et al., 2021). The flowers of B. napus are monoecious, and the creation and application of male sterile materials in rapeseed can greatly enrich the utilization of heterosis. The feml phenotype, similar to the female flower, was found in the T0 generation after editing BnaAG. Although the carpel was deformed, a few seeds could be obtained by cross-pollination, implying that the feml was a female flower without stamens. Thus, BnaAG may be a sex-determining marker gene in B. napus.