Identification of tapetum-specific genes by comparing global gene expression of four different male sterile lines in Brassica oleracea

Plant materials

Cytological analysis using microscopes

Light microscopy and transmission electron microscopy (TEM) were used to investigate the cytological differences between the mutants and wild-type. Different sized buds were fixed overnight in 4 % glutaraldehyde with 200 mM phosphate buffer (pH 7.0) and then rinsed overnight in 200 mM phosphate buffer (pH 7.0). Next, buds were post-fixed in a solution containing 1 % osmium tetroxide for 2 h and washed in a dehydrated ethanol series for 30 min at each step (2 × 50, 60, 70, 80, 90, and 2 × 100 %). Dehydrated tissues were embedded in Spurr’s epoxy resin (Spurr 1969; Ted Pella, Redding, CA) for 3 days at 60 °C, and sectioned at 60–90 nm using a microtome (Leica Ultratome V, LKB, Bromma, Sweden). Anther transverse sections were stained in 4 % uranyl acetate for 20 min and in lead citrate for 3 min. All specimens were analyzed using TEM (H-8100, Hitachi, Tokyo, Japan). For the light microscopy analysis, buds were fixed overnight in FAA (50 % ethanol, 5.0 % glacial acetic acid, 3.7 % formaldehyde), dehydrated in a graded ethanol series (2 × 50, 60, 70, 85, 95, and 3 × 100 %), embedded in Spurr’s epoxy resin, and sectioned in 1 µm thick slices using a microtome. Anther transverse sections were stained in 1 % toluidine blue at 42 °C for 1–2 h and observed under a compound microscope (Olympus Model BH2, Tokyo, Japan).

RNA preparation, probe labeling, and microarray hybridization

For total RNA isolation, all flower buds above the last opened flowers of three flowering branches from three MS and three MF plants were collected in duplicate and combined to reduce plant-to-plant variation. Total RNA was isolated using TRIzol following the manufacturer’s instructions (Invitrogen, Beijing, China). RNA was further purified using an RNeasy Mini kit (Qiagen China Co., Ltd., Shanghai, China) and the quality was checked using an Agilent 2100 bioanalyzer (Agilent Technologies China Co., Ltd., Shanghai, China) and RNA 6000 Nano kit (Agilent Technologies China Co., Ltd., Shanghai, China) before labeled cRNA was synthesized. Cy3- and Cy5-labeled cRNA was synthesized from 400 ng total RNA using a Low RNA Input Linear Amplification and Labeling kit Plus (Agilent Technologies China Co., Ltd., Shanghai, China) following manufacturer’s protocol. Swap labeling of the other MS and MF (three plant pools) RNAs was conducted for the replications. Labeled cRNA was hybridized onto a 22 K Arabidopsis oligo microarray (Agilent Technologies China Co., Ltd., Shanghai, China) using the In situ Hybridization kit Plus (Agilent Technologies China Co., Ltd., Shanghai, China).

Data acquisition, normalization, and gene annotation analysis

Hybridized microarrays were scanned sequentially for Cy3- and Cy5-labeled probes with a laser scanner (G2655AA, Agilent Technologies China Co., Ltd., Shanghai, China) at a resolution of 10 µm and a PMT of 100. The intensities were normalized by linear LOWESS (Yang et al. 2002). The signal was considered positive when the signal/noise value was >5. To determine pollen stage specificity and co-expression information, our data were compared with array data from two other studies. The pollen transcriptome refers to the dataset from (Honys and Twell 2004), and the stamen transcriptome refers to the dataset from (Wellmer et al. 2004). Microsoft Office Excel (Excel 2010, Microsoft China Co., Ltd., Beijing, China) was used to manage and filter the microarray data. Differently expressed genes (DEGs) were functionally categorized based on the ontology annotation of the Arabidopsis genome from the Arabidopsis Information Resource (http://www.arabidopsis.org). Pollen expression type was determined based on the maximum expression value (MaxP) from the four pollen stages (uninucleate microspore, bicellular pollen, tricellular pollen, and mature pollen), the maximum expression value (MaxS) from seven sporophytic tissues (cotyledon, leaves, petiole, stems, roots, root hair zone, and suspension cell cultures), or the value of MaxP/MaxS according to the pollen transcriptome dataset (Honys and Twell 2004). Late pollen genes were defined as those expressed after microspore mitosis during male gametophyte development, which had continued transcript accumulation during pollen maturation (McCormick 1993).

Reverse transcription-polymerase chain reaction (RT-PCR)

In situ hybridization

Arabidopsis Col-0 inflorescences were embedded in Paraplast (Sigma-Aldrich, Shanghai, China), sectioned at 8-μm thickness and mounted onto precharged slides. For sense and antisense probe synthesis, five coding regions of the NPGs, MEE48, A9, CYP98A8, EXL6, and GGPS5, resulting in 990-, 895-, 749-, 552-, and 656-bp DNA templates, were PCR amplified from flower cDNA using gene-specific forward and reverse primers. A T7 polymerase binding site was incorporated into the forward primer for sense probe amplification and in the reverse primer for antisense probe amplification. Digoxigenin-labeled probes were transcribed off the template using T7 polymerase (Roche, Shanghai, China). Probes were shortened to 200-bp fragments by limited carbonate hydrolysis, and then quantified and hybridized to slides. Tissue fixation, embedding, hybridization, and signal detection were performed as described by (Hooker et al. 2002).

Cytological defects in the four B. oleracea MS lines

The correct spatiotemporal expression of genes in the anther is required for normal tapetum development. We clarified the sequential appearance and characteristics of the cytological defects of the four B. oleracea MS lines by comparing them with the wild-type (Fig. 1I). Light microscopy of the main anther developmental stages revealed that the abortive phenotypes appeared successively in the NiCMS line (Fig. 1I-7), the RGMS line (Fig. 1I-14), the OguCMS line (Fig. 1I-21), and finally in the DGMS line (Fig. 1I-28). We performed a TEM analysis to characterize the defective tapetum development in the four MS mutant lines (Fig. 1II). In the wild-type line, the sporogenous cells, which give rise to pollen, are visible within locules of sectioned anthers. Concentric rings of other cell types associated with pollen development and release are differentiated around the sporogenous cells during the sporogenesis cell stage (Fig. 1II-1). Sporogenous cells develop into microspore mother cells and four single distinguishable layers of anther wall and microsporangium could be observed during the microspore mother cell stage (Fig. 1II-2). Tetrads and tapetum with normal structures, as well as a single microspore tetrad with a central large nucleus, thick cytoplasm and abundant mitochondria develop during the tetrad stage (Fig. 1II-3).Vacuolated epidermal and endothelial cells, degenerating tapetum, and a free uninucleated microspore, containing a central nucleus, clear nuclear membrane, thick cytoplasm, and abundant plastids, appear in the post-tetrad stage (Fig. 1II-4). We compared the cytological features of the four male sterility types with those of wild-type. For each, we observed unique defective features. In the NiCMS line, the tapetal cells differentiated inconspicuously, with an indistinguishable middle layer at the sporogenesis cell stage (Fig. 1II-5). In the RGMS line, the tetrad aborted once it was formatted and the tapetal separated from anther wall at the microspore mother stage (Fig. 1II-6). In the OguCMS line, the tapetums were abnormally activated and thickened continuously when meiosis finished during the early tetrad stage (Fig. 1II-7). In the DGMS line, the morphology of the tapetum was not affected (Fig. 1II-8), as reported by (Lou et al. 2007). The development of microspores in the four MS lines was affected at different stages because of the abnormal tapetum development.

Open image in new window

Genes with depressed expression levels became the focus of the research because the cytological observations indicated that MS lines were blocked by separate MS proteins. We hypothesized that a sequential developmental interruption model would clarify the gene expression sequence and be in accordance with the cytological results (Fig. 2). Based on the appearance point of the four MS phenotypes (NiCMS earlier than RGMS, RGMS earlier than OguCMS, OguCMS earlier than DGMS), the genes down-regulated only in the NiCMS lines were considered to express earlier than the genes down-regulated in both NiCMS and RGMS lines, and the genes down-regulated in both NiCMS and RGMS lines were considered to express earlier than the genes down-regulated in NiCMS, RGMS, and OguCMS lines. The latest expressing genes would be those that were down-regulated in all four MS lines. Genes involving in anther development mainly express in time series. The accumulation of products produced by early-expressing genes, such as transcription factors and secreted proteins., play important roles in expression of late-expressing genes (Wilson and Zhang 2009).

Open image in new window

Signal extraction of the microarrays

To reduce plant to plant and inflorescence branch-to-branch variations, we pooled the RNA obtained from three different plants. Pooling RNA before labeling has the advantage of reducing the variation due to biological replication and sample handling. Two replicate hybridizations were performed using pooled RNA from three different sterile or fertile plants and the labels were swapped for the second slide, with biological replicates. After the quantification of the signal intensities, the data were normalized to compensate for the nonlinearity of intensity distributions and differences in probe labeling (Fig. 3). Using a signal/noise value >5 to select for positive signals, we detected 12,838 positive signals both in NiCMS control samples and NiCMS samples; 13,037 positive signals both in RGMS control samples and RGMS samples; 13,083 positive signals both in OguCMS control samples and OguCMS samples; and 11,581 positive signals both in DGMS control samples and DGMS samples. The reproducibility was determined by calculating the coefficient (R²) of the Log₂ normalized signal values of all detected signals. The R²-value between replicas were as follows: 0.9847 for the NiCMS control (Fig. 3a) and 0.9808 for the NiCMS samples (Fig. 3b); 0.9754 for the RGMS control samples (Fig. 3c) and 0.9663 for the RGMS samples (Fig. 3d); 0.9708 for the OguCMS control samples (Fig. 3e) and 0.9834 for the OguCMS samples (Fig. 3f); and 0.9772 for the DGMS control samples (Fig. 3g) and 0.9738 for the DGMS samples (Fig. 3h). The consistency of the two slides, together with the large number of detectable genes, indicates the feasibility of using the Agilent Arabidopsis 2 Oligo array to analyze the B. oleracea transcriptome.

Open image in new window

Microarray analysis

An Agilent Arabidopsis 2 Oligo array was used to compare gene expression profiles of the four B. oleracea MS lines to their MF control lines. The array contained 21,500 probes for genes or transcripts of Arabidopsis. Hybridizations of two replicates were performed using pooled RNA from each of three different sterile or fertile plants, and labels were swapped for the second slide with biological replicates. A signal/noise value of >5 was used to select positive signals; this identified 12,837, 13,036, 13,082, and 11,580 genes in NiCMS, RGMS, OguCMS, and DGMS lines, respectively. After combining these four datasets 13,984 genes (65.0 %) were detected as positive signals in at least one of the MS lines. This percentage of identified genes was similar to previous research that identified 14,660 (64.5 %) genes in six organs and structures, including the inflorescences, at different Arabidopsis floral stages (Zhang et al. 2005). Genes with differential mRNA abundance levels (ratios >3, or <0.33) in the two replicate slides were selected for further analysis. To confirm the microarray profiling data, nine genes were randomly selected for semi-quantitative RT-PCR analyses, and their expression patterns were found to be consistent with the microarray results (Fig. 4, Supplementary Table S1).

Open image in new window

There were 544, 338, 526, and 209 down-regulated genes, and 5, 45, 151, and 15 up-regulated genes detected in the NiCMS, RGMS, OguCMS, and DGMS lines, respectively (Supplementary Table S2 and S3). After redundant genes were merged, 838 DEGs were down-regulated (group 1) and 188 up-regulated (group 2) in at least one of the four MS lines. The overlap of group 1 and group 2 contained 21 DEGs, including the two tapetum-related genes TDF1 and MYB 103. These genes play vital roles in tapetum synthesis and degeneration, and the sporopollenin monomer biosynthesis process (Higginson et al. 2003). These genes were down-regulated in the NiCMS line but up-regulated in RGMS and OguCMS lines. Following the removal of these 21 DEGs, 1,005 non-redundant DEGs were identified in this research.

Identification of tapetum-specific genes

The tapetum has been studied in many kinds of plants using MS defects, and, due to the difficulty of tapetum isolation, only four genes have been reported to be tapetum specific by MS mutant gene cloning (Suwabe et al. 2008). Therefore, we developed an approach to identify tapetum-specific genes on a large-scale, and at the developmental stages in which they are expressed. Fortunately, the separation of male gametophytes is easy, and numerous male gametophyte-specific genes have been identified from multiple MS mutants and global transcriptome analyses.

To identify genes specifically involved in tapetum development, we removed genes known to be expressed in sporophytic tissues based on the findings of Honys and Twell 2004. Their research identified 17,677 sporophytic genes, of which 725 genes overlapped with the 1,005 DEGs identified in our research (Supplementary Table S4). This left 280 DEGs thought to be expressed specifically in the anthers. Furthermore, 176 male-gametophyte expressed DEGs identified by (Honys and Twell 2004) were removed from the 280 DEGs specifically expressed in the anthers, leaving 104 NPGs (Fig. 5). As the four MS lines had tapetums aborted at successive developmental stages, we were able to identify 104 NPGs as anther wall-specific genes, the vast majority of which were considered tapetum-specific genes because their anther walls developed normally with the exception of the distinct abortion of the tapetums (Table 2). We cannot completely rule out that some genes expressed from other tissues are included in the 104 NPGs, although this probability is very low.

Open image in new window

Male sterility genes block the developmental pathways of anthers at a certain stage, and lead to abnormal anther development. Because each of the four MS lines showed distinct tapetum abortive phenotypes, and their abnormal characteristics appeared successively during anther development, we could deduce the expression sequence of the NPGs by their expression patterns. The expression sequence of the 55 NPGs could be determined according to the sequential appearance of the tapetum abortive phenotypes (Table 2). First, 28 NPGs that were only down-regulated in the NiCMS line were expressed, followed by seven NPGs down-regulated in both the NiCMS and RGMS lines, then 16 NPGs down-regulated in the NiCMS, RGMS, and OguCMS lines, and finally four NPGs down-regulated in the four MS lines were expressed. It is thought these 55 NPGs constitute the main stream of tapetum development, while the remaining 49 NPGs belonged to bypass ways which also play roles in tapetum development. These results showed that the tapetum development is strongly correlated with gene expression patterns and anther developmental timing. In the eight previously reported tapetum-specific genes, five (ACOS5, A6, LTP12, LAP5, and A9) were detected in our filtered microarray results, although the other three tapetum-specific genes (TSM1, TAP35, and TAP44) were detected in our microarray analysis at variable expression levels in the four MS lines. Because of the strict filter these three tapetum-specific genes were not included in our set of 104 NPGs. There were many noticeable features in the up-regulated genes, with 13 NPGs up-regulated in the OguCMS line, five NPGs up-regulated in both the RGMS and OguCMS lines, and one NPG up-regulated in both the RGMS and DGMS lines (Table 2). This is consistent with the cytological observations of different abnormal tapetum development in the four MS lines.

The tapetum developmental network is regulated by many genetic pathways (Wilson and Zhang 2009). MS mutants occurring at different developmental stages lead to abnormal downstream reactions, including altered tapetum structure and gene expression patterns. These changes are evoked by the presence of MS-associated proteins (Fujii et al. 2010). Only a few previous studies have analyzed gene expression patterns in B. oleracea MS lines (Kang et al. 2008). These studies compared anther gene expression profiles in MS lines with their corresponding fertile lines, allowing for the preferentially expressed anther genes to be identified. Despite their clear contribution to anther developmental pathways, downstream expressed anther wall-specific genes have been overlooked because of the removal of male gametophyte-specific expressed genes.

Function of genes arrested by the four types of B. oleracea MS lines

The distribution of the NPGs was determined in the gene ontology data set (MAS 3.0, http://bioinfo.capitalbio.com/mas3/) and found to cover virtually all functional categories (Fig. 6). The classification of functional categories revealed that some were enriched in DEGs that had reduced expression levels, including structural molecules, transporters, and physiological processes. These categories are associated with metabolic activities that are dynamic in the tapetum, suggesting a positive role of the tapetum in the regulation of metabolic functions. As we are interested in the genetic mechanism of tapetum abortion in the four MS lines, genes specifically expressed in the tapetum were further analyzed to identify tapetum abortive phenotypes.

Open image in new window

Special attention was paid to abortive mechanisms leading to tapetum dysfunction by gene regulation networks within the 1,005 DEGs in any of the MS lines. According to previous studies, 22 DEGs were proven to be related to tapetum development (Supplementary Table S6). Among these 22 genes, 11 of the DEGs were reported to play important roles in tapetum development (ULT2, TDF1, PGA4, PAB3, TKPR2, PAB5, SHT, ACT12, LAP6, HMA4, and ATA20). More importantly, the other 11 DEGs were identified as NPGs in this study (LTP12, CYP703A2, CYP704B1, LAP5, ACOS5, ABCG26, MYB103, MYB99, WBC27, ATBHLH089, and ATBHLH091) (Table 2). To determine the expression patterns of NPGs, we compared them with other mutant transcriptomes that have been analyzed by bioinformatics filtering. The results showed that 27 (26 %) NPGs were also detected in an anther-specific expressed gene set (Xu et al. 2014), 32 (31 %) NPGs were detected in a stamen-specific expressed gene set that excluded pollen-specific expressed genes (Ma et al. 2012). To validate microarray results in anthers, we performed in situ hybridization using five randomly chosen NPG-derived probes and developing wild-type A. thaliana flowers (Fig. 7). We used MEE48, A9, CPY98A8, and EXL6 probes that hybridized with anthers when they had developed to the uninucleated microspore stage, respectively. GGPS5 probes were hybridized with anthers when they developed to the pollen mother cell stage. No hybridization signal was observed in locules with microspores and tapetum using any of the NPG sense control probes. However, NPG anti-sense probes resulted in varying degrees of hybridization to the tapetum. These results showed that our data set, which was generated by bioinformatics filtering, was reliable. Therefore, the NPGs can be seen as potential tapetum-specific expressed genes.

Open image in new window