Krakauer and Mira reply

Many of the issues raised by Perez et al. concern molecular mechanism rather than evolutionary function. Thus ‘cause’ does not carry the same meaning in evolution as in physiology. The evolutionary cause or function of sexual reproduction or two-step meiosis can be seen as a means of fending off parasites or selfish genetic elements: the developmental or mechanistic causes of these are different and involve precise chemical and genetic processes. We propose that the evolutionary function of atresia has been to eliminate mildly deleterious mutations from mitochondria, thereby retarding Müller's ratchet1. It is not necessary for all of the oocytes undergoing apoptosis in each generation to contain defective mitochondrial genomes, much as individuals do not need to be teeming with parasites or sister-killer mutants to engage in sex, or for cells to undergo meiosis. Once these adaptations arise, organisms are often committed to their implementation2.

We suggest that the long-term evolutionary persistence of atresia requires that germ cells with healthy mitochondria represent statistically superior competitors in the ovary and are thus more likely to escape cell death to seed successive generations. The result of this competition is to eliminate the worst cells (when present) and retain a sufficiently large population of cells to produce the maximum number of offspring. This is rather like a game of musical chairs, where the worst competitors are eliminated early on and there is only one winner among many equally matched contestants. Thus many egg cells compete for survival factors through metabolism, but only a few will be successful. The mechanism remains unclear, although oocyte elimination must involve initiation of apoptosis by some somatically derived signal.

A key part of our theory is that competition for survival factors be promoted among germ cells carrying a small number of mitochondrial genomes. Cells with large numbers of defective mitochondria can metabolize at almost 100% efficiency, providing at least 10% of wild-type mitochondrial genomes are present3 (the ‘threshold effect’), so selective differences among cells only become apparent when wild-type genomes are rare in the germ cell. This is the function of the bottleneck — it decreases mitochondria to a level that makes the functional variance among cells selectively detectable (Fig. 1b of ref. 1).

The results of Perez et al. agree with our hypothesis and match our comparative data, showing that the number of mitochondria present during the bottleneck can predict the level of atresia. Both experimental and comparative data indicate that small numbers of healthy mitochondria can inhibit atresia, which in our view is evidence that mitochondria can indirectly influence oocyte fate. Mitochondrial genome-deletion experiments confirm the importance of genome quality for mitochondrial persistence.

Perez et al. suggest that atresia in the next generation is redundant, as the founder population of oocyte mitochondria have previously experienced selection. What is important selectively is the stage in development when mitochondria are sequestered into the oocytes and the reproductive lifespan of the mother. Assuming that all mitochondria in the oocyte at reproductive maturity are a subset of those present in pre-atretic fetal development (not the case in all species), then mutations could accumulate (through division or the effects of environmental mutagens) from the time of atresia up until the formation of the next-generation fetus. This can be between 13 and 40 years in humans. There is good evidence that ontogenetic mutations accumulate in mitochondria to reconstitute polymorphism4, which might contribute by a similar mechanism towards the ongoing postnatal atresia discussed by Perez et al. In the absence of a germline bottleneck, ontogenetic ageing of mitochondria should proceed more rapidly.

As Perez et al. point out, many more cells are eliminated during early atresia than might plausibly contain defective mitochondrial genomes. This is surprising, but selection may incur some continuous or facultative response proportional to the mutational load per generation. As we noted earlier, however, this is not a feature of mammalian meiosis or sex that is obligate for development. The adaptively salient outcome of atresia is that it retains sufficient eggs for the female's reproductive lifetime. This might simply be a matter of insurance: evolution cannot always discover the globally optimal solution.

We have offered a purely functional theory of atresia, one that has the merit of addressing an evolutionary puzzle, Müller's ratchet in mitochondria. We believe that evolution (function) and molecular biology (mechanism) must work together in this area.

See also — Perez et al.