BcMF26a and BcMF26b Are Duplicated Polygalacturonase Genes with Divergent Expression Patterns and Functions in Pollen Development and Pollen Tube Formation in Brassica campestris

Polygalacturonase (PG) is one of the cell wall hydrolytic enzymes involving in pectin degradation. A comparison of two highly conserved duplicated PG genes, namely, Brassica campestris Male Fertility 26a (BcMF26a) and BcMF26b, revealed the different features of their expression patterns and functions. We found that these two genes were orthologous genes of At4g33440, and they originated from a chromosomal segmental duplication. Although structurally similar, their regulatory and intron sequences largely diverged. QRT-PCR analysis showed that the expression level of BcMF26b was higher than that of BcMF26a in almost all the tested organs and tissues in Brassica campestris. Promoter activity analysis showed that, at reproductive development stages, BcMF26b promoter was active in tapetum, pollen grains, and pistils, whereas BcMF26a promoter was only active in pistils. In the subcellular localization experiment, BcMF26a and BcMF26b proteins could be localized to the cell wall. When the two genes were co-inhibited, pollen intine was formed abnormally and pollen tubes could not grow or stretch. Moreover, the knockout mutants of At4g33440 delayed the growth of pollen tubes. Therefore, BcMF26a/b can participate in the construction of pollen wall by modulating intine information and BcMF26b may play a major role in co-inhibiting transformed plants.

B. campestris ssp. chinensis var. parachinensis is a subspecies of B. campestris L. ssp. chinensis, and a fast growing inbred with a growth period of only 50 d. This subspecies was used for our genetic transformation experiment. The transgenic plants were cultivated in a 22°C to 24°C growth chamber for phenotype observation.
The seeds of wild-type Arabidopsis (Col-0) and two At4g33440 mutants (SALK_028430 and SALK_013967) were obtained from the ABRC database. The A. thaliana plants were grown under long-day conditions (16 h light/8 h dark) in a 22±1°C growth chamber for phenotype observation.
DNA and RNA extraction, cDNA synthesis, and sequence amplification Genomic DNA was isolated from leaves using the cetyltrimethylammonium bromide method [28]. Total RNA was purified from different tissues using TRIzol Reagent. The first strand cDNA was synthesized using a PrimerScript RT reagent Kit. The gene-specific primers (Primers 1 to 4, S1 Table) designed based on the sequence information of BcMF26a (gene ID: Bra011440) and BcMF26b (gene ID: Bra037005) in BRAD were used for homolog amplification of their DNA and cDNA sequences in B. campestris. The high fidelity thermostable DNA polymerase KOD was used for PCR. The PCR products were verified by sequencing.

Gene structure and phylogenetic analysis
Gene structure was analyzed by FGENESH website. Sequence similarity alignment was carried out using ClustalX software. Molecular characteristics of the deduced protein were determined using ExPASy website. The signal sequence was analyzed with Signal P4.1 server. Collinear analysis was performed with GEvo website. Amino acid sequences in S2 Table from different plant species were retrieved from GenBank database or published references. The phylogenetic tree was generated with Mega6.0 software.

QRT-PCR analysis
Two total RNAs were independently used for qRT-PCR to examine the transcript levels of BcMF26a and BcMF26b in B. campestris. The gene-specific primer pairs (Primers 5 and 6, S1 Table, S1 Fig) were designed based on the diverged sequences upstream of 'ATG', respectively, for independent expression examination of the two genes. The constitutively expressed UBC-10 gene (Primer 7, S1 Table) was selected as an internal control. The CFX96 Real Time System machine and SYBR Premix Ex Taq Kit were used for the amplifications. The PCR products were sequenced to confirm the two genes. The 2 -ΔΔCt method was used to calculate the relative expression levels [29].
putative promoter fragments included 50 bp to 150 bp downstream of 'ATG'. The obtained putative promoter fragments were used to construct the fusion vectors pBGWFS7.0-proBcMF26a: GUS-GFP and pBGWFS7.0-proBcMF26b: GUS-GFP, with the destination binary vector pBGWFS7.0: GUS-GFP [30]. To examined the activity of the obtained promoter fragments, the two plasmids were transiently transformed into onion epidermal cells, respectively, by particle bombardment [31] using the Biolistic PDS-1000/He gene gun system. After 24 h of incubation, GFP-dependent fluorescence signals in the onion epidermal cells were observed by Fluorescent Microscope. Meanwhile, plasmids were introduced into Agrobacterium tumefaciens strain GV3101 and transformed into Arabidopsis wild-type plants using the floral dipping method [32] to produce transgenic GUS-expressing lines. The empty vector pBGWFS7.0: GUS-GFP was used as control.
The seedlings of T 1 transgenic Arabidopsis plants were subjected to selection by spraying with 0.1% BASTA. The T 2 transgenic lines were selected and used for histochemical GUS assay. The T 2 seeds were grown on agar medium [consisting of half MS salts, 1.0% (w/v) sucrose, and 0.8% agar] for 4 d for seedling staining, as well as in soil for 35 d for inflorescence staining. Staining was performed as described by Sun et al. [33]. Subsequently, photographs of the seedlings and flower buds were taken using a stereomicroscope. Moreover, the floral buds at different development stages were dehydrated and embedded in Tissue-Tek -OCT compound, and then frozen and sliced into 10μm sections. The sections were observed under Fluorescent Microscope. Populations derived from at least six independently transgenic lines were analyzed for each transformed construct.

Subcellular localization of BcMF26a and BcMF26b proteins
The ORF fragments with specific restriction enzyme sites of BcMF26a and BcMF26b were amplified with gene-specific primer pairs (Primers 10 and 11, S1 Table). The resulting fragments were cloned into the pFGC: GFP vector to create the fusion vectors pFGC-BcMF26a: GFP and pFGC-BcMF26b: GFP. The two constructs were verified by sequencing and transformed into onion epidermal cells. The transformation process was consistent with the method of promoter activity analysis. Fluorescence signal was analyzed after 16 h of incubation. Moreover, the onion epidermal cell was plasmolyzed in 0.3 g mL −1 sucrose for 3 min and photographed by Fluorescent Microscope to investigate the GFP signal distribution.

Artificial miRNA (amiRNA) construction and plant transformation
The amiRNA co-inhibiting BcMF26a and BcMF26b was designed as the procedure introduced in the MicroRNA Designer (WMD3) website according to the multiple target mode [34]. The amiRNA was constructed based on the miR164a backbone. The obtained amiRNA fragment was cloned into a modified pCAMBIA1301 vector, which was added with another CaMV35 promoter between the KpnI and SmaI restriction sites. The successful construct was transferred into B. campestris ssp. chinensis var. parachinensis though the A. tumefaciens-mediated transformation system described by Yu [35]. Meanwhile, empty vector pCAMBIA1301 was transformed as the negative control. Hygromycin was used to screen the transformed plants. With an identical primer (Primer 16, S1 Table), the mRNA levels of BcMF26a and BcMF26b in the transgenic lines were measured simultaneously by qRT-PCR analysis.

Pollen analysis and pollen germination observation
The pollen grains from the dissected anthers of the flower buds about to open were collected and stained to analyze pollen viability. Alexander was used to assay the cytoplasm and cell wall [36]. 4', 6-Diamidino-2-phenylindole (DAPI) was used to stain the nuclei.
The samples for scanning electron microscopy (SEM), transmission electron microscopy (TEM), and semi-thin section observation were prepared using the procedure described by Huang [14]. The digital images of the pollen grains were obtained with a SEM. Semi-ultrathin sections (2 μm) were obtained, stained with dimethyl blue, and photographed with a Fluorescent Microscope. Ultrathin sections (50 nm) were stained with uranyl acetate and lead citrate, and then viewed in a TEM operating at 80 kV.
To observe pollen tube germination in vitro, pollen grains were collected and cultured as described by Jiang et al. [37]. After 4 h of incubation, the germinating percentage was calculated and the pollen tube morphologies were observed with Alexander staining under Fluorescent Microscope. Three independent transgenic lines and the corresponding control line were examined. More than six flowers buds from each plant were tested. At least 300 pollen grains were observed to calculate the average germination rate.

Verification of T-DNA insertion mutants and in vivo pollen tube germination in Arabidopsis
Homozygous T-DNA insertion mutants of At4g33440, namely, SALK_013967 and SALK_028430, were identified using PCR with gene-specific primers (Primers 12 and 13, S1 Table). The coding region of At4g33440 for RT-PCR was amplified using the specific-primer Primer 14 (S1 Table). Tubulin (Primer 15, S1 Table) was used as an internal control.
Preemasculated mature flowers were pollinated according to the following combinations: SALK_013967 × SALK_013967, wild-type × SALK_013967. Pistils were collected at 4, 12, and 24 h after hand-pollination, fixed with ethanol:acetic acid (v/v) = 3:1 solution, and softened with 8 M NaOH overnight. The pistils were washed in distilled water and stained with aniline blue solution in the dark for 3 h to 4 h. The stained pistils were photographed with Fluorescent Microscope under UV light. The same treatments were performed for the SALK_028430 mutant lines. Each combination was observed in triplicate.

Sequence characterization analysis of BcMF26a and BcMF26b
BcMF26a and BcMF26b, the orthologous genes of At4g33440, were found in BRAD using the basic local alignment search tool service. The phylogenetic tree analysis of PG genes from A. thaliana and B. campestris genomes also proved the high homology between BcMF26a, BcMF26b, and At4g33440 ( Fig 1A, part of data displayed). Collinear analysis was performed on BcMF26a, BcMF26b, and At4g33440 ( Fig 1B). BcMF26a was localized on Chromosome 1 (chr1) and BcMF26b was localized on Chromosome 3 (chr3) in the B. campestris genome. Meanwhile, BcMF26a and BcMF26b were aligned to the same locus, the PG gene At4g33440, in the A. thaliana genome. The regions immediately surrounding BcMF26a and BcMF26b on chr1 and chr3, respectively, exhibited the reverse orientation. The collinearly duplicated chromosomal blocks of the two genes were interspersed with chromosomal fragments of no apparent similarity. Moreover, the collinear conserved blocks between the two chromosomal regions preserved their orientations with respect to the target genes.
Gene sequences of BcMF26a and BcMF26b were isolated; both genes contained five exons and four introns (S1A Fig). The ORFs of BcMF26a and BcMF26b were 1425 and 1434 bp, encoding 474 and 477 amino acid residues, respectively. Their calculated molecular masses were 51.899 and 52.473 kDa, and their estimated pIs were 6.79 and 7.7. BcMF26a and BcMF26b had high similarities for ORFs and amino acid sequences (S1B and S1C Fig Based on peptide sequence homology, PG genes from other plant species were found in the NCBI database by BLASTX. The phylogenetic tree was constructed using full-length amino acid sequences of BcMF26a, BcMF26b, and PG polypeptides from different plant species in S2 Table. The result revealed that BcMF26a, BcMF26b, and At4g33440 were grouped into Clade E ( Fig 1C). Furthermore, multiple sequence alignment with these PG proteins showed that they all contained three out of the four typical conserved domains of PG protein (S4 Fig).
BcMF26a and BcMF26b had differential expression patterns The transcripts of BcMF26a and BcMF26b were detected in different organs and tissues of B. campestris using qRT-PCR analysis (Fig 2). In the five different organs of heterozygous male fertile 'Bcajh97-01B' plants, BcMF26b expression in stems, leaves, inflorescences, and siliques was 2.46, 1.87, 2.22 and 3.32 times higher than that in roots, respectively (Fig 2A). In the flower buds at five pollen developmental stages from the homozygous male sterile 'Bcajh97-01A' plants (A1-A5) and heterozygous male fertile 'Bcajh97-01B' plants (B1-B5) (Fig 2B), the relative expression levels of BcMF26b in B1, B2, B4, and B5 were higher than that in A1, A2, A4, and A5; the minimum expression level divergence was 1.38 times between B2 and A2. Moreover, the expression of BcMF26b in 'Bcajh97-01B' was much higher in B1, B2, B4, and B5, which were 3.59, 2.13, 3.45, and 5.13 times higher than that in B3, respectively. For the four floral parts of B5 (Fig 2C), BcMF26b was mainly expressed in stamens and pistils, with relative expression values of 259.57 and 409.20, respectively, but those in sepals and petals were only  40.69 and 39.48, respectively. BcMF26a was hardly expressed in all the tested organs and tissues compared with BcMF26b (Fig 2). The highest expression value was 43.11 in pistils, which was only 0.1 times that of BcMF26b in the same tissue. However, BcMF26a and BcMF26b had similar expression trends (Fig 2). The main expression divergences between BcMF26a and BcMF26b were the differences in expression levels in the tested organs and tissues.
'Bcajh97-01A' is a genetic male-sterile line that lacks mature pollen; 'Bcajh97-01B' is the fertile maintain line [38]. Given the different expression patterns in 'Bcajh97-01A' and 'Bcajh97-01B' and the high similarity of the coding sequences, we renamed Bra011440 and Bra037005 as Brassica campestris Male Fertility 26a (BcMF26a) and BcMF26b, respectively, in this study.
The fusion vectors pBGWFS7.0-proBcMF26a: GUS-GFP and pBGWFS7.0-proBcMF26b: GUS-GFP, were transiently transformed into onion epidermal cells to detect the activity of the promoter fragment. Typical GFP transient expression signals were observed in the onion epidermal cells (S5A and S5C Fig). By contrast, no signal was observed when the empty vector pBGWFS7.0: GFP was transformed. These results showed that the amplified promoter fragments demonstrated promoter activity. Meanwhile, GUS activity in the seedlings and inflorescences of the Arabidopsis T 2 transgenic lines were examined (Figs 3 and 4). At the seedling stages, GUS activity driven by the BcMF26a promoter was observed in the shoot apical meristem, developing leaf, and root tip, no activity was observed in the hypocotyl (Fig 3A to 3E). The BcMF26b promoter drove GUS expression in all parts of the seedling (Fig 4A to 4E), although the signal in the shoot apical meristem was weak (Fig 4B). At the reproductive stages, obvious differences were observed in inflorescences. BcMF26a promoter drove GUS expression only in the pistils at the early stages of flower bud development, and the expression signal decreased with the growth of the pistils (Fig 3F to 3L). Whereas, BcMF26b promoter drove GUS expression in pedicels and anthers at the early stages of flower bud development, as well as in the pistils and filaments at the late stages of development (Fig 4F to 4M). The transverse sections of the flower buds at different flower bud developmental stages for BcMF26a (Fig 3M  and 3N) and BcMF26b (Fig 4N to 4P) promoter-transformed plants also confirmed the results described above. BcMF26a promoter activity was still only observed the in pistils at the early developmental stages (Fig 3M). However, BcMF26b promoter activity was also detected in the tapetum at the binucleate microspore stage ( Fig 4O) and in pollen grains at the mature pollen stage ( Fig 4P). These results illustrate the partial overlapping at the seedling stages and spatial BcMF26a and BcMF26b proteins could be localized to the cell wall No signal peptide cleavage site was observed in the amino acid sequences of both BcMF26a and BcMF26b as predicted by SignalP 4.1 server. In the subcellular localization experiment, the cells transformed with pFGC-BcMF26a: GFP, pFGC-BcMF26b: GFP, and pFGC: GFP vectors showed obvious GFP signal in the whole cells. Then, the epidermal cells were plasmolyzed to determine whether the proteins can be localized to the cell wall. The GFP signal in the cells transformed with empty vector plasmid was only observed in the cytoplasm (Fig 5E). Whereas, the fluorescence signals in the cells transformed by the pFGC-BcMF26a: GFP and pFGC-BcMF26b: GFP vectors were observed in the cytoplasm, cell wall, and space between the cell membrane and cell wall (Fig 5A and 5C). Moreover, the fluorescence signals between the cell membrane and cell wall were distributed radially, linking the cell membrane and cell wall. These results indicated that BcMF26a and BcMF26b proteins can localize in cell wall.  the control plants (Fig 6), indicating that the expression of the two genes was co-inhibited simultaneously and fairly.
The transgenic lines, bcmf26a/b-3, bcmf26a/b-4, and bcmf26a/b-5, which represent the average inhibitory level, were used for further analysis. The   Alexander staining was used to analyze pollen viability (Fig 7A and 7B). The statistical results showed that 37.5% to 47.8% of the bcmf26a/b pollen grains were nonviable, whereas this value was only~2.1% for the control plants (Fig 8). Meanwhile, DAPI staining showed that the nonviable pollen grains of bcmf26a/b plants contained neither vegetative nuclei nor reproductive nuclei (Fig 7E and 7F). By contrast, the pollen grains from the control plants contained normal vegetative nuclei and reproductive nucleus (Fig 7C and 7D). SEM examination was used to detect the surface phenotype of pollen grains. The results revealed that mature pollen grains of the control plants were uniformly spheroid with finely reticulate ornamentation (Fig 7G and  7H). On the contrary, most of the pollen grains in the bcmf26a/b lines formed irregular clumps and exhibited abnormal reticulate ornamentation and germinal furrows (Fig 7I to 7N). Statistically, 38.9% to 71.1% of bcmf26a/b pollen grains exhibited irregular shapes, but this percentage was only~3.75% for the control plants (Fig 8).
In vitro pollen germination analysis showed none of the germinated pollen tubes of the control pollen grains burst or displayed abnormal shapes after 4 h of incubation (Fig 9A). The germination rate was~92.8% (Fig 9C). Whereas, in the bcmf26a/b lines (Fig 9B), the nonviable pollen grains could not germinate; the germinated abnormal pollen tubes exhibited two defective morphologies: pollen wall burst at the beginning of germination and pollen tube tip burst during the elongation process. The percentage of normal pollen tube was only 27.7% to 35.1% for the bcmf26a/b lines (Fig 9C).
BcMF26a and BcMF26b co-inhibition led to pollen deformities with abnormal intine development Semi-thin section analysis of the anthers for the control and bcmf26a/b lines was performed to further clarify pollen development of the transgenic plants (Fig 10). At Stage I and Stage II, no obvious differences were observed in the anthers between the control plants (Fig 10A and 10B) and the bcmf26a/b lines (Fig 10F and 10G). At Stage III, uninucleate microspores with large vacuoles were distributed in the anther locules of the control anthers (Fig 10C), whereas most of the microspores shrunk and deformed in the bcmf26a/b anthers (Fig 10H). At Stage IV, the control anthers were filled with well-developed pollen grains, and the tapetum began to degenerate ( Fig 10D). However, the bcmf26a/b pollen grains exhibited abnormal shapes, indicating shriveled, vacuolated, and swollen phenotypes (Fig 10I). The anthers became dehiscent and mature pollen grains were densely stained at Stage V in the control plants ( Fig 10E). The surviving defective pollen grains in the bcmf26a/b lines were shriveled, swollen with dense staining, or vacuolated and could not be stained (Fig 10J).
Anther transverse sections were observed by TEM. The differences between the control and bcmf26a/b pollen were observed at uninucleate, binucleate, and trinucleate stages during intine formation. For the control pollen at uninucleate stage (Fig 11A), the exine layers took shape and underwent further thickening, whereas no exine deposition occurred in the regions of the three prospective germinal furrows; meanwhile, the outer and inner layers of intine began to form, which were much thicker in the germinal furrows than that in the other regions. At the binucleate stage (Fig 11B), intine within the germinal furrows underwent further thickening. At the trinucleate mature pollen stage (Fig 11C), the shape of intine both outside ( Fig 11D) and inside ( Fig 11E) the germinal furrow regions was normal; the vegetative nuclei and generative nucleus also formed normally. Compared with the control pollen, the bcmf26a/b pollen mainly exhibited two kinds of deficient phenotype: vacuolated pollen grains (Fig 11F to 11H) and swollen pollen grains (Fig 11K to 11M), with percentages of~43.7% and~10%, respectively. For the formation of the vacuolated pollen grains, the pollen cytoplasmic inclusions began to degrade at the uninucleate stage (Fig 11F), and the nuclear membrane was shriveled. Continuously, the nuclei was disappeared at binucleate stage (Fig 11G), although some cytoplasmic residues still could be observed. At the mature pollen stage (Fig 11H), the pollen cytoplasmic inclusions completely disappeared. Moreover, the development of intine was disorderly. Intine disappearance (Fig 11I) or abnormal thickening (Fig 11J) out of the germinal furrows was detected. For the formation of the swollen pollen grains, the development of pollen cytoplasmic inclusions and the differentiation of the vegetative nuclei and generative nucleus    were abnormal. At the binucleate stage (Fig 11L), the pollen grain was filled with dense cytoplasm, no obvious vegetative nuclei and generative nucleus was observed. Moreover, the germinal furrows were pointing outward. At the trinucleate stage (Fig 11M), pollen cytoplasmic inclusions further changed, with number of spherical substances. Meanwhile, the intine outside ( Fig 11N) and inside ( Fig 11O) the germinal furrow regions exhibited abnormal thickening compared with the control pollen. No obvious difference in the bcmf26a/b anther was observed at Stage I and Stage II compared with that in the control: the nuclei exhibited normal division and differentiation, and the cell organelles formed normally.

At4g33440 was required for normal pollen tube growth
At4g33440 is the orthologous gene of BcMF26a and BcMF26b. In the wild-type plants of A. thaliana, At4g33440 had higher expression levels in inflorescences compared with those in roots, stems, and rosette leaves, as confirmed by RT-PCR analysis (Fig 12B). The two knockout mutants of At4g33440 (SALK_028430 and SALK_013967) were characterized (Fig 12A and  12C) to better investigate the biological functions of BcMF26a and BcMF26b and take advantage of the T-DNA pool. The growth status of plants, Alexander staining of pollen grains, and seed numbers in the mature siliques showed no obvious differences between the SALK_013967 mutant and wild-type plants ( S6 Fig).
In vivo pollen germination experiment was performed to characterize pollen tube movement. The pollen grains from wild-type plants and SALK_013967 mutant were hand-pollinated on the pistils of the SALK_013967 mutant. After aniline blue staining, the pollen tubes in the pistils were visualized by fluorescent microscopy. After 4 h of pollination, pollen tubes that germinated from wild-type pollen grains traveled a substantial distance into the transmitting tract in the SALK_013967 pistils (Fig 12D), whereas, those that germinated from the SALK_013967 mutant pollen grains only slightly penetrated the ovary chamber (Fig 12E). After 12 h of incubation, the pollen tubes that germinated from the SALK_013967 mutant pollen grains penetrated approximately half of the way into the ovary chamber (Fig 12F), whereas those that germinated from the wild-type pollen grains were penetrated longer (Fig 12G). However, the pollen tubes that germinated from the wild-type and SALK_013967 mutant pollen grains could reach the bottom of the transmitting tracts at 24 h after pollination (Fig 12H and 12I). These findings indicated that the absence of At4g33440 may delay pollen-tube growth to some extent, but not affect the final fertilization and seed numbers. Similar results were also observed in the SALK_028430 mutant.

Discussion
BcMF26a and BcMF26b are PG genes that originated from a segmental chromosomal duplication PG genes contain four motifs (motif I, SPNTD; motif II, GDDC; motif III, GPGGHG; motif IV, RIK) and belong to the glycosyl hydrolase family 28 [39]. They were classified as endo-PGs, exo-PGs, and rhamno-PGs. A total of 225 PG genes were grouped as endo-PGs in Clades A and B, exo-PGs in Clades C and D, and rhamno-PGs in Clade E, the Clade F members could not be clearly defined as either endo-or exo-PGs [22]. PG genes belonging to Clade E are ubiquitously expressed and only contain three typical structure motifs (I, II, and IV) [22,23]. BcMF26a and BcMF26b are orthologous genes of the PG gene At4g33440 that contain motifs I, II, and IV of PG protein. Moreover, phylogenetic analysis indicated that BcMF26a and BcMF26b are clustered into Clade E with other PG genes from different species. Thus, BcMF26a and BcMF26b are two putative PG genes and may be with rhamno-PGs activity. These characteristics are similar to the PG gene BcMF24 previously cloned in B. campestris [40].
The mechanisms of gene duplications are as follows: local (tandem) duplication, tetraploidy (or polyploidy), chromosomal segmental duplication, and single gene transposition-duplication [41,42]. Among these mechanisms, chromosomal segmental duplication covers 89% of the Arabidopsis genome [43]. It is defined as a number of genes of a whole chromosome are duplicated as a segment and they begins with the fractionation process [42]. The retained, segmentally duplicated genes are called syntenic paralogs. According to the results, the regions closely surrounding BcMF26a and BcMF26b on chr1 and chr3 are syntenic conserved chromosomal blocks. These regions are interspersed and rearranged as fragments with high similarity, suggesting that BcMF26a and BcMF26b originated as a result of a chromosomal segmental duplication.
Differential expression characteristics of BcMF26a and BcMF26b are probably caused by the differences in promoter and intron sequences The BcMF26a and BcMF26b promoters can drive the GUS signal in leaves and root tips of seedlings. In addition, their orthologous gene At4g33440 can express in roots, stems, leaves, and inflorescences. Thus, both BcMF26a and BcMF26b retained some expression characteristics of At4g33440 during the evolutionary process. The expression level of BcMF26b was much higher than that of BcMF26a in all the organs and tissues. Moreover, BcMF26a was hardly expressed in all the detected organs and tissues. Meanwhile, at the seedling stages, the BcMF26b promoter could drive GUS expression in hypocotyl, whereas the BcMF26a promoter could not. At the reproductive stages, GUS activity of BcMF26b was observed in the pedicels, pollen grains, tapetum, pistils, and filaments, whereas the BcMF26a promoter only drove GUS signal in the pistils. Therefore, BcMF26a and BcMF26b are two duplicated genes with expression divergences.
Duplicated genes initially have identical sequences and functions but tend to diverge in regulatory and coding regions. Divergence in regulatory regions can result in shifts in expression pattern [44]. Genes with the same biochemical functions may also be expressed at different times or in different places [45]. ATX1 and ATX2 were two highly conserved duplicated genes in Arabidopsis. Although structurally similar, their regulatory sequences differed, resulting in distinct temporal and spatial expression patterns [46]. BcMF26a and BcMF26b had the same number of exons and introns. They shared significant homology in their ORFs and amino acid sequences. However, the regulatory sequences upstream of 'ATG' were significantly different. Furthermore, changes in cis-regulatory modules (promoters) of duplicated genes may lead to specific shifts in expression patterns between duplicated genes [47,48]. Therefore, the differential expression patterns between BcMF26a and BcMF26b are probably caused by the differences between their prompter regions. Moreover, duplicated genes show significant homology within coding regions, but none within introns, may also have differences in expression patterns and functions. For example, two maize duplicated FIE genes (FIE1 and FIE2) share significant homology over their coding regions, but their expression pattern significantly differ because of the differences in their cis-acting elements and introns [49]. Meanwhile, Xu et al. found that divergences in exon-intron structure have been very prevalent in duplicated genes and, in many cases, have led to the generation of functionally distinct paralogs [44]. BcMF26a and BcMF26b have the same numbers of exons and introns, and the intron phases are also consistent. However, the lengths and sequence similarities of the four introns between the two genes were largely divergent. Therefore, the divergences of introns may also contribute to the generation of different expression patterns between BcMF26a and BcMF26b. For these hypothesizes, more experiments will be needed.
BcMF26a and BcMF26b proteins may play roles for cell wall construction and secret by unconventional secretory mechanism PG proteins involving in the degradation of pectin are considered to be localized in the cell wall [50]. Compared with the control, subcellular localization of BcMF26a and BcMF26b proteins showed that the fluorescence signals could be localized to the cell wall. Therefore, BcMF26a and BcMF26b may play roles in cell wall construction. The processes of eukaryotic protein secretion include two mechanisms: signal peptide-dependent secretory transport and unconventional protein export [51]. Although a number of secretory proteins with defined extracellular functions do not possess functional signal peptides, they are transported to extracellular space by unconventional protein export [52]. In the present study, no signal peptide cleavage site was present in amino acid sequences of BcMF26a and BcMF26b proteins by Sig-nalP 4.1 server prediction. Thus, they may be released by unconventional protein export. Furthermore, in the plasmolyzed onion epidermal cells, the fluorescence signal was also observed in the cytoplasm. It can be inferred that the proteins were transported by vesicles which fused with the plasma membrane. The fluorescence signal existed in the cell space between the cell membrane and cell wall. Moreover, it was radially distributed, linking the cell membrane and cell wall. This may be the method by which BcMF26a and BcMF26b proteins move from the plasma membrane to the cell wall.

BcMF26a/b impairs pollen development and pollen tube growth by modulating intine information
The analysis of complete eukaryotic genome sequences has revealed that gene duplication is indeed thriving [53]. Furthermore, many eukaryotic organisms have had their whole genome duplicated, sometimes more than once, such as Brassiceae [54]. Functional redundancy or complementarity has occurred among many genes. Natural plant miRNAs have a very narrow action spectrum and target only mRNAs with few mismatches [34]. With highly specific gene silencing by amiRNAs, single-, double-, and triple-mutant analysis have been used for redundancy studies [55]. Such studies inspired us to inhibit BcMF26a and BcMF26b independently and specifically using amiRNA technology so that their functions can be studied. Unfortunately, BcMF26a and BcMF26b shared high similarity over their ORFs, and the length of continuous differential bases in the ORFs could not meet the designed requirements for singleinhibition. Considering that PG was encoded by a large gene family in plants, we designed the multiple-target amiRNA to inhibit the expression of BcMF26a and BcMF26b simultaneously. This design excluded the functional redundancy of other PG genes in the family.
The PG genes involving in pollen wall development are considered to be necessary for pollen development and pollen tube formation [56]. In the T-DNA insertion mutants of At4g33440, pollen tubes growth was delayed. A similar phenomenon was observed in the mutants of COBRA-LIKE 10 (COBL10), which caused gametophytic male sterility because of reduced pollen tube growth and compromised directional sensing in the female transmitting tract [57]. Obvious defective phenotypes were also observed in bcmf26a/b, with pollen viability of only 37.5% to 47.8% and the germination rate of normal pollen tube was only 27.7% to 35.1%. Therefore, the absence of BcMF26a/b seriously impaired pollen development and pollen tube growth.
Pectin is a major component of pollen intine, and it is also enriched at the germinal furrows [3,58]. With LM5 mAb, pectin was immunolocalized in pollen and pollen tube of olive, especially at all the three apertures [59]. As a hydrolase and loosening enzyme, PG disintegrates the cell wall structure through pectin degradation [6]. According to anther semi-thin section analysis, the most dramatic aberrations in the bcmf26a/b transgenic plants were pollen deformities characterized by swelling with dense staining and vacuoles that could not be stained. TEM revealed that the pollen grains underwent a disordered formation in the intine region. Abnormal thickening of the intine inside and outside the germinal furrows was observed. Meanwhile, the development of cytoplasmic materials and nucleus was also disorderly. Given that intine synthesis is largely under the control of the microspore [60], it is credible that the cytoplasmic inclusions and intine are disordered simultaneously. Therefore, BcMF26a/b may affect pollen development by disrupting intine formation. These aforementioned phenotypes are also similar to the osgt1 and cap1 mutants, which affected pollen development and maturation by modulating the abnormal intine structure. GLYCOSYLTRANSFERASE1 (OsGT1) encodes glycosyltransferase, which is essential for intine construction and pollen maturation; in osgt1 pollen, the intine structure was disrupted and the grains were shrunken [61]. The collapsed abnormal pollen1 (cap1) mutant produces abnormal pollen grains, which lack almost all cytoplasmic materials, nuclei, and intine, and cannot germinate [62].
During pollen germination, the outer layer of the pollen tube wall appears to be an extension of the pectinaceous intine of the pollen grain [63]. The apical wall layer is rich in methylated pectic to allow cell expansion, and the shank wall consists of demethylated pectin [64]. Shortly after pollen grains are exposed to water, the exine abruptly breaks with pressure accumulated from the expansion of the intine; in optimal conditions, the germinating pollen cell emerges from one end of the pollen grain and then travels through the appropriate germinal furrow of the intine [13]. During in vitro germination, the bcmf26a/b pollen tubes burst at the beginning of the germination, or the pollen tube's tip burst during the elongation process. Those abnormal morphologies are possibly caused by the thickened intine, which may increase the pressure for pollen germination and pollen tube growth.
BcMF2 and BcMF9 are two PG genes in B. campestris. Based on our obtained phylogenetic tree, these two genes both belonged to Clade C. BcMF2 was specifically expressed in the tapetum and pollen after the tetrad stage; in bcmf2 plants, mature pollen presented a distorted morphology in the intine, leading to abnormal pollen tube growth and a consequent reduction in male fertility [13]. BcMF9 was also expressed in the tapetum and microspores during the late stage; it affected pollen and pollen tube development by playing a role in intine and exine formation [14]. In the present study, BcMF26a and BcMF26b are also PG genes in B. campestris, but they belong to Clade E. In bcmf26a/b plants, mature pollen also presented a distorted morphology in the intine. However, the expression of BcMF26b in anthers was at the early stages of pollen development. Moreover, BcMF26b did not specifically express in anther; for example, the activity of its promoter was also observed in pedicels. Whether the expression of the target genes in other organs also contributes to the phenotypes of bcmf26a/b plants has not been verified.
BcMF26b may play a major role in pollen development of the bcmf26a/b plants In the T-DNA insertion mutants, the absence of At4g33440 could only delay the growth rate of pollen tubes, whereas, both the pollen grains and pollen tubes all showed defective phenotypes in bcmf26a/b transgenic plants. Therefore, BcMF26a/b may be the result of neofunctionalization during evolution. According to our results, the expression level of BcMF26b was much higher than that of BcMF26a in all the tested organs and tissues. The BcMF26a promoter activity was only observed in pistils at the early stages of flower buds development, while BcMF26b promoter could drive GUS signal in anther and pollen. Therefore, BcMF26b may play major roles in pollen development and pollen tube formation in bcmf26a/b plants, though we designed multiple-targeted amiRNA to realize co-inhibition. In addition, bcmf26a/b plants had normal pistils consistent with the control plants during the flowering period. Thus, BcMF26a gene may have a weaker function that could not affect the growth and development of plants. Furthermore, cell separation has been recognized for its involvement in plant development, which includes pod abscission and dehiscence [65]. Although the BcMF26b promoter could drive GUS expression in pistils and filaments at the late stages of flower buds development, the morphologies of pistils and filaments in bcmf26a/b plants were normal. This finding may be due to the possible that BcMF26b also participates in the senescence and abscission of pistils and filaments as a cell wall hydrolytic enzyme, so it cannot affect the early growth and development of the two organs.