|Home | About | Journals | Submit | Contact Us | Français|
Dietary restriction (DR) can extend lifespan and reduce disease burden across a wide range of animals and yeast but the mechanisms mediating these remarkably protective effects remain to be elucidated despite extensive efforts. Although it has generally been assumed that protective effects of DR are cell-autonomous, there is considerable evidence that many whole-body responses to nutritional state, including DR, are regulated by nutrient-sensing neurons. In this review, we explore the hypothesis that nutrient sensing neurons in the ventromedial hypothalamus hierarchically regulate the protective responses of dietary restriction. We describe multiple peripheral responses that are hierarchically regulated by the hypothalamus and we present evidence for non-cell autonomous signaling of dietary restriction gathered from a diverse range of models including invertebrates, mammalian cell culture, and rodent studies.
Dietary restriction (DR) has been shown to extend lifespan and slow aging in highly divergent taxa including yeast, worms, spiders, flies, fish, mice, and rats . DR can also protect against age-related diseases, including cancer [79, 201, 210–212], kidney disease , and models of Alzheimer's disease        , Parkinson's disease   , Huntington's disease  , stroke  and other neurodegenerative diseases [27, 179]. Whether lifespan extension will extend to primates is currently unclear. Nevertheless, many of the physiological responses to DR are shared between humans, non-human primates, and other mammals including decreased cardiovascular disease risk factors, improved glucose regulation, decreased body temperature, and decreased inflammation     .
While all cells in principle are capable of responding to nutritional availability, animals optimize responses to nutritional deficits by coordinating organ and tissue responses. In mammals the most important cells organizing responses to nutritional deficit are arguably the nutrient-sensing neurons of the ventromedial hypothalamus (VMH) which includes both the ventromedial nucleus and the arcuate nucleus[1, 53]. Neurons in these areas of the brain regulate neuroendocrine and autonomic nervous system control of glucose homeostasis, energy balance, and growth hormone (and thus insulin-like growth factor). They may also regulate the response of body temperature to nutritional deprivation through signaling with the preoptic area/anterior hypothalamus . All of these systems are altered during nutritional deficit and potentially contribute to the protective effects of DR. In this review, we develop the hypothesis that nutrient-sensing neurons in the VMH mediate protective effects of DR.
DR elicits complex interacting pathways, many of which have been proposed to mediate protective effects . For example, several signaling pathways are decreased including growth hormone (GH) and its downstream effector insulin-like growth factor-1 (IGF-1) , insulin , glucose metabolism , body temperature , inflammation , and mTOR activity . In contrast, glucocorticoid secretion is increased [75, 97]. Since glucorticoid secretion is also induced by other forms of stress in addition to DR , the latter observation has been used to support the hypothesis that protective effects of DR are due to hormesis, the phenomenon by which low levels of toxic insults can actually enhance functionality, presumably by inducing protective mechanisms . While there may be some merit to this hypothesis, the majority of responses to DR are not generally induced by other stressors, so we do not further explore the applicability of hormesis.
For example, several signaling pathways are decreased including growth hormone (GH) and its downstream effector insulin-like growth factor-1 (IGF-1) , insulin , glucose metabolism , body temperature , inflammation , and mTOR activity . In contrast, glucocorticoid secretion is increased [75, 97]. Since glucorticoid secretion is also induced by other forms of stress in addition to DR , the latter observation has been used to support the hypothesis that protective effects of DR are due to hormesis, the phenomenon by which low levels of toxic insults can actually enhance functionality, presumably by inducing protective mechanisms . While there may be some merit to this hypothesis, the majority of responses to DR are not generally induced by other stressors, so we do not further explore the applicability of hormesis.
Age-related impairments occur in probably all cell types and tissues mediated by mechanisms largely specific to those tissues (e.g., ovarian atresia and thymic involution [137, 213]. Furthermore mechanisms to respond to nutrient supplies also probably exist in all cell types and tissues (e.g., as with the lac operon, the presence of metabolites often induces the machinery necessary for their metabolism and reduces alternative metabolic pathways . Similarly, as described in detail below, mechanisms producing circadian rhythms are also present in most cell types. Nevertheless these mechanisms may be highly influenced and effectively controlled by neuroendocrine mechanisms, as is the case with circadian rhythms. Herein we make the case that nutrient-sensing hypothalamic neurons mediate many effects of dietary restriction on cellular function during aging.
As with aging mechanisms in general, much of the most compelling evidence for mechanisms mediating protective effects of DR derives from elegant experiments in non-mammalian model organisms. These studies have demonstrated that extension of lifespan by DR either requires or is heavily modulated by small populations of nutrient sensing cells.
The best evidence that nutrient-sensing neurons mediate protective effects of DR derives from studies in C. elegans focusing on two ASI sensory neurons that regulate endocrine signals and peripheral metabolism in response to food availability . Based on the known protective effects of the mammalian anti-oxidant transcription factor Nrf-1 , Bishop et al. hypothesized and demonstrated that DR lifespan extension in C. elegans requires Skn-1, the C. elegans homolog of Nrf-1 . Remarkably, DR only induced Skn-1 in 2 neurons, the ASI chemo-nutrient sensory neurons. Ablation of these neurons blocked effects of DR to increase lifespan, and expressing Skn-1 only in these neurons completely supported protective effects of DR to increase lifespan . Skn-1 specifically in these neurons alone also mediated effects of DR on whole-body oxygen consumption . This seminal study demonstrated in C. elegans that nutrient sensing neurons regulate the metabolism of the entire body and are sufficient to mediate effects of DR on lifespan.
Further evidence that nutrient-sensing neurons mediate protective effects of DR comes from a series of elegant studies in Drosophila by Partridge and colleagues. In the adult brain of Drosophila, median neurosecretory (MNC) cells respond to nutrient availability by producing insulin-like peptides (DILPs). Although ablation of these cells does not affect feeding, it does block the effects of food availability on lifespan and fecundity . Partial ablation of MNC cells increases lifespan, increases fasting glucose, increases lipid and carbohydrate storage in peripheral tissues, and reduces fecundity [24, 25]. Furthermore, although the loss of DILPs from MNCs is partially compensated for by up-regulation of DILP6 in fat bodies, the effects of DR on longevity and fecundity require DILPs produced by neurons . Thus MNCs in Drosophila and ASI neurons in C. elegans share many features and suggest that such sensory cells can regulate lifespan independent of food intake.
Sensory cells controlling lifespan may sense nutritional state through both caloric and chemosensory mechanisms. In both Drosophila and C. elegans, exposure to nutrient-derived odors partially blocks the protective effects of DR on lifespan without affecting feeding   . Furthermore, DR increases lifespan when initiated in aged worms but this effect is independent of calorie intake because animals at this age no longer consume bacterial food . The non-caloric signal of these effects is unknown, but several studies have demonstrated that a defect in sensory signaling can extend lifespan without altering food intake. These defects have included defective sensory cilia , a mutation in the odorant receptor Or83b  or the carbon dioxide olfactory sensor Gr63a , loss of sensory G proteins , or ablation of sensory cells [24, 25]  . Although the ablation of either gustatory or olfactory neurons extends lifespan in C. elegans, the effects of gustatory neurons are completely daf-16 dependent while the olfactory neuron effects are not, indicating that sensory cells in C. elegans can modulate longevity via more than one circuit .
Several of the studies discussed above report that, in invertebrates, nutrient sensing cells may regulate longevity and DR through effects on whole-body metabolism   , nutrient storage and availability  , and stress resistance  . Together, these invertebrate studies indicate that small populations of nutrient sensing cells regulate peripheral responses to nutritional state as well as lifespan extension by DR.
As with all model organism experiments, it is unclear whether these principles from invertebrate animals will be translatable to humans and mammals. Nevertheless, many components of the responses to DR, including lifespan extension and/or health benefits, are similar across invertebrates and mammals . Mammalian in vitro studies support the idea that intracellular responses to DR can be elicited in a non-cell-autonomous manner. Culturing mammalian cells with serum from rats, monkeys, or humans subjected to DR activates intracellular responses that mimic some effects of DR, including reduced cell proliferation, increased oxidative stress tolerance, increased stress response genes, increased Sirt1 expression, and decreased responses to TNFalpha [5, 44, 46].
If nutrient sensing cells mediate lifespan extension by DR, as suggested by the invertebrate studies, which mammalian cell types correspond to ASI neurons in C. elegans and MNC cells in Drosophila? A major clue is that ASI neurons (along with a few others) promote satiety in C. elegans . In mammals, the classic “satiety” center of the brain is the ventromedial hypothalamus (VMH; including neurons in the ventromedial nucleus as well as the arcuate nucleus). Damage to this area causes loss of satiety and obesity (the classic “VMH” or simply “hypothalamic” obesity syndrome ). We have demonstrated that VMH neurons also exhibit many molecular and functional similarities to pancreatic beta cells . For example, VMH neurons are uniquely sensitive to stimulation by glucose [148, 218], and stimulation of VMH neurons stimulates peripheral glucose metabolism [77, 131, 132, 173, 186, 194–196]. Interestingly, both ASI and MNC cells also exhibit many molecular and functional similarities to pancreatic beta cells including sensitivity to nutrients, production of insulin-like peptides, and promotion of glucose utilization . We hypothesize that ancestral cellular precursors to VMH neurons and pancreatic beta cells supported functions similar to those of ASI and MNC neurons. However, over the course of evolution (perhaps due to greater body size of vertebrates), the cells diverged into two populations of glucose stimulated cells: one in the pancreas relatively near the intestinal site of nutrient absorption, and one in the hypothalamus at the site of hierarchical control of glucose homeostasis.
Lesion studies have supported that the protective effects of DR require nutrient sensing neurons in the VMH. DR normally suppresses tumorigenesis in rodents but this protective effect was blocked in two separate mouse models of impaired function of the VMH arcuate nucleus: lesion of the arcuate nucleus with monosodium glutamate or genetic deletion of a key arcuate neuropeptide NPY . Arcuate nucleus ablation also blocked the DR suppression of fasting blood glucose in the monosodium glutamate model and the DR-induced increase of blood adiponectin in the NPY knockout model . In another study, VMH lesions blocked the effect of DR to reduce resting energy expenditure . These lesion studies, although not conclusive, indicate that the VMH is required for DR to protect against cancer and elicit major systemic changes in energy balance and glucose homeostasis.
Genetic manipulations in mice have also demonstrated that hypothalamic gene expression can be required for some protective responses to either DR or putative pharmacological mimetics of DR like resveratrol and rapamycin. The deacetylase Sirt2 is reported to mediate the protective effects of DR (glucose deprivation) in yeast  and its homolog sirtuins have been implicated in mediating metabolic responses to DR and fasting in mammals . Two weeks of DR induced Sirt1 in the dorsomedial and lateral hypothalamus, though not the VMH, and transgenic mice with overexpression of SIRT1 in neurons enhanced the neurobehavioral response to fasting and DR . In another study, improvements of glucose homeostasis by central or peripheral resveratrol, a DR mimetic and a putative activator of SIRT1, were blocked by the inhibition of SIRT1 in the hypothalamus, possibly via glucose-sensing neurons . Based on these data, it has been speculated that SIRT1 specifically in the hypothalamus may regulate aging .
Hypothalamic neurons may also mediate the protective effects of another putative DR mimetic, rapamycin, which extends lifespan in mice [78, 129] and protects against tumorogenesis  and age-related cognitive deficits . The protective effects of rapamycin are believed to occur through its inhibition of the mTOR nutrient-sensing pathway  Gene expression analysis indicates that DR suppresses hypothalamic mTOR  while feeding and central leptin treatment activate hypothalamic p70 S6 kinase 1a (S6K), an effector of mTOR. Over-activation of S6K in the mediobasal hypothalamus (that includes the VMH) was shown alter energy metabolism and cold tolerance. Importantly, experimental over-activation of S6K in the mediobasal hypothalamus blocked the metabolic and feeding changes responses to fasting and the onset of metabolic syndrome and sustained overeating in response to a high-fat diet . To our knowledge, no studies have similarly tested the role of hypothalamic S6K on responses to DR. However, the data suggest that hypothalamic mTOR and S6K are critical for metabolic responses to altered nutritional state.
The extent to which hypothalamic neurons mediate responses to dietary restriction in humans remains to be fully addressed, but the role of the hypothalamus in coordinating the body's responses to nutritional status appear to be largely similar to other mammals. For example, the role of hypothalamic neurons in regulating neuroendocrine functions and energy balance was first discovered in humans with hypothalamic tumors, and such tumors, called craniopharyngioma, are well-documented to produce similar metabolic impairments as occurs in animals with hypothalamic lesions . Similarly deficiency in leptin, which largely acts through nutrient-sensing neurons in the VMH, produces largely the same metabolic and neuroendocrine impairments in humans as observed in leptin-deficient mice . Since some recent studies failed to observe protective effects of dietary restriction in mice  and non-human primates , and since despite likely caloric restriction in many human populations throughout history there has never been a documented case of the expected life extension, it seems quite likely that the optimum caloric restriction for humans has never been discovered, and likely never will be. Nevertheless in view of the conservation of the relevant regulatory mechanisms, we hypothesize that pharmacological approaches activating these mechanisms may still be protective in humans.
A major motivation to propose that the VMH is a hierarchical mediator of lifespan extension by DR is that it is a hierarchical regulator of many systemic responses to nutrient intake and energy balance. An example of this top-down control is glucose homeostasis.
Much of the evidence for how mammals maintain glucose homeostasis derives from studies of responses to acute peripheral hypoglycemia (e.g., blood glucose below 50 mg/dl). Acute peripheral hypoglycemia produces many of the same neuroendocrine responses as DR including reduction of reproductive hormones , reduced thyroid hormone , reduced IGF-1 , and increased glucocorticoids as well as glucagon, and epinephrine [10, 43, 80, 117]. The combined effect of all these responses is to produce a “counterregulatory” response to hypoglycemia that increases blood glucose, by increasing hepatic glucose production and reducing peripheral (e.g., muscle) glucose metabolism as well .
In contrast to neurons in most areas of the brain, the activity of neurons in the VMH is sensitive to variations within the physiological range of glucose concentrations [52, 148, 149, 191, 218]. Neuroendocrine and autonomic responses to hypoglycemia appear to be mediated largely or entirely by these glucose-sensing neurons in the VMH. For example, ablation of VMH neurons blocks counterregulatory responses to hypoglycemia . Moreover, reducing glucose metabolism in the VMH induces peripheral counterregulatory responses  while increasing glucose  or lactate  levels in the VMH prevents counterregulatory responses to hypoglycemia. Furthermore, blocking K-ATP channels mimics the effect of glucose in glucose-excited cells  and selectively blocking K-ATP channels in the VMH blocks counterregulatory responses to hypoglycemia  whereas activating K-ATP channels in the VMH enhances those responses . The activation of counterregulatory responses by VMH neurons appears to be mediated in part by disinhibition, since hypoglycemia reduces VMH GABA, GABA agonists block counterregulatory responses, and GABA antagonists enhance counterregulatory responses . These data support the hypothesis that counterregulatory responses during hypoglycemia are mediated in part by reduced activity of glucose-stimulated GABA-ergic neurons in the VMH.
Counterregulatory responses to hypoglycemia were also impaired by genetic inhibition of glutamate transmission in the Sf-1 neurons of the VMN . Of particular importance, the blood glucose levels in these mice were relatively low after 24 hours of fasting and this change was associated with impaired induction of hepatic gluconeogenic gene expression . Since acute hypoglycemia is probably rarely encountered under normal circumstances, these observations suggest that neuroendocrine and autonomic responses to acute hypoglycemia probably reflect systems evolved to adapt to more commonly encountered caloric deficits. This is consistent with the observation that some mechanisms mediating responses to hypoglycemia and DR overlap (see below for molecular evidence supporting this hypothesis).
Another line of evidence for the hypothalamic control of systemic glucose homeostasis originates from our studies in leptin-deficient mice. These mice are characterized by reduced expression of hypothalamic POMC . Transgenic restoration of POMC completely normalized blood glucose, associated with normalization of hepatic gluconeogenesis . Furthermore, infusion of glucose into the hypothalamus of normal mice acutely reduces hepatic glucose output . Taken together these studies demonstrate that activity of glucose-stimulated hypothalamic neurons promotes peripheral glucose metabolism through hepatic glucose output.
The hypothalamus can maintain blood glucose homeostasis and increase blood glucose availability by increasing hepatic glucose production but also by reducing metabolism in peripheral organs to preserve blood glucose for brain use. VMH neurons plausibly mediate these effects on glucose metabolism since electrical stimulation of the VMH but not the lateral hypothalamus increases glucose metabolism in peripheral tissues, mediated by enhanced sympathetic activity [194, 195]. Consistent with the results of genetic inhibition of VMH glutamate release , infusion of glutamate into the VMH also enhances peripheral glucose metabolism . Increased glucose metabolism from VMH stimulation does not entail increased glucose transporters, in contrast to insulin-induced glucose uptake . Associated with increased glucose metabolism, VMH stimulation also increases brown adipose tissue temperature , likely reflecting hypothalamic mediation of hypothermia induced by 2DG  (see below). Infusion of leptin specifically into the VMH also induces peripheral glucose metabolism  through enhancement of sympathetic nervous activity . Thus, the VMH can regulate whole-body glucose homeostasis both through hepatic glucose output and through peripheral cellular glucose metabolism, at least in rats, the species in which these studies were carried out.
While VMH neurons clearly regulate peripheral glucose metabolism, glucose metabolism is also regulated by insulin in some tissues and locally (cell autonomously) in all tissues. Thus an important question is the quantitative importance of the VMH in regulating peripheral glucose metabolism compared to these other factors. Many processes are regulated hierarchically both locally and by the hypothalamus, yet hypothalamic regulation is often dominant. For example, insulin secretion is regulated autonomously by glucose metabolism in the pancreatic beta cell in vitro . Nevertheless, in vivo VMH neurons exert major control via the autonomic nervous system over pancreatic secretion of insulin. Insulin secretion increases 4-fold within 30 minutes after VMH lesions, with no change in blood glucose, dependent on the vagus nerve . The magnitude of this change in insulin secretion is comparable to or greater than that produced by glucose alone in vivo  suggesting that VMH neurons play an important role in regulating regulation of insulin secretion. Moreover, VMH neurons can regulate insulin sensitivity in peripheral tissues. For example, mice lacking Fox01 specifically in SF-1 neurons of the VMH had improved glucose tolerance due to improved insulin sensitivity in skeletal muscle and heart . Similarly, even though glucose metabolism is generally driven locally by cellular demand (e.g., reduction of ATP), the effect of insulin on glucose metabolism is dominant in insulin-sensitive cells. Therefore it is plausible that VMH neurons contribute significantly to regulating glucose metabolism during aging and under conditions of dietary restriction.
Another example of the hypothalamic top-down control of peripheral cell-autonomous signaling is leptin. Leptin is key hormone for the regulation of energy balance and metabolism. Although peripheral fat cells release the hormone and peripheral tissues contain leptin receptors that elicit intracellular responses in metabolism and other pathways, leptin receptors in the VMH have critical functions in leptin-regulated energy balance. The loss of leptin receptors from AgRP/NPY neurons of the VMH of young mice caused obesity due to transient hyperphagia and decreased energy expenditure  and restoration of leptin signaling in leptin-deficient mice was sufficient to reverse the phenotypes of leptin deficiency including obesity, diabetes, and infertility [47, 202]. People and animals with obesity are often leptin resistant and intranasal delivery of leptin in rats has been shown to reduce food intake and increase energy expenditure in lean and diet-induced obese rats without increasing serum leptin levels [181, 182].
A third example of hypothalamic hierarchical control is the circadian clock. Most tissues of the body have an internal “clock” with intracellular circadian rhythms of gene expression and activity that can maintain rhythmic behavior in vitro. However, the internal clocks in these tissues oscillate with different phases and have been referred to as “slave oscillators” because they require the hypothalamus to synchronize tissues of the body and maintain population-based amplitude of oscillation , reviewed by Froy and Miskin (2010)  and by Reppert and Weaver (2002) . For example, ablation of the suprachiasmatic nucleus (SCN) of the hypothalamus blocks circadian rhythm in behavior and grafting SCN tissue into arrhythmic hamsters with SCN ablation restores a circadian rhythm that is dependent on the genotype of the grafted SCN tissue rather than the genotype of the host animal . The circadian rhythm can be altered by changes in nutritional state [35, 128] and coordinated whole-body responses to nutritional state and timing appear to require regions of the hypothalamus  [35, 128].
DR effects on circadian rhythm may contribute to the health and lifespan consequences of DR . DR increases the amplitude of circadian changes in locomotor activity in older rhesus monkeys . Changes in the circadian rhythm or wheel running and body temperature have also been shown in rodents [35, 128]. Epidemiological evidence from humans suggests that shift work increases the risk of obesity and age-related disease including diabetes, cardiovascular disease, and cancer (reviewed in [8, 61]. In humans and mice, disruption of circadian rhythm has metabolic consequences including altered leptin and insulin levels  . Lifespan in hamsters and mice is decreased by disruption of circadian rhythm     and lifespan can be extended by successful grafting of fetal SCN into intact adult hamsters that have age-related impairment in circadian rhythm .
Of the systemic responses to DR, which ones are hierarchically regulated by the hypothalamus? One likely candidate is IGF-1. The potential role of IGF-1 and related hormones and homologs in lifespan and the protective effects of DR have been covered by many extensive reviews   . Here we note only that the hypothalamus regulates GH release from the pituitary, which in turn hierarchically regulates the release of IGF-1 from the liver (Figure 1). Hypothalamic regulation of GH release occurs through several pathways. The most commonly known circuits involve the hypothalamic expression of growth-hormone releasing hormone (GHRH) and somatostatin. During DR, serum GH in rats and sheep corresponds with GHRH mRNA levels in the arcuate nucleus of the mediobasal hypothalamus [81, 187]. GH release is also regulated by other hypothalamic peptides including ghrelin, thyrotropin-releasing hormone, pituitary adenylate cyclase-activating peptide, and gonadotropin-releasing hormone (reviewed in ). These systems may be complemented by hypothalamic control of the autonomic nervous system, such that resection of the vagus nerve was shown to blunt GH release in response to GHRH and decreased basal IGF-1 and GH levels . In summary, if the protective effects of DR are mediated by reduced GH/IGF-1 signaling in mammals, this reduction is probably achieved through hypothalamic responses to DR.
DR decreases body temperature in humans , non-human primates, and some though not all mice [166, 167]. Similarly, hypoglycemia and glucopenia can reduce body temperature in humans [60, 140] and other species  [20, 31, 153]. Reducing temperature dramatically increases lifespan in poikilothermic (cold-blooded) animals [40, 84, 112–116, 206, 220] but the effect in mammals has been challenging to determine. The most convincing evidence comes from Conti and colleagues, who produced transgenic mice with overexpression of uncoupling protein 2 (UCP2) specifically in the lateral hypothalamus . This manipulation led to chronic reduction of body temperature by about 0.3 °C because the elevation of UCP2 produced a local increase in temperature at the area of the hypothalamus that hierarchically regulates body temperature based on local thermosensory neurons and peripheral temperature information. Although the transgenic mice consumed equal calories as wild-type controls, they lived 12% (males) and 20% (females) longer .
The degree to which reduced body temperature mediates the protective effects of DR is unclear. One group of rhesus macaques on DR exhibited decreased body temperature  and health benefits with young-onset CR such as decreased cancer incidence but not increased lifespan , strong evidence that it is not a major contributor to lifespan in these animals. On the other hand, analysis of mortality curves indicated that, in mice, lifespan extension by DR resembles lifespan extension by decreased body temperature in that both reduce the age-associated acceleration of mortality rate . One rodent study reports that increasing ambient temperature blocked lifespan extension by DR, primarily through effects on lymphoma rates in a lymphoma-prone mouse strain . Moreover, body temperature correlates with longevity in humans  and across genders in C57Bl/6J mice , and two long-lived rodent models exhibit decreased body temperature [86, 215]. Thus although the evidence clearly indicates that reduction in temperature is not the major mechanism by which DR increases lifespan, a role for temperature cannot yet be completely ruled out.
The circuitry by which DR reduces body temperature is uncertain but, in homeotherms like mammals, the hypothalamus is almost certainly involved. There are numerous pathways by which nutritional status influences whole-body temperature through hypothalamic mechanisms. These have been recently reviewed in depth by Bartfai & Conti so we will not review these here other than to state that several of these proposed pathways involve nutrient sensing and/or glucose-sensing neurons, including those in the arcuate nucleus of the hypothalamus .
Whether or not thermoregulation is a critical protective component of DR, it is certainly an important yet often overlooked variable in experimental design of DR studies. DR of rats in a cool environment (12°C) caused greater weight loss and suppressed oxygen consumption compared to DR rats kept at a thermoneutral environment (30°C) . Standard laboratory environments are usually below the thermoneutral zone of most rodents  and thus require additional energy costs to maintain core temperature by shivering or non-shivering thermogenesis. Since lifespan extension by DR correlates with fuel efficiency across strains of mice, defined as the ability to maintain growth and body weight independently of calories , it is possible that some discrepancies in lifespan extension by DR may be due to variations in laboratory conditions that alter heat exchange including animal size and strain as well as environmental factors such as ambient temperature, air flow, bedding, and individual versus group housing .
During DR, there is a relative metabolic shift from glucose to the use of other fuel substrates including free fatty acids and ketone bodies . Reducing glucose is sufficient to increase lifespan in flies  while reducing glucose metabolism increases lifespan in C. elegans . In mammals, DR reduces average plasma glucose levels  and genes that increase lifespan entail reduction in whole-body glucose homeostasis . Conversely elevated glucose reduces lifespan in humans  and C. elegans [107, 183] suggesting that the protective effects of DR may be mediated by a reduction in glucose metabolism [122, 136]. On the other hand, genetic reduction of plasma glucose in mice by overexpression of glucose transporters in muscle does not increase lifespan . Therefore the reduction in glucose levels by DR is not sufficient, though it may be necessary, to mediate the full protective effects of DR on lifespan in animals. We suggest however that DR also reduces glucose metabolism via neuroendocrine and autonomic regulation by glucose-sensing VMH neurons.
In support of this hypothesis, many studies indicate that glucose toxicity is mediated by glucose metabolism . The metabolism of glucose, relative to alternative substrates, preferentially produces oxidative stress, in part because it preferentially produces NADH (relative to FADH2 and NADPH), which is oxidized at Complex 1, the main mitochondrial source of reactive oxygen species . Glucose metabolism is also more efficient than alternative energy sources because NADH metabolism produces more ATP per carbon bond . This may explain why increased respiration has been reported to be associated with the protective effects of DR in some [14, 111], though not all  studies. During DR, the metabolism of glucose is decreased relative to other substrates, supporting that DR does not increase lifespan by reducing metabolic rate per se .
Although the shift in metabolism from glucose to other substrates could be protective through decreased oxidative stress, recent studies indicate that increased oxidative stress does not always reduce lifespan . Our own studies indicate that increasing oxidative stress by inactivation of SOD isoforms has no effect on lifespan and specifically no effect on life extension by DR . However, these studies do not address if DR extends lifespan by specifically reducing oxidative stress produced by glucose metabolism, which will in general entail different compartmentalization and cellular targets than oxidative stress produced by other processes. Reducing glucose metabolism also has other protective effects independent of oxidative stress , including AMPK induction of autophagy .
Glucose sensitive neurons in the VMH are ideally situated to hierarchically regulate the systemic metabolic shift from glucose to other substrates during DR. Much of the evidence for this is described above. Glucose sensing neurons in the VMH share similarities with the ASI neurons in C.elegans and the MNC cells in Drosophila that mediate effects of DR on lifespan, as described above. VMH nutrient sensing neurons regulate peripheral glucose metabolism through the sympathetic nervous system [77, 195], and reduced hepatic glucose production. VMH control of neuroendocrine systems could also contribute to this effect, as sex steroids , thyroid hormone , and IGF-1  stimulate glycolysis, whereas glucocorticoids inhibit glycolysis . VMH neurons are required for peripheral metabolic responses to fasting  and hypoglycemia [16, 18] (see above). Many of these peripheral metabolic responses also occur during DR [72, 73], fasting  and hypoglycemia .
This review argues that glucose-sensing neurons in the VMH hierarchically mediate many key protective mechanisms produce by DR. If this hypothesis is correct, then a major question is what are the components of hypothalamic nutrient sensing and molecular signaling that mediate the key responses to DR? In this section, we will discuss evidence for a metabolic shift from glucose to other fuel substrates. In addition, we will discuss several molecular candidates that may be involved in hypothalamic responses to dietary restriction such as peroxisome proliferator-activated receptor alpha (Ppar-alpha), Fox01, mammalian target of rapamycin (mTOR) and Creb-binding protein (Cbp) (Figure 1).
As described above, there is strong evidence that the glucose-stimulated neurons in the VMH mediate neuroendocrine responses to hypoglycemia and that many of these neuroendocrine responses also occur during DR [16, 18, 19, 54]. However, DR and hypoglycemia are not identical and the reduction in blood glucose during DR is not sufficient to activate the neuroendocrine responses to DR . Nevertheless, glucose-sensing VMH neurons are positioned to respond to other plasma signals of energy balance that are altered during DR such as decreased leptin and ketone bodies [16, 18, 19, 54, 127]. Reduced leptin and glucose may synergize to affect VMH neuronal activity since leptin enhances hypothalamic responses to glucose  and hypothalamic glucose metabolism is required for leptin signaling  . Specifically, glucose deprivation or blocking glucose metabolism specifically blocks molecular responses (phosphorylation in the JAK-STAT pathway) to leptin but not GH .
Free fatty acids are elevated by DR  and may heighten the VMH neuronal response to decreased glucose by reducing the hypothalamic glucose metabolism that is a critical component of glucose sensing in these neurons [218, 219]. Specifically, free fatty acids can activate Ppar-alpha, although this effect varies by the chain length and saturation of the free fatty acids impacts the activation of Ppar-alpha. We have recently observed that both fasting  and hypoglycemia  induce Ppar-alpha target genes. In turn, activation of Ppar-alpha promotes fatty acid beta-oxidation and reduces glucose metabolism [33, 165]. Furthermore, activation of hypothalamic Ppar-alpha reduces peripheral glucose metabolism  and activation of hypothalamic Ppar-alpha reduces glucose metabolism in the periphery . The complex potential role of Ppar-alpha and free fatty acids in impacting hypothalamic glucose metabolism was covered in a recent review .
In addition to free fatty acids, DR increases plasma ketones in humans  and we have shown that ketones block glucose metabolism in hypothalamic neurons . Thus, an increase in either fatty acids or ketone bodies would be expected to decrease hypothalamic glucose metabolism which would in turn be expected to reduce peripheral glucose metabolism [77, 131, 132, 153, 186, 194, 195, 199]. This circuit would tend to be self-perpetuating, possibly explaining why some long-term protective metabolic effects of DR persist even after return to ad lib feeding .
Evidence for the metabolic shift from glucose to fatty acid metabolism in the hypothalamus originates, in part, from studies on a large panel of hypothalamic genes induced both by acute nutritional deprivation  and acute hypoglycemia . Some of these responses overlap with the genes induced by chronic DR . The gene expression pattern indicated reduced hypothalamic glucose metabolism relative to other fuel sources (e.g. increased fatty acid beta oxidation) [157, 158]. Chronic DR produces a similar metabolic shift away from glycolysis and toward the use of alternate fuels in heart , liver , as well as the hypothalamus and other brain regions . This molecular profile is also reflected in whole-body metabolism, since chronic DR produces a rapid shift toward greater 24-hour fat oxidation (accompanied by increased fatty acid synthesis) relative to ad lib fed animals .
The hypothalamic expression of Creb-binding protein (Cbp) may be an important mediator of the protective effects of DR . We have demonstrated in two independent groups of animals that expression of Cbp in the VMH, but not liver or fat accounts for over 80% of lifespan in 5 strains of mice . In contrast, hypothalamic expression of other genes implicated in aging did not significantly correlate with lifespan (mammalian orthologs to sirtuins, daf-16/Fox03a, pha-4/FoxA, rheb-1, hsf-1, skn-1/NFE2, hif-1, and aak-2/ampk). The relevance of this robust correlation of Cbp is supported by our studies in C.elegans that showed that the induction of Cbp is required for protective effects of DR, including lifespan extension, and also for the animals' metabolic switch away from glycolysis and toward beta-oxidation .
Cbp is a co-activator of Ppar-gamma  and Ppar-alpha  as well as FoxO1/daf-16 . The FoxO1/daf-16 transcription factor, which is inhibited by insulin-like signaling and whose induction is required for the protective effects of insulin-like signaling, is also necessary for the protective effects of some, though not all, protocols of DR [67, 223]. FoxO1 appears to mediate some neuroendocrine hypothalamic responses to fasting [30, 96]. In addition, the loss of Fox01 specifically from SF-1 neurons of the ventromedial hypothalamus resulted in lean mice with increased energy expenditure and improved glucose tolerance with improved insulin sensitivity in skeletal muscle and heart . The response to 24-hour fasting was also altered in these mice with increased weight loss and a blunted decrease in energy expenditure during the fast . The role of hypothalamic FoxO1 the responses to specifically DR remain to be assessed.
Cbp is also induced by reduced insulin-like signaling and is required for the protective effects of this reduction [174, 223]. Cbp is also a co-activator of cyclic AMP-response element binding protein (Creb), a transcription factor that directs hepatic counter-regulatory responses to hypoglycemia by promoting gluconeogenesis and glycogenolysis . In contrast to the co-activity of Cbp with these transcriptional co-activators, Cbp activity is inhibited by CtBP through an NADH-dependent mechanism [94, 185] that may mediate molecular effects of glucose and the inhibition of those effects by ketones . Interestingly, Creb in the brain is required for DR to induce Sirt-1 and associated metabolic and neuroprotective responses. Since Cbp is required for Creb activity , these results suggest that Cbp also mediates effects of DR on Sirt-1 transcription. Of further interest, Ctbp represses Sirt1 through a nutrient-dependent pathway . Mechanisms mediating effects of DR on Sirt-1 transcription are of increasing interest since recent studies suggest that sirtuin activity may be independent of NAD+ availability . Recent studies also suggest that protective effects of resveratrol, previously thought to act through Sirt-1, may actually be mediated through the elevation of cAMP  which would enhance Cbp activity.
The protective effects of DR may also involve hypothalamic expression of mTOR (mammalian target of rapamycin). Inhibition of the TOR pathway increases lifespan in C. elegans [89, 203], flies  yeast , and mice [78, 129]. DR inhibits TOR activity and produces many effects (e.g., reduced protein synthesis) similar to inhibiting TOR activity . Of particular interest, inhibition of TOR activity and DR did not produce additive effects on lifespan , suggesting that protective effects of DR are mediated by inhibition of mTOR. The effects of DR on hypothalamic gene expression are consistent with a role for mTOR in protective effects of DR . Selective inhibition of hypothalamic mTOR activity produced similar molecular and behavioral effects as DR [15, 42]. One mechanism by which DR may inhibit hypothalamic mTOR is through the activation of hypothalamic AMPK, which mediates some effects of glucose and leptin on hypothalamic function . Induction of AMPK also appears to be required for some, though not all, protective effects of DR [67, 68]. Finally, as indicated above, DR induces Sirt1 in the hypothalamus and hypothalamic Sirt1 appears to mediate at least some effects of DR in mice [99, 177].
One major candidate involved in inhibition of molecular effects of glucose is the transcriptional co-repressor C-terminal binding protein (CtBP) , which is regulated by NADH . Free fatty acids inhibit glucose metabolism and induce beta oxidation mainly by activating the peroxisome proliferator-activated receptor (Ppar) family of transcriptional co-activators , which we observed also seems to mediate some molecular responses to hypoglycemia . Indeed, chronic DR also induces the Pparcoactivator PGC-1alpha . Finally, glucocorticoids inhibit brain glucose utilization , possibly in part by inducing Ppar activity . Thus hypothalamic responses to fasting and DR may involve several key protein molecular mechanisms involving Cbp, Creb, CtBP, mTOR, sirtuins, FoxO, and Ppars (Figure 1).
A small number of nutrient-sensing cells in C. elegans (ASI neurons)  and Drosophila (MNC cells)  largely or entirely regulate the protective effects of DR to increase lifespan. A similar system has been neither proven nor disproven yet in mammals, but it is known that mammalian hypothalamic neurons mediate many responses to nutritional state and deprivation, including some responses to DR. VMH neurons share features of ASI and MNC neurons and mediate neuroendocrine and autonomic responses to low glucose and leptin. Several lines of evidence suggest that glucose metabolism in the hypothalamus (primarily VMH) drives glucose metabolism and glucose levels in the brain and periphery. Based on these and other data, we hypothesize that DR decreases glucose metabolism in the hypothalamus and that this effect mediates the reduction of glucose metabolism in the periphery, thus slowing aging in these tissues. It's possible that this system involves hypothalamic expression of Cbp, which has been shown in mice to predict lifespan across strains. On the other hand, hypothalamic neurons also mediate effects of nutrients to reduce temperature and the GH/IGF-1 pathway. Taken together these observations support that hypothalamic neurons have the potential to regulate many metabolic and protective effects of DR during aging. If so, then the VMH nutrient sensing neurons may eventually be a useful target for the development of calorie restriction mimetics to protect against age-related disease. Advances in molecular technology may enable the studies needed to fully elucidate the role of the hypothalamus and nutrient sensing cells as top-down hierarchical regulators of the protective responses of DR.
We would like to thank Dr. Andrew Dacks for his feedback on the manuscript and his skilled editing of our figures. This work was supported by the NIH National Institute of Diabetes and Digestive and Kidney Diseases and The Klarman Family Foundation, Boston, MA. PA Dacks was supported by the Hilda and Preston Davis Foundation.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.