Mitochondrial DNA duplication, recombination, and introgression during interspecific hybridization

mtDNA recombination events in yeasts are known, but altered mitochondrial genomes were not completed. Therefore, we analyzed recombined mtDNAs in six Saccharomyces cerevisiae × Saccharomyces paradoxus hybrids in detail. Assembled molecules contain mostly segments with variable length introgressed to other mtDNA. All recombination sites are in the vicinity of the mobile elements, introns in cox1, cob genes and free standing ORF1, ORF4. The transplaced regions involve co-converted proximal exon regions. Thus, these selfish elements are beneficial to the host if the mother molecule is challenged with another molecule for transmission to the progeny. They trigger mtDNA recombination ensuring the transfer of adjacent regions, into the progeny of recombinant molecules. The recombination of the large segments may result in mitotically stable duplication of several genes.

www.nature.com/scientificreports/ that was quickly fixed, except for S. kudriavzevii × S. mikatae and S. cerevisiae × S. uvarum allotetraploids and the six species hybrids, which were all heteroplasmic. In addition, in the S. kudriavzevii × S. mikatae hybrid, the portion of the cox1 gene from S. mikatae seemed to have been introduced into the S. kudriavzevii mtDNA. Unfortunately, these recombined mitochondrial genomes were not completed 9 .
In summary, several mtDNA recombination events were reported, but altered mitochondrial genomes were not assembled. Therefore, the aim of this work was to analyze mtDNA transmission and the rate of recombination in S. cerevisiae × S. paradoxus hybrids in detail.

Results
Interspecific hybrids. To avoid any alteration due to auxotrophic mutations and to mimic the condition in nature (prototrophy), our first intention was to mate type strains of S. cerevisiae CBS 1171 to S. paradoxus CBS 432 31 as well as to S. paradoxus CBS 2908. However, CBS 1171 is an extremely poor sporulant; therefore, we used S. cerevisiae CCY 21-4-96 instead. S. cerevisiae and S. paradoxus differ in two basic phenotypic traits. The S. cerevisiae CCY 21-4-96 strain grows at 37 °C and produces rough colonies, whereas S. paradoxus strains do not grow at 37 °C and produce smooth colonies. Upon spore-to-spore mating, smooth colonies were prepared capable to grow at 37 °C from two independent crosses. The presence of both nuclear genomes in the hybrids was confirmed by the AluI polymorphism of the PCR D1/D2 domain products (Fig. S1). The ability to grow on YPGE plates demonstrated the presence of complete mtDNA. In addition, all hybrids were excellent sporulants, but the viability of their spore was low, from 2 to 14%. Eleven F1 hybrids, six from S. paradoxus CBS 432 × S. cerevisiae CCY 21-4-96 crosses and five from S. paradoxus CBS 2908 × S. cerevisiae CCY 21-4-96 crosses, were selected for further analysis. The presence of both parental chromosome sets in all F1 hybrids was confirmed by the karyotype analysis using pulsed field gel electrophoresis (PFGE) (Fig. S2).
Analysis of mitochondrial DNA (mtDNA). Restriction endonuclease analysis of mtDNA is widely used to distinguish between different yeast strains 32,33 . The S. cerevisiae and S. paradoxus mitochondrial gene orders differ from each other by two rearrangements, and they can be distinguished by HinfI or EcoRV digest 32,34 . mtDNA from 11 hybrids was subjected to EcoRV restriction analysis. The obtained patterns were compared for those of the parental strains (Fig. 1). Four hybrids originating from S. cerevisiae CCY 21-4-96 and the S. paradoxus strain CBS 432 showed mtDNA restriction patterns identical to those exhibited by the parental strains (Fig. 1A). However, two hybrids showed mixed restriction profiles (recombination rate 33.3%, hybrids 3-R1 and 6-R2; Fig. 1A). The 3-R1 mtDNA exhibited the same pattern as S. paradoxus CBS 432, but it contained some additional bands from S. cerevisiae CCY 21-4-96. The 6-R2 pattern showed a majority of bands identical to those of S. cerevisiae CCY 21-4-96, with one or two new larger bands and at least one band absent.  www.nature.com/scientificreports/ Four of the S. cerevisiae CCY 21-4-96 × S. paradoxus CBS 2908 hybrids acquired S. cerevisiae CCY 21-4-96 mtDNA, as can be inferred from their electrophoretic profiles (hybrids 1, 2, 3, and 5; Fig. 1B). Only one hybrid, 4-R3 (72 K), exhibited a mixture of bands belonging to both parents, with some parental bands missing (recombination rate 20.0%).
Rearranged mtDNA forms are stable. To exclude a possibility of heteroplasmic mtDNA populations present in zygotes with mixed mtDNA profiles, the hybrids were grown for approximately 20 generations, and mtDNA from 10 single colonies was purified and subjected to HinfI or EcoRV restriction analysis. All recombined mtDNAs were stable since their unchanged restriction profiles were found in all single colonies (Figs. S3, S4).
Characterization of mtDNA rearrangements. To determine the exact organization of the rearranged mitochondrial molecules, purified mtDNAs were directly sequenced with primers specific to the conserved regions of several mitochondrial genes (cox1, cox2, cox3, cob, rns, rnl, atp9, trnM). However recent advances in the assembly of mtDNA sequenced by Illumina MiSeq 34 allowed us to obtain complete sequences of recombined mitochondrial genomes (Table S1) that were confirmed by RFLP. Generally, the rearranged molecules were composed of a major skeleton from one parental molecule with small regions substituted with segments from another parental molecule (Figs. 2, 3). This observation suggested that recombination between two mtDNA molecules had occurred.
Hybrid 6-R2 molecule. The recombined 6-R2 mtDNA molecule received a majority of genes from S. paradoxus ( Fig. 2), but three genes, atp6, atp8, and trnE, were introgressed from S. cerevisiae. The recombination sites were identified in the cob and cox1 genes. The 5′ junction was located in the last exon of the cox1 gene, and the 3′ junction was located in the cob gene in the Spcob-I3 intron that is homologous to Sccob-I2 intron (Fig. 2).
The duplication of rns, cox3, and trnM (adjacent to rnpB) genes was also confirmed by Southern blot analysis (Fig. 4). The HinfI restriction pattern of the 4-R3 (72 K) mtDNA molecule was hybridized with radioactively labeled probes and compared to those of the parental strains. All three probes were hybridized to combination of parental bands in the restriction pattern of the 4-R3 (72 K) molecule.

How frequent is duplication in mtDNA?
We addressed the following question. At what frequency do mtDNA variants with duplicate regions arise? Selection based on colony appearance (smooth colonies) and the ability to grow at 37 °C provides only a limited number of hybrids. To analyze multiple recombined mitochondrial genomes, the mtDNA from S. cerevisiae CCY 21-4-96 was transferred via cytoduction to the auxotrophic laboratory strain 3C with geneticin resistance resulting in the strain 3C-CCY. S. cerevisiae 3C-CCY/S. paradoxus CBS 2908 hybrids were then selected on minimal medium with geneticin, and a larger set of sixty hybrids was analyzed. Although some recombined mtDNA were found, none of them contained the restriction profile typical for 4-R3 (72 K) (Fig. S5). Apparently, gene duplication in mtDNA is a rare event. To identify other recombination variants in hybrids, we sequenced three complete recombined mitochondrial genomes of hybrids (ZAN) by Illumina MiSeq. All recombined molecules are the result of transposition from S. cerevisiae to S. paradoxus (Figs. 3,4). The ZAN15 molecule received about a 23,840 kb segment from the beginning of cox1 up to the end of the cob gene. The precise recombination sites were at the beginning of the first exon of cox1 and the last exon of the cob gene. The ZAN31 molecule contains about 3280 bp long DNA containing ORF4 and atp6 from S. cerevisiae. The recombination sites were mapped in the region 251-332 downstream the atp6 start codon. The mtDNA in the ZAN37 hybrid results from two recombination events. The first one was triggered by the homing and co-conversion of flanking exons. The recombination sites were identified in the flanking exons. The second recombination event was also initiated by ORF4, and due to the mobility of cob-I2 it is about 14,760 bp long. The recombination sites were mapped downstream of the atp6 start codon and at the beginning of the last cob exon.

Discussion
In Saccharomyces yeasts, the diverse and highly reticulated mtDNAs show signatures of recombination and horizontal gene transfer within and between species [27][28][29][30][36][37][38] . To understand mtDNA transmission and the rate of recombination in interspecific Saccharomyces hybrids, we studied the progeny from the mating of two very closely related species: S. cerevisiae and S. paradoxus. The gene order of the S. paradoxus mitochondrial genome (European lineage) differs from that of S. cerevisiae by only two rearrangements and is well preserved within the species. In addition, there is wide variability in intron content and intergenic sequences, even among isolates www.nature.com/scientificreports/ paradoxus hybrids, two inherited mtDNA from S. paradoxus and six inherited mtDNA from S. cerevisiae, and in three hybrids we found mixed mtDNA restriction patterns, indicating recombination events. These hybrids were homoplasmic as the single colonies, and after approximately 20 generations, they provide the same restriction profiles. The frequency of recombination inferred from the restriction analysis of 60 S. cerevisiae 3C-CCY/S. paradoxus CBS 2908 hybrids was 15% (Fig. S5, Table S2). The exact organization of the rearranged mitochondrial molecules was determined by genome-wide sequencing, and the recombination sites were confirmed by primer walking sequencing. All rearranged molecules were composed of a major skeleton from one parental molecule, with small regions introgressed from the second parental mtDNA. Sequence differences between interspecific mtDNA molecules allowed us to precisely map the recombination spot within the 16-200 bp region (Fig. S6).
Specific mtDNA regions play a crucial role in recombination, namely mobile introns [44][45][46] and GC clusters 34,47 , where the insertion sites act as recombination "hot spots". Generally, the mobilization of group I (GI) introns coding open reading frame (ORFs) and free-standing ORFs are initiated by double-strand break formation in an intronless gene or ORF-free allele 15,45,48 . The intron or ORF-coded endonuclease (HOE) cleaves DNA, generating DNA ends that invade the homologous exon or ORF-lacking sequences of an element-containing allele. The process is called homing and is accompanied by the co-conversion of flanking sequences. Group II (GII) introns also possess a reverse-transcription-independent homing pathway that is initiated by the intron-encoded www.nature.com/scientificreports/ endonuclease (cleaves the antisense strand) and completed by the double-strand break repair (DSBR) recombination system of yeast mitochondria 44,49 . In addition, there is wide variability in intron content and intergenic sequences, even among isolates of the same species 28,34,36,39,40 . Therefore, if two isolates having different intron composition should mate, the homing/transposition pathway of numerous introns is potentially initiated. It was believed that a consequence of this process should be strong activation of the recombination system, associated with replication and repair machinery 26,50,51 . However, the role of repair machinery is questionable because the absence of certain genes from this group (NTG1, MGT1) does not affect the frequency of mtDNA recombination 26 . In S. cerevisiae mtDNA three genes (rnl, cox1,cob) are interrupted by introns. Among these (GII) I1 and I2, and group I introns (GI) I3α, I4α and I5α in cox1 are known to be mobile 34,35,52,53 . In many strains they are inactive due to the interruption with GC clusters or mutations introducing premature stop codons 28,34,35 . In cob gene GI I2, I3 and I4 introns code for maturases and only in some strains I2 might be mobile, owing to only 2 amino acid substitutions 34,35,54 . GI intron known as ω interrupts rnl gene and often contains "freestanding" open reading frame coding for I-SceI endonuclease necessary for its mobility 35,55 . Free standing ORFs were found during early sequencing experiments according to LAGLI-DADG motif reminiscent to the intron coded ORFs. HOE-like reading frames were found downstream cox2 gene continuous open reading ORF1, downstream cox3 ORF2 and ORF4, residing at the 3′ end of the atp6 gene (Fig. 5) 34,35 . Product of ORF4 (also referred to as ENS2), is the subunit of Endo.SceI endonuclease, whereas second subunit is coded in nucleus and imported to the mitochondria.
The occurrence of mobile elements near many recombination junctions suggests causal relationship. Five out of 14 junctions are located in the cox1 gene (Figs. 5, S6).
Three 5′ recombination junctions are in the sequence of exon 1 of the cox1 gene and are apparently the result of homing of mobile introns cox1-I1 and cox1-I2. Two 5′ recombination sites are in the same position between atp8 and atp6 genes, resulting from the transplacement of the ORF4 element. Most of the 3′ ends of novel recombination sites were found in the cob gene and are seemingly the result of the homing/transposition of the cob-I2. In summary, from seven identified recombination events in six different hybrid mtDNAs, three can be considered as the synergic result of the ORF4 and cob-I2 transplacement. One can be attributed to cooperative transposition of ORF4 with cox1-I, and the largest segment was mediated by their cooperative transfer with the ORF1 element. The most plausible explanation of all recombination events is homing/transposition of mobile elements accompanied by the co-conversion of flanking sequences, which can be several kbp long 56,57 . Apparently, ORF4 is the strongest element that triggers the recombination. ORF4 (ENS2) codes for Endo.Scel endonuclease www.nature.com/scientificreports/ that cleaves mtDNA every 2-3 kb, but the main cleavage site is at the end of the atp6 gene, allowing homing into the vacant site 58 . The main recombination site was in the region that was 700-900 bp upstream of the recognition site, seemingly due to co-conversion of flanking sequences. In addition, the recombination strength is elevated by the absence of ORF4 in both S. paradoxus strains used for hybrid construction 34,39 . Evidence of ORF4 involvement in the introgression of atp6 has been observed in the study of complex atp6 phylogeny 59 . In one event, the recombination of two large segments-trnE-cox2 from S. paradoxus and ORF1-atp8 from S. cerevisiae-yields duplication of the genes for cox3, rnpB, and rns and six tRNAs. Duplications in the mitochondrial genome are relatively rare and have been observed predominantly in plants [60][61][62] . Large-scale duplications of human mitochondrial DNA were found associated with Kearns-Sayre syndrome 63 . Data concerning duplication in yeast are scarce. To our knowledge there is only one paper describing the duplication from Clark-Walker laboratory 64 . The reported strain with duplication did not exhibit any growth phenotype. The same we observed in the case of our hybrid with duplication.
A characteristic feature of duplicate sections in animal mtDNA is tandemly arranged repetitive sequences resulting from slipping of the synthesized slipped strand mispairing fiber [65][66][67] . Major events in the evolution of animals, such as multicellularity and the emergence of symmetry, are accompanied by changes in the organization of mtDNA 68 . The most profound feature of Saccharomyces species is the conservative and species-specific gene order in their mtDNAs 34 . In addition, they offer a great opportunity to study evolutionary processes experimentally. The process called "hybrid speciation" implies that hybridization has been involved in the origin of new species 69,70 . All models of gene reordering in mtDNA consider the formation of duplicate regions 66,67 . A tandem duplication followed by random gene loss (TDRL model) is the most important mechanism of gene order rearrangements in mitochondrial genomes 68,71,72 . Consequently, the duplication of mtDNA may lead to changes in the gene order in the hybrid progeny and accompany the speciation process. Groth et al. 73 already proposed that a change in the mitochondrial gene order may be a step in sexual isolation and the generation of new species. If two isolates having different gene orders should mate, homologous recombination in the zygote would create a number of mtDNA molecules that would lack a complete set of mitochondrial genes or contain a gross duplication. Indeed, in our study when S. cerevisiae and S. paradoxus, having a different mitochondrial gene order, were crossed, a rare genome with the duplication of several genes arose, which can be used to test this hypothesis.
Duplication and recombination are mediated by the translocation of selfish mobile elements. Their extent of benefit and harmfulness for the host is still discussed 74 . Selfish mitochondrial introns, as well as free-standing ORFs, are beneficial to the host if the mother molecule is challenged with another molecule for transmission to

Methods
Yeast strains and media. Restriction analysis of 26S rRNA. Total DNA was extracted from each hybrid and parental strain 75 . The D1/D2 domain of the 26S rRNA gene was amplified by PCR using primers NL1 and NL4 76 . The PCR products were digested to completion with AluI restriction enzyme. The restriction fragments were resolved on 1% agarose gel.
mtDNA purification, restriction analyses, sequencing, assembly, and annotation. mtDNA was purified by bisbenzimide/CsCl buoyant density centrifugation 77 or by differential centrifugation 78 . Restriction analyses of mtDNA were performed using EcoRV and HinfI restriction enzymes 32 . Whole genome Illumina MiSeq sequencing and assembly was performed as previously described 34 . Briefly, paired reads were trimmed and assembled into individual contigs using CLC Genomics Workbench 9.5 (Qiagen, Hilden, DE). Contigs containing mtDNA were selected by comparison (BLASTN) with the known mtDNA from S. cerevisiae and S. paradoxus and assembled into a single molecule using the Vector NTI v.9.0 (v.10) software package from InforMax, Inc. Gene annotation was carried out using MFannot (http:// megas un. bch. umont real. ca/ cgi-bin/ mfann ot/ mfann otInt erface. pl) as described previously 34 . Novel junctions were sequenced directly by dye terminator sequencing chemistry with a Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) (ABI310 and ABI3100-Avant) as described in other work 39 .
Southern blot analysis. CsCl-purified mtDNAs digested with Hinf1 restriction endonuclease were analyzed by standard gel electrophoresis through 1% agarose gel at 2.0 V/cm in 1% TBE buffer. The DNA in the gel was denatured (1.5 M NaCl, 0.5 M NaOH) for 30 min, neutralized (1.5 M NaCl, 1 M Tris-HCl, pH 7.5) for 15 min, and transferred to a Hybond™-N + membrane (GE Healthcare) in 20 × SSC (1.5 M NaCl, 0.15 M sodium citrate) for 2 h by vacuum blotting (VacuGene TM XL). Finally, the membrane was washed in ddH 2 O, and DNA was fixed by UV light. Cox3 and rns genes were detected by Southern blot analysis using 32 P-labeled PCR products as probes (GE Healthcare). PCR products were purified with the QIAquick PCR extraction kit (Qiagen, Dorking, UK). Unincorporated nucleotides were removed by gel filtration through G-25 columns (GE Healthcare). After prehybridization, the membrane was hybridized (0.25 M Na 2 HPO 4 , 7% SDS, 1 mM EDTA) at 60 °C for 12 h. The membrane was washed twice at room temperature for 5 min and once at 60 °C for 20 min with 2% SDS and 100 mM Na 2 HPO 4 . The membrane was stripped (0.4 M NaOH) for 2 h at 40 °C and re-hybridized more than once. The trnM gene adjacent to the rnpB gene was detected using the Fmet oligonucleotide probe, of which the 5′ end was labeled with γ 32 P-ATP by a T4 polynucleotide kinase (ABgene). The membrane was pre-hybridized and hybridized as described above at 40 °C and washed twice at room temperature for 10 min. Signals were detected using Phospho-Screen (imaging Screen-K, 32*43 cm, catalog #170-7841, Bio-Rad) and Personal Imager Fx (Bio-Rad).