Cytoplasmic male sterility (CMS) is a maternally inherited trait that fails to produce viable pollens, and has been reported in a large number of plant species. CMS is encoded in the mitochondrial genome and can arise spontaneously due to mutation in the genome (autoplasmy) or can be expressed following cytoplasmic substitutions due to nuclear-mitochondrial incompatibility (alloplasmy) (Prakash et al., 2009). A CMS fertility restoration system has been used as a suitable pollination control system to produce commercial hybrid seed for many crops, because it is easy to maintain.

Cytoplasmic male sterility phenotypes encompass a wide range of reproductive abnormalities, including degenerate anthers, aborted pollen, carpelloid and petaloid stamens (for review see Chase, 2007). Many CMS genes result from mtDNA rearrangements, as at least 10 essential mitochondrial genes are involved in the origination of CMS genes among 28 types of CMS from 13 crop species (Chen and Liu, 2014). Intriguingly, sequences of the known functional mitochondrial genes are not necessarily included in some CMS genes, for several rice CMS-associated ORFs consist of sequences of putative mitochondrial ORFs (Chen and Liu, 2014). Additionally, some CMS genes (such as orf125 in radish CMS-Kos and its variant orf138 in Brassica CMS-Ogu) are non-chimeric genes with the sequences from single source (Bonhomme et al., 1992; Iwabuchi et al., 1999). CMS genes have been shown to cause mitochondrial dysfunction by disrupting the mitochondrial membranes, reducing the mitochondrial ATP level and increasing the reactive oxygen species content (Hanson and Bentolila, 2004; Chen and Liu, 2014; Horn et al., 2014). Up to now, nine Rf (restorer of fertility) genes have been isolated in seven plant species, such as Rfo (Rfk1) in radish and Brassica (Chen and Liu, 2014). Most of the identified Rf genes encode PPR (pentatricopeptide repeat) proteins, but Rf genes are also highly multifarious and the respective restoration of fertility in CMS/Rf systems may be realized by various mechanisms at genomic, post-transcriptional, translational, or post-translational, and metabolic levels (Chen and Liu, 2014).

Stable CMS lines of crops could be produced by introducing the alien cytoplasm from the relatives through interspecific-/intergeneric hybridizations, or by mediating the mitochondrial rearrangement via somatic fusions (Prakash et al., 2009). For Brassica crops, since a sterility inducing cytoplasm identified in a wild population of Raphanus sativus (Ogura CMS) was introduced into B. napus and B. oleracea (Bannerot et al., 1974), a spectrum of alloplasmic CMS lines of diverse origins have been obtained by combining the cytoplasm of Brassica coenospecies with crop nuclei, particularly in B. juncea (see reviews by Delourme and Budar, 1999; Budar et al., 2004; Prakash et al., 2009; Yamagishi and Bhat, 2014). More recently, a novel CMS system in B. juncea incorporating the cytoplasm of B. fruticulosa was developed (Atri et al., 2016).

Somatic hybridization via protoplast fusion is a possible alternative for gene transfer from wild relatives to crops, by combining both nuclear and cytoplasmic genomes of two distantly related species, genera or even tribes. In somatic hybrids from protoplast fusion, the chloroplasts are usually inherited from one of the parents, whereas the mitochondrial DNA (mtDNA) is rearranged and may include DNA from both parents (Prakash et al., 1999). Intergenomic mitochondrial recombination with high frequency was found to occur in various somatic hybrids from the combinations of different species within Brassicaceae (Landgren and Glimelius, 1994; Dieterich et al., 2003; Leino et al., 2003; Oshima et al., 2010; Liu et al., 2015), leading to novel CMS genes. In the B. napus CMS Tournefortii-Stiewe produced by protoplast fusion with B. tournefortii (Dieterich et al., 2003), the mitochondrial rearrangement at upstream of the gene atp9 generated a chimeric orf193 that was co-transcribed with atp9. But another B. napus alloplasmic CMS Tournefortii by sexual hybridization with the same species had a chimeric orf263 at the vicinity of the atp6 (Landgren et al., 1996). Furthermore, recombinant mitochondrial genomes were also generated when protoplasts of fertile and sterile cytoplasm were fused (Sakai and Imamura, 1990; Wang et al., 1995; Meur et al., 2006).

From the mitochondrial genomes of nap, pol, ogu, and hau CMS sequenced (Handa, 2003; Chen et al., 2011; Tanaka et al., 2012; Heng et al., 2014), comparative analysis revealed that the CMS associated genes were localized close to the edge of syntenic sequence blocks, together with short repeats and overlapped repeats. These repeats were responsible for reorganization of mitochondrial genomes via homologous recombination. Overlapping homologous sequences from the fusion parents were produced when the parental mitochondria fused and the recombination occurred. In the cybrid derived from the somatic hybrid between Nicotiana tabacum and Hyoscyamus niger after being backcrossed with N. tabacum, the mitochondrial genome was highly recombinant with at least 35 intergenomic rearrangement events detected between two parents (Sanchez-Puerta et al., 2015). Recombination occurred via homologous mechanisms involving the double-strand break repair and/or break-induced replication pathways.

Currently, the autoplasmic pol CMS system from natural mutations in B. napus is still predominant for rapeseed hybrid production in China, but the alloplasmic Ogura CMS-fertility restoration system is widely used in Europe and Canada. Although several CMS systems have been reported, they are not used commercially due to undesirable CMS phenotype and lack of fertility restoration genes. In our previous study, the intertribal somatic hybrids between B. napus and Isatis indigotica Fort. (Chinese woad) of the Isatideae tribe within the Brassicaceae family were obtained and backcrossed successively to B. napus (Du et al., 2009; Kang et al., 2014), resulting in the development of one novel B. napus CMS line with carpelloid stamens (named as inap CMS). Herein, the histological and genetical characterizations of inap CMS were described, together with the mitochondrial rearrangements.

The novel inap CMS in B. napus selected from the intertribal somatic hybrids contained the recombinant mtDNA of both parents but biased to woad, and had the stamens converted into carpelloid structures. The sterile phenotype was stable during years and under different ecological environment. The developmental conversion of stamens occurred at the stage of stamen primordium differentiation. It was different from other CMS lines in phenotype, CMS associated genes, and restorer and maintainer relationships.

Homeotic conversion of stamens into carpelloid structures were described in several CMS lines of wheat (Ogihara et al., 1997), tobacco (Farbos et al., 2001), carrot (Linke et al., 2003), rapeseed (Leino et al., 2003), cauliflower (Kamiñski et al., 2012), and broccoli (Shu et al., 2015). Stamens of the CMS lines were replaced by carpelloid organs, thus resembled the B-class genes APETALA3 (AP3) and PISTILLATA (PI) mutants. In B. napus CMS line with the rearranged Arabidopsis mitochondria, downregulation of AP3 in the stamen primordia and subsequent repression of PI resulted in the conversion of stamens into carpelloid organ (Carlsson et al., 2007a). Similarly, the B-class genes were transcriptionally downregulated in carpelloid CMS flowers of tobacco, carrot and wheat (Farbos et al., 2001; Linke et al., 2003; Hama et al., 2004). The mitochondrial background had a distinct influence on nuclear gene expression (Carlsson et al., 2007b). This indicated that CMS-associated mitochondria genes regulated the expression of these nuclear MADS-box genes by mitochondrial retrograde signaling (MRS). Wheat Calmodulin-Binding Protein 1 (WCBP1) significantly upregulated in young spikes of the pistillody line (Yamamoto et al., 2013), indicating that CMS- associated genes played a role in development of pistil-like stamens by MRS involved Ca²⁺ signaling pathway in wheat CMS.

The most mitochondrial genes of inap CMS were found to originate from I. indigotica (Table 4), which was likely caused by the asymmetric fusion to produce the hybrids (Du et al., 2009). The protoplasts of B. napus were treated by 3 mM iodoacetate (IOA), and those of I. indigotica were irradiated with UV (Du et al., 2009). IOA could inactivate the glyceraldehyde-3-phosphate dehydrogenase, and resulted in the prevention of the glycolytic pathway (Zhao et al., 2008). Only when B. napus protoplasts for fusion contained the normal cytoplasm, the fused cells could divide, formed calli and regenerated plants. Thus, the mitochondria from I. indigotica were retained, and those from B. napus eliminated but several segments were introgressed into the mitochondria of the hybrid by recombination (Table 4). Such situation was also reported in other asymmetric somatic hybrids (Morgan and Maliga, 1987; Akagi et al., 1995). MtDNA sequencing of an Ogura-CMS cybrid derived from somatic fusions between B. napus and sterile radish showed that the mtDNA component mainly inherited from radish (Wang et al., 2012). The B. napus origin of the chloroplast DNA of inap CMS was in accord with the conclusion that the chloroplasts were generally from the iodoacetate-treated parent (Morgan and Maliga, 1987; Leino et al., 2003). This might be due to the better compatibility of recipient chloroplasts with the recipient nucleus, or the loss of the irradiated chloroplasts.

Two copies of cox2 gene in Brassica species (Handa, 2003; Yamagishi et al., 2014) arose from duplication, for it was involved in a large repeat sequence. Via this large repeat, the Brassica mitochondrial genome could be recombined into two subgenomic circles. I. indigotica contained only one copy of this gene. Of two copies of cox2 gene in inap CMS, cox2-1 was from I. indigotica, but novel cox2-2 was the recombinant between cox2 of woad and cox2-2 of rapeseed (Figure 4 and Supplementary Figure S1). Rearrangements in the cox2 region were also reported in somatic hybrids between B. napus and A. thaliana (Landgren and Glimelius, 1994; Leino et al., 2003), B. napus and R. sativus Kosena CMS (Sakai and Imamura, 1990). CMS-associated genes identified in many CMS systems (Wang et al., 1995; Chen et al., 2011; Tanaka et al., 2012; Heng et al., 2014) were located in flanking sequence of the known mitochondrial genes, and co-transcribed with it. Protoplast fusion offered an opportunity for mitochondrial genome recombination, leading to appearance of novel CMS genes. Three atp9 genes were present in the alloplasmic Tournefortii-Stiewe CMS of B. napus, which was generated by protoplast fusion with B. tournefortii (Dieterich et al., 2003). MtDNA rearrangements happened in the upstream of one of these genes and generated a chimeric gene orf193 that was co-transcribed with atp9. An alloplasmic B. juncea CMS line derived from somatic hybridization with Diplotaxis catholica had two copies of coxI gene, and the recombinant coxI-2 gene was suggested to cause CMS (Pathania et al., 2007). For the novel cox2-2 gene was expressed in the inap CMS flowers weakly and insignificantly from restorer plant, it was likely not responsible for male sterility.

The identification of the restorer and maintainer relationship is one of the most classical methods to distinguish different types of sterile cytoplasm (Wan et al., 2008). The test results showed that the fertility of inap CMS failed to be restored by restorer lines from Ogura, pol, nap and Tournefortii-Stiewe CMS systems. Budahn et al. (2008) found that rapeseed plants with the alien radish cytoplasm showed pistilloid stamens with ovules and stigmatoid tips, and the radish chromosome F could eliminate the disadvantageous effect of alien cytoplasm in B. napus. Additionally, the chromosome F did not restore Ogura CMS. In spite of the similar phenotype with inap CMS, the radish chromosome F was unable to restore its fertility, indicating that these two CMS lines had different CMS-associated genes and mechanisms of male sterility. No restorers were screened among 112 B. napus cultivars. This further testified that the fertility restoration gene(s) of alloplasmic CMS lines generally existed in nuclear genome of cytoplasm donor species. The restoration gene of a B. napus CMS line with rearranged A. thaliana mtDNA was located on chromosome III of A. thaliana (Leino et al., 2004). The restorer line for Nsa CMS of B. napus developed by the somatic hybridization with Sinapis arvensis was one disomic addition line with the restoration gene on the alien chromosome (Wei et al., 2010). Similarly, the restoration gene(s) for our inap CMS also existed on one specific woad chromosome, because only the monosomic alien addition line of B. napus with woad chromosome e among the complete set of such lines and also the same cytoplasm as inap CMS produced the normal flowers with fertile pollen grains (Kang et al., 2014). To breed the restorer line for inap CMS, the chromosomal segment carrying the fertility restoration gene(s) should be introgressed into one of rapeseed chromosome and the lines (2n = 38) with high restoration capability selected, as excellently exemplified by the long-term development of the radish introgression carrying the Rfo restorer gene for the Ogu-INRA CMS in B. napus (Delourme et al., 1991, 1998, 1999; Primard-Brisset et al., 2005). The identification of some progeny plants (2n = 38) with the restored flower development and pollen fertility from the addition line gave the expectation that the new CMS and fertility restoration system should be developed for rapeseed hybrid production in near future.

ZL and LK conceived and designed the experiments, analyzed the data and wrote the paper. LK performed the experiments. PL participated in mtDNA analysis. AW performed the histological section experiments. XG provided technical expertise for molecular analysis and scientific discussions.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.