The increase in life expectancy has boosted the incidence of age-related pathologies beyond social and economic sustainability. Consequently, there is an urgent need for interventions that revert or at least prevent the pathogenic age-associated deterioration. The permanent or periodic reduction of calorie intake without malnutrition (caloric restriction and fasting) is the only strategy that reliably extends healthspan in mammals including non-human primates. However, the strict and life-long compliance with these regimens is difficult, which has promoted the emergence of caloric restriction mimetics (CRMs). We define CRMs as compounds that ignite the protective pathways of caloric restriction by promoting autophagy, a cytoplasmic recycling mechanism, via a reduction in protein acetylation. Here, we describe the current knowledge on molecular, cellular, and organismal effects of known and putative CRMs in mice and humans. We anticipate that CRMs will become part of the pharmacological armamentarium against aging and age-related cardiovascular, neurodegenerative, and malignant diseases.

 As a catabolic process, autophagy is induced upon nutrient deprivation and plays an important role in the beneficial effects exerted by CR and fasting regimens. CR modulates several molecular key players involved in the regulation and execution of autophagy, nutrient signaling, and energy metabolism (Figure 1). For instance, CR activates AMP-activated protein kinase (AMPK) (Canto´ and Auwerx, 2011). AMPK is an energy sensor that inhibits the kinase activity of mechanistic target of rapamycin (mTOR), an autophagy repressor, under CR. Furthermore, CR directly and indirectly activates sirtuins (SIRTs), which are nicotine adenine dinucleotide (NAD+ )-dependent lysine deacetylases (KDACs) and play central roles during aging and autophagy (Guarente, 2007). SIRT1 and AMPK may engage in a positive feedforward loop to amplify the response to CR. Protein acetylation is a major regulator of autophagy. The N 3-acetylation of lysines is a phylogenetically conserved, posttranslational protein modification that is catalyzed by lysine acetyltransferases (KATs) and reversed by KDACs. N 3-acetylation regulates multiple metabolic enzymes, facilitating the adaptation to nutrient availability. Of note, N 3-acetylation may occur in a non-enzymatic fashion in the presence of AcCoA, especially at an acidic pH (James et al., 2017). There are four ways to diminish N 3-acetylation of proteins: (1) by reducing the concentration of cytosolic AcCoA, the sole donor of acetyl groups used by KATs, e.g., via inhibition of its synthesis from glycolysis, b-oxidation of fatty acids, or the catabolism of branched amino acids, or via increase of its consumption, for instance by carnitine acetyltransferases that transfer AcCoA acetyl groups on carnitine;

 (2) by degrading S-acetyl glutathione by mitochondrial thioesterase glyoxalase 2, GLO2, or cytosolic GLO1, thus reducing intermediates for non-enzymatic N 3-acetylation; (3) by activating specific KDACs, mostly SIRTs; and (4) by inhibiting KATs such as E1Abinding protein p300 (EP300). Notably, SIRT1 activity is low in aged and obese mice. This correlates with the inhibitory hyperacetylation of SIRT3, and transgenic activation of SIRT3 may improve the hepatic consequences of obesity including glucose intolerance (Kwon et al., 2017). Moreover, in mice, transgeneenforced overexpression of SIRT6 (Kanfi et al., 2012) or brainspecific expression of SIRT1 (Satoh et al., 2013) is sufficient to extend lifespan. In an earlier study, however, whole-body overexpression of SIRT1 did not extend lifespan (Herranz et al., 2010). Similarly, another report observed no lifespan extension upon overexpression of the SIRTs sir-2.1 and dSir2 in the nematode Caenorhabditis elegans and the fruit fly Drosophila melanogaster, respectively (Burnett et al., 2011), thus contradicting previous results (Bauer et al., 2009; Rogina et al., 2002; Tissenbaum and Guarente, 2001). While it seems clear that SIRTs exert important functions related to healthy aging, their specific role in promoting longevity remains to be clarified (Dang, 2014). Interestingly, autophagy and protein acetylation are subjected to circadian fluctuations (Sato et al., 2017). This oscillation is lost with aging and has been proposed as a modulatory target of CR (Sato et al., 2017). 

The maintenance of rhythmic (de)acetylation by CR is hypothetically linked to increased NAD+ levels, coupled to SIRT1 activation and rhythmic changes in the inhibitory acetylation of acetyl-CoA-generating acyl-CoA synthase short-chain family member 1 (ACSS1) (Sato et al., 2017). In aged flies, protein acetylation is increased, a phenomenon that can be attenuated by reducing the AcCoA-generating enzyme ATP citrate lyase (ACLY) or by mutating the KAT Chameau, resulting in an extended lifespan (Peleg et al., 2016). Similarly, the inhibitory hyperacetylation of the pro-autophagic transcriptional factor Foxo1 has been observed in aged mouse hearts (Ren et al., 2017). Moreover, CR deacetylates histones H3 and H4 in mouse fat pads (Xu et al., 2015) and reduces the levels of biotin, which acts as an endogenous inhibitor of SIRT1 (Xu et al., 2015). Both histone deacetylation and deacetylation of cytosolic proteins may affect the expression and activity, respectively, of autophagy-relevant proteins (Eisenberg et al., 2009; Marin˜ o et al., 2014). Both in mice and in humans, acute starvation causes a reduction of the acetylation of cytoplasmic proteins in peripheral blood mononuclear cells (Pietrocola et al., 2017). However, every-other-day fasting increases histone acetylation in the mouse retina (Guo et al., 2016), and acetylation is reduced in aged mouse livers, a phenomenon that is reversed by CR, which causes hepatic protein hyperacetylation (Sato et al., 2017). This is at odds with chronic alcohol abuse, which leads to NAD+ depletion and SIRT inhibition, resulting in hyperacetylation of multiple proteins in the liver (such as AMPK, b-catenin, histone H3, and the transcription factors SREBB2, PPARa, FOX01, NFkB, and NFAT) (French, 2016). 

 We previously defined CRMs as compounds that activate autophagy by promoting the deacetylation of cellular proteins (Madeo et al., 2014), by (1) depleting AcCoA, (2) inhibiting acetyltransferases, and/or (3) stimulating deacetylases (Figure 2). This definition reflects the fact that protein acetylation usually inhibits autophagy, while protein deacetylation favors autophagy. For instance, starvation is coupled to the inhibition of the acetyltransferase EP300 (due to the depletion of AcCoA), as well as to the activation of the deacetylase SIRT1 (due to the increase of the NAD+ /NADH ratio and the activation of AMPK). This results in the deacetylation of hundreds of cellular proteins (Morselli et al., 2011), reflecting multipronged regulatory effects on cell metabolism and the autophagic cascade. A systematic screen for KATs, the inhibition of which would induce autophagy, led to the identification of EP300 as a major negative regulator of autophagy that acts epistatic to starvation (Marin˜ o et al., 2014). Interestingly, EP300 is subjected to activating phosphorylation by mTORC1 (Wan et al., 2017), while conversely, inhibition of EP300 generally results in mTORC1 inhibition (Pietrocola et al., 2015), suggesting that both regulatory systems are intertwined. Similarly, protein deacetylation may be connected to the activation of AMPK, a potent autophagy inducer. Thus, deacetylation of liver kinase B1 (LKB1), for instance by SIRT2, favors the LKB1- mediated activation of AMPK (Tang et al., 2017). Likewise, EP300 inhibition results in AMPK activation (Pietrocola et al., 2015). 

These examples illustrate how protein deacetylation may initiate autophagy, correlating with mTORC1 inhibition and AMPK activation. However, EP300 inhibition results in the induction of autophagy even in conditions in which AMPK is deleted, mTORC1 is artificially activated, or ULK1 is inhibited (Pietrocola et al., 2018; Su et al., 2017). This suggests that protein deacetylation can set off the autophagic cascade in a dominant fashion that is largely independent of other regulatory systems. In accord with this interpretation, EP300 inhibition or SIRT activation may favor autophagy through deacetylation reactions that affect multiple autophagy-executory proteins (Pietrocola et al., 2015). For instance, EP300 inhibition results in the deacetylation of phosphatidylinositol 3-kinase catalytic subunit type 3 (PI3K3) at K29 and K771, favoring its interaction with allosteric activators contained in the pro-autophagic Beclin 1 (BECN1) complex and its substrate phosphatidylinositol, respectively (Su et al., 2017). BECN1 itself is also a substrate of EP300 and SIRT1 (at K430 and K437), and deacetylation of BECN1 favors the dissociation of its inhibitory interactor Rubicon (Sun et al., 2015). Of note, pro-autophagic derepression of BECN1 has been recently shown to promote longevity in mice (Marin˜ o et al., 2014). Furthermore, SIRT1 deacetylates nuclear microtubule-associated proteins 1A/1B light chain 3B (hereafter referred to as LC3) (at K49 and K51), stimulating its interaction with the nuclear protein DOR and its export to the cytoplasm, where it acts as a key initiator of autophagy (Huang et al., 2015). 

EP300 can also acetylate ATG5 and ATG7, both of which are involved in a conjugation system that promotes LC3 lipidation, which is required for autophagy induction. Of note, ATG5 is also deacetylated by SIRT2, supporting the notion that many autophagy regulators are substrates of both EP300 and SIRTs (Liu et al., 2017a). While the link between deacetylation of cytoplasmic proteins and autophagy seems rather unambiguous, it appears less straightforward with respect to nuclear proteins. On the one hand, the pro-autophagic transcriptional response has been linked, for example, to SIRT1- and spermidine-induced deacetylation of histones H4 and H3 (Eisenberg et al., 2009), respectively. On the other hand, glucose deprivation stimulates AMPK activation with the final result that acetyl-CoA synthetase 2 (ACSS2) phosphorylated by AMPK translocates to the nucleus where it interacts with the transcription factor EB (TFEB) and binds to promoter regions of autophagy genes, locally producing acetyl-CoA and favoring pro-autophagic H3 hyperacetylation (Li et al., 2017b). These divergent outcomes may reflect feedback loops that impose a self-limitation on the autophagic process. For instance, rapamycin-induced autophagy is coupled to the hypoacetylation of H4K16 following the downregulation of lysine acetyltransferase 8 (KAT8), thereby reducing the transcription of pro-autophagic genes (Fullgrabe et al., 2013 € ). 

Besides autophagy-regulatory and executory proteins, deacetylation may also affect autophagic substrates. Depletion of general control of amino acid synthesis 5 (GCN5) like-1 (GCN5L1), a component of the mitochondrial acetyltransferase machinery, leads to mitochondrial protein deacetylation catalyzed by SIRT3, thus favoring mitophagy (Webster et al., 2013). SIRT1 deacetylates mitofusin-2, a protein tethered to the mitochondrial membrane, facilitating SIRT1-induced autophagy and mitophagy (Biel et al., 2016). As a further example, EP300 inhibition reduces the acetylation of Tau (a protein that forms pathogenic intraneuronal aggregates in Alzheimer’s disease), which favors its clearance by autophagy (Min et al., 2015).Bona fide CRMs and Candidate CRMs Several agents may be considered as CRMs since they cause protein deacetylation deriving in autophagy induction (Figure 2). We suggest that CRMs should also have the capacity to reproducibly extend lifespan and/or healthspan in model organisms, hence extending the functional definition of CRMs by another criterion. Here, we enumerate compounds that either fully comply with these stringent criteria (bona fide CRMs) or that do so at least partially according to the current state-of-the-art (potential CRMs) (Table 1). Resveratrol and Other SIRT1 Activators Resveratrol is a polyphenolic phytoalexin that is particularly abundant in the skin of grapes and in red wine. It has been shown to promote longevity across species and to improve age-related parameters in mice. 

However, resveratrol seems to only prolong the lifespan of mice on a high-fat diet (HFD) (Baur et al., 2006), but not on regular chow. Still, resveratrol exerts a number of protective effects in mammalian models of metabolic syndrome, type 2 diabetes (an effect that is enhanced when resveratrol is combined with metformin), cancer, neurodegeneration, and CVD (Rajman et al., 2018). However, contrary findings have been reported recently on its efficacy against metabolic syndrome (Kjær et al., 2017). Interestingly, resveratrol can counteract the reduction of duodenal SIRT1 levels in rats fed an HFD, which is accompanied by improved insulin sensitivity (Coˆ te´ et al., 2015). This indicates the potential of resveratrol as an agent to counteract obesity- and diabetes-induced insulin resistance as well as dysregulated glucose homeostasis. Moreover, resveratrol induces a CR-like transcriptional signature in mice and recapitulates metabolic changes of CR in humans (Timmers et al., 2011). Several studies have examined resveratrol on primates, also showing SIRT1 induction, NF-kB repression, improved insulin signaling, and attenuated inflammation in adipose tissue of high-fat, high-sugar (HFS)-fed animals (Rajman et al., 2018), coupled to reduced CVD risk parameters induced by HFS (Mattison et al., 2014). A large number of clinical trials assessing its effects on cancer, diabetes, obesity, non-alcoholic fatty liver (NAFL), neurological disease, and CVDs have been performed with mostly beneficial outcomes. Resveratrol targets a number of stress-related cellular components, including AMPK (Rajman et al., 2018), which might represent a major molecular target, and the NAD+ -dependent deacetylase SIRT1. 

 For instance, genetic ablation of autophagy abrogates spermidine-mediated lifespan extension in yeast, nematodes, and flies and attenuates cardioprotective effects (Eisenberg et al., 2016) in mice. Spermidine inhibits the activity of several acetyltransferases (Eisenberg et al., 2009), including EP300, and this suffices for autophagy induction (Pietrocola et al., 2015). Intriguingly, these pro-autophagic deacetylation effects are synergistic with those of resveratrol (Morselli et al., 2011), which instead promotes the deacetylase activity of SIRT1 (see above). Moreover, spermidine has been shown to inhibit mTORC1 and activate AMPK (Marin˜ o et al., 2014). It has also been speculated that spermidine might posttranslationally hypusinate the translation factor eIF5A, which leads to the synthesis of the pro-autophagy transcription factor TFEB, at least in immune cells (Zhang et al., 2018). Moreover, spermidine can promote mitophagy (a specialized form of autophagy that eliminates damaged or dysfunctional mitochondria) in cell culture (Qi et al., 2016) and mice (Eisenberg et al., 2016). In human cells, this depends on ataxia-telangiectasia mutated protein kinase (ATM) and consequently on the phosphatase and tensin homolog (PTEN)-induced putative kinase 1 (PINK1) (Qi et al., 2016), which has been linked to the promotion of mitophagy (Eiyama and Okamoto, 2015). Spermidine is naturally produced in the body by cellular biosynthesis as well as by the intestinal microbiota.