|Home | About | Journals | Submit | Contact Us | Français|
The Roman poet Horace was among the first to recognize that when “clogged with yesterday's excess, the body drags the mind down with it.” Although considerable attention has been paid in neuroscience to the enhancement of neuronal function by wheel running and caloric restriction, far less is known about the other side of this issue. What are the consequences of unhealthy habits to central nervous system function? Prolonged exposure to excessive caloric intake impairs neuronal function, and also contributes to obesity and other risk factors for diabetes. Diabetes, a disease characterized by reduced sensitivity to glucose and insulin, is also associated with deficits in brain structure and function. In contrast, enhancement of somatic metabolism by wheel running or caloric restriction improves central neuroplasticity. Generalizing across studies reveals a relationship between global metabolic efficiency and neuroplasticity in the hippocampus, a brain region that is essential for learning and memory. The specific principles upheld by these findings are suggestive of a continuum, with global metabolic alterations fluctuating in concert with neuroplasticity in the hippocampus.
The hippocampus is essential for learning, memory, and mood regulation. Neurons in the hippocampus have among the highest energy requirements, as demonstrated by studies of regional glucose uptake (LaManna and Harik, 1985; McEwen and Reagan, 2004). Given their metabolic demands, it is not surprising that the function of hippocampal cells would be influenced by energy expenditure and intake (Cotman et al., 2007; Hillman et al., 2008; Lee et al., 2002). Studies using animal models have shown that excessive caloric intake is detrimental for neuronal plasticity (Molteni et al 2002), while caloric restriction enhances structural plasticity in this region (Fontan-Lozano et al 2007). Increasing energy expenditure through wheel running also enhances hippocampal structure and function (van Praag et al., 1999a, 1999b; Stranahan et al., 2006). Insulin resistant diabetes, a condition in which prolonged exposure to elevated glucose and insulin levels leads to impaired cellular responses to both factors, also exerts detrimental effects on the hippocampus (Jackson-Guilford et al., 2000; Zhang et al., 2007; Stranahan et al., 2008).
Findings from studies of human populations and clinical studies of subjects with diabetes support the hypothesis that excessive energy intake leads to deficits in neuroplasticity, while energy restriction or increased expenditure enhances neuronal function. Excessive caloric intake contributes directly to obesity, and obesity is a major risk factor for insulin resistant diabetes (Tatarinni 2000; Dandona et al., 2004). Individuals with diabetes are at increased risk for age-related cognitive impairment and Alzheimer's disease, both of which involve hippocampal atrophy (Korf et al., 2006; Pasquier et al., 2006). Although adverse effects of diabetes on the cerebral vasculature may contribute to cognitive impairment, brain imaging studies suggest that hippocampal and amygdalar atrophy occurs in diabetes regardless of vascular pathology (den Heijer et al., 2003). Individuals with relatively low dietary energy intakes are at reduced risk for Alzheimer's disease (Mayeux, 2003), whereas those who exercise regularly are protected against age-related cognitive impairment (Larson et al., 2006).
In this review, we explore the co-regulation of different mechanisms of plasticity in the adult hippocampus by energy intake and expenditure. Initially, we present neuroimaging data to suggest that caloric intake and insulin sensitivity are relevant to neuronal function and neuroendocrine stress systems in humans. Next, we discuss potential interactions between different plasticity mechanisms under baseline conditions, and under conditions of altered energy intake, cellular availability and expenditure. Lastly, we present the broader implications of a metabolic spectrum, with reference to the possibility of optimizing brain function and durability.
Diabetic individuals exhibit deficits in insulin production (Type 1) or impaired glucose and insulin sensitivity (Type 2). Diabetes reduces hippocampal volume, and impairs performance in behavioral tests of hippocampal function (Convit et al., 2003; Gold et al., 2007; den Heijer et al., 2003; Korf et al., 2006; but see Burns et al 2007). Obesity arises from a combination of excessive caloric intake and genetic susceptibility (Tatarinni 2000; Barsh et al., 2000). Although the mechanisms linking obesity and diabetes are still being elucidated, the association is sufficiently strong to suggest that obesity is a risk factor for diabetes (Dandona et al., 2004). Waist-to-hip ratio, an indicator of obesity, is negatively correlated with hippocampal volume in humans (Jagust et al., 2005). Functionally, obese patients exhibit reduced hippocampal activation following a liquid meal (DelParigi et al., 2004), but show increased activation following gastric distension (Wang et al., 2006). These data suggest that excessive caloric intake, combined with a genetic predisposition leading to obesity, is associated with changes in hippocampal structure and alterations in feeding-related activity in the hippocampus.
Although investigations of the relationship between exercise and brain structure in humans are scarce, preliminary studies indicate that physical activity enhances regional cerebral blood volume in the hippocampus, in parallel with improvements in declarative memory (Pereira et al., 2007). Other studies at the behavioral level have upheld this relationship (reviewed in Hillman et al., 2008). The paucity of studies surrounding the effects of caloric restriction on brain structure makes it difficult to draw conclusions about energy deficit. However, insofar as exercise and caloric restriction result in similar effects on metabolism, it is possible to suggest that caloric restriction may also promote similar changes in the hippocampus.
The spectrum of effects described above fits the concept of a continuum. A continuum is, by definition, a graded and progressive relationship. The opposing central consequences of metabolic impairment induced by diabetes and obesity, and increased metabolic efficiency induced by exercise can be considered a continuum. On one side of the balance, excessive energy intake leads to obesity, and obesity contributes to the pathogenesis of diabetes; on the other side, increased energy expenditure through exercise enhances metabolism. This sequence is recapitulated at the level of hippocampal structure and function, but the underlying mechanisms for this relationship are still being elucidated.
The hippocampus and temporal lobe structures are critical relays for learning and memory, and studies of behavior in rodent models have contributed extensively to the elucidation of specific associational mechanisms. The Morris water maze is the most well-characterized model of hippocampus-dependent learning (Morris et al., 1986). In this paradigm, the animal swims through a pool of opaque water. Because rodents find swimming aversive, they are motivated to move through the water until they locate the hidden platform beneath its surface. While they initially locate the platform by chance, with repeated trials, they begin using visual cues to orient themselves to the platform location. This task requires allocentric encoding, because it relies on external cues; additionally, because the task requires that the animal remember the previous location where it was able to escape from the water, the water maze is thought to represent episodic-like memory in rodents (Morris 2001).
Performance in the water maze is impaired by hippocampal lesions, as well as by pharmacological, genetic, and viral-vector-induced interference with excitatory neurotransmission in the hippocampus (Nakazawa et al., 2004). Conversely, enhancement of hippocampal synaptic function by environmental and genetic manipulations also improves performance in the water maze (Maher, Akaishi and Abe 2006; Tang et al., 1999; van Praag et al., 1999b). The idea of a metabolic continuum is upheld in the literature describing impaired water maze performance in diabetic animals (Li et al., 2002; Biessels et al., 1996) and in animals maintained on a high-fat, high-glucose diet (Molteni et al., 2002); the opposite end of the spectrum is best represented by the effects of wheel running and caloric restriction, with have both been shown to enhance hippocampus-dependent memory (Fontan-Lozano et al., 2007; van Praag et al., 1999a, 1999b).
The hippocampus is also essential for object recognition memory, in both rodents and primates (Squire, Wixted and Clark 2007). Rodents show an innate bias for exploration of novel objects, and the novel object preference task takes advantage of this tendency. During training, the animal is allowed to explore two objects freely, under low-stress conditions. The objects are then removed, and some time later, the animal is presented with one familiar, and one novel object. Because the animal will then prefer to explore the novel object, the index of the percent time spent exploring the novel object provides an index of memory for the familiar one.
Object recognition memory is impaired by hippocampal lesions (Clark, Zola and Squire 2000), and enhanced by many of the same manipulations that enhance performance in the water maze (Maher, Akaishi and Abe 2006; Tang et al., 1999). However, because the animals exhibit a bias for the novel object regardless of its spatial location, this represents recognition memory, rather than memory for a spatial context. Additional distinctions between this task and the water maze are that the object preference test does not rely on aversive motivation and that, unlike the water maze, the object recognition task does not involve the formation of a novel association between a location and an action. The lack of demand for associational learning may be reflected in the fact that greater damage to the hippocampus is required in order to observe impairment in recognition memory, relative to spatial memory impairment in the water maze (Broadbent, Squire and Clark 2004). Object recognition most likely involves comparisons between the strength of firing or firing rate among different groups of neurons in the hippocampus and perirhinal cortex – although the significance of neuronal firing in terms of “familiar” versus “unfamiliar” memories, or “strong” versus “weak” memories, is still a matter of debate (Sauvage et al., 2008; Squire, Wixted and Clark 2007).
The opposing effects of diabetes and wheel running are also evident in recognition memory tasks. Diabetic animals are impaired in the novel object preference task (Stranahan et al., 2008), while running enhances recognition memory (Van der Borght 2007). No studies to date have explored the consequences of excessive caloric intake on recognition memory. Because maintenance on a high-fat, high-glucose diet is likely to induce changes that are similar, but less severe than those observed in diabetic animals, it is possible that objection recognition memory might be spared with excessive energy intake.
When Timothy Bliss and Terje Lomo discovered the stable modification of synaptic strength that would come to be known as long-term potentiation (LTP; (Bliss and Lomo 1973), neuroscientists rapidly came to accept this phenomenon as a candidate cellular mechanism for learning and memory. While most research on LTP in the hippocampus has focused on the Schaffer collateral pathway between the axons from neurons in CA3 and the apical dendrites of neurons in CA1, the synapses that connect layer II stellate cells in the medial entorhinal cortex and dentate gyrus granule neurons (Figure 1) are of particular interest because this pathway represents the first excitatory input to the developing dendrites of adult-generated neurons (Schmidt-Hieber, Jonas and Bischofberger 2004). Moreover, medial perforant path LTP is enhanced by running (van Praag et al., 1999a; Farmer et al., 2004), and impaired in diabetes (Stranahan et al., 2008; Kamal et al., 1999).
Studies of excessive energy intake have focused primarily on the CA1 field, and results have been equivocal. Although there are some reports of decreased expression of proteins associated with synaptic plasticity in the hippocampus of animals fed a high-fat, high-glucose diet (Molteni etal., 2002), there was no effect of long-term excessive energy intake on CA1 LTP (Mielke et al 2006). However, it remains to be determined whether synaptic plasticity at medial perforant path synapses in the dentate gyrus might be differentially regulated by excessive caloric intake.
Paired-pulse depression is another specific characteristic of medial perforant path synapses (Asztely et al., 2000; Colino and Malenka 1993; Hanse and Gustafsson 1992). Because there is a high probability of synaptic vesicle release, the pool of available vesicles is rapidly depleted. This decreases the strength of the dendritic response when two stimuli are delivered in rapid succession. This form of short-term plasticity is not observed in the lateral perforant path, which permits the functional identification of medial perforant path synapses based on this property. Running does not influence paired-pulse depression (van Praag et al., 1999a); diabetes also did not influence this form of presynaptic plasticity (Stranahan et al., 2008). This suggests that the metabolic continuum is detectable primarily in postsynaptic mechanisms of plasticity.
Ramon y Cajal was among the first to speculate that associational learning might induce the growth of neuronal processes (Ramon y Cajal, 1893; cited in (Yuste and Bonhoeffer 2001). Apart from his prescient statements about neuronal connectivity, Donald Hebb also suggested that changes in synaptic strength might be accompanied by alterations in neuronal morphology (Hebb 1949). Total dendritic length, as well as the complexity of the dendritic tree, can be altered by experience (Fiala, Joyce, and Greenough, 1978; Pysh and Weiss, 1979; Juraska et al., 1985; Watanabe, Gould, and McEwen 1992; Eadie, Redila and Christie 2005; Kozorovitskiy et al., 2005; Stranahan, Khalil and Gould 2007). These features influence the conductance properties of the neuron, thereby regulating its function.
Structural plasticity among dendritic spines has become accepted as an additional mechanism for learning and memory (Fedulov et al., 2007; Lang et al., 2004; Zhou, Homma and Poo 2004). The spatial and temporal pattern of spine formation closely follows that of LTP; moreover, pharmacological treatments and genetic manipulations that block LTP also block spinogenesis (Engert and Bonhoeffer 1999). In the hippocampus, bidirectional synaptic plasticity occurs at the level of spine structure, as well as in synaptic function, because LTD has been correlated with loss and/or shrinkage of spines (Zhou, Homma and Poo 2004). If one accepts LTP as an underlying mechanism for learning and memory, then one must also accept a role for alterations in the number and structure of spines – because while the exact mechanisms of functional alterations in spine structure are still being elucidated, it is clear that these two processes are dependent on one another.
Contrasting effects of excessive energy intake and reduced energy availability are also evident in the regulation of dendritic spine density. While diabetes is associated with reduced dendritic spine density and neuronal atrophy (Magarinos and McEwen, 2000; Martinez-Tellez et al, 2005), increased energy expenditure following wheel running enhances neuronal structure (Stranahan et al 2007; Eadie and Christie 2006; Redila and Christie 2006). Caloric restriction also upregulates factors that support structural plasticity (Lee et al 2002), but data on the regulation of hippocampal neuron structure by caloric restriction remain equivocal. Although some studies have demonstrated reduced synaptic density and neuronal arborization following food restriction (Andrade et al., 2002), the diets used in these studies were low in protein, and feeding a smaller amount of a low-protein diet results in protein deficiency, which is known to inhibit synaptogenesis (Andrade et al., 1996). It remains to be determined whether caloric restriction with maintenance of adequate macronutrient content might exert different effects.
Spines are morphologically heterogeneous. The shape of the spine head alters local calcium dynamics, and spine head volume is linearly correlated with the area of the post-synaptic density (Nikonenko et al., 2002). There are a number of theories regarding the importance of different spine morphologies in learning and memory; some researchers have suggested that thin, filopodia-like spines, which turn over frequently, may be critical to learning (Holtmaat et al., 2005), while other researchers suggest that larger, mushroom-headed spines may be more important (Sorra and Harris 2000, Matsuo et al., 2008).
Although the relative contributions of different morphological classes of spines are still a matter of debate, it is clear that neurons exhibit experience-dependent functional adaptations that are likely to play a role in learning and memory. Again, the opposing effects of diabetes-induced hyperglycemia, and phasic reductions in glucose levels following wheel-running, are evident in the regulation of spine structure. Diabetes reduced synaptic vesicle density at mossy fiber synapses in hippocampal area CA3 (Magarinos and McEwen 2000), while wheel running induced a bias towards longer, thinner spines in hippocampal area CA1 (Stranahan et al 2007). Additional studies of the effect of excessive energy intake and caloric restriction will be required in order to disprove or uphold the existence of a metabolic spectrum at the level of synaptic structure.
Ten years before Bliss and Lomo discovered LTP, Joseph Altman made the heretical suggestion that new neurons could be added to the adult brain (Altman 1962). However, compared to LTP, the phenomenon of adult neurogenesis took far longer to be accepted. The full extent of new neuron addition in the adult brain is still being explored, but within the hippocampus, the turnover of dentate gyrus granule neurons has been accepted and proposed as an additional mechanism for learning and memory (Leuner, Gould and Shors 2006).
Resident progenitor cells in the hippocampal dentate gyrus undergo asymmetric cell division, producing a transit-amplifying cell, and another progenitor. Transit amplifying cells undergo one to two additional rounds of cell division before terminal differentiation (Kempermann et al., 2004). The estimated cell cycle duration for hippocampal progenitors in the rat is approximately 24 hours (Cameron and McKay 2001); in the mouse, cell cycle duration is approximately 14 hours (Mandyam, Harburg and Eisch 2007). Pulse-chase studies using the thymidine analog bromodeoxyuridine and double-labeling for cell-cycle stage-specific markers indicate that S-phase duration is approximately 9 hours in the rat, and 7 hours in the mouse (Cameron and McKay 2001; Hayes and Nowakowski 2002; Mandyam, Harburg and Eisch 2007).
In terms of adult neurogenesis, the consequences of running and caloric restriction are again in opposition to those of diabetes and excessive caloric intake. Running and caloric restriction enhance progenitor cell proliferation and survival, respectively, in the adult dentate gyrus (van Praag et al., 1999a, 1999b; Farmer et al 2004; Stranahan et al 2006). In contrast, diabetes and high-fat feeding inhibit the proliferation and survival of new neurons and astrocytes (Lindqvist et al., 2006; Jackson-Guilford et al., 2000; Zhang et al 2007; Stranahan et al 2008). Adult neurogenesis is a metabolically expensive process, and it is therefore not surprising that the metabolic spectrum would be observable at the level of adult neurogenesis.
New neurons migrate a short distance from the dentate subgranular zone into the dentate granule cell layer, where they extend dendrites and axons, both of which form functional contacts with medial perforant path axons and CA3 pyramidal neurons, respectively (Hastings and Gould 1999; Toni et al., 2007). The extension of processes and formation of synaptic contacts occurs during the first one to three weeks after the cell's ‘birthday;’ by three to four weeks post-mitosis, new neurons have fully formed dendrites and have integrated into the hippocampal circuitry (van Praag et al., 2002).
The synaptic properties of newly generated neurons recapitulate those in the developing brain. Initially, adult-generated neurons are excited by GABA (Esposito et al., 2005; Ge et al., 2006); after they transition to glutamatergic excitation, they retain a higher resting membrane potential, and exhibit greater excitability (Schmidt-Hieber, Jonas and Bischofberger 2004; Ge et al., 2007). These properties persist for the first four to six weeks, and there are reports that morphological plasticity is greater among newly generated granule neurons for up to four months after mitosis (Zhao et al., 2006). New neurons born in the brains of animals that run exhibit a higher density of large-headed spines (Zhao et al., 2006), but show synaptic functional properties that are indistinguishable from new neurons born in the brains of sedentary animals (Jakubs et al., 2006). No studies have yet investigated the effects of diabetes, excessive caloric intake, or caloric restriction on the function and morphology of newly generated neurons in the dentate gyrus.
The opposing effects of diabetes and running on hippocampal neuroplasticity suggest that peripheral metabolic alterations can impact the brain. Neurons are protected by the blood-brain barrier, but a variety of signals can cross the barrier, including corticosterone, insulin, and glucose, and several different growth factors and cytokines. These peripherally derived signals coordinate central adaptations, bringing central nervous system function in line with overall somatic health.
A number of candidate mechanisms have been proposed for the effects of metabolic perturbations on the brain. By comparing the literature surrounding the regulation of brain function by caloric deficit and excess, it is possible to identify common factors that may mediate these opposing effects. The neurotrophin BDNF has been recognized as a potent modulator of neuronal excitability, synaptic function, and hippocampal morphology for over ten years. Levels of BDNF are increased by exercise and caloric restriction (Lee et al., 2002; Neeper et al., 1995), and reduced in animal models of diet-induced obesity (Molteni et al., 2002). Although BDNF also regulates energy intake, pair-feeding experiments following central administration of BDNF in insulin resistant mice indicate that effects on energy metabolism are independent of effects on energy intake (Nakagawa et al., 2000). BDNF heterozygous knockout mice are obese and insulin resistant (Duan et al., 2003), but both their diabetes and brain BDNF levels can be normalized by dietary energy restriction suggesting a functional connection between central BDNF levels and peripheral energy metabolism.
Central elevations in corticosterone suppress neuronal glucose metabolism (Sapolsky 1986), while BDNF and other neurotrophic factors exert the opposite effect (Burkhalter et al., 2003; Yeo et al., 2004). This opens the possibility that one mechanism underlying the lack of any negative consequences of exposure to running-induced elevations in corticosterone may involve neurotrophic factors. We and others have observed increases in hippocampal BDNF levels with running (Cotman, Berchtold, and Christie 2007); in fact, running has previously been demonstrated to protect against the stress-induced downregulation of BDNF (Adlard and Cotman 2004). These findings suggest that runners may be protected from the deleterious effects of exposure to elevated corticosterone levels through increases in BDNF. However, the effects of running on BDNF are likely to be secondary to changes in levels of another growth factor, insulin-like growth factor 1 (IGF-1).
Insulin-like growth factor (IGF-1) is produced in a variety of organs, including the liver, muscle, and brain (Dore, Kar, and Quirion 1997). Serum IGF-1 is reduced in animal models of insulin-deficient and insulin-resistant diabetes (Kim et al., 2006; Kumari et al., 2007); in contrast, running and caloric restriction both enhance production of IGF-1 (Niedernhofer et al., 2006; Carro et al., 2001). A causal relationship for running-induced alterations in serum IGF-1 and hippocampal neurogenesis was proposed, based on experiments using peripheral blockade of IGF-1 (Trejo, Carro, and Torres-Aleman 2001). Administration of exogenous IGF-1 ameliorates cognitive deficits in diabetic animals, and enhances hippocampal learning in non-diabetic animals (Lupien, Bluhm, and Ishii 2003). Additionally, the effects of running on hippocampal BDNF levels are prevented when peripheral upregulation of IGF-1 is blocked (Chen and Russo-Neustadt 2007). Thus, it is possible that alterations in IGF-1 may be upstream of BDNF in the opposing effects of caloric excess and deficit in the hippocampus.
The studies presented in this review are meant to explore the relationship between global somatic energy metabolism, hippocampal neuroplasticity, and behavioral alterations in learning and memory. The overall trend of these findings supports a continuum, with diabetes- and stress-induced impairments on the low end, and running-induced enhancement at the higher end. This continuum was observable across a variety of structural, synaptic, behavioral, and biochemical measures. The relationship between central nervous system function and global energy metabolism has been described in terms of ‘allostatic load,’ by Bruce McEwen; ‘allostasis’ is a process in which the sum of the competing mediators of enhanced and impaired cognitive function determines the outcome (2001). While the direction of the findings described above fits well with the theory of allostasis, a number of questions remain regarding the mechanisms and brain regions that are sensitive to the opposing effects of diabetes and wheel running.
Most studies of the association between cardiovascular fitness and cognition in humans have observed improvement on tasks typically related to prefrontal and cortical structures (Hillman, Erickson, and Kramer 2008). Similar correlations have been observed for glycemic control and cognition; most of the reported deficits in the human literature involve working memory and executive function (Messier 2005). However, the bulk of the studies in animal models have focused on the hippocampus. It would be interesting and relevant to evaluate whether the opposing effects of exercise and diabetes represent a general phenomenon in the brain, or whether this relationship is specific to the hippocampus.
Running and diabetes also oppose each other in regulating disease progression across a number of neurodegenerative disorders. In humans, regular cardiovascular exercise is correlated with reduced risk of Alzheimer's dementia (Cotman, Berchtold, and Christie 2007; Hillman, Erickson, and Kramer 2008); in contrast, type 2 diabetes is correlated with increased risk (Messier 2005). In animal models, running and caloric restriction attenuate the accumulation of amyloid-beta in animal models (Adlard et al., 2005; Halagappa et al., 2007; but see Wolf et al., 2006, for an alternative viewpoint on the time-dependence of the effects of wheel running on Aβ plaque accumulation), while diet-induced insulin resistance accelerates disease progression (Ho et al., 2004). Elevations in corticosterone levels also contribute to Alzheimer's disease progression (Green et al., 2006), and it would be worthwhile to determine whether running-induced elevations in corticosterone levels might act as a pre-conditioning mechanism in animal models of neurodegenerative disease.
The regional extent and mechanisms for the opposing effects of exercise and diabetes are still being characterized. Understanding the relationship between energy metabolism and plasticity in the central nervous system has significant implications for understanding human brain development, aging, and disease. Given the explosion in juvenile-onset insulin resistance,(Bloomgarden 2004) it would be relevant assess whether the results observed in the adult brain might be exacerbated or attenuated at other points during the lifespan. Identifying and characterizing relationships between peripherally derived metabolic signals and plasticity will help to further an integrated theory of somatic and cognitive functioning.
This work was supported by a NRSA predoctoral fellowship to A.S. and by the National Institute on Aging Intramural Research Program.