Systematic Analysis of BELL Family Genes in Zizania latifolia and Functional Identification of ZlqSH1a/b in Rice Seed Shattering

The loss of seed shattering is an important event in crop domestication, and elucidating the genetic mechanisms underlying seed shattering can help reduce yield loss during crop production. This study is the first to systematically identify and analyse the BELL family of transcription factor-encoding genes in Chinese wild rice (Zizania latifolia). ZlqSH1a (Zla04G033720) and ZlqSH1b (Zla02G027130) were identified as key candidate genes involved in seed shattering in Z. latifolia. These genes were involved in regulating the development of the abscission layer (AL) and were located in the nucleus of the cell. Over-expression of ZlqSH1a and ZlqSH1b resulted in a complete AL between the grain and pedicel and significantly enhanced seed shattering after grain maturation in rice. Transcriptome sequencing revealed that 172 genes were differentially expressed between the wild type (WT) and the two transgenic (ZlqSH1a and ZlqSH1b over-expressing) plants. Three of the differentially expressed genes related to seed shattering were validated using qRT-PCR analysis. These results indicate that ZlqSH1a and ZlqSH1b are involved in AL development in rice grains, thereby regulating seed shattering. Our results could facilitate the genetic improvement of seed-shattering behaviour in Z. latifolia and other cereal crops.


Introduction
Seed shattering is an important trait that helps wild plants adapt to the natural environment and maintain population growth. In addition, seed shattering is considered one of the most important events in crop domestication [1]. Crop domestication began approximately 10,000 years ago [2][3][4] and was accompanied by a reduction in seed shattering, changes in seed shape, reduction of dormancy, increase in grain number, and improvement of plant type and fertility-all of which are important factors in artificial selection [5,6]. Research on the genetic regulatory mechanisms underlying plant seed shattering has important theoretical significance for revealing the molecular mechanisms underlying the domestication and improvement of wild plants. In addition, it has extremely high application value for the rapid domestication of wild plants and the enhancement of their yield [7,8].
In plants, the abscission zone (AZ) consists of plant organs and several adjacent cell layers separated via abscission. During plant organ shedding, the AZ differentiates into one or several layers of parenchyma cells to form an abscission layer (AL) [9]. The AL is In recent years, several domestic and international scholars have studied plant seed shattering [10,[27][28][29], providing an important reference for research on the genetic regulation of this behaviour [30,31]. Compared with wild rice, the loss of seed shattering is one of the biggest trait changes in cultivated rice and an important marker of the transition from wild to cultivated rice [8,32]. The form of the AL differs between rice cultivars: in cultivars prone to seed-shattering behaviour, the AL develops from the epidermis to near the vascular bundles; in contrast, cultivars with low seed-shattering behaviour have no AL or have an AL that is difficult to degrade. Cultivars with moderate seed-shattering behaviour either have a partially developed AL on the glumelle side of the branches or form a complete AL on the glumelle side and an irregular AL on the lemma side [11]. In these cultivars, the AL (composed of non-degradable tissue) is wider than that of cultivars whose seeds are easy to shed and narrower than that of cultivars whose seeds are difficult to shed. To date, 12 genes related to rice seed shattering have been cloned [33]. These genes encode transcription factors (TFs) or kinases that regulate seed shattering by affecting the development or degradation of the AL [32]. The qSH1 (LOC_Os01g62920) gene was identified by cloning the hybrid of the Indica rice cultivar 'Kasalath' (whose seeds shed easily) and the Japonica rice cultivar 'Nipponbare' (with reduced seed shedding). The qSH1 gene explains 68.6% of the phenotypic variation in this rice population, and it is considered the main quantitative trait locus/gene regulating AL development and seed abscission in rice. qSH1 encodes a BELL-type homeotic protein, and its coding sequence is identical between 'Kasalath' and 'Nipponbare'. However, a single-nucleotide polymorphism (SNP; G/T) in the 12-kb regulatory region upstream of the initiator codon of qSH1 changes the expression pattern of the protein-coding sequence in the AL region. Because of this change, the AL of Japonica rice plants cannot develop normally, resulting in the loss of seed shattering [27]. Gehring (1987) [34] found that mutations in some genes-such as the extremely conserved homeobox genes-can produce homologous phenotypes. Homeobox genes encode a type of transcription regulator called a homeodomain (HD), and TFs regulate the specific expression of plant genes. The TF family in plants includes NAC (NAM, ATAF, and CUC), MADS (MADS-box), ERF (ethylene response factor), and TALE (3-aminoacid-loop extension) [35][36][37][38]. The typical homeobox domain consists of 60 amino acids (except the TALE gene family, which encodes an atypical DNA-binding domain consisting of 63 amino acids) [39,40]. Based on differences in gene sequences and evolutionary history, the TALE gene family in plants can be further divided into two subfamilies: BELL (BEL1-like HD) and KNOX (KNOTTED1-like homeobox). With the rapid development of molecular biology, bioinformatics, and next-generation sequencing technology, genes in the KNOX and BELL families have been cloned and verified in different species [35]. The KNOX gene family is mainly involved in cell differentiation and maintenance of the shoot apical meristem [36], whereas the BELL gene family helps regulate plant development, hormone responses, and stress responses [41][42][43]. However, the number of members in the BELL gene family varies across plant species. For example, there are 4 BELL family genes in the bryophyte genome [38], 13 in Arabidopsis [44], 15 in maize [45], and 14 in rice and potato [46]. Nevertheless, the proteins encoded by these genes are highly conserved in structure, and all contain the SKY, BEL, and HD domains [47]. Various combinations of heterologous BELL-KNOX dimers with different activities are involved in regulating plant growth and development, including factors such as the maintenance of the shoot apical meristem and its boundaries, leaf development, and flowering [48][49][50]. In rice, the BELL genes qSH1 and SH5 participate in rice grain abscission by promoting the development of the AZ and inhibiting lignin synthesis. In addition, SHAT1 (shattering abortion 1) and SH4 (shattering 4) are essential regulators of AZ formation. Yoon et al. (2014) also reported that SH5 induces the expression of SHAT1 and SH4, promoting AL formation and seed shattering [51]. Yan et al. (2022) [52] performed genome collinearity analysis in Z. latifolia and rice to show that the ZlqSH1a (Zla04G033720) and ZlqSH1b (Zla02G027130) genes in Z. latifolia are orthologous to qSH1 (LOC_Os01g62920) in rice. To date, no studies have analysed the BELL family TFs in the Z. latifolia genome or reported the function of any BELL genes. This study is the first to perform bioinformatic analysis of BELL TFs in the entire Z. latifolia genome using subcellular localisation and RNA-seq. We verified the functions of ZlqSH1a and ZlqSH1b by over-expressing these genes in transgenic rice. In addition, we measured the breaking tensile strength (BTS) of ZlqSH1a and ZlqSH1b over-expressing (OE) and wild type (WT) seeds after maturity, observed the structural differences in AL tissue and fracture surfaces in transgenic and WT plants, and performed RNA-seq analysis of transgenic and WT plants to determine the influence of ZlqSH1a or ZlqSH1b over-expression on gene expression in the rice grain AL. We identified and verified the functions of ZlqSH1a and ZlqSH1b in rice and discussed their effects on the development of the AL in rice grains. Our results provide theoretical support and new genetic resources for selecting Z. latifolia cultivars with appropriate seed-shattering behaviour that makes them suitable for commercial cultivation.

Identification of ZlBELL Family TFs and Chromosome Localisation
A total of 48 ZlBELL genes were identified in the Z. latifolia genome. Based on their location on the chromosomes, they were sequentially named ZlBELL01-48 ( Figure 1, Table S1). Notably, ZlBELL05 (Zla02G027130) and ZlBELL13 (Zla04G033720) were orthologous to qSH1 (LOC_Os01g62920) in rice and were named ZlqSH1b and ZlqSH1a, respectively [52]. The chromosome localisation analysis of ZlBELL genes showed that the 48 ZlBELL genes were distributed on 14 chromosomes in the Z. latifolia genome. A total of 12 genes (accounting for 25% of all ZlBELL genes) were located on chromosome 5. Moreover, 14.58% (7) and 12.50% (6) of the genes were located on chromosomes 9 and 3, respectively. Chromosomes 16 and 17 contained the fewest ZlBELL genes (only one gene each, accounting for 2.08% of the total number). The cluster of ZlBELL genes was located in the middle of chromosome 12 and at the top of chromosome 9 in Z. latifolia. rice grains. Our results provide theoretical support and new genetic resources for selecting Z. latifolia cultivars with appropriate seed-shattering behaviour that makes them suitable for commercial cultivation.

Identification of ZlBELL Family TFs and Chromosome Localisation
A total of 48 ZlBELL genes were identified in the Z. latifolia genome. Based on their location on the chromosomes, they were sequentially named ZlBELL01-48 ( Figure 1, Table  S1). Notably, ZlBELL05 (Zla02G027130) and ZlBELL13 (Zla04G033720) were orthologous to qSH1 (LOC_Os01g62920) in rice and were named ZlqSH1b and ZlqSH1a, respectively [52]. The chromosome localisation analysis of ZlBELL genes showed that the 48 ZlBELL genes were distributed on 14 chromosomes in the Z. latifolia genome. A total of 12 genes (accounting for 25% of all ZlBELL genes) were located on chromosome 5. Moreover, 14.58% (7) and 12.50% (6) of the genes were located on chromosomes 9 and 3, respectively. Chromosomes 16 and 17 contained the fewest ZlBELL genes (only one gene each, accounting for 2.08% of the total number). The cluster of ZlBELL genes was located in the middle of chromosome 12 and at the top of chromosome 9 in Z. latifolia.  Table S1. ZlBELL05 (ZlqSH1b) and ZlBELL13 (ZlqSH1a) are marked in red.  Table S1. ZlBELL05 (ZlqSH1b) and ZlBELL13 (ZlqSH1a) are marked in red.

Phylogenetic Analysis of ZlBELL Genes
The protein sequences of 48 BELL genes in Z. latifolia (Table S1), 13 in Arabidopsis, and 14 in rice were used to analyse the BELL phylogenetic relationship between Arabidopsis, Z. latifolia, and rice. Compared with the six subfamilies of BELL genes in Arabidopsis (A, D, E, F, H, and I subfamilies) and seven subfamilies of BELL genes in rice (A, D, E, F, G, H, and J subfamilies), the 48 ZlBELL genes in Z. latifolia were distributed in nine subfamilies (A, B, C, D, E, F, G, H, and J subfamilies) ( Figure 2). The I subfamily had the largest number (five) of ZlBELL genes in Arabidopsis, and the J subfamily had the largest number (three) in rice. The C subfamily had the largest number (16) of ZlBELL genes in Z. latifolia but did not contain Arabidopsis and rice BELL genes. The E subfamilies had fewer ZlBELL members (only two), and subfamily I did not contain any ZlBELL genes. Notably, ZlBELL05 (ZlqSH1b) and ZlBELL13 (ZlqSH1a) belonged to the D subfamily.

Phylogenetic Analysis of ZlBELL Genes
The protein sequences of 48 BELL genes in Z. latifolia (Table S1), 13 in Arabidopsis, and 14 in rice were used to analyse the BELL phylogenetic relationship between Arabidopsis, Z. latifolia, and rice. Compared with the six subfamilies of BELL genes in Arabidopsis (A, D, E, F, H, and I subfamilies) and seven subfamilies of BELL genes in rice (A, D, E, F, G, H, and J subfamilies), the 48 ZlBELL genes in Z. latifolia were distributed in nine subfamilies (A, B, C, D, E, F, G, H, and J subfamilies) ( Figure 2). The I subfamily had the largest number (five) of ZlBELL genes in Arabidopsis, and the J subfamily had the largest number (three) in rice. The C subfamily had the largest number (16) of ZlBELL genes in Z. latifolia but did not contain Arabidopsis and rice BELL genes. The E subfamilies had fewer ZlBELL members (only two), and subfamily I did not contain any ZlBELL genes. Notably, ZlBELL05 (ZlqSH1b) and ZlBELL13 (ZlqSH1a) belonged to the D subfamily.

Gene Structure and Conserved Motif Analysis of Genes in the ZlBELL Family
Phylogenetic analysis of ZlBELL protein sequences revealed that they could be divided into eight subfamilies ( Figure 3A). Analysis of the intron/exon structure of each ZlBELL gene indicated that the exon number in ZlBELL genes varied from four to six; most of the genes (77.08%) contained four to five exons with a length of more than 1.6 kb. However, the intron number ranged from three to five, and did not vary greatly ( Figure 3B). Protein domain analysis was performed using the MEME software to elucidate the functional diversity of ZlBELL proteins in Z. latifolia. The results showed that ZlBELL proteins had a relatively conserved structure. Three relatively conserved motifs (Motifs 1-3) were identified, of which Motifs 1 and 2 were the most conserved and were shared across all subfamilies ( Figure 3C). In addition to the shared motifs, each subfamily showed some specificity; for example, Motif 3 was only present in subfamilies I, II, III, IV, and V. Notably, the typical sequence characteristics analysis of BELL family in Z. latifolia are shown in Figure S1. It can be seen from Figure 3 and Figure S1 that the conserved motif of the ZlBELL family did not completely overlap with its typical sequence.

Subcellular Localisation of the ZlqSH1a and ZlqSH1b Proteins
Considering the homology of ZlqSH1a (ZlBELL13), ZlqSH1b (ZlBELL05), and qSH1 (LOC_Os01g62920) [52], we chose the ZlqSH1a and ZlqSH1b for subcellular localization. To clarify the specific expression sites of the ZlqSH1a and ZlqSH1b proteins in cells, we used the PSORT program (http://psort.hgc.jp/) to predict whether they contained nuclear localisation signals. To test this prediction, we first constructed the ZlqSH1a-GFP and ZlqSH1b-GFP fusion protein over-expression constructs. The fusion expression vector, empty vector, and nuclear marker were co-transformed into rice protoplasts and cultured for 8-10 h under low light. Following this, the fluorescent signal was observed using confocal microscopy ( Figure 4). The red fluorescent (RFP) regions of the nuclear marker overlapped with the green fluorescent (GFP) regions of ZlqSH1a-GFP and ZlqSH1b-GFP, indicating that both ZlqSH1a-GFP and ZlqSH1b-GFP were localised in the nucleus ( Figure 4). That is, both proteins were nuclear TFs, which is consistent with the predicted results.

Seed Shattering Phenotype and AL-Based Histological Analysis of WT Plants and ZlqSH1a and ZlqSH1b Over-Expressing Rice
To verify the function of the ZlqSH1a/b gene, we constructed its over-expression vector and transformed it into a non-shattering rice cultivar, L422, to obtain positive transgenic rice plants. Phenotypic observations showed that at the harvest stage, ZlqSH1a and ZlqSH1b over-expressing plants had fewer seeds remaining in the rice panicle than WT plants. Thus, ZlqSH1a/b are speculated to be key genes for seed shattering in Z. latifolia. Thirty days after flowering and pollination, the BTS between the grain and the pedicel were measured in the WT and over-expressing lines using a digital tension meter. The results showed that the maximum tensile strength of the ZlqSH1a/b over-expressing lines was lower than that of the WT ( Figure 5A). Compared with the WT, the ZlqSH1a/b over-expressing lines showed significantly higher natural seed shedding after maturity ( Figure 5(B1,C1,D1)). Scanning electron microscopic (SEM) observations of the grain and pedicel fracture surfaces revealed that both ZlqSH1a and ZlqSH1b were involved in AL development and that the fracture surfaces were relatively smooth in the gene over-expressing plants ( Figure 5(C2-C5,D2-D5)). In contrast, the fracture surfaces of the WT were rougher and more fragmented ( Figure 5(B2-B5)). The results suggested that seed shattering after seed maturity was higher in ZlqSH1a and ZlqSH1b over-expressing plants, where the abscission was more complete than in the WT. To accurately determine the histological differences in the AL between ZlqSH1a/b over-expressing plants and the WT, we used laser scanning confocal microscopy (LSCM) to compare the longitudinal sections of spikelets at the flowering stage. Samples from the ZlqSH1a and ZlqSH1b over-expressing plants showed a complete AL composed of small flat parenchymal cells of equal diameter between the pedicel and grain. The longitudinal section showed continuous lines, indicating cell abscission between the pedicel and the grain ( Figure 5(C6,D6)). In contrast, no abscission was formed in the WT, and no single obvious line indicated cell abscission ( Figure 5(B6)). These anatomical features of the AL were observed 3-5 days after the rice plant flowered, and similar features were maintained in the mature grain. Considering the natural grain shedding behaviour of ZlqSH1a and ZlqSH1b over-expressing plants, the results of BTS and SEM analysis indicated that seed shattering was significantly higher in the transgenic plants than in the WT. This was consistent with the observations of the longitudinal AL sections, indicating that ZlqSH1a and ZlqSH1b were involved in AL tissue development.

Effects of ZlqSH1a and ZlqSH1b Over-Expression on the Transcriptome of the AL in Rice
To analyse the gene functions of ZlqSH1a and ZlqSH1b, we performed RNA-seq experiments using AZ tissues from WT and ZlqSH1a and ZlqSH1b over-expressing plants. The unit structure of the AL consists of a 1 mm pedicel region and a 1.5 mm grain. We analysed the differential expression of genes in the two groups of samples and visualised the statistically significant differences using a volcano plot ( Figure 6A,B). Hierarchical cluster analysis was performed using the selected differentially expressed genes (DEGs), such that genes with similar expression patterns were clustered ( Figure 6C,D). Compared with the WT, 327 DEGs were identified in ZlqSH1a over-expressing plants, of which 155 were up-regulated and 172 were down-regulated (Table S2). Compared with the WT, 696 DEGs were identified in ZlqSH1b over-expressing plants, out of which 323 were up-regulated and 373 were down-regulated (Table S3).
The two sets of DEGs shared 172 genes. Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) enrichment analyses were performed on these DEGs. The DEGs were searched against the GO database to classify the normalised gene function and to elucidate the cell components, biological processes, and molecular functions involved in AL development. Pathway annotation and enrichment analysis of DEGs helped further interpret the gene function. A total of 562 DEGs were assigned to three classes (Biological Process, Molecular Function, and Cellular Component) and 44 subclasses of GO ( Figure 6E,F). The main subclasses were "cell part", "cell", and "organelle" in the Cellular Component; "metabolic process", "cellular process", and "single-organism process" in Biological Process; and "binding", "catalytic activity", and "transporter activity" in Molecular Function. All DEGs were subjected to KEGG pathway enrichment analysis to understand the complex biological processes of the transcriptome ( Figure 6G,H). The results showed that 411 DEGs were annotated and enriched in several KEGG pathways, including "MAPK signalling pathway-plant", "Starch and sucrose metabolism", "Carbon metabolism", and "Plant hormone signal transduction". In this study, we mainly focused on "Carbon metabolism" and "Plant hormone signal transduction".  The two sets of DEGs shared 172 genes. Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) enrichment analyses were performed on these DEGs. The DEGs were searched against the GO database to classify the normalised gene function and to elucidate the cell components, biological processes, and molecular functions involved in AL development. Pathway annotation and enrichment analysis of DEGs helped further interpret the gene function. A total of 562 DEGs were assigned to three classes (Biological Process, Molecular Function, and Cellular Component) and 44 subclasses of GO ( Figure 6E,F). The main subclasses were "cell part", "cell", and "organelle" in the

Expression of Seed Shattering-Related Genes and Validation by qRT-PCR
As an important plant hormone, ethylene can regulate the abscission of seeds and flowers [53]. Moreover, increased polygalacturonase activity is related to cell separation during fruit shedding [54]. The previous results showed that ZlqSH1a and ZlqSH1b over-expressing plants were significantly different from WT plants in terms of seed-shattering-related morphology and histology. Based on the annotations of the two sets of DEGs, we selected three seed-shattering-related genes for qRT-PCR validation. The three candidate genes-including two down-regulated genes and one up-regulated gene-were regulated in the same direction in the ZlqSH1a and ZlqSH1b over-expressing plants. The following candidate genes were included: polygalacturonase 1 (PG1), polygalacturonase 2 (PG2), and ethylene-responsive transcription factor (ERF). All candidate genes were expressed in the AL of WT and transgenic plants. We used the rice housekeeping gene UBQ5 as the housekeeping gene to normalise the gene expression levels. Compared with the WT, the expression levels of PG1 and PG2 were significantly down-regulated in ZlqSH1a and ZlqSH1b over-expressing plants, whereas ERF was significantly up-regulated (p < 0.05) (Figure 7). This was consistent with the transcriptome sequencing results (Figure 6, Tables S2 and S3).

Full-Genome Analysis of BELL Family TFs in Z. latifolia
Studies have shown that approximately 300 million years ago, at least one BELL gene was already present in the common ancestor of extant spermatophytes and that this gene domain is highly conserved [55]. To date, members of the BEL1-like family of TFs have been identified in every plant species studied. With the advent of full-genome sequences, the breadth and potential functions of this key family of DNA-binding proteins can now be fully understood. Genes in the BELL family are involved in plant vegetative growth, fruit development, and seed abscission. In apples, the BELL1-like gene MDH1 is mainly expressed in developing and mature flowers and fruits. In Arabidopsis lines, the ectopic expression of MDH1 leads to dwarfing, reduced fertility, deformation of the carpel, and deformed silique folds [56]. In rice, the qSH1 and SH5 genes (homologs of the Arabidopsis

Full-Genome Analysis of BELL Family TFs in Z. latifolia
Studies have shown that approximately 300 million years ago, at least one BELL gene was already present in the common ancestor of extant spermatophytes and that this gene domain is highly conserved [55]. To date, members of the BEL1-like family of TFs have been identified in every plant species studied. With the advent of full-genome sequences, the breadth and potential functions of this key family of DNA-binding proteins can now be fully understood. Genes in the BELL family are involved in plant vegetative growth, fruit development, and seed abscission. In apples, the BELL1-like gene MDH1 is mainly expressed in developing and mature flowers and fruits. In Arabidopsis lines, the ectopic expression of MDH1 leads to dwarfing, reduced fertility, deformation of the carpel, and deformed silique folds [56]. In rice, the qSH1 and SH5 genes (homologs of the Arabidopsis PNY) participate in seed shattering by promoting AZ development and inhibiting lignin synthesis [51].
This study is the first to systematically identify BELL TFs in Z. latifolia. We identified 48 ZlBELL genes (Figure 1, Table S1). The chromosome localisation analysis revealed that the ZlBELL genes were unevenly distributed on the chromosomes of Z. latifolia. Some ZlBELL genes were located at the top and bottom of the chromosome (Figure 1), which is consistent with the findings of Sharma et al. (2014) [57]. This indicates that the ZlBELL genes underwent a certain degree of contraction and expansion. Introns and exons can evolve through gain, loss, insertion, or deletion [58]. Here, we found that the number of exons in the ZlBELL gene family ranged from four to six. Most of the genes (77.08%) contained four to five exons with a length of more than 1.6 kb, whereas the number of introns ranged from three to five and did not vary greatly. Analysis of the structure of this gene family revealed the length of the introns and exons encoding key functional domains and showed the degree of conservation of their splicing mode (Figure 3). The results indicated that in ZlBELL genes, the insertion or loss of introns and exons was relatively stable during the evolutionary process, resulting in relatively small differences in the number of exons and introns. This ensures stability in the biological functions of these genes to a certain extent. The conserved motifs of 48 ZlBELL genes in Z. latifolia were analysed, revealing three highly conserved amino acid motifs. Most of the conserved motifs in the same subfamily were similar, indicating stability in the function of the coding protein in each subfamily. Almost all ZlBELL proteins contained Motifs 1 and 2, which were always adjacent to each other and together constituted the ZlBELL domain. The uniqueness and conserved nature of these motifs in the same subfamily also supports the evolutionary classification of the ZlBELL gene family ( Figure 3C). The 48 ZlBELL genes in Z. latifolia were distributed in 9 subfamilies (Figure 3). The C subfamily contained the largest number of members (16), whereas the E subfamily contained only two members. This indicates that the members of this subfamily evolved relatively slowly, and we speculate that the function of this gene family is relatively conserved.

Histological Comparison of the AL between ZlqSH1a/ZlqSH1b Over-Expressing Plants and WT
Rice grains are shed at the AL, a junction between small branches. The AL consists of one or several layers of dense cells with similar morphology. Seed shattering in rice is a complex biological trait mainly controlled by the formation and fracture of the AL [28]. qSH1 is one of the major genes controlling seed shattering in rice [27], and the locus containing this gene (LOC_Os01g62920) explains 68.6% of the phenotypic variation in this species. Continuous expression of qSH1 (encoding a BELL-type homeoprotein) at the base of spikelets promotes the formation of the AL. As explained above, an SNP causes changes in the expression pattern of qSH1 in the abscission region of 'Nipponbare' plants, resulting in the loss of seed shattering. Yan et al. (2022) [52] performed genome collinearity and homology analysis in Z. latifolia and rice and found that the ZlqSH1a (Zla04G033720) and ZlqSH1b (Zla02G027130) genes in Z. latifolia were homologs of the qSH1 (LOC_Os01g62920) gene in rice. Based on this, we speculated that ZlqSH1a and ZlqSH1b might be related to seed shattering in Z. latifolia.
In the present study, we transformed ZlqSH1a and ZlqSH1b into the rice cultivar L422 (which does not exhibit seed-shattering behaviour) and constructed ZlqSH1a and ZlqSH1b gene over-expressing rice plants. We then used a digital tension meter to measure the BTS between the grain and pedicel of WT and ZlqSH1a and ZlqSH1b over-expressing plants at the same developmental period. The results showed that the maximum tensile strength was lower in ZlqSH1a/b over-expressing lines than in the WT ( Figure 5A). In addition, the number of naturally shed seeds after maturity was significantly higher in the ZlqSH1a/b over-expressing lines than in the WT ( Figure 5(B1,C1,D1)). SEM analysis showed that the phenotype of the AL was very different between ZlqSH1a and ZlqSH1b over-expressing lines and the WT. In particular, the fracture surfaces of the grain and pedicel were relatively smooth in ZlqSH1a and ZlqSH1b over-expressing plants ( Figure 5(C2-C5,D2-D5)) and rougher and more broken in the WT (Figure 5(B2-B5)). LSCM observation of the longitudinal section of the AL revealed a complete AL composed of small and flat parenchymal cells of equal diameter between the pedicel and grain of the ZlqSH1a and ZlqSH1b gene over-expressing plants. (Figure 5(C6,D6)). In contrast, no AL had been formed in the WT plants, and no obvious line indicated cell abscission (Figure 5(B6)). These results suggest that ZlqSH1a and ZlqSH1b affect seed shattering in rice by controlling the formation of the AL. To verify the soundness of this inference, we used RNA-seq to compare and analyse the gene expression levels of ALs in ZlqSH1a and ZlqSH1b gene over-expressing plants and the WT.

ZlqSH1a and ZlqSH1b Are Involved in the Regulation of AL Growth and Development
We selected two groups of DEGs based on RNA-seq analysis, which revealed that 172 genes were shared among the DEGs (Tables S2 and S3). GO and KEGG enrichment analysis of these DEGs can shed light on the cellular components, biological processes, and molecular functions involved in the development of the AL. Pathway annotation and analysis of DEGs can also help interpret the functions of the genes. We selected three genes related to seed shattering for qRT-PCR validation. Phytohormones are signal molecules produced by plants that play important roles in the regulation of plant growth and development. Ethylene-a phytohormone-regulates the shedding of flowers and seeds, and elevated levels of ethylene are often associated with tissue ageing and cellular stress [53]. Ethylene is an effective inhibitor of auxin, and the auxin content of the AZ regulates the sensitivity of this site to ethylene [59]. In the present study, expression levels of the ERF gene were significantly higher in ZlqSH1a and ZlqSH1b over-expressing plants than in the WT (Figure 7), suggesting that ERF may be related to enhanced seed shattering in the transgenic plants. In addition to plant hormones, some cell wall hydrolases also influence plant organ abscission. Some studies suggest that the increased activity of polygalacturonase is related to cell separation during fruit shedding. This is because polygalacturonase promotes the hydrolysis of pectin in the cell wall, thus changing its texture and hardness [54,60,61]. The AL of several plant species show increased polygalacturonase activity before the shedding of leaves, flowers, and fruits [62], and polygalacturonase-induced cell wall degradation is an important step in the loss of cell-cell adhesion in the abscission layer [59]. In the present study, GO functional annotation revealed that the candidate genes PG1 and PG2 were involved in regulating polygalacturonase activity. The expression levels of PG1 and PG2 were significantly lower in ZlqSH1a and ZlqSH1b over-expressing plants than in the WT, indicating that these genes are regulated by ZlqSH1a and ZlqSH1b and are involved in the development of abscission. The findings suggest that ZlqSH1a and ZlqSH1b are involved in AL development in rice grains and thereby regulate seed shattering. Hence, the ZlqSH1a and ZlqSH1b genes can be exploited as key target genes in molecular breeding to cultivate new Z. latifolia varieties resistant to seed shattering.

Samples
The Z. latifolia plant used in this study was collected from Baimahu Village in Jinhu County, Huai'an City, Jiangsu Province (33 • 11 9 N; 119 • 9 37 E). Oryza sativa cv. L422 was provided by Hefei Jian Gu Biotechnology Co., Ltd. Escherichia coli-competent DH5α and T vector were purchased from Sangon Biotech (Shanghai, China) Co., Ltd. The overexpression vector 1390-UBI was provided by Hefei Jian Gu Biotechnology Co., Ltd. (Hefei, China), and the pCAMBIA1390-ubiquitin over-expression vector was provided by Wuhan Edgene Biotechnology Co., Ltd. Equal amounts of soil and water were added to each plant box, and the experiment was repeated three times.

Full-Genome Analysis of BELL TFs in Z. latifolia 4.2.1. Identification of the BELL TFs Family in Z. latifolia
The protein sequence of rice BELL was downloaded from PlantTFDB (v5.0) (http: //planttfdb.gao-lab.org). The full genome data of Z. latifolia (genomic sequence, coding sequence, and protein sequence) were downloaded from the Genome Warehouse accessed on 4 December 2021 (https://ngdc.cncb.ac.cn/gwh/Assembly/22880/show). Using the rice BELL sequence as the query sequence, the candidate BELL gene in the Z. latifolia genome was identified using BLAST. Subsequently, the hidden Markov model (HMM) of the BELL family gene was downloaded from the PFAM database (http://pfam.xfam.org). Finally, the PFAM, NCBI conserved domains (http://www.ncbi.nlm.nih.gov/Structure/ cdd/wrpsb.cgi) and SMART (http://smart.embl-heidelberg.de/) databases were consulted to verify the presence of the BELL conserved domain.

Chromosome Localisation Analysis and Phylogenetic Analysis
The location information of all BELL genes in the Z. latifolia genome was extracted, and the online tool Map Gene 2 Chromosome (v2.0) (http:/mg2c.iask.in/mg2c_v2.0) was used to draw the chromosome localisation map of the ZlBELL family gene. The neighbourjoining method in MEGA7 was used to construct the phylogenetic tree for the protein sequences of Arabidopsis, Z. latifolia and rice with default parameters (Test of Phylogeny: Bootstrap method; No. of Bootstrap Replications: 1000; Gaps/Missing Data Treatment: Pairwise deletion; Multiple Sequence Alignment: MAFFT v 7.5).

Total RNA Extraction and Reverse Transcription
The total RNA of Z. latifolia leaves was extracted with the FastPure ® Universal Plant Total RNA Isolation Kit (RC-401, Vazyme, Nanjing, China) according to the manufacturer's instructions. The concentration of the extracted RNA was measured using an ultramicro spectrophotometer (OSE-260, Tiangen, Beijing, China). The extracted RNA was used as a substrate for reverse transcription using the PrimeScript TM II 1st Strand cDNA Synthesis Kit. The cDNA products obtained were stored at −20 • C for later use.

Amplification, Transformation, and Sequencing of ZlqSH1a and ZlqSH1b
The prepared cDNA was used as a template for PCR amplification with the following primers: ZlqSH1a-F and ZlqSH1a-R for ZlqSH1a; and ZlqSH1b-F and ZlqSH1b-R for ZlqSH1b (Table S4). The PCR product was subjected to agarose gel electrophoresis (AGE) to analyse the fragment sizes and to verify whether the presence of a heteroband. Subsequently, the PCR product was recovered by DNA precipitation; 10 µL of the mixture was added to 10 µL of Taq enzyme, heated at 72 • C for 15 min, and ligated to a ploy(A) tail. After AGE, the product was cut from the gel and recovered. The target gene was ligated to a T vector as follows: 0.5 µL of T vector, 4.5 µL of DNA, and 5 µL of the reaction buffer were added to a centrifuge tube and incubated at 4 • C overnight. The ZlqSH1a and ZlqSH1b genes were ligated to the T vector and transformed into E. coli-competent DH5α cells. Positive clones were screened and sequenced.

Subcellular Localisation of ZlqSH1a/ZlqSH1b
The ZlqSH1a/ZlqSH1b constructs and the empty vector PC2300S-GFP were propagated in a bacterial solution, and the plasmids were extracted. The plasmids of ZlqSH1a/ZlqSH1b and the empty vector PC2300S-GFP were used to co-transform the rice protoplast with the nuclear marker. After 16 h of incubation at 28 • C in the dark, GFP and RFP fluorescence was detected using an Olympus FV1000 laser scanning microscope (Olympus, Tokyo, Japan) at 488 and 543 nm.

Construction of ZlqSH1a/ZlqSH1b
Over-Expressing Rice 4.5.1. Construction of ZlqSH1a/ZlqSH1b Over-Expressing Carriers DNA fragments containing the full-length ZlqSH1a/ZlqSH1b genes were amplified with primers for seamless cloning and amplification (ZlqSH1a/ZlqSH1b-FP, ZlqSH1a/ZlqSH1b-RP) ( Table S4). The amplification products were detected by AGE, and the target fragments were cut under UV light and purified. The 1390-ubi vector was digested with the restriction enzyme BamHI, the digested product was separated by AGE, and the linearised 1390-ubi large fragment was recovered. After recovery, the digested ZlqSH1a/ZlqSH1b gene fragment was ligated with the 1390-ubi vector. A 5 µL aliquot of the product was used to transform DH5α-competent cells. The next day, monoclonal colonies were selected, and the PCR primers 1390-FP and ZlqSH1a/ZlqSH1b-RP were designed for the bacterial solution (Table  S4). The monoclonal colonies with positive PCR results were sent for sequencing-based verification. The single colonies with correct sequencing results were placed in LB culture medium containing kanamycin and shaken overnight at 37 • C at 200 r/min. The overexpression vector was extracted for vector transformation.

Agrobacterium-Mediated Transformation of Rice
The activated Agrobacterium strain EHA105 and previously prepared rice calluses were co-cultured to facilitate infection with Agrobacterium. A 25 mL aliquot of the Agrobacterium resuspension was added to the callus and the solution was incubated for 15 min, during which the solution was shaken gently from time to time. After soaking, the resuspension was poured out, and the callus was placed in a Petri dish lined with several sterile filter papers (which absorbed the residual bacterial liquid on the surface of the callus). Next, the disposable Petri dish was lined with three sterile filter papers, and 2.5 mL of inoculating culture medium was added. The dried callus was spread evenly on the filter paper and cultured at 23 • C for 48 h in the dark. The co-cultured callus was spread evenly and sparsely in the recovery culture medium. After approximately 5 days, the callus was transferred to the screening culture medium. After 3-4 weeks, 3-5 small, fresh, resistant callus particles in good growth condition were selected from each independent transformant and transferred to the regenerative culture medium. When the seedlings grew to 2-5 cm in length, one seedling in good growth condition from each independent transformant was transferred to a root-based culture medium. The culture conditions from recovery to root-based culture were as follows: incubation at 30°C under a light incubator with a 16 h light/8 h dark photoperiod.

Identification of ZlqSH1a/ZlqSH1b Over-Expressing Rice Plants
The following primers were designed for each promoter region to identify positive transformation by the over-expression vectors: ZlqSH1a-F and ZlqSH1a-R for ZlqSH1a and ZlqSH1b-F and ZlqSH1b-R for ZlqSH1b (Table S4). The PCR reaction mixture was prepared as follows: KOD One PCR Master Mix, 25 µL; 1390-FP, 2 µL; RP-1390, 2 µL; template, 1 µL; ddH 2 O, 20 µL. The PCR program was as follows: 94 • C for 5 min; followed by 30 cycles of 94 • C for 30 s, 60 • C for 30 s, and 72 • C for 30 s; followed by 72 • C for 5 min ( Figure S2).

Measurement of Seed Shattering
The flowering time of the rice was set after the heading had occurred. This was done to accurately quantify the changes in the seed-shattering phenotype of T2 generation of the ZlqSH1a/ZlqSH1b over-expressing rice plants A digital tension meter was used to measure the BTS between the grains and pedicels of the WT and over-expressing lines at the same period 30 days after flowering and pollination. A small clip was suspended on the small hook of the tension meter, and the clip was clamped to the middle part of the seed glume and pulled horizontally so that the seed and the branch separated. At this time, the tension meter showed the maximum tensile strength required in the process, and this value was recorded. The measurement was repeated 50 times, and the average value was recorded as the tensile strength of the seed.

Laser Scanning Confocal Microscopy
Approximately 20 spikelets were sampled from each plant at the flowering stage to observe the structure of the AZ. A longitudinal section was made manually through the junction between the flower and pedicel, and the sections were stained with acridine orange. The sections were observed using a Leica SP8 laser scanning microscope (Leica SP8; Leica Biosystems, Nussloch, Germany) at 488 and a 543 nm.

Scanning Electron Microscopy
Thirty days after pollination, the florets (grains) in the middle area of the inflorescence were selected. The grains (approximately 2-4 mm in length) were cut from the pedicel near the branches and stored at 4 • C in the fixative solution used for electron microscopy. The fixed samples were taken out, and the connection between the grain and the pedicel was gently broken with tweezers. The samples were glued to the metal block and observed under a JEOL JSM-840 scanning electron microscope (JEOL Ltd., Tokyo, Japan).

RNA-seq Analysis
The AL of WT and ZlqSH1a/ZlqSH1b over-expressing plants were collected 3-5 days after flowering, and RNA was extracted from the AL tissue. This step was performed for three biological replicates, with three plants in each replicate. Paired-end libraries were constructed and sequenced using Illumina HiSeq 2500. The raw reads were mapped to the reference genome (Os-Nipponbare-Reference-IRGSP-1.0, MSU7) using TopHat2 with the default parameters [63]. Cuffdiff was used to calculate the fragments per kilobase of exon of the mapped reads per million genes, and the DEGs (fold change ≥ 2) between WT and ZlqSH1a/ZlqSH1b over-expressing plants were identified [64]. The DEGs were functionally analysed using the agriGO and KEGG databases [65,66].

Data Validation by qRT-PCR
Three seed-shattering-related genes were analysed in the ZlqSH1a/ZlqSH1b overexpressing and WT plants using qRT-PCR analysis. cDNAs were reverse transcribed using residual RNA from the RNA-seq analysis. The synthesised cDNAs were amplified using SYBR Premix Ex Taq (TaKaRa Biotechnology Co., Ltd., Dalian, China). Quantitative RT-PCR was performed using the SYBR Green QPCR mix (Bio-Rad Laboratories, Hercules, CA, USA), and the UBQ5 gene (NCBI accession number: LOC4341684) was used as an endogenous control to normalise gene expression. The cycling conditions included incubation for 30 s at 95 • C, followed by 40 cycles of amplification (95 • C for 5 s and 60 • C for 30 s). All samples were repeated at least three times. The primers used in the qRT-PCR are listed in Table S4.

Statistical Analysis
The experiments were arranged in a randomised block design. Data were collected from each experimental group containing different numbers of plants, and the mean ± SD was calculated for the total number of plants in each experiment. Data were analysed with analysis of variance (ANOVA) in IBM SPSS 18 IIBM Corp., Armonk, NY, USA) statistical software. Tukey's multiple comparison test was used at a significance level of p < 0.05 to determine any statistically significant differences between treatments.

Conclusions
The present study explored the functions of two genes in the BELL family of TFs (ZlqSH1a and ZlqSH1b) in Z. latifolia. We found that over-expression of ZlqSH1a and ZlqSH1b was responsible for seed-shattering behaviour after the maturation of rice grains. Our findings shed light on the mechanisms underlying seed shattering in a crop plant and can facilitate the genetic improvement of Z. latifolia and other cereal crops.