Rapid intergenerational segregation of mtDNA is a poorly studied phenomenon despite its importance for understanding principles of the fundamental biology and management of clinical syndromes caused by mtDNA mutations. Such unpredictable intergenerational mtDNA fluctuations appear to be even more extreme and frequent in humans than in the mouse (Wonnapinij et al., 2010
). To explain the segregation of mtDNA variants between generations, a genetic bottleneck hypothesis in the female germline has been proposed. According to this model, segregation and rapid shift of mtDNA occurs during oogenesis resulting in partitioning of mtDNA haplotypes between individual oocytes (Jenuth et al., 1996
; Olivo et al., 1983
Although, the existence of mtDNA segregation during early embryogenesis and subsequent shift from transiently heteroplasmic embryos to the homoplasmic offspring was hypothesized (Laipis et al., 1988
), experimentally this was never demonstrated. We have now tested this hypothesis by generating heteroplasmic rhesus oocytes with equal mixture of two wild-type mtDNA haplotypes and followed the transmission of mtDNA to preimplantation embryos, fetuses and germ cells. Using a single cell analysis, we demonstrate rapid segregation of mtDNA variants between daughter blastomeres of cleaving preimplantation embryos, similar to that seen in germ cells, i.e. in individual oocytes (). Intriguingly, three ESC lines and one fetus produced from these heteroplasmic embryos were nearly homoplasmic suggesting that return to the homoplasmic condition can occur during development of an individual organism from a zygote to birth, without actual passage of mtDNA through the germline (). The fact that early passage ESCs derived from ICMs of heteroplasmic blastocysts have already shifted towards homoplasmy suggests that strong segregation and bottleneck occurs during early epiblast lineage specification rather than during later fetal development. Our study also supports the model that possibly few cells within epiblast progenitors give rise to the somatic cell lineage of embryo proper. This concept is supported by evidence in the mouse that whole somatic lineage of the embryo proper can be derived from just one founder epiblast cell (Wang and Jaenisch, 2004
). Another interesting point is that only progenitors with low heteroplasmy (towards either resident or alien mtDNA) contributed to the embryo proper lineage while intermediate variants where lost. This observation supports the notion that the genetic bottleneck at this stage may not be random but rather preferentially selects homoplasmic conditions.
In contrast to relatively homogeneous fetal somatic tissues with low heteroplasmy levels, we found a wide range of heteroplasmic variants distributed in individual oocytes. These data support the existence of a secondary mtDNA bottleneck responsible for segregation of mtDNA variants in the female germline. Previous studies concluded that much of the mtDNA segregation occurs during oocyte development (Jenuth et al., 1996
). However, based on our results, we cannot exclude possibility that this segregation takes place much earlier, in precursors of germ cells within the epiblast. This model postulates that the majority of cells in the ICM may contribute to the germline.
The biological mechanism underlying the mtDNA bottleneck is not clear yet. Earlier mouse studies suggested that the basis of the mtDNA bottleneck may be a significant reduction in mtDNA copy number per cell (about 200) during a specific point in oocyte development (Cree et al., 2008
). However, recent evidence suggests that segregation occurs without a reduction of mtDNA contents but rather by preferential amplification of specific mtDNA variants (Cao et al., 2007
). Neither of these models seems to be able to explain the mtDNA segregation between individual blastomeres in cleaving monkey embryos seen in our study. First, mtDNA copy numbers in blastomeres of cleaving embryos could be in the tens of thousands, well above of estimated minimum (≈200) to be effective for segregation. Secondly, there is no known replication of mtDNA during preimplantation development excluding the possibility of selective amplification. Although, a very short period of mtDNA synthesis immediately after fertilization has been described (McConnell and Petrie, 2004
). Mitochondria are not passive organelles, but rather exist as dynamic and interacting structures maintained through balance of fusion and fission. Hence, the distribution of mitochondria and mtDNA to daughter cells during cell division seems to be regulated by mechanisms responsible for mitochondrial fusion and fission (Kashatus et al., 2011
). It would be important to investigate the extent of mitochondrial dynamics in preimplantation embryos and their role in regulation of mtDNA segregation and bottleneck.
In our study, we utilized mtDNA differences between Indian- and Chinese-origin rhesus macaques to generate heteroplasmic embryos and offspring. It is believed that sequence polymorphisms between these two subpopulations of macaques are as great as those between some primate species (Smith, 2005
). Transcription and replication of mtDNA is tightly controlled by the nuclear genome, suggesting that the resident mtDNA may have a replicative advantage over the alien mtDNA during development (Meirelles and Smith, 1998
). However, segregation towards either resident or alien mtDNA haplotypes in fetuses or ESCs, irrespective of the nuclear genetic background seen in our study, either conflicts with this assumption or suggests that the nuclear control of mtDNA replication is highly conserved.
Our study also has far reaching clinical implications for genetic management of mtDNA diseases. Currently, PGD is actively pursued for monitoring mtDNA disorders by sampling one or two blastomeres and selecting embryos for transfer with low mtDNA mutation loads (Poulton and Bredenoord, 2010
; Thorburn et al., 2009
). However, based on our results, heteroplasmy in biopsied blastomeres from cleaving embryos may not be fully predictive of total mutation load in remaining blastomeres and thus in the embryo. Moreover, heteroplasmy levels in fetal tissues may change drastically compared to preimplantation embryos due observed bottleneck. In contrast, chorionic villus sampling (Poulton et al., 2010
) could be more reliable assays based on observations that mtDNA segregation patterns between placenta and fetus were relatively low. It is worth to note that our study is based on heteroplasmy using wild-type mtDNA haplotypes. However, segregation patterns for pathogenic mutations in human embryos could be different (Marchington et al., 1997
; Marchington et al., 1998
; Steffann et al., 2006
). Available clinical observations are often made on limited number of blastomeres or on arrested embryos (Monnot et al., 2011
) suggesting that more extensive human studies for each mutation type are needed.
In our follow up studies with the ST approach designed for efficient replacement of mtDNA in oocytes, the amount of mtDNA carryover was insignificant or undetectable in major organs and tissues of offspring. However, a few oocytes recovered from the ST female offspring carried mtDNA from spindle donor oocytes at levels reaching 16.24%. This was unexpected since carryover levels during ST procedure were estimated to be below 1%. While these heteroplasmy levels are below known mutation thresholds for phenotypic expression of many mitochondrial diseases, the possibility exists that mtDNA heteroplasmy may change in subsequent generations through the bottleneck. This observation also suggests that PGD selection for embryos carrying 30% or less mutation loads most likely will not eliminate the possibility of recurrence of mitochondrial diseases in subsequent generations.