Our sequence analysis of mtDNA control region (cloning–sequencing) and whole mtDNA (MPS) of MZ twins revealed several novel insights into one of the most fundamental characteristics of mitochondrial genetics—heteroplasmy. First, the higher level of heteroplasmy in skeletal muscle versus blood samples supported previous suggestions that the level of heteroplasmy is higher in post-mitotic tissues (1
). Second, the repertoire and level of heteroplasmy varied widely among different individuals. Third, a large portion of the HM pattern was inherited rather than de novo
among twins. This was more evident in twins with many heteroplasmic mutations. Finally, a clear pattern of negative selection was observed among both highly abundant and low prevalence heteroplasmic mutations. All these observations underline both the usage of MPS and the choice of identical twins as a successful approach to investigate the nature of heteroplasmic mutations.
T2DM concordant and discordant twins were utilized to compare the etiological role of acquired HM pattern differences within a constant genetic background. Although our study enabled high resolution and identification of multiple heteroplasmic mutations, no clear T2DM-associated pattern was identified. Two obvious reasons could underlie these findings: (i) T2DM is an extremely complex disorder and is frequently referred to as a family of disorders. The disease is caused by the interplay between multiple genetic and environmental factors resulting in an estimated heritability of ~50% (25
). Thus, although MZ twins notably reduce the genetic variability between the twin members, the genetic and epigenetic variability across individuals is still likely to be great. (ii) Although multiple sources point to the role that mitochondrial dysfunction plays in the etiology of T2DM, it is possible that any effect is dependent on small effect sizes and the sample size should be increased to exclude these and possible clinical subgroups (26
That many heteroplasmic changes were either common in both twins within a pair or shared across both blood or skeletal muscle samples from the same individual suggest that many of the heteroplasmic mutations are inherited rather than accumulated over the lifetime of the individual. Although previous reports suggested the inheritance of certain discovered heteroplasmic mutations (reviewed in 27
), especially disease-causing mutations, the extent of this phenomenon was not thoroughly addressed. In mice, apparently neutral heteroplasmic variants could persist for at least 14 generations (29
). Recent MPS analysis (SOLiD ABI) of whole mtDNA sequences from several generations of mice either carrying a mutated DNA PolG or wild-type revealed among other findings a notable proportion of heteroplasmic mutations sharing among sibs, suggesting their inheritance from the maternal egg cytoplasm (20
). Similarly, MPS analysis (Illumina) of whole mtDNA sequences from a single individual revealed mutations shared among different tissues (18
). These findings, along with our observation of notable sharing of HM repertoires even among twin members having multiple heteroplasmic mutations, raise the possibility that transmission of multiple heteroplasmic mutations could constitute a common phenomenon, despite the possible action of the mitochondrial bottleneck during the development of the female germline (30
We noticed that the vast majority of the mutations identified by the PCR–cloning–sequencing technique were present in the MPS data, but were removed from our final analysis by our stringent filters. Notably, the purpose of our filters was to reduce the effect of false positive results. Therefore, we decided against retaining these particular mutations and removing others just because these mutations were also observed in the cloning–sequencing technique. We are aware that by taking this approach, many true heteroplasmic mutations will be overlooked, thus increasing the rate of false negative results. An excellent example for this outcome is the mutation at position 15 132 which was identified by the MPS and confirmed by Sanger sequencing in several samples but was removed from the MPS analysis of the skeletal muscle of sample 68 841 by our filter D. Hence, utilizing our filters could reduce false positive results but may lead to overlooking true heteroplasmic mutations. This is one of the reasons why PCR–cloning–Sanger sequencing was used in addition to MPS, as these techniques are complementary: the drawbacks of one are relieved by the other. As an example, low complexity repeat regions could be analyzed by the cloning–sequencing approach but currently cannot be analyzed by MPS, at least not with high certainty. Second, the short reads generated by SOLiD interfere with identifying mutational combinations, which could be partially identified by the cloning–sequencing approach. Third, mutations in the MPS are discovered at much higher resolution than by the cloning–sequencing approach and as a result many mutations are seen in the MPS but not by cloning. Thus, until long high quality reads are generated by any of the MPS platforms there is benefit in using the current MPS platforms in combination with complementary techniques, such as the cloning–sequencing approach.
Inheritance of HM patterns may have serious implications on evolution, since these could set the basis for recombination, thus resulting in new genetic backgrounds (27
). Nevertheless, recombination events among mtDNA molecules which differ in a single mutation, which constitute most events of heteroplasmy (3
), are likely to have relatively little influence on evolution. Alternatively, recombination among mixed populations of mtDNA molecules that differ in their type (different haplogroups) would likely influence population genetic variation. Such a mixture of two different mtDNA types could occur upon rare paternal leakage of mtDNA (32
). Our results, both using the cloning–sequencing and MPS techniques, did not reveal clear evidence for such a pattern. As SOLiD analysis produces short reads, combinations of mtDNA mutations within a single molecule could not be easily detected. This may be overcome by introduction of MPS techniques generating long reads of single molecules, such as the recently developed technique by Pacific biosciences (www.pacificbiosciences.com
While investigating the distribution of heteroplasmic mutations we found significant over-representation of mutations in the control region, suggesting the action of negative selection on the mtDNA-coding region. Recent studies in mice used experimental designs that could adequately differentiate between drift and selection (20
). These studies generated maternal mice heteroplasmic for pathogenic mutations, and then tracked their progression into future generations. All three studies presented evidence for purifying selection during oogenesis within few generations. Two plausible mechanisms of selection, working at the organelle level, have been put forward by Shoubridge and Wai (36
): (i) pathogenic mtDNA mutations may make the mitochondria prone to autophagy, or (ii) non-silent mutations will simply make these mitochondria less efficient at replication, leading to them being outnumbered by healthy organelles. A recent study suggested that mitophagy factors actively select against mutated mtDNA molecule in human cells, thus supporting the autophagy mechanism (37
). Additionally, we found that the effect of negative selection was evident both when analyzing low abundance (1.6–5%) or high abundance (>5%) mutations, which argues for the existence of an active selective mechanism that sifts through the repertoire of heteroplasmic mutations. This is further supported by the fact that 85.9% (55/64) out of the identified heteroplasmic mutations were already present in the public databases, thus reflecting selection against rarer variants which are expected to be enriched with mutations with deleterious potential.
In summary we performed deep sequence analysis of HM patterns throughout the human mtDNA in MZ twins either discordant or concordant for the T2DM phenotype. Although no clear correlation between heteroplasmic mutations and T2DM could be observed, we identified evidence both for the inheritance of heteroplasmic mutations and for the action of negative natural selection even over low abundance heteroplasmic mutations.