We present experimental evidence for strong purifying selection against nonsynonymous mutations in protein-coding genes during maternal transmission of mutated mtDNA in the mouse. The drastic reduction of mutations in the amino acid changing first and second codon positions of protein-coding genes are a direct result of purifying selection against deleterious mtDNA mutations at some stage within the reproductive cycle of these mice. This bias occurs rapidly and is evident as early as the N2 generation. These findings have profound implications for our understanding of how mutated mtDNA is transmitted between generations. It is important to recognize that this strong purifying selection against nonsynonymous changes that we observe is likely to be a universal phenomenon in mammals, but the rapid nature of this selective force would render these mutations difficult to detect in population studies. These findings have profound implications for our understanding of how mutated mtDNA is transmitted between generations.
Within studies of human mtDNA evolution, the observation is that many substitutions are not ancient changes shared deep within human haplogroups, but rather are new variants clustered within the tips of phylogenetic networks and found only in a small number of individuals. This implies they are mildly deleterious variants not yet selected against [23
]. Studies of disease-causing mtDNA mutations show they are often heteroplasmic, and can be present at high levels without consequence for the carrier. However, once the levels exceed a specific threshold, the respiratory chain function will be impaired, causing a clinical phenotype [11
]. Based on these observations, when using the mtDNA mutator mouse to study germ line transmission of mtDNA mutations, one could expect to observe the inheritance of high numbers of mutations at all sites in the early generations, which would eventually be removed from the mouse lines once their phenotypic thresholds had been crossed. Whereas the inheritance of the tRNA, rRNA, and third codon position mutations appear to be following this expected pattern (see and ), this is not the behaviour of mutations at the nonsynonymous first and second codon positions in our mouse lines ().
The strongest signature of purifying selection can be observed within mt-CO1
, consistent with the very high levels of sequence conservation in these genes. The strength and speed of this purifying selection could have other effects on the mutation patterns observed in our model. The consensus view is that bi-parental recombination of mammalian mtDNA is at most extremely rare [3
] and therefore selection acting at any one site in the mtDNA will affect the entire mtDNA molecule. The observed strong and rapid selection of mtDNA mutations could therefore also reduce the number of neutral variants observed, due to their linkage to deleterious mutations. This means that 4-fold degenerate sites or even noncoding mutations might not be the reliable measure of the mitochondrial neutral mutation rate. Such an underestimation of the mtDNA mutation rate using phylogenetic or population methods relative to pedigree-based observation has been reported previously [45
]. If this is the case, the models based on this assumption require recalibration.
This point is also important in interpreting the excess change observed for the mt-CYB
, and mt-ATP8
genes in mtDNA mutator lines. Similar gene-specific increases of mutations have been reported in human mtDNA, especially in mt-ATP6
]. Though some argue that this signifies positive selection, the pattern may also be due to less-intense purifying selection on these specific genes. If mutations at mt-ATP6
experience less-selective constraint, mutations at these sites will be allowed to accumulate and persist in the mtDNA pool. Meanwhile, mutations at strongly selected sites, such as mt-CO1
, are eliminated, leading to the relative increase in the observed frequency of mt-ATP6
mutations in our model organisms.
In contrast to the rapid selection against nonsynonymous changes, rRNA and tRNA genes experienced less-intense purifying selection in our mtDNA mutator lines. Though tRNA genes also have high levels of sequence conservation, the frequency of observed mutations at these sites in our mouse lines was quite similar to the rate observed at third codon positions (A and A). Some of the identified tRNA mutations, e.g., the deletion of one base in the anticodon loop of mt-TM
(3873delC mutation) can be predicted to have a biochemical effect if present at high levels. Previous models have mainly been based on observational studies of transmission of mutated mtDNA in human pedigrees affected by mitochondrial disease. Such threshold-mediated protection from selection should lead to slower purifying selection of the mtDNA variant, which may be reflected in the essentially neutral segregation patterns observed for disease-causing mutations prior to clinical manifestation [11
It is plausible that these tRNAs, as well as a number of the nonsynonymous changes in the protein-coding genes in our model system, may eventually behave like human mtDNA disease mutations in that these mutations are transmitted and cause no obvious phenotype at low levels, but may be selected against or cause a disease-like phenotype at higher levels. We will continue sampling our lines to investigate the long-term fate of the observed tRNA gene mutations, as well as the stably transmitted nonsynonymous protein-coding gene changes.
In the mtDNA mutator mice, the mutations within protein-coding genes are equally distributed across all three codon positions [38
], whereas the pattern of mutation accumulation is different in the mtDNA mutator lines. It has previously been proposed that mitochondrial fitness may be selected for during oocyte development [50
], and it is therefore quite possible that mtDNA in germ cells is under a different selective regime than the mtDNA in somatic cells. There is a massive proliferation of mtDNA during oogenesis, whereby a small number of mtDNA copies in the primordial germ cells are extensively amplified to generate the approximately 105
mtDNA copies in the mature oocyte [2
]. This mechanism provides ample opportunities for functional testing of mtDNA during female germ-cell development, and future research is required to unravel molecular mechanisms responsible for this selection.
Our experimental strategy has allowed us to look at the fate of a broad spectrum of mtDNA variation, and we report evidence of strong purifying selection in the mouse female germ line. All of the generated mtDNA mutator mouse lines showed the same strong reduction in nonsynonymous substitutions, exemplified by the reduction in first and second codon position mutations. This pattern is also seen in human populations and implies that purifying selection has a similar, drastic impact on the mtDNA variation in humans despite different demographic histories. The RNA genes, in contrast, appear to accumulate at levels approximating the synonymous third codon positions. These mutations are expected to eventually raise to high-enough levels and lead to impaired mitochondrial function in a manner similar to the threshold effect seen in human mtDNA disease. The data generated from this experimental model will allow us to build more accurate molecular models of mtDNA evolution and aid the understanding of inheritance patterns of human mtDNA disease mutations.