|Home | About | Journals | Submit | Contact Us | Français|
Environmental enrichment, i.e., increased intellectual, social, and physical activity makes brain more resilient to subsequent neurological disease. The mechanisms for this effect remain incompletely defined, but evidence shows tumor necrosis factor-alpha (TNF-α) is involved. TNF-α, at acutely high levels, possesses the intrinsic capacity to enhance injury associated with neurological disease. Conversely, the effect of TNF-α at low-levels is nutritive over time, consistent with physiological conditioning hormesis. Evidence shows that neural activity triggers low-level pro-inflammatory signaling involving TNF-α. This low-level TNF-α signaling alters gene expression, resulting in an enhanced resilience to disease. Brain-immune signaling may become maladaptive when increased activity is chronic without sufficient periods of reduced activity necessary for nutritive adaptation. Such tonically increased activity may explain, for example, the transformation of episodic to chronic migraine with related increased susceptibility to spreading depression, the most likely underlying cause of this malady. Thus, TNF-α, whose function is to alter gene expression, and its principal cellular source, microglia, seem powerfully positioned to orchestrate hormetic immune signaling that establishes the phenotype of neurological health and disease from brain activity.
Brain is unique among organ structures in that it exists to predict (Llinás 2002), a capacity that requires constant “reprogramming.” Importantly, reprogramming increases the brain’s resistance to neurological disease. Neural reprogramming is classically evidenced by Hebbian synaptic plasticity, and extends to environmental enrichment (EE; i.e., increased intellectual, social, and physical activity), which is well-known to be protective (for review see van Praag et al. 2000; Will et al. 2004).
Reprogramming is also an inherent capacity of the immune system (Waldmann 2002; Graca and Waldmann 2006; Waldmann et al. 2008). In addition, immune system reprogramming may be a well-conserved process by which ischemic preconditioning stimuli applied to brain prompt subsequent neuroprotection (Marsh et. al. 2009).
Our focus is to illustrate how immune stimuli within brain, associated with neural activity, similarly reprogram gene function to create the phenotype of increased neurological health through interactive signaling between neurons and glia. Increasing evidence indicates that neural and immune systems interact in response to increased activity to establish the phenotype of enhanced brain health. This activity-dependent cooperative reprogramming may be driven by tumor necrosis factor-alpha (TNF-α), an innate immunity cytokine that emanates from microglia under normal circumstances (Hulse et al. 2008).
Classically, microglia, principal immune signaling cells within brain, have been recognized for their powerful destructive capacities to enhance brain injury via mechanisms that include increased expression of TNF-α (Chao and Hu 1994; Dawson et al. 1996; Barone et al. 1997; Meistrell et al. 1997; Aggarwal et al. 2001; Zou and Crews 2005). Accordingly, why would the brain use lethal signaling systems to enhance its strength? Developing evidence indicates the answer lies in the old saying “…what doesn’t kill you makes you stronger …” (Hadley 2003). This notion termed “physiological conditioning hormesis” (Calabrese et al. 2007) is increasingly recognized as a well-conserved capacity of biological systems, including brain (Mattson et al. 2002; Mattson 2008a, 2008b). Mounting evidence indicates that TNF-α and microglia can be neuroprotective (Cheng et al. 1994; Wilde et al. 2000; Hallenbeck 2002; Stoll et al. 2002; Streit 2005, 2006; Turrin and Rivest 2006; Carson et al. 2007; Hanisch and Kettenmann 2007; Sriram and O’Callaghan 2007; Hulse, et al. 2008; Salmina 2009).
Hormesis is a dose-response pattern involving low-level stimulation and high-level inhibition (Calabrese and Baldwin 2003). Hormesis is seen with neuroprotective treatments for stroke and brain trauma (Calabrese 2008) and also may play a role in the brain’s intrinsic capacity to protect itself against injury. Importantly, such irritative signaling that initiates neuroprotection requires time to develop.
We present original data and related literature to support and extend the suggestion that brain uses TNF-α-dependent signaling (Arumugam et al. 2006) from microglia to initiate subsequent adaptive changes that provide the neuroprotection consistent with physiological conditioning hormesis (Calabrese et al. 2007). Furthermore, we propose that this nutritive scenario may become maladaptive if initiating stimuli (i.e., neural activity) rise to a frequency that precludes sufficient recovery periods needed for adequate adaptive responses. We suggest the latter may exemplify the transformation of episodic to chronic migraine with related increased susceptibility to spreading depression, the most likely underlying cause for this headache disorder (Lauritzen and Kraig 2005). Similar considerations are likely to apply to other neurodegenerative disorders, including temporal lobe epilepsy.
The mechanisms of endogenous neuroprotection are commonly studied in reference to “ischemic tolerance,” a phenomenon where a lethal stimulus brought below threshold imparts neuroprotection against subsequent injury (Kirino 2002; Dirnagl et al. 2003, Dirnagl et al. 2009). As noted above, such neuroprotection most often includes the study of stimuli applied to brain. Deciphering the means by which biologically driven mechanisms mitigate brain injury will lead to improved treatment strategies that may result in fewer negative sequelae and greater physiological effects (Kraig and Kunkler 2002). The notion that treatments should follow naturally evolved pathways has considerable merit. Our interest in defining the mechanisms of EE-based neuroprotection extends beyond traditional ischemic tolerance studies. This emphasis has two distinct advantages. First, defining mechanisms will provide important insights for new acute treatment strategies derived from stimuli with no potential for confounding effects associated with lethal stimuli. Second, detailed knowledge of underlying EE mechanisms will establish the physiological bases (and therefore rationale) for neurological preventative health strategies designed to lessen the impact of neurological diseases. The latter would empower patients and healthcare providers with evidence-based strategies to improve their health before the onset of brain disease.
Considerable clinical evidence indicates that moderately increased exercise lessens the severity of neurodegenerative disease. For example, Chen and coworkers (2005) noted that higher levels of physical activity may lower the risk of Parkinson’s disease development in men prone to develop this disorder. Curiously, women in the study showed no benefit. Rovio and coworkers (2005) also noted that physical activity at midlife correlated with decreased risk of dementia and development of Alzheimer’s disease in later life. Van Gelder and coworkers reported a similar positive impact of physical activity begun later in life (2004).
Importantly to our thesis, Rovio and colleagues (2005) observed the greatest reduction in disease development, due to increased physical activity, in individuals with a genetic marker for Alzheimer’s disease, APOE4. This finding begins to illustrate the concepts put forward by Silverman (2004) in a recent article entitled, “Rethinking Genetic Determinism.” Silverman notes inadequacies of the traditional assumptions that “…the gene is deterministic in gene expression and can therefore predict disease propensities.” He suggests instead that more attention should be given to how “environment” alters genetic propensity to influence phenotype. While genetic determinism may be valid in specific circumstances (e.g., lethal genetic changes), our research focus is greatly influenced by the therapeutic potential of Silverman’s suggestion, which we believe EE-based neuroprotection powerfully exemplifies.
A second caveat of EE-based neuroprotection is whether physical activity alone is sufficient to induce the protection. For example, exercise is sufficient to retard development of cardiovascular disease (Thompson et al. 2003). However, physical activity alone may not be adequate for neuroprotection from EE (Sturman et al. 2005). Instead, learning (van Praag et al. 2000) may be the key component by which EE (i.e., increased intellectual, physical, and social activity) triggers neuroprotection.
Evidence from experimental animal studies supports the neuroprotective capacity of EE (Will et al. 2004). EE initiated after brain trauma (Wagner et al. 2002) or stroke (Dahlqvist et al. 2004) enhances cognitive function involving the hippocampus when compared to non-enriched animals. EE also slows cognitive decline in ageing rodents (Kempermann et al. 2002).
EE is also effective in enhancing brain function after injury to developing brain. Early enrichment in perinatal rats a day after hypoxia/ischemia results in partial recovery of memory deficits (Pereira et al. 2008). However, this effect was only seen in females. A common problem in studies of EE is so-called dominant male effect that stresses other enriched male animals (see below). This prompted many to include only females in enrichment studies. Alternatively, when EE starts two weeks after neonatal ischemia, male rats show significant improvement compared to non-enriched counterparts (Pereira et al. 2007). Furthermore, postnatal EE can ameliorate behavioral effects of fetal alcohol syndrome (Hannigan et al. 2007) and the consequences of other neurological disease models (Nithianantharajah and Hannan 2006). Accordingly, EE and experiential interventions may reduce or ameliorate the neuronal and non-neuronal abnormalities caused by injury to the developing brain (Dong and Greenough 2004). Remarkably, the positive effects of EE are seen when EE is initiated long after perinatal injury (i.e., after weaning).
EE is also effective when initiated before the onset of neurological disease. For example, EE reduces temporal lobe seizures in adult rats and reduces cognitive deficits compared to non-enriched counterparts (Young et al. 1999).
Our studies using in vivo and in vitro models of excitotoxic hippocampal injury confirm and extend the evidence that learning from EE is neuroprotective. First, we did not observe any clinical evidence of a dominant male effect with our enrichment cage conditions (Figure 1), perhaps because the cage was sufficiently large (Marashi et al. 2003). This conclusion is further supported by the fact that we did not find any outlying serum corticosterone values within animal groups (see below). Second, 28-day exposure to EE triggered a significant increase in hippocampal (i.e., contextual) learning compared to non-enriched counterparts (Figure 2). Third, this EE paradigm significantly reduced CA3 area hippocampal pyramidal neuron loss from kainic acid-induced temporal lobe epilepsy, and significantly improved animal survival compared to non-enriched counterparts (Figure 3).
Initially, serum corticosterone levels were measured to probe for potential stress reaction from exposure to EE. Our results indicate what may be an important signaling pattern associated with increased learning from EE. We, like others (Kempermann et al. 2002), found that morning corticosterone levels from EE animals were modestly elevated compared to non-enriched counterparts (Figure 4). However, we also found that evening levels were significantly higher in EE animals compared to non-enriched mice with nearly two-fold increase in circadian rhythm change (Figure 4). A similar effect of EE on circadian corticosterone was reported by others studying growing pigs (de Jong et al. 2000).
Phasically elevated corticosterone can initially enhance hippocampal function, whereas chronically elevated corticosterone inhibits hippocampal function (de Kloet et al. 1999; Zoladz and Diamond 2008). Diamond and coworkers showed that corticosterone enhances primed burst electrical potentiation of hippocampal pyramidal neurons (Diamond et al. 1992; Kim and Diamond 2002). Moderate and intermittent corticosterone levels enhance electrical activity, while extreme and chronically elevated levels (e.g., consistent with clinical depression) inhibit electrical responses. These effects are consistent with physiological hormesis.
While corticosterone is a peripheral, and potentially systems-wide, modulator of EE-based neuroprotection, our research focuses to intrinsic brain signaling. There, TNF-α, and the cellular source of this innate cytokine, microglia, are likely key mediators of EE-based neuroprotection. This focus stems from data indicating that learning from EE initiates neuroprotective and hormetic TNF-α signaling from microglia.
First, EE (i.e., increased neural activity) triggered a significant increase in expression of TNF-α within hippocampus, the site of excitotoxic neuroprotection (Figure 5). Prior evidence shows that brain TNF-α rises in response to treadmill activity (Ding et al. 2005).
Second, TNF-α modulates synaptic activity. Within minutes of exposure, TNF-α enhances synaptic efficacy by increasing the exocytosis of AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazoleproprionic acid) receptors in cultured neurons (Beattie et al. 2002) as well as the endocytosis of inhibitory GABAA (γ-amino butyric acid type A) receptors (Stellwagen et al. 2005). TNF-α also triggers increased expression of AMPA receptors as a synaptic scaling response to activity deprivation by tetrodotoxin (Stellwagen and Malenka 2006). Increased AMPA receptor expression is seen with long-term potentiation (LTP) (Park et al. 2004), a well-accepted cellular model of learning, and EE (Naka et al. 2005). While TNF-α may not be necessary for LTP induction (Albensi and Mattson 2000) and may retard it (Cunningham et al. 1996), TNF-α does play a role in related homeostatic plasticity changes (Kaneko et al. 2008). This non-Hebbian capacity of neurons (and neural tissue) to modulate their overall sensitivity to activity may play a critical role in maintaining network stability during development and in the mature, learning brain (Turrigiano et al. 1998; Goldberg et al. 2002).
Recent work from Wheeler and coworkers (2009) shows that acute exposure to TNF-α also leads to prompt increased neuronal excitability that involves NMDA (N-methyl-d-aspartate) receptor expression. Importantly, these workers studied in detail the molecular events that lead to fusion of NMDA containing vesicles with the plasma membrane. Their work demonstrates that TNF-α-induced activation of neutral sphingomyelinase-2 and increased expression of ceramide are essential for increased surface membrane expression of NMDA receptors and increased synaptic efficacy. This is an important step forward for defining the signaling syntax by which TNF-α from glia can acutely alter neuronal excitability.
Third, neuroprotection from increased neuronal activity involves TNF-α. Reduced synaptic efficacy from long-term depression has no impact on CA1 area excitotoxic injury in hippocampal slices. On the other hand, increased synaptic efficacy from LTP triggers a significant reduction in injury that requires TNF-α (Kraig et al. 2006) (Figure 6). A similar neuroprotective effect requiring TNF-α is seen with spreading depression (Kraig et al. 2005). Spreading depression is non-injurious and has many cellular and molecular features similar to learning and memory (Kraig and Kunkler 2002; Kunkler et al. 2005), including transient elevation of TNF-α (Kunkler et al. 2004). Finally, Lambertsen and coworkers (2009) used traditional knockout animals for TNF-α (and its cognate receptors) to show that stroke injury and behavioral deficits are increased in the absence of TNF-α signaling. Potentially confounding adaptive effects may obscure observations using traditional knockout/in animals (Gao et al. 1999); however, here TNF-α related adaptive changes may account for the observations.
Fourth, monomeric immunoglobin G (IgG) may help interlink neuronal activity to TNF-α production (Hulse et al. 2008) (Figure 7). EE activates microglia to a pro-inflammatory state (Ziv et al. 2006). While all brain cell types plus adjacent endothelial cells are capable of releasing TNF-α, microglia are recognized as the likely predominant source of this pro-inflammatory cytokine, especially in response to disease (Hopkins and Rothwell, 1995; Allan and Rothwell, 2001). Microglia are also likely the predominant, if not sole, source of TNF-α under physiological conditions including EE (Hulse et al. 2008). Monomeric IgG is commonly regarded as a quiescent signaling molecule poised only to initiate potentially injurious inflammatory reactions in response to disease via immune complex formation and associated phagocytosis and TNF-α production. Instead, IgG signaling within non-injured brain results in neuroprotection when it follows a physiological conditioning hormesis dose-response pattern and takes time to develop. This IgG-mediated neuroprotection involves enhanced microglial recycling endocytosis and TNF-α production—changes that are quickly evident after IgG exposure. Importantly, minocycline, known for its anti-inflammatory and neuroprotective effects when administered after the onset of brain disease, abrogates IgG-mediated neuroprotection and related microglial enhanced recycling endocytosis and TNF-α production. Furthermore, E-prostanoid receptor subtype 2 activation by prostaglandin E2, which is released by hippocampal pyramidal neurons proportionate to their electrical activity, amplifies IgG-mediated effects on microglia. This suggests paracrine mediators released from neurons, due to their increased activity, can affect microglial TNF-α release via amplification of recycling endocytosis. These results confirm and extend evidence indicating that microglial pro-inflammatory activation over time is nutritive. Furthermore, they begin to illustrate specific immune signaling mechanisms by which neurons and microglia interact to enhance brain function.
Neuroprotection from EE requires time to develop. This suggests that EE, like other preconditioning stimuli (Barone et al. 1998; Nishio et al. 2000), requires new protein synthesis. The predominant effect of TNF-α is to alter gene expression (Abbas and Lichtman 2003). Thus, TNF-α is well positioned to orchestrate the activity-dependent structural and functional adaptation necessary for EE-based neuroprotection. However, the precise nature of these adaptive processes and their relation to TNF-α and microglial activation remain undefined. Nonetheless, evidence from physical activity, dietary restriction, and ageing research provide provocative clues.
Radak and colleagues (2008) review evidence that moderate physical activity improves general health by triggering adaptive responses to reactive oxygen species (ROS). They note that physical activity extremes (i.e., inactivity and excessive activity) lead to deterioration in health, consistent with the basic tenet of hormesis. Exercise must be sufficiently high to initiate “stress” so that adaptive responses will occur. Importantly, exercise needs to be intermittent with sufficient periods of rest to allow for the development of adaptive responses associated with improved muscle (and organismal) health. We suggest this requirement not only applies to brain and EE (Figure 8), but may also be involved in the maladaptive effects of excessively increased brain activity (see below). The authors point to literature (Gomez-Cabrera et al. 2005) indicating that ROS inhibition via exogenous scavengers abrogates the positive effects of physical exercise, a result now extended to humans (Ristow et al. 2009).
Dietary restriction may also improve organismal health. Schulz and coworkers (2007) show that glucose restriction increases the lifespan of Caenorhabditis elegans. This caloric restriction paradigm triggers increased ROS formation, and importantly, increased ROS scavenging via heightened catalase activity. Similar to physical activity, exogenous antioxidants and vitamins that scavenge ROS may reduce the worm lifespan initially increased by dietary restriction.
The Mattson laboratory emphasizes that these peripheral preconditioning strategies for exercise and dietary restriction hormesis involving ROS are likely to apply to the ageing brain (Arumugam et al. 2006). This suggestion is supported by data showing that NMDA receptor activation enhances neuronal antioxidant defenses (Papadia et al. 2008).
Modestly increased ROS production and the pronounced impact of intrinsic antioxidants on neurological preventative health may be linked to upstream microglial activation that includes production of TNF-α and oxidants. Several lines of evidence support this suggestion. First, TNF-α stimulates glucose uptake into astrocytes (Yu et al. 1995; Véga et al. 2002), which in turn provide high-energy substrates to neurons (Barros and Deitmer 2009). Second, learning increases brain oxidative metabolism (Heiss et al. 1992). Third, microglia are activated by synaptic activity (Ziv et al. 2006; Hulse et al. 2008) and activated microglia release ROS (Block et al. 2007; Innamorato et al. 2009), which can prompt increased generation of neural tissue antioxidants. Fourth, oxidant stress plus antioxidant responses can modulate synthesis of proteins involved in activity-dependent neuroprotection, most likely via the “translational switch,” eukaryotic translational initiation factor 2α (Costa-Mattiolo et al. 2009; Tan et al. 2009).
Fifth, Radak and colleagues (2008) emphasize that not only must exercise be sufficiently stressful, but adequate rest periods are also essential to improve health. While sufficiently increased brain activity (i.e., EE) is necessary to initiate improved health, sleep may provide the essential rest period required for adaptive changes. Vyazovskiy and colleagues (2008) note that while the wake state is associated with potentiation of synaptic function, sleep reverses this to depotentiation. This scenario is schematized in Figure 8. Importantly, although synaptic potentiation falls with sleep, protein synthesis rises (Ramm and Smith, 1990). Seemingly paradoxical to our overall thesis involving neural activity and TNF-α, hippocampal TNF-α mRNA rises with sleep (Cearley et al. 2003). However, TNF-α changes associated with LTP in the wake state may be localized to only the synaptic regions activated by learning and thus changes would not be detectable in whole brain regions used for the circadian mRNA measurements. Finally, brain catalase activity (and possibly the activities of other antioxidants which collectively can influence protein synthesis), shows phasic behavior (Sani et al. 2006).
In summary, mounting evidence indicates that physiologically increased neural activity begets increased neuronal excitability via mechanisms that include TNF-α from microglia (Figure 9). When this increased neural activity is sufficiently phasic, brain becomes more resilient. In contrast, heightened neural activity and associated glial activation, without adequate periods for adaptation, may leave brain more susceptible to disease (see below).
Considerations of brain physiological conditioning hormesis have largely focused on the nourishing and restorative capacities of this signaling response pattern. However, examination of mechanisms by which increased brain activity can become maladaptive may help explain select brain diseases. For example, spreading depression is a non injurious and transient perturbation of brain that, like EE, triggers microglial activation (Caggiano and Kraig 1996), increased TNF-α (Kunkler et al. 2004) (Figure 10), and neuroprotection that depends on TNF-α (Kraig et al. 2005). Spreading depression is also the most likely underlying cause for episodic migraine aura and pain (Moskowitz et al. 1993; Kunkler and Kraig 2003; Lauritzen and Kraig 2005).
Reduced inhibitory synaptic function plays a key role in the mechanisms of spreading depression. Our work, and that from the Somjen laboratory (for review see Kunkler et al. 2005), show that a synchronous reduction in synaptic inhibitory drive occurring over a sufficiently large brain volume initiates spreading depression (Figure 10). In addition, as spreading depression subsides, inhibitory synaptic function is last to recover. We speculate that as spreading depression (i.e., migraine) becomes more frequent, the lack of sufficient periods for compensatory adaptive recovery leads to increased brain excitability from reduced inhibition, which may account for the transformation of episodic to chronic migraine (Figure 8).
Chronic migraine is a prevalent healthcare burden (Lipton et al. 2004) whose pathogenesis remains incompletely defined. Evidence suggests that chronic migraine occurs with central sensitization of sensory pathways that involves increased expression of pro-inflammatory mediators and alterations in the periaqueductal gray (Aurora 2009). However, increased migraine frequency also correlates with the transformation of episodic migraine to chronic migraine (Silberstein and Olesen 2005), suggesting the frequency of spreading depression may be an initiating cause of chronic migraine pain. Accordingly, central sensitization of sensory pathways and alterations of the periaqueductal gray may be “downstream” signaling phenomena of chronic migraine while recurrent spreading depression is the “upstream” neural signaling causal change.
Spreading depression initiates pro-inflammatory changes. Astrocytes (Kraig et al. 1991) and microglia (Caggiano and Kraig 1996) show reactive changes for weeks after spreading depression. Importantly, spreading depression also triggers increased TNF-α production (Kunkler et al. 2004) and microglial activation (Caggiano and Kraig 1996) which would include increased production of tissue oxidants. Together these factors, as well as release of other factors such as brain derived neurotrophic factor from microglia (Coull et al. 2005), can reduce neural network inhibitory synaptic drive, which would promote recurrent spreading depression (i.e., chronic migraine). Evidence supports this maladaptive potential of physiological conditioning hormesis. Repetitive spreading depression triggers a selective suppression of inhibitory function (Kruger et al. 1996) and cortical hyperexcitability in migraineurs stems from reduced inhibition (Palmer et al. 2000). Accordingly, microglial activation from increased brain activity that occurs without sufficient time to permit adequate adaptation, may initiate maladaptive consequences to brain.
Physiological conditioning hormesis characterizes the signaling response patterns by which increased brain activity from EE generates enhanced resilience to brain disease. Two basic tenets of hormesis require that initiating irritative, but not injurious, stimuli are sufficiently robust and that adequate periods of recovery from stressful initiating stimuli occur to allow adequate time for adaptive processes to take effect. Microglia, because of their activation from neural activity and their associated production of TNF-α and oxidant irritants, are well-positioned to orchestrate hormetic immune signaling that establishes the phenotype of neurological health and disease from brain activity.
This work was supported by grants from the National Institute of Neurological Disorders and Stroke (NS-19108), the National Institute of Child Health and Human Disorders (5 PO1 HD 09402), the Migraine Research Foundation and the White Foundation. BC-P was supported by a Campbell McConnell Fellowship from Cornell College. Ms. Marcia P. Kraig assisted in the preparation and maintenance of culture systems and in the mouse enrichment studies. We thank Yelena Grinberg for reading and commenting on a final version of the manuscript.