A proposed refinement of the mitochondrial free radical theory of aging

A proposed refinement of the mitochondrial free radical theory of aging
Aubrey D. N. J. de Grey

Summary
 

Over recent years, evidence has been accumulating in favour of the free radical
theory of aging, first proposed by Harman(1,2).

Despite this, an understanding of the mechanism by which cells might succumb to the effects of free radicals has proved elusive.

This paper proposes such a mechanism, based on a previously unexplored hypothesis for the proliferation of mutant mitochondrial DNA:

that mitochondria with reduced respiratory function, due to a mutation or deletion affecting the respiratory chain, suffer less frequent lysosomal degradation, because they inflict free radical damage more slowly on their own membranes.

mi·to·chon·dri·on   Pronunciation Key  (mt-kndr-n) n. pl. mi·to·chon·dri·a (-dr-) A spherical or elongated organelle in the cytoplasm of nearly all eukaryotic cells, containing genetic material and many enzymes important for cell metabolism, including those responsible for the conversion of food to usable energy. Also called chondriosome.   Source: The American Heritage® Dictionary of the English Language, Fourth Edition Copyright © 2000 by Houghton Mifflin Company. Published by Houghton Mifflin Company. All rights reserved. Karl Simplification:  mitochondria is the name given to that part of a cell which converts food into energy which is then stored inside the cell and used inside the cell for the purposes of cell function, and ultimately that means for body function.

ly·so·some   Pronunciation Key  (ls-sm) n. A membrane-bound organelle in the cytoplasm of most cells containing various hydrolytic enzymes that function in intracellular digestion.     Source: The American Heritage® Dictionary of the English Language, Fourth Edition Copyright © 2000 by Houghton Mifflin Company. Published by Houghton Mifflin Company. All rights reserved. Karl Simplification:  Cells in the body need energy for their own activity, and that energy comes by way of oxygen coming into the cell, combining with the sugar (glucose) which also comes into the cell.  These two items are "burned" to create energy (heat stored, actually, as a type of electrical energy).  The "Lysome" is a part of the cell that helps in this process.

Karl Loren:  Here is my simplified understanding of the above. Inside the human cell is something called the "mitochondria."  There are actually several of these inside each cell. Their job is to control the production of energy inside the cell. There are parts of the cell which contain the "blueprint" or "instructions" on how the cell functions and how it reproduces.  When and as this part gets damaged by a random free radical, the damage might be to alter the normal function of the larger part in which it exists. Thus, there is

Once such a mutation occurs in a mitochondrion of a non-dividing cell, therefore, mitochondria carrying it will rapidly populate that cell, thereby destroying the cell's respiratory capability.

The accumulation of cells that have undergone this transition results in aging at the organismal level.

The consistency of the hypothesis with known facts is discussed, and technically feasible tests are suggested both of the proposed mechanism and of its overall contribution to mammalian aging.

Introduction
 

In 1956, Harman(1) published the first version of the free radical hypothesis of aging, which will be the topic of this paper. His idea was that a major mechanism of aging might be the accumulation of damage to macromolecules, inflicted by free radicals (that is, molecules which carry an unpaired electron). Subsequently he extended the idea(2) with the proposal that mitochondria, which had by then been shown to be a major site of production of oxygen free radicals(3), were also the major site of free radical damage causing aging. The following observations are among those supporting this proposal: (a) that the pigment lipofuscin, which is believed to be the indigestible remnant of phagocytosed cellular detritus, seems mainly to derive from mitochondrial membranes(4); (b) that the most harmful free radicals must, on thermodynamic grounds, do most of their damage very near to their site of production(5); (c) that aged muscle has reduced mitochondrial activity(6,7,8); (d) that, to a first approximation, longevity of different homeotherms varies inversely with their metabolic rate(9); (e) that species which live longer than their metabolic rate would predict suffer unusually little mitochondrial lipid peroxidation, and their mitochondrial membranes also possess unusually few highly unsaturated fatty acids(10). (These last results are linked because oxygen free radicals only attack lipids' C=C bonds, not C-C bonds.)

To this list should be added an item not explicitly supportive of the above idea, but on which the present paper will draw heavily: that mitochondria replicate more frequently than their host cells(11,12). The extreme case is that of cell types which never divide, such as in heart, skeletal muscle and brain; in rats, mitochondrial generation time even in these tissues is under a month(12). This does not result in ever-increasing numbers of mitochondria per cell, so we can infer that mitochondria are also broken down at a similar rate. Why is this turnover so rapid? Since mitochondrial biogenesis is an expensive task, this is a valuable question to explore.

The remainder of this paper falls into three parts: (i) a summary of the three main subsequent ideas that have led to one current version of the free radical hypothesis; (ii) a hypothesis, which accepts the first and second of those ideas, suggests an alternative to the last, and extends it; and (iii) a discussion of the plausibility and testability of that hypothesis.

Idea 1:

Mitochondrial decline results from increasing mtDNA dysfunction Mitochondria are alone among mammalian subcellular organelles in possessing their own DNA (mtDNA), present in a few copies per mitochondrion. It was not long before researchers began to ask whether mtDNA mutations play the central role in the mitochondrial decay that Harman had proposed. Several items of evidence suggest this: (a) mammalian mtDNA is far more vulnerable to mutation than nuclear DNA, because (among other reasons) it has a repeated risk of replication error, which ceases for nuclear DNA of a postmitotic cell, and it is closer to the main source of mutagenic free radicals; (b) mt-coded genes are essential in the respiratory chain and mitochondrial ATP synthesis, but are unnecessary for mitochondrial or mtDNA replication, so a mutation can theoretically become homozygous in one or many mitochondria, through repeated replication and division; (c) no mt-coded proteins or mRNAs are transferred between mitochondria, so each mitochondrion's respiratory ability depends on its own mtDNA; (d) mtDNA deletions and point mutations are both found to accumulate with age in humans(13,14). (A recent, conflicting report(15) regarding point mutations is noted below in relation to a proposed sequence analysis experiment.)
 

Idea 2:

Increasing mtDNA dysfunction results from preferential amplification of mutations For researchers who pursued the idea that free radicals cause mitochondrial decline by damaging the mtDNA, a natural next question was: does mitochondrial "competition" play a role in this process? The mitochondria of a single cell are an isolated population undergoing replication and destruction, so there is the opportunity for selective pressure to exist, and supportive evidence has been reported(16). An alternative hypothesis(17) is that initial mutations cause a tolerable but significant reduction in the cell's ATP synthesis capacity, which accelerates the mtDNA mutation rate, so that eventually all mtDNA suffers deleterious mutations. Subsequent work lends weight to the selection-based hypothesis(18). The mitochondria of a single respiration-deficient muscle fibre are found all to carry the same mutation, and different fibres carry different mutations. If the means whereby a cell succumbs to mtDNA mutations were by accelerated mutation rate, we should see the same range of mutations in each affected cell, irrespective of which mutation occurred first in a given cell.
 

Idea 3:

Preferential amplification of mutations results from their preferential replication Selection for a deleterious mtDNA mutation can, formally, be achieved in either (or both) of two ways:
 

preferential replication of mutant mitochondria,

or preferential lysosomal digestion of non-mutant ones.
 

Hitherto, researchers have focused on the preferential replication option, but great difficulty has been encountered in identifying a model, consistent with the evidence, for how such an advantage might accrue(19,20).

The preferential digestion option is the alternative which underlies the hypothesis that will now be outlined.
 

The hypothesis not only seems consistent with known facts (a question explored in detail in the "plausibility" section below), but also addresses why, in the first place, cells engage in the expensive process of frequent mitochondrial turnover.
The hypothesis: survival of the slowest

The peroxidation and polymerisation of the lipid molecules of the inner mitochondrial membrane by free radicals is substantially not repaired. A respiring mitochondrion will therefore accumulate such damage to its inner membrane. In due course, the membrane will become unable to perform its main function, which is the maintenance of the proton gradient created by the respiratory chain. This will not affect the operation of the respiratory chain itself, however, only the production of ATP, so damage will continue. It will eventually be so severe that proteins escape from the mitochondrion: the outer membrane will not suffice to prevent this, since

(a) it may itself be somewhat damaged, and

(b) even if it is not, the smaller mitochondrial proteins can pass
through its pores.

Many such proteins may be toxic to cytosolic processes -- they were constructed in the cytosol, but at that time they were rendered inert by the presence of a leader peptide, which will by now have been cleaved. The cell removes these toxins at source, by lysosomal degradation of the damaged mitochondrion.
 

Therefore, in order to have any mitochondria left in the long term, the cell must avert the above process by maintaining the degree of contamination of its mitochondrial membranes at a stable level. It achieves this by degrading some of its mitochondria and replicating others. This works because the degraded membrane is recycled, and the new membrane (added to the parent mitochondrion in order to bring it to a size ready to divide) is pristine. Turnover thus acts to dilute the existing membrane damage.

It is proposed that this turnover is driven by the toxicity hypothesised above. Mitochondria accumulate damage until they become poisonous, and are then digested. Replication occurs when the cell detects a shortage of ATP, caused directly by the diminished numbers of mitochondria. Of necessity, the mitochondria that are replicated are those that have not already been digested.
 

This situation is stable while all mitochondria are fully functional. At some point, however, a mtDNA mutation will inevitably occur that lowers the respiratory capability of its host mitochondrion. That mitochondrion's lower level of respiration will translate into a lower concentration of harmful free radicals in its immediate environment. This will result in a slower accumulation of damage to its inner membrane than is occurring in properly respiring ones. Such a mitochondrion will therefore still be intact when many of the cell's non-mutant mitochondria have succumbed to the degradation process hypothesised above. Thus, it will be preferentially replicated. Repetition of this process will rapidly divest the cell of all its properly respiring mitochondria (Fig. 1).
 

Dividing cells must replicate their mitochondria at least once per cell division, so as to maintain the number of mitochondria per cell. If cellular division is rapid enough, there will thus be no significant digestion of mitochondria, because membrane damage is diluted as fast as it is inflicted. Aging at the organismal level will therefore not result from this mechanism in tissues composed of such cell types. In non-dividing or rarelydividing cell types, however, there will be an accumulation of ATP-deficient cells leading to aging at the organismal level.

How plausible is this hypothesis?
 

The compatibility of this hypothesis with known facts is now discussed, by analysing several potential counterarguments.
 

Antioxidant drugs

A difficulty that has dogged the free radical hypothesis from early days is the very modest increase in maximum lifespan which results, in mammals, from antioxidant therapy(2). Recent data give a hint of why this may be. Simultaneous up-regulation of two major antioxidant enzymes has been found to bestow longevity benefits in flies(21), substantially greater than has yet been achieved in mammals; but previous studies from the same laboratory found that up-regulation of either one of those enzymes without the other has little such effect (noted in ref. 21). The authors point out that an organism's antioxidant system, viewed as a whole, comprises a subtle interplay of components, whose balance may be crucial to the system's overall efficacy. This system is much more complex in mammals than in insects (for example, insects entirely lack glutathione peroxidase(22)), so the destabilising effect in mammals of introducing a single antioxidant drug may mask that drug's direct benefit; it may be very difficult to improve on the defences that mammals already possess.

Recessiveness of mutations
 

A second argument is genetic: it concerns the polyploidy of mitochondria. A single mutated copy of mtDNA in a mammalian mitochondrion would reduce its copy number of wildtype DNA only by 10-20%; would this really affect its respiratory capacity? Indeed, the effect on respiration rate of a single-copy mutation should be slight. There are two problems with this challenge, however. Firstly, a small proportion of mutations will
survive long enough to become homozygous in at least one mitochondrion, purely through genetic drift, so even complete recessiveness would not stop the process from taking hold eventually. Secondly, even a very slight slowing of respiration should be enough to activate the above mechanism, since the relationship between respiration rate and chance of survival will not be linear: in fact, to a first approximation it will be a step function, as the mitochondrion only needs to survive that little bit longer than its sisters before being rescued by damage-diluting division.

Membrane repair in situ
 

A further potential challenge to this hypothesis focuses on its initial assertion -- that membrane damage is substantially not repaired. Indeed, it has been shown(24) that phospholipase A2 (EC 3.1.1.4), which detaches a fatty acid from the glycerol of a phospholipid, preferentially targets the more oxidised regions of membranes. This detached fatty acid can then be freed from the membrane for degradation or re-reduction elsewhere, while the lysophospholipid that remains is reacylated by the attachment of a new, unoxidised fatty acid. Other processes may exist that excise (or re-reduce in situ) various products of lipid peroxidation in other ways. However, all such mechanisms can be only partial solutions, because they must rely, at root, on the recognition of particular species or families of damaged molecules. The huge variety of peroxidation products will always include inter-linked and intra-linked molecules that are intractable to any of the available selective processes. Such molecules will only succumb to the blunderbuss approach applied inside a lysosome -- an approach that would be fatal if attempted in the cytosol. They will thus inevitably accumulate during a mitochondrion's life, even if other peroxidation products do not.

Free radical production vs. free radical damage
 

Another challenge concerns whether a respiration-deficient mitochondrion will in fact have a diminished exposure to free radical damage. This is not obvious, because certain classes of blockage of the respiratory chain may actually increase the production of one reactive oxygen species, superoxide. This is because the leakage of electrons from some of the enzyme complexes rises when they are in a very reduced state, as those upstream of the defect will be(25). However, the thermodynamics of lipid peroxidation, which are well worked out, indicate that superoxide concentration per se is not relevant. Three major pathways for initiation of lipid peroxidation are thought to occur, as summarised in Fig. 2; the reader is referred to ref. 5 for a thorough overview and to refs. 26-30 for the specifics of certain reactions. It will be seen that superoxide only initiates lipid peroxidation when protonated to become perhydroxy radical, a reaction whose rate varies with the acidity of the medium near the membrane and is very low at cytosolic pH(5,26). Respiration, however, creates an excess of protons -- that is, low pH -- on the outside of the inner membrane. Thus, perhydroxy levels will necessarily be lower near a mitochondrion that is respiring less. Another major pathway for initiating lipid peroxidation is controlled by the availability of ubisemiquinone, which will likewise be diminished in respiration-deficient mitochondria, in favour of a predominance of either ubiquinone (if the respiratory chain components downstream of it are fully functional) or ubiquinol (if not).
 

The only pathway that may be unaffected is the metal-catalysed one, since the relatively stable H2O2 intermediate has time to diffuse from far away so should be locally undiminished. Even this may not be so:

when ubiquinol is high, all pathways may in fact be attenuated by ubiquinol's capacity to reduce lipid peroxyl radicals(27); conversely, when ubiquinone is high, complex III will be starved of electrons so less able to leak them to O2.
 

How can the hypothesis be tested?

Since the ideas presented here apply most forcefully to non-dividing cells, D. melanogaster and C. elegans might be thought suitable model organisms for testing this hypothesis, given that almost all cells in adults are postmitotic. However, there is reason for caution. The mechanism presented here can only have an effect if there are enough mitochondrial generations in the lifetime of the organism to allow a cell to be taken over by copies of a mutant mtDNA molecule. Minimally, this is the logarithm to base 2 of the number of
mitochondrial genomes per cell. Since both the above species live only a few weeks, a sufficiently short mitochondrial lifetime seems unlikely; a greater role may be played by the simple efficacy of antioxidant enzymes, as recent work(21) suggests. Therefore, before one embarks on tests of whether the process hypothesised in this paper plays a role in the aging of such short-lived organisms, there should be a measurement of their rate of mitochondrial turnover, to establish whether the process can occur at all.
 

Experiments on dividing tissues may be able to test this hypothesis, so long as conditions are used in which the cellular generation time considerably exceeds the mitochondrial. Chambers and Gingold found in 1986 that suppressiveness in yeast, which is a notably similar phenomenon to the process discussed here (in that it entails the preferential amplification of mitochondria mutant or deficient for respiratory chain genes), is often not associated with faster replication of the mutant mtDNA than the wild-type(31). They proposed an explanation based on non-random segregation in the asymmetric division of a yeast zygote, but preferential degradation of non-mutants would also explain this observation. In the same paper, they also reported (data not shown, but confirmed by Gingold, pers. comm.) that starvation of a yeast culture markedly increased the levels of
suppressiveness which it exhibited thereafter, independent of genetic factors. If mitochondria continue to turn over during starvation, then lysosomal degradation of mitochondria must also continue, so a mechanism such as proposed here would indeed increase suppressiveness. It would be valuable to repeat both the above experiments with human tissue culture cells, to discover whether the suppressive behaviour already reported in them(16) shares these features.
 

A different type of test involves sequence analysis. Scant attention has been paid in this paper so far to complex V, the ATP synthetase complex (EC 3.6.1.34). Two of its components are mitochondrially coded.
 

If one of them suffers a mutation, will it be selected for? This hypothesis predicts that it will not, in view of the pathways initiating lipid peroxidation that were discussed above and are depicted in Fig. 2. The proton gradient across the inner membrane of a mitochondrion with an ATP synthetase mutation (but with an intact respiratory chain) will presumably be undiminished, so just as much lipid peroxidation will be initiated by perhydroxy radical. The rate of respiration may be reduced (if the mutation blocks the passage of protons into the mitochondrion), but this slow-down would be uniform along the whole respiratory chain, so the proportion of coenzyme Q that is in the ubisemiquinone state should be unchanged, so its contribution to lipid damage (via reaction with H2O2) should also be undiminished. Therefore, this hypothesis predicts that mutations local to the ATP synthetase subunits will not be selected for, and thus will not accumulate with age (unless they somehow block their own transcription: this would knock out other genes too, since the whole mtDNA is transcribed as one primary transcript of each strand), whereas mutations in all other mitochondrially coded genes will accumulate. This test is particularly strong, because various other theories of mitochondrial free radical damage predict a uniform distribution of mutations across all genes. While this paper was in review, a careful study was published(15) which provisionally confirms the above prediction; however, a systematic test of it is still motivated.

How can the effect on aging be tested?
 

One experiment, of which a variant has already been suggested by Hoeben(32), stands out as uniquely direct. It is undoubtedly very ambitious, but the past decade has brought us a great wealth of experimental evidence on all aspects of aging at the cellular and macromolecular levels, so its time may have come. It is to transfect mice with copies of all 13 protein-coding genes of the mtDNA, thereby incorporating them into the nuclear genome, where they would (as noted earlier) enjoy a hugely greater degree of protection from mutation. This hypothesis predicts that mice with these transgenes will, once pitfalls such as those noted below have been circumvented, exhibit dramatically slowed aging of muscles, nervous tissue and other non-dividing or rarelydividing cell types.
 

This would be a big project, but it is difficult to dismiss on grounds of practicality.

The disparity of genetic codes can be corrected by base-pair substitutions, and the regulation of the products' expression and transport to mitochondria can be addressed by mimicking the nuclear-coded subunits of the corresponding enzyme complexes. Nor can the treatment be dismissed as inevitably having harmful side-effects. The fact that we retain mtDNA does not prove that we need to, since the differences of genetic code make gene transfer during evolution extremely difficult. Problems of stoichiometry may arise, due to simultaneous expression of the mtcoded and transgenic copies of genes, but this can if necessary be tackled by, for example, disruption of nuclear-coded components of the mitochondrial transcription and translation machinery. In view of this, such a project may realistically be able to create mice whose mtDNA is both superfluous and harmless to them.

Conclusion
 

This paper has presented a hypothesis that extends a long-standing and prominent theory of aging. Several challenges to the idea's plausibility have been anticipated and explored, and tests of it proposed. The last of these tests is a highly ambitious project, which many will consider wildly premature, but which, if and when even provisionally successful, would profoundly alter our view of the inevitability of old age.

Acknowledgements

I thank Bernhard Kadenbach for his encouragement and support during the refinement of the hypothesis presented here, and Adelaide Carpenter, Bernard Daunas, Jonathan Ewbank, Peter Hoeben and Rachel Drysdale for critical reading of the manuscript. Dr. Carpenter also helped design the figures.

References
 

1 Harman, D. (1956). Aging: a theory based on free radical and radiation chemistry. J. Gerontol. 11, 298-300.

2 Harman, D. (1972). The biologic clock: the mitochondria? J. Am. Geriatr. Soc. 20, 145-147.

3 Pullman, B. (1963). Electronic aspects of biochemistry. Academic Press, New York.

4 Brunk, U. T. (1989). On the origin of lipofuscin; the iron content of residual bodies, and the relation of these organelles to the lysosomal vacuome. A study on cultured human glial cells. Adv. Exp. Med. Biol. 266, 313-320.

5 Kappus, H. (1985). In Oxidative Stress (ed. H. Sies), pp. 152-195. Academic Press, New York.

6 Syrovy, I. and Gutmann, E. (1977). Changes in speed of contraction and ATPase activity in striated muscle during old age. Exp. Gerontol. 12, 31-35.

7 Muller-Hocker, J. (1989). Cytochrome-c-oxidase deficient cardiomyocytes in the human heart - an age-related phenomenon. A histochemical ultracytochemical study. Am. J. Pathol. 134, 1167-1173.
 

8 Boffoli, D., Scacco, S. C., Vergari, R., Solarino, G., Santacroce, G. and Papa, S. (1994). Decline with age of the respiratory chain activity in human skeletal muscle. Biochim. Biophys. Acta 1226, 73-82.
 

9 Schmidt-Nielsen, K. (1970). Energy metabolism, body size, and problems of scaling. Fed. Proc. 29, 1524-1532.
 

10 Pamplona, R., Prat, J., Cadenas, S., Rojas, C., Perez-Campo, R., Torres, M. L. and Barja, G . (1996). Low fatty acid unsaturation protects against lipid peroxidation in liver mitochondria from long-lived species: the pigeon and human case. Mech. Ageing Dev. 86, 53-66.
 

11 Gross, N. J., Getz, G. S. and Rabinowitz, M. (1969). Apparent turnover of mitochondrial deoxyribonucleic acid and mitochondrial phospholipids in the tissues of the rat. J. Biol. Chem. 244, 1552-1562.
 

12 Menzies, R. A. and Gold, P. H. (1971). The turnover of mitochondria in a variety of tissues of young adult and aged rats. J. Biol. Chem. 246, 2425-2429.
 

13 Munscher, C., Muller-Hocker, J. and Kadenbach, B. (1993). Human aging is associated with various point mutations in tRNA genes of mitochondrial DNA. Biol. Chem. Hoppe Seyler 374, 1099-1104.
 

14 Kadenbach, B., Munscher, C., Frank, V., Muller-Hocker, J. and Napiwotzki, J. (1995). Human aging is associated with stochastic somatic mutations of mitochondrial DNA. Mutat. Res. 338, 161-172.
 

15 Pallotti, F., Chen, X., Bonilla, E. and Schon, E. A. (1996). Evidence that specific mtDNA point mutations may not accumulate in skeletal muscle during normal human aging. Am. J. Hum. Genet. 59, 591-602.
 

16 Yoneda, M., Chomyn, A., Martinuzzi, A., Hurko, O. and Attardi, G. (1992). Marked replicative advantage of human mtDNA carrying a point mutation that causes the MELAS encephalomyopathy. Proc. Natl. Acad. Sci. USA 89,
11164-11168.
 

17 Arnheim, N. and Cortopassi, G. (1992). Deleterious mitochondrial DNA mutations accumulate in aging human tissues. Mutat. Res. 275, 157-167.
 

18 Muller-Hocker, J., Seibel, P., Schneiderbanger, K. and Kadenbach, B. (1993). Different in situ hybridization patterns of mitochondrial DNA in cytochrome c oxidase-deficient extraocular muscle fibres in the elderly.
Virchows Arch. (A) 422, 7-15.
 

19 Gadaleta, M. N., Rainaldi, G., Lezza, A. M., Milella, F., Fracasso, F. and Cantatore, P. (1992).  Mitochondrial DNA copy number and mitochondrial DNA deletion in adult and senescent rats. Mutat. Res. 275, 181-193.
 

20 Wallace, D. C., Bohr, V. A., Cortopassi, G., Kadenbach, B., Linn, S., Linnane, A. W., Richter, C. and Shay, J. W. (1995). In Molecular Aspects of Aging (ed. K. Esser and G.M. Martin), pp. 199-225. John Wiley & Sons, Chichester, UK.
 

21 Orr, W. C. and Sohal, R. S. (1994). Extension of life-span by overexpression of superoxide dismutase and catalase in Drosophila melanogaster. Science 263, 1128-1130.
 

22 Smith, J. and Shrift, A. (1979). Phylogenetic distribution of glutathione peroxidase. Comp. Biochem. Physiol. B 63, 39-44.
 

23 Clayton, D. A., Doda, J. N. and Friedberg, E. C. (1974). The absence of a pyrimidine dimer repair mechanism in mammalian mitochondria. Proc. Natl. Acad. Sci. USA 71, 2777-2781.
 

24 Sevanian, A. and Kim, E. (1985). Phospholipase A2-dependent release of fatty acids from peroxidized membranes. J. Free Radic. Biol. Med. 1, 263-271.
 

25 Sohal, R. S. and Brunk, U. T. (1992). Mitochondrial production of pro-oxidants and cellular senescence. Mutat. Res. 275, 295-304.
 

26 Bielski, B. H., Arudi, R. L. and Sutherland, M. W. (1983). A study of the reactivity of HO2/O2- with unsaturated fatty acids. J. Biol. Chem. 258, 4759-4761.
 

27 Frei, B., Kim, M. C. and Ames, B. N. (1990). Ubiquinol-10 is an effective lipid-soluble antioxidant at physiological concentrations. Proc. Natl. Acad. Sci. USA 87, 4879-4883.
 

28 Cadenas, E. and Boveris, A. (1980). Enhancement of hydrogen peroxide formation by protophores and ionophores in antimycin-supplemented mitochondria. Biochem. J. 188, 31-37.
 

29 Nohl, H. (1990). Is redox-cycling ubiquinone involved in mitochondrial oxygen activation? Free Radic. Res. Commun.  8, 307-315.
 

30 Nohl, H., Gille, L., Schonheit, K. and Yang, L. (1996). Conditions allowing redox-cycling ubisemiquinone in mitochondria to establish a direct redox couple with molecular oxygen. Free Radic. Biol. Med. 20, 207-213.
 

31 Chambers, P. and Gingold, E. (1986). A direct study of the relative synthesis of petite and grande mitochondrial DNA in zygotes from crosses involving suppressive petite mutants of Saccharomyces cerevisiae. Curr. Genet. 10, 565-571.
 

32 Hoeben, P. (1993). Possible reversal of ageing and other mitochondrial deficiencies through retroviral transfection of mitochondrially encoded proteins to the nucleus. Med. Hypotheses 41, 131-133.
 

Aubrey de Grey is at the Department of Genetics, University of Cambridge, Downing Street, Cambridge CB23EH, UK. E-mail: ag24@gen.cam.ac.uk
 

Fig. 1. Overview of the hypothesis. Once a mutation arises, it may become fixed; if it does, it is amplified by rarer than average lysosomal degradation, to the detriment of the cell's overall supply of ATP. Red circles:
mtDNA molecules. L: lysosome. N: nucleus. Black spots: membrane lesions. Crosses: mtDNA mutations.

Blue dots: ATP density.
 

Fig. 2. Principal paths of unpaired electrons from the respiratory chain to lipids. The steps marked "*" are those most certain to be attenuated by the localised chemical changes in and near the inner membrane of a
non-respiring mitochondrion. Only the step marked "*!*" may be amplified.