Increasing longevity through caloric restriction or rapamycin feeding in mammals: common mechanisms for common outcomes?

Rapamycin, a macrolide antibiotic, is known to act by non-competitive inhibition of the TORC1 complex, a crucial serine-threonine kinase-containing complex which integrates cellular signals from the insulin/IGF1 axis, growth factors, and nutrient sensors and promotes increased protein synthesis via phosphorylation of ribosomal S6 kinase and 4E-BP1, the binding protein of translation initiation factor eIF4E (reviewed by Wullschleger et al. 2006 ; Stanfel et al. 2009 ). But is rapamycin actually extending lifespan or merely preventing death? For instance, the closely related antibiotic rifamycin, increases lifespan in certain human subjects by treating otherwise fatal tuberculosis, but this is not an ageing-related intervention, rather prevention of premature mortality. Mice in the ITP study ( Harrison et al. 2009 ) were maintained in a specific pathogen free environment so not subject to potential life-shortening infections, thus the rapamycin effect is probably acting via age-related mortality. The fact that rapamycin acts on the TOR pathway, and the TOR pathway is intimately linked to lifespan (reviewed by Stanfel et al. 2009 ), further suggests that its effect is direct rather than indirect. For instance, inhibition of insulin signalling by genetic mutation of pathway components (eg Daf-2 in C. elegans and chico in Drosophila ), of the TOR complex itself by dominant negative mutation, rapamycin treatment or overexpression of cellular inhibitor TSC1, or of downstream effectors such as S6K1 by mutation, all result in increased lifespan in experimental systems from yeast ( Kaeberlein et al. 2005 ; Powers et al. 2006 ) through worms ( Kenyon et al. 1993 ; Lithgow et al. 1994 ; Vellai et al. 2003 ) to flies ( Clancy et al. 2001 ; Kapahi et al. 2004 ). Harrison et al. (2009) have now extended this to mice, an important model mammalian species, suggesting conservation of the role of the TOR pathway in regulating lifespan. What has not been addressed, and will have obvious implications in human studies, is the basis of the sex differences in response to rapamycin, where female lifespan extension was greatly elevated compared with males. Female humans generally live longer than males, and while this is not observed in inbred laboratory mouse populations, a gender difference has been reported in several genetic models of lifespan extension ( Bartke & Brown-Borg 2004 ). The basis for this gender effect remains under investigation.

Lifespan extension without either a delay or compression of morbidity may not be an appropriate population intervention in ageing humans, as age-related morbidity carries huge financial and personal costs. Do the current studies suggest that compression of morbidity can be achieved? Although rapamycin treatment resulted in a greater age at survival into the 90 th percentile (ie when 90% of animals had died), indicative of an increase in maximal lifespan, the mice eventually succumbed to many of the same types of diseases as in the control groups, particularly cancer and cardiovascular disease ( Harrison et al. 2009 ). Although not a significant difference, that the rapamycin mice actually had a slightly higher incidence of these two diseases suggests that it will be necessary to explore the pathology in much greater detail before a health benefit can be concluded. Whether the rapamycin-treated mice acquired age-related disease at the same time as control mice but the disease progressed more slowly, or whether onset of disease was delayed with progression at the same rate as in controls, is not distinguishable at this stage. Additionally, it will be necessary to assess function at the cellular, organ, physiological, and behavioural level as mice progress to old age. Such issues are likely to become important if and when pharmacological intervention in human ageing becomes a reality: the jury is still out on whether rapamycin or similar interventions can deliver compression of morbidity. By contrast, the monkeys in the Wisconsin study ( Colman et al. 2009 ) show evidence of lower morbidity as well as increased longevity. Age-related sarcopenia was decreased (as measured by X-ray absorptiometry ( Colman et al. 2008 ), cancer incidence was halved (though numbers are too small to be statistically certain that this difference is significant), and progression to pre-diabetic or diabetic states was completely abolished in the CR monkeys ( Colman et al. 2009 ). It is likely that such pathologies will also be decreased in humans undergoing CR. For example, significant decreases in inflammatory markers, LDL cholesterol and systolic and diastolic blood pressure, together with marked drop in BMI to ~19 is observed after as little as a year of CR in volunteer human studies (e.g. Holloszy & Fontana 2007 , reviewed by Fontana & Klein 2007 ; Fontana 2009 ): however, adverse effects such as decreased physical work capacity ( Weiss et al. 2007 ) and loss of bone mineral density at key sites such as head of the femur ( Villareal et al. 2006 ) are also reported. These studies examined humans with a starting BMI of ~25 and an equivalent control group; for those with a lower starting BMI, CR may be immediately detrimental to health rather than a positive benefit. Perhaps the decrease in caloric intake, currently to 70–80 % of the ad libitum start point, may need to be tailored to initial body mass and BMI, rather than ad lib appetite, if it is to be usefully translated to humans. In addition to the benefits reported by Colman et al. (2009) , in the cohort of rhesus monkeys maintained by the NIA, CR monkeys showed marked improvement in immune function. Messaoudi et al. (2006) showed that CR initiated during adulthood can delay T-cell aging and preserve naïve CD8 and CD4 T cells into advanced age. However, when CR was started very early in life, prior to puberty, or late in life, the beneficial effects were lost ( Messaoudi et al. 2008 ). Thus it may be that there is a key window of opportunity for initiating CR to obtain optimal immune function benefits. This contrasts with rapamycin treatment in mice where both mid-life (from 270 days) and late-life (from 600 days) intervention showed significant benefits in terms of lifespan extension ( Harrison et al. 2009 ). Inflammation is also reduced in CR monkeys, as detected by reduced inflammatory cytokine production ( Messaoudi et al. 2006 ). In experimental induction of oral inflammation via ligature-induced periodontitis, CR attenuated the inflammatory response and slowed the progression of periodontal destruction ( Branch-Mays et al. 2008 ). Although these findings are not directly linked to a lifespan extension, they are indicative of an improvement in the aging phenotype. “Inflammageing” ie the shift to secretion of high levels of circulating pro-inflammatory cytokines such as IL-6 and TNF-α has been suggested as a major factor in ageing ( Franceschi et al. 2000 ), so its attenuation by CR may actually be one mechanism by which dietary restriction improves longevity. Since white adipose contributes to a pro-inflammatory environment (e.g. see Yudkin et al. 2000 ; Cao et al. 2008 ) it will be very interesting to study the contribution of visceral fat deposits to the inflammatory and pro-ageing phenotype ( Gustafson et al. 2009 ; Starr et al. 2009 ). CR monkeys and people have much reduced abdominal fat deposition, while patients with the progeroid Werner’s syndrome, though with greatly reduced overall BMIs, show lipodystrophy with large visceral fat deposits and high levels of IL-6 and TNF-α ( Honjo et al. 2008 , reviewed in Cox & Faragher, 2007 ). This finding supports the suggestion that increases in abdominal fat correlate with and contribute to increased inflammation and accelerated ageing. Any treatment aimed at the overweight and obese (an increasingly important target group as human lifespan in western populations may be about to plummet due to the obesity epidemic) may have marked benefits in terms of improved lifespan and health span. It will be interesting to determine the cytokine profiles of rapamycin treated mice to test whether the drug has a beneficial effect in decreasing inflammation. Anecdotally, 6/31 control ITP mice died of inflammatory-type conditions while only 1/40 of the rapamycin mice assessed by necropsy did, though numbers are far too small to derive robust conclusions.

Lifespan extension in important mammalian species is of great interest to the ageing research community. The two new papers showing significant lifespan extension in mice on rapamycin feeding ( Harrison et al. 2009 ) and in rhesus monkeys on CR ( Colman et al. 2009 ) merit serious consideration in terms of what they add to understanding of ageing processes and molecular mechanisms of ageing, together with enhancing prospects for modification of human ageing. The major findings of the new study by Harrison et al. are that (i) a lifespan extending drug, rapamycin, can be fed in the diet, and (ii) late life intervention can have a marked impact on increasing lifespan in mice. In addition to these key important results, rapamycin is a valuable start point for further studies as its cellular target and downstream effects have been intensively studied and newer rapamycin analogs (rapalogs) are in development (reviewed by Easton & Houghton 2006 ; Sabatini 2006 ; Abraham & Gibbons 2007 ; Stanfel et al. 2009 ). By contrast, Colman et al. (2009) show evidence that (i) adult onset CR leads to increased longevity in rhesus monkeys, and (ii) that this CR intervention also markedly reduces the incidence of age-associated pathologies such as cancer and glucoregulatory defects. Excitingly, they also show a decrease in brain grey matter atrophy in the aged CR monkeys. Apart from the obvious similarity of increasing lifespan in mammals, do these two very different studies have any points of commonality, and can they be assessed together as enhancing our understanding of ageing processes?

Some critics have argued that publication of the Wisconsin study ( Colman et al. 2009 ) is premature; it is still unknown if these findings will translate to a longer maximal lifespan, since around half of the monkeys are still alive 20 years into the project (compared with ~2% of the mice in the rapamycin study ( Harrison et al. 2009 )). Although reduction in age-related mortality is statistically significant in the CR group of monkeys, there is as yet no significant difference in mortality from all causes, though the data suggest a trend in favor of CR prolonging life. We believe the significance of lifespan extension on CR should not be over-shadowed by the distinction between age-related and overall mortality. Such concerns are less overt for the mouse study, which was conducted as part of the NIA’s Interventions Testing Program across three sites in the USA, and used 1960 mice in total. The large sample size for controls and intervention cohorts allowed rigorous statistical analysis of any observed effects (see Ladiges et al. (2009) for the basis of power calculations), and the split site nature of the project ensured that sufficient experimental “replicates” were performed to ensure robustness and reproducibility of data. The observation of 9% (male) and 14% (female) increases in lifespan at the 90 th percentile is thus highly significant, despite losses from the population through non-age related causes and accidents. Even potentially confounding variables, such as different diets in the rapamycin cohort prior to drug feeding compared with controls, could be excluded because at one site (TJL) diet was correctly controlled throughout life, and numbers were still large enough to derive statistical validity. (As an aside, it will be interesting to determine what aspects of the Teklad 7912 or Purina 5008 chow promoted increases in male mouse longevity - slight differences in palatability may lead to unintended CR, and thus such studies in future should include measurements of food intake in addition to animal body mass). By contrast, monkey studies generally do not have the advantage of overly large groups and thus even single deaths for reasons not germane to the study could potentially mask the effect of the treatment. This strategy can be likened to a human clinical trial in which an early death caused by an accident or unrelated injury would not be attributed to the intervention. Because lifespan studies in long-lived primates are rare, there is not a widely accepted standard practice in attributing causes of death. Thus, the Wisconsin investigators led by Rick Weindruch have collaborated with those from the related NIA primate longevity study ( Ingram et al. 1990 ; Mattison et al. 2003 ) to establish criteria to categorize causes of death. Both studies perform a preliminary survival analysis which includes all deaths followed by a secondary analysis of age-related deaths which may better represent the effect of the intervention. Some of the deaths reported by Colman et al. (2009) might be categorized as non-age related and therefore less relevant to the scientific question, but are they truly unrelated to age? In the mouse study, causes not associated with ageing were obvious: fighting and cage flooding, though regrettable, are not generally considered to be age-related. Deaths from other causes including cancer were included in the Kaplan-Meier curves for mouse mortality. In the monkey study, some deaths were due to a poorly understood condition, gastric dilatation-volvulus (GDV), or gastric bloat with torsion ( Van Kruiningen 1975 ), which also occurs in dogs, horses, pigs, rabbits, and rodents. Although the precise cause is uncertain, several factors have been associated with its occurrence, including: gastric motility dysfunction, aerophagia, gastric content fermentation, and gas diffusion and seasonal barometric changes ( Levine & Moore 2009 ), none of which are age-associated. A few deaths in the monkey study were due to endometriosis, which is considered non-age related because it occurs in young adults through to middle-aged animals, but then the incidence is naturally halted at the onset of menopause. It is a multifaceted condition with many contributing factors ( Mattison et al. 2007 ). Despite careful monitoring, monkeys often mask symptoms until endometriosis is more advanced. Thus in human medicine, endometriosis is not likely to lead to a moribund condition whereas the complications that occur at advanced stages in monkeys often necessitate euthanasia. It is possible that CR might influence the incidence and/or the outcome of endometriosis cases; this information will not be lost in these initial mortality analyses. Finally, it is necessary to anesthetize the monkeys for routine health screening or for clinical or experimental interventions, and unfortunately, but rarely, this resulted in death. In such cases, at both Wisconsin and the NIA primate facility, a thorough necropsy is performed and tissues are examined histologically. For any documented accidental death, other contributing factors are ruled out before the cause of death is categorized. Thus we believe it is valid to assess longevity in terms of age-related deaths rather than overall mortality.

The question of quality of life (QOL) is often posed by critics of the CR field: is it the case that although you might not live to be 120, the loss of enjoyment involved will make it feel that way? In both the Wisconsin and NIA nonhuman primate studies, a battery of behavioral analyses to evaluate indices that may indicate improved quality of life have been performed. Behaviorally, CR monkeys are no meaner, more aggressive, or more lethargic than their control counterparts. In addition to decreased morbidity which is likely to contribute to improved quality of life, perhaps the most exciting finding from the study by Coleman et al . (2009) is that CR results in either a delay or decrease in atrophy of the grey matter of the brain. One of the biggest concerns for most people approaching older age is loss of cognitive ability, most seriously manifested in senile dementia, which affects 13.9% of those over 71 years in the human population ( Plassman et al. , 2007 ). As Weindruch and colleagues point out, modelling age-related decline in higher cognitive centres is far from satisfactory in rodents, so their study in monkeys, which show very similar age-related brain changes to humans, is especially interesting. A short-term 6-month CR regimen also showed beneficial brain effects in adult male rhesus monkeys injected with the neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), which induces a Parkinson’s-like syndrome by selective degeneration of dopamine neurons ( Maswood et al. 2004 ). Less severe effects of MPTP treatment were observed in the CR monkeys, who were more active post-injection, with higher levels of dopamine and dopamine metabolites in the affected brain regions compared to the controls. Interestingly, rapamycin inhibits Parkinson’s disease (PD) dyskinesia in mice ( Santini et al. 2009 ), suggesting that both CR and inhibition of the TOR pathway can be beneficial in PD. It is uncertain whether attenuated brain atrophy contributes to improved function in the monkeys, though it will be interesting to revisit the CR human cohorts in later years to determine whether brain function, and particularly ability to conduct complex cognitive tasks, is maintained. Current analyses are suggestive of protective effects of CR even against Alzheimer’s disease, but randomized clinical studies do not show significance ( Gillette-Guyonnet & Vellas 2008 ). Retrospective cohort studies (e.g. the Lothian birth cohort study, Deary et al. 2007 ) may contribute to this question. Interestingly, factors in the TOR pathway such as FOXO transcription factor, together with lipid modulators such as PPAR and statins, do appear to have mild effects in improving cognition on ageing ( Starr et al. 2004 , reviewed by Contestabile 2009 ). It is possible that rapamycin, by acting on TOR and possibly agonistically with FOXO, may also have a neuroprotective effect, but indirectly through modulating circulating lipids and decreasing the risk of cerebrovascular accidents. Unfortunately, the side-effects of rapamycin when used as an immunosuppressant are severe, so the drug is not suitable for improving quality of older life in otherwise healthy humans. However, it is unknown if the dose used by Harrison et al. (2009) affected immune function in these mice. Rapalogs or agents that can spare immune function yet target ageing-critical organs may thus provide better prospects of QOL enhancement in humans.

What are the prospects for human intervention based on these recent rapamycin and CR studies? CR is not without risk in humans (reviewed by Fontana & Klein 2007 ), and rapamycin cannot be used to extend human longevity because of its serious immunosuppressive and other side effects, including increased risk of some neoplasias. However, rapalogs (eg. CCI-779/temsirolimus (Wyeth), RAD-001/everolimus (Novartis) and AP23573 (Ariad)), or development of competitive TOR pathway inhibitors (Akt or PI3K inhibitors, for instance) may be a potential route to developing agents with fewer side effects. Drug dose is also an issue, as the doses used by Harrison et al. (2009) were extremely high compared with clinical use in human transplant patients. However, as a highly lipophilic drug, it is likely that rapamycin will accumulate in body fat. Since inhibition of TOR signalling in the metabolically active fat body of flies is sufficient to increase lifespan ( Kapahi et al. 2004 ), it is attractive to imagine that fat accumulation could lower the dose of rapamycin required to initiate a longevity effect in humans while minimizing any immunosuppressive risk. Further studies in Ob mice may be informative. Design of CR mimetics ( Chen & Guarente 2007 ) to provide the compression of morbidity and lifespan extension benefits observed by Colman et al. (2009) may be complex, given both an absence of a defined single pathway by which CR mediates its longevity effects, and in the light of recent findings that CR has different effects on different tissues. For instance, Sirt1 activity is actually decreased in liver of CR mice, while a high calorie diet increases its expression in the liver ( Chen et al. 2008 ). What will be very interesting to test in an ITP study is whether resveratrol (or other Sirt-activators) acts epistatically, additively or synergistically with rapamycin in mice. Overall, we do not advocate novel human interventions based solely on the findings of these two new papers ( Colman et al. 2009 ; Harrison et al. 2009 ). They do however, reinforce other studies on model eukaryotes and vindicate the use of inveterrbrates in studying processes relevant to mammalian, and by extrapolation, human ageing. Current advice on sensible balanced diet and regular exercise, combined with high quality health care, is still likely to provide the best current route to maintaining quality of life throughout the human lifespan and of decreasing risks of morbidity and early mortality. Further understanding of the pathways involved in CR and rapamycin inhibition, and the mode of action of any rapalogs or CR-mimetics, will be necessary to exclude unforeseen consequences before it is possible to contemplate pharmacological intervention in ageing at a population level in humans.

There are likely to be many pathways at work keeping CR monkeys (and yeast, worms, flies, fish, and rodents) healthy and extending lifespan and the molecular mechanisms remain under investigation (reviewed by Bishop & Guarente 2007 ). Since mid-life or late-life rapamycin treatment of rodents similarly increases lifespan, are CR and rapamycin effects related mechanistically? Rapamycin binds to FRAP/FKBP512 and forms a ternary inhibitory complex with TOR kinase (mTOR in mammals) in TORC1 and thus prevents downstream phosphorylation events that would normally promote protein synthesis and cell growth (reviewed by Wullschleger et al. 2006 ; Stanfel et al. 2009 ). While the immediate mode of action of rapamycin is therefore clear, the non-competitive nature of the inhibition leads to other indirect effects such as enhanced insulin signalling through abrogation of IRS down-regulation normally mediated through S6K1, and diminution of TORC2 signalling through sequestration of TOR protein (reviewed by Abraham & Gibbons 2007 ). Nutrient sensing through TORC2 is thus impaired on rapamycin treatment, though TORC2 is not itself directly inhibited by rapamycin: could this relate to CR? Moreover, a decrease in dietary intake on CR leads to lower insulin levels and thus a decrease in the insulin/IGF1 signalling (IIS) pathway, which feeds into TORC1. Furthermore, TOR pathway inhibition leads to increased density of oxidative phosphorylation components within the mitochondria ( Pan & Shadel 2009 ), suggesting a metabolic shift, at least in yeast, from anaerobic to aerobic respiration, though membrane potential uncoupling may also be occurring ( Pan & Shadel 2009 ). Whether this aspect is specific to yeast, or applicable to obligate aerobic organisms, has yet to be determined, though membrane potential uncoupling has also been associated with longevity in worms ( Lemire et al. 2009 ). There is a well-established relationship between CR and the TOR pathway in lower organisms such as worms ( Hansen et al. 2008 ), yeast ( Kaeberlein et al. 2005 ), and flies ( Kapahi et al. 2004 ). Possible mechanisms for TOR’s effect are reviewed in Stanfel et al. (2009) and gene interaction studies in these species leave little doubt that reduced TOR signalling mediates many effects of CR. Thus the impact of rapamycin is more varied than would at first sight have been anticipated, and overlaps with some CR pathways. But as evidenced in ( Harrison et al ., 2009 ) a body weight effect may not be essential to the lifespan extension that does occur with CR (e.g. Roth et al. , 2004 ). Perhaps site of action as well as molecular target is important for rapamycin’s effects on TOR signalling. For instance, inhibition of TOR signalling in the highly metabolically active fat body is sufficient to promote longevity in flies ( Kapahi et al. 2004 ), and Harrison et al. (2009) found significantly decreased S6 kinase phosphorylation in visceral fat pads of mice treated with rapamycin from 270 days. However, the marked decrease in age-related morbidity on CR is not mirrored on rapamycin treatment in mice, so perhaps the pathways of action are parallel rather than intertwined. Epistasis analysis of other longevity factors in lower organisms, such as the sirtuins (activated by resveratrol, Howitz et al. 2003 ) suggest that they may strongly impact the CR pathway (e.g. Rogina & Helfand 2004 ; Wood et al. 2004 ). The Sirt1 pathway, activated with resveratrol feeding ( Baur et al. 2006 ; Pearson et al. 2008 ) or BAC-based transgene expression ( Pfluger et al. 2008) show changes consistent with improved healthspan of mice fed a standard diet but impacted longevity of the mice only when placed on a high fat diet. As ageing is a complex, polygenic process, it is likely to involve multiple molecular mechanisms and pathways, modification of one of which may delay death without affecting morbidity, while the more pleiotropic CR would be expected to impact on a greater range of age -associated phenotypes, as is observed in the low morbidity in aged CR monkeys.