Genome‐wide analysis of the callose enzyme families of fertile and sterile flower buds of the Chinese cabbage (Brassica rapa L. ssp. pekinensis)

Callose is a β‐1,3‐glucan commonly found in higher plants that plays an important role in regulating plant pollen development. It is synthesized by glucan synthase‐like (GSL) and is degraded by the enzyme endo‐1,3‐β‐glucosidase. However, genome‐wide analyses of callose GSL and endo‐1,3‐β‐glucosidase enzymes in fertile and sterile flower buds of Chinese cabbage have not yet been reported. Here, we show that delayed callose degradation at the tetrad stage may be the main cause of microspore abortion in Chinese cabbage with nuclear sterility near‐isogenic line ‘10L03’. Fifteen callose GSLs and 77 endo‐1,3‐β‐glucosidase enzymes were identified in Chinese cabbage. Phylogenetic, gene structural and chromosomal analyses revealed that the expansion occurred due to three polyploidization events of these two gene families. Expression pattern analysis showed that the GSL and endo‐1,3‐β‐glucosidase enzymes are involved in the development of various tissues and that the genes functionally diverged during long‐term evolution. Relative gene expression analysis of Chinese cabbage flowers at different developmental stages showed that high expression of the synthetic enzyme BraA01g041620 and low expression of AtA6‐homologous genes (BraA04g008040, BraA07g009320, BraA01g030220 and BraA03g040850) and two other genes (BraA10g020080 and BraA05g038340) for degrading enzymes in the meiosis and tetrad stages may cause nuclear sterility in the near‐isogenic line ‘10L03’. Overall, our data provide an important foundation for comprehending the potential roles of the callose GSL and endo‐1,3‐β‐glucosidase enzymes in regulating pollen development in Chinese cabbage.

Callose is a b-1,3-glucan commonly found in higher plants that plays an important role in regulating plant pollen development. It is synthesized by glucan synthase-like (GSL) and is degraded by the enzyme endo-1,3-bglucosidase. However, genome-wide analyses of callose GSL and endo-1,3-bglucosidase enzymes in fertile and sterile flower buds of Chinese cabbage have not yet been reported. Here, we show that delayed callose degradation at the tetrad stage may be the main cause of microspore abortion in Chinese cabbage with nuclear sterility near-isogenic line '10L03'. Fifteen callose GSLs and 77 endo-1,3-b-glucosidase enzymes were identified in Chinese cabbage. Phylogenetic, gene structural and chromosomal analyses revealed that the expansion occurred due to three polyploidization events of these two gene families. Expression pattern analysis showed that the GSL and endo-1,3-bglucosidase enzymes are involved in the development of various tissues and that the genes functionally diverged during long-term evolution. Relative gene expression analysis of Chinese cabbage flowers at different developmental stages showed that high expression of the synthetic enzyme BraA01g041620 and low expression of AtA6-homologous genes (BraA04g008040, BraA07g009320, BraA01g030220 and BraA03g040850) and two other genes (BraA10g020080 and BraA05g038340) for degrading enzymes in the meiosis and tetrad stages may cause nuclear sterility in the near-isogenic line '10L03'. Overall, our data provide an important foundation for comprehending the potential roles of the callose GSL and endo-1,3-b-glucosidase enzymes in regulating pollen development in Chinese cabbage.
Pollen development is an extremely important biological process in flowering plants and an indispensable life process for plant genetic breeding [1]. Pollen development abnormalities make the formation of functional pollen difficult, leading to male sterility, which seriously affects the reproduction and evolution of plants [2]. As the utilization of crop heterosis continues to increase, male sterility has become more widely used in crop breeding practices to make it easier to achieve heterosis in crops that use hybrid pollination, which can greatly increase crop yields and resistance [3]. Chinese cabbage has two types of male sterility: cytoplasmic male sterility (CMS) and genetic male sterility (GMS) [4]. Previous studies [5] have shown that CMS is the result of gene interaction between the nuclear and cytoplasmic genomes, and the interaction between the nuclear genome and mitochondrial genome leads to nuclear-cytoplasmic inconsistency, thus disturbing the normal metabolic process [6]. The main cytological characteristic is tapetum cell death, and the physiological response is the disorder of mitochondrial energy metabolism. GMS inheritance is relatively simple, generally controlled by one or several pairs of nuclear genes, and has nothing to do with cytoplasmic genes. But a population of 100% male sterile plants could not be found when a test cross was employed in selecting its maintainer line, whether GMS is controlled by dominant or recessive genes [7]. The male sterility could be maintained at only 50% by the sibling in an AB line in which the separation ratio of sterile and fertile plants is 1 : 1. There are two kinds of AB lines, AB line 1 and AB line 2. Twenty-five percent of the progeny from selfing of male fertile plants in an AB line 1 are male sterile, and all progeny from selfing of male fertile plants in an AB line 2 are male fertile [8]. Fertile plants must be identified and rogued before flowering when AB lines are utilized to produce hybrid seed, which is more costly and labor intensive. GMS lines with a 100% sterility rate were obtained for the first time by crossing between male sterile plants in AB line 1 and male fertile plants in AB line 2 by Zhang et al. in 1990 [9]. In 1996, the multiple allele hypothesis was raised to explain the inheritance of this male sterility, which was controlled by three alleles at a single locus including male sterile allele 'Ms', male fertile allele 'ms' and a restorer allele 'Ms f '. The dominantrecessive relationship of these three alleles was Ms f >Ms>ms [10]. Recessive GMS is sensitive to environment, while dominant GMS is insensitive to environment [11]. Compared with CMS, GMS has the advantages of good stability, incompleteness, no negative impact on cytoplasm, and wide distribution of fertility restoration genes. Studies of the mechanism of nuclear sterility in Brassica napus [12], tobacco (Nicotiana tabacum L.) [13,14] and wheat (Triticum aestivum L.) [15] have shown that most plant sterility originates from the abnormal development of pollen at different stages. Because of the significant production and theoretical research value of male sterility, it has long been a subject of interest. However, pollen development is a very complex process involving the regulation and expression of many genes. Mutation or abnormal expression of any of the key genes may cause abnormal pollen development and lead to male sterility. Indepth studies of pollen development-related genes and their molecular regulatory mechanisms will provide an important theoretical foundation for understanding plant male sterility and creating superior male sterile lines.
Callose is an important feature of sexual reproduction in plants, especially in sexual cells that are about to divide, as it forms the mother cell wall and the zygote wall of large spores [16], and plays a protective role by providing mechanical isolation, a chemical barrier, or a molecular sieve, and prevents cell aggregation [17]. Callose prevents agglomeration and fusion between cells when microspores are released, protects developing microspores from premature expansion and rupture, and acts as a 'mold' for the development and shaping of pollen outer wall patterns [16,18]. The callose of the external body wall of the plant anther tetrad plays an important role in the development and maturation of the microspores [13]. Decomposition at the proper time is critical; premature callose decomposition leads to incomplete growth of the microspores, and a lack of decomposition or delayed decomposition results in problems with the normal release of the microspores [14,19], which can lead to pollen sterility.
Studies on callose have shown that the main component is b-1,3-glucan, which is synthesized by the enzyme glucan synthase-like (GSL) and degraded by endo-1,3-b-glucosidase enzymes [20,21]. Twelve GSL genes (AtGSL1-AtGSL12) have been found in Arabidopsis that are involved in the synthesis of callose [22]. AtGSL2 is a major callose synthesis gene in pollen development [23]. Knock-out mutation of AtGSL2 has little effect on vegetative growth, but severely affects pollen development and callose deposition in pollen tube germination, resulting in the callose wall being completely absent and collapse of the pollen wall [24]. Overexpression of AtGSL2 in Arabidopsis resulted in abnormal deposition of callose during microsporogenesis and promoted early release of pollen before flowering [25]. AtGSL1 and AtGSL5 are not required for callose synthesis in pollen mother cells, but are essential for the synthesis of tetrad callose. The mutant gsl1-1/+gsl5-2/gsl5-3 exhibited tetrad callose depletion, tetrad malformation, and incomplete isolation of microspores [26]. AtGSL10 and AtGSL8 cause callose deposition on cell plates, cell walls and intercellular filaments. The AtGSL8 mutant shows delayed callose deposition at the cell plate and suffers from cytokinesis defects that cause seedling death [27]. The microspores of AtGSL10 and AtGSL8 mutants are unable to enter mitosis, which eventually leads to male gametophyte death [28]. In rice, the OsGSL5 gene plays a regulatory role in the late meiotic stage. In OsGSL5 mutants, the callose synthesis catalyzed by GSL5 is greatly reduced, and the development of mitotic and tetrad abnormalities leads to a decrease in the fertility of the mutant plants [29]. Plant anther endo-1,3-b-glucosidase genes have currently only been cloned in Arabidopsis (AtA6) and B. napus (BnA6) and are localized in tapetum cells [21]. Endo-1,3-b-glucosidase enzymes are highly expressed just before the disintegration of the tetrads and then disappear rapidly [18]. This shows that endo-1,3-b-glucosidase enzymes play an important role in the development of anthers in plants and are closely related to the development and maturation of male gametes. However, to our knowledge, a genome-wide analysis of callose GSL and endo-1,3-b-glucosidase enzymes in fertile and sterile flower buds of Chinese cabbage has not yet been reported.
In this study, we observed callose degradation by paraffin sections and light microscopy, then we identified 15 callose GSL and 77 endo-1,3-b-glucosidase enzymes in Chinese cabbage by genome-wide analysis of the callose enzyme families. Furthermore, we performed phylogenetic, gene structure, chromosomal location, expression profiling, and relative expression analyses. Our results provide an important foundation for understanding the potential roles of callose GSL and endo-1,3-b-glucosidase enzymes in regulating pollen development in Chinese cabbage.

Plant materials
Chinese cabbage nuclear sterility near-isogenic line (NIL) '10L03' was used in this study. It was bred from saturated backcross progenies of the inbred '02s005' (backcross parent, Ms f Ms f genotype) and the male sterile line (Msms genotype). Genotypes of fertile and sterile plants are Ms f Ms and MsMs respectively in '10L03'. Siblings between sterile and fertile plants of '10L03' could reproduce, and sterile and fertile plants in their progeny maintain 1 : 1 separation. '10L03' was cultivated in the experimental plot in Yunnan Agriculture University (Kunming, Yunnan Province, China). Fertility of plants was identified when flowering. Buds from fertile and sterile plants were collected for paraffin embedding. With regard to Arabidopsis pollen development [30], stages were identified and microspore developmental stages of different bud sizes were investigated by microscopy. Fertile and sterile buds at meiosis (stage 6, fertile: 1.40 9 1.10 mm, sterile: 1.30 9 1.00 mm), tetrad (stage 7, fertile: 2.10 9 1.30 mm, sterile: 2.00 9 1.20 mm), mononuclear (stage 8, fertile: 2.80 9 1.60 mm, sterile: 2.60 9 1.40 mm), and mature (stage 13, fertile: 6.00 9 2.70 mm, sterile: 5.30 9 2.20 mm) were chosen and stored at À80°C for RNA extraction.

Paraffin sectioning
Fertile and sterile buds were fixed in FAA fixative solution (every 100 mL of which contains 90 mL of 50% ethanol, 5 mL of acetic acid, and of 5 mL formalin) for 24 h at room temperature, which was followed by dehydration in increasing grades of ethanol. Ethanol and xylene in a volume ratio of 2 : 1, 1 : 1 and 1 : 2 were used for clearing, and the samples were embedded in paraffin. Paraffin sections were cut to a thickness of 8 lm on an semi-automatic microtome (YD-335, Jinhua Yidi Medical Equipment Co., Ltd., Jinhua, China) followed by dewaxing with xylene. Sections were stained with haematoxylin and eosin, visualized with a Leica DM2000 (Beijing, China) optical microscopy and photographed.

Aniline blue staining for callose
For callose staining, transverse anther sections and microspores released from the anthers were stained with 0.1% aniline blue solution in 0.077 M phosphate buffer (pH 8.5) [31] for about 2 h at room temperature and visualized using a Leica fluorescence microscope.
The GenBank non-redundant protein database was used to search against the candidates. DNAMAN software (LynnonBiosoft, San Ramon, CA, USA) was used for the homology analysis between B. rapa and Arabidopsis. WOLF PSORT (http://wolfpsort.org) [32] was used to predict protein subcellular localization. The TMHMM server (http://www.cbs.dtu.dk/services/ TMHMM/) was used to estimate the number of transmembrane helical (TMHs) domains. The molecular mass, theoretical pI and grand average of hydropathicity (GRAVY) were calculated using the tool of EXPASY [33] (http://web.expasy.org/protparam/).

Phylogenetic analyses of the callose GSL and endo-1,3-b-glucosidase enzymes
The full-length Chinese cabbage callose GSL and endo-1,3-b-glucosidase protein sequences were aligned using the program MAFFT 7.0, and phylogenetic reconstruction used the neighbor-joining method with MEGA7 software [34]. Bootstrap values for each branch were estimated (with 1000 replicates) to assess the relative support. A diagram of the intron/exon structures of the callose GSL and endo-1,3-b-glucosidase genes was generated using the online Gene Structure Display Server (http://gsds.cbi.pku.edu.cn/) [35]. Subsequently, the MEME program was used to search for conserved motifs in the B. rapa callose GSL and endo-1,3-b-glucosidase protein sequences [36].
Expression profiles of callose GSL and endo-1,3b-glucosidase genes in B. rapa Transcriptome data for four tissues in B. rapa were obtained from previous research [37] and reanalyzed. From the samples were obtained the transcriptome data of callose GSL genes at different floral developmental stages in Arabidopsis [38]. The expression levels of the callose GSL and endo-1,3-b-glucosidase genes were calculated using values of fragments per kilobase of exon model per million mapped reads in the root, stem, leaf, and flower in B. rapa. Finally, the expression data were normalized gene-wise and hierarchically clustered based on Pearson's coefficient with average linkage in the GENESIS (version 1.7.6) program [39].

RNA extraction and quantitative real-time PCR
Total RNA was extracted from Chinese cabbage flowers at four developmental stages using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) following the manufacturer's instructions. RNA quality was determined with a Nano-Drop ND1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). A total of 2 lg of total RNA per sample was reverse-transcribed using oligo (dT) and Superscript II reverse transcriptase (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). All of the primers (Tables S1 and S2) were designed by PRIMER-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/ index.cgi?LINK_LOC=BlastHome), using the following parameters: 150-200 bp PCR product size, Nr database, 58-62°C primer melting temperature, and B. rapa as the organism (taxid: 3711). All PCR reactions were performed under the following conditions: 40 cycles of 5 s at 94°C, 15 s at 60°C, and 34 s at 72°C. Using FastStart Universal SYBR Green Master (Rox; Roche, Indianapolis, IN, USA) and a 7500 Sequence Detection System (Thermo Fisher Scientific), quantitative real-time PCR was conducted in triplicate with different cDNAs synthesized from three biological replicates of different tissues and development stages. For each analysis, a linear standard curve, the threshold cycle number versus log (designated transcript level) was constructed using a serial dilution of a specific cDNA standard. The levels of the transcript in all unknown samples were determined according to the standard curve. Brassica rapa tubulin b-2 chain-like (LOC103873913) was used as an internal standard, and t-test statistical analysis was performed using the software IBM SPSS STATISTICS 20.0 (IBM Corp., Hong Kong, China) [40].

Phenotypic characterization of fertile and sterile floral buds
Different developmental stages (meiosis, tetrad, mononuclear and mature stages) of floral buds are shown in Fig. 1. Sterile flowers were visually smaller than fertile flowers, and no pollen grains were observed in the sterile flower buds (Fig. 1). During the developmental process, the anthers and filaments of the sterile flowers remained shorter than those of the fertile flowers.
An investigation using paraffin sections of fertile and sterile anthers is shown in Fig. 2 [30]. Microspores developed normally ( Fig. 2A-C, stage 6, 7, 8) and numerous mature pollen grains could be observed (Fig. 2D, stage 13) in fertile anthers. Compared with fertile anthers (Fig. 2A,B), in sterile anthers there were no obvious differences in the cellular morphology of the epidermis, endothecium, middle layer, and microsporocytes at stage 6 ( Fig. 2E), and at stage 7 normal tetrads could be formed (Fig. 2F), but we could observe vacuolization in tapetal cells of sterile anthers at stage 7 (Fig. 2F). Sterile tetrads were malformed and abortion occurred while the vacuolization of tapetal cells was increased (Fig. 2G) compared with stage 7. The pollen sac abortion was different in one anther (Fig. 2H,I). Some tapetal cells enlarged and occupied locules and some degraded (Fig. 2). Some tetrads seemed normal, others were abortive (Fig. 2H). Tetrads could not develop into mononuclear microspores (Fig. 2) and some became abortive microspores without viability (Fig. 2K) while the tapetum structure finally disappeared (Fig. 2J). At the late pollen developmental stage, sterile anther locules had sequentially shrunk and shriveled (Fig. 2L). We could not observe any microspores released from tetrads ( Fig. 2G-L), and the possible reason for microspore abortion was that callose surrounding tetrads could not degrade in time and tetrads could not develop into mononuclear microspores further.

Abnormal callose deposition during the abortion of microspores in Chinese cabbage
To test our hypothesis, callose deposition was examined during microspore development in fertile and sterile anthers. Results are shown in Fig. 3, where panels F1-F4 show normal fluorescence signals emitted by callose at meiosis, tetrad, mononuclear and mature pollen stages in fertile microspores. Some short linear fluorescence emitted by pollen mother cells could be observed at the meiosis stage from fertile anthers (Fig. 3, F1). Fluorescence signals were the strongest at tetrad stage, and appeared on the surface and the template of tetrads (Fig. 3, F2). Then, the callose degraded, the fluorescence signals on the surface and the template of tetrads disappeared, and the tetrads released mononuclear microspores. After the tetrads had released mononuclear microspores, callose accumulated on the pollen exine and fluorescence signals were detected there (Fig. 3, F3 and F4). Compared with fertile anthers, sterile anthers showed similar callose deposition to fertile anthers at meiosis and tetrad stage, but the tetrads were malformed (Fig. 3, F1 and S1; F2 and S2). However, the fluorescence response was stronger in sterile than in fertile anthers at mononuclear stages (Fig. 3, F3 and S3) and was expressed similar to the tetrads stage (Fig. 3, S2 and S3), indicating that the callose on the surface and the template of tetrads in sterile a anther could not be degraded in time. Microspores were not released and gradually broke down, and no mature pollen was produced ( Fig. 3, S4). Until anther complete abortion, we could still detect fluorescence on aborted microspores (Fig. 3, S4), indicating that callose accumulation accompanied microspore abortion.

Identification of the callose GSL and endo-1,3-bglucosidase enzyme families in Chinese cabbage
To determine which gene was responsible for the abnormal degradation of callose, a genome-wide analysis of the GSL and endo-1,3-b-glucosidase enzyme families was performed in Chinese cabbage. We used 12 GSL and 25 endo-1,3-b-glucosidase coding sequences from Arabidopsis as queries to search against the published genome of Chinese cabbage. After the search, we identified 15 GSLs and 77 endo-1,3-b-glucosidases (Tables 1  and 2). The full-length GSL proteins in Chinese cabbage ranged from 1768 (BraA09g025290) to 2290 (BraA07g023720) amino acids and shared a high level of similarity to the Arabidopsis thaliana proteins. In terms of biochemical properties, all of the GSL proteins were alkaline, with pI values ranging from 8.55 (BraA05g-008500) to 9.24 (BraA05g012120). The polypeptides were also predicted to contain 14-20 TMHs. Furthermore, we predicted the probable protein localization for each of the candidate callose GSL and endo-1,3-b-glucosidase enzymes in Chinese cabbage using the protein subcellular localization prediction software WOLF PSORT (http://wolfpsort.org). All candidate callose GSL enzymes were most likely localized in the plasma membrane ( Table 1). The endo-1,3-b-glucosidases identified in our study had isoelectric points (pI) ranging from 4.61 (BraA07g008730) to 9.73 (BraA07g022070), with coding sequences of 172-962 amino acids (Table 2). These ranges suggested the biochemical properties of the endo-1,3-b-glucosidase enzyme family in Chinese cabbage were diverse. All candidate callose endo-1,3-bglucosidase enzymes were most likely localized in the chloroplast and plasma membrane ( Table 2).

Phylogenetic and gene structural analyses of the callose GSL and endo-1,3-b-glucosidase enzymes in Chinese cabbage
We generated phylogenetic trees for each family to gain insight into the phylogenetics of both the GSL and endo-1,3-b-glucosidase enzyme families. We also compared the evolutionary relationships between Chinese cabbage and Arabidopsis GSLs and endo-1,3b-glucosidase enzymes (Figs 4 and 5). All GSLs in Chinese cabbage were clustered closely with AtGSLs. In addition, we compared intron-exon numbers, lengths, distribution and other genetic structures to understand the evolutionary history of the gene families [41]. The results showed that all the GSLs in Chinese cabbage were intron-rich, with 24-57 introns (except BraA09g025290 with one intron), and that the Arabidopsis and B. rapa callose GSL enzymes had similar coding sequences and very similar exon-intron structures, strongly supporting their close evolutionary relationships (Fig. 4). The loss of BraA09g025290 introns might have resulted in functional differences.
In addition, all members of the callose GSL enzymes possessed 20 motifs (Fig. 4, Table S3). As expected, all of the most closely related members had common motif compositions, suggesting functional similarities among the callose GSL enzyme proteins. According to the phylogenetic tree of the endo-1,3-bglucosidase enzymes, the number of exons ranged from 1 to 5. The structural differences in the enzymes might allow them to perform different functions. Gene structure diversification is a direct expression of gene family expansion [41]. The structural diversity of callose endo-1,3-b-glucosidase enzyme family members in Chinese cabbage provides a mechanism for gene evolution, and exon loss or gain can be an important step in generating structural diversity and complexity [42]. In this study, 13 exons were found in BraA01g009790 (Fig. 5), and BraA01g003770 had 20 exons, which indicates that the gain of exons might have resulted in functional divergence. In addition, we searched for conserved motifs within the Chinese cabbage endo-1,3-b-glucosidase enzymes using online MEME tools to analyze their motif compositions. A total of 20 conserved motifs, designated motif 1 to motif 20, were identified within the genes. These motifs may help to predict the genes' functions. As shown in Fig. 5 and Table S4, the number of motifs of the endo-1,3-b-glucosidase enzyme family ranged from 5 to 18, and most had 14. Differences in motif composition may have resulted in functional divergence between different genes. However, it is uncertain whether such differences have led to loss of function and this needs further investigation and molecular experimental verification. Motifs 6 and 15 were shared by all the Chinese cabbage callose endo-1,3-b-glucosidase enzyme members and were relatively conserved. The phylogenetic and structural analyses presented here may facilitate the functional annotation and study of callose GSL and endo-1,3-b-glucosidase enzymes in Chinese cabbage.

Chromosomal locations of the callose GSL and endo-1,3-b-glucosidase genes in the genome
To investigate the genomic distribution of the predicted callose GSL and endo-1,3-b-glucosidase genes, the DNA sequences were used to search the B. rapa genome database. We found that the callose GSL genes were distributed unevenly among the 10 chromosomes of the Chinese cabbage genome (Fig. S1). Four callose GSL genes were found on chromosomes chr5 and chr9, three on chromosome chr10, and one on each on chromosomes chr1, chr2, chr3, and chr7. No callose GSL genes were found on chr4, chr6 or chr8. The callose endo-1,3-b-glucosidase genes were also unevenly distributed on the Chinese cabbage chromosomes (Fig. S1). Chromosomes 9 and 6 contained the maximum and minimum numbers, respectively, of callose endo-1,3-b-glucosidase genes.

Callose GSL and endo-1,3-b-glucosidase gene expression profiles in different tissues of Chinese cabbage
Expression profiles give useful information on gene functions [43]. Here, we reanalyzed the expression levels of callose GSL and endo-1,3-b-glucosidase genes using publicly available RNA sequence data of four different tissues [37]. We selected the B. rapa accession Chiifu-401-42 to investigate the callose GSL and endo-1,3-b-glucosidase genes in the root, stem, leaf and flower. We used hierarchical clustering of the gene expression profiles to create heat maps (Fig. 6A,B). The results showed, in different tissues, 15 GSL genes had different expression patterns (Fig. 6A). In certain tissues, some genes were highly expressed. For example, BraA09g010050 and BraA05g012120 showed peak transcript levels in the flower. Similarly, the 77 endo-1,3-b-glucosidase genes in Chinese cabbage were differentially expressed in different tissues (Fig. 6B). In certain developmental stages, some genes were highly and specifically expressed. For example, BraA04g003430 and BraA09g008430 were only expressed and showed peak transcript levels in the flower. Therefore, these genes may play roles that cannot be neglected during flower development. In addition, the BraA03g025370, BraA05g031250 and BraA05g042280 genes showed particularly high levels of expression in the flower. Some genes (including BraA07g022770, BraA08g031380, BraA02g034940, BraA01g008280 and BraA07g020840) were highly transcribed in the stem, whereas BraA02g011800, BraA09g047910 and BraA04g003450 were highly transcribed in the leaf. Conversely, transcripts of some genes, especiallyBraA04g008040 and BraA07g009320, were not detected in any of the tissues. The transcriptional patterns indicate that these genes may be involved in organ development and growth. The different patterns also suggest the functional divergence of callose endo-1,3b-glucosidase enzymes.   Callose is synthesized by GSL and degraded by endo-1,3-b-glucosidase enzymes [1,44,45]. To determine the reason for the delayed degradation of callose in the tetrad stage, we further investigated the expression profiles of the callose GSL and endo-1,3-b-glucosidase genes in four developmental stages (meiosis, tetrad, mononuclear, and mature) of flowers in Chinese cabbage by conducting real-time quantitative PCR (qRT-PCR) analysis. Among the 15 GSL genes assayed in the four flower development stages, BraA01g041620 showed the most significant increases in the meiosis and tetrad stages in sterile plants (Fig. 7). Among the 77 endo-1,3b-glucosidase genes, BraA04g008040, BraA07g009320, BraA01g030220, BraA03g040850, BraA10g020080 and BraA05g038340 exhibited remarkable downregulation in the meiosis and tetrad stages in sterile plants (Fig. 8).
These results suggest the reason for the delayed degradation of callose may be a combination of high expression of the synthesis enzyme BraA01g041620 and low expression of the degradation enzymes BraA04g008040, BraA07g009320, BraA01g030220, BraA03g040850, BraA10g020080 and BraA05g038340.
The expression divergence of callose GSL and endo-1,3-b-glucosidase genes family in Arabidopsis and B. rapa To observe clearly the expression divergence between different members in the same gene family, we reanalyzed the expression levels of callose GSL and endo-1,3-b-glucosidase genes using publicly available RNA sequence data of different stages of flower in Arabidopsis [38], and our data of the expression of callose GSL and endo-1,3-b-glucosidase genes in fertile and sterile flower buds of the Chinese cabbage were reanalyzed. The results showed callose GSL genes in the same clades have similar expression trends, strongly supporting their close evolutionary relationship (Fig. 4). For example, the GSL genes AtCAL7, BraA10g005220 and BraA09g063900 are highly expressed during the sterile meiotic stage. But there are still some genes that have undergone functional differentiation (Fig. 4). The genes of the same clades, the BraA01g041620 and AtCAL9 genes, were highly expressed at the sterile meiotic stage, but the BraA03g033140 and BraA05g038460 genes were highly expressed in the fertile mononuclear stage. Furthermore, differences were found between clades in the endo-1,3-b-glucosidase enzyme family. In the same clade, At4G14080 (At-A6) and BraA04g008040 were detected at a higher transcription level in the sterile meiotic stage; however, the BraA08g009700 and  Table S3.  Table S4.
BraA05g015330 transcript levels were found to be greater in the fertile tetrad stage.

Discussion
Pollen development is an extremely important biological process in flowering plants and an indispensable life process for plant genetic breeding [2]. Abnormal development of pollen affects its function and can lead to male sterility, which seriously affects the reproduction and evolution of plants [46]. In this study, we compared the morphology of fertile and sterile flower buds of cabbage, and found that sterile stamens had no pollen grains when compared with those in fertile flower buds (Fig. 1). Microscopic examination of thin paraffin sections showed that, compared with fertile anthers, sterile anthers began to show abnormalities at the tetrad stage in which tapetal cells were highly vacuolate (Fig. 1B, F2 and S2). In later stages, tetrads could not develop into mononuclear microspores (Fig. 2H,I), and some became abortive microspores without viability (Fig. 2K). The four anther locules in sterile plants had sequentially shrunk and shriveled with the late pollen developmental stage, which led to pollen sterility (Fig. 2L).
There is an important link between pollen development and the deposition of callose [20]. The process of callose deposition is regulated by many factors. Changes in some of these factors lead to abnormal deposition of callose. Low deposition, premature degradation or delayed degradation of callose leads to abnormal pollen development and causes abnormal Fig. 7. Expression patterns of callose GSL genes at different floral developmental stages in Chinese cabbage and Arabidopsis. Relative expression analysis of callose GSL genes in the meiosis, tetrad, mononuclear and mature stages. qPCR analyses were performed, and expression values were calculated using the 2 ÀDDCT method. Data are mean values AE SE obtained from three replicates. pollen fertilization and subsequent pollen abortion [47]. The blue fluorescence of aniline blue was used to detect the synthesis and degradation during microspore abortion of callose in a male sterility NIL of Chinese cabbage. The misshapen form of many tetrads at the tetrad stage (Fig. 3, S2) and the fact that the tetrads in sterile microspores still exhibited a fluorescence response after tetrad phase (Fig. 3, F3 and S3) suggested that delayed callose degradation could be the important factor that determines pollen sterility in the Chinese cabbage line '10L03'.
To determine the main reason for the delayed degradation of callose, we performed a genome-wide analysis of the GSL and endo-1,3-b-glucosidase enzyme families in Chinese cabbage. Callose biosynthesis is catalyzed by GSL enzymes, which are located on the plasma membrane with the substrate on the cytoplasm side [48]. The callose synthesized is deposited between the plasma membrane and the cellulose of the cell wall [20]. Previously, 12 GSL genes (AtGSL1-AtGSL12) were found in Arabidopsis [22] and identified to participate in the synthesis of callose [49]. Callose is a glucan that contains b-1,3 bonds [16], and 1,3-b-glucosidase enzymes hydrolyze b-1,3-glucan (callose), which is widely present in viruses, bacteria, fungi [50,51] and seed plants [52]. In vitro experiments confirmed that endo-1,3-b-glucosidase enzymes play a major role in the process of microspore release [45]. Stieglitz and Stern (1977) studied the development of microspores in Lilium and found that 1,3-bglucosidase enzymes had a peak before the microspores were released from the tetrad, and then the callose walls of the microspores were dissolved to release mature pollen [53]. It has been demonstrated that endo-1,3-b-glucosidase enzymes play a key role in this process. However, the mechanism by which this enzyme is accurately induced before microspore maturation is not known [13]. The known endo-1,3-b-glucosidase enzymes belong to the seventeenth family of glycosyl hydrolases and their members share a common amino acid sequence structure: (LIVM)-X-(LIVM-FVW)-(STAG)-E-(ST)-GW-P-(ST)-X-G [52].
In this study, based on conserved domains and sequence similarities to known Arabidopsis callose GSL and endo-1,3-b-glucosidase enzymes, a genome-wide database were searched, revealing 15 GSL and 77 endo-1,3-b-glucosidase enzymes in the Chinese cabbage genome (Tables 1 and 2). The B. rapa has undergone the c triplication (135 MYA) and the b (90-100 MYA) and a (24-40 MYA) duplications that previous studies revealed [54]. In the evolutionary history of Chinese cabbage, three polyploidization events occurred and led to chromosomal reduction, rearrangement, and numerous gene losses. As a result, highly complex gene families have evolved. In this study, the 15 GSL and 77 endo-1,3-b-glucosidase members all clustered closely with their Arabidopsis orthologs. Many GSL and endo-1,3-b-glucosidase enzymes in Arabidopsis also had two or more counterparts in Chinese cabbage. This result showed that the expansion of the GSL and endo-1,3-bglucosidase families in Chinese cabbage may have been caused by genome duplication events, including transposition events, tandem duplication, multiple segmental duplication, and entire-genome duplication [55,56].
Polyploidization events also cause structural and functional domain diversification. By comparing intron/exon numbers, lengths, distribution and other genetic structures of a gene family, we can realize the diversity of genetic structure [41]. In coding region BraA07g023720 contained the largest number of introns (57). Structural diversification may be caused by a large number of introns (Fig. 4B). Correspondingly, BraA07g023720 had the largest protein length (Table 1). Variations in intron lengths also occurred in the endo-1,3-b-glucosidase enzymes (Fig. 5). The structural diversity of endo-1,3-bglucosidase family members in Chinese cabbage provides an evolutionary mechanism for gene duplication, an important step that with exon loss or gain generates structural diversity and complexity [42]. The longest exon length was found in BraA01g003770 relative to other endo-1,3-b-glucosidase enzymes in Chinese cabbage, which indicates that gain of exons might have resulted in functional diversification. As shown in Fig. 5, most endo-1,3-b-glucosidase enzymes had 14 motifs. This difference in motif composition among endo-1,3-bglucosidase enzymes may potentially have caused functional divergence among different genes. The polyploidization events that occurred in the evolution of the B. rapa were very important and gave the plant the ability to diversify and respond to changing habitats. New functions have developed for the genes generated from polyploidization events [54].
The large size of these two gene families in Chinese cabbage indicate their importance in the regulation of cabbage-specific processes. Here, the expression profiles of the 15 GSL and 77 endo-1,3-b-glucosidase genes in different tissues showed that different genes were different expression levels in different organs (Fig. 6A,B). By contrast, BraA09g010050 and BraA05g012120 showed peak transcript levels in flowers. Similarly, among the 77 endo-1,3-b-glucosidase genes in Chinese cabbage, BraA04g003430 and BraA09g008430 were only expressed and showed peak transcript levels in the flower. Additionally, BraA03g025370, BraA05g031250 and BraA05g042280 showed high levels of expression in the flower. Therefore, these genes may play a potentially important part during flower development.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Chromosomal locations of the callose GSL and endo-1,3-b-glucosidase genes in Chinese cabbage. Table S1. Callose GSL enzyme primers used for qPCR analysis. Table S2. Callose endo-1,3-b-glucosidase enzyme primers used for qPCR analysis. Table S3. The MEME results of callose GSL enzymes in Chinese cabbage .  Table S4. The MEME results of callose endo-1,3-b-glucosidase enzymes in Chinese cabbage.