Mosaic Origins of a Complex Chimeric Mitochondrial Gene in Silene vulgaris

Chimeric genes are significant sources of evolutionary innovation that are normally created when portions of two or more protein coding regions fuse to form a new open reading frame. In plant mitochondria astonishingly high numbers of different novel chimeric genes have been reported, where they are generated through processes of rearrangement and recombination. Nonetheless, because most studies do not find or report nucleotide variation within the same chimeric gene, evolution after the origination of these chimeric genes remains unstudied. Here we identify two alleles of a complex chimera in Silene vulgaris that are divergent in nucleotide sequence, genomic position relative to other mitochondrial genes, and expression patterns. Structural patterns suggest a history partially influenced by gene conversion between the chimeric gene and functional copies of subunit 1 of the mitochondrial ATP synthase gene (atp1). We identified small repeat structures within the chimeras that are likely recombination sites allowing generation of the chimera. These results establish the potential for chimeric gene divergence in different plant mitochondrial lineages within the same species. This result contrasts with the absence of diversity within mitochondrial chimeras found in crop species.


Introduction
Chimeric genes are formed through processes of recombination and duplication that result in the fusion of segments derived from DNA fragments in different genomic positions.Because processes generating chimeric genes can result in novel protein structures, they are significant sources of evolutionary innovation in humans, insects, and plants [1,2,3,4].Formation of adaptive chimeric genes in plant mitochondria has been well known for over 2 decades [5] and is related to the unique tempo and mode of evolution of the plant mitochondrial genome.Unlike animal mitochondria, in most flowering plant genera mitochondrial coding regions evolve very slowly relative to those in chloroplast and nuclear genomes [6,7,8,9], whereas rearrangements and intragenomic recombination can be quite dynamic, even within species [10,11,12,13].Patterns of rearrangement and duplication within plant mitochondria are complex, but can be conceptually categorized into common and rare events [10,11,14].Common rearrangements are associated with large repeats in the mitochondrial genome (.4 kb in Arabidopsis [15]).These generate mini-molecules, which are in constant flux and can be assembled into mitochondrial master molecules according to the Multipartite model [8,16,17].Rearrangements associated with shorter repeats (,500 bp) are much less common, perhaps because nuclear surveillance loci such as MSH1 inhibit development of the new conformation of DNA [18,19,20].These rare recombination processes may occasionally produce novel chimeric genes that are expressed and have phenotypic effects in mature plants.The rate at which rare rearrangements arise is currently unknown, and it is unlikely that all chimeric genes are expressed; therefore the rate at which expressed chimeric recombinants arise is likely even lower.
Because the creation of each chimeric gene involves unique situations, each gene is assumed to have originated only once.The fate of these genes, however, is mysterious because few studies report allelic variation within chimeric genes (but see [21]).Chimeric mitochondrial genes composed of the same genic segments are not shared across plant species, suggesting that they do not persist long in the mitochondrial genome.Also, multiple different mitochondrial chimeric genes have been identified in several crop species (e.g.maize [1]).These patterns support a process whereby some mitochondrial chimeric genes originate and go extinct at a rate faster than speciation occurs.
Species with gynodioecious breeding systems that carry mitochondrial chimeric genes controlling male-sterility exhibit higher diversity at mitochondrial housekeeping genes (chimeric genes were not studied) than congeners with hermaphroditic breeding systems [22].This pattern is consistent with longer-term maintenance of the same mitochondrial types within gynodioecious species, perhaps due to balancing selection maintaining cytoplasmic male sterility (CMS).Insight into the processes generating chimeric genes has been gained through studies of CMS genes, which act to block the production of viable pollen and render female an otherwise hermaphroditic plant.CMS genes are usually composed of segments of mitochondrial housekeeping genes such as ATP synthase or cytochrome oxidase subunits that are spliced together with a region of unknown origin (reviewed in [1]).Transcriptional start positions and regulatory regions are recruited from the housekeeping genes [23,24], thus the regulatory region does not have to evolve de novo.Chimeric CMS genes from different species, and even within the same species, do not share structural homology, reflecting their independent generation [1].The presence of the region of unknown origin underscores the unique origin of each CMS gene because these unknown regions cannot be found in mitochondria that lack the particular CMS gene.Therefore, it is assumed that the proper circumstances for the origination of a new CMS must be fleeting.
The genus Silene is the subject of numerous studies of mitochondrial genomics and the population genetics and evolution of mitochondrial genes in natural populations [25,26,27,28,29].
Here we document a chimeric mitochondrial open reading frame that was first identified in Silene vulgaris when it was PCR-amplified along with ATP synthase 1 (atp1) when using primers designed to amplify atp1 (Fig. 1).The initial fragment, the transcribed RNA, and the gene have been named ''bobt.''Bobt is not present in all individuals.An analysis of the within and among population patterns of S. vulgaris plants that carried the bobt PCR fragment in a small region of western Virginia can be found in [28] and [30], wherein bobt co-segregates with haplotype b.
Here we analyze differences between two allelic versions of the mitochondrial chimeric gene bobt with highly divergent nucleotide sequences and expression variation, a unique phenomenon that has not yet been reported.We also show that one version of this chimeric gene is co-transcribed with cytochrome b (cob), whereas the other version is not adjacent to another coding region.Finally, qRT-PCR experiments indicated that these chimeric genes exhibit significantly higher transcription levels in female plants than in hermaphrodites, which supports the candidacy of at least one variant as a CMS gene.These discoveries indicate that mitochondrial chimeric genes have residence times in natural populations that are sufficiently long to allow nucleotide and expression divergence among alleles.

Plant material and PCR screening
A small sample of Silene vulgaris plants grown from seed were screened for the presence of bobt using a PCR assay described below.These seeds were collected from natural populations in Virginia, Austria, Germany, and the Czech Republic; collection sites were the same as in [28] and [31].Additional seeds were sent to us by Dr. David McCauley from 2 populations in New York, two in Vermont, and one population in Broadway, Virginia; locations of these sites can be found in [32].Finally, seeds were sent to us by Dr. Natalia I. Tiupitzina from a site near Krasnoyarsk, Russia.All collections were made along roadsides in public right of ways or public areas that were not subject to permitting.The field studies also did not involve endangered or protected species.Plants were propagated from seed in the greenhouse in pots filled with perlite, vermiculite and coconut coir (1:1:1) and fertilized 2-3 times per week depending on the season.To ensure controlled growth conditions during expression studies for plants from Mountain View Virginia (MV) and Krasnoyarsk, Russia (KR), we transferred the plants to a Percival growth chamber and cultivated them in a 19 h light/5 h darkness photoperiodic regime (23uC, 80% humidity, light intensity 80 mM/m 2 /sec).
Genomic DNA was isolated from fresh leaf material using either Qiagen Plant Mini kits or according to [33].The presence or absence of bobt in Silene vulgaris individuals from natural populations can be detected as a 2-band pattern when PCRamplified with atp1 primers 1 and 2 (Table S1); one band is atp1 and the second band is bobt.All individuals were screened for the presence of bobt using this PCR assay.

Southern hybridization
For Southern analysis, 1 mg of genomic DNA was digested with restriction enzyme EcoRI, electrophoresed overnight on a 0.7% agarose gel, and transferred to a positively charged membrane (Hybond N+, Amersham) by capillary blotting.A 1.5 kb fragment of the cox1 gene, a 1.4 kb fragment of the atp1 gene and a 220 bp fragment of the bobt gene were amplified and labeled with digoxigenin (DIG) using PCR labeling kit (Roche) according to the manufacturer.The PCR bobt probe included a portion of the ORF of unknown origin, which was not similar to any sequence available in GenBank.The primers used to generate the probes are provided in Table S1.The membranes were hybridized with DIG-labeled probes and visualized as described by [31].
Figure 1.Structure and genomic context of the two variants of bobt, bobt_KR (A) and bobt_MV (B).bobt_KR and bobt_MV differ in 43 nucleotide sites, but share the same general atp1-cox2-unknown ORF structure between the start and stop sites.They share a segment of atp1 at the 39 end downstream of the stop codon, a segment of cox2, and a segment of unknown origin.The fourth segment, located just downstream of stop codon corresponds to atp1.Its size differs between both copies, because bobt_KR is adjacent to cob, whereas bobt_MV is not.Horizontal arrows designate the locations of the primers (atp1 lo, atp1 up) originally used to discover bobt via PCR co-amplification with the functional copy of atp1.A full-length atp1 gene is shown below the bobt variants, the homologous regions are indicated by dotted lines.A scale in bp is given above.doi:10.1371/journal.pone.0030401.g001

Gene expression estimation using qRT PCR
Total RNA extracted by means of RNeasy Plant Mini Kit (Qiagen) was treated with DNase I (Ambion, Europe).One mg of DNAse I treated total RNA and 3.22 mg of random hexamers (Roche Applied Science, Germany) were heated 10 minutes at 65uC, chilled on ice and mixed with RT buffer, 0.5 ml of Protector RNase Inhibitor (Roche Applied Science, Germany), 1.2 ml of 10 mM dNTPs and 40 units of Transcriptor Reverse Transcriptase (Roche Applied Science, Germany), final volume 25 ml.The first strand cDNA was synthesized at 50uC for 30 min.RNA samples were reverse transcribed in two independent RT reactions and each cDNA was measured twice.All repeated measurements were consistent.
First strand cDNA was diluted 106 and qPCR was performed using the Light Cycler 480 SYBR Green I Master (Roche Applied Science, Germany) in a final volume of 10 ml with 300 nM of each of the HPLC purified primers (Table S1) supplied by Metabion.The LightCycler LC 480 (Roche Applied Science, Germany) was programmed as follows: 10 min of initial denaturation at 95uC, then 40 cycles for 10 s at 95uC, 10 s at 58uC, and 15 s at 72uC.PCR efficiencies were estimated from calibration curves generated from serial dilution of cDNAs.The relative ratio of the target gene was calculated as follows:

E CpT
T Where E T /E R represents the efficiency of target/reference amplification and CpT/CpR represents cycle number at target/ reference detection threshold (crossing point).A negative control consisting of PCR mixture with RNA instead of cDNA was performed for each RNA sample to confirm the efficiency of the DNAseI treatment.Transcript levels were normalized with mitochondiral (mt) 18S rRNA of S. vulgaris.The invariant levels of this reference transcript were confirmed by direct quantification of cDNA, as described by [34].A t-test implemented in Excel was used to estimate significance of transcript level differences between hermaphrodites and females.
Finally, quantitative PCR was adopted to compare the abundance of genomic regions under the same conditions as qRT PCR with total DNA used as a template instead of cDNA.

Inverse PCR
Seven ml of genomic DNA solution (70 ng/ml) was digested with EcoRI or HindIII in 50 ml final volume.The digestion mixture was heat inactivated, 106 diluted and ligated with 8 units of T4 ligase (MBI Fermentas) in 100 ml final volume at 5uC overnight.These conditions supported self-ligation of EcoRI fragments [35].A self-ligated fragment was digested by HindIII or BglII to provide a linear template for the subsequent PCR amplification.PCR was performed in Biometra T Gradient thermocycler with 5 ml of ligation mixture, 0.16 mM each of primers, 2.5 mM MgCl 2 , 320 mM dNTP and 1 unit Taq DNA polymerase (Promega) per reaction.PCR conditions were: 2 min at 94uC; 36 cycles for 1 min at 93uC, 1 min at 58uC and 2 or 3 min at 72uC, with a final extension of 5 min at 72uC.The resulting fragments were cloned in pGEM-T Easy vector (Promega) and sequenced.

Sequencing of bobt and atp1
The sequencing strategy is depicted in Figure S7.The central portion of bobt was amplified and sequenced using primers 1 and 2 (Table S1).The 59 and 39 portions of bobt_MV were amplified and sequenced from the product of inverse PCR of HindIII or EcoRI fragment with the primers 8 and 9.The sequence was confirmed by resequencing using primers targeted to regions specific for bobt (18/22 and 21/2).The 59 and 39 portions of bobt_KR were sequenced in the same manner, but using different primers: 6 and 7. Sequence of the cob gene adjacent to bobt_KR was obtained on the basis of PCR fragment generated by primers 3 and 5 using additional internal primers.The sequence of an entire bobt_KR-cob region was confirmed by using specific primers 10/2, 3/26 and 4/5 (Table S1) The atp1 variants (except their short 39 extremities) were amplified using specific primers (Table S1).The atp1.1_KR variant was amplified using the primers 16 and 17, atp1.2_KR by the primers 10 and 11 atp1.1_MVby the primers 18 and 17, atp1.2_MV by the primers 18 and 20.Internal primers 12 and 13 were applied to fill the gaps.To sequence 39 ends of atp1 variants the primers 14 and 15 were employed.To provide resolution within the Silene vulgaris clade, atp1.2 was sequenced in two individuals that originated form a population near Kovary, Czech Republic.Atp1.1 was not found in these individuals from Kovary.
The sequences from this study were deposited in GenBank under accession numbers HQ437988-437995 and JF343540-343541.

Phylogenetic analyses
Phylogenetic analyses were conducted in GARLI v0.96 [36], which simultaneously searches for the topology and tree lengths that maximize the likelihood of the data.The nucleotide substitution model was determined using Akaike Information Criteria using a correction for small sample size (AICc) in JMODELTEST v 0.1.1 [37,38].Optimal mutation models used for maximum likelihood analyses differed among data sets.For both the complete atp1 dataset and the bobt-atp1 homologous region data set, TPMuf+G [39] had the lowest AIC, so this model was used for all analyses.Six replicate runs were conducted for each dataset with random starting trees.Each replicate was run for 20,000 iterations.To determine clade support, 1000 ML bootstrap replicates were produced in GARLI, each was run for a length of 3,000 iterations.To check the results maximum likelihood analyses also were conducted in PAUP* 4.0b [40].Results from GARLI and Paup* analyses were identical.
Partition homogeneity tests were implemented in PAUP* to compare the phylogenetic signal from two regions of bobt with homology to atp1.The null hypothesis for a single origin of the bobt_MV and bobt_KR genes was tested using Kishino-Hasegawa [41] and Shimodaira-Hasegawa [42] tests implemented in PAUP* 4.0b [40].

Complete sequence of bobt in two genomes
Within a small haphazard sample of plants originating from the eastern US, Europe, and eastern Russia screened using a PCR based assay, we found the chimeric bobt gene only in plants from the region around Mountain Lake, Virginia and in plants from Krasnoyarsk, Russia.Further investigation of the Virginian and Russian copies of bobt using PCR-amplification with atp1-specific primers, inverse PCR, and direct sequencing revealed two divergent versions (alleles) of the bobt gene from the two regions: Mountain View, Virginia (bobt_MV) and Krasnoyarsk, Russia (bobt_KR).Although they shared the same overall structure and are composed of homologous genic segments, the two bobt alleles differed considerably in nucleotide sequence and genomic context.The structure of the bobt coding region was composed of three segments: a 59 192 bp segment with homology to atp1, an 81 bp segment with homology to cox2, and a region of 443 bp with unknown homology (Fig. 1).An 8 bp nucleotide region at the junction of the atp1and cox2-derived portions of bobt shared homology with both genes and is a likely region for intragenomic recombination that generated the chimera.The nucleotide sequences between the start and stop codons of bobt_MV and bobt_KR differed in 37 positions (8.4% of the sites; Fig. S1); 20.5 of these were at synonymous sites (Ks = 0.13) and 16.5 were at nonsynonymous sites (Ka = 0.03; Fig. S1).The regulatory/ promoter region upstream from the start codon was composed primarily of mono-and di-nucleotide repeats, with bobt_MV and bobt_KR sharing a region that included a 16 bp motif similar to the conserved region in angiosperm atp1 59UTR (positions 2181 to 2166 Fig. S1, [43]).The primary difference between the regions 59 to bobt_MV and bobt_KR was the presence of an 8 bp sequence (ATTTTAAT) in bobt_MV, which was shared with the MV copy of atp1.2 (positions 29 to 224 in the alignment).Following the stop codon was a 39 segment with homology to atp1; in the KR version this region was 85 bp, whereas in the MV version this region was 163 bp (Fig. 1).The KR version of bobt was located directly 59 to the cob gene, whereas cob was not in this position for the MV version (Fig. 1).

Inheritance of bobt
We conducted two series of reciprocal crosses to assess the inheritance of bobt.One set included crosses between individuals carrying bobt_MV and individuals without bobt, whereas the other set included crosses between individuals carrying bobt_MV and bobt_KR (Table 1).Presence or absence of each bobt allele was assessed using a PCR assay and allele specific primers.In all cases, bobt alleles were maternally inherited, reflecting their mitochondrial origin.The conclusive evidence for the mt origin of bobt has been obtained recently by 454 sequencing of purified mtDNA in both MV and KR.The bobt-related reads were as abundant as the reads derived from the mt genes, whereas all chloroplast-specific reads were very rare (Muller and Storchova, in preparation).

Copy numbers of mitochondrial genes
Southern hybridization with a probe derived from the coding region unique to bobt MV revealed one copy of bobt in both the MV and KR mitochondrial (mt) genomes.Stronger hybridization to the MV than the KR genome was caused by higher affinity of the bobt_MV probe to MV genome than to KR genome.In S. vulgaris DNAs for which we were unable to amplify bobt with bobt-specific primers, we also found no bands after Southern hybridization (Fig. 2A).When the same membrane was hybridized with an atp1 specific probe, at least two copies of atp1 were detected in the KR genome and 3 copies were visible in the MV genome (Fig. 2B); because the conditions were high stringency, none of the atp1 copies corresponded to the bobt gene.As a control, Southern blot studies revealed a single copy of cytochrome oxidase 1 (cox1) in both the KR and MV mt genomes (Fig. S2).The intensities of atp1and bobt-specific bands were similar, indicating similar copy numbers for both genes as expected if bobt was localized in the mt genome.
A single copy was also confirmed for the cox2 gene in the KR and MV genomes (data not shown).cDNA sequences revealed two different atp1 variants (atp1.1 and atp1.2) present in both the MV and KR individuals.We tested whether atp1 was duplicated in two individuals that did not carry bobt, Kov4 & Kov5; only atp1.2, and not atp1.1,was present in Kov4 & Kov5.Although this is not an exhaustive sample, it does show polymorphism for the duplication.Atp1.1 and atp1.2 exhibited considerable sequence differences (Fig. S3).A 6 bp insertion/deletion at 39end of the coding region in both MV and KR backgrounds differentiated atp1.1 and atp1.2 and was used to develop version-specific primers for sequencing.Atp1.1 and atp1.2 within the MV mitochondria differed in 27 nucleotides, whereas in KR they differed in 35 nucleotides (both excluding the 6 bp indel at 1452-1457; Fig. S3).Phylogenetic analysis indicates that the S. vulgaris atp1 duplication occurred after divergence from S. latifolia (Fig. 3A), but atp1.1 appears not to be present in all individuals (e.g.Kov4 & Kov5).

Relationship of bobt_KR and bobt_MV
The portion of bobt homologous to atp1 coding sequence differed in 10 nucleotides between bobt_KR and bobt_MV (Fig. 1, both atp1 segments concatenated; Fig. S1).We anticipated one of two outcomes when we placed the sequences from bobt onto a tree with the homologous regions from atp1: 1) either both bobt alleles would be sister to one another, reflecting a single evolutionary origin and subsequent divergence of bobt_MV and bobt_KR, or 2) both bobt sequences (MV or KR) would be more similar to the homologous sequence in atp1 from the same individual (MV or KR), a pattern reflecting either recent independent origins of the two bobt alleles or persistent and ongoing gene conversion.
Most relationships among sequences were not resolved when homologous regions in bobt and atp1 were analyzed together (Fig. 3B), indicating the introduction of substantial homoplasy into the dataset.To address the origins of this homoplasy, we analyzed the two atp1 segments present in bobt chimeric gene separately.
The phylogenetic tree based only on the 59 atp1 region of bobt (Fig. 1, first atp1 segment) is depicted in Figure 3C, and the tree based only on the 39 atp1 region (Fig. 1, second atp1 segment) is shown in Figure 3D.Partition homogeneity tests revealed strong incongruence between these regions of atp1 homology in bobt (P = 0.01) and neither phylogeny was consistent with predictions from a single origin of bobt without gene conversion (KH tests P,0.001).Interestingly, bobt_MV and atp1.2_MVshared 78 bp of identical sequence at the end of the region of atp1 homology 39 to the bobt stop codon (Fig. 1), a likely footprint of recent gene conversion in this small region.The 39 atp1 portion of bobt_KR is shorter than that of bobt_MV, thus any information from this region concerning the relationship of bobt_KR to the atp1 variants has been lost, or was never present.The cox2-derived portions of bobt_KR and bobt_MV were shorter than portions derived from atp1 and differed from one another by 4 substitutions (Fig. S1).Whereas three of these were unique to bobt (i.e.not present in the functional copies of cox2; Fig. S4), the fourth substitution exhibited a difference between bobt alleles such that each shared alleles with the corresponding functional cox2 region in each respective genome (KR or MV, position 231, Fig. S1).These relationships favored either very localized gene conversion or the independent origins of the two allelic versions of bobt.

Expression of bobt
Quantitative RT PCR indicated that bobt is expressed in both the MV and KR mitochondria, but its expression is over two orders of magnitude higher in KR than in MV (Fig. 4, Fig. S5).PCR experiments indicated that the bobt gene is co-transcribed with cob in the KR mt genome.Primers targeted to the coding region unique to bobt_KR and to cob generated a 1220 bp PCR fragment, as expected from the sequence of the bobt-cob region (Fig. S6).The cob gene is not adjacent to bobt in the MV genome, and was not co-transcribed.
Quantification by qRT PCR indicated that bobt transcript levels were twice as high in females compared to hermaphrodites in both MV and KR (Fig. 4).Co-transcription of bobt and cob in the KR genome allowed us to address the stage at which expression was differentially regulated in females and hermaphrodites.qRT PCR studies indicated that cob transcript levels were approximately 5 times higher than those of bobt.Whereas females had higher bobt transcript levels than did hermaphrodite sibs, we detected no significant difference in cob transcript levels between female and hermaphrodite sibs (Fig. 5).Moreover, the relative transcript levels measured with qRT PCR primers that targeted the junction between bobt and cob were similar to the relative transcript levels measured by the primers targeted to the coding region of cob.These results indicate that the vast majority of the cob transcript in the KR mt genome is derived from the cob gene copy cotranscribed with bobt and not from other copies of cob in the KR genome, if they are there.To exclude the possibility that differences in transcript levels reflected differences in gene copy number and not altered expression, we performed qRT-PCR with genomic DNA as a template under exactly the same conditions as measurements with cDNA.Copy numbers of all the analyzed regions appeared to be equal (Fig. 5B).Taken together, these data indicate that bobt expression is lower in hermaphrodites than in females in KR, whereas cob transcript levels are approximately equal in both genders.We speculate that a post-transcriptional process is responsible for this phenomenon.

Post-transcriptional editing of bobt
Comparing direct sequences from 3 cDNA PCR products to the genomic sequence in bobt (also directly sequenced) revealed that one site was edited in the atp1 region (position 784 Fig. S1) and one site was edited in the cox2 region of bobt (position 249 Fig. S1) for both the bobt_KR and bobt_MV alleles.These sites also were edited in the functional copies of atp1.2 and cox2 (Figs.S1 and S4) indicating that the signal for editing was transferred with the sequence when intragenomic recombination created bobt.

Discussion
This study provides unique insight into evolution of chimeric genes in plant mitochondria.Our observation of high divergence in nucleotide sequence of the two alleles of a complex chimeric gene, bobt_MV and bobt_KR, is surprising.We know of one other study, in Raphanus raphanistrum [21], that observed two alleles in a chimeric mt gene, but these differences were much less striking than what we found in S. vulgaris.One characteristic of S. vulgaris that may have contributed to high divergence between these alleles was the high mitochondrial DNA mutation rate in this species, which also influences the high diversity found in the atp1 gene [29,44,45].Overall, 43 single nucleotide differences were observed between bobt_MV and bobt_KR, counting from the start codon through the 39 regions homologous to atp1; 10 of these differences were within the segments homologous to atp1 (Figs. 1 and S1).
The divergence between the two bobt alleles and the atp1 variants in S. vulgaris allowed us to test whether bobt had a single evolutionary origin or two independent origins.The phylogenetic analysis of atp1 genic segments of bobt and the corresponding regions of the atp1 genes did not provide an unequivocal solution.Although it is possible that bobt evolved twice independently, given the complex structure of this chimeric gene and the presence of a large region with unknown homology, it is more parsimonious to assume a single origin of the gene.Assuming a single origin, the two bobt alleles have taken remarkably different evolutionary routes that include nucleotide divergence likely influenced by gene conversion and structural re-arrangement of flanking regions that have influenced expression patterns.In particular, our study contributes to the recent realization of the commonness of gene conversion in angiosperm mitochondria [46] that is motivating reassessment of the factors contributing to this genomes' evolution [47].
Gene conversion also likely contributed to divergence between bobt_MV and bobt_KR.Its affect can be most easily seen in the nearly identical sequences from sites 780-848 for atp1.2_MV and bobt_MV in Figure S1.Several instances of gene conversion in plant mitochondrial genes have been described recently [46,48,49] and patterns in bobt indicate that gene conversion must be a persistent characteristic of its evolution in Silene vulgaris.Moreover, independent gene conversion events appear to have occurred in each bobt lineage, generating patterns consistent with the phylogeny in Figure 3C.
The overall structure of bobt reveals clues regarding its origin.Chimeric ORFs are created from the fusion of DNA fragments via recombination across shared short repeats (,30 bp).Recombination at these short repeat sites in plant mitochondria is considered very rare and generally they are not considered to facilitate homologous recombination (Marechal and Brisson 2010), despite some recent findings that even very short repeats (6 bp) can mediate recombination [50].We found an 8 bp region of complete identity at the junction between the atp1 and cox2 segments of bobt, which is a likely site for repeated recombination between the atp1 and cox2 genes.Because the origin of the internal part of the bobt genes is unknown, we are unable to identify possible regions of similarity and potential recombination between the internal unknown region and the segments of cox2 and atp1.The presence of hotspots of recombination may increase the probability of independent origins of the two bobt alleles (bobt_KR and bobt_MV).We note that the same cox2 segment has been independently incorporated into two different chimeric regions in petunia and wheat, although the wheat region is likely a pseudogene [51,52,53].
Although not detected in previous studies [44,54], here we documented multiple copies of atp1 within the MV and KR genomes.The dynamic nature of angiosperm mt genomic rearrangements has been shown to result in dramatically different copy numbers and relative positions of genes within genomes [55,56,57].It is likely that Silene also harbors high variation in gene copy number among lineages and this is a likely explanation for the absence of previous detection.Nonetheless, caution is justified when assuming homology among mt coding regions in Silene and other plant species.
Bobt has many of the hallmarks of CMS genes.One commonality among most chimeric CMS genes is the presence of portions of ATP synthase subunits [1]; portions of these subunits are found in both bobt alleles.CMS genes also tend to be transcribed; they are not loss-of-function genes.Both bobt genes are transcribed, albeit at a comparably low level in the case of bobt_MV.Finally, nuclear restorers of fertility (Rf) often posttranscriptionally interfere with the products of CMS genes, so that hermaphrodites accumulate less transcript than females [58,59,60].Both bobt alleles exhibited lower transcript levels in hermaphrodite than in female plants.The much higher transcript level for bobt_KR than for bobt_MV in both female and hermaphrodite plants, however, may indicate significant differences in the functionality of these alleles including the possibility that KR_bobt is a CMS-related gene but MV_bobt is not.The different genomic context between bobt_KR and bobt_MV is likely related to the difference in transcript levels.bobt_KR is cotranscribed with cob, which lacks its own promoter, although a start of transcription from a sequence motif located upstream of the cob coding region, including the unknown region of bobt, cannot be completely excluded.In contrast, bobt_MV is not cotranscribed with cob.Although additional cob copies may be present in Krasnoyarsk plants, the high transcript level detected at the junction between bobt_KR and cob suggests that the cotranscript is a major source of cob mRNA.We may speculate that co-transcription plays a role in the inability to independently regulate transcription of cob and bobt_KR, however, our studies also show that bobt might be post-transcriptionally regulated independently from cob.We are curious to know whether the cotranscription may limit the options for new mutations to arise that down-regulate transcription of bobt_KR and other CMS-associated genes.Co-transcription of CMS-associated genes with essential protein coding genes is not unique to the present study.For example, orf456 is co-transcribed with cox2 in chili pepper [60], orf256 is co-transcribed with cox1 in wheat [61], and a CMSrelated factor is co-transcribed with nad6 in Mimulus [62].Molecular options for suppression of expression may be more limited when CMS genes are co-transcribed with housekeeping genes.
Our study indicates that highly divergent copies of complex chimeric genes with similar structures as CMS genes are present within the same species, albeit in individuals with wide geographic separation.The ecology and evolution of gynodioecious mating has been studied in S. vulgaris for many years and multiple studies have remarked on the high level of mitochondrial gene polymorphism [22,44,45], especially when compared to hermaphroditic congeners [22].Discussion of the source of mitochondrial divergence has been primarily whether it results from high mutation rates within gynodioecious species or balancing selection maintaining divergent copies over long time periods.Our study suggests that gene conversion may also be a process contributing to divergence.Combined with the increasingly convincing evidence for low frequency paternal inheritance of mitochondrial variation [63], gene conversion between two copies of a duplicated gene provides increasingly complex possibilities for the origins of plant mitochondrial genes.We should not let this complexity divert us from conducting carefully studies to determine the tempo and mode of plant mitochondrial genome evolution.
In summary, we have discovered two divergent allelic versions of a complex mitochondrial chimeric ORFs in different S. vulgaris individuals.Comparisons between homologous regions between the chimeric genes and atp1 are consistent with a history of independent gene conversion events in each lineage.Gene expression also has diverged both in quantity and the cotranscription of an adjacent cob gene in the bobt_KR version.These observations indicate that chimeric genes may persist for sufficiently long periods within species for significant evolution in nucleotide sequence, expression and genomic environment.

Figure 2 .
Figure 2. Southern hybridization results with bobt-specific and atp1-specific probes.Genomic DNA of Silene from various sites was digested with EcoRI and hybridized with probes derived from atp1 and bobt_MV.The same membrane was hybridized with both (A) an bobt_MV probe and (B) an atp1 probe.Lanes 1-5 S. vulgaris Mt.View; 6-8 S. vulgaris Krasnoyarsk; 9-11 S. vulgaris Beagle; 12 S. vulgaris Krasnoyarsk; 13 S. vulgaris Beagle; 14-16 S. latifolia Prague.Individuals from the Beagle population differ in Southern-RFLP pattern of atp1 region, which is very common in S. vulgaris populations.Faint, but visible, bands visible in Beagle and Krasnoyarsk DNA hybridized with a bobt_MV probe may correspond to weakly homologous regions in nuclear or mt DNA.The bands corresponding to bobt_KR are a bit weaker than bobt_MV bands due to nucleotide divergence between the two variants.One band detected by bobt probes suggests the existence of only one bobt copy in the mt genome, whereas two or three atp1-specific bands reflect the existence of two or three atp1 copies.Marker sizes are shown along the right hand side of each blot.The results of PCR with bobt specific primers are shown below (+,2).doi:10.1371/journal.pone.0030401.g002

Figure 3 .
Figure 3. Maximum likelihood phylogenies of atp1 and homologous regions of bobt from Silene vulgaris accessions.(A) Phylogeny using the complete sequence of atp1.The same topology was recovered when only the regions with homology to bobt were used.Numbers above lineages indicate bootstrap support for nodes in the phylogeny with the complete sequence, whereas numbers below the line represent support for node using only regions with homology to bobt.(B) Phylogeny including the shared regions of bobt, the atp1 genes found in the same individuals as bobt (KR & MV), and two additional atp1 copies from individuals from the Kovary CR population.Silene latifolia, Vitis vinifera, and Beta vulgaris atp1 sequences were used as outgroups.(C) Phylogeny of first 192 bases of the shared regions of bobt and atp1 homologous region found at the 59 end of bobt.(D) Phylogeny of the region 39 to the stop codon of bobt and the homologous region in atp1.This phylogeny includes 78 bases than are not found in bobt_KR, but are fond in the other accessions.doi:10.1371/journal.pone.0030401.g003

Figure 4 .Figure 5 .
Figure 4. Transcript levels of bobt_MV and bobt_KR in flower buds estimated by qRT PCR.Expression is presented as a ratio of the accumulation of PCR product for bobt and the mt 18S rRNA reference gene.Ratios for bobt_MV are multiplied by 100 to allow visualization of difference on the same scale as bobt_KR.The standard deviation was calculated on the basis of 3-5 sibling pairs F and H.The differences in bobt expression between genders were significant in both MV and KR plants (t-test P,0.05).doi:10.1371/journal.pone.0030401.g004Figure 5. Relative transcript levels in flower buds assessed at different regions of the bobt_KR-cob co-transcript.(Relative) transcript level at the atp1-cob junction and cob is more than twice that in the unknown region of bobt_KR.No signal was detected when RNA instead of cDNA was added to qPCR reaction mixture, which excluded contamination of RNA samples with genomic DNA.(A) Expression is presented as the ratio of the accumulation of PCR product for the specific region and the mt 18S rRNA reference gene and (B) copy numbers of different regions of bobt_KR-cob DNA as a ratio of the accumulation of PCR product for the specific region and the mt 18S rRNA reference gene.Whereas target gene and reference copy numbers are approximately equal, atp1-cob cotranscript and cob transcript levels are more than one order of magnitude lower than mt 18S rRNA.The black lines below the gene indicate the positions of PCR products used for quantitation.The standard deviation was calculated on the basis of 5 sibling pairs F and H.The significant differences (t-test P,0.05) between F and H are marked by **.doi:10.1371/journal.pone.0030401.g005
The presence of bobt was determined using a PCR assay with allele specific primers.doi:10.1371/journal.pone.0030401.t001