The oxidative stress theory of aging - concepts and limitations

What do we know about the involvement of oxidative stress in disease?
Definitely not enough. The speakers at MiP-School London reminded us that researchers use the term "ROS" as explanation to hide behind with little explanatory value. Berry Halliwell (3,4) and Mike Murphy (6) gave fascinating talks on this topic. I would like to add a more biogerontologic perspective.

A deep mechanistic understanding of how ROS regulate lifespan might be called for, but I don't think we should over-emphasize detailed understanding in the field of aging as a whole. Gerontology is difficult. There are enough basic questions still unanswered. (Is calorie restriction a universal mechanism? Are current models of mouse lifespan extension redundant/convergent? etc.)

Historically, progress has come from rather crude phenotypic screens. Calorie-restriction was discovered before the Second World War, using mice without fancy molecular biology. To me, the whole Interventions Testing Programme is one big, naive phenotypic screen (readout = dead mice). The rationale for some of the tested compounds is almost cringe-worthy. Guess what? It's the most successful programme in the world, yielding multiple promising leads. Particularly, Rapamycin. To propose a test of rapamycin as a CR-mimetic we do not need to understand the underlying biology in any detail. Unsurprisingly, the first data on the mTOR pathway was generated using simple phenotypic screens in yeast and 6 years after the rapamycin lifespan study in mice, we still do not understand the mTOR pathway very well. The big invention by the ITP, and others like Spindler, were perfect animal husbandry combined with a modicum of understanding.

Given how little we understand about aging, the naive phenotypic screen will remain a worthwhile strategy. Knowing "just enough" is often quite useful. However, the naive approach to the oxidative stress theory of aging has produced conflicting results. Is it time to re-consider some of the basics?

What is the problem with the oxidative stress theory of aging?
The dilemma is simple: oxidative damage accumulates with age and long-lived species resist such damage. Nevertheless the overwhelming majority of interventional studies using knockouts or overexpression of antioxidants show no effects on lifespan.

In this blogpost I want to provide some background and discuss the limitations of our methods. My focus will be on the mitochondrial oxidative stress theory, but the ideas can apply more broadly. It's not as simple as proclaiming the theory dead, since this would fail to explain the comparative data.

I. Limitations of mouse lifespan studies: are they able to prove or refute the oxidative stress theory?
Let's stipulate oxidative stress is a driver of cancer and hence long-lived organisms must develop mechanisms to reduce such damage. This is quite plausible, but could it be shown using the gold standard, lifespan studies? Probably not. The argument is straight-forward, yet I have had problems putting it into words and finding mention of it in the literature (maybe these two are connected). To quote a colleague on this (7b):

… taking an organism that is aging normally and slowing or even stopping just ONE mechanism of aging in an otherwise normally-aging organism ought not to have much effect on lifespan or overall age-related health, because all the other mechanisms are still proceeding mostly apace. Per contra, when you take an animal that has already been pushed out of its normal, adaptive homeostatic poise and has an induced, abnormal slow-aging metabolic state like CR or Ames dwarves, and then interfere with one of its mechanisms of slowed aging, you might expect more of a 'weakest link in the chain' effect....
However, now that I think about it, the idea effortlessly follows from the demographics of aging. This problem has been known since the 70s and is called Taeuber paradox (7a): "A cure for cancer would only have the effect of giving people the opportunity to die of heart disease." As far as human lifespan is concerned, the net gain from eradicating cancer is at most 4 years of life expectancy (roughly 5% of mean lifespan).

A thought experiment, imagine rodents behaved similarly (they may not) and our neat antioxidant intervention cuts down cancer mortality by 50% or it reduces another specific age-related pathology by a similar margin. That's extremely powerful. But can we detect the resulting 2-3% change in mean lifespan? Optimistically, lifespan studies with n~40 are only powered to detect changes in the range of 10-30%.

Why then are there any studies showing mouse lifespan extension? Taeuber's paradox would predict that successful life extension studies must have addressed multiple causes of aging. Indeed, the best validated models, e.g. calorie or methionine restriction, dwarfism or rapamycin treatment all have multiple effects contributing to lifespan extension and converge on a few very large pleiotropic pathways. Usually this is GH/IGF-1/mTOR signalling. The converse case should be illuminating. Are there any successful "single factor" interventions? Not to my knowledge. The study on mitochondrially targeted catalase is promising, but the controls were a little short-lived and it was never replicated. (CAVE: alterations in mitochondrial ROS homeostasis may trigger large adaptive changes through retrograde signalling)

Discussion of potential solutions are beyond the scope of this blog post, although, they will certainly involved the study of healthspan, specific pathology, biomarkers as well as loss of function/unmasking study designs (7b).

II. Stress resistance is a side-effect, or a co-variate, of long lifespans
In this case, the comparative studies would suggest an effect and interventional studies should come up empty. So fat this is consistent with the data we have, but I have not seen many testable hypotheses regarding said covariate. It could be any number of things. Perhaps autophagy covaries with stress resistance, and autophagy is essential to prevent the buildup of non-oxidative protein damage with aging (e.g. protein aggregation).

One often discussed covariate is bodymass and BMR, but modern studies usually rule this out by bodymass-adjustment (Lambert 2007; ref. 8) and (Pickering 2014; ref. 9).

III. Spatial considerations: Are antioxidants even theoretically able to prevent oxidative damage to mitochondrial DNA and other vulnerable sites?

Barja 2013:
"the source of mtROSp at the inner mitochondrial membrane is very near or even likely in contact with the initial target for aging, the mtDNA. Such contact ensures that antioxidants will not interfere with the rate of free radical damage generation in mtDNA, because it spatially prevents the interception of the ROS before they can damage the mtDNA. Oxidative damage to mtDNA just at the places of mtROS generation makes longevity dependent on the rate of ROSp but not on the antioxidant levels. "

Kirkwood and Kowald 2012:
"Positional effects are of major importance...Interestingly, it is known that nucleoids, and thus mtDNA, are attached to the inner mitochondrial membrane, which is exactly where most ROS are generated via oxidative phosphorylation. Since ROS are generated at the membrane and diffuse into the mitochondrial matrix, a gradient is created with the highest ROS concentration at the inner membrane and the lowest in the matrix centre... it is clear that any addition of external oxidative stress will have a significant relative effect on ROS steady-state concentrations in shallow areas of the intrinsic ROS gradient, but not on ROS concentrations (and thus damage rates) close to the membrane."

Given this gradient, the concentration of a diffusible antioxidant (protein or small molecule) would always be too low towards the electron transport chain and too high away from it. Put another way, a concentration high enough to protect the mtDNA might be toxic elsewhere, e.g. by interfering with oxidative stress signaling.

Iron - site-directed damage?
Reading some old literature I recognized the same could be true for iron. 1. it seems difficult to intercept Fenton reaction products in all three dimensions, 2. iron may be specifically bound by some biomolecules limiting access of antioxidants (5, 10).

"iron ions cannot exist free in aequous solution...must...bind...the damage due to OH[-radicals] will thus be 'directed onto' the site of iron binding... if ... catalytic iron salts are bound to DNA...[peroxide and superoxide]...will fragment the DNA"
"Hydroxyl radicals in free solution can attack most biological molecules at almost diffusion-controlled rates, but have little or no specificity in Fenton chemistry, they have considerable site-specificity because the hydroxyl radical is formed close to where the iron is located (reviewed by Symons &C Gutteridge)"
Also (10, p. 337):
"Transition metal ions may greatly influence the yields and types of DNA damage due to the production of hydroxyl  radicals in close proximity to the DNA.[10] In fact, the  proximity of these site-specific reactions to DNA targets makes scavenging extremely unfavorable."

The implication: exogenous antioxidants should be relatively ineffective at reducing oxidative stress
Indeed this is often the case (4):
There is an additional explanation of the failure of antioxidants in human intervention trials. We were one of the first groups to discover that supplements of the commonly used antioxidants (ascorbate, β-carotene, α-tocopherol) are generally unsuccessful in decreasing levels of oxidative damage in healthy humans (reviewed in [28], [37], [42] and [43]), and many others have found the same or similar results [199], [200], [201], [202], [203] and [204]. If oxidative damage facilitates disease development and oxidative damage is not decreased by the administered antioxidants, then they will not affect disease occurrence—it is as simple as that [39]! 
1. Updating the mitochondrial free radical theory of aging: an integrated view, key aspects, and confounding concepts.
Barja G.
Antioxid Redox Signal. 2013 Oct 20;19(12):1420-45. doi: 10.1089/ars.2012.5148. Epub 2013 Jul 3. Review.

2. The free-radical theory of ageing--older, wiser and still alive: modelling positional effects of the primary targets of ROS reveals new support.
Kirkwood TB, Kowald A.
Bioessays. 2012 Aug;34(8):692-700. doi: 10.1002/bies.201200014. Epub 2012 May 29.

3. Halliwell, B. (2012). Free radicals and antioxidants: updating a personal view. Nutrition reviews, 70(5), 257-265.

4. Free Radic Biol Med. 2009 Mar 1;46(5):531-42. doi: 10.1016/j.freeradbiomed.2008.11.008. Epub 2008 Dec 3.
The wanderings of a free radical.
Halliwell B.

5. Gutteridge, J. M., & Mitchell, J. (1999). Redox imbalance in the critically ill. British Medical Bulletin, 55(1), 49-75.

6. Biochim Biophys Acta. 2014 Feb;1840(2):923-30. doi: 10.1016/j.bbagen.2013.05.026. Epub 2013 May 30.
Using exomarkers to assess mitochondrial reactive species in vivo.
Logan A1, Cochemé HM, Li Pun PB, Apostolova N, Smith RA, Larsen L, Larsen DS, James AM, Fearnley IM, Rogatti S, Prime TA, Finichiu PG, Dare A, Chouchani ET, Pell VR, Methner C, Quin C, McQuaker SJ, Krieg T, Hartley RC, Murphy MP.

7a. Demography. 1977 Nov;14(4):411-8.
What difference would it make if cancer were eradicated? An examination of the Taeuber paradox.
Keyfitz N.


8. Aging Cell. 2007 Oct;6(5):607-18. Epub 2007 Jun 27. Low rates of hydrogen peroxide production by isolated heart mitochondria associate with long maximum lifespan in vertebrate homeotherms.
Lambert AJ1, Boysen HM, Buckingham JA, Yang T, Podlutsky A, Austad SN,Kunz TH, Buffenstein R, Brand MD.

9. J Gerontol A Biol Sci Med Sci. 2014 Jul 28. pii: glu115. [Epub ahead of print]
Fibroblasts From Longer-Lived Species of Primates, Rodents, Bats, Carnivores, and Birds Resist Protein Damage.
Pickering AM1, Lehr M1, Kohler WJ1, Han ML, Miller RA.

10. Iron and Human Disease. Randall B. Lauffer. 1992.

10b. Chattopadhyaya, R., & Goswami, B. (2012). Oxidative damage to DNA constituents by iron-mediated Fenton reactions: the deoxyadenosine family. Journal of Biomolecular Structure and Dynamics, 30(4), 394-406.

"It was also proposed (Imlay & Linn, 1988; Henle, Luo, Gassmann et al., 1996; Luo et al., 1994, 1996) that DNA damage by the Fenton reaction is mediated by an oxidative agent unlike a freely diffusible [OH radical] due to the Fe2+ being bound to DNA. "

10c. "DNA is a charged molecule that attracts positively charged molecules, like Fe2+; hence, charge-charge interaction brings iron in close proximity to the DNA phosphodiester backbone. The close proximity of Fe2+ to DNA means that HO· generated by Fenton chemistry will likely react with DNA, inducing lethal or non-lethal mutations"

10d. "Another important hypothetical concept regarding iron-catalyzed oxidative damage was that of a ‘site-specific’ mechanism that was raised in the 1980s. Fe(III) bound loosely to biological molecules such as DNA and proteins may undergo cyclic reduction and oxidation."