*2.3. Identification of BrBGLU Genes Involved in Pollen Development*

1 Rice *TDR* (*Tapetum Degeneration Retardation*) mutant alters *BGLU1* expression with flower specificity [28], and *BGLU1* and *BGLU13* are found to be related to male organ development in *Calamus palustris* [34]. These previous reports lead to a hypothesis that *BrBGLU*s are involved in pollen development. To test this hypothesis, the previously published microarray data relating to male sterility in *B. rapa* [29] were re-annotated, using the improved *B. rapa* genome (version 3.0) [35] and analyzed based on pollen development (Table S1). A total of 36 *BrBGLU*s, represented by 88 probes (or 88 ESTs) showed significant hybridization values, of which 12 *BrBGLU*s showed over two-fold change in expression levels between fertile and sterile floral buds: six members were upregulated, and members were downregulated in fertile buds. Among these genes, four upregulated genes (*BrBGLU10/AtBGLU20, BrBGLU15/AtBGLU3, BrBGLU16/AtBGLU4*, and *BrBGLU64/AtBGLU41*) and

two downregulated genes (*BrBGLU2/AtBGLU46* and *BrBGLU19/AtBGLU30*) were described as good candidates that were associated with pollen development. The function of all four upregulated genes has not been known up to now, but at least three, *BrBGLU10*, *BrBGLU15*, and *BrBGLU64* appeared to be related to pollen wall development. In particular, we further analyzed *BrBGLU10/AtBGLU20*, as these showed hundred-fold changes between fertile and sterile buds. *BrBGLU64/AtBGLU41*) and two downregulated genes (*BrBGLU2/AtBGLU46* and *BrBGLU19/AtBGLU30*) were described as good candidates that were associated with pollen development. The function of all four upregulated genes has not been known up to now, but at least three, *BrBGLU10*, *BrBGLU15*, and *BrBGLU64* appeared to be related to pollen wall development. In particular, we further analyzed *BrBGLU10/AtBGLU20*, as these showed hundred-fold changes between fertile and sterile buds.

upregulated genes (*BrBGLU10/AtBGLU20, BrBGLU15/AtBGLU3, BrBGLU16/AtBGLU4*, and

#### *2.4. Analysis of the Putative Functions of BrBGLU10/AtBGLU20 in Pollen Development 2.4. Analysis of the Putative Functions of BrBGLU10/AtBGLU20 in Pollen Development*

To gain more insights into the functions of the *BrBGLU*s during pollen development, *BrBGLU10,* which was highly and specifically expressed in fertile buds, was selected for further analysis. *AtBGLU20,* the *Arabidopsis* ortholog of *BrBGLU10,* was initially named as *ATA27,* which is one of the *A. thaliana* anther-specific expressed genes [36]. To confirm the expression patterns of *BrBGLU10* and *AtBGLU20,* RT-PCR was conducted (Figure 4A,B). The expression level of *BrBGLU10* was specifically detected at the F1–F3 stages, with highest levels at the F2 stage, representing the tetrad stage, and *AtBGLU20* was specifically expressed before floral stage 12. The RT-PCR results might imply its important role in pollen development. To gain more insights into the functions of the *BrBGLU*s during pollen development, *BrBGLU10,* which was highly and specifically expressed in fertile buds, was selected for further analysis. *AtBGLU20,* the *Arabidopsis* ortholog of *BrBGLU10,* was initially named as *ATA27,* which is one of the *A. thaliana* anther-specific expressed genes [36]. To confirm the expression patterns of *BrBGLU10* and *AtBGLU20,* RT-PCR was conducted (Figures 4A, B). The expression level of *BrBGLU10* was specifically detected at the F1–F3 stages, with highest levels at the F2 stage, representing the tetrad stage, and *AtBGLU20* was specifically expressed before floral stage 12. The RT-PCR results might imply its important role in pollen development.

**Figure 4.** Analysis of expression of *BrBGLU10* and *AtBGLU20,* and Gene Ontology (GO) enrichment of co-expressed genes. **A**, Expression of BrBGLU10 in different tissues and floral bud stages in *B. rapa*. **B**, Expression of AtBGLU20 in different tissues and floral bud stages in *Arabidopsis*. **C**, Expression patterns of *BrBGLU10* and its co-expressed genes in sterile and fertile *B. rapa* floral buds, based on previously published microarray data [29]**. D**, GO enrichment analysis of genes co-expressed with *BrBGLU10*. **E**, Expression pattern of *AtBGLU20* and its co-expressed genes in various tissues of *Arabidopsis*, which was performed using the Arabidopsis eFP Browser (http://bar.utoronto.ca/efp/cgibin/efpWeb.cgi). **F**, GO enrichment analysis of genes co-expressed with *AtBGLU20.* S1–S3 represent the floral buds from male-sterile *B. rapa*. S1, before the tetrad stage. S2, after the tetrad stage. S3, containing aberrant pollen grains. F1–F4 indicate fertile *B. rapa* floral buds before the tetrad stage (F1), at the tetrad stage (F2), after the tetrad stage, but before containing mature pollen (F3), and containing mature pollen (F4). For *Arabidopsis*, FS1–12, flower stage 1 to stage 12; FS13–14, flower stage 13 to stage 14. PI, probe intensity. **Figure 4.** Analysis of expression of *BrBGLU10* and *AtBGLU20,* and Gene Ontology (GO) enrichment of co-expressed genes. **A**, Expression of BrBGLU10 in different tissues and floral bud stages in *B. rapa*. **B**, Expression of AtBGLU20 in different tissues and floral bud stages in *Arabidopsis*. **C**, Expression patterns of *BrBGLU10* and its co-expressed genes in sterile and fertile *B. rapa* floral buds, based on previously published microarray data [29]. **D**, GO enrichment analysis of genes co-expressed with *BrBGLU10*. **E**, Expression pattern of *AtBGLU20* and its co-expressed genes in various tissues of *Arabidopsis*, which was performed using the Arabidopsis eFP Browser (http://bar.utoronto.ca/efp/cgibin/efpWeb.cgi). **F**, GO enrichment analysis of genes co-expressed with *AtBGLU20.* S1–S3 represent the floral buds from male-sterile *B. rapa*. S1, before the tetrad stage. S2, after the tetrad stage. S3, containing aberrant pollen grains. F1–F4 indicate fertile *B. rapa* floral buds before the tetrad stage (F1), at the tetrad stage (F2), after the tetrad stage, but before containing mature pollen (F3), and containing mature pollen (F4). For *Arabidopsis*, FS1–12, flower stage 1 to stage 12; FS13–14, flower stage 13 to stage 14. PI, probe intensity.

To demonstrate similar or conserved functions between *BrBLU10* and *AtBGLU20*, we isolated the co-expressed genes of *BrBGLU10*, using microarray data [29] and *AtBGLU20* from the *Arabidopsis* eFP Browser (http://bar.utoronto.ca/efp/cgi-bin/efpWeb.cgi) [37]. With the Pearson's correlation coefficient (PCC) value above 0.90, 183 probes (107 genes) and 25 genes were determined to be co-expressed with *BrBGLU10* and *AtBGLU20,* respectively (Figure 4C, E; Tables S2–S3). *BrBGLU10* and its co-expressers were upregulated at the fertile floral bud stage, and the highest expression level was detected at the F2 stage (Figure 4C), suggesting that *BrBGLU10* plays a role during pollen development, especially from the tetrad stage to that before the mature pollen stage. In *Arabidopsis*, flower and stamen development processes were divided into 14 stages and 12 stages, respectively [37–39]. *AtBGLU20* and its co-expressers were represented by a high probe intensity (PI) value at flower stages (FS) 9 to 12, indicating that *AtBGLU20* plays a role in *Arabidopsis* pollen development (Figure 4E). We also conducted Gene Ontology (GO) enrichment analysis to provide more information on the function of *BrBGLU10* and *AtBGLU20* (Figure 4D,F). The results showed that genes involved in pollen exine formation and pollen wall assembly were highly over-represented among genes co-expressed with *BrBGLU10* and *AtBGLU20.* Taken together, our analysis indicated that *BrBGLU10* and *AtBGLU20* may be required for pollen development in both *B. rapa* and *Arabidopsis*.

To validate *AtBGLU20* function in the pollen development, we generated knockdown mutants of *AtBGLU20* by introducing antisense constructs under the control of the *CaMV35S* promoter (Figure 5A). After screening, four independent knockdown lines were obtained with expression levels ranging from 55% to 85% (Figure 5B). However, the *AtBGLU20* downregulated plants showed normal vegetative growth based on morphology (Figure 5C), but produced defective pollen grains relative to the wild-type plants (Figure 5D). These results indicated that normal pollen development in *Arabidopsis* requires sufficient amounts of *AtBGLU20*. All data obtained from gene expression, co-expression analysis, and transgenesis led to the conclusion that *AtBGLU20* and *BrBGLU10* may have indispensable functions in pollen development in both *Arabidopsis* and *B. rapa*, respectively.

**Figure 5.** Analysis of WT and *AtBGLU20* antisense knockdown mutant *Arabidopsis* plants. **A**. Schematic representation of the *AtBGLU20* gene structure and DNA fragment regions for antisense constructs. The white box indicates the UTR region; gray boxes are exons; lines represent introns. The single arrow indicates the antisense orientation of the fragments in the constructs. F and R indicate the primer

1

positions used in qRT-PCR analysis. **B**, Analysis of the expression levels of *AtBGLU20*. Expression was normalized to that of *At*ACT*7*, and represented relative to the expression levels of the WT. Error bars represent the SD of three biological replicates. **C,** Morphologies of wild-type Arabidopsis plants and *AtBGLU20* knockdown transgenic plants, which showed no obvious differences in vegetative growth. Bar = 20 mm. **D**, Mature pollen grain of WT and *AtBGLU20* transgenic plants stained with modified Alexander solution (Peterson et al., 2010). The well-developed pollen grains were stained red. Bar = 20 µm. WT, wild-type. 10, 17, 20, and 30 indicate four independent transgenic lines. The number in the parentheses indicate the percentages of defective pollen grains.

#### **3. Discussion**

#### *3.1. Identification and Analysis of BrBGLUs*

GH1 family genes play an important role in regulating abiotic and biotic stress responses, as well as various developmental processes in plants [9,12,14,18,23,40]. Based on the results of an increase in the number of whole genome sequences from a large number of species, genome-wide analysis of various gene families has been extensively performed. However, genome-wide identification and characterization of the GH1 gene family has only been reported in a few plant species, and there is no information on *Brassica* species, which are important crops for production of functional foods, as well as health-promoting compounds. In this study, the isolation of *BrBGLU*s from *B. rapa* genome (Figure 1), the distribution of *BrBGLU* genes on chromosomes (Figure 1), phylogenetic analysis (Figure 2), and exon–intron structures (Figure 3) provides substantial information on the functions and roles of these genes.

Compared with the 49 *AtBGLUs* and 37 *OsBGLUs* in Arabidopsis and Rice, respectively [9], 64 *BrBGLUs* were isolated from the *B. rapa* genome, which is the largest number so far that has been reported in plants (Figure 1). The high number of *BGLU* family members in *B. rapa* could be related to the genome triplication event in this lineage [41]. To adapt different new functions that are suitable for changes in the environment, gene structure was commonly diversified during the evolution of multigene families [42]. For *BGLUs*, 13 exon–12 intron organization was considered as the ancestral gene structure, with the loss of certain introns leading to other gene structures [6]. The exons present in *BrBGLUs* varied from 2 to 13, and the most common organization was 11 exons (Figure 3). The introns in Arabidopsis vary from 0 to 13 [6]. This results suggested that little diversity exists in the gene structure of *BrBGLUs* when compared to *AtBGLUs*.

*BrBGLU*s may have originated from *Arabidopsis*, although duplication, gene loss, and functional diversification may have also occurred. This is supported by the fact that *BGLU*s from both species could be grouped into 10 subfamilies, with tandem arrays, as defined by Singh et al., 2013 [39], although some families were re-grouped or diverged into other subgroups. Figure 2 shows that *AtBGLU* subfamilies 8 and 9 [6] were incorporated into one *B. rapa* subfamily, GH1-c, and *BrBGLU51* is composed of GH1-d with six *AtBGLUs* (*AtBGLU34*/*35*/*36*/*37*/*38*/*39*), indicating the loss of some *BGLU*s in *B. rapa.* This phenomenon may result from the rapid evolution of genes similar to that previously observed between *Arabidopsis* and rice [5]. One more interesting finding was that *AtBGLU48* (*SFR2*) was incorporated into the GH1-j subgroup, with *BrBGLU8* and *BrBGLU42* (Figure 2). *AtBGLU42* is a β-glucosidase, but it is divergent from all other *AtBGLUs*, and more similar to several β-glycosidases from thermophilic archea and bacteria [32]. *SFR2* is involved in the lipid remodeling of the outer chloroplast membrane during freezing tolerance [43,44]. Because two *BrBGLU*s in the GH1-j subgroup had identities between 85% and 87% with *AtBGLU2*, *Brassica* genes may have a similar function of freezing tolerance as that in *AtSFR2*, although this requires further investigation.

On the basis of *Arabidopsis* study, most subfamilies of BGLUs in Figure 2 may be associated with specific functions: GH1-a for flavonoid and anthocyanin metabolism, GH1-e for flavonoid utilization, GH1-d for myrosinases, and GH1-i for scopolin hydrolysis. At least 12 genes are known to be involved in flavonoid metabolism in GH1-a: *AtBGLU1-6* for flavonol accumulation [10,11], *AtBGLU7-11* as anthocyanin glucosyltransferases [10–12], and *AtBGLU15* for flavonol bisglycoside

catabolism under abiotic stress [13]. *AtBGLU12*-*17* in the GH1-e subgroup code for flavonoid-utilizing BGLUs in legumes [10]. An examination of the functions of *BrBGLUs* that are clustered with *AtBGLUs* in subgroups GH1a and GH1-e may provide information and understanding into the regulation of flavonoid biosynthesis in *Brassica* species.

Several subfamilies may be related to abiotic and biotic stress resistance, such as GH1-b, GH1-c, GH1-d, GH1-f, and GH1-i. Myrosinases hydrolyze glucosinolates into active forms that are involved in plant defense against herbivory and pathogens, and in human health promotion [45–48]. *AtBGLU26* and *AtBGLU34*-*39* function as myrosinases [14–16]. Except for *AtBGLU26* (GH1–h), most genes belong to the GH1-d subgroup (Figure 2). Understanding myrosinase function in Brassicaceae, which is rich in glucosinolates, may provide an excellent strategy for breeding health-promoting *Brassica* crops [49,50]. ABA also functions in stress responses, including drought stress. *AtBGLU18* [17] and *AtBGLU33* [18] hydrolyze glucose-conjugated ABA, thereby increasing ABA levels and inducing ABA responses such as drought tolerance. However, these two proteins are separated into two subfamilies, implying the presence of more BGLUs for the regulation of ABA levels. Scopolin is one of the coumarins produced in roots [51], and it plays a role in the defense against pathogen attack and abiotic stresses [19,20]. Three β-glucosidases that hydrolyze scopolin and their encoding genes (*AtBGLU21, 22* and *23*) have been characterized [21,22]. The GH1-i subfamily includes these three genes and 11 *BrBGLU*s, which should be examined in relation to scopolin production. The GH1-b subfamily includes two monolignol glucoside hydrolases (AtBGLU45 and AtBGLU46) that control lignin content [23]. Because *OsBGLU14*, *16*, and *18* are involved in lignin biosynthesis with monolignol β-glucosidase activity and compensate for the *Arabidopsis bglu45* mutant [52], BrBGLUs in this subfamily may play similar roles. *AtBGLU42* in GH1-c is involved in the induction of systemic resistance to bacterial disease, and the release of iron-mobilizing phenolic metabolites under iron deficiency [24]. Several genes in this subfamily would thus be expected to contribute to eliciting defense responses. All of this information may contribute to future research directions in relation to *BrBGLUs*.

## *3.2. The Potential Functions of BrBGLUs During Pollen Development*

Previous studies on rice and other plant species have indicated that β-glucosidases play roles in pollen development [34,36,53]. To identify the BrBGLUs responsible for pollen development, the previously published microarray data relating to male sterility in *B. rapa* were re-annotated and re-analyzed. Among the 36 *BrBGLU*s, 12 *BrBGLUs* showed over a two-fold change between fertile and sterile floral buds (Table S1). However, six genes (four upregulated and two downregulated genes) were more extensively studied in terms of their role in pollen development. We selected one *BrBGLU10* for investigation, the homolog of *AtBGLU20*, which showed hundreds-fold changes in its expression.

We examined the expression levels of *BrBGLU10/AtBGLU20* and analyzed the co-expressed genes in both *B. rapa* and *Arabidopsis* (Figure 4). An assessment of expression levels strongly suggests that *BrBGLU10/AtBGLU20* are involved in pollen development. The cellular contents from the degeneration of the tapetum supports pollen wall formation and subsequent pollen release [39]. Mutations in polysaccharide metabolism-related genes lead to defective pollen wall formation [26]. Glycoside hydrolase has been reported to be involved in the cell wall polysaccharide degradation [27]. The expression patterns of *BrBGLU10* and *AtBGLU20* suggest that they might play a role from the tetrad stage to mature pollen grains (Figure 4A, B), which corresponding to the tapetum degradation stage [29,39]. Co-expression analysis is a valuable approach for classifying and visualizing transcriptomic data to identify genes with similar cellular functions and regulatory pathways [54–56], although this is not always the case [57,58]. In plants, co-expression analysis under various experimental conditions has been used for predicting gene function [55,59]. Figure 4C,D shows that this gene possibly regulates pollen development. In particular, GO annotation of co-expressed genes reflects that pollen wall and exine formation are influenced by *BrBGLU10/AtBGLU20*, indicating that the hydrolysis of glucose moieties is necessary for proper pollen development.

Because BrBGLU10 had a high sequence identity with AtBGLU20 (87% at the nucleotide level and 84% at the amino acid sequence level), both genes may thus have similar functions. Therefore, knocking down *AtBGLU20* may provide direct evidence for its function in pollen development. Figure 5 shows that the suppression of *AtBGLU20* expression had no effect on plant growth and development, although this aborted pollen production. This result implies that BrBGLU10/AtBGLU20 are critical to pollen grain development.

## **4. Materials and Methods**

#### *4.1. Plant Materials and Growth Conditions*

Seeds of *B. rapa* subsp. *pekinensis* (Chiifu) were germinated in Petri dishes in the dark at 23 ◦C for two days, then the germinated seeds were transferred to a 4 ◦C growth chamber with 16 h of light for 25 days to induce vernalization. After vernalization, the seedlings were transplanted into 15 cm × 15 cm × 18 cm pots containing potting soil and grown in a 23 ◦C growth chamber with 16 h of light. The floral buds were collected from 10 plants with three biological replicates, as previously described [29], and stored at −70 ◦C until use. Root and shoot tissues were collected from three-week-old seedlings without vernalization. Stem tissue was sampled from the plants one week after bolting.

*A. thaliana* (L.) Heynh var. Columbia (Col-0) plants were grown under 140 µmol/m2/s light intensity at 23 ± 1 ◦C with a long day cycle with 16 h of light for plant transformation. Seeds were sown in 55 mm × 55 mm pots in potting soil, stratified for three days at 4 ◦C, and then transferred to the growth room. The plants were then kept under a transparent polythene lid for one week to increase humidity and support equal germination. The plate-cultured seeds were sterilized with 30% bleach and 0.1% Triton X-100 (Sigma, St. Louis, MO, USA), stratified for three days at 4 ◦C, and sown in Petri dishes with dimensions of 100 mm × 100 mm × 20 mm. The dishes contained half-strength MS media (Duchefa Biochemie, Netherlands) supplemented with 0.8% phytoagar and 1% sucrose.

#### *4.2. Antisense Constructs and Plant Transformation*

The full-length coding sequence of *AtBGLU20* was cloned from first-strand complementary DNA (cDNA), using the primers BGLU20F (Table S4). Then, the fragments were inserted into T&A cloning vectors (RBC T&A cloning kit, Real Biotech Corporation, Taiwan). After confirmation of the *AtBGLU20* sequence in the T&A vector by sequencing, the fragment was cloned into pCambina 3300-35S binary vectors and used in plant transformation. Col-0 were used for transformation with *Agrobacterium tumefaciens* GV3101 carrying the above binary plasmid using the floral dip method [60]. The transformants were selected on plates containing 25 mg/mL glufosinate in MS medium (Sigma, St. Louis, MO, USA), and also confirmed by genomic DNA PCR analysis.
