Implications of human evolution and admixture for mitochondrial replacement therapy

Mutations to mitochondrial DNA (mtDNA) have been associated with a wide range of human diseases [1, 2]. Since mitochondria are maternally inherited, mitochondrial genetic disorders will be passed from mothers to their children. Effective treatments for mitochondrial disease are rare, and patients are often faced with limited therapeutic options. Furthermore, the ability to accurately assess the risk of inheriting a mitochondrial genetic disorder can be complicated by the co-occurrence of wild-type and mutated mtDNA (i.e., heteroplasmy) in a single female [3]. Mitochondrial replacement (MR) therapy is a promising new assisted reproductive technology that could allow women with mitochondrial disorders to give birth to healthy children to whom they are closely genetically related. MR therapy works by combining nuclear DNA (nDNA) from a mother who has a mitochondrial disorder together with healthy mitochondria from an egg donor. For MR-assisted in vitro fertilization (IVF), the nuclear genome is removed from a fertilized oocyte with diseased mitochondria and injected into an enucleated donor egg that contains healthy mitochondria. This process results in so-called ‘three-person babies’ since children born from MR therapy will have genetic contributions from two mothers and one father.

Studies in mammalian systems over the last decade have underscored both the surmountable technical challenges, and the considerable promise, associated with MR therapy. The nuclear transplantation procedure that underlies MR therapy was first shown to be possible in mice [4]. The progeny of nuclear transplantations from mouse oocytes with mitochondrial disease into healthy oocytes were found to be viable and disease-free. Later in primates, MR therapy was used to produce four Macaque offspring [5] that showed healthy development to 3 years of age [6]. Nuclear transfer in this case occurred prior to fertilization, a technique that has not proven to be equally effective in humans. The MR procedure was first developed in humans using abnormally fertilized zygotes [7]. Human MR experiments rely on pronuclear transfer, whereby the nuclear genome is removed from a newly formed human embryo shortly after fertilization. This approach showed promise with respect to both the small amount of diseased mitochondria that are carried over to the healthy donor egg and in terms of normal in vitro development through the blastocyst stage. More recently, the same group demonstrated even greater efficacy of the pronuclear transfer technique for MR therapy with normally fertilized embryos by transferring the pronuclei compartments containing maternal and paternal haploid genomes almost immediately after they first appear [8].

These crucial experimental advances in MR therapy have occurred against a backdrop of substantial regulatory investigation related to the technique’s desirability, safety and potential efficacy [9]. Most of the effort and progress on this front has occurred in the United Kingdom (UK). The UK’s Human Fertilisation and Embryology Authority (HFEA) was initially charged with evaluating MR therapy, and they recommended further studies before the technique could be adopted as a clinical practice. Subsequently, several independent UK science agencies supported the bioethics of MR therapy, and the public was found to be largely in favor of its use. These findings ultimately led the UK government to draft a set of regulations for the technique, and parliament approved MR therapy as an assisted reproductive technology in February of 2015. Regulations were to be enacted by October of the same year, with clinics able to apply for a license by November. Initial attempts to conceive via MR-assisted IVF could have begun by the end of that same year. When this manuscript was written, there was no record of any human birth resulting from MR therapy in the UK. However, during the manuscript review process, news broke of a ‘three-person baby’ resulting from MR therapy born in Mexico [10]. The United States (US) doctors who led the procedure chose Mexico to avoid US regulations that do not yet permit MR therapy, leading to charges of unethical and irresponsible behavior.

Despite the considerable technical and regulatory progress that has been made on the issue, substantial concerns have been raised about the safety of MR therapy [9, 11]. These concerns rest largely on the notion of potential incompatibility (mismatches) between nuclear and mitochondrial genomes from different populations of the same species. We refer to this idea here as the ‘nuclear-mitochondrial mismatch hypothesis’. Incompatibility between nuclear and mitochondrial genomes from divergent populations would most likely be predicated upon interactions between proteins encoded by each [11]. While the human mitochondrial genome only encodes 37 protein coding genes, there are more than a thousand nuclear genes that encode proteins involved in mitochondrial function. These include 76 nuclear encoded proteins that directly bind mitochondrial counterparts. Sequence variations that change the binding affinities between nuclear and mitochondrial proteins could have deleterious effects, which would jeopardize healthy outcomes from MR therapy. This possibility has been supported by comparative sequence analyses showing the importance of compensatory sequence changes that serve to maintain physical interactions between nuclear and mitochondrial encoded proteins [12, 13].

A number of studies from model organisms have provided even more direct evidence for incompatibility between nuclear and mitochondrial genomes brought together from different populations of the same species. For example, mice with mismatched mitochondrial and nuclear genomes were able to survive to adulthood but showed stunted growth and reduced physical performance [14] as well as reduced learning and exploratory behavior [15]. The neurological effects observed in the latter study increased with age. Analogous studies in invertebrates have also turned up numerous deleterious effects of combining divergent nuclear and mitochondrial genomes. Such effects include changes in aging [16–19], survival [20] and fertility [21–24] along with impaired mitochondrial function [25, 26]. It should be noted that all of these studies entailed repeated genetic backcrosses whereby divergent mitochondria were introduced into highly inbred lines. As such, they represent extremes of genetic divergence between lineages and are not likely to accurately reflect human populations that routinely interbreed [27]. Nevertheless, these findings do point to a number of possible complications arising from nuclear-mitochondrial genome mismatch.

While the aforementioned studies have revealed instances of nuclear-mitochondrial incompatibility by studying the progression of chimeric individuals into adulthood, MR studies in humans have been conducted in vitro and only followed embryos through the blastula stage of development. This has led to calls for additional preclinical trials of MR therapy with a much longer time horizon [9]. However, it has occurred to us that long term experiments of this kind have already been conducted in nature via the process of human evolution. The evolutionary history of anatomically modern humans has been characterized by relatively long periods of isolation and genetic divergence interspersed with migrations and genetic admixture between previously isolated populations [28–30]. Admixture between genetically distinct human populations should have brought together divergent nuclear and mitochondrial genomes. Evidence that this is in fact the case could be taken to refute the nuclear-mitochondrial genome mismatch hypothesis for humans and to thereby support the feasibility of MR therapy.

In light of this realization, we systematically evaluated the extent of naturally occurring nuclear-mitochondrial genome swapping that has occurred among human populations. The goal of our survey was to get an idea of the extent of nuclear-mitochondrial divergence that can be tolerated within any single individual as well as a sense of how often swapping has occurred in human evolution. To do this, we evaluated the distribution of nuclear genomic diversity and mtDNA haplotypes among the 26 human populations, representing five major continental groups, which were characterized via whole genome sequencing as part of the 1000 Genomes Project (1KGP) [31]. We also performed a confirmatory analysis of mtDNA sequence variation among 58 populations from the Human Genome Diversity Project (HGDP), which characterized mitochondrial genomes to a lower level of resolution using SNP arrays [29]. We reasoned that since the donors for these human genome diversity projects are (apparently) healthy individuals, they should not bear incompatible nuclear-mitochondrial genome combinations. In addition, since these diverse populations have been shaped by millennia of natural selection, deleterious combinations of nuclear-mitochondrial genomes should have been eliminated long ago and would not be observed in extant populations. In this sense, our survey can be considered as a test of the nuclear-mitochondrial genome incompatibility hypothesis in humans.

The results of our analysis on naturally occurring human genetic variation show that nuclear and mitochondrial genomes from divergent human populations can co-exist within presumably healthy individuals, indicating that such mismatched nDNA-mtDNA combinations are not deleterious and have not been eliminated by purifying selection. In other words, the long and ongoing experiment of human evolution provides no evidence whatsoever in support of the nuclear-mitochondrial mismatch hypothesis. These results can be taken to support the feasibility, and potential safety, of MR-assisted in vitro fertilization, at least with respect to the compatibility of human nDNA and mtDNA genomes. Of course, our results do not bear on any potential complications related to the technical implementation of such a complicated procedure. For example, it is extremely difficult to ensure that none of the defective mitochondria are transferred along with the nuclear genome. Indeed, it was recently shown that even when only a small percentage of defective mitochondria are carried over in the nuclear transfer process, they can increase in copy number and eventually replace most or all of the healthy mitochondria from the egg donor [57]. Such technical hurdles will need to be addressed to ensure the maximum safety of MR therapy.

Our results are in conflict with a number of previous studies on model organisms, which provide numerous lines of evidence in support of nuclear-mitochondrial genome incompatibility. For example, studies in mice have shown physical and neurological deficits related to nuclear-mitochondrial mismatches [14, 15]. In addition, the co-occurrence of divergent nuclear and mitochondrial genomes in invertebrates has been associated with diminished mitochondrial function [25, 26] along with deleterious effects on aging [16–19], survival [20] and fertility [21–24]. When considered together, these previous studies have been taken to issue a strong note of caution against MR therapy [9, 11].

It is interesting to note that much of the resistance to MR therapy has been articulated by evolutionary biologists who emphasize the extent to which nuclear and mitochondrial genomes co-evolve along population lineages [11]. This realization has raised the seemingly legitimate concern that advocates of MR therapy, and/or the regulatory bodies that are charged with evaluating its safety, may not have adequately considered the implications of evolution for the implementation of this new technology. However, the results of our study suggest that the model organism studies that have been used to argue against the safety of MR therapy do not accurately reflect the nature of human evolution [27]. For the most part, these model organism studies relied on backcrossing and the generation of inbred lines, and they also involved relatively divergent populations. Experiments of this kind can be expected to result in extremes of nuclear-mitochondrial genome divergence. Human populations, on the other hand, tend to show both low levels of genetic divergence and low inbreeding. Accordingly, one may expect to see less pronounced effects of nuclear-mitochondrial mismatch in human populations, and that is exactly what we observed in our study.