Autophagy mediates pharmacological lifespan extension by spermidineand resveratrol

Thecomplex regulation of autophagy in response to stress

One of the key regulators ofautophagy in human and murine cells is the mammalian target of rapamycin (mTOR,whose yeast ortholog is TOR) kinase, which suppresses autophagy in conditionsof nutrient and growth factor repletion. Signal transducers including class Iphoshatidylinositol-3-kinases (PI3Ks) and Akt link receptor tyrosine kinases tomTOR activation, thereby repressing autophagy in response to insulin,insulin-like growth factor (IGF) and other growth signals [11]. Activationof the mTOR complex 1 (mTORC1) - and consequent repression of autophagy - canalso be mediated by mitogen-activated protein kinases (MAPKs) includingextracellular signal-regulated kinases (ERKs) [12], byRas-dependent activation of the p90 ribosomal S6 kinase [13], as well asby the Wnt signaling pathway [14]. Otherprominent regulators of autophagy include (but are not limited to):AMP-activated protein kinase (AMPK), which inhibits mTOR in response to reducedATP levels [15]; eukaryotictranslation initiation factor 2α (eIF2α), which responds to nutrientdeprivation as well as to double-stranded RNA (dsRNA) [16], ERN1 (whoseyeast ortholog is known as IRE1), an endoplasmic reticulum (ER)-associatedprotein possessing intrinsic kinase and endoribonuclease activities and playingan important role in the alteration of gene expression upon ER stress [10,17]; andc-Jun N-terminal kinase (JNK), which is involved in multiple signaling cascadesactivated by stressful conditions [18].

Our own work in this field hasadded to this list of autophagy regulators: members of the Bcl-2 protein familythat contain a single Bcl-2 homology (BH) domain, the so-called BH3-onlyproteins, which displace (and hence derepress) the essential autophagymodulator Beclin 1 from inhibitory complexes with Bcl-2 or Bcl-X_L[19,20]; Sirtuin1, which responds to high NAD⁺ levels, de facto acting as asensor of nutrient availability [21]; the oncosuppressorprotein p53, which inhibits autophagy when present in the cytoplasm [22]; theIκB kinase (IKK) complex, which is also essential for the activation ofNF-κB by stress [23,24]; as wellas the inositol 1,4,5-trisphosphate (IP₃) receptor (IP₃R)at the level of the ER [20,25].Finally, autophagy is positively regulated by the transcription factor activityof E2F1 [26], FoxO3a [27,28],NF-κB [29] and p53 [30,31], amongothers. The apical events of the phylogenetically ancient molecular pathway forautophagy involve ULK1 and ULK2 (the mammalian orthologs of Atg1) as well asBeclin 1 (the human ortholog of Atg6). Beclin 1 functions as an allostericactivator of the class IIIPI3K hVps34 (which promotes phagophore nucleation/elongationvia its product phosphatidylinositol-3-phosphate), and is part of a highlydynamic multiprotein complex that can incorporate various autophagicstimulators (e.g., UVRAG, Bif-1/endophilin B1, Ambra 1) and/orinhibitors (e.g., Bcl-2, RUBICON) [32-36].

In synthesis, autophagy is connected to multiple stress pathways. Insome cases, specific proteins and organelles are "tagged" for autophagicsequestration, implying that intrinsic features of the cargo determine itselimination by autophagy. This has been documented for proteotoxins [37-39], uncoupled mitochondria (aprocess that has been dubbed "mitophagy") [40-42], peroxisomes (for which the term"pexophagy" has been introduced) [43]; damaged ER (which is eliminatedby "reticulophagy") [44]; and invading pathogens (whichactivate "xenophagy") [45]. In many other instances,autophagy occurs in a rather unselective fashion and represents a generalresponse of the cell that transits from baseline to induced, yet limitedlevels. It is tempting to speculate that this change from basal to upregulatedautophagy may involve the activation of a general "switch" that can respond to multiple distinct stress-responsive and damage-sensing pathways. Complex molecular switches regulatethe clear-cut separation between discrete cellular states including thetransition from an undifferentiated to a more differentiated state, theadvancement of the cell cycle, or the "decision" to activate the apoptoticcascade [46,47]. Usually, such switchesintegrate diverse signals that are transmitted through negative feedback loops(which maintain homeostasis and keep cells in a defined state) and positivefeedback loops (which mark the rapid evolution between two states) [48]. There is abundant evidence thatthe induction of autophagy involves positive feedback loops. For example, wehave documented (i) that autophagy induced by rapamycin (which inhibits mTOR)is accompanied by the degradation of p53 and the activation of IKK; (ii) thatpharmacological inhibition of p53 with pifithrin-α leads to mTORinhibition and IKK activation; and that (iii) transgene-enforced activation ofIKK stimulates p53 degradation at the same time as it inhibits mTOR [22-24,49]. This implies that mTORinhibition, IKK activation and the degradation of cytoplasmic p53 arecross-linked through a network of self-amplifying feedforward loops (Figure 1),although it remains elusive how this occurs in molecular terms.

Figure 1.

Molecular composition of the hypothetical autophagic switch.

Irrespective ofthe primary stress signal, a homogeneous response would be obtained by theindependent activation of molecular complexes organized around the IκBkinase (IKK), cytoplasmic p53, mTOR and Beclin 1. Within these complexes(and perhaps others), proteins would undergo reversible post-translationalmodifications and/or shuttle from one complex to another, therebydetermining the function of the integrator/switch that activates autophagy.Please note that the autophagic switch is expected to contain severalpositive feedback loops that determine its (in)activation.

Autophagyas a cytoprotective and anti-aging mechanism

Cells that are stressed andon the verge of death frequently manifest the cytoplasmic accumulation ofautophagosomes and auto(phago)lysosomes, an observation that has been(mis)interpreted as if autophagy would contribute to the cellular suicide [50]. Thus,hundreds of papers have described "autophagic" (also dubbed "type 2") celldeath, a cell demise subroutine preceded by massive autophagic vacuolizationthat is morphologically distinct from apoptosis ("type 1") and necrosis ("type3") [51-54]. Although "autophagic cell death" undoubtedly exists as amorphological entity [53], this onlyexceptionally (at least in mammalian models) reflects the execution of cells byautophagy [50]. Rather,autophagy most frequently constitutes a (sometimes futile) mechanism ofcellular adaptation to a diverse range of adverse conditions includinghypoglycemia, hypoxia, lack of essential amino acids, absence of obligategrowth factors or sublethal damage to cytoplasmic organelles includingmitochondria and the ER [4,55,56].Accordingly, the genetic inhibition of autophagy by knockout or knockdown of ATGgenes often precipitates the apoptotic or necrotic death of cells thatotherwise would survive nutrient depletion, growth factor withdrawal, hypoxia,ionizing radiation or anticancer chemotherapy [11,50,57-60].Deficient autophagy is directly involved in a number of pathologies includingneurodegenerative diseases, heart failure, hereditary myopathies,steatosis/steatohepatitis and other chronic inflammatory states [6,61-64].Genetic and pharmacological manipulations designed to induce autophagy havebeen shown to protect cells against otherwise lethal damage in vitro[5,6]. Autophagyfavors the maintenance of high intracellular ATP levels [22,65],increases the capacity of cells to resist metabolic stress (hypoxia combinedwith nutrient deprivation) [22,66],prevents genomic instability [60,67] andlimits the accumulation of potentially toxic proteins including proteotoxinsthat are responsible for neurodegeneration [10,38]. From aphysiological point of view, aging can be viewed as a continuous decline incellular and organismal functions that (at least partially) reflects theaccumulation of misfolded proteins, oxidized lipids, as well as mutatedmitochondrial and nuclear DNA.

The sole regimen leading to lifespanextension in every organism tested to date is dietary restriction, a reductionof the organism's caloric intake not associated to malnutrition [68]. Dietaryrestriction is a potent inducer of autophagy in virtually all species includingmammals [69-71]. In thenematode Caenorhabditis elegans, autophagy is required for lifespanprolongation mediated by caloric restriction [72-74] or p53depletion [22,49,75-77].Thus, worms undergoing dietary restricttion do not live longer than controlanimals if concomitantly subjected to RNA interference (RNAi) against atggenes [72-74].

Rapamycin,which activates autophagy via inhibition of (m)TOR, has also been ascribed withprominent anti-aging properties, in various model organisms. However, rapamycincannot extend the chronological lifespan (i.e., the time post-mitoticcells survive during the stationary phase [78]) of yeastmutants that lack functional Atg1, Atg7 or Atg11 [79]. In C.elegans, the beneficial effects of rapamycin on longevity are lost when theessential autophagy modulator BEC-1 (the worm ortholog of mammalian Beclin 1and yeast Atg6) is knocked down [74]. Thus,autophagy is required for rapamycin-mediated lifespan extension and delay ofchronological aging in yeast and nematodes. Although it has not been formallydemonstrated that rapamycin prolongs the lifespan of mice by inducingautophagy, even the treatment of pre-aged, genetically heterogeneous (out-bred)mice has been shown to increase longevity [80]. In mice,rapamycin avoids the age-related decline in hematopoietic stem cells function [81], ananti-senescence effect that has also been described in vitro[82,83].

Altogether,these results suggest that whole-body induction of autophagy by pharmacologicalagents may prolong the healthy lifespan, at least in laboratory conditions,supporting the idea that autophagy does not only confer cytoprotection but thatit also has anti-aging effects at the organismal level.

Autophagymediates lifespan extension by resveratrol

Driven bythe aforementioned considerations, we launched the working hypothesis thatautophagy constitutes (one of) the major mechanism(s) through whichlongevity-extending drugs operate. We thus studied whether resveratrol, awell-studied anti-aging agent [84], wouldextend the lifespan of model organisms via the induction of autophagy. Althoughit also affects mitochondrial functions [85],resveratrol prominently acts as an allosteric activator of Sirtuin 1, aphylogenetically conserved deacetylase that senses the NAD⁺/NADHratio [84].Resveratrol increases the longevity of yeast, nematodes, and flies (Drosophilamelanogaster) and also exerts anti-aging effects on mice kept on a high-fatdiet [84,86]. Circumstantial evidence indicates that resveratrolcan induce autophagy in yeast (although this was attributed to the oxidation ofmitochondrial lipids [87]) and inhuman cancer cells (in which resveratrol-induced autophagy often precedes celldeath [88]). Sirtuin 1is the first protein that has been demonstrated to prolong lifespan in yeast(and then in animals including C. elegans and flies) [89], and hasalso been shown to trigger autophagy in human and murine cultured cells [90].

Weconfirmed that Sirtuin 1 overexpression increased the autophagic flux in humancancer cells in vitro, and that this effect was abolished by theaddition of EX527, a pharmacological inhibitor of its catalytic activity [91,92].Similarly, a transgene coding for SIR-2.1 (the C. elegans orthologof human Sirtuin 1) caused autophagy in nematodes, suggesting that the linkbetween Sirtuin 1 activation and autophagy is evolutionarily conserved [91,92].Importantly, Sirtuin 1 was required for the induction of autophagy by nutrientdeprivation (that was achieved by culturing cells in the absence of serum,amino acids and glucose) but not by other stimuli. Thus, in human cells, thedepletion (by RNAi) or inhibition (with EX527) of Sirtuin 1 fully prevented theproautophagic effects of nutrient starvation, yet failed to affect thestimulation of autophagy by mTOR inhibition (with rapamycin), p53 inhibition(with pifithrin-α) or ER stress (triggered by the addition oftunicamycin). Similarly, loss-of-functionmutations of sir-2.1 abolishedautophagy induced by caloric restriction but not that promoted by rapamycin ortunicamycin in C. elegans [91,92].Transgenic overexpression of sir-2.1 increased the median and maximumlifespan of nematodes as compared to non-transgenic control strains with thesame genetic background. This gain in longevity was lost when the essentialautophagic modulator BEC-1 was depleted by RNAi [91,92].RNAi-mediated knockdown of the C. elegans p53 ortholog CEP-1, amanipulation that extends longevity through the stimulation of autophagy [77], failedto further ameliorate the beneficial effects of sir-2.1 overexpressionon longevity [91,92]. Thisepistatic analysis suggests that SIR-2.1 accumulation and CEP-1 depletionextend lifespan through a common final pathway that relies on the induction ofautophagy.

Another genetic intervention designed toindirectly activate Sirtuin 1 (or its worm ortholog SIR-2.1) consists in thetransgenic overexpression of the gene coding for the pyrazinamidase/nicotinamidasePNC-1, which depletes nicotinamide, a negative regulator of Sirtuin 1/SIR-2.1.Transgenic overexpression of pnc-1 did indeed induce autophagy in worms,and this response was abolished by RNAi-mediated depletion of SIR-2.1. Accordingly,the longevity-extending effects of PNC-1 were lost upon the knockdown ofSIR-2.1, as well as upon that of either of the two essential autophagymodulators BEC-1 or ATG-5 [77]. Thus, boththe overexpression and the metabolic activation of Sirtuin 1/SIR-2.1 increaselifespan through the induction of autophagy.

Next, weinvestigated whether resveratrol would induce autophagy in C. elegansvia the activation of SIR-2.1. Addition of resveratrol to the worm culturemedium did indeed stimulate autophagy, and this effect was lost uponRNAi-mediated depletion of SIR-2.1. Similarly, resveratrol reduced theaging-associated mortality of C. elegans, unless the products of sir-2.1or bec-1 were knocked down [77]. Weconcluded from these experiments that resveratrol prolongs lifespan in humanand nematode cells by inducing autophagy, which results fromresveratrol-mediated activation of Sirtuin 1/SIR-2.1 (rather than from anoff-target effect).

Autophagymediates lifespan extension by spermidine

Driven bythe fact that the intracellular level of polyamines declines in (otherwisehealthy) aging humans [93], weinvestigated whether the polyamine spermidine display anti-aging properties. Toaddress this question, we first took advantage of a yeast strain that isdeficient in the ornithine decarboxylase SPE1, which catalyzes the first stepof polyamine biosynthesis. In chronological aging experiments, Δspe1 yeastcells exhibited an increased mortality (and hence a shortened lifespan), whichcould be restored to normal levels by supplementation with low doses (0.1 mM)of spermidine or its precursor putrescine [94]. Surprisingly,we found that higher concentrations of spermidine were able to increase thelifespan of wild type yeast cells with different genetic backgrounds. Thus,both chronological aging (which constitutes a model of post-mitotic aging) andreplicative aging (which constitutes a model of stem cell aging) of yeast cellswere significantly inhibited by spermidine supplementation. Lifespanprolongation in spermidine-treated yeast cells could be correlated with thereduced acetylation of several lysine residues located at the N-terminal tailof histone H3 (i.e., Lys9, Lys14 and Lys18) [94]. Deletion ofsir2 (the yeast ortholog of Sirtuin 1) or any other sirtuin didnot affect the ability of spermidine to extend chronological lifespan. Instead,epistatic analyses revealed that the anti-aging effect of spermidine wasphenocopied by the knockout of histone acetylases, which hence were shown toregulate the same longevity-increasing pathway than spermidine does [94]. Moreover,spermidine efficiently inhibited general histone acetylase activity in extractsfrom purified yeast and mammalian nuclei in an in vitro assay [94]. Theseresults suggest that spermidine acts differently from resveratrol. Thus, whilethe former inhibits histone acetylase(s), the latter stimulates the deacetylaseactivity of Sirtuin 1. However, formal evidence that the (de)acetylation ofhistones rather than that of other proteins (either in the nucleus or in thecytoplasm) account for the anti-aging properties of spermidine is stillmissing.

Microarrayprofiling of spermidine-treated yeast cells revealed the transcriptionalactivation of several autophagy genes including atg7, atg11 and atg15,and we indeed found that spermidine induces autophagy in yeast cells.Similarly, spermidine was highly efficient in upregulating the autophagicpathways when it was added to the culture medium or solid food of C. elegansor D. melanogaster, respectively. The same concentrations ofspermidine that exerted pro-autophagic effects also had a markedlifespan-extending effect on yeast, nematodes and flies. The genetic inhibitionof essential ATG genes (i.e., knockout of atg7 in yeastand flies, RNAi-mediated silencing of bec-1 in nematodes) abrogatedlongevity extension induced by spermidine, indicating this polyamine canprolong lifespan by the induction of autophagy [94].

Openquestions

Theaforementioned results indicate that resveratrol and spermidine can prolong thelifespan of model organisms through the induction of autophagy (Figure 2). Inaddition, our work raises at least three issues that must be addressed byfuture investigation.

First, do resveratrol and spermidineextend longevity by acting on the same molecular pathway? While resveratrol canprolong lifespan through the activation of the deacetylase activity of Sirtuin1 (or its non-mammalian equivalents SIR2 in yeast and SIR-2.1 in C. elegans),spermidine inhibits the general histone acetylase activity of yeast and mouseliver extracts. Clearly, histone (de)acetylation has been recognized as animportant epigenetic regulator of longevity [95,96].However, a fraction of Sirtuin 1 is present in the cytoplasm, from where it candirectly deacetylate essential autophagic proteins (including ATG5, ATG7 andATG8/LC3) [90], suggestingthat (at least part of) the pro-autophagic effects of resveratrol derive fromextranuclear, transcription-independent events. It will be important to knowwhether polyamines (like spermidine) and Sirtuin 1 activators (includingresveratrol) can exert additive or synergistic effects on autophagy andlongevity or whether these agents exactly activate the same molecular pathway.Moreover, the precise mechanisms by which spermidine and resveratrol controlthe autophagic switch awaits further exploration.Careful mechanistic and epistatic analyses are required to address thisproblem.

Figure 2.

Hypothetical mode of action of resveratrol and spermidine as autophagy inducers.

Whileresveratrol functions as an activator of the deacetylase Sirtuin 1,spermidine inhibits one or several histone acetylases. Therefore, bothresveratrol and spermidine are expected to favor protein hypoacetylation.However, the autophagy-relevant substrates whose deacetylation is inducedby resveratrol and spermidine are not fully characterized and it is evennot known if they are completely distinct, partially overlapping oridentical. For further details, please consult the main text.

Second, doall longevity-prolonging manipulations induce autophagy? And is autophagyrequired for all such intervention to extend lifespan? Current results clearlyindicate that autophagy is indispensable for the anti-aging action ofrapamycin, resveratrol and spermidine. Moreover, it has been suggested thatautophagy is required for longevity extension by dietary restriction in C.elegans, although this has not been tested for all caloric restrictionprotocols [73]. It remainsan ongoing conundrum whether an increased level of autophagy is required in C.elegans for longevity extension conferred by the deficiency of GTPaseRHEB-1 [97], thetranscription faction hypoxia-inducible factor 1 (HIF-1) [98] and itsnegative regulator VHL-1 [99], theubiquitin ligase WWP-1 [100], as wellas the chaperones CCT4 and CCT6 [101]. Apositive response to this question could establish a new paradigm in longevityresearch.

Third, andmost important, can the data that we discuss here, which have mostly beenobtained in simple model organisms and in laboratory conditions (where, forinstance, the immunosuppressive side effects of resveratrol are certainly lessincisive), be extrapolated to humans and to real life? Although rapamycin andpolyamines can increase lifespan in mice [80,102],resveratrol only extends the longevity of mice that are kept on a high-caloricdiet [86]. Clearly, rapamycin and resveratrol can induce autophagy in vivo,in mice [23,24,103].However, it is thus far unknown whether there is indeed a cause-effect relationshipbetween increased autophagy and healthy aging in mammals and in particular inhumans. Such a causal relationship would revolutionize the entire field ofaging research.