Introduction

Mitochondrial genomes in angiosperms are mostly maternally inherited. Despite a large variation in size and structure, the gene content of plant mitochondrial genomes is limited (~ 60) and shows narrow variation (Gualberto et al. 2014; Chen et al. 2017). Further, the nucleotide sequences of essential genes are highly conserved (Gualberto et al. 2014). As each cell contains several copies of the mitochondrial genome, mitochondrial mutations are rarely found in nature and are difficult to obtain through induced mutagenesis. Mitochondrial genetics in plants is mostly studied in the context of cytoplasmic male sterility (CMS), the only agronomic trait encoded by the mitochondrial genome, and is widely employed in crop breeding (Kubo et al. 2013; Yamagishi and Bhat 2014; Chen et al. 2017). While uniparental inheritance of mitochondria precludes conventional genetic analysis, somatic hybridization where organellar genomes of both the parents are brought together in the hybrid cell has been found to offer unique opportunity to study organellar genetics including recombination (Pelletier et al. 1983). Previous studies have relied mostly on restriction fragment length polymorphism (RFLP) analysis to identify mitochondrial recombinations (Pelletier et al. 1983; Kirti et al. 1995a, b; Vasupalli et al. 2017). However, these studies examined only small regions and were unsuitable for fine analysis of recombination throughout the entire mitochondrial genome. The availability of high-throughput genome sequencing technologies has accelerated the sequencing of many plant mitochondrial genomes. But, until now, complete mitochondrial genomes have been reported only for three cybrids (Wang et al. 2012; Sanchez-Puerta et al. 2015; Garcia et al. 2019). However, the actual parents that contributed to the cybrid were not used for comparative analyses in these studies.

To correct the defects such as leaf chlorosis, petaloid anthers, poor female fertility of the radish ‘Ogura’ CMS line of B. juncea (OgRLM), protoplast fusions were made between OgRLM plants and normal B. juncea (RLM198), and the cybrids Og1 and Og2 lacking leaf chlorosis were selected (Kirti et al. 1995b). Og1 has sterile anthers, shows good female fertility but Og2 still carries petaloid anthers and poor female fertility, which was also reflected in the mitochondrial genome profiles of the cybrids; Og1 is closer to B. juncea whereas Og2 is closer to OgRLM (Meur et al. 2006). While OgRLM and Og2 are very stable CMS lines, Og1 shows a frequent reversion to male fertility, especially when maximum day temperature rises above 28 °C (Bhat unpublished). Such male fertile plants do not restore fertility to CMS lines of Og1, Og2 or OgRLM. Further, we did not observe any male sterile progeny from such revertant plants over several generations which suggests some permanent genetic change in the revertants.

Ogura CMS has been ascribed to the mitochondrial gene orf138, which is co-transcribed with the atp8 (orfB) gene (Bonhomme et al. 1992). While both Og1 and Og2 have recombined mitochondrial genome and also carry male-sterility-inducing orf138 sequence (Kirti et al. 1995b; Meur et al. 2006), the instability of CMS in Og1 is not yet understood. Breakdown of male sterility in CMS lines is undesirable from commercial seed production viewpoint but is found in several instances; maize CMS-S cytoplasm (Small et al. 1988), pearl millet A1 cytoplasm (Smith and Chowdhury 1991; Feng et al. 2009), sugarbeet (Dudareva et al. 1990) and rice (Nawa et al. 1987). In all those cases, loss of specific mitochondrial fragments was detected. In the present study, we determined the complete mitochondrial genome sequences of the cybrid Og1, its parents, and the male fertility revertant line (Og1-rt) with the twin objectives, namely, (i) to obtain complete details of mitochondrial recombinations that gave rise to Og1 cybrid, and (ii) to understand the exact changes in the mitochondrial genome that accompanies breakdown of male sterility.

Materials and methods

Plant materials

Male sterile lines of B. juncea carrying Ogura cytoplasm (OgRLM), chlorosis-corrected cybrids Og1 and Og2 (Kirti et al. 1995b), and euplasmic B. juncea cv. RLM198 and male fertile revertants of Og1 (Og1-rt) were used in this study. Plants were raised in the field during October–March season following standard agronomic practices.

Nucleic acids isolation

Total and mitochondrial DNA (mt-DNA) were isolated from unopened flower buds using the protocol described in Kirti et al. (1995a) and Pathania et al. (2003), respectively. Total RNA was isolated from unopened flower buds with TRIzol reagent (Invitrogen) as per manufacturer’s protocol.

Northern and Southern hybridization

Southern and northern hybridizations were carried out as per Kirti et al. (1995a) using 4 µg total DNA and 12 µg of total RNA, respectively. Nucleic acids were transferred to Amersham HybondTM–NX membrane using nuclease-free 20X SSC. Probes were prepared using either 32P-labeled dCTP (Kirti et al. 1995a, b) or non-radioactive DIG labeling kit (Roche). Probes were PCR-amplified using gene-specific primers (Supplementary Table 1). Hybridization was performed as per Kirti et al. (1995a, b) or Roche protocols. FUJI medical X-ray film was used to develop the blots.

PCR, long PCR, RT-PCR

PCR (Taq DNA polymerase, Sigma-Aldrich) and long PCR (AccuTaq™ LA DNA Polymerase, Sigma-Aldrich) were performed as per the manufacturer’s protocols. Prior to cDNA synthesis, 1 µg of total RNA was treated with RNase-free DNase1 (Thermo Scientific) to remove DNA contamination. Two-step RT-PCR was performed using SuperScript™ First-Strand Synthesis System of Invitrogen as per manufacturer’s instructions.

Illumina next generation sequencing and assembly of mitochondrial genomes

The mt-DNA isolated from Og1 and Og1-rt, OgRLM and B. juncea cv. RLM198 was used for NGS sequencing. The mt-DNA was sheared using Covaris S220 and fragments of 300–400 bp were selected. The selected fragments were used for library preparation for paired-end sequencing as per the protocol “TruSeq® DNA Sample Preparation Guide” and run on Illumina MiSeq DNA Sequencing Platform. CLC Genomics workbench V7.5.1 software was used for the analysis of reads. Low-quality reads were discarded after trimming the reads using default parameters. The quality trimmed reads were subjected to de novo assembly with default parameters and K-mer -40 to generate contigs. First, the mitochondrial genomes of parents OgRLM and RLM198 were assembled using the B. juncea (NC_016123.1) and Ogura (AB694744.1) mitochondrial genomes as reference sequences. Subsequently, Og1 contigs were BLAST searched against the parental mitochondrial genomes. Contigs that matched either of these two parental genomes were selected for further analysis. To link the contigs, outward orientated primers were designed from all selected contigs (Supplementary Table S1) and PCR was performed using all possible combinations of primers. The amplified products were directly sequenced. By aligning all the sequences the complete mitochondrial genome sequences were assembled.

Gene annotation

The coding sequences of the B. juncea (NC_016123.1) and Ogura (AB694744.1) mitochondrial genomes were used in Chlorobox GeSeq (Tillich et al. 2017) for Og1 and Og1-rt mitochondrial genome annotation. The mitochondrial genome pictorial representations were also generated using this software.

Identification of repeat regions

For the identification of repeat regions, assembled mitochondrial genome was matched against itself (BLASTN) on the NCBI web server using default parameters, and repeats of > 50 bp were considered.

The mitochondrial genome sequences of B. juncea cv. RLM198 (Accession No. MT675103), OgRLM (Accession No. MT675104) and Og1 (Accession No. MT675105 & MT675106) have been submitted to NCBI.

Results

Mitochondrial genome assembly

The number of raw reads obtained ranged from 4,343,816 in Og1 to 515,244 in OgRLM. After quality trimming, the reads were reduced to 4,337,989 and 507,738, respectively (Table 1). Average read length after trimming ranged from 223 to 238 nucleotides. The quality-trimmed reads of OgRLM and RLM198 lines were mapped to the mitochondrial genome sequences of Ogura (AB694744.1) and Brassica juncea (NC_016123.1), respectively. A large proportion of reads of RLM198 (83%) and OgRLM (73%) mapped to the respective mitochondrial genome in the NCBI database (Table 1). Barring one A to G change, eight single nucleotide (A or T) insertions and one dinucleotide (AT) insertion, RLM198 reads mapped perfectly with the NC_016123.1 sequence. Nine of these changes were in the intergenic region and one was in the intron of rpl12 gene. Similarly, except for 48 single nucleotide insertions and eight deletions (3 dinucleotide and the rest single nucleotide), OgRLM reads matched with the AB694744.1 sequence. In almost all cases, the indel nucleotide was identical to the adjacent repeating mononucleotide, suggesting that indels arose through replication slippage. Further, the indels were located in the intergenic or non-coding regions. Interestingly, orf138 in OgRLM showed one nucleotide deletion and one nucleotide insertion in close proximity to each other, which alteration of reading frame leading to two amino acids changes at the C-terminal end (Supplementary Fig. S1). Thus, the mitochondrial genomes of RLM198 and OgRLM were determined to be 219,776 bp and 258,462 bp, respectively.

Table 1 Details of mapped reads of mitochondrial DNA of the male sterile Brassica juncea cybrid Og1, parents and the fertile revertant (Og1-rt)
Full size table

The quality reads of Og1 and Og1-rt were subjected to de novo assembly on a CLC workbench using k-mer 20, 25, 30, 35, 40, 45, 50 and 60. As 40 k-mer gave longer contigs compared with others, assembly obtained with 40 k-mer was used for further analysis. The number of contigs after de novo assembly in Og1 and Og1-rt were 894 and 117, respectively. The assembly covered 1,096,220 and 453,850 bp in Og1 and Og1-rt, respectively. The longest contigs in Og1 and Og1-rt were 44,370 bp and 84,133 bp, respectively (Table 2). Although mt-DNA used for sequencing was extracted from isolated mitochondria, it is unlikely to be entirely free of plastid and nuclear DNA contaminations. Therefore, the genome assembly was BLAST searched and contigs showing high match with the nuclear or plastid genome sequences were removed. As Og1 carries a recombinant mitochondrial genome, the remaining contigs were subjected to BLASTN against OgRLM and RLM198 mitochondrial genome sequences (determined as above) to identify matching contigs. This returned a total of 14 and 12 contigs for Og1 and Og1-rt, respectively (Table 2). Size of the contigs ranged from 2418 to 44,370 bp and from 2306 to 84,133 bp in Og1 and Og1-rt, respectively (Supplementary Table S2). The total number of base pairs in BLASTN-selected contigs of Og1 and Og1-rt added up to 242,893 and 236,232, respectively. Although the proportion of reads of Og1 and Og1-rt mapped to these contigs was lower than that of RLM198 and OgRLM (Table 1), there were still sufficiently large number of reads to give good coverage of the genome. The contigs of Og1 and Og1-rt having identical sequences are denoted by the same number but are distinguished by the suffix –st and –rt, respectively (Supplementary Table 2). Minor differences in size between some of the contigs bearing the same number are due to truncated ends of the contigs. Contig 4-8-rt is a compound contig that resulted from joining of Contigs 4-st and 8-st. Contig10-st had no counterpart in the Og1-rt. These results indicated that the mitochondrial genome of Og1 and Og1-rt are largely identical except for a 6.0 kb Contig 10.

Table 2 General information of contigs selected for mitochondrial genome assembly of B. juncea CMS line Og1 and its fertile revertant (Og1-rt)
Full size table

To join the contigs and to fill the gaps, outward primers were designed from regions close to the ends of each contig and PCR was performed using all possible pairs of primers from different contigs. Assuming that the mitochondrial genome of the cybrid is circular and almost all mitochondrial genome sequences are represented in the contigs, each primer is expected to give amplicon with only one primer designed from another contig. However, the outward primers (reverse and forward primers in Supplementary Table S1 refer to outward primers corresponding to 5′ and 3′ ends of contigs, respectively) of Contig-6, -7, and -9 of Og1 and Og1-rt gave PCR amplicons with primers designed from two different contigs (Supplementary Table S3). Further, primer 8F gave amplification with two different primers 4R and 10F in Og1, while the reverse primer 8R gave amplification with only one primer (8R-5F). Similarly, primer 10R did not show amplification with any of the contigs suggesting some conflict/gap in the assembly. The amplification patterns in Og1-rt were identical to Og1 except for 8F (Supplementary Table S4). The amplified PCR products (Supplementary Tables S3 and S4) were got sequenced on the Sanger sequencing platform. The fragments of identical size obtained with the same primer combinations in Og1 and Og1-rt showed identical sequences.

Mitochondrial genomes are known to contain repeat sequences and exist as subgenomic circles. If primers are designed from repeats, multiple PCR amplicons are expected. To resolve this, additional PCR reactions were set up involving primers designed from contigs giving alternative amplicons. For example, primer 6R gave amplicons with primer 5R and 11R, while 6F gave amplicons with 1F and 12F. Hence, we tested 1F-5R, 1F-11R, 12F-5R and 12F-11R primer combinations. Likewise, 7R gave amplicons with 1R and 3F while 7F gave amplicons with 3R and 13F. So, PCR was carried out using primer pairs 1R-3R, 1R-13F, 3F-13F and 3F-3R. Since Contigs 7 and 9 are more than 4 kb, long PCR Taq polymerase was used. Results of PCR are summarized in Supplementary Table S5 and S6. An amplicon of 9.1 kb was detected with both 1R-3R and 3F-13F, while the other two combinations failed to amplify any PCR product (Supplementary Fig. S2A). This confirmed that Contig 7 is duplicated in between 1R-3R and 3F-13F. Similarly, we checked the Contig 9 adjacent sequences by PCR with primer combinations of 4F-11F, 4F-14F, 2R-11F and 2R-14F (Supplementary Fig. S2B). Amplicons of 4.4 and 4.5 kb obtained with 4F-11F and 2R-14F confirmed that Contig 9 lies between 4F-11F and 2R-14F. Likewise, Contig 6 was also confirmed by PCR. Reverse primer from Contig 10 of Og1 did not give any amplification with any combinations of primers from other contigs. But this primer is located downstream of the orf138 coding region. We know that orf138 and atp8 are present side by side in the Ogura mitochondrial genome. But these two genes appeared in different contigs in Og1 assembly. Therefore, PCR was done with inverse primer of atp8 and Contig 10 reverse primer, which gave an amplicon of 1169 bp. This confirmed Contig 10 lies between 8F and atp8, and further confirmed a new context for atp8 in Og1 (Supplementary Tables S3 and S4). Mitochondrial genome assembly was finalized by aligning the sequences obtained from Sanger sequencing and contig adjacent sequences.

After assembly, Og1 and Og1-rt mitochondrial genomes formed a circle of 250,999 bp. In addition, Og1 was inferred to carry an additional circular molecule of 96,185 bp containing male-sterility-associated orf138 sequence. We designate the mitochondrial genome of Og1 to include both Og1a and Og1b molecules while Og1a forms the mitochondrial genome of Og1-rt (see later).

Gene content and repeats

After annotation, total of 52 genes were found in Og1 cybrid (Fig. 1; Table 3). These included three rRNA genes, 15 tRNA genes, 8 genes coding for proteins of the mitochondrial ribosomes, and 25 genes involved in oxidative phosphorylation and other essential mitochondrial functions and CMS-associated orf138. Except orf138, all the other genes were present in Og1a molecule. Further, four protein coding genes, namely, atp4, cox1 nad4L and trnM were duplicated in Og1a and both the copies were derived from B. juncea. The Og1b molecule contained 17 genes, including orf138, three rRNA, two tRNA and two mitochondrial ribosomal protein coding genes. However, nad1, nad2 and nad5 were partial genes in Og1b. The male-sterility-inducing gene orf138 was derived from the OgRLM parent (Table 3). Three copies of trnfM gene were found in the cybrid; one copy from RLM198 parent was present on Og1a whereas Og1b had two copies, one each derived from RLM198 and OgRLM. Thus, the sterile and fertile lines differed only with respect to orf138, and loss of orf138 bearing Og1b molecules appeared to lead to the breakdown of male sterility in the Og1 line. Expression of atp6 differs between OgRLM and Og1 (Muer et al. 2006). Therefore, we compared atp6 gene of the cybrid with its parents, and found that Og1 has inherited atp6 from the B. juncea parent.

Fig. 1
figure 1

Mitochondrial genome of CMS Brassica juncea cybrid Og1. Genes are shown in outer or inner circle to indicate the coding strands. a The larger Og1a molecule, b The smaller Og1b molecule carrying CMS-causing orf138 gene

Full size image
Table 3 Mitochondrial genes in the Og1a and Og1b molecules and their parental origin
Full size table

We searched repeat regions (> 50 bp) within each molecule and also between the two molecules, Og1-a and Og1-b. Repeats within each mt-DNA molecule were detected by BLASTN using the same sequence in both query and subject fields. Repeats between the two mt-DNA molecules were detected by BLASTN of the two molecules. Og1a and Og1b molecules share a large stretch (63.634 kb) of identical sequences which can be regarded as a direct repeat. Likewise, two other fragments of 26.61 kb and 2.431 kb were also shared by these two molecules (Supplementary Table S7). In addition, 35 repeats (18 direct repeats and 17 inverted repeats) were detected in Og1a, whereas Og1b contained three direct and one inverse repeats. Three large repeats of > 2 kb in Og1a were duplicated genome copies arising from mitochondrial genome recombination. Two repeats in Og1a had three copies. Except for the large repeats stated above, other repeats were below 313 bp. Only four perfect repeats were found in Og1a; other repeats contained 1–17 bp mismatches (indel and SNP).

Comparison of mitochondrial genome structure and composition of Og1 cybrid with its parents

To know the exact derivation of the recombinant mitochondrial genome, the Og1 mitochondrial genome was compared with the mitochondrial genomes of its parents RLM198 and OgRLM. Og1a and Og1b molecules share two large identical fragments of 63.634 kb and 26.61 kb oriented in the same direction Hence, they can undergo recombination within this region to give a single master circular molecule. However, with the short read sequencing adopted here, such a recombinant molecule cannot be detected through genome assembly. Thus, total size of the mitochondrial genome of the cybrid Og1 was determined as 347,184 bp, which is 58% and 34% larger than its parents, RML198 (219,776 bp) and OgRLM (258,462 bp), respectively. To identify the fragments derived from the two parents, we separately aligned the sequences of Og1a and Og1b molecules to the parental mitochondrial genomes. The aligned sequences were examined for the presence/absence of parent-specific SNP/Indels and the fragments coming from the two parents were demarcated. Overlapping identical sequences between the two parental fragments were designated as homologus regions (Tables 4 and 5). This, however, does not mean that the immediate upstream and downstream sequences were non-homologus; in most cases at least ~ 300 bp stretch was very similar to permit pairing and recombination. Overall, 83.6% of the cybrid genome (excluding sequences shared by both RLM198 and OgRLM) was derived from the RLM198 parent.

Table 4 Details of fragments contributed by parents RLM198 and OgRLM to Og1a
Full size table
Table 5 Details of fragments contributed by parents RLM198 and OgRLM to Og1b
Full size table

A careful examination revealed 11 and 4 recombinations in Og1a and Og1b molecules. All recombinations except two occurred in homologous regions shared between RLM198 and OgRLM mitochondrial genomes (Fig. 2, Table 4). An intramolecular recombination within RLM198 mitochondrial genome involving a repeat region led to duplication of 8959 bp fragment leading to duplication of four protein coding genes in Og1a. Likewise, one of the four recombinations in Og1b involved a 163 bp repeat region. Thus, Og1 cybrid genome arose through both intra- and inter-molecular recombination between the parental genomes. As stated above, the longest, perfectly identical region between Og1a and Og1b molecules spanned 63,634 bp. Og1a molecule contains eight fragments derived from RLM198, of which two were in reverse orientation as compared to the parent. On the other hand, four of the six fragments inherited from OgRLM parent were in reverse orientation as compared to the parent. Og1b molecule comprises of five fragments, of which, three are derived from RLM198. Further, both the OgRLM fragments were in reverse orientation as compared to the parent. Thus, structurally, mitochondrial genome of the cybrid was largely comparable to the RLM198 parent (Supplementary Fig. S3). A close examination of alignment of Og1a and Og1b molecules against the parental genomes indicated two and one instances, respectively, of likely gene conversion (Table 4). We inferred gene conversion when the middle of three closely located SNPs was derived from one parent while the flanking SNPs were derived from the other parent.

Fig. 2
figure 2

Recombinant mitochondrial genome sequence of CMS Brassica juncea cybrid Og1. Fragments derived from the parents RLM198 and Ogura are depicted in white and grey boxes, respectively. a Og1a molecule, b Og1b molecule

Full size image

Confirmation of genome assembly results with RFLP analysis

To validate the results of genome sequence assembly, RFLP analysis was done for orf138 and atp8 genes. When total DNA was digested with EcoRI, HindIII or BamHI and hybridized with radiolabeled orf138 probe, a clear signal corresponding to size 2.9, 3.3 or 2.3 kb, respectively, was observed in Og1, Og2 and OgRLM but no signal was detected in Og1-rt and RLM198 (Fig. 3). However, compared with Og1, the signal was stronger in Og2 and OgRLM. In addition, some weak signals were also detected in Og2 and OgRLM. Enzyme-probe combination EcoRI-atp8 revealed a single 1.2 kb fragment in all the lines. Similarly, HindIII-atp8 gave a 3.3 kb band in all lines except euplasmic line RLM198 where a 1.3 kb band was recorded. BamHI released a 2.3 kb fragment hybridizing with the atp8 probe in OgRLM and Og2 lines. In Og1 line, a faint signal was detected corresponding to a 2.3 kb fragment. Further, a 1.3 kb atp8 fragment was visualized in Og1 and Og1-rt (Fig. 3). Euplasmic line RLM198 gave a much longer 13.7 kb atp8 fragment with BamHI restriction enzyme. The Southern patterns are in total agreement with that expected based on genome assembly, and reversion to male fertility in Og1-rt line appears to be associated with loss of orf138-containing Og1b molecules.

Fig. 3
figure 3

Mitochondrial RFLP analysis of Brassica juncea CMS lines. a Schematic diagram depicting the restriction sites and nucleotide positions of the mitochondrial genome sequences. Numbers in bold indicate the start and end nucleotides of gene. Locations of probes are indicated by soild bars. b Southern blots of B. juncea lines Og1, Og1 revertant (Og1-rt), Og2, OgRLM and RLM198, probed with orf138 (upper panel) and atp8 (lower panel). Fragment sizes (kb) are indicated on the right

Full size image

Additional RFLP analysis was done using non-radio labeled probes to confirm the presence of two different mt-DNA circles in the Og1 cybrid. As shown earlier (Fig. 3), atp8 (probe A, present in both Og1a and Og1b molecules) detected two BamHI fragments (2.3 and 1.3 kb) in Og1, whereas a single 1.3 kb fragment was visualized in Og1-rt (Fig. 4a). The signal intensity of the 2.3 kb fragment was much less than the 1.3 kb fragment indicating differences in the abundance of Og1a and Og1b molecules. Additional four probes were designed corresponding to different locations in the two molecules (Fig. 4). Probes B and C were specific to Og1b while probes D and E were specific to Og1a. Southern blots prepared using mt-DNA restricted with BamHI, EcoRV or EcoRI restriction enzyme were hybridized with B, C, D or E probes. With probe B, a single 2.3 kb BamHI fragment was found in Og1, but no signal was detected in Og1-rt. Similarly, probe C revealed a 4.9 kb EcoRV fragment, in Og1 while Og1-rt gave no signal (Fig. 4b, C). When probe D was hybridized to EcoRV-digested DNA, a single 6.4 kb fragment of comparable intensity was detected in both Og1 and Og1-rt (Fig. 4d). Similarly, probe E revealed a single 2.3 kb EcoRI fragment in both Og1 and Og1-rt lines (Fig. 4e). Thus, the Southern patterns strongly supported the presence of two different mitochondrial genome circles in Og1, whereas Og1b molecule was not detectable by Southern blotting in the revertant line, Og1-rt.

Fig. 4
figure 4

Southern blots (a–e) of Og1 and its male fertile revertant line (Og1-rt) hybridized with probes A, B, C, D and E derived from different regions of Og1a and Og1b molecules. Blot A is a cropped image from Fig. 3. Fragment sizes (kb) are shown on the right. Schematic diagram showing restriction sites, nucleotide positions and locations of probes (A-E) in Og1a and Og1b molecules is depicted on the top. Dotted lines indicate shared regions whereas bold lines represent unique regions of Og1a and Og1b molecules

Full size image

Expression analysis of orf138-atp8

As orf138 cotranscribed with atp8 is associated with male sterility (Bonhomme et al. 1992), we examined the expression of these transcripts in Og1 and Og1-rt lines. Northern blot revealed a 1.4 kb orf138 transcript in all CMS lines and in Og1-rt, but no signal was detected in RLM198 (Fig. 5). The intensity of 1.4 kb band, however, was very faint in Og1-rt. With atp8 probe, a single 1.4 kb transcript was detected in OgRLM and Og2, whereas Og1 and Og1-rt revealed two transcripts of 0.8 kb and 1.4 kb. Again, the intensity of 1.4 kb fragment was faint in Og1-rt line. The euplasmic line RLM198 gave two fragments of 0.8 kb and 1.0 kb. The occurrence of 0.8 kb transcript in Og1 and Og1-rt supported genome assembly results that Og1 cybrid has inherited atp8 copy from B. juncea. These results show that orf138-atp8 transcripts are detectable in og1 revertants although orf138 is not detectable in Southern blot. Therefore, we conclude that a drastic reduction of Og1b molecule is associated with reversion to male fertility in the Og1 line.

Fig. 5
figure 5

Northern blots of Brassica juncea lines RLM198, Og1, Og1-rt, Og2 and OgRLM probed with atp8 and orf138. Fragment sizes (kb) are indicated on the right

Full size image

Genome coverage and abundance of two molecules

The overall mitochondrial genome coverages of the parents RLM198 and OgRLM were 2064 × and 326 ×, respectively. The results of Southern analysis and sequence assembly taken together suggested that Og1 mitochondrial genome exists mainly as two molecules, and the abundance of these two molecules is different. To find the relative abundance of the two molecules, we determined coverage of orf138 (present only in Og1b) and atp8 (present in both Og1a and Og1b) sequences. Coverage of orf138 was 357 × whereas that of atp8 was 1215 × giving Og1b:Og1a ratio of 1:2.4. The abundance of two molecules is also expected to reflect in the coverage of other regions including contigs. Further, duplicated sequences are expected to show higher coverage than unique sequences. Contig-wise coverage values were determined and the corrected values were derived after taking into consideration the shared contigs within or between molecules. We used coverage of Contig 10-st to adjust coverage of contigs shared by Og1a and Og1b. For example, Contig 6-st is present in Og1b and is present as two copies in Og1a. Therefore, adjusted coverage of Contig 6-st was calculated as (3016–459)/2 = 1278. The corrected coverage values showed 1.5- to 2-fold variation among contigs in both Og1 and Og1-rt (Supplementary Table S2). On the whole, coverage values of Og1 contigs were 3–4 times that of Og1-rt contigs, which is expected as more sequence reads were available in Og1 than Og1-rt. The mean coverage fell into two ranges for contigs of Og1 (1019–1069 and 1278–1595) whereas such a distinction was not evident for contigs of Og1-rt. These coverage values, however, cannot be taken at face value because contigs that contain repeats shared by other contigs would show higher coverage when coverages are calculated separately for each contig. We did de novo assembly of OgRLM and calculated contig-wise coverage of the 13 contigs totalling 247.7 kb (data not shown). Coverage of contigs showed variation from 244 × to 384 × indicating that coverage of different fragments varies. Thus, it is clear that Og1a is the abundant molecule, but it is not possible to conclude whether Og1b exists as a separate molecule or as a compound molecule of Og1a and Og1b. Further, we tested whether orf138 is altogether lost in og1-rt or diluted to a level where it would go undetected. We checked the reads of Og1-rt for orf138 and found that it has 29 × coverage as against the 188 × coverage of atp8 in Og1-rt. Thus, based on coverage, the relative abundance of Og1b and Og1a molecules in Og1-rt works out to be 1:5.5. Hence, non-detection of Og1b molecule in Og1-rt is mainly attributable to relatively low coverage of Og1-rt as compared to Og1 in our experiment.

Discussion

Somatic cell hybridization provides a unique opportunity to study recombination between mitochondrial genomes. Previous reports documenting high-frequency mitochondrial recombination in such hybrids/cybrids (Pelletier et al. 1983; Kirti et al. 1995a, b; Vasupalli et al. 2017) relied on RFLP analysis which precluded fine analysis of mitochondrial recombination. Detailed analysis of mitochondrial recombination based on whole mitochondrial genome sequence data of cybrids are limited (Wang et al. 2012; Sanchez-Puerta et al. 2015; Garcia et al. 2019) and did not use the original parents for comparison. Further, two cybrids came from somatic hybridization between very diverse species Nicotiana tabacum + Hyoscyamus niger. Thus, there is inadequate data for generalisation of nature and frequency of mitochondrial recombination in cybrids. This is the first study of mitochondrial recombination using cybrid and its parents.

Mitochondrial genome sequences of the parents of Og1 showed a near-perfect match with sequences of their counterparts (i.e. B. juncea and Raphanus sativus ‘Ogura’) deposited in the NCBI database. B. juncea cv. RLM198 showed only ten Indels/mismatches (all in the non-coding regions) with the B. juncea sequence in the database. In contrast, Ogura displayed 51 Indels including those that led to two amino acid substitution in the male-sterility-associated ORF138 protein. This is a novel allele of orf138 but retains the male-sterility-inducing capability. Thus, the sequence deviations observed were minor (indels of 1–2 nucleotide) arising mostly due to differences in the number of the mononucleotide repeats. The high depth of sequencing of these parents (326 × to 2064 ×) suggests that these differences are not due to sequencing or assembly error. Wang et al. (2012) reported 40 point mutations, mostly transitions, in as many as 22 genes in an Ogura-based cybrid of B. napus. However, they did not use the actual parents for comparison with the cybrid. The Og1 cybrid sequence, on the other hand, was identical to its parental fragments which indicated that cybridization has not created additional sequence variation.

Og1 mitochondrial genome assembly yielded two circular molecules, Og1a and Og1b sharing two large fragments of 63.6 kb and 26.6 kb which could serve as direct repeats to facilitate recombination to yield one master circle. However, with the strategy adopted here, such a molecule cannot be detected. The larger Og1a molecule contained all the essential mitochondrial genes whereas the Og1b molecule had the CMS-causing orf138 gene. Thus, Og1b was dispensable for the survival of the cybrid. The total size of the Og1 mitochondrial genome was larger than either of its parents. Similar findings in previous studies (Wang et al. 2012; Sanchez-Puerta et al. 2015; Garcia et al. 2019) suggest a general trend towards genome enlargement in cybrids. In particular, duplication of protein coding genes was common to all the cybrids sequenced so far. In contrast, no gene was duplicated in RLM198 and only one gene trnFM was duplicated in OgRLM. Although mitochondrial genomes of angiosperms mostly map as a single circular molecule, multipartite mitochondrial genomes with circular and linear molecules have also been described (Alverson et al. 2011; Sloan 2013). In literature, the phrase ‘multipartite mitochondrial genome’ is used rather loosely to include subgenomic circles arising from recombination between repeats. Although Og1 mitochondrial genome can theoretically form a single master circle, our results indicate that the cybrid genome exists predominantly as two circular molecules (Og1a and Og1b or Og1a and Og1a + Og1b) occuring at different frequency. The Og1 mitochondrial genome is comparable to the tripartite mitochondrial genome of cucumber where the largest molecule carries all the essential genes while the other two molecules contain repeat sequences and duplicated genes (Alverson et al. 2011).

The cybrid has inherited all essential mitochondrial genes from B. juncea while Ogura contributed only orf138 and trnfM genes. These results confirm our previous report based on RFLP analysis (Muer et al. 2006) that Og1 mitochondrial genome is highly similar to its B. juncea parent. Structurally, Og1a molecule closely resembles the RLM198 parent but with two inversions. However, there were at least 11 recombinations in the Og1a molecule whereas four recombinations were detected in Og1b molecule. The majority of recombinations occurred in homologous regions shared by the parents. In Nicotiana-Hyoscyamus cybrids as many as 30 recombinations were recorded (Sanchez-Puerta et al. 2015; Garcia et al. 2019), which were also confined to homologous regions. RLM198 and OgRLM mitochondrial genomes share several long stretches of sequence and a double crossover in such regions would go detected. Therefore, the number of recombinations recorded in our study could be an underestimate. It is generally held that repeats present in the mitochondrial genomes of plants facilitate recombination and contribute to structural diversity (Chang et al. 2011). However, except two instances, the recombination points identified in our study were not containing repeats. Thus, our results corroborate previous studies that recombinations in cybrids occur predominantly in homologous regions. The two male sterile tobacco cybrids derived from Nicotiana + Hyoscyamus mitochondrial genome recombination had predominantly tobacco mitochondrial genome (Sanchez-Puerta et al. 2015 and Garcia et al. 2019) suggesting influence of nuclear genes on the overall mitochondrial genome of the cybrids. In our study, the two cytoplasms were in isonuclear background. Yet, Og1 and Og2 cybrids derived from the same experiment had very contrasting mitochondrial genome constitution (Muer et al. 2006). In fact, Og2 had much poor floral phenotype than the parents indicating that B. juncea nuclear genome did not favour preferential retention of recombinant molecules bearing mainly B. juncea mitochondrial fragments.

We noticed three instances of likely gene conversion events in the Og1 cybrid (Tables 4 and 5) as against five reported in Hyoscyamus + Nicotiana cybrid (Sanchez-Puerta et al. 2015). Gene conversions are relatively easy to determine for nuclear genes by comparing SNP differences between parents and progenies, as recombination occurs between homologous chromosomes only once at meiosis. However, mitochondrial genome occurs in multiple copies within a cell and is capable of undergoing multiple cycles of recombination during the life of a cell. The extent of mitochondrial recombination and sorting of the recombinant molecules in early cell cycles following somatic hybridization/cybridization is still not well understood. Hence, gene conversions in mitochondrial genomes need to be viewed with caution.

Fertility reversion is not uncommon in CMS lines and factors such as nuclear background, temperature, water stress, tissue culture etc. have been shown to influence this reversion (Earle et al. 1987; Newton 1988; Elkonin et al. 2005). The reversion is often accompanied with either the loss of expression of CMS-associated gene or loss/rearrangement of CMS-inducing mitochondrial genome fragment (Fauron et al. 1990; Bonhomme et al. 1991; Gourret et al. 1992; Zabala et al. 1997). In the present investigation also, in Og1-rt plants, orf138 genomic copies were reduced to below the detection level of Southern blots. Nevertheless, orf138 transcripts were detected in northern blots. Also, examination of reads of Og1-rt line revealed a much lower (29 ×) coverage of orf138 region as comapred to the rest of the region (270 ×). The contrast between northern and sequencing results with respect to orf138 could be due to the source plants; DNA/RNA for Southern and northern blotting was isolated from revertant plants of the first generation, whereas for mitochondrial genome sequencing, fourth-generation selfed progenies of revertants were used. The relative abundance of Og1a and Og1b molecules in Og1 and Og1-rt calculated based on coverage of different regions clearly showed that orf138-bearing molecules were greatly reduced in these later generation revertants. Fertility reversion due to loss of orf138 has also been reported in a B. napus cybrid containing recombinant mitochondria of Ogura and B. napus (Bonhomme et al. 1991; Gourret et al. 1992). It has been found that mitochondrial genome exists as subgenomic circles and the relative abundance of different molecules is variable between tissues and at different times (Bellaoui et al. 1998; Janska et al. 1998; Mackenzie et al. 1988). In B. napus, Chen et al. (2011) reported the co-occurrence of both pol and nap mitotypes, and stoichiometry of the two mitotypes varied among genotypes. In common bean, fertility restorer gene Fr2 was found to lower the male-sterility-inducing pvs molecules to less than one copy per mitochondria (Janska et al. 1998; Mackenzie et al. 1988). Nevertheless, the pvs-containing molecules were not completely lost in the fertility-restored plants. Recently, Shen et al. (2019) showed that in mesophyll cells of cucumber not all mitochondria contain full copy of mt-DNA. Thus, fertility reversion of Og1 resembles the common bean example cited above except that the stoichiometric shift in this case, occurs due to external factors. Stoichiometric differences in mitochondrial subgenomic circles are well reported (Feng et al. 2009). In cucumber, the smaller circles (45 and 84 kb) were 1.5–2.0 times less frequent than the larger circle (1556 kb) (Alverson et al. 2011).

Og1 bears sterile anthers while OgRLM shows petaloid anthers. We earlier reported that atp6 transcript of Og1 is similar to RLM but differs from OgRLM (Muer et al. 2006). Mitochondrial genome sequence of Og1 shows that it inherited atp6 from B. juncea, which explains our previous results. Tanaka et al. (2012) reported that atp6 gene of Ogura radish differs from normal radish at the 5′ region. The deduced ATP6 polypeptides of Ogura and B. juncea differ by 14 amino acids (247 versus 261 amino acids). However, except for the nine N-terminal amino acids, B. juncea ATP6 is identical to the Ogura counterpart. Perhaps, Ogura atp6 transcripts are not efficiently translated in B. juncea nuclear background leading to petaloidy in Og2 and Og-RLM. Og1 contains B. juncea-derived atp8 in two contexts in Og1a and Og1b molecules giving a normal atp8 transcript besides orf138-atp8 bicistronic transcript. The male sterility of Og1 plants, therefore, indicates that orf138 directly causes male sterility rather than through interference in oxidative phosphorylation.

In conclusion, our study showed that Og1 cybrid genome resulted from inter- and intra-molecular recombination between the parental molecules and exists mainly as two circular molecules of different stoichiometry. Further, all essential mitochondrial genes were present on the larger molecule while the smaller molecule carried the male-sterility-inducing orf138. Our findings showed a clear mechanistic relation between mitochondrial stoichiometry and male sterility/fertility.