Abstract
Recombination has been proposed as a possible mechanism to explain mitochondrial (mt) gene rearrangements, although the issue of whether mtDNA recombination occurs in animals has been controversial. In this study, we sequenced the entire mt genome of the megaspilid wasp Conostigmus sp., which possessed a highly rearranged mt genome. The sequence of the A+T-rich region contained a number of different types of repeats, similar to those reported previously in the nematode Meloidogyne javanica, in which recombination was discovered. In Conostigmus, we detected the end products of recombination: a range of minicircles. However, using isolated (cloned) fragments of the A+T-rich region, we established that some of these minicircles were found to be polymerase chain reaction (PCR) artifacts. It appears that regions with repeats are prone to PCR template switching or PCR jumping. Nevertheless, there is strong evidence that one minicircle is real, as amplification primers that straddle the putative breakpoint junction produce a single strong amplicon from genomic DNA but not from the cloned A+T-rich region. The results provide support for the direct link between recombination and mt gene rearrangement. Furthermore, we developed a model of recombination which is important for our understanding of mtDNA evolution.
Introduction
The mitochondrial (mt) genome is one of the most highly used sources of molecular markers for reconstructing phylogeny, both through alignment of orthologous sequences (Simon et al. 1994) and through comparison of gene position (Boore et al. 1998). Analysis of aligned sequences indicates that there are inherent biases in the number and types of nucleotide substitutions that occur, and that incorporating this information into phylogenetic analysis improves the power of the analysis (Hillis et al. 1994). To appreciate whether these patterns are general or specific to the taxon being studied, it is also important to understand the mechanism of nucleotide substitution (Bakker et al. 2000). For the same reasons, it is critical to understand the mechanism of gene rearrangement. But knowledge on the mechanism and biases of gene rearrangement has remained elusive, because they occur much more sporadically. Information on the mechanism and frequency of different types of rearrangement can be best inferred by characterizing lineages with high rates of gene rearrangement (Dowton and Austin 1999).
There are three main models proposed to explain gene rearrangement: duplication/random loss (Moritz et al. 1987; San Mauro et al. 2006), duplication/nonrandom loss (Lavrov et al. 2002), and recombination (Dowton and Campbell 2001). In the duplication/random loss model, slipped-strand mispairing or inaccurate termination during replication can cause duplication of part of the mt genome. The supernumerary gene copies are subsequently eliminated due to selection, which will either restore or alter the original gene order. The duplication/random loss model is consistent with the observed gene rearrangements in the vertebrates (San Mauro et al. 2006). However, it cannot explain gene inversions and long-range translocations, which are common in invertebrate mt genomes (Dowton et al. 2002). A defining feature of the duplication/nonrandom loss model is the duplication of the entire mt genome, resulting in a dimeric molecule linked head to tail. Such a dimeric molecule would contain two copies of transcriptional promoters. Mutation to one copy of the promoters would result in the genes under the control of the disabled promoter becoming pseudogenes; eventually, these would be eliminated. After rearrangement, genes with identical transcriptional direction will be clustered in the mt genome. However, the duplication/nonrandom loss model also cannot explain gene inversions and translocations with different transcriptional direction to genes close to the translocation point. Recombination was considered to be absent in animal mtDNA based on early studies. For example, recombinant haplotypes of mtDNA were not detected in somatic cell hybrids; sequestration of mtDNA molecules into clusters prevents physical contact of unrelated molecules and there is also evidence that excision repair activity and crossover products are absent in mammalian mitochondria (Clayton et al. 1974; Zuckerman et al. 1984; Satoh and Kuroiwa 1991). However, both direct and indirect evidence of animal mtDNA recombination has been demonstrated more recently. In 1996, the end products of homologous recombination were detected in human mitochondria (Thyagarajan et al. 1996). Lunt and Hyman (1997) provided the most convincing evidence by characterizing the end products of recombination in the nematode Meloidogyne javanica. A number of subsequent studies also provided evidence of recombination in other animal species (Awadalla et al. 1999; Ladoukakis and Zouros 2001; Hoarau et al. 2002; D'Aurelio et al. 2004; Kraytsberg et al. 2004; Ujvari et al. 2007). However, it is not yet widely accepted that mt recombination occurs (e.g., Jorde and Bamshad 2000; Kivisild and Villems 2000; Kumar et al. 2000; Parsons and Irwin 2000), while there is very little information on how frequently it occurs.
The occurrence of recombination in animal mtDNA has substantial impact on evolutionary studies that utilize mt genes. First, the traditional methods for phylogenetic analysis are based on the assumption that animal mtDNA does not recombine; ignoring recombination might mislead phylogenetic reconstruction (Rokas et al. 2003; Slate and Gemmell 2004). Second, it offers an intriguing understanding of how mt genes might rearrange. As the duplication/random loss model cannot provide a full explanation of gene rearrangements across the Metazoa, recombination might explain the gene inversions and long-range translocations that are often observed in invertebrate mt genomes (Dowton and Campbell 2001). The key component of this mechanism is the introduction of double-strand breaks (DSBs) in the mt circle, with subsequent rejoining of the broken ends. Two recent studies provided support for this mechanism: DSBs were demonstrated to play an important role in the generation of mtDNA rearrangements in mice (Bacman et al. 2009). A more recent study speculated that the presence of mtMUTS in the octocoral mt genome might promote recombination which accounts for the gene inversions observed in that genome (Brockman and McFadden 2012).
Here, we report the discovery of a minicircular mtDNA molecule in the megaspilid wasp Conostigmus sp., similar to that reported by Lunt and Hyman (1997) in the nematode M. javanica. We used a similar approach to investigate the presence of the end products of recombination: the maxicircle and minicircle. We detected such end products, although we caution that polymerase errors can produce a range of apparent recombination products. Furthermore, the possibility that the minicircle comes from nuclear copies of mt sequences (numts; nuclear pseudogenes) is eliminated in our study. We then investigate the structure of the minicircles produced and develop a model of recombination that is consistent with the sequence data. We suggest that the presence of minicircles is associated with the most highly rearranged invertebrate mt genomes (the Hymenoptera and Phthiraptera) (Shao et al. 2009), lending credibility to the notion that recombination is an important component of the mt genome rearrangement mechanism.
Results
The mt Genome of Conostigmus sp.
The entire mt genome of Conostigmus sp. is 16,315 bp long. The overall A+T content is 82.9%. Like other insect mt genomes, 13 protein-coding genes, 22 tRNA genes, 2 rRNA genes, and the A+T-rich region were identified. A series of gene rearrangements are evident relative to the putative ancestral pancrustacean mt genome (Cook 2005). Although tRNA genes are highly rearranged among wasps (Dowton et al. 2009), the protein-coding genes are highly conserved (but see Oliveira et al. 2008; Xiao et al. 2011). In Conostigmus sp., 11 tRNA and 2 protein-coding genes are rearranged (fig. 1). Cox1 and ND2 are translocated into two junctions (Cox3-ND3 and rrnL-rrnS), respectively. Interestingly, the gene orientations are almost identical to that of the ancestral organization, except a single tRNA inversion of trnW.
Mitochondrial genome organization of Conostigmus sp., compared with the ancestral pancrustacean mt genome organization (Cook 2005). tRNA genes are indicated by single-letter amino acid codes, L1, L2, S1, and S2 denote trnLCUN, trnLUUR, trnSAGN, and trnSUCN. Genes are transcribed from left to right except those indicated by underline. Gene movements, relative to the ancestral organization, are indicated with arrows.
The A+T-rich region is 1,447 bp long with an A+T content of 86.8%. It can be divided into three subsections (Sec I, Sec II, and Sec III) (fig. 2A). Sec I contains seven copies of tandem repeat A (the last one is a partial copy), nine copies of tandem repeat B, and one copy of nontandem repeat C and D. In the second subsection, several conserved elements (poly T stretch, [TA (A)]n-like stretch, TATA motif, and stem-loop structure) can be identified, which are putatively involved in the initiation of replication and transcription (Zhang and Hewitt 1997). Sec III possesses four copies of tandem repeat B and the second copy of nontandem repeat C and D. D overlaps with trnG by 12 bp.
Organizations of the entire mt genome, maxicircle, and minicircles that were initially characterized by amplification. (A) Organization of the entire mt A+T-rich region of Conostigmus sp. Different repeats are shown in different colors. PCR primer sites and orientations are indicated by arrows. Breakpoint junctions are indicated by black arrows. Regions with hatching indicate nonrepeat regions. (B) The real minicircles (L2) from different individuals and the artifactual minicircles (L1, S1, and S2) generated by PCR errors. (C) Organization of the maxicircle A+T-rich region.
Initial Characterization of the Putative Maxicircle and Minicircles of Conostigmus sp.
The presence of distinct repeats within the A+T-rich region was reminiscent of those described by Lunt and Hyman (1997) in the nematode M. javanica, in which they characterized both minicircular and maxicircular products of recombination. For this reason, we designed outward facing amplification primers (85-ATF and 85-ATR, fig. 2A) within the A+T-rich region in an attempt to amplify any minicircles that might be present in Conostigmus sp. A short extension time should preclude the amplification of the entire mt genome, while any amplicons produced from the entire mt genome would be easily distinguishable (based on their size of 16,100 bp) by agarose gel electrophoresis. Thus, any amplicons smaller than 16,100 bp would be produced from minicircles. In one reaction, four different sizes of minicircles (L1, L2, S1, and S2) were detected (fig. 3A). To test whether these putative minicircles could be amplified from a range of individuals, we amplified them using the same primers in another two individuals caught at the same location (85B and 85C). Four and three polymerase chain reaction (PCR) products (L1 was lacking in 85C) were obtained respectively, with the sizes of the various amplicons consistent with the minicircles from 85A (data for 85B-L1, 85B-L2, and 85C-L2, see supplementary fig. S1A, Supplementary Material online; data not shown for 85B-S1, 85B-S2, 85C-S1, and 85C-S2). We obtained sequence data from the most strongly amplified product (L2) from each individual, the two shorter products (S1 and S2) from 85A, and the longest product (L1) from 85B (fig. 2B). However, we were not able to purify and sequence all products from each individual, due to the low levels of some amplicons in some reactions. For example, we failed to purify and sequence L1 from 85A.
Size separations of amplicons on 1% agarose gels. (A) Putative minicircular amplicons obtained using outward-facing primers and genomic DNA (extracted from individual 85A). Primers were 85-ATF and 85-ATR. (B) Minicircular PCR artifact obtained from plasmid DNA containing the A+T-rich region of the entire mt genome using primers 85-ATF and 85-ATR. Although the size of the amplicon is similar to L2, sequencing indicated that the two amplicons were distinct. (C) Minicircular amplicon obtained from genomic DNA extracted from individual 85A using primers 85-ATF2 and 85-ATR. M, molecular weight marker (lambda DNA digested with EcoRI and HindIII).
After amplification of these putative minicircles, we attempted to amplify the maxicircle that would be formed as the product of reciprocal recombination. If reciprocal recombination was responsible for production of the minicircle, we should be able to detect a maxicircle; together, the minicircle and maxicircle sequences should precisely match the sequence of the entire mt genome. To detect the maxicircle, we designed primers that would amplify the A+T-rich region of the entire mt genome (85-MR1 and 85-36R1, see fig. 2A). Any maxicircles would be evident by the presence of shorter than expected amplicons (as they would be missing the section encompassed by the minicircle). The PCR amplification was performed and the only amplicon that was produced was one that represented the A+T-rich region of the entire mt genome. To investigate whether there were low levels of maxicircles present, we designed two internal primers (85-MR2 and 85-36R2) and performed nested PCR. In this experiment, two PCR products, representing the A+T-rich region of the entire mt genome and the maxicircle, were obtained. Sequence analysis of the maxicircle amplicon indicated that it did correspond to the A+T-rich region of the entire mt genome (but lacking the minicircle fragment), although two sites differed with the corresponding sequence of the A+T-rich region of the entire mt genome (fig. 2C). These sites were not close to the breakpoint junction. The inability to amplify the maxicircle from genomic DNA, despite its smaller size compared with the entire mt genome, might suggest that the subgenomic circles are not formed by reciprocal recombination. We attempted to design a primer that straddles the putative breakpoint junction of the maxicircle in order to perform reactions that would only amplify the maxicircle. However, this was unsuccessful due to the lack of any sequence differences between the maxicircle breakpoint junctions. These junctions occurred at two perfect (i.e., without any mismatches) microhomologies (21 bp) of nontandem repeat D, with one complete copy of repeat D kept in the maxicircle after formation. The breakpoint junction primer would therefore anneal at three places: to the two copies of repeat D in the entire mt genome and the repeat D of the maxicircle (supplementary fig. S2, Supplementary Material online).
We then carried out sequence analysis of the most strongly amplified PCR product L2 amplified from 85A, representing the minicircle. The sequence of L2 corresponded to the fragment deleted from the mt genome, based on the maxicircle sequence information. It comprised Sec II and part of Sec I and Sec III, although some variation existed when compared with the A+T-rich region of the entire mt genome. In Sec II, the main difference was the length of the “TA” repeats. In Sec III, the first copy of repeat B contained one site substitution while the last three copies were smaller; repeat B is 10 bp long in the entire mt genome, but the last three copies are 9, 7, and 7 bp in the putative minicircle. In addition, repeat D in Sec I was shorter, only containing the last 14 bp of that repeat and with one site substitution. The breakpoint junctions occurred at two imperfect microhomologies (14 bp) of nontandem repeat D (supplementary fig. S2, Supplementary Material online). This observation indicates that the minicircle and maxicircle sequences do not precisely match the entire mt genome. This suggests that the subgenomic circles are not formed by reciprocal recombination, unless the recombination occurred some time ago, with the subgenomic circles accumulating mutations since that time. Alternatively, the level of variation might be consistent with the notion that direct, precisely matching repeat sequences are not necessary for the repair of DSBs (Mita et al. 1990; Samuels et al. 2004; Srivastava and Moraes 2005; Bacman et al. 2009).
We then compared the sequence of the fragments of L2 amplified from each individual. They were mostly identical, with the main difference among them being the length of the TA repeats in Sec II. The fragments S1 and S2 were 58% and 22% of the length of L2 in 85A. L1 was 50 bp longer than L2 in 85B, and most of the length difference was due to the number of TA repeats (21 repeats). All the alignment data are shown in supplementary figure S1A and Supplementary Data, Supplementary Material online.
Are the Minicircles Amplification Artifacts?
As the A+T-rich region contained a complex series of repeats, the possibility remained that some of these putative minicircles were artifacts of amplification. For example, PCR jumping, in which the newly synthesized DNA strand jumps from one repeat to another, could produce small amplicons that resembled those that would be produced by minicircles (or maxicircles). To investigate whether the putative minicircles were artifacts caused by PCR jumping, we used plasmid DNA containing cloned fragments of the A+T-rich region of the entire mt genome. In this way, we could examine whether amplification of a template known not to contain minicircles could produce smaller amplicons (i.e., artifactual minicircles), while sequencing would indicate whether they coincided precisely with the putative minicircle amplicons characterized above (L1, L2, S1, and S2). In the first set of experiments, we used a plasmid containing the A+T-rich region of the entire mt genome as template, together with the same primers used to produce the minicircle amplicons (L1, L2, S1, and S2). This reaction produced a single amplicon of size ∼600 bp (fig. 3B), but sequencing of this fragment indicated that it was 639 bp and not similar to any of the minicircle amplicons. It contained one copy of repeat C and D and nine copies of repeat B—this organization was not evident in either the entire mt genome or in any of the minicircle amplicons. The amplification of a unique product was surprising as we expected that molecules amplified from the cloned A+T-rich region should also be amplified from genomic DNA: both templates contain the same sequence. We then investigated whether the cloned A+T-rich region had rearranged during bacterial growth, but the sequence precisely resembled the full-length A+T-rich region. This does not rule out the possibility that the cloned DNA contained a small proportion of rearranged molecules, but we think it is more likely that PCR jumping occurred during the amplification of the cloned A+T-rich region to a greater degree compared with the genomic DNA amplification. This may be attributable to the much higher concentration of target DNA. In the amplification utilizing the cloned A+T-rich region, the target is not diluted by nuclear DNA, whereas in the genomic DNA amplification, the target (mtDNA) represents only a fraction of the total DNA.
To investigate whether the presence of minicircles might also lead to PCR jumping, we used another plasmid DNA containing the strongest amplified fragment 85A-L2. If PCR jumping did not occur, we would expect to produce a single amplicon using the same PCR primers and conditions as used in the original amplification. However, this amplification produced four bands, which corresponded (in size) to each of the putative minicircles. Two of these amplicons were subcloned and sequenced and found to be identical to the sequences of 85A-L2 and 85A-S1. We consider this strong evidence that S1 is an artifact, produced by PCR template switching between the repeats present on the minicircle. Although we could not purify and sequence the band corresponding to L1 and S2, the coincidence in size leads us to suspect that L1 and S2 are also artifacts of PCR jumping.
To further test whether the stongest amplicon (L2) was a PCR artifact, we performed two experiments. First, we designed a primer (85-ATF2) that straddled the putative “breakpoint junction” sequence and together with 85-ATR performed an amplification using genomic DNA (fig. 2B). The breakpoint junction primer should only anneal to a minicircle, as this sequence is absent in the entire mt genome. A single sharp band was obtained (fig. 3C) and the sequence data showed that it was identical to the corresponding part of L2. In the second experiment, we used the plasmid containing the entire A+T-rich region as template, together with the same primer pair to perform amplification. As expected, no PCR product was obtained, indicating that the breakpoint junction primer does not misprime somewhere in the A+T-rich region. The two experiments showed that L2 is unlikely to be an artifact.
Elimination of the Possibility of numts
numts commonly exist in metazoans (Bensasson et al. 2001). In our protocol, the amplification of minicircles was carried out using genomic DNA. Therefore, the possibility that the minicircle came from numts cannot be excluded. In order to investigate this possibility, we used a primer pair (85-ATF2 and 85-MRF2) that would amplify the remainder of L2, if it was present on a minicircle. These primers are inward facing on a minicircle but would be outward facing on a nuclear chromosome. As expected, this reaction produced an amplicon with sequence identical to a section of sec II of the A+T-rich region and, together with L2, conceptually comprises a circular molecule of about 700 bp. The amplified fragment length varied slightly among different clones (see alignment data in supplementary fig. S1C, Supplementary Material online). These differences might be caused by polymerase error or represent minor variations present on different minicircles. The use of outward-facing primers reduces the risk of amplifying numts but does not completely exclude it. If a numt is present as a tandem copy, outward-facing primers would produce an amplicon. Indeed, if a tandemly repeated numt existed which had a junction that precisely matched the breakpoints identified in the putative minicircle L2, this template would produce amplicons with both of the pairs of outward-facing primers described earlier (85-ATF and 85-ATR and 85-ATF2 and 85-MRF2). However, a tandemly repeated numt would not produce the heterogeneity observed between clones (all copies would be the same) nor the variation in repeat B and D that was observed in the 85-ATF/85-ATR amplification. Although the existence of a tandemly repeated numt cannot be absolutely excluded, we consider its existence unlikely.
Discussion
Importance of Control Reactions
In this study, we report convincing evidence of the product of recombination (the minicircle) in the mt genome of the parasitic wasp Conostigmus sp. However, we found that it was critical to establish appropriate control reactions to identify and characterize artifacts produced by PCR jumping. We suggest that this is especially important when repeats exist within the DNA template. Our experiments indicated that it was necessary to perform control amplifications using both the isolated (cloned) A+T-rich region and the putative minicircle in order to identify all nonbiological products of amplification. PCR artifacts were especially prevalent when using the cloned minicircle fragment, suggesting that this sequence is more prone to PCR jumping. The main difference between these two control sequences is the presence of the breakpoint junction sequence in the minicircle fragment. Finally, our results suggest that it is critical to use breakpoint junction primers—primers that straddle the putative breakpoint junction—in order to firmly establish the biological nature of putative minicircles. Once the biological origin of the minicircle was confirmed, it was also necessary to demonstrate that the minicircle was derived from a circular (mt) molecule, as the existence of numts might interfere with our identification of real recombination end products.
Possible Explanations for Low Copy Number of the Maxicircle
The maxicircle detected in this study appeared to be present at very low copy number, as it could only be detected by nested PCR. A low copy number of the maxicircle was also observed in human tissues (Kajander et al. 2000). However, we interpret the existence of the maxicircle in our study with some caution. As it could only be detected after nested PCR, we cannot exclude the possibility that it was caused by PCR error. If, as we suspect, the minicircle contains the origin of replication (discussed later), the absence of an origin of replication in the maxicircle would explain its presence at very low copy number.
Comparison with Minicircles from Other Metazoa
The discovery of minicircles in a hymenopteran expands the number of taxonomic groups in which minicircles have been characterized. Previously they have been characterized in humans (Holt et al. 1988), the Phthiraptera (Shao et al. 2009; Cameron et al. 2011), and in nematodes (Lunt and Hyman 1997; Armstrong et al. 2000; Gibson et al. 2007), although the characteristics of the minicircles vary between these groups. In humans, minicircles are generated by partial mt genome deletion, and these minicircles coexist with the entire mt genome. In the Phthiraptera, three types of minicircles have been identified: the first type is similar to the minicircles in humans (Cameron et al. 2011), the second type has multiple genes and a short noncoding region (Cameron et al. 2011), and the third type has one to three genes and a large noncoding region (Shao et al. 2009; Cameron et al. 2011). In the nematodes, the minicircle of M. javanica is excised from the entire mt genome and coexsits with both its reciprocal partner (maxicircle) and the entire mt genome (Lunt and Hyman 1997); minicircles found in the potato cyst nematode range from 6 to 9 kb (Armstrong et al. 2000; Gibson et al. 2007). Some of the potato cyst nematode mincircles are mosaics, comprising multigenetic fragments derived from other minicircles. The minicircle reported in the present study is most similar to those reported in the nematode M. javanica, and the discovery of minicircles in four taxonomically divergent groups indicates that they are likely to occur much more commonly than is currently appreciated.
We found that there were both similarities and differences between the minicircle of Conostigmus sp. and the minicircle reported by Lunt and Hyman (1997) in the nematode M. javanica. The minicircles were similar in that they were 1) both excised from the A+T-rich region of the entire mt genome and 2) comprised of different repeats flanked by unique sequences. The differences were mainly in two aspects. First, the breakpoints in the mt genome of Conostigmus sp. occurred in two copies of a nontandem repeat, and imperfect microhomologies (14 bp) could be found around the breakpoints (supplementary fig. S2, Supplementary Material online). In contrast, only short (3 bp) or no direct repeats could be found around the breakpoints in the entire mt genome of the nematode M. javanica. Second, nucleotide deletion occurred upstream of the second breakpoint in the minicircles of all individuals examined in this study, when compared with the entire mt genome. No such deletions were evident in M. javanica. Two candidate explanations might be responsible for the nucleotide variation within an individual. First, the minicircles were not recently formed but instead were generated long ago and transmitted to subsequent generations. Another explanation for the nucleotide variation is that mismatching might be tolerated during recombination as perfect microhomologies are not necessary for the repair of DSBs. Both explanations might also be responsible for the observation that the minicircles are highly similar in related individuals. Unfortunately, it is very difficult to experimentally distinguish between these two scenarios.
We suspect that the minicircles contain origin of replication information, and that these smaller molecules are likely to replicate faster than larger ones (Diaz et al. 2002). This could explain the higher levels of minicircle compared with the maxicircle.
Possible Mechanism Yielding the Minicircle and Maxicircle
The mechanism that produces mtDNA minicircles is still unclear. The characterization of a suite of minicircles, from evolutionarily divergent organisms, is perhaps the best approach for identifying the key mechanistic components. Thus far, the production and repair of DSBs appears to be associated with the existence of sublimons (ΔmtDNA) in humans (Krishnan et al. 2008). Recent research of rodent mtDNA showed that DSBs promoted recombination, resulting in the formation of sublimons (Srivastava and Moraes 2005; Bacman et al. 2009).
Mitochondria have been shown to have an effective system of DSB repair in Drosophila (Morel et al. 2008). Krishnan et al. (2008) proposed that 3′→5′ exonucleases could generate single-stranded regions at DSBs, and these exposed single-strand regions could subsequently anneal at regions containing microhomologous sequences, and that this would result in the formation of sublimons. We speculate that the minicircle and maxicircle of Conostigmus sp. were probably generated by a similar mechanism. However, in order to generate both a minicircle and a maxicircle, we propose that two breakpoints must occur simultaneously near or within the direct repeats (14 bp) (fig. 4). After the minicircle sequence was excised, another pair of microhomologous sequences (up to 21 bp, the sizes varies depending on the position of the breakpoints) could be detected around the breakpoints (supplementary fig. S2, Supplementary Material online), which probably facilitated the formation of the maxicircle. Ligase III has been identified to play an important role in maintaining mtDNA integrity (Gao et al. 2011); the ends of the linearized DNA around the breakpoints were probably joined by ligase III.
Model for the generation of minicircle and maxicircle mtDNA by recombination during repair of DSBs. Direct repeats which facilitate the formation of minicircle and maxicircle mtDNA are shown in red and purple regions, respectively.
Direct Link between Intramitochondrial Recombination and mt Gene Rearrangement
The Hymenoptera is now the second insect order after the Phthiraptera in which minicircles have been detected (Shao et al. 2009). Coincidently, both groups possess the most rearranged mt genomes among insects. We suggest that there might be a direct link between these two phenomena. Although intramitochondrial recombination has been considered as a possible mechanism for gene rearrangement in animal mt genomes (Dowton and Campbell 2001), there has been very little direct evidence to support this. One reason for this is that the evidence, in the form of excised partial mt genomes, direct repeats, or inverted repeats, may be erased soon after (in evolutionary terms) the recombination event (Brockman and McFadden 2012). The presence of minicircles, combined with the highly rearranged mt genome reported in this study, supports the notion that recombination is an important component of the mt gene rearrangement mechanism. Our working hypothesis is that the generation of minicircles is a crucial first step in the rearrangement mechanism, but that these minicircles may have to be long-lived (i.e., contain an origin of replication). The integration of these minicircles back into the full-length mt genome, via homologous recombination, then generates gene rearrangements. This hypothesis not only gives a good explanation to the existence of minicircles but also accommodates the existence of the nontandem repeat fragments that are found in some taxa, such as those whose mt genome contains two or three copies of the control regions (Kumazawa et al. 1996; Mueller and Boore 2005; Kurabayashi et al. 2008). These duplicated control regions have been postulated to be generated and maintained by recombination (Mueller and Boore 2005; Kurabayashi et al. 2008).
Materials and Methods
DNA Extraction
Genomic DNA was extracted from 100% ethanol-preserved specimens of Conostigmus sp. (collected from Mount Barker, South Australia) using the “salting out” protocol (Aljanabi and Martinez 1997). The DNA was resuspended in 100 μl of fresh TE solution (1 mM Tris–HCl, 0.1 mM ethylenediaminetetraacetic acid [pH 8]) and stored at 4 °C.
Amplification, Sequencing, and Annotation of the Entire mt Genome
Initially, small gene fragments were amplified and sequenced using either published universal primers (Simon et al. 1994; Simon et al. 2006) or using primers designed from consensus hymenopteran mt sequences. Using the sequence information obtained, specific primers were designed for the amplification of the remaining regions. Short and long PCRs were performed using BIOTAQ DNA Polymerase (Bioline, Australia) and Takara LA Taq (Takara Biomedical, Japan), respectively, following the manufacturers’ instructions. For short PCRs, cycling condition consisted of 94 °C for 2 min, 35 cycles of 94 °C for 30 s, 45–65 °C for 1 min, 72 °C for 2 min, and a final elongation for 10 min at 72 °C. Long PCRs were performed with the following conditions: 94 °C for 2 min, 30 cycles of 94 °C for 1 min, 45–65 °C for 30 s, 68 °C for 12 min, and a final elongation for 10 min at 68 °C. All PCR products were resolved by agarose gel electrophoresis and purified with ExoSAP-IT (GE Healthcare, Bucks, UK). The primer walking method was used for direct sequencing, using the ABIPRISM BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Australia). Amplicons that were difficult to sequence directly were ligated into the pGEM-T Easy Vector System (Promega, Madison, Wisconsin) and cloned into Escherichia coli JM109 cells (Promega, Madison, Wisconsin). The sequences of both strands were determined for the entire mt genome. Genes were identified using methods as previously described (Mao et al. 2012).
Amplification and Sequencing of Minicircles and the Maxicircle
To detect the existence of minicircles, two specific, outward-facing primers (85-ATF and 85-ATR) were designed, based on the A+T-rich region sequence, following the strategy of Lunt and Hyman (1997). PCR amplification was carried out using Takara LA Taq. PCR products were purified by agarose gel electrophoresis and cloned as described above. A total of 3–5 clones per product were sequenced. The putative maxicircle was amplified using the primers 85-MR1 and 85-36R1 followed by a nested PCR with two internal primers 85-MR2 and 85-36R2. The PCR product was purified and sequenced directly. The primers are indicated in figure 2.
Acknowledgments
The authors thank Xiushuai Yang for his helpful discussion and assistance with preparing the figures. This study was supported by the Center for Medical Bioscience, University of Wollongong, and in part by a National Science Foundation grant (No. DEB-0614764) to N.F.J. and A.D.A.
References
Author notes
Associate editor: Gregory Wray
The annotated sequence of Conostigmus sp. mitochondrial genome has been deposited in GenBank with accession number KF015227.
