1. Cell Biology
  2. Genetics and Genomics
Download icon

Selections that isolate recombinant mitochondrial genomes in animals

  1. Hansong Ma
  2. Patrick H O'Farrell  Is a corresponding author
  1. University of California, San Francisco, United States
Tools and Resources
  • Cited 10
  • Views 1,998
  • Annotations
Cite as: eLife 2015;4:e07247 doi: 10.7554/eLife.07247

Abstract

Homologous recombination is widespread and catalyzes evolution. Nonetheless, its existence in animal mitochondrial DNA is questioned. We designed selections for recombination between co-resident mitochondrial genomes in various heteroplasmic Drosophila lines. In four experimental settings, recombinant genomes became the sole or dominant genome in the progeny. Thus, selection uncovers occurrence of homologous recombination in Drosophila mtDNA and documents its functional benefit. Double-strand breaks enhanced recombination in the germline and revealed somatic recombination. When the recombination partner was a diverged Drosophila melanogaster genome or a genome from a different species such as Drosophila yakuba, sequencing revealed long continuous stretches of exchange. In addition, the distribution of sequence polymorphisms in recombinants allowed us to map a selected trait to a particular region in the Drosophila mitochondrial genome. Thus, recombination can be harnessed to dissect function and evolution of mitochondrial genome.

https://doi.org/10.7554/eLife.07247.001

eLife digest

Animals store the main part of their DNA—including all of the genes that are required to build and maintain an individual—inside their cells in a structure called the nucleus. Most of the information stored in the DNA is stored in duplicate, with one copy inherited from the individual's mother via the egg, and the other from the father via the sperm. This duplicate storage allows a very important damage repair process to occur, where undamaged sequences in one copy can be used to repair damage in the other. This process of homologous repair uses mechanisms that are also used in another important genetic process. When sperm and egg cells are formed, the parental DNA goes through a process called homologous recombination, in which DNA molecules are cut and reassembled into new arrangements. This recombination process ‘shuffles’ genetic combinations, making every individual unique—a process of great evolutionary importance that allows natural selection to act on distinct traits.

The structures inside cells that generate energy—called mitochondria—also contain DNA, which is inherited only from mothers. Little is known about whether recombination is possible in the mitochondrial DNA of animals.

Ma and O'Farrell used genetics techniques to investigate recombination in the mitochondria of fruit flies. One experiment tracked how a mutation that makes flies less healthy at high temperatures spread as flies were bred for several generations. When the mutation was associated with a mitochondrial genome that had a strong drive towards replication, the mutation became more widespread over time, and in most cases, this eventually resulted in the mutation killing the flies. In rare cases, however, a few flies survived, giving rise to a healthy population. Molecular analyses revealed that, in these survivors, the defective genome had recombined with the other mitochondrial DNA to produce a new genome that lacked the mutation but retained the high replicative drive. This new recombinant genome worked normally and was able to resist the spread of the defective genomes.

In addition, by artificially cutting mitochondrial DNA, Ma and O'Farrell show that such ‘double-strand breaks’ lead to recombination, signaling a role for homologous repair in the repair of damaged, in this case broken, DNA. Recombination is also possible between the mitochondrial DNA of two different fruit fly species, and this recombination process can assemble long stretches of DNA.

Now that the recombination of animal mitochondrial DNA is known to be possible, future work will be required to understand how it works and how it affects evolution.

https://doi.org/10.7554/eLife.07247.002

Introduction

Homologous recombination operates in organisms from bacteriophage to human. This includes the mitochondrial genomes in many plant and fungal species (Rank and Bech-Hansen, 1972; Dujon et al., 1974; André et al., 1992; Zassenhaus and Denniger, 1994; Shedge et al., 2007). Nonetheless, there is very little support for recombination in animal mitochondria (Elson et al., 2001; Berlin et al., 2004; Hagström et al., 2013); lack of an identified mitochondrial RecA homolog, evidence of continuous lineages of mitochondrial haplotypes and a failure to detect recombinants in propagated heteroplasmic lines are taken as indications that it does not occur.

Despite evidence arguing against recombination of animal mitochondrial genomes, a variety of exceptional reports suggest that it can occur. The remarkable ‘double uniparental inheritance’ pattern of mitochondrial genomes in some bivalve mollusks has been associated with rare recombination events on at least an evolutionary time scale (Ladoukakis and Zouros, 2001; Ladoukakis et al., 2011). One human patient was reported to carry recombinant genomes (Kraytsberg, 2004), and there have been reports of recombinant mitochondrial genotypes in some species like lizard and fish (Guo et al., 2006; Ciborowski et al., 2007; Ujvari et al., 2007), but these reports are based on single individuals without documentation of parents or origin of the genomes presumed to have recombined. Given opposing observations, such as the introgression of intact genomes from one species into another (Solignac, 2004), it is not clear whether the cases reported are exceptional, or whether we have simply lacked the experimental power to directly demonstrate recombination in animals.

Recombination between small regions of nonallelic homology has been proposed to underlie deletion and insertion mutations (Mita et al., 1990; Bacman et al., 2009; Fukui and Moraes, 2009). However, when precise, recombination can generate favorable combinations of alleles, which, when coupled to the action of purifying selection, could increase in abundance to restore function (Muller, 1932). These positive or negative impacts on gene function could influence evolution, the behavior of disease mutations, and the age-associated degenerative changes of the mitochondrial genome.

Until recently, several factors have hindered detection of recombination of mitochondrial sequences. Chief among these, uniparental inheritance largely limits exposure of mitochondrial genomes to sibling genomes differing only at newly mutated sites (Birky, 1995; DeLuca and O'Farrell, 2012; Sato and Sato, 2013). Additionally, rare recombinant genomes can be difficult to detect: they can be stochastically lost during the random segregation, and if transmitted, they can be hard to track amid the chaotically segregating genomes. Finally, there are few markers suitable for design of conditions that would select for a rare recombinant genome.

Previous work in Drosophila showed that germline expression of a restriction enzyme targeted to mitochondria results in potent selection against mitochondrial genomes carrying a cognate cleavage site (Xu et al., 2008). Using this selection, a number of variant genomes lacking a particular site have been selected. In addition to removing a restriction site, these selected changes often also alter an encoded gene product (Figure 1A). One of the variants that lost a XhoI site is a temperature-sensitive lethal mutation of mt:CoI that can be counter-selected at high temperature (Hill et al., 2014; Ma et al., 2014). Moreover, one can transfer cytoplasm between early Drosophila embryos of different mitochondrial genotypes to create heteroplasmic lines that carry both the recipient and donor genomes for multiple generations (Matsuura et al., 1989; Ma et al., 2014). In addition to these tools, characterization of diverged mitochondrial genomes has revealed marked differences in their abilities to compete for transmission when combined in heteroplasmic combinations, a feature that we have been able to use as another selectable trait. In this work, we have combined these tools to create a powerful system in which we can test for the existence of homologous recombination and select for recombinant genomes.

Without selection, no recombination was detected in a stable heteroplasmic line after 60 generations.

(A) Mutants at the BglII and XhoI sites of mtDNA used in this study (Xu et al., 2008; Ma et al., 2014). The mitochondrial genome of Drosophila resembles that of mammals. It has little intergenic spacing, and encodes 13 polypeptides, all of which are involved in oxidative phosphorylation as well as 22 tRNAs and two rRNAs required for mtDNA translation. A single non-coding region (∼5 kb) called the ‘AT-rich region’ (dark brown) as it contains >90% A and T residues includes origins of replication and some repeated sequences of unknown function (Lewis et al., 1994). The genome contains one BglII and one XhoI site in the coding regions of mt:ND2 and mt:CoI, respectively. (B) No wild-type genome was detected by Southern blotting analysis in heteroplasmic lines where mt:ND2del1 and mt:CoIT300I were maintained in the same population for more than 60 generations at 29°C. In the heteroplasmic lane, 40 adults were sacrificed and their mtDNA were cut with both BglII and XhoI, and probed by a DIG-labeled sequence that hybridizes to mt1579–mt2369. The sensitivity of the Southern analysis was measured by loading a series of dilutions of wild-type mtDNA cut with both enzymes from 40 adult flies.

https://doi.org/10.7554/eLife.07247.003

Here, we provide clear evidence for homologous genetic exchange between Drosophila mitochondrial genomes under various conditions. The complete genomic sequence of parental and recombinant molecules details exchange events, and several recombinants are shown to involve transfer of a substantial segment of sequence from one genome to the other. We also show that exchange is stimulated by double-strand breaks (DSBs), as is recombination in many systems. Importantly, the success of the selections that we have applied shows that production of favorable combinations of alleles by recombination, even if rare, can have a profound benefit.

Results

Direct screening for recombinant genomes in a heteroplasmic line

We used Southern analysis to test for recombination between genomes distinguished by differences in restriction sites. A line heteroplasmic for mt:ND2del1, which lacks the BglII site, and mt:CoIT300I, which lacks the XhoI site, carries both genomes stably at 29°C (Figure 1B) (Ma et al., 2014). Each of the genomes of this heteroplasmic line has a defect complemented by the other so that the line is healthy but neither genomes is lost, a balancing selection (Ma et al., 2014). Recombination should produce wild-type genomes distinguished by the presence of both restriction sites and consequent production of a ∼1.6-kb fragment upon cutting with both BglII and XhoI. Even after maintenance of this line for more than 60 generations, this 1.6 kb band was not detected (Figure 1B). Reconstruction shows that this Southern assay can detect recombinant molecules at the level of 1 in 1000 (Figure 1B). Thus, like related experiments by others (Hagström et al., 2013), this physical assay failed to detect recombination. We conclude that recombination in this situation is not frequent (Sato et al., 2005).

Selection reveals recombination between mitochondrial genomes

We then developed methods to genetically select for recombinant mitochondrial genomes in the hopes that such approaches would both detect its occurrence and allow isolation of the recombination product. To give recombinant genomes an advantage, we produced a heteroplasmic line wherein one genome was compromised by a temperature-sensitive mutation and the other was compromised in its ability to compete for transmission. A temperature-sensitive genome with two mutations, mt:ND2del1 + mt:CoIT300I, was introduced into a strain with a genome named ATP6[1] (Celotto et al., 2006, 2011). High temperature, 29°C, selected against the mt:CoIT300I allele, and in previously analyzed heteroplasmic lines where the partner genome was a closely related wild-type genome, this selection resulted in a multi-generational decline of the temperature-sensitive genome and its eventual elimination (Ma et al., 2014). To our surprise, when in competition with the ATP6[1] genome, which is distinguished by numerous sequence polymorphisms and a shorter AT-rich region, the temperature-sensitive genome displaced the ATP6[1] genome over a few generations at 29°C, even though the flies homoplasmic for ATP6[1] are relatively healthy and apparently more robust than mt:ND2del1 + mt:CoIT300I flies at both temperatures. The decline of ATP6[1] leaves the temperature-sensitive genome without complementing mt:CoI activity, and the entire population dies after several generations (Figure 2A).

Figure 2 with 1 supplement see all
Selection revealed homologous recombination in a heteroplasmic line containing the ATP6[1] genome and the temperature-sensitive double-mutant: mt:ND2del1 + mt:CoIT300I.

(A) The abundance the ATP6[1] genome declined when co-resident with mt:ND2del1 + mt:CoIT300I at 29°C. After several generations, the flies at 29°C started to die. (B) A combination of a restriction fragment length polymorphism and a restriction site difference reveals the emergence of a recombinant genome. mtDNA isolated from 40 adults from each generation was cut with EcoRI in the presence or absence of XhoI. The schematics show the distribution of the EcoRI and XhoI sites on the whole parental genomes (left) and a detail of the largest EcoRI fragment with the position (purple bar) of a hybridization probe (center). Southern analysis shows single and double cut samples taken at different generations during the selection. Only the two parental bands were detected early, from G0 to G5 (shown for G3). From G6 onward, a third EcoRI fragment appeared that had a length characteristic of the mt:ND2del1 + mt:CoIT300I genome but with a XhoI site. By G7, the ATP6[1] specific fragment was not detected while a new apparently recombinant genome dominated the population. (C) Detailed maps of three genomes sequenced by PacBio SMRT. Red lines indicate mismatches between ATP6[1] and mt:ND2del1 + mt:CoIT300I sequences. The ATP6[1] genome also lacks ∼1.6 kb of the AT-rich region (two type I repeats and two type II repeats, see Figure 2—figure supplement 1A for details). Pink arrows indicate approximate points of exchange with the given range defined by the nearest neighboring polymorphisms (see deposited full sequences in GenBank as KT174472, KT174473 and KT174474). (D) Proposed progression leading to the recombinant. The original ATP6[1] genome is still displaced, but the newly emerged recombinant competes effectively and persists. By generation 6, the abundance of the recombinant genome is sufficient to complement the temperature-sensitive genome so that some viable flies sustain the line. Over subsequent generations at 29°C, the recombinant genome increases in relative abundance because of selection against the temperature-sensitive genome.

https://doi.org/10.7554/eLife.07247.004

To select for possible recombinant genomes, we followed five independently established lines with a high starting proportion of ATP6[1] genomes (50–90%) at 29°C. Initially, all the lines were healthy and grew productively, but with decline in ATP6[1] abundance, the health of the lines held at 29°C declined abruptly after five or six generations and died out within the next few generations. One line went through a similar crisis with diminished survival, recovered and continued to produce viable progeny in subsequent generations. Southern analysis showed emergence of a new genotype, which was cut by XhoI (like the ATP6[1] genome), but had a long AT-rich region (like the double mutant) (Figure 2B). To map recombination sites, we sequenced the parental and recombinant genomes using the PacBio single molecule real-time sequencing technique (SMRT), whose long reads gave us unambiguous sequence that included the repeats of the 5 kb highly AT-rich noncoding region (Figure 1A and Figure 2). Complete sequences of these genomes revealed that the two parental genomes differed by more than 100 SNPs plus >20 indels, and the ATP6[1] genome lacked ∼1.6 kb of the AT-rich region that was present in the mt:ND2del1 + mt:CoIT300I genome (Figure 2C and Figure 2—figure supplement 1A). The recombinant genome was the result of an exchange of a large continuous segment to produce an ∼60%/40% chimera of the ATP6[1] and mt:ND2del1 + mt:CoIT300I genomes that lacks the mt:ND2 and mt:CoI mutations but contains the non-coding region of the mt:ND2del1 + mt:CoIT300I genome (Figure 2C).

Southern analysis also showed the abundance of the recombinant genome at later stages of the selection. After generation 6, the original ATP6[1] genome was no longer detected (Figure 2B), and a new heteroplasmic line was formed with the other parental genome mt:ND2del1 + mt:CoIT300I and the recombinant genome. This line was viable at 29°C as the temperature-sensitive defect of the double mutant is complemented by the ATP6[1] mt:CoI of the recombinant genome (Figure 2D). Over subsequent generations, a multigenerational selection for function caused an increase in the proportion of the recombinant genome (Figure 2B), showing that it had lost the transmission disadvantage of the parental ATP6[1] genome. This led us to conclude that the ability to compete for transmission is localized to the sequences of the recombinant that came from the temperature-sensitive genome. This includes the entire AT-rich regulatory region and some flanking sequences distinguished by three SNPs (Figure 2C).

We later isolated two other recombinant genomes by following another 46 heteroplasmic lines as above. Both recombinants had a size (∼19.5 kb) similar to the temperature-sensitive genome, implying that mt:ND2del1 + mt:CoIT300I was the source of the regulatory region. By sequencing the coding region, we show that one recombinant has the entire coding sequence of the ATP6[1] genome, whereas the other contains a much smaller segment of ATP6[1] extending at least from mt671 to mt5978, with the rest of the coding region belonging to the mt:ND2del1 + mt:CoIT300I genome (Figure 2—figure supplement 1B). We conclude that our selection has isolated recombinant genomes including extensive stretches of sequence originating from the parental genomes and exhibiting functional traits of these parental genomes.

Homologous exchange upon introduction of reciprocal DSBs

Normal meiotic recombination is induced by DSBs and studies in many contexts have suggested that DSBs, whether experimentally produced or secondary to DNA damage, greatly stimulates recombination. Indeed, DSBs are thought to be the key initiators of homologous exchange. To test the importance of DSBs for mtDNA recombination, we set up a condition to select for recombination in conjunction with restriction cutting of the mtDNA.

We had produced genomes that lacked the XhoI site, mt:CoIR301Q, or the BglII site, mt:ND2del1. While each genome is resistant to one enzyme, expression of mito-BglII and mito-XhoI simultaneously in the germline should cut either of these genomes. Indeed, germline expression of both enzymes in flies that were homoplasmic for either mt:ND2del1 or mt:CoIR301Q led to sterility in females (Figure 3A). As previously shown, resistance ought to emerge at a low frequency due to mutations at the second restriction site (Xu et al., 2008). Indeed, upon selection ∼1% of the females were weakly fertile, giving a few escaper F1 progeny harboring a variety of mutant alleles that inactivate the remaining restriction site. In contrast, even though the combination of enzymes was able to select against both genotypes, when flies were heteroplasmic for the two genomes (in a 50:50 ratio), fertile females were very frequent (>90%), and most (60%) were as fecund as wild type. PCR, DNA sequencing, and phenotypic analysis showed that these progeny carried a single genome with the restriction-resistant alleles found in the two parents (Figure 3A).

Introducing DSBs into both parental genomes vastly induces homologous recombination in two heteroplasmic lines.

(A) Expression of both mito-BglII and mito-XhoI enzymes in the germline effectively sterilizes most females carrying either mt:ND2del1 or mt:CoIR301Q genome, resistant to only one of the enzymes, but resistant progeny appear at a much higher frequency if the female carries both types of sensitive genomes. The resistant progeny from the heteroplasmic parent are homoplasmic for a newly generated recombinant genotype carrying both parental alleles: mt:ND2del1 + mt:CoIR301Q. (B) The same level of rescue was observed when a different heteroplasmic line was used, which had a different starting ratio of two parental genomes. Results are means ±SD (n = 4 × 100 for each heteroplasmic line).

https://doi.org/10.7554/eLife.07247.006

A second heteroplasmic line where the mt:CoIR301Q allele was replaced with the mt:CoIT300I allele gave a similar frequency of recombinant progeny, even though the starting ratio for the two parental genomes was not equal (Figure 3B). The resulting recombinant lines exhibited the phenotypes expected for the input alleles. For instance, mt:ND2del1 + mt:CoIR301Q flies were male sterile, and mt:ND2del1 + mt:CoIT300I flies were temperature lethal.

The success of a second and very different selection confirms that homologous exchange between mitochondrial genomes can occur. Most females exposed to the restriction enzyme selection were fertile, while in the selection without DSBs (above, Figure 2), the recombinant emerged from a population over several generations of selection. The difference suggests that the frequency of recombination upon expression of the two restriction enzymes is much higher. We propose that DSBs produced by restriction enzyme cutting induces exchange as well as selecting for the products that are resistant to cutting.

To test whether recombination could also occur in somatic tissues, we examined the consequence of expressing both restriction enzymes in the developing eye under the control of ey:GAL4 driver. The high level of expression that occurs at 29°C results in pupal lethality (headless phenotype) in 99% flies that are homoplasmic for mitochondrial genomes mutant at only one site and the few survivors are eyeless or have a small eye phenotype. Similar expression in flies heteroplasmic for the two mutant genomes gave survivors at 10% with most survivors showing well-developed eyes (Supplementary file 1), suggesting that recombination of mtDNA also occurred in somatic tissues exposed to the double restriction enzyme selection. Somatically, active recombination could impact the stability of the mitochondrial genome in the soma, a factor thought to be important in aging (Cortopassi et al., 1992; Liu et al., 1998; Cao et al., 2001; Bender et al., 2006).

Recombination upon introduction of a single DSB

Given the high frequency of recovered recombinants when both genomes of a heteroplasmic strain were cut, we asked whether we could detect recombination if DSBs were introduced into only one genome. To do this, we made a heteroplasmic line containing the wild-type genome and the temperature-sensitive double-mutant mt:ND2del1 + mt:CoIT300I, and used expression of restriction enzyme to select against wild-type genome and used the temperature-dependent selection against the double mutant. When we expressed mito-BglII in the germline while keeping the flies at 29°C, about 14.6% of the F1 females were viable and fertile (Figure 4A). PCR analysis and restriction digestion showed that the progeny had lost the wild-type allele of mt:ND2, which was targeted by the restriction enzyme, but they were heteroplasmic for a new mt:ND2del1 genome and the double-mutant parental mt:ND2del1 + mt:CoIT300I genome (Figure 4A). Apparently, the wild-type mtDNA was cut and homologous repair or exchange using the mutant mt:ND2del1 sequence led to loss of the site. Initially, the abundance of the newly generated mt:ND2del1 genome was very low in all the progeny (Figure 4B), probably because only a small fraction of the wild-type genomes underwent homologous exchange. However, because a low level of the recombinant genome is sufficient to rescue the temperature-sensitive phenotype of mt:ND2del1 + mt:CoIT300I, this low level of recombinant genome supported production of fertile heteroplasmic flies. Since the more functional mt:ND2del1 genome has a selective advantage at 29°C, its abundance increased over subsequent generations (Figure 4B), and in some lineages, the mt:ND2del1 + mt:CoIT300I genome was completely eliminated after 18 generations (Ma et al., 2014).

Homology-dependent conversion of the wild-type BglII site into the sequence of the mt:ND2del1 allele following cutting of the BglII site.

(A) Expression of mito-BglII in the germline of flies heteroplasmic for wild-type and mt:ND2del1 + mt:CoIT300I genomes at 29°C led to isolation of progeny with repaired wild-type genome at the BglII site and converted it to mt:ND2del1. (B) The starting abundance for the newly generated mt:ND2del1 was low, but it increased over generations as the genome possessed a selective advantage at 29°C. The abundance of mt:ND2del1 was measured by PCR amplifying a mtDNA region (mt1826–mt2496) using mtDNA from 40 adults as template followed by restriction digestion using XhoI in four heteroplasmic lineages from generation 1 to 3.

https://doi.org/10.7554/eLife.07247.007

Again, we isolated recombinant genomes at a relatively high frequency, which leads us to propose that a single DSB is sufficient to promote recombination (see ‘Discussion’). However, we note that without additional markers, we could do little to characterize the nature of the exchange events.

Exchange between the mitochondrial genomes of two different species

In order to study whether DSB-induced exchanges involve restricted local repair events, or exchange of longer stretches of DNA, we needed more markers. To achieve this, we examined recombination between the diverged mitochondrial genomes of Drosophila melanogaster and Drosophila yakuba. We made a heteroplasmic D. melanogaster line in which the D. yakuba mtDNA is sensitive to BglII and the D. melanogaster genome is sensitive to XhoI (see ‘Materials and methods’) (Figure 5A). Within the coding sequences, the D. yakuba genome shared about 93% sequence identity with mt:ND2del1, while the non-coding AT-rich sequence was highly diverged and reduced to about 1 kb in length. We introduced DSBs into both genomes by germline expression of mito-BglII and mito-XhoI and found that about 10% of the females were fertile. Progeny homoplasmic for recombinant genomes were recovered. All propagating recombinants contained the AT-rich region of the D. melanogaster genome (assessed by Southern analysis, Figure 5—figure supplement 1), which we expected because D. melanogaster genomes outcompete D. yakuba genomes very quickly when co-resident (H Ma and PH O'Farrell, unpublished). We characterized the mitochondrial genomes of three progeny lines by PCR and standard sequencing. All the recombinant genomes had one crossover point very close to but upstream of the XhoI site of the mt:ND2del1 genome (Figure 5B). A second crossover is considerably downstream: in two of the recombinants, the other crossover occurred between markers at mt3124 and 3184, and between markers at mt3379 and 3445. In the third case, the other crossover is further downstream and we obtained mixed sequencing signals for a region between mt5524 and mt6291 (Figure 5C). Flies were likely to be heteroplasmic for that particular region as if multiple recombinants are carried in this line. Based on these recombinants, we concluded that homology-based repair of DSBs promoted exchange of a substantial uninterrupted stretch of sequence. We also note that there were no discontinuities in the mapped region of exchange as might occur if repair randomly converted mismatches in heteroduplex in one direction or the other (see below for more discussion).

Figure 5 with 1 supplement see all
Exchange between two mitochondrial genomes of different species that share 93% sequence homology.

(A) A heteroplasmic line containing both a Drosophila yakuba (mt:ND1R274W) genome and a Drosophila melanogaster (mt:ND2del1) genome was made. 90% of heteroplasmic females induced to express both mito-BglII and mito-XhoI enzymes in the germline were sterile, but the escaper progeny from the fertile females contained newly generated recombinant genotypes. (B) Sequences of the parental and three recombinant genomes reveal that one recombination junction is close to the XhoI site of the D. melanogaster genome. The XhoI site (yellow highlight) present in the D. melanogaster parent sequence (top line) is absent in all the other sequences. Pink rectangles indicate the interval in which recombination occurred. (C) A schematic illustration of where the second crossover was for the recombinant genomes. Arrows indicate the approximate position and the numbers, which are based on the D. melanogaster sequence, given the positions of bounding polymorphisms.

https://doi.org/10.7554/eLife.07247.008

Discussion

This study, which provides direct evidence for homologous recombination in Drosophila mtDNA, opens up the possibility of recombinational mapping of functions on the mitochondrial genome in animals. While the details of recombinational mechanism remain to be worked out, our analysis outlines some of its features. If widespread in animals, the recombinational activity we have observed is likely to have a pervasive influence on the genetics and evolution of metazoan mitochondrial genomes.

Features of mitochondrial recombination

Multiple modes of homology-dependent exchange are used in different organisms and different situations. Though differing in their details, these different modes often involve DNA breaks of one or both strands, resection of ends, heteroduplex formation, local replication, and resolution. Some features of the processes contributing to mtDNA recombination can be inferred from the sequence of three D. melanogaster/D. yakuba recombinants that we isolated following introduction of DSBs.

In all cases, one crossover point was very close to the cleavage site of XhoI, indicating stimulation of exchange by DSBs. Some resection (8–44 bp) must have occurred as the sites of exchange are located two or three SNPs upstream of the cleavage site (Figure 5B). Surprisingly, the actual exchange events occur in sequences with frequent interruptions of homology such that one of the exchanges occurred within a stretch of only 11 bases of homology. In contrast, some of the more thoroughly characterized homology-dependent mechanisms such as those catalyzed by RecB-C and RecF require significantly greater homology (23–27 bp and 44–90 bp, respectively) (Shen and Huang, 1986). This discordance favors other mechanisms such as break-induced recombination (BIR), a template switching mechanism of recombination, which can use very short regions of homology in an exchange process referred to as microhomology-mediated BIR (Hastings et al., 2009; Anand et al., 2014), to be the underlying mechanism.

The second site of exchange is far from either cleavage site, but seems to occur in regions with high homology. It is not clear whether both exchanges occur at the same time: since the genomes are circular, some types of exchange will produce a dimer circle that might resolve by an intramolecular recombination event at a later time. Our finding that one of our recombinant lines carried more than one recombinant genome with a common point of exchange near the XhoI cleavage site but different downstream points of exchange is consistent with this possibility. If separated in time, propagation of the unresolved product of the first recombination could give a population of molecules and independent resolution events could give a mixed population. While these analyses give us only a limited perspective on the mechanism of recombination, it is notable that the abundant SNPs suggest that each of the characterized events involved exchange of a substantial segment without interruptions (at least 1 kb), rather than restricted local repair.

When no DSBs were introduced by restriction enzymes in lines heteroplasmic for ATP6[1] and mt:ND2del1 + mt:CoIT300I, all of the recombinants again involved the exchange of a substantial and continuous stretch of sequence (at least 5 kb), indicating either the exchange of extended region of duplex or priming of an extended stretch of replication template by a second genome. We also noted that two of the three recombinants had a crossover site very close to where the Drosophila mTERF (DmTTF) binds (mt6314–mt6341 and mt11698–mt11725, respectively [Roberti et al., 2003]). Although a relatively weak association, we call attention to it because mtDNA replication pauses at the DmTTF binding sites (Jõers et al., 2013), and such pauses might destabilize forks and promote exchange.

The purpose of mitochondrial recombination

Homologous recombination plays major roles in repairing DSBs and in genetic exchange. Below, we discuss the significance of these two roles for maintaining mitochondrial genomes in natural populations.

Our finding that restriction cutting gives frequent recombinant products demonstrates that this process is active in the repair of these DSBs in mtDNA. While genetically distinct genomes in heteroplasmic flies provide the means to detect homologous exchange, sibling molecules ought to have a greater opportunity to guide homology-dependent repair as they will occur, at least transiently, in the same mitochondrion or nucleoid. In homoplasmic flies, accurate repair using sibling molecules will leave no trace. Only misaligned intra-molecular recombination between repeats would give a clear signature. Occurrence of recombination junctions in regions of low homology in this study supports suggestions that non-allelic recombination can contribute to formation of deletions and duplications in the mitochondrial genomes—a frequent cause of mitochondrial defects during aging. Such a recombination could also explain several observations found in natural population (Rand and Harrison, 1989; La Roche et al., 1990; Lunt and Hyman, 1997; Ludwig et al., 2000). For instance, mtDNA of European rabbit (Oryctolagus councils) has repeated 153-bp motifs in the vicinity of the replication origin of H strand and very individual carries genomes with different numbers of these repeats (Casane et al., 1994). In Drosophila, mtDNA varies in size from 16 to 20 kb (Solignac et al., 1986; Townsend and Rand, 2004; Rand, 2011), mainly due to the variation in the length of the noncoding region, which contains two types of tandemly repeated elements (Lewis et al., 1994). Rand has shown that spontaneous changes in length in populations as if repeats can come and go at a relatively rapid pace (Rand, 2011). While accurate repair using sibling molecules is likely to be the main purpose of mitochondrial recombination, more rare non-allelic events can have a major impact on the production of variants, and exchanges between distinctly marked genomes, as reported here, provide a means to characterize the process.

Recombination is detected by and well known for its ability to promote genetic exchange. However, animal mtDNA is restricted in the types of exchange that can occur because uniparental inheritance prevents intermingling of genomes. The mitochondrial genome of each female is sheltered from encounters with foreign genomes and will be passed on in an isolated lineage or clone. Genetic exchange between distantly related genomes might occur as a result of occasional paternal transmission that introduces a foreign genome (e.g., Kraytsberg, 2004), but the significance of such violations of uniparental inheritance are matters of an ongoing debate. For the most part, the integrity of mitochondrial haplotypes suggests that the contribution of exchange among distantly related genomes is relatively small (Elson et al., 2001).

The limitations on genetic exchange imposed by uniparental inheritance, which are similar to those in asexually propagating organisms, do not mean that there is no meaningful genetic exchange (Hurst and Peck, 1996). It has become increasingly apparent that organisms can carry more than one mitochondrial genotype, that is, they are naturally heteroplasmic because of new mutations, stably transmitting heteroplamic combinations, or a breakdown of uniparental inheritance (Solignac et al., 1983; Ladoukakis et al., 2011; Tsang and Lemire, 2011; Payne et al., 2012; Ma et al., 2014; Ye et al., 2014). In this sense, the situation we created by introducing the temperature-sensitive genome into the ATP6[1] line, might be a model for things that can happen naturally. Emergence of a new genome with strong selfish drive, but with a defect in function, or introduction of such genome from another lineage will create a natural situation analogous to our experiment. The outcome of our experimental manipulation shows that this situation has generally detrimental consequences. Indeed, the temperature-sensitive mutation together with the drive advantage created a sort of ‘population time bomb’: the lineage remained healthy for multiple generations allowing the population to expand greatly, and then it collapsed upon elimination of the functional mt genome. Even rare recombination can uncouple a positively selected drive mutation from detrimental mutations, and as in our experiment, selection can then restore health. Thus, even occasional genetic exchange would prevent rogue genomes from wiping out lineages.

In conclusion, we show that recombination among mitochondrial genomes occurs in Drosophila and that this recombination can be used to manipulate genomes for functional mapping. We suggest that recombination will influence evolution of the mitochondrial genome in animals and impact the genetic behavior of mitochondrial disease mutations.

Materials and methods

Experimental procedures

Fly stocks

The homoplasmic stocks used in this study include the following mutant alleles: mt:ND2del1, mt:CoIR301Q, mt:CoIT300I and mt:ND2del1 + mt:CoIT300I. w1118 was used to provide wild-type mtDNA. Flies homoplasmic for ATP6[1] mtDNA was kindly provided by Michael Palladino (University of Pittsburgh, U.S.). D. yakuba flies were obtained from Drosophila species stock center, San Diego. Other strains used included UAS-mito-BglII, UAS-mito-XhoI, UAS-mito-PstI, nos-Gal4, and ey-Gal4. The stocks were cultured at 18–25°C on standard fly medium.

Generation of heteroplasmic flies

Poleplasm transplantation was used to generate heteroplasmic flies, and the method was described in Ma et al. (2014). Basically, a portion of the poleplasm was sucked out from donor embryos and transferred into the posterior end of the recipient embryos. The injected recipient embryos were kept in a humidified chamber at 22°C, and hatched larvae were transferred to vials with yeast paste in the next 2 days and incubated at 22°C until eclosion. Lines were established from the females obtained from the injected embryos, which were systematically mated to males with the recipient mtDNA genotype to ensure that the only source of newly introduced mitochondrial genomes was the injected material (i.e., that it did not arise from the purported possibility of exceptional paternal transmission). For each of these females, 10–30 F1 females were isolated to establish sublines. When not specified, each generation was derived from at least 50 individuals belonging to the previous generation.

The procedures to make the heteroplasmic line with both D. melanogaster (mt:ND2del1) and D. yakuba (mt:ND1R274W) genomes involve the following four steps. Initially, in order to make a D. melanogaster line with the wild-type D. yakuba mitochondrial genome, cytoplasm from D. yakuba embryos was transplanted into the temperature lethal mutant (mt:ND2del1 + mt:CoIT300I) embryos, and eclosed adults were kept at 29°C to select for flies with the D. yakuba genome. By doing this, two independent lines were established and both stably transmitted D. yakuba mitochondrial genome (3–4%) from generation to generation at 29°C. Secondly, a mitochondrially targeted restriction enzyme, mito-PstI, was expressed in the germline mitochondria of the two heteroplasmic lines to eliminate the mt:ND2del1 + mt:CoIT300I genome, as PstI site was only present in the D. melanogaster genome. Through this, several lines with only wild-type D. yakuba mtDNA were established. Surprisingly, D. melanogaster flies homoplasmic for D. yakuba mtDNA were as healthy as wild-type flies at both 22°C and 29°C. Similar to the D. melanogaster genome, wild-type D. yakuba mtDNA has one BglII site and one XhoI site. The BglII site is located in the same place as in D. melanogaster's genome (mt:ND2), whereas the XhoI site, unlike the D. melanogaster genome (which is in the mt:CoI coding region), is located in the mt:ND1 coding region further downstream. Thirdly, expression of mito-XhoI enzyme in the germline of flies with the wild-type D. yakuba genome was used to select a mutant derivative with an allele mt:ND1R274W that removed the XhoI site. This led to a line that is sensitive to BglII cut but not to XhoI. Finally, this line was then used as a recipient for making the mt:ND2del1/mt:ND1R274W heteroplasmic line.

DNA isolation

Genomic DNA was extracted from adults as described in Ma et al. (2014). In brief, flies were mechanically homogenized with a plastic pestle in homogenization buffer (100 mM Tris-HCl [pH 8.8], 0.5 mM EDTA, 1.0% SDS). The homogenate was incubated at 65°C for 30 min, followed by addition of potassium acetate (to 1 M) and incubation on ice (30 min) to precipitate protein and SDS. Subsequently, the homogenate was centrifuged at 20,000×g for 10 min at 4°C. DNA was recovered from the supernatant by adding 0.5 vol of isopropanol and centrifuging at 20,000×g for 5 min at room temperature. The resultant pellet was washed with 70% ethanol and suspended in 100 μl of ddH2O per fly. mtDNA genotype frequencies were measured in individual founding females and their further generations via qPCR. When populations were analyzed, we extracted DNA from groups of 40 individuals.

mtDNA isolation

Unfertilized eggs were the best source for isolating pure mtDNA without much contamination from genomic DNA: each egg contains 10 ± 2 million copies of mtDNA, which provides about 0.2 ng mtDNA but very little nuclear DNA. In order to isolate a large number of unfertilized eggs, virgins (XX) collected from flies with a certain mtDNA genotype were crossed to males with compound XY (C[1:Y]). Because the progeny of such a cross were male sterile, bulk mating of the progeny gave an abundant source of unfertilized eggs. The unfertilized eggs were then collected and dechorionated with 50% bleach and washed before being homogenized in STE buffer (100 mM NaCl, 50 mM Tris-HCl, pH = 8.5, 10 mM EDTA) using Kimble-Chase Kontes 2-ml glass Dounce tissue grinder. The homogenate was supplemented with 1% SDS and proteinase K and incubated at 55°C for 1 hr. The sample was further treated with RNase A at 37°C for 30 min before phenol–chloroform extraction and ethanol precipitation.

Southern blotting

Southern blotting was used to detect the recombinant mitochondrial genome. It was performed as described in Ma et al. (2014). Basically, digested DNA was separated on a 0.8% agarose gel by electrophoresis and transferred to Hybond N+ membrane by the capillary method. DNA transferred to the membrane was fixed by UV cross-linking (Stratalinker 120 mJ). The blot was hybridized with PCR-generated probes (mt1577–mt2365) that were labeled with DIG-11-dUTP using 0.35 mM DIG-11-dUTP, 1.65 mM dTTP, and Bioline Velocity Taq DNA polymerase, following the manufacturer's instructions (Roche, Germany). Prehybridization and hybridization were carried out at 41°C overnight in DIG Easy Hyb buffer solution (Roche). The membrane was washed two times with 2 × SSC + 0.1% SDS at room temperature for 5 min and twice for 15 min with 0.1 × SSC + 0.1% SDS at 65°C. Hybridized membrane was visualized with NBT/BCIP following the manufacturer's instructions (Roche). mtDNA primers used to generate the DIG-labeled probe are listed in Supplementary file 2.

Sequencing mtDNA via PacBio SMRT technique

Due to a long repetitive sequence in the AT-rich non-coding region of Drosophila mtDNA, PacBio SMRT was used to sequence the whole mitochondrial genome for three genotypes. SMRT sequencing can generate extraordinary long reads (>30,000 bp) with extremely high consensus accuracy (>99.999%). Thus, the whole AT rich region could be covered in a single read without the trouble of re-assembly. Purified mtDNA (isolated as described above) was linearized by restriction cutting at the PstI site and a ∼20-kb band was isolated and gel purified after electrophoresis. About 500 ng of gel purified linear mtDNA sample was used for library preparation using a modified 10 kb Template Preparations Protocol (PacBio, University of Washington, Seattle). Basically, the blunt hairpin SMRTBell adaptors were ligated to the repaired ends of the double-strand DNA fragments. Failed ligation products were removed by adding ExoIII and ExoVII exonucleases. The attached templates were further purified with 0.5× and 0.45× bead-washes for sequencing. We conducted the sequencing reactions on a PacBio RSII system using one SMRT cell for each genome. All libraries were sequenced using P4/C2 chemistry. mtDNA samples from mt:ND2del1 + mt:CoIT300I, and the recombinants were run with 120 min movie time and ATP6[1] sample was run with 180 min. Raw reads were analyzed following either HGAP (for de novo assembly) and BLASR protocol (for resequencing) in SMRT Portal 2.2. The coverage for each samples ranged from 1000× to 35,000×.

When only the coding region of some genomes were sequenced, two long-range PCR reactions using Expand Long Template PCR system (Roche) was performed: mt186–mt7502, mt6905–mt14797 with the following program: 1 cycle of 93°C for 3 min, 30 cycles of 93°C 15 s, 50°C 30 s, 60°C 8 min, and 1 cycle of 60°C for 10 min. Primers were designed all around the coding region (Supplementary file 2) for sequencing by QuintaraBio (Albany, CA).

References

  1. 1
  2. 2
    Small repeated sequences and the structure of plant mitochondrial genomes
    1. C André
    2. A Levy
    3. V Walbot
    (1992)
    Trends in Genetics 8:128–132.
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
    Genetic analysis of systematic mitochondrial heteroplasmy in rabbits
    1. D Casane
    2. N Dennebouy
    3. H de Rochambeau
    4. JC Mounolou
    5. M Monnerot
    (1994)
    Genetics 138:471–480.
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
    Mitochondrial genetics IX: a model for recombination and segregation of mitochondrial genomes in Saccharomyces cerevisiae
    1. B Dujon
    2. PP Slonimski
    3. L Weill
    (1974)
    Genetics 78:415–437.
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
    Molecular characterization of a repeat element causing large-scale size variation in the mitochondrial DNA of the sea scallop Placopecten magellanicus
    1. J La Roche
    2. M Snyder
    3. DI Cook
    4. K Fuller
    5. E Zouros
    (1990)
    Molecular Biology and Evolution 7:45–64.
  25. 25
  26. 26
  27. 27
    Sequence, organization, and evolution of the A+T region of Drosophila melanogaster mitochondrial DNA
    1. DL Lewis
    2. CL Farr
    3. AL Farquhar
    4. LS Kaguni
    (1994)
    Molecular Biology and Evolution 11:523–538.
  28. 28
  29. 29
    Heteroplasmy in the mtDNA control region of sturgeon (Acipenser, Huso and Scaphirhynchus)
    1. A Ludwig
    2. B May
    3. L Debus
    4. I Jenneckens
    (2000)
    Genetics 156:1933–1947.
  30. 30
  31. 31
  32. 32
    Induction of mitochondrial DNA heteroplasmy by intra- and interspecific transplantation of germ plasm in Drosophila
    1. ET Matsuura
    2. SI Chigusa
    3. Y Niki
    (1989)
    Genetics 122:663–667.
  33. 33
  34. 34
    JSTOR: The American Naturalist
    1. HJ Muller
    (1932)
    118–138, JSTOR: The American Naturalist, Volume 66.
  35. 35
  36. 36
  37. 37
    Molecular population genetics of mtDNA size variation in crickets
    1. DM Rand
    2. RG Harrison
    (1989)
    Genetics 121:551–569.
  38. 38
    Somatic segregation, recombination, asymmetrical distribution and complementation tests of cytoplasmically-inherited antibiotic-resistance mitochondrial markers in S. cerevisiae
    1. GH Rank
    2. NT Bech-Hansen
    (1972)
    Genetics 72:1–15.
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
    Homologous recombination in Escherichia coli: dependence on substrate length and homology
    1. P Shen
    2. HV Huang
    (1986)
    Genetics 112:441–457.
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52

Decision letter

  1. Jodi Nunnari
    Reviewing Editor; University of California, Davis, United States

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for sending your work entitled “Selections that isolate recombinant mitochondrial genomes in animals” for consideration at eLife. Your article has been favourably evaluated by Detlef Weigel (Senior Editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors. One of the three reviewers, David Rand, has agreed to share his identity.

The manuscript describes a Drosophila animal model that can generate genetic exchange between mitochondrial genomes, which the reviewers felt was fully supported by evidence. The reviewers also unanimously agreed that the development of a system that overcomes the existing mtDNA haplotype “barrier” represents a significant advance in the field. However, the mechanistic basis of the genetic exchange observed was not sufficiently elucidated. In addition, the reviewers noted that the genetic exchange events observed in the studies were rare and the frequency of such events was only observed under conditions of strong selection, raising the concern that the system does not inform the mechanism of native events. On a related note, the reviewers also noted that that most of the recombination observations reported in the manuscript have been previously published, which was not adequately acknowledged. Given these points, it was decided that the advance lies in the development of a workable system for studying mtDNA genetic exchange, which could be used, for example, to map regions that are causative of disease in an animal by the community. As such, the consensus was to ask the authors to revise their study as a “Tools and Resources” manuscript.

Comments on mechanistic aspects of the work:

Data clearly support the appearance of mtDNA haplotypes containing tracts of sequence from the two parental mtDNA molecules following selection. However, whether these haplotypes are products of homologous recombination is not sufficiently clear. For example, the authors have not definitively excluded gene conversion as a possible explanation for their results. The length of the putative crossover tract is not itself sufficient to distinguish between gene conversion and true crossover events. Accordingly, the authors should revise their title to refer to these events more generally as genetic exchange.

There are well-established statistical tests for gene conversion that should be applied, especially considering that in some cases the putative recombinant haplotypes have already been sequenced.

The authors have not established that parental haplotypes are in close proximity required for recombination (such as by imaging heteroplasmic Drosophila nucleoids labelled by FISH), despite the ability to do so as demonstrated in previous manuscripts from this group.

The authors have not provided any evidence that mtDNA of Drosophila is capable of forming true Holliday junctions, and failed to discuss relevant literature on the structure of mtDNA replication intermediates.

Crossover points in putative recombinants are often in the vicinity of the heavy or light strand replication origins or the mTERF replication pausing site. How did the authors exclude the possibility that novel haplotypes may arise via strand-switching of stalled forks? Can other restriction sites be used to induce DBSs that are further from these sites?

The authors' assertion that the use of restriction-targeted enzymes actually informs us about the mechanism of the unassisted recombination events is not necessarily true. Introduction of the strand breaks may also result in exchange, but this correlation does not prove that the same mechanisms actually apply under the restriction digestion-free situations. As such, the authors should revise this claim. In addition, the authors also imply that recombination is somewhat common, and that the selection provides a means to “amplify” and detect these events. Have the authors considered that the negative physiological consequences of mitochondria under these selection regimes may be required to generate the recombination positive molecules?

Comments on the novelty and presentation of the work:

The reviewers felt that while data clearly demonstrate that selection is critical for exchange in the fly, most of the observations on exchange have been previously published. Additionally, these observations have been made before, in hybrid systems, but used amplification-based methods which may have induced artifacts (e.g. Ujvari, B et al., 2007, Biol. Lett., 3, 189-192; Guo et al., 2006, Genetics, 172, 1745-1749), but were robustly demonstrated in the bi-uniparental mitochondrial systems (e.g.. Ladoukakis et al., 2011, Mol. Biol. Evol., 28: 1847-1859), at a very low rate. On a related note, the reviewers also request that the Discussion be revised, specifically to eliminate overstatements of significance. The reviewers noted that a significant fraction of the Discussion was dedicated to a potential explanatory mechanism of exchange that may not actually be occurring in a native system as the double-strand break model engineered with restriction enzymes is highly artificial, and may not actually be analogous to the natural mechanism.

Minor comments:

The discussion of the results in the beginning of the subsection “Screening for recombination without selection” (Results) is confusing. In their Ma et al., Nature Genetics 2014 paper, the authors write: “Lines in which the temperature-sensitive genome was paired with the mt:ND2del1 genome also showed an early decrease in the abundance of the temperature-sensitive genome. However, the decrease in abundance did not continue to 0% but asymptotically approached ∼8%.” Does this observation mean that there actually is selection at the mtDNA level under the “selection free” scenario? If true, this implies that these events are only observable under scenarios of near-lethal conditions, not simply under selection.

The paragraph “If it were to occur […] degenerative changes of this genome” is somewhat overstated. Given the current estimated rates of mutation, any two mtDNAs would differ by only a single SNP.

The authors overstate the results of this work (for example, in the sentences “The finding means that recombination […] mitochondrial disease” and “We suggest that recombination […] mitochondrial disease mutations”. The events observed occur only under extreme selection regimes, and between highly divergent molecules.

In the subsection “The purpose of homologous recombination” of the Discussion, the authors state: “… rabbit (Oryctolagus cuniculus) mtDNA have repeated 153 bp motifs in the vicinity of the replication origin of H strand…”. Would slip-strand mispairing not fully explain this pattern? This activity is quite common in animal mtDNA.

In the same subsection, the authors state that: “recombination among sibling molecules may be prevalent”. Why claim recombination is prevalent? The data in the Results show the opposite.

In your manuscript, you suggest that: “the main role of recombination in mitochondria is homology dependent DNA repair, which may be especially important in light of high level of DNA damage inflicted on the mtDNA by its oxidative environment”. There is growing skepticism regarding the amount of ROS damage that mtDNA experiences (example: PLoS Genet. 2014 Feb; 10(2): e1003974).

Please clarify the following note (in the subsection “Recombination upon introduction of a single DSB”): “genome was completely eliminated after 18 generations (not shown)”. Why “data not shown”? Why not show data?

Figure 1B: The size of the band in Figure 1B looks bigger than 21.2kb. The literature (and this paper) indicate that D. melanogaster mtDNA is ∼19kb. Clarify this discrepancy. D. melanogaster mtDNA size does vary, but >21.2 kb is an unusually large size.

Figure 2B: Where are the other digestion products on the gel. There should be a ∼3kb band produced by this digest. It looks like the probe does not overlap this smaller band. It would be reassuring that all products from the digest are demonstrated on the Southern blot. Can these blots be re-probed with another probe? A PCR product for the targeted region to the right of the Xhol site would confirm this.

Discussion: Can some estimates be made about the frequency of exchange in the heteroplasmic cultures? A big issue in this new finding is just how frequent this is in nature. The Discussion does a good job of clarifying that recombination will not happen without paternal leakage, mtDNAs being of different haplotypes, and being in the same organelle or ‘nucleoid’ of mtDNAs. There are two issues of interest and impact here. First, population geneticists might want to know: how much recombination is enough to purge deleterious mutation accumulation? Second, molecular geneticists will want to know how often these occur for disease issues or for constructing novel genotypes for experimental work. This frequency estimate is a rough guess, but that would increase the impact.

https://doi.org/10.7554/eLife.07247.012

Author response

The manuscript describes a Drosophila animal model that can generate genetic exchange between mitochondrial genomes, which the reviewers felt was fully supported by evidence. The reviewers also unanimously agreed that the development of a system that overcomes the existing mtDNA haplotype “barrier” represents a significant advance in the field. However, the mechanistic basis of the genetic exchange observed was not sufficiently elucidated. In addition, the reviewers noted that the genetic exchange events observed in the studies were rare and the frequency of such events was only observed under conditions of strong selection, raising the concern that the system does not inform the mechanism of native events. On a related note, the reviewers also noted that that most of the recombination observations reported in the manuscript have been previously published (see below for details), which was not adequately acknowledged. Given these points, it was decided that the advance lies in the development of a workable system for studying mtDNA genetic exchange, which could be used, for example, to map regions that are causative of disease in an animal by the community. As such, the consensus was to ask the authors to revise their study as a “Tools and Resources” manuscript.

This overview suggests that “the mechanistic basis of the genetic exchange observed was not sufficiently elucidated”. We acknowledge that there is much to learn about the mechanism of recombination. However, we report unprecedented advances documenting and characterizing recombination in animal mitochondria and fully document our findings.

The overview also notes that the recombinants were rare and observed “under conditions of strong selection”. Recombination has been studied for decades by selection for recombinant genotypes and in many circumstances the events are rare (e.g. somatic recombination leading to loss of heterozygosity and cancer progression is fortunately particularly rare). For reasons that are not clear to us, our application of a genetic approach raises a heretofore never mentioned concern “that the system does not inform the mechanism of native events”. Is this an indictment of the field of genetics or just our work? We fail to understand the logic. Neither rarity nor detection by genetic selection should disqualify recombination from being a natural event. Recombination is often a rare event, but nonetheless has profound consequences and is thought to be an essential to avoid genomic deterioration by “Muller’s ratchet” (Muller, 1932). Furthermore, selection is usually thought to act on pre-existing events, which should qualify as natural, after all Darwin called it natural selection. Nonetheless, the possibility that stress would induce recombination is real, but it adds to, rather than diminishes, the biological impact of the process. Recombination among fully functional genomes can generate non-allelic events that cause mutageneic rearrangements, so it might be something to avoid in the absence of problems. In contrast, recombination can restore function to genomes with distinct defects and so might be productively induced under circumstances of insufficient function or DNA damage. We consider this further below in responding to a specific comment in the review.

The reviewers “noted that most of the recombination observations reported in the manuscript have been previously published, which was not adequately acknowledged”. Firstly, the reviewers mention papers that we did reference. While there are many additional and unreferenced papers in the literature that discuss recombination in animal mitochondria, we believe that we appropriately represented the controversial status of the literature. Secondly, the review does not point out any observation that anticipated a key finding of this manuscript. The only reported observation that has a parallel in the literature is one that we presented for the express purpose of showing a parallel. We showed that heteroplasmic strains can faithfully propagate two genomes for multiple generations. Like a similar result in mice, our analysis failed to detect recombinants in flies without selection. It was presented to suggest potential parallels between systems and to highlight the importance of the special tools that we developed to take the analysis of recombination to a new and unprecedented level. The central approaches, the outcomes and the characterizations that are presented in this paper are unique. We discuss the specific papers below, but note here that they show little consensus about the existence of recombination in animal mitochondria and have little to say about its characteristics.

Comments on mechanistic aspects of the work:

Data clearly support the appearance of mtDNA haplotypes containing tracts of sequence from the two parental mtDNA molecules following selection. However, whether these haplotypes are products of homologous recombination is not sufficiently clear. For example, the authors have not definitively excluded gene conversion as a possible explanation for their results. The length of the putative crossover tract is not itself sufficient to distinguish between gene conversion and true crossover events. Accordingly, the authors should revise their title to refer to these events more generally as genetic exchange.

Recombination is a genetic term that refers to the appearance of non-parental combinations of input alleles in progeny organisms or cells. This is what is reported in the manuscript. Like recombination, gene conversion has a genetic definition: when the products of meiosis reveal allele frequencies that are inconsistent with Mendel’s rules, one allele is said to be converted to another. Gene conversion represents a type of recombination, i.e. nonparental combinations of alleles are found in progeny. For example, Chen et al. (Nature, 2007) state: ‘In eukaryotes, gene conversion constitutes the main form of homologous recombination that is initiated by DNA double-strand breaks.’ Additionally, according to Molecular Biology of the Cell. 4th edition (Bruce Alberts et al): ‘In the process of gene conversion, DNA sequence information is transferred from one DNA helix that remains unchanged (a donor sequence) to another DNA helix whose sequence is altered (an acceptor sequence). There are several different ways this might happen, all of which involve the following two processes: (1) a homologous recombination event that juxtaposes two homologous DNA double helices, and (2) a limited amount of localized DNA synthesis, which is necessary to create an extra copy of one allele.’ The fact that recombination is a term that encompasses gene conversion is nicely highlighted by the fact that Robin Holliday’s presentation of his model for recombination (1964 Genet. Res., 5, pp. 282–304) was entitled “A mechanism for gene conversion in fungi”. We thus do not agree with the reviewers’ concern about revising the title.

We are not quite sure what is meant by the comment that “whether these haplotypes are products of homologous recombination is not sufficiently clear”. Again, recombination is defined by recovery of non-parental recombinant progeny and we demonstrate this at a phenotypic and genetic level. Homologous recombination refers to events in which the products are joined in regions of homology. The precise allelic junctions in the multiple recombinants isolated show that these recombinants are the result of homology directed exchange. We believe that this is clear. If the issue is that we have not defined the detailed enzymology and mechanism, we agree, but do point out that in those systems where recombination has been studied for decades, mechanisms were worked out over many years by many laboratories, and still recent work is refining our views of mechanism in even the most studied systems. Additionally, we emphasize that recombination is not defined by a unique mechanism. Indeed, it occurs by multiple mechanisms. For example, bacteriophage Lambda recombination, which is one of the pillars of modern genetic engineering by recombination, recombineering (e.g. Pines et al., 2015), does not use the same mechanism as E. coli RecA mediated recombination. Lambda relies heavily on a processive exonuclease that produces extended tracks of single stranded DNA that interact, while RecA mediates pairing between duplex molecules. We argue that we have made unprecedented advances in documenting and characterizing animal mitochondrial recombination and acknowledge that the system is in its infancy and there is still more to learn.

The review suggests that an example of the insufficiency of the work is the failure to definitively exclude gene conversion as a source of the recombinants. This focus on gene conversion appears to be based on misunderstanding that gene conversion is a mechanism. As mentioned above, gene conversion is defined by change of one allele into another, and it can occur by multiple mechanisms. For example, translation of Holliday junction results in heteroduplex formation with mismatches at points of sequence discordance: Mismatch repair can rectify the mismatches, most frequently by excision and repair, resulting in the loss of one allele (Holliday, 1964). Alternatively, in double-strand break induced recombination, heteroduplex can also be produced in other ways and similarly be repaired by mismatch repair mechanisms (Kobayashi, 1992). Additionally, a synthesis dependent strand annealing mechanism (McMahill et al., 2007) provides yet another mechanism’s generating heteroduplex. In sum, gene conversion is not defined by mechanism, and there exist multiple mechanisms leading to conversion. Furthermore, the mechanisms do not have a simple one to one relationship with genetic outcomes, with one producing recombination and the other gene conversion. Importantly, all events leading to recombinant genomes are appropriately referred to as recombination. In reviewing our text, we realized that in several places we unintentionally suggested the dichotomy between recombination and gene conversion that we now challenge. We have re-written several paragraphs and changed a section subtitle in the Discussion to avoid this.

There are well-established statistical tests for gene conversion that should be applied, especially considering that in some cases the putative recombinant haplotypes have already been sequenced.

There have been detailed characterizations of gene conversion in particular biological settings, but the measured parameters do not result in a “test for gene conversion” as the values are largely dictated by conversion track length, a parameter that varies greatly with context and specific mechanism underlying the conversion process. For this reason, we do not feel that the suggested analysis would be valid. Nonetheless, it is true that the commonly studied forms of gene conversion have relatively short track lengths and we have considered whether our results are congruent with other cases of gene. The mean tract length for meiotic gene conversation has been estimated by several experimental studies for organisms including yeast, human and Drosophila to be in the range of 350-2000bp (Padhukasahasram and Rannala, 2013, Chen et al, 2007). In Drosophila, it is estimated to be ∼352bp (Hilliker et al., 1994). The recombinants we isolated between D. mel and D. yak all involve at least 1kb of continuous exchange. Furthermore, the recombinant we isolated between the temperature sensitive mutant and ATP6[1] showing exchange of a continuous 7.5kb fragment, which is well above the mean tract length of nuclear meiotic gene conversion (using the data and model of Hilliker et al., 1994 for meiotic nuclear events, the probability of a gene conversion track of this length would be less than 10-9). More recently, by following more lines heteroplasmic for ATP6[1] and temperature sensitive genomes, we recovered two more recombinants (this finding and sequence data have been added to the manuscript as part of Figure 2–figure supplement 1), both involved exchange of a long stretch of continuous sequence (5-12kb), again well outside of the range characteristic of most gene conversion systems. However, since different mechanisms of homologous recombination produce products that are conservative exchange events as well as products that convert one allele to another, we suggest that efforts to distinguish these processes without the genetic test that defines them is not terribly important. Instead, the important aspects that we report are the features of the recombination products. We have re-written the Discussion to maintain a more direct connection to the data provided and believe that these data provide meaningful insights into processes involved in generating the recombinants that we describe.

The authors have not established that parental haplotypes are in close proximity required for recombination (such as by imaging heteroplasmic Drosophila nucleoids labelled by FISH), despite the ability to do so as demonstrated in previous manuscripts from this group.

We do not consider that it is necessary to show that parental haplotypes are in close proximity required for recombination, because we documented that they did recombine. We have never imaged heteroplasmic Drosophila nucleoids labelled by FISH in any of our previous manuscripts. We also note that the interaction that gave rise to the recombinants might have been transient, and might have occurred at any time during the life cycle of the flies. Lastly, at this point in time, we believe the imaging criterion that would define the proximity for recombination is not practical and we do not see how its absence calls into question the existence of the recombinants that we have characterized.

The authors have not provided any evidence that mtDNA of Drosophila is capable of forming true Holliday junctions, and failed to discuss relevant literature on the structure of mtDNA replication intermediates.

These junctions have seldom been visualized outside of their generation in in vitro reactions and the lack of direct demonstration in most systems has not prevented study of recombination. Additionally, Holliday junctions are not used in all forms of recombination and the structure can also form without generating a recombinant. We feel that a demonstration of Holliday junctions is impractical at this stage and largely tangential to the advances made in the paper.

We were reticent to discuss the structure of mtDNA replication intermediates because there is little consensus in Drosophila even about major aspects of the mechanism of replication. For instance, a recent paper by Joers and Jacob, 2013, argues that replication is unidirectional (starting within the non-coding region and proceeding in the direction of the rRNA locus) and mainly strand-coupled (as almost all intermediates that were detected by 2D electrophoresis were found to be fully double-stranded). However, their data disagree with the previously proposed strand-displacement model for mtDNA replication, which is based on transmission electron microscope imaging of replicative forms (e.g. Goddard and Wolstenholme, 1978). We remain of the opinion that there is little that would be gained from an extensive discussion of potential interplay with replication at this time, but in response to the next question we have added specific considerations and tried to make more clear that we have not excluded interactions with replication.

Crossover points in putative recombinants are often in the vicinity of the heavy or light strand replication origins or the mTERF replication pausing site. How did the authors exclude the possibility that novel haplotypes may arise via strand-switching of stalled forks? Can other restriction sites be used to induce DBSs that are further from these sites?

Our manuscript does not eliminate particular mechanisms, and we do not exclude strand-switching mechanisms. Indeed, we suggested “break induced replication” or BIR as a possible mechanism. This mechanism posits that breaks induce invasion of homologous duplex, followed by replicative extension of the invading strand, that is strand switching (see Arand et al., 2013). Although, the name emphasizes double-strand breaks and we advanced this particular idea because of data obtained in double restriction enzyme selection, the mechanism also appears to be used to repair stalled forks (Constantino et al., 2014).

The suggestion that we examine possible coincidence of recombination junctions with special sequences such as replication origins is interesting, but without tests of causality it is not clear that much meaning can be attached to it. Nonetheless, we examined this issue. Most of our sites of recombination have not been localized near an origin of replication, but this might be influenced by rapid divergence of sequences around the replication origins so that homology is reduced. None of the most accurately mapped sites of recombination, those between D. melanogaster and D. yakuba, map to any of the mentioned sites. However, the location of sites of recombination in these cases might be special because sites are restricted by the frequent interruptions of homology in this pairing. One junction of the recombinant between ATP6[1] and the temperature sensitive genome was mapped within a 531 bp region (1154-12085) that encompasses a mTERF binding site downstream of ND1. Additionally, one of two other newly isolated recombinants between these genomes was mapped within a 631 bp region (5978-6619) that encompasses the other site of mTERF binding at 6314. While possibly consistent with a role of replication fork stalling near mTERF sites in triggering recombination in two cases, the correlation is weak and several steps of inference are involved in making such a suggestion. We have added a short discussion in the manuscript.

The authors' assertion that the use of restriction-targeted enzymes actually informs us about the mechanism of the unassisted recombination events is not necessarily true. Introduction of the strand breaks may also result in exchange, but this correlation does not prove that the same mechanisms actually apply under the restriction digestion-free situations. As such, the authors should revise this claim. In addition, the authors also imply that recombination is somewhat common, and that the selection provides a means to “amplify” and detect these events. Have the authors considered that the negative physiological consequences of mitochondria under these selection regimes may be required to generate the recombination positive molecules?

The issue that restriction enzyme cutting might induce a distinct process is a reasonable caution and we made changes to in the discussion to accommodate this criticism. Additionally, we have deleted the word natural in the Abstract, since it had inappropriately implied that the restriction cuts were directly stimulating the normal event.

We totally agree that stress might induce recombination, but it is difficult to test. We did attempt to do so as follows. For the ATP6[1] and temperature sensitive selection, the five lines that we followed at 29°C were also followed at 22°C (Populations in each line were split into half at generation 1 and kept at either 29 or 22°C for the following generations). At 22°C there is little selection against the temperature sensitive genome and presumably little stress as the flies are healthy. At generation 5 (when the average abundance of the ATP[1] genome had dropped to about 10% in most lines), populations growing at 22°C were transferred back to 29°C and kept there for the subsequent generations. We had reasoned that if the recombinants were pre-existing, then one would expect the same outcome whether the selection was imposed from the beginning or only after the decline the ATP6[1] genome. Perhaps supporting such a notion, none of the lines that were initially at 22°C produced survivors, suggesting the possibility of induced recombination by negative physiological consequences at high temperatures. However, as we considered events during the selection, we realized the finding might be explained without assuming induced recombination. Because there is huge growth of the population during the early generations, we can only analyze a small fraction of the potential progeny (we cull the population at each generation). Hence, the temperature might also have had an affect on the efficiency of recovery of preexisting recombinants by giving them an early selective advantage so that they were less likely to be culled in the early generations. Since the above experiment does not provide a definitive conclusion, we did not include it in the manuscript. Thus, the idea while perhaps weakly supported by our experiment, remains hypothetical. As emphasized in our comments above, stress-induction of recombination might add greatly to its impact. Consequently, it is a factor we do not want to ignore but will continue to investigate.

The reviewers felt that while data clearly demonstrate that selection is critical for exchange in the fly, most of the observations on exchange have been previously published. Additionally, these observations have been made before, in hybrid systems, but used amplification-based methods which may have induced artifacts (e.g. Ujvari, B et al., 2007, Biol. Lett., 3, 189-192; Guo et al., 2006, Genetics, 172, 1745-1749), but were robustly demonstrated in the bi-uniparental mitochondrial systems (e.g.. Ladoukakis et al., 2011, Mol. Biol. Evol., 28: 1847-1859), at a very low rate. On a related note, the reviewers also request that the Discussion be revised, specifically to eliminate overstatements of significance. The reviewers noted that a significant fraction of the Discussion was dedicated to a potential explanatory mechanism of exchange that may not actually be occurring in a native system as the double-strand break model engineered with restriction enzymes is highly artificial, and may not actually be analogous to the natural mechanism.

The reviewers challenged the novelty of the work because they felt that most of the observations on exchange have been previously published. Except for Guo et al., 2006, the work they quote is work that we reference. Here we examine these reports to consider whether they demonstrate that our observations “have been previously published”.

The Larsson group presents its conclusion in the title ‘No recombination of mtDNA after heteroplasmy for 50 generations in the mouse maternal germline’. Clearly this paper did not report genetic selections for organisms carrying characterized recombinant genomes. Indeed, just the opposite, it argued that there is no recombination. As mentioned above, we did report a parallel analysis in Drosophila with a similar outcome. We reported this experiment because we suspect that that there is little difference between systems other than the fact that we developed uniquely powerful tools that allowed us to go beyond this limited experiment to do heretofore never reported experiments directly selecting and isolating animals with recombinant mitochondrial genomes.

A. Sato et al., 2005, in contrast with S. Berlin et al., 2004, reports mitochondrial recombination in mice. Sato et al. made heteroplasmic tissue culture cells and heteroplasmic mice and retrieved mitochondrial genomes from the heteroplasmic lines by cloning. Many independent clones were sequenced. In one of three experimental settings, they identified two rare clones having short stretches of sequence with SNIP markers from the complementary genotype. This paper nicely shows rare somatic recombinant molecules that include between 36 bp to 120 bp of transferred sequence. Notably, the sequences transferred are short. There is no demonstration of germline events, no transmission of the recombinants and no enrichment of the recombinant. Furthermore, the rare events they detect have no demonstrated functional consequence. Outside of an evolutionary time scale, we consider this the clearest published experimental test for recombination in animals prior to our work, but none of the reported experiments resemble ours and their results are more limited and different.

H. Fukui et al., 2009 made heteroplasmic tissue culture cells and mice and used PCR reactions to support claims for rare recombinant molecules provoked by expression of ScaI restriction enzyme. Unfortunately, PCR reactions are themselves recombinogenic and both of the above papers show that this gives rise to artifactual results. In addition to uncertain validity, this study only detected genomes with big deletions close to the cut site and all but one of the apparent deletions was interpreted as intramolecular. Instead of full genomes resulting from exchange of sequence between the parental genomes, rearranged (deleted) sequences that were recovered by PCR support this recombination event. There is no evidence of germline transmission, no evidence of functional contribution, and uncertainty about the contribution of multiple restriction sites to the generation of the reported fragments. While this report might overlap in its intent, it does not report observations comparable to those in our paper.

B. Ujvari et al., 2007 reports an analysis of the mitochondrial genome sequence of wild lizards captured at a boundary between two different populations of lizards, which have mitochondrial genomes marked by many sequence differences. A single lizard was found to have a genome with a somewhat intermediate sequence. Analysis of the distribution of polymorphisms in this intermediate genome suggests an evolutionary connection to the two dominant haplotypes, where one segment was highly related to one haplotype and the other segment was related to the other haplotype. This strongly suggests that a recombination event occurred at some time in the evolution of the intermediate haplotype. However, numerous sequence differences indicate that the dominant existing haplotypes were not the immediate parents of the intermediate haplotype. Thus, the study does not identify the parental genomes, but infers that a recombination event at some time in evolution (apparently long ago) was the source of the exceptional genome. Such population studies of wild populations, while of value and supporting the idea that some genetic exchange exists, bare little resemblance to our analysis, and do not report observations that we have made.

The Guo et al. paper has some parallels to B. Ujvari et al., 2007, but we feel is much less definitive. The Guo paper examines the sequences of carp that are used in farmed fish production in China. An interesting set of cross-species crosses generate hybrid fish with useful attributes. The paper reports analysis of the mitochondrial DNA from a number of fish selected from the populations used in the final cross as well as several hybrid fish. One of the hybrid fish has a mitochondrial genome sequence unlike that of the other hybrids or of the presumed parents. An argument is presented that this genome is recombinant between the two parental types. However, the relationship of the sequence to presumed parents is very inexact with many polymorphisms distinguishing the “recombinant” genome from either parent. As result, it is clear that the genome was not produced as an immediate product of recombination and given the lack of knowledge of the lineage of the fish analyzed and cross species cross used to produce the populations examined, it seems likely that this diverged genome was introduced by introgression form another strain or species.

A. Sato et al., 2005 reports another analysis of wild caught organisms. In this case, the clams that were examined exhibit an exceptional and interesting type of inheritance pattern of their mitochondrial genomes called doubly uniparental. Here the male mitochondrial genome in the sperm is transmitted to the progeny, but its contribution is kept compartmentalized and separate from that of the mitochondrial genome from the egg. The male mitochondrial genome contributes only to the mitochondria in the male germline. The separation between the female lineage and male lineage is so thorough that they evolve along different paths and have dramatic differences in sequence suggesting isolation for hundreds of thousands to millions of years. Nonetheless, some populations have male genomes with segments having a high similarity to the female genome, a finding that is taken as indication of recombination event in more recent evolutionary history. This is a very interesting biological system and the study supports rare recombination in this unusual setting. However, it implicates a recombination event in the distant past, albeit more recent than the original separation between the paternal and maternal lineages. As in the above cases, we disagree that this paper can be presented as evidence that “most of the observations on exchange have been previously published”.

The cited work do not report experiments analogous to ours, or show findings analogous to ours. If the reviewers’ point is that there was prior evidence for recombination in mammals, we do not dispute this. Indeed, we described the literature as controversial, so of course it contains papers arguing for both interpretations. Finally, we would like to declare that, to our knowledge, there has been no previous case in animals in which recombinant mitochondrial genomes have been selected in animals and progeny recovered with a dominant recombinant genome.

Minor comments:

The discussion of the results in the beginning of the subsection “Screening for recombination without selection” (Results) is confusing. In their Ma et al., Nature Genetics 2014 paper, the authors write: “Lines in which the temperature-sensitive genome was paired with the mt:ND2del1 genome also showed an early decrease in the abundance of the temperature-sensitive genome. However, the decrease in abundance did not continue to 0% but asymptotically approached ∼8%.” Does this observation mean that there actually is selection at the mtDNA level under the “selection free” scenario? If true, this implies that these events are only observable under scenarios of near-lethal conditions, not simply under selection.

This is a very good point, and there is likely some truth in the inference that is made. In the line heteroplasmic for mt:ND2del1 and mt:CoIT300I genomes, there appears to be is selection against both genomes to achieve the balanced ∼8% ratio. However, this balanced heteroplasmic line behaved, if not better, at least as well as the flies homoplasmic for wild type mitochondrial genome (See supplementary Table 1b in Ma et al., Nature Genetics 2014). Therefore, there is no functional advantage to the organism incurred by generating a wild type recombinant mtDNA under such a condition. Thus, from an organismal perspective the condition appears to be ‘selection free’, but there does seem to be selection acting on the mitochondrial genomes to maintain the ratio (as pointed out). Because there is some level of selection ongoing in this heteroplasmic line, we have changed the title of the first section of the results as the previous wording had implied otherwise.

Why wouldn’t this selection on the mitochondrial genomes act to promote formation of recombinants during this “selection free” situation? One reason is that recombination might not occur at the same stage of the life cycle as the purifying selection, which is limited to oogenesis. Indeed, in order to give functional genomes an advantage over less functional genomes during oogenesis, they must be autonomous and presumably in different mitochondria, while recombination requires that they reside in the same mitochondria. In accord with the idea that different types of selection can occur at different times, we find that replicative drive imparts a selective advantage at stages other than oogenesis (Ma and O’Farrell, submitted). Hence, the likely answer to this vexing issue is that purifying selection and recombination may not happen at the same time.

The paragraph “If it were to occur […] degenerative changes of this genome” is somewhat overstated. Given the current estimated rates of mutation, any two mtDNAs would differ by only a single SNP.

Apparently our wording suggested more than we intended. We have changed it.

We do agree with the reviewers that more recent and seemingly more definitive measures have suggested rather modest frequencies of SNPs (e.g. Kennedy et al., 2013) compared to some suggestions in the literature. However, this does not preclude important actions of recombination either somatically or in the germline. There are strong data indicating that deletions of mtDNA increase in incidence with age, that they accumulate to high levels in affected cells, and that the end points of the deletions tend to lie in regions of local homology suggesting that recombination contributes to their formation (e.g. Solignac, 2004; S. Mita et al., 1990; Bacman, Williams, and Moraes, 2009). Additionally, direct analysis of mtDNA from individual intestinal crypts shows that these rapidly dividing cells accumulate a few SNIPs to relatively high abundance and likely carry many more at lower levels (Taylor et al., 2003).

The authors overstate the results of this work (for example, in the sentences “The finding means that recombination […] mitochondrial disease” and “We suggest that recombination […] mitochondrial disease mutations”. The events observed occur only under extreme selection regimes, and between highly divergent molecules.

For the first passage, the concern that we made statement about normal processes from an experiment that involves a perturbation applies, and we have changed it. We have altered the wording to make clear that it is a possibility but not a conclusion from the results.

Regarding the second sentence, it is not a conclusion. It is a suggestion that recombination will be influential as it is throughout biology, and we do not regard this as an overstatement.

In the subsection “The purpose of homologous recombination” of the Discussion, the authors state: “… rabbit (Oryctolagus cuniculus) mtDNA have repeated 153 bp motifs in the vicinity of the replication origin of H strand…”. Would slip-strand mispairing not fully explain this pattern? This activity is quite common in animal mtDNA.

Thanks for pointing this out. According to the literature, slipped-strand mispairing typically occurs in repeated motifs of 1-10 bases. Therefore, it seems an unlikely explanation for deletion or addition of the longer repeats present in D. melanogaster mitochondrial regulatory region (∼350bp or 450bp), or the 153 bp repeats in rabbit. Nonetheless, we have changed our wording in the Discussion to make it clear that recombination is a possible explanation, not that it is the known explanation, for the shifts in repeat number.

In the same subsection, the authors state that: “recombination among sibling molecules may be prevalent”. Why claim recombination is prevalent? The data in the Results show the opposite.

The sentence reads, “Since recombination between sibling molecules may be prevalent…”. This is not a claim. This is a possibility, and it follows form the preceding paragraph. It also is restricted. It suggests that recombination might be prevalent between sibling molecules. Sibling molecules are the products of one replication event and are necessarily, at least transiently, co-localized, and hence not subject to all of the other barriers that might hinder interaction and recombination between two independent genomes. Consequently, there is no contradiction data from the Results, which refer to recombination between independent genomes in a heteroplasmic line. In any case, out of fear that this was not made clear by the text, we have entirely revised the presentation of these issues and left out use of the word prevalent.

In your manuscript, you suggest that: “the main role of recombination in mitochondria is homology dependent DNA repair, which may be especially important in light of high level of DNA damage inflicted on the mtDNA by its oxidative environment”. There is growing skepticism regarding the amount of ROS damage that mtDNA experiences (example: PLoS Genet. 2014 Feb; 10(2): e1003974).

True, and the skepticism appears to be founded on good data. We have re-written the discussion without such a statement.

Please clarify the following note (in the subsection “Recombination upon introduction of a single DSB”): “genome was completely eliminated after 18 generations (not shown)”. Why “data not shown”? Why not show data?

Because the data have been shown as a supplementary figure in Hill, Chen, and Xu, 2014, which was also referenced. To clarify this, we omitted the “data not shown” statement and simply referred to reference.

Figure 1B: The size of the band in Figure 1B looks bigger than 21.2kb. The literature (and this paper) indicate that D. melanogaster mtDNA is ∼19kb. Clarify this discrepancy. D. melanogaster mtDNA size does vary, but >21.2 kb is an unusually large size.

Our concern in this experiment was the presence or absence of the small doubly cut band. To optimize detection we ran a gel (1.2%) more suitable for detection of a 1.6 kb band than a 19+ kb band. Additionally, we loaded a large amount of sample extracted from whole flies. Under the circumstances, we did not consider the gel an accurate way to assess the size of the genome. On the other hand, we have the full-length sequence of this genome and this sequence is deposited in conjunction with this paper. The genome is 19,542bp (Figure 2).

Figure 2B: Where are the other digestion products on the gel. There should be a ∼3kb band produced by this digest. It looks like the probe does not overlap this smaller band. It would be reassuring that all products from the digest are demonstrated on the Southern blot. Can these blots be re-probed with another probe? A PCR product for the targeted region to the right of the Xhol site would confirm this.

Indeed, as indicated in the schematic accompanying the figure, the probe is limited to a single diagnositic band, and that is why the other bands are not visible. We disagree with the idea that it would be reassuring to see all the bands. It would be confusing. The point of this gel is not to define the distinctions in the genomes. The fully sequenced genomes do that (Figure 2). The point of the blot was to illustrate the changes in the levels of the different genomes in successive generations using a diagnostic signal and the probe was designed to provide that.

Discussion: Can some estimates be made about the frequency of exchange in the heteroplasmic cultures? A big issue in this new finding is just how frequent this is in nature. The Discussion does a good job of clarifying that recombination will not happen without paternal leakage, mtDNAs being of different haplotypes, and being in the same organelle or ‘nucleoid’ of mtDNAs. There are two issues of interest and impact here. First, population geneticists might want to know: how much recombination is enough to purge deleterious mutation accumulation? Second, molecular geneticists will want to know how often these occur for disease issues or for constructing novel genotypes for experimental work. This frequency estimate is a rough guess, but that would increase the impact.

We are conflicted about this request to estimate a frequency of recombination. While we are in better position than a reader to assess this, we cannot make a realistic estimate of frequency of a molecular event. Additionally, we strongly suspect that recombination is regulated and genetically modulated, and that a single estimate might be more misleading than informative. We can of course note the frequency with which females are fertile or fraction of vials yielding progeny when subjected to selection and we do report this. We also have noted what we feel is an enormous difference in frequency in the presence and absence of restriction enzyme cutting, perhaps an indication of range of regulation.

Here we list the problems in extrapolating to a molecular estimate:

1) We do not know the efficiency with which a recombination event can result in escape of selection: If recombination produces a single resistant (or functional) genome in the germline, what is the likelihood that we will recover it? We don’t know how fast the mitochondrial genomes were cut and then disappear, how long the germline cells can survive with mostly disrupted mtDNA, and how many copies of a recombinant genome is required in order to make the germline cell viable and capable of subsequent development. Maybe one recombinant genome from one successful recombination event is sufficient to survive the selection, but it has to be generated early during oogenesis, so it has the chance to replicate itself and then repopulate the whole egg, which often contains ∼10 million copies of mtDNA. Since most of the progeny we recovered are homoplasmic, it appears that only a single event was recovered, but for every event recovered perhaps there were many thousands that failed to make it.

2) Even if we knew the efficiency with which a recombinant could rescue, we need a denominator to turn the frequency of fertile females into a more meaningful frequency of recombination than the one we have already reported, and we do not know what this denominator is. We can estimate the number of germline stem cells (about 64 or perhaps twice of this if we include the first wave of direct developing stem cells), but we actually don’t know if recombination occurs in these cells or their progenitors. So we cannot accurately report a frequency on a per cell basis. Furthermore, we do not know the number of mitochondria in a cell, so we cannot change this to recombination frequency per genome.

3) Different selections are likely to result in different frequencies of recovered recombinants. The survivors of restriction enzyme selection are homoplasmic reflecting a strong selection against the starting genomes that acts to completion in a single generation with little opportunity for a slow amplification of the resistant genomes. In contrast, the temperature sensitive selection “against” the mutant allele of mt:CoI is actually a selection for the function of the wild-type allele and even a low abundance of functional recombinant genome can rescue survival of heteroplasmic line and gradually accumulate over multiple generations.

As a result of these uncertainties, we feel that we cannot offer a frequency. We do, however, point out here that when we induced complementary DSBs, which gave an immediate and strong selection for homoplasmic recombinant progeny, most of females had high fertility. This means that many of the germline stem cells had recombinant genomes. This requires that many recombination events occur in most females. Furthermore, preliminary data suggests that rescue of fertility by having ∼2.5% of resistant genome upon cutting, is inefficient, meaning that many recombinant genomes are required for in each stem cell to produce progeny. Given that this was in response to DSBs, we suggest that it is an indication that recombination pathways will efficiently repair DSBs. Indeed, it is likely that the recombinants observed represent a small fraction of amount of homologous repair, since the sibling genomes will presumably be the more available partners for repair but will not contribute restriction resistant genomes.

At the other extreme, when the no restriction cutting was used in the selection (in the combination of ATP6[1] and the temperature sensitive genome), recombinants were observed in the minority of vials (3 out of 51 vials), each with a small population of females (about 40), and the recombinants appeared only after several generations. While this still represents a reasonably high likelihood of observing a recombinant in a vial, we suspect that the frequency of underlying event is considerably more than two orders of magnitude less frequent than that observed in the double restriction experiment.

We would also like to note that if one mixed two genetically marked populations of E. coli or of the yeast S. cerevisiae (which in the wild is homothallic), one is unlikely to see any significant recombination, and that the intensive use of these organisms for genetics followed the discovery of specialized approaches to manipulate them to reveal their innate capacities for recombination. We now know that mitochondria can recombine and hopefully we can uncover the factors that modulate their capacity to do so.

https://doi.org/10.7554/eLife.07247.013

Article and author information

Author details

  1. Hansong Ma

    Department of Biochemistry and Biophysics, University of California, San Francisco, San Francisco, United States
    Contribution
    HM, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  2. Patrick H O'Farrell

    Department of Biochemistry and Biophysics, University of California, San Francisco, San Francisco, United States
    Contribution
    PHO'F, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    ofarrell@cgl.ucsf.edu
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0003-0011-2734

Funding

NIH Office of the Director (GM086854)

  • Patrick H O'Farrell

Human Frontier Science Program (LT000138/2010-L)

  • Hansong Ma

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This research was supported by NIH (GM086854) funding to PHO'F. HM was supported by the Long-term postdoc fellowship from Human Frontiers Science Program. We thank Hong Xu for generously sharing unpublished information and reagents, and Michael J Palladino at University of Pittsburgh for kindly providing us flies with ATP6[1] genome. We also are grateful for support from the PacBio team, particularly Nicole Rapicavoli (Melon Park, CA), Maika Malig (University of Washington, Settle), and Roberto Lleras.

Reviewing Editor

  1. Jodi Nunnari, University of California, Davis, United States

Publication history

  1. Received: February 28, 2015
  2. Accepted: August 1, 2015
  3. Accepted Manuscript published: August 3, 2015 (version 1)
  4. Version of Record published: September 28, 2015 (version 2)

Copyright

© 2015, Ma and O'Farrell

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,998
    Page views
  • 480
    Downloads
  • 10
    Citations

Article citation count generated by polling the highest count across the following sources: Scopus, PubMed Central, Crossref.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)