In this paper, the author argues for greater emphasis to be placed on identifying rate-limiting processes in aging, here termed "flux-controlling" processes. One can tinker with various aspects of cellular metabolism connected to any one given molecule or class of molecules, and do so in many different ways, but any given approach may or may not interact with a rate-limiting step. If it doesn't, then the outcome will not tell us all that much about whether or not this molecule, this process, is important in aging.
Let us imagine that Gustav Embden (1874-1933), one of the ingenious discoverers of glycolysis, would have had modern transgenic techniques at hand and intended to use them to investigate the role of phosphoglycerate kinase (PGK) in the biochemical degradation of glucose to pyruvate. He would have probably overexpressed the enzyme 10x first, and he would not have seen any relevant change in the rate of pyruvate formation in the perfused working heart, whereas the addition of insulin would have shown a clear effect. He then would have generated 90% knockdown animals and again would not have seen any decrease in the rate of glycolysis. Hence, he would have confidently concluded that PGK was not involved in glycolysis. Thus, he would have arrived at an overtly wrong conclusion (merely hypothetical; sorry, Gustav!).
In essence, this is what we do today when we conclude from unsuccessful overexpression or knockdown studies of antioxidant enzymes that free radicals were not involved in aging. We arrive at a wrong conclusion.
What is the mistake here, and what did Embden and his successors do better? First, they looked at the intrinsic chemical logic of the overall system. This should also be done in the study of aging. In particular, they recognized that steps can be involved and essential in a causal chain of (chemical) events even without ever being rate-limiting (or "flux-controlling") for the overall passage through the chain of events. This principle applies to linear chains, branched chains, branching-converging chains and even cyclic chains. Because aging certainly represents an arrangement of causally chained elementary steps (of whatever type and complexity), the decisive point will be to identify the flux-controlling steps of aging as narrowly as possible and then determine their control coefficients for the overall process.
Thus, the only thing we can learn from the fact that superoxide dismutase (SOD) modulation does not influence lifespan is that superoxide degradation is not flux-controlling for aging (in mice). This is still a valuable conclusion, even if it may not be particularly surprising: flux control is usually exerted by low-level, low-efficiency, or highly regulated enzymes, none of which applies to SOD. Moreover, if simple overexpression of SOD indeed would have had a measurable effect on lifespan, one might wonder why evolution has not yielded such a parsimonious solution before. Hence, it is quite unlikely that any isolated enzyme overexpression approach will ever substantially extend life in a species in which longevity is under positive selective pressure (like, arguably, in mammals). Extensive data support this generalization. We have to grab for higher-hanging fruit.