Learn more: PMC Disclaimer | PMC Copyright Notice
Torpor induction in mammals: Recent discoveries fueling new ideas
Abstract
When faced with a harsh climate or inadequate food, some mammals enter a state of suspended animation known as torpor. A major goal of torpor research is to determine mechanisms that integrate environmental cues, gene expression and metabolism to produce periods of torpor lasting from hours to weeks. Recent discoveries spanning the metazoa suggest that sirtuins, the mammalian circadian clock, fibroblast growth factor 21 (FGF21) and lipids are involved in torpor induction. For example, sirtuins link cellular energy status to the mammalian circadian clock, oxidative stress, and metabolic fuel selection. In this review, we discuss how these recent discoveries form a new hypothesis linking changes in the physical environment with changes in the expression of genes that regulate torpor induction.
Torpor in mammals
All animals require energy from food to sustain normal physiology and behavior. The quantity and quality of food provided by the environment might fluctuate day-to-day or season-to-season such that natural selection has favored animals adapted to these fluctuations. Torpor, the controlled lowering of metabolic rate, body temperature, and physical activity is a highly successful adaptation that various mammals use to cope with periods of low food availability (Figure 1 and Box 1).
Environmental influence on the induction of torpor in mammals. Torpor has evolved as an adaptation for conserving energy during periods of inadequate food. Many regions of the earth are subject to extreme fluctuations in day length, temperature, and rainfall. These environmental changes directly affect the quantity and quality of food that is available to resident animals. When an animal’s food becomes scarce, it is forced to enter a period of fasting. Some animals conserve energy during periods of low food availability by reducing their metabolic rate and entering torpor. In mammals, torpor conserves fuel reserves by drastically reducing body functions that depend on the continuous input of energy. The most recognizable characteristics of torpor in mammals are profound reductions in heart rate, body temperature and physical activity.
As homeotherms, mammals maintain a constant warm body temperature (approximately 35 to 38°C) despite continuous temperature variation in their environment. The ability to maintain a constant body temperature has contributed to the evolutionary success of mammals, allowing the group to invade unoccupied niches and become distributed worldwide. Some mammals periodically turn down their internal thermostat and enter torpor as a means to survive periods of low food availability and then re-warm and return to a normal level of activity when the environment becomes favorable. Various degrees of torpor are observed over a phylogenetically wide range of mammals (Figure 2 and Box 1) -- from the shallow torpor seen in starved laboratory mice (Mus musculus) [1] to daily torpor in Siberian hamsters (Phodopus sungorus) [2] to the champion of all torpid mammals, hibernating Arctic ground squirrels (Spermophilus parryii) that exhibit body temperatures as low as −2.9 °C [3] (Box 1).
Mammalian groups that display torpor. Torpor is widespread among mammals, suggesting that the relevant metabolic pathways and physiological processes are ancestrally derived. The dendrogram shows the phylogenetic relationships of families (each leaf of the dendrogram) within eight mammalian groups in which some members exhibit torpor. The type(s) of torpor exhibited by a specific family is indicated by color on the leaves of the dendrogram (see figure key). Representative animals that exhibit a form of torpor are depicted for each group except the Xenarthra (anteaters and armadillos). Groups of land-dwelling mammals for which torpor has not been described in any member are excluded from the dendrogram. Excluded groups include the superorder Cetartiodactyla (cattle, camels, and deer) and the orders Proboscidae (elephants), Dermoptera (flying lemurs), Lagomorpha (hares and rabbits), Scandentia (tree shrews), Perissodactyla (horses and rhinoceroses), and Pholidota (pangolins). The dendrogram was prepared using the interactive tree of life (iTOL) [103].
The expression of torpor is responsive to the availability of food. It is well established that the laboratory mouse can enter daily torpor when fasted [1]. Variable expression of torpor in response to the availability of food was elegantly demonstrated in the Eastern chipmunk (Tamias striatus) [4, 5]. Chipmunks are food-storing hibernators that hoard food within their burrow prior to the onset of torpor [6]. During the hibernation season, chipmunks whose food hoard was supplemented above the level available by normal foraging had a higher frequency of normothermy with minimum body temperatures 5 to 10 °C higher than that of chipmunks whose food stores were not supplemented [5].
Many natural compounds have been shown to induce a torpor-like phenotype in mammals (reviewed in [7]). However, the mechanisms that integrate the environmental, physiological, metabolic, and molecular changes associated with torpor are largely unknown. Interestingly, much has been learned from genetic mutations in both invertebrate and vertebrate organisms that phenocopy the biochemistry and physiology of mammalian torpor. In this review, we highlight recent discoveries that demonstrate how nutrient-sensing molecules integrate gene expression with cellular energy status, and the increasingly recognized role of lipids as signaling molecules. We also discuss potential mechanisms of mammalian torpor induction that are beginning to emerge based on these new discoveries.
Timing of torpor
Recent studies in mice demonstrate a connection between cellular nutrient status and the core molecular machinery of the mammalian circadian clock [8, 9]. Discovery of the link between environmental changes, metabolism and cellular time-keeping has great potential for determining the mechanism(s) that induces torpor. Both hibernation and daily torpor are often preceded by a period of fasting or reduced food consumption that alters cellular nutrient status [10]. Indicators of cellular nutrient status are the ratios of intracellular [5′-AMP] to [ATP] and [NAD+] to [NADH] [11, 12]. Fasting reduces cellular nutrient status and increases both ratios. The cell can attempt to recover nutrient status by remodeling metabolism so that stored lipid becomes the primary source of fuel [13].
Circadian oscillation of cellular nutrient status and metabolic substrate use has been observed in mammals and is tightly coupled to feeding schedule [8, 11, 14]. When the timing of food availability is altered, the metabolic cycle is shifted to correlate with the new feeding schedule [15–17]. There is substantial evidence supporting the existence of a clock mechanism that paces circadian oscillations in response to regularity of feeding (reviewed in [18]). These results suggest that availability of food, in addition to other environmental factors such as day length and temperature, might strongly regulate the timing of torpor induction.
The expression of genes that encode metabolic enzymes also follows a circadian pattern in the brain and liver of mouse [19]. A key mitochondrial enzyme involved in shutting down carbohydrate oxidation during the deep torpor associated with mammalian hibernation, pyruvate dehydrogenase kinase-isoenzyme 4 (PDK4; [20, 21]), also shows a circadian pattern of expression in rat (Rattus norvegicus) cardiomyocytes [22, 23] and in mouse liver, aorta, and kidney [19]. Genetic mutations that disrupt function of the core circadian clock proteins, circadian locomotor output cycles kaput (CLOCK) and brain and muscle aryl hydrocarbon receptor nuclear translocator (BMAL1), are associated with metabolic disorders in mice and humans (reviewed in [24]). Such a synergistic relationship between the mammalian circadian clock and metabolism is likely required to integrate an animal’s changing environment with molecular and metabolic cycles to induce torpor (Figure 1). An integrative mechanism that accounts for this synergy has been suggested for mammals [25], but its experimental demonstration has only just begun.
Nampt and SIRT1 link cellular nutrient status to the circadian clock and torpor
Recent studies by Ramsey et al. [8] and Nakahata et al. [9] reveal a direct link between metabolism and the core transcription-translation feedback loop of the mammalian circadian clock in mice. At the center of this link are nicotinamide phosphoribosyltransferase (Nampt), the rate limiting enzyme in the major mammalian NAD+ salvage pathway [26, 27], and silent mating type information regulation 2 - homolog 1 (SIRT1), an NAD+-activated deacetylase [28]. These studies demonstrate that cellular nutrient status reflected by intracellular [NAD+] has a strong influence on transcription of genes at the core of circadian rhythms and metabolism (reviewed in [24, 29–31]). These discoveries suggest a new and exciting avenue of investigation that is likely to reveal mechanisms that integrate nutrient status, biological timing, and metabolic substrate use during torpor. In fact, the concentration of NAD in liver of thirteen-lined ground squirrels (Spermophilus tridecemlineatus) entering torpor is on average two to three times higher than that found in summer active ground squirrels, or ground squirrels that are about to exit torpor and return to activity [32], suggesting that Nampt and SIRT1 are likely to be activated during torpor induction.
A hypothetical role for Nampt, SIRT1 and the core circadian clock genes in the induction of torpor can be easily envisioned (Figure 3). First, as a result of reduced food supply due to a changing environment, an animal enters a period of fasting. During a fast, cellular [NAD+] increases due to nutrient depletion [16, 17, 19, 22], causing SIRT1 deacetylase activity to increase. SIRT1 activation of Nampt activity upregulates the NAD+ salvage pathway, maintaining elevated [NAD+] and reinforcing SIRT1 activity. Activated SIRT1 also deacetylates the circadian clock protein BMAL1, inactivating the heterodimeric transcription activation complex that BMAL1 forms with CLOCK. Inactivation of the BMAL1:CLOCK complex inhibits expression of the circadian clock genes Per and Cry [33, 34], disrupting the normal circadian feedback loop and causing the clock to either slow or stand still. This would prolong a transcription profile typical of an animal’s inactive phase (e.g. during sleep) for genes regulated by the circadian clock.
A model for torpor induction in mammals that links reduced availability of food in the environment with the energy status of the cell, the mammalian circadian clock, and response to oxidative stress. An animal begins to fast as a result of reduced availability of food in the environment. Fasting results in depletion of cellular nutrients and increased intracellular concentrations of 5′-adenosine monophosphate (AMP) and oxidized nicotinamide adenine dinucleotide (NAD+). As intracellular [NAD+] increases, the activity of silent mating type information regulation 2 - homolog 1 (SIRT1) deacetylase is also increased. Activated SIRT1 can enter the nucleus and deacetylate many targets including the core circadian clock protein, brain and muscle aryl hydrocarbon receptor nuclear translocator (BMAL1). Deacetylation of BMAL1 inhibits transcriptional activation of genes that are regulated by the heterodimeric transcription factor complex formed between BMAL1 and circadian locomotor output cycles kaput (BMAL1:CLOCK). This might prolong a gene expression profile that is typical of that during an animal’s inactive period, e.g. when the animal is normally resting. SIRT1 might also activate nuclear receptor proteins that promote transcription of enzymes that function in lipolysis and in resistance to hypoxic stress (hypoxia inducible factor-2α; HIF-2α). Activation of lipolytic enzymes is part of the switch from carbohydrate to lipid-based metabolism. SIRT1 might increase biosynthesis of its own activating cofactor, NAD+, by deacetylation and activation of nicotinamide phosphoribosyltransferase (NAMPT). NAMPT is the rate-limiting enzyme in the major NAD+ salvage pathway of mammals that reinforces SIRT1 activity. Fasting induced nutrient depletion also causes increased intracellular [AMP] and results in increased activation of adenosine monophosphate kinase (AMPK). AMPK acts to promote the switch to lipid-based metabolism by down-regulating glycolysis via phosphorylation of the nuclear receptor, carbohydrate responsive element binding protein (ChREBP). Small molecules are indicated by green hexagons. Proteins are indicated by ovals of various colors. Biochemical processes are indicated in purple boxes. Whole organism phenotypes are indicated in light-blue boxes. ↑ = increase and ↓ = decrease.
The molecular details of gene regulation by the mammalian circadian clock have been reviewed recently [24]. In mice, genes encoding key, rate-limiting enzymes of glycolysis, lipolysis, and cholesterol metabolism are regulated by the circadian clock in liver [19], and there is evidence that lipid metabolism in adipocytes is regulated by BMAL1 [35]. Additionally, SIRT1 activation of lipolytic pathways would shift metabolism away from carbohydrates and toward catabolism of lipid [36, 37]. The combined effects of prolonged fasting, SIRT1 activity, inactive/night-time mode circadian clock, and a switch to lipolysis would ultimately poise the animal for entry into torpor (Figure 3).
The nuclear receptors REV-ERBα, a negative regulator of BMAL1:CLOCK, and RORα, a positive regulator of BMAL1:CLOCK, might also play a role in regulating mammalian torpor [38–41]. Supporting the model in Figure 3, the expression of REV-ERBα in the heart of Siberian hamsters, an animal that exhibits daily torpor, was higher in torpid than in normothermic animals [42]. Carey and coworkers also observed that the liver concentration of cholesterol sulfate, a ligand of RORα [43], was low during torpor and rose slightly during interbout arousals (IBAs; brief, periodic arousals from hibernation torpor) and in summer active animals [44]. Activation of BMAL1:CLOCK by RORα might be necessary for IBAs to occur. These results suggest that REV-ERBα and RORα function to integrate cellular nutrient status, the circadian clock and metabolism (reviewed in [41]). Further investigation of these nuclear receptors in active and torpid animals will possibly reveal more about the regulation and maintenance of torpor.
In addition to SIRT1 activation in response to fasting (reviewed in [29, 45]), SIRT1 activity might also blunt oxidative stress encountered during torpor or arousal from torpor by its direct activation of hypoxia-inducible factor-2 alpha (HIF-2α) and the resulting expression of HIF-2α target genes [46] (Figure 3). It is not currently understood how animals enter and exit from torpor with no apparent damage due to reduced oxygen utilization during torpor, or due to rapid reperfusion of tissues during arousal. A recent study of Arctic ground squirrels found no evidence of oxidative stress in brain cortex or liver tissues of torpid or cold-adapted squirrels [47]. In the same study, brown adipose tissue showed limited oxidative stress in association with arousal from torpor. It has been hypothesized that antioxidant defense mechanisms are set in place prior to torpor induction [47]. SIRT1 activation of hypoxia inducible factors might provide a mechanism for preparation of these defenses.
FGF21 induces lipolysis in response to long-term fasting and promotes torpor
In mouse hepatocytes, SIRT1 interacts directly with peroxisome proliferator activated receptor α (PPARα) to activate genes involved in lipolysis [48]. Overexpression of SIRT1 activates PPARα targets including the gene encoding fibroblast growth factor 21 (FGF21) [48]. FGF21 is a hormone that promotes lipolysis and predisposes 24-hour fasted mice toward entry into torpor [49–51]. In mice, torpor is a natural response to starvation conditions [1], and circulating FGF21 levels rise after a 12-hour fast in mice and after 7 days of fasting in humans [49, 50, 52].
FGF21’s promotion of torpor, activation of lipolysis, reduction of serum glucose, and increase in serum ketones shown in mice resembles the torpor phenotype of natural hibernators. Interestingly, in FGF21 transgenic mice, liver mRNA levels for pancreatic triacylglycerol lipase (PTL), PTL-related protein 2, and the PTL cofactor colipase were increased. These same lipolytic genes were also induced in wild type mice when PPAR agonists were administered [50]. PTL is a common gastrointestinal enzyme that is expressed in non-pancreatic locations such as heart and white adipose tissue in hibernating thirteen-lined ground squirrels [20, 53]. We have shown that both human and ground squirrel PTL hydrolyze triacylglycerols at temperatures as low as 0 °C [54]. Lipolysis at low body temperatures is important for the survival of mammals that rely on stored lipids throughout the hibernation season.
It was recently shown that FGF21 overexpression in the mouse liver stimulates transcription of peroxisome proliferator-activated receptor gamma (PPARγ) coactivator-1-α (PGC1-α) [55]. Induction of PGC1-α results in regulation of essential genes for lipolysis, the tricarboxylic acid cycle, oxidative phosphorylation and gluconeogenesis. Fasted FGF21 knock-out mice have reduced PGC1-α expression and impaired lipolysis [55]. However, the ability of FGF21 knock-out mice to enter torpor has not been reported. PGC1-α is also transcriptionally activated in liver by SIRT1 (reviewed in [45]). These results in mice suggest a mechanism of torpor induction that connects SIRT1, PPARs, and FGF21 that deserves further investigation.
Adenosine monophosphate-activated protein kinase (AMPK)
Proteins that sense and respond to cellular nutrient status are also likely to play an important role in the induction of torpor. Such sensing mechanisms ensure that preparedness for entry into torpor is coordinated with environmental conditions. Adenosine monophosphate-activated protein kinase (AMPK) is highly sensitive to cellular nutrient status in terms of the [5′-AMP] to [ATP] ratio (Figure 3; reviewed in [56]). AMPK is activated when cellular [5′-AMP] rises and is antagonized by increased cellular [ATP]. Indeed, raising the level of 5′-AMP by injection induces hypothermia in mice [57], although the relevance of increasing extracellular [5′-AMP] to the activity of AMPK is unclear [58].
The traditional role of AMPK in the cell is to maintain a low [5′-AMP] to [ATP] balance by down-regulating pathways that consume ATP and up-regulating those that produce ATP. In rat hepatocytes, AMPK conserves carbohydrates by inhibiting glycolysis under conditions that mimic the fasting-induced increase of cellular [5 -AMP] [59] (Figure 3). Kawaguchi and coworkers [59] showed that AMPK down-regulates transcription of the key glycolytic enzyme, L-pyruvate kinase, through phosphorylation and inactivation of the transcription factor, carbohydrate responsive element binding protein (ChREBP) [60]. AMPK also influences energy homeostasis at the level of the whole organism and might be regulated by leptin [61, 62], adiponectin [63], and ghrelin – a known inducer of hypothermia and torpor in starved mice [64]. Taken together, these studies suggest that 5′-AMP plays a broad role in torpor induction.
AMPK might play an important role in regulating the flow of metabolic fuel during torpor. Many hibernating mammals switch to lipid catabolism and survive soley on their own lipid stores for up to six months [3]. It is therefore important that enough fat is stored as white adipose tissue and that fat stores are mobilized no faster than is necessary to maintain a low metabolic rate. Similar to hibernating mammals, survival of the metabolically suspended dauer larvae of the nematode Caenorhabditis elegans depends on frugal use of lipid reserves. Larvae carrying a null mutation in the alpha 2 subunit of AMPK are unable to inactivate adipose triacylglycerol lipase (ATGL), an enzyme that catalyzes the initial step of triacylglycerol mobilization in mammals [65], and therefore rapidly burn through their lipid stores, resulting in death before exiting dauer stage [66]. It is not known if this mechanism is conserved in mammals.
Lipids and torpor
The switch from carbohydrate to lipid catabolism is one of the most striking changes observed during torpor. This switch is essential for animals that enter long-term torpor such as hibernating mammals (reviewed in [67]) and those that estivate (a state of torpor induced by high temperatures or dehydration) (Box 1). In many hibernating species, the animal will not ingest food for several months and must strictly conserve fuel reserves. In Siberian hamsters, the switch to lipid catabolism is also seen during daily torpor where glucose is catabolized for the first few hours followed by a gradual switch to lipids as torpor continues [2]. We have shown in hibernating ground squirrels that the heart and brain preferentially use lipid-derived ketones as fuel [68]. At low body temperatures, glucose and the ketone D-β-hydroxybutyrate (BHB) are both transported across the blood-brain barrier, but BHB is used preferentially, highlighting the strict conservation of carbohydrates [68]. In the heart, both substrates are catabolized, but only carbon atoms derived from BHB enter the tricarboxylic acid (TCA) cycle [68]. This is because increased expression of PDK4 in the heart blocks conversion of pyruvate to acetyl-CoA, while catabolism of BHB results in two molecules of acetyl-CoA that directly enter the TCA cycle [21, 68, 69]. Among substances capable of inducing hypothermia and a torpor-like phenotype, the thyroid hormone derivative 3-iodothyronamine [70] also promotes lipolysis in Siberian hamsters [71]. Lipid catabolism was increased for several hours following administration of 3-iodothyronamine [71] suggesting that this naturally occurring hormone has an important role in torpor maintenance.
Adipose tissue also functions as an endocrine organ (reviewed in [72, 73]). Lipid signaling might function to sense when lipid stores are sufficient for entry into torpor and/or initiate the switch from carbohydrate to lipid-based catabolism. For example, we hypothesized that a seasonal increase in circulating lipids liberated from white adipose tissue, combined with a decrease in insulin signaling, is responsible for inducing PDK4 expression in hibernators [21]. This hypothesis is supported by a study showing that synthetic agonists and long-chain fatty acids activate PPAR and increase PDK4 expression in heart and skeletal muscle [74].
Ceramides
As stated earlier, the mechanisms that protect mammals from oxidative damage during repeated cycles of torpor and arousal are unknown. The ceramides, a group of sphingolipid derivatives (Box 2), might have an important role in this protection (reviewed in [75]). A study in C. elegans demonstrated that specific ceramide species containing fatty acid chains of 20 to 22 carbons (C20 to C22 ceramides) protect against exposure to anoxia [76]. Menuz and colleagues identified a mutation in the homolog of yeast longevity assurance gene 2 (Hyl-2) that lowers survival during anoxia [76]. The investigators determined that tolerance of anoxia required functional Hyl-2, which encodes a ceramide synthase. The Hyl-2 ceramide synthase preferentially synthesized C20 to C22 ceramides. A second ceramide synthase, Hyl-1, was unable to confer anoxia tolerance as well as Hyl-2 and preferentially produced C24 to C26 ceramides.
The specificity of the hypoxia-resistant phenotype to a specific set of ceramides suggests that the C20 to C22 ceramides function in a hypoxia protective pathway. Mammals have six isoforms of the ceramide synthetic enzyme, dihydroceramide synthase (CerS), which produce C16 to C24 ceramides. All mammalian CerS enzymes produce multiple ceramide species, with CerS1 and CerS2 producing C20 ceramides and CerS2, CerS3, and CerS4 producing C22 ceramides [77]. In a study of ceramide species composition in growing, hypoxic mouse hearts, Noureddine and coworkers observed a significant increase in concentration of dihydro-N-palmitoyl-D-erythro-sphingosine (DHC-16-Cer, the precursor of C16 ceramide), and a significant decrease in C20 ceramide concentration in the right ventricle [78]. These changes in ceramide composition occurred in association with adaptation of the right ventricle to hypoxia and suggest that shorter chain ceramides might participate in the cellular response to low oxygen stress.
The mechanistic role of ceramides in mammalian torpor is currently unknown. Recent studies show that levels of the ceramide breakdown product sphingosine are reduced in the liver of torpid compared to active thirteen-lined ground squirrels [44]. Also, sphingomyelins, of which ceramides are a component, are reduced in the cerebral cortex of hibernating Yakutian ground squirrels (Spermophilus undulatus) [79]. It is currently unclear if these results indicate a preference for specific, protective ceramides, or a general economy of lipid molecules in these hibernating species.
Future directions and medical applications
Management of food resources is a critical concern across the metazoa and the mechanisms involved are likely to be highly conserved. In fact, these same mechanisms might play a role in conserving energy by inducing torpor. In this review, we have emphasized the potential roles of recently discovered, highly conserved mechanisms of managing cellular energy status including SIRT1, the circadian clock, FGF21, and lipid molecules. The recent discovery of these mechanisms and their striking ability to reproduce many of the phenotypic changes observed in naturally-induced torpor, present a new context for investigating some of the major questions of torpor research (Box 3). The roles of Nampt, SIRT1, FGF21, ceramides and other lipid molecules have yet to be thoroughly studied in mammals that naturally enter torpor. Doing so will likely reveal how environmental cues are interpreted and translated, and how the metabolic fuel switch is achieved. Understanding these mechanisms is also likely to accelerate the development of pharmaceutical interventions for obesity, diabetes, and the metabolic syndrome.
Acknowledgments
During the writing of this review, the authors were supported by NIH grants 5 K 12 GM07 4628 (RGM) and by R15 HL08110 (MTA). We thank Steven A. Kliewer, Bethany T. Nelson, Sarah J. Timm, George Trachte and the University of Minnesota IRACDA postdoctoral fellows for comments and discussion of the draft manuscript. Carrie Bethel and James Petite helped us create the figures.
Footnotes
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.