2.1. The Floral Bud Size Is Associated with the Microspore Developmental Stage
No morphological difference between the CMS line and the corresponding maintainer line was observed except for the anther. CMS lines have empty yellow anthers whereas maintainer lines have full and dehiscent anthers (
Figure 1). No histological differences were observed between the CMS line 9802A1 and the corresponding maintainer line 9802B1 in the anther at the tetrad stage when tetrads were enclosed by a thick callose wall in our previous study [
29]. After the tetrad stage, microspores released from tetrads underwent cell expansion and cell wall synthesis in male fertile anthers. In contrast, most microspores from CMS anthers grew slowly and were malformed, and the tapetal cell grew abnormally large with increased vacuoles [
29]. The cytoplasm of 9802A1 was introgressed into HYBP-B and YH-B at more than 16 times of backcrossing. The resulting CMS lines, HYBP-A and YH-A, display different morphologies, indicating that they differ in their nuclear genomes. In this work, the stage of pollen abortion in HYBP-A and YH-A is the same with that of 9802A1 suggesting that different nuclear backgrounds could not change the pollen abortion phase from the same CMS cytoplasm. In the four lines, HYBP-A/B and YH-A/B, when the buds reached about 0.8 mm, anthers were at the microspore mother cell stage (
Figure 2A). Microspore mother cells were undergoing meiosis when the buds were about 1.1 mm in length (
Figure 2B). When the bud length was about 1.5 mm, individual tetrads could be seen (
Figure 2C). In HYBP-B and YH-B, the phase of the vacuolate microspore appeared in buds about 2.0 mm in length (
Figure 2D), with a single large vacuole displacing the single nucleus to one side of the cell.
Figure 1.
Partially dissected radish flowers. Male fertile plant (A); and male sterile plant (B). Bar = 1 mm.
Figure 1.
Partially dissected radish flowers. Male fertile plant (A); and male sterile plant (B). Bar = 1 mm.
Figure 2.
Development of anthers in the maintainer line HYBP-B (A–D) and the CMS line HYBP-A (E–H). Anther from floral bud at about 0.8 mm long (A,E); Anther from floral bud at about 1.1 mm long (B,F); Anther from floral bud at about 1.5 mm long (C,G); Anther from floral bud at about 2.0 mm long (D,H). Bar = 20 μm.
Figure 2.
Development of anthers in the maintainer line HYBP-B (A–D) and the CMS line HYBP-A (E–H). Anther from floral bud at about 0.8 mm long (A,E); Anther from floral bud at about 1.1 mm long (B,F); Anther from floral bud at about 1.5 mm long (C,G); Anther from floral bud at about 2.0 mm long (D,H). Bar = 20 μm.
2.3. De Novo Assembly and Functional Classification
High-quality clean reads were de novo assembled into 130,240 contigs with an average length of 570 bp and a N50 length of 821 bp. Sequences that did not extend on either end were defined as unigenes. In total, 67,140 unigenes showed similarity to known genes in at least one of the five public databases including NR, Swiss-Prot, GO, KOG (eukaryotic ortholog groups) and KEGG.
GO annotation including three ontologies (molecular function, cellular component and biological process) is an international gene classification system that provides a dynamically-updated standardized vocabulary for assigning functions of genes and their products [
30]. BLAST2GO program was employed to find GO terms for assembled unigenes and 29,031 unigenes were assigned at least one GO term in 55 functional groups. Of these, most were assigned to the biological process ontology (86,785; 41.2%), followed by cellular component (78,749; 37.4%) and molecular function (45,090; 21.4%) ontologies (
Figure 3).
The KOGs, based on orthologous relationships between genes, include proteins encoded by seven sequenced eukaryotic genomes [
31]. In total, 26,527 unigenes were assigned to the KOG classification. Because some unigenes were annotated with multiple KOG functions, a total of 29,353 functional annotations were obtained. The four largest groups annotated were “posttranslational modification, protein turnover, chaperones” (3171 unigenes), “general function prediction only” (3110 unigenes), “signal transduction mechanisms” (2893 unigenes) and “Translation, ribosomal structure and biogenesis” (2237 unigenes) (
Figure 4).
Figure 3.
Gene ontology classifications of unigenes from the radish floral bud transcriptome. The x-axis indicates the sub-categories; the y-axis indicates the number of unigenes in a sub-category.
Figure 3.
Gene ontology classifications of unigenes from the radish floral bud transcriptome. The x-axis indicates the sub-categories; the y-axis indicates the number of unigenes in a sub-category.
Figure 4.
KOG function classification of unigenes from the radish floral bud transcriptome. The x-axis indicates the number of unigenes in a function class; the y-axis indicates the function classes.
Figure 4.
KOG function classification of unigenes from the radish floral bud transcriptome. The x-axis indicates the number of unigenes in a function class; the y-axis indicates the function classes.
The KEGG pathway records molecular interaction networks such as pathways and complexes. Sequence similarity search against KEGG GENES identified 5706 unigenes assigned to 323 KEGG pathways. The result showed that the three largest pathway categories were “carbohydrate metabolism” (803 members), “translation” (595 members) and “signal transduction” (563 members) suggesting that these pathways could be crucial for radish anther development (
Figure 5).
Figure 5.
Classification based on categories of KEGG pathways. Cellular Processes (A); Environmental Information Processing (B); Genetic Information Processing (C); and Metabolism (D). The x-axis indicates the number of unigenes in a function class; the y-axis indicates the function classes.
Figure 5.
Classification based on categories of KEGG pathways. Cellular Processes (A); Environmental Information Processing (B); Genetic Information Processing (C); and Metabolism (D). The x-axis indicates the number of unigenes in a function class; the y-axis indicates the function classes.
2.4. Comparison of Gene Expression Levels between CMS and Maintainer Lines
RNA-Seq gene expression levels were measured using reads per kilobase of exon model per million mapped sequence reads (RPKM) [
32]. RPKM values for each gene were compared between the CMS and corresponding maintainer lines to screen for DEGs. We identified 3843 DEGs (2487 up-regulated and 1356 down-regulated genes) in HYBP-A floral buds compared with HYBP-B and 2041 DEGs (676 up-regulated and 1365 down-regulated genes) in YH-A floral buds compared with YH-B. Of these, 539 DEGs exhibited the same direction of expression change in the two CMS lines relative to their corresponding maintainer lines. The transcription level of a total of 127 transcripts was increased and 412 transcripts was decreased in both HYBP-A and YH-A compared with their corresponding maintainer lines (
Table S2). These data suggest that combination analysis of two groups of CMS/maintainer lines with different nuclear background was useful to decrease false positives. GO annotation for these DEGs showed that 268 genes were categorized into 50 functional groups with 1629 functional terms indicating that some genes were annotated with multiple terms (
Table S3). The DEGs were assigned into nine subcategories in the “molecular function” category, and the two most abundant terms were ‘binding” (149 members) and “catalytic activity” (122 members) (
Figure 6). Among “cellular component”, “cell” (199 members), “cell part” (199 members) and “organelle” (108 members) were the three most strongly represented terms (
Figure 6). For “biological process”, the three predominant terms for these DEGs were “metabolic process” (141 members), “cellular process” (135 members), “response to stimulus” (83 members) followed by “biological regulation” (50 members) (
Figure 6).
There are a total of 79 protein-encoding genes from the radish chloroplast (cp) genome [
33], and we retrieved transcripts for 72 cp protein-encoding genes from radish floral buds at this work (transcripts for cp protein genes,
psaJ,
petG,
ndhK,
psaB,
rps14,
psbC and
psbD, were not detected). In contrast, mitochondrial (mt) genome from radish contains 34 protein genes [
34], but only 16 protein transcripts (
atp1,
atp4,
atp6,
atp9,
ccmFC,
cob,
cox2,
cox3,
nad4,
nad4L,
nad5,
nad9,
orf138,
rpL16,
rps3 and
rps4) were detected from the transcriptome. No difference in levels of transcripts between CMS and maintainer lines was found from these annotated mt and cp transcripts in the present work, except CMS-inducing mt
orf138, which was only expressed in the CMS lines.
Figure 6.
Gene Ontology classification of down- and up-regulated unigenes in floral buds in CMS lines compared with maintainer lines. The x-axis indicates the sub-categories; the y-axis indicates the number of unigenes in a sub-category.
Figure 6.
Gene Ontology classification of down- and up-regulated unigenes in floral buds in CMS lines compared with maintainer lines. The x-axis indicates the sub-categories; the y-axis indicates the number of unigenes in a sub-category.
Transcription factors (TFs) play important roles in complex biological processes under a wide range of environmental signals by regulating gene transcription through binding to specific DNA sequences in the promoters of multiple target genes [
35]. At least 608 transcription factors (TFs) from
Arabidopsis divided into 34 families were identified in the male gametophyte [
36]. Here, 25 TF genes from 11 gene families were down regulated in the radish CMS lines. The tapetum directly surrounding the developing male gametophyte plays a crucial role in the microspore development and maturation by supplying necessary nutrients and structural components. In both rice and
Arabidopsis at least two bHLH factors (UDT1 and TDR, and DYT1 and AMS, respectively) were shown to be important in tapetum development [
37,
38,
39,
40]. The timing of tapetal PCD is crucial for microspore development. T-DNA knockout of
Arabidopsis MYB80 (encoding an MYB transcription factor) triggers premature tapetal PCD resulting in male sterility, indicating that
MYB80 is important for tapetal PCD at the proper developmental stage [
41]. In the present study, three bHLH and three MYB genes were found to be down-regulated in CMS lines. Since functions of their corresponding
Arabidopsis genes (AT1G26610, AT3G19580, AT5G04340, AT3G50060, AT5G52660, and AT4G37260) are still unkown, it is worth studying whether these genes are involved in radish CMS.
To date, two C
2H
2 zinc-finger transcription factors (DAZ1 and DAZ2) and two WRKY transcription factors (WRKY34 and WRKY2) in
Arabidopsis are responsible for male gametogenesis. DAZ1 and DAZ2 promote G2- to M-phase transition in pollen [
42]. WRKY34 and WRKY2 are required for pollen development at early stage [
43]. Since the abortive phenotype for our radish CMS is cytologically observed shortly after the tetrad stage, it remains to be determined whether three DEGs encoding a C
2H
2 zinc-finger transcription factor and two WRKYs are involved in early events of the radish CMS.
Heat shock transcription factors (HSFs) and heat shock proteins (HSPs) play key roles in signal transduction in the heat stress response. Several studies suggest that some HSFs and HSPs are related to gametophyte development in the absence of heat stress [
44,
45,
46]. In current work, five HSF and 19 HSP genes were down-regulated in radish CMS lines. Especially, AtHSP23.6 is targeted to mitochondria [
47], and the maize homolog
hsp81 is strongly expressed at the pre-meiotic and meiotic stages of microspore development without heat shock [
45].
Many fertility restorer genes for the plant CMSs gene were identified as belonging to the PPR family [
8,
17,
18,
19,
20,
21,
22]. In this work, three PPR genes were downregulated in the CMS lines. Their orthologs (GenBank accession numbers AT2G20540, AT1G33350 and AT1G71060) from
Arabidopsis were predicted to target to the mitochondrion. Interestingly, one ortholog
MEF21 from AT2G20540 is involved in the C-to-U transcript editing of the mitochondrial gene
cox3 [
48]. These PPR DEGs would be very valuable for further study of their roles in the radish CMS.