Roles of Nuclear Factor κB in Neuronal Survival and Plasticity


  • Mark P. Mattson,

    1. Sanders-Brown Research Center on Aging and Department of Anatomy and Neurobiology, University of Kentucky, Lexington, Kentucky, U.S.A.
    Search for more papers by this author
  • Carsten Culmsee,

    1. Sanders-Brown Research Center on Aging and Department of Anatomy and Neurobiology, University of Kentucky, Lexington, Kentucky, U.S.A.
    Search for more papers by this author
  • ZaiFang Yu,

    1. Sanders-Brown Research Center on Aging and Department of Anatomy and Neurobiology, University of Kentucky, Lexington, Kentucky, U.S.A.
    Search for more papers by this author
  • Simonetta Camandola

    1. Sanders-Brown Research Center on Aging and Department of Anatomy and Neurobiology, University of Kentucky, Lexington, Kentucky, U.S.A.
    Search for more papers by this author

  • International Society for Neurochemistry

  • Abbreviations used: AMPA, 2-amino-3-hydroxy-5-methylisoxazole-4-propionate; APP, amyloid precursor protein; IAP, inhibitor of apoptosis; IKK, IκB kinase; LTP, long-term potentiation; MnSOD, manganese superoxide dismutase; NF-κB, nuclear factor κB; NMDA, N-methyl-D-aspartate; sAPPα, α-secretase-derived form of secreted β-amyloid precursor protein; TNF, tumor necrosis factor-α.

Address correspondence and reprint requests to Dr. M. P. Mattson at his present address: Laboratory of Neurosciences, National Institute on Aging, 5600 Nathan Shock Drive, Baltimore, MD 21224, U.S.A. E-mail:


Abstract: The transcription factor nuclear factor κB (NF-κB) is moving to the forefront of the fields of apoptosis and neuronal plasticity because of recent findings showing that activation of NF-κB prevents neuronal apoptosis in various cell culture and in vivo models and because NF-κB is activated in association with synaptic plasticity. Activation of NF-κB was first shown to mediate antiapoptotic actions of tumor necrosis factor in cultured neurons and was subsequently shown to prevent death of various nonneuronal cells. NF-κB is activated by several cytokines and neurotrophic factors and in response to various cell stressors. Oxidative stress and elevation of intracellular calcium levels are particularly important inducers of NF-κB activation. Activation of NF-κB can interrupt apoptotic biochemical cascades at relatively early steps, before mitochondrial dysfunction and oxyradical production. Gene targets for NF-κB that may mediate its anti-apoptotic actions include the antioxidant enzyme manganese superoxide dismutase, members of the inhibitor of apoptosis family of proteins, and the calcium-binding protein calbindin D28k. NF-κB is activated by synaptic activity and may play important roles in the process of learning and memory. The available data identify NF-κB as an important regulator of evolutionarily conserved biochemical and molecular cascades designed to prevent cell death and promote neuronal plasticity. Because NF-κB may play roles in a range of neurological disorders that involve neuronal degeneration and/or perturbed synaptic function, pharmacological and genetic manipulations of NF-κB signaling are being developed that may prove valuable in treating disorders ranging from Alzheimer’s disease to schizophrenia.


Nuclear factor κB (NF-κB) was originally identified in B lymphocytes, where it stimulates transcription of the immunoglobulin κ light chain (Sen and Baltimore, 1986; Baeuerle and Henkel, 1994; Baeuerle and Baltimore, 1996). Various genes have since been shown to be responsive to NF-κB, including those for cytokines, cell surface receptors, and antioxidant enzymes (Table 1). In its inactive form NF-κB consists of a three-subunit complex consisting of two (prototypical) subunits of 50 kDa (p50) and 65 kDa (p65; RelA), and an inhibitory subunit called IκB (IκBα or IκBβ). However, depending on cell type, developmental stage, and environmental factors, cells may express other NF-κB DNA-binding subunits, e.g., p52, c-Rel, and RelB, and IκBs, e.g., Bcl-3 and IκBε (Table 2). Additional κB-binding proteins are only partially characterized but appear to exhibit cell type specificity, with a recent example being the neuron-specific κB-binding factor NKBF (Moerman et al., 1999). The NF-κB complex is located in the cytosol and is activated when IκB is induced to dissociate from the complex (Fig. 1). The p50-p65 dimer then translocates to the nucleus and binds to 5′ regulatory elements consisting of the decameric sequence 5′-GGGPuNNPy-PyCC-3′ (Pu, purine; Py, pyrimidine; N, any base) in genes responsive to NF-κB.

Table 1. Examples of NF-κB responsive genes.Thumbnail image of
Table 2. Examples of NF-κB proteins, IκB proteins, and their cellular expression.Thumbnail image of
Figure 1.

Mechanisms of regulation of NF-κB activity. The inactive form of NF-κB exists in the cytosol as a three-subunit complex, with the prototypical components being p65 and p50 (transcription factor dimer) and IκBα (inhibitory subunit). Signals that activate NF-κB do so by inducing phosphorylation of IκBα (which may target IκBα for degradation in the proteosome), which, in turn, causes its dissociation from the p65-p50 dimer. The p65-p50 dimer then translocates to the nucleus and binds to consensus κB sequences in the enhancer region of κB-responsive genes. Antiapoptotic genes induced by NF-κB include those encoding MnSOD, calbindin, IAPs, and some Bcl-2 family members. Many different stimuli result in the activation of NF-κB, including activation of the TNF receptor (p55), which occurs when TNF binds and induces receptor trimerization. A dimer of the protein TRADD (TNF receptor-associated death domain) then associates with a “death domain” in the cytoplasmic portion of the receptor, which, in turn, causes association of TRAF2 (TNF receptor-associated protein 2) and IAP with TRADD. A kinase cascade is thus activated that results in phosphorylation of IκB. Increases in levels of intracellular calcium and reactive oxygen species, e.g., H2O2, resulting from glutamate receptor activation, for example, are important inducers of NF-κB in cells exposed to various apoptotic insults. JUN kinase, c-Jun N-terminal kinase; MAP kinase, mitogenactivated protein kinase; SM, sphingomyelin.



FIG. 1.

The molecular events that lead to the activation of NF-κB are beginning to be elucidated. Phosphorylation and ubiquitination of IκB are necessary for dissociation of IκB from the transcription factor dimer; IκB is then degraded in the proteosome (Woronicz et al., 1997). IκB is phosphorylated by a protein kinase complex [IκB kinase (IKK)] that consists of two catalytic subunits (IKKα and IKKβ) and a regulatory subunit called IKKγ (Rothwarf et al., 1998; Yamaoka et al., 1998). IKK phosphorylates IκBα at Ser32 and Ser36, and phosphorylation at both residues appears to be necessary for activation of NF-κB (Verma and Stevenson, 1997; Zandiet al., 1998). Studies of IKK phosphorylation and activation in tumor necrosis factor-α (TNF)-stimulated tumor cells have provided evidence that IKK is activated by phosphorylation of IKKβ (Delhase et al., 1999). Attempts at identifying kinases that act upstream of IKK have revealed two candidates, NF-κB-inducing kinase [NIK (Malinin et al., 1997)] and mitogen-activated protein kinase kinase kinase-1 [MEKK1 (F.S. Lee et al., 1998)]. Autophosphorylation of IKKβ on several carboxyl-terminal serine residues decreases IKK activity and may thereby provide a feedback mechanism to prevent prolonged activation of NF-κB.

Gene knockout studies in mice are revealing the roles of specific NF-κB subunits and regulatory proteins in the development and functions of different organ systems. Targeted disruption of the p65 gene results in embryonic lethality due to liver failure (Beg et al., 1995). In contrast, disruption of the p50 gene results in mice that appear to develop normally; however, such p50—/— mice exhibit altered lymphocyte responses when challenged with lipopolysaccharide and infectious agents (Sha et al., 1995; Snapper et al., 1996) and altered responses of neurons to excitotoxic and ischemic brain injury (Yu et al., 1999; also see below). The latter findings demonstrate that p65 is essential for normal development and that κB-binding proteins other than p50 may heterodimerize with p65 and thereby substitute for p50 during development (Baeuerle and Baltimore, 1996). Recent studies in which mice lacking either IKKα or IKKβ were generated have revealed intriguing roles for these NF-κB-regulating kinases. Mice lacking IKKα die soon after birth and exhibit impaired limb outgrowth and hyperproliferation of epidermal cells but no apparent abnormalities in TNF-induced NF-κB activation (Takeda et al., 1999). Mice lacking IKKβ die as embryos as the result of liver damage resulting from increased apoptosis; IKKβ-deficient mice can be rescued by inactivation of the gene for the p55 TNF receptor (Li et al., 1999).

The gene targets of NF-κB are many and, in general, reflect the important roles of NF-κB in cellular responses to tissue injury (Table 1). Genes encoding several different cytokines, including TNF and interleukin-6, are induced by NF-κB (Tomita et al., 1998). These cytokines play central roles in coordinating the immune response to tissue infection and injury (Baeuerle and Henkel, 1994). The IκBα gene is induced by NF-κB and so mediates a negative feedback pathway to suppress prolonged activation of NF-κB (Le Bail et al., 1993). One of the first genes shown to be responsive to NF-κB was manganese superoxide dismutase (MnSOD), a mitochondrial antioxidant enzyme that protects cells against apoptosis (Wong et al., 1989; Mattson et al., 1997). Other genes induced by NF-κB include the cell adhesion molecules such as ICAM-1 (Collins et al., 1995; S. J. Lee et al., 1998), the inducible form of nitric oxide synthase (Taylor et al., 1998), tissue transglutaminase (Mirza et al., 1997), and the Bcl-2 homologue Bfl-1/A1 (Zong et al., 1999). Another example of a gene responsive to NF-κB that is known to be induced in response to brain injury is that encoding the astrocytic glial fibrillary acidic protein (Krohn et al., 1999).


NF-κB is activated by various intercellular signals, including cytokines, neurotrophic factors, and neuro-transmitters (Table 3). Activation of glutamate receptors and membrane depolarization lead to activation of NF-κB in hippocampal pyramidal neurons and cerebellar granule neurons in cell culture (Guerrini et al., 1995; Kaltschmidt et al., 1995). Several different cytokine receptors that are expressed in neurons are linked to NF-κB activation, with the p55 TNF receptor being most intensively studied to date (Cheng et al., 1994; Barger et al., 1995; Bruce et al., 1996). The low-affinity nerve growth factor receptor, which is a member of the TNF receptor family, is also linked to NF-κB activation (Carter et al., 1996).

Table 3.. Examples of stimuli that activate NF-κB in the CNS.Thumbnail image of


Although this proposal is largely unexplored, it is likely that NF-κB plays important roles in development of the nervous system. As evidence, the Drosophila NF-κB homologue “dorsal” plays a pivotal role in the establishment of dorsoventral polarity in the developing embryo (Hoch and Jackle, 1993). Levels of NF-κB activity change during development of the nervous system, with levels peaking in the rat cerebellum during the early postnatal period (when synaptogenesis is ongoing) and decreasing thereafter (Kaltschmidt et al., 1995). Inducible p50 dimers and p65/cRel dimers were shown to be present in brains of young rats but not in adults (Bakalkin et al., 1993). These kinds of data suggest roles for NF-κB in synaptogenesis and natural cell death during development of the nervous system.

NF-κB activity is increased in neurons and glial cells both in acute neurodegenerative conditions and in chronic age-related neurodegenerative disorders. Several studies have documented increased levels of NF-κB activation in brain tissues in rodent models of stroke or cardiac arrest. For example, Clemens et al. (1997) provided evidence that NF-κB is activated in CA1 hippocampal neurons following transient global forebrain ischemia in rats. A delayed increase in NF-κB activation, occurring several days following focal ischemia-reperfusion, appears to occur in association with reactive glial cells (Gabriel et al., 1999). NF-κB activity is rapidly increased in hippocampal neurons within 4-16 h following kainate-induced seizures (Rong and Baudry, 1996). The rapid increase in NF-κB activity in neurons is followed by a delayed and sustained increase in NF-κB activity in glial cells (Matsuoka et al., 1999). In a rat model of traumatic brain injury levels of NF-κB DNA-binding activity were increased in the traumatized cortex within 1 day of injury (Yang et al., 1995).

Studies of postmortem brain tissue from Alzheimer’s disease patients have revealed increased NF-κB activity in association with the neurodegenerative process. For example, Kaltschmidt et al. (1997) found increased p65 immunoreactivity in neurons and astrocytes in the immediate vicinity of amyloid plaques in brain sections from Alzheimer’s patients consistent with NF-κB activation in those cells. The latter study and other studies (see, e.g. Guo et al., 1998a) have shown that amyloid β-peptide can induce NF-κB activation in cultured neurons, suggesting a role for amyloid in NF-κB activation in Alzheimer’s disease. More recently Lukiw and Bazan (1998) reported a strong correlation between increased NF-κB activity and cyclooxygenase-2 gene transcription in superior temporal lobe gyrus of Alzheimer’s patients. Immunohistochemical studies suggest that levels of NF-κB activity are increased in cholinergic neurons in the basal forebrain of Alzheimer’s patients (Boissiere et al., 1997). Analyses of brain sections from Parkinson’s patients revealed a 70-fold increase in the proportion of dopaminergic neurons (in the substantia nigra) with p65 immunoreactivity in their nuclei compared with age-matched controls (Hunot et al., 1997). Additional data suggest increased activation of NF-κB in astrocytes, but not in motor neurons, in spinal cords of amyotrophic lateral sclerosis patients (Migheli et al., 1997).

The kinds of data just described clearly show that NF-κB is activated under conditions where neurons are degenerating and glial cells are reacting to the neurodegenerative process. However, such studies of postmortem tissue do not allow firm conclusions to be reached as to the roles of NF-κB in the injury and/or recovery process. The remaining sections of this article describe data from experiments that directly address the roles of NF-κB in neuronal survival and plasticity.


For many years it was thought that because NF-κB is activated in cells under conditions where many cells die, NF-κB plays a role in killing the cells. For example, NF-κB activity is greatly increased in neurons following seizure activity (Prasad et al., 1994; Grilli et al., 1996a; Rong and Baudry, 1996; Matsuoka et al., 1999) and ischemia (Salminen et al., 1995; Clemens et al., 1997; Carroll et al., 1998; Zhang et al., 1998). However, the interpretation that NF-κB plays a role in killing neurons proved to be flawed because it was based on “guilt-by-association” rather than on concrete data addressing cause—effect relationships. We had been studying cytoprotective signaling pathways that are activated following injury to the nervous system and had developed the concept of “programmed cell life” to describe the involvement of growth factor signaling pathways in preventing cell death (for review, see Mattson and Furukawa, 1996; Mattson and Lindvall, 1997). During the course of our work we found that TNF can protect cultured embryonic rat hippocampal neurons against excitotoxic and apoptotic insults, including exposure to glutamate, glucose deprivation, and amyloid β-peptide (Cheng et al., 1994; Barger et al., 1995). The latter cell culture studies demonstrated that TNF exerts a direct neuroprotective action on neurons. We found that TNF induced NF-κB activation in hippocampal neurons, that the antiapoptotic action of TNF was mimicked by treatment with IκB antisense oligonucleotides (Barger et al., 1995), and that treatment of the neurons with κB decoy DNA (which selectively blocks NF-κB activity) abolished the cytoprotective effect of TNF (Mattson et al., 1997). Moreover, treatment of hippocampal cultures with C2-ceramide, which activates NF-κB, protected neurons against apoptosis induced by oxidative and excitotoxic insults (Goodman and Mattson, 1996). Tamatani et al. (1999) recently confirmed and extended our findings by showing that expression of an unresponsive mutant form of IκB in cultured hippocampal neurons increases their vulnerability to hypoxia-induced cell death. On the other hand, high concentrations of ceramide can induce apoptosis (Hartfield et al., 1998; Irie and Hirabayashi, 1998), although it is unclear whether levels of intracellular ceramide reach such levels under pathophysiological conditions.

Additional studies support an antiapoptotic function of NF-κB (Table 4). Taglialatela et al. (1997) reported that agents that inhibit NF-κB [pyrrolidinedithiocarbamate and the proteosome inhibitor N-benzyloxycarbonyl-Ile-Glu(O-tert-butyl)-Ala-leucinal] induce apoptosis of PC12 cells and that such apoptosis is not prevented by treatment with nerve growth factor. Studies of sympathetic neurons that used a peptide inhibitor of NF-κB (SN50) provided evidence that nerve growth factor-dependent activation of NF-κB prevents apoptosis of cultured sympathetic neurons (Maggirwar et al., 1998). Activation of NF-κB in response to nerve growth factor appears to be mediated by the low-affinity neurotrophin receptor (Carter et al., 1996). Another example of the neuroprotective function of NF-κB comes from studies of the mechanism of action of the α-secretase-derived form of secreted β-amyloid precursor protein (sAPPα), which is known to exhibit potent excitoprotective and antiapoptotic actions in CNS neurons (for review, see Mattson, 1997). Treatment of cultured hippocampal neurons with sAPPα results in NF-κB activation that is correlated with increased resistance to metabolic and excitotoxic insults (Barger and Mattson, 1996). Moreover, by activating NF-κB, sAPPα can counteract the proapoptotic actions of mutations in the presenilin-1 gene (mutations that are causally linked to early-onset inherited forms of Alzheimer’s disease) (Guo et al., 1998a).

Table 4.. Chronology of evidence that activation of NF-κB can prevent apoptosis.Thumbnail image of


Although the cell culture data just described provided direct evidence that activation of NF-κB in cultured neurons increases their resistance to excitotoxicity and apoptosis, the roles of TNF and NF-κB in vivo remained unclear. To determine whether endogenous TNF serves a neuroprotective function following brain injury in vivo, we generated mice lacking one or both TNF receptors. Two different TNF receptors, p55 and p75, are expressed in neurons throughout the brain (Kinouchi et al., 1991; Cheng et al., 1994). Activation of the p55 receptor results in recruitment of TNF receptor-associated proteins and a phosphorylation cascade that results in activation of the transcription factor NF-κB (Fig. 1) (Smith et al., 1994); the signaling pathway(s) of p75 is not yet established. We found that the vulnerability of hippocampal neurons to seizure-induced injury and the vulnerability of cortical and striatal neurons to focal ischemic injury were significantly increased in mice lacking the p55 receptor or both TNF receptors (Bruce et al., 1996; Gary et al., 1998). More recently we have found that brain lesion size and blood—brain barrier breach following traumatic brain injury are increased in mice lacking TNF receptors compared with wild-type mice (Sullivan et al., 1999). NF-κB activation and MnSOD expression were delayed and reduced following traumatic brain injury in the TNF receptor knockout mice. Collectively, these findings provide direct evidence that endogenous, injury-induced TNF serves a neuroprotective function.

The role of NF-κB activation in neuronal death following brain injury was directly examined using two different approaches in a seizure model of excitotoxic death of hippocampal neurons. The first approach was to administer κB decoy DNA via intraventricular infusion before administration of the excitotoxin kainate. Examination of brain sections from mice that had been administered fluorescently tagged κB decoy DNA showed that the DNA was taken up by cells throughout the ipsilateral hippocampus, including pyramidal neurons in CA1 and CA3 (Fig. 2). (Yu et al., 1999). The extent of neuronal death in regions CA1 and CA3 of hippocampus was significantly increased in mice administered κB decoy DNA compared with mice administered scrambled control DNA. The second approach was to examine NF-κB activation and neuronal loss following kainate administration in mice lacking the p50 subunit of NF-κB (Yu et al., 1999). Gel-shift analyses showed that p50 is required for the vast majority of κB DNA-binding activity in hippocampus. Mice lacking p50 exhibited increased damage to hippocampal pyramidal neurons and attenuated increases in levels of TNFα and MnSOD following kainate administration. Cultured hippocampal neurons from p50-deficient mice exhibited enhanced elevations of intracellular calcium levels following exposure to glutamate and were more vulnerable to excitotoxicity than were neurons from wild-type mice (Fig. 3). These data suggest that the p50 subunit of NF-κB plays a major role in protecting neurons against excitotoxic cell death. Further evidence supporting a role for NF-κB in promoting neuronal survival in vivo comes from a study in which a proteosome inhibitor was infused into the lateral ventricles of adult rats; the authors provided evidence that this proteosome inhibitor suppressed NF-κB activity and induced DNA fragmentation indicative of apoptosis in neurons in several brain regions (Taglialatela et al., 1998).

Figure 2.

Evidence that NF-κB activation following epileptic seizures serves a neuroprotective function. Upper panel: A micrograph shows κB decoy DNA-associated fluorescence in CA1 neurons in the hippocampus of a mouse 2 h following intraventricular administration of 60 μg of fluorescein-labeled κB decoy DNA. Middle and lower panels: Nissl-stained sections of hippocampus 24 h following intrahippocampal administration of kainate in a control mouse (middle panel) and a mouse that had been administered κB decoy DNA before kainate administration (lower panel). Note increased degeneration of hippocampal neurons in the mouse that had been administered κB decoy DNA.

Figure 3.

Increased vulnerability to excitotoxicity and perturbed calcium homeostasis in hippocampal neurons from mice lacking the p50 subunit of NF-κB. A: Hippocampal cultures from p50+/+, p50+/-, and p50-/- mice were exposed for 24 h to the indicated concentrations of glutamate, and neuronal survival was quantified. *p < 0.05, **p < 0.01 compared with corresponding value for p50+/+ mice. B: The intracellular free Ca2+ concentration ([Ca2+]i]) was measured by imaging of the dye fura-2 in neurons from p50+/+, p50+/-, and p50-/- mice at the indicated time points before and following exposure to 5 μM glutamate. Data are mean values of 10-15 neurons. Modified from Yu et al. (1999).

FIG. 2.

FIG. 3.

More recent studies have provided evidence that, as previously demonstrated in neurons, NF-κB also serves an antiapoptotic role in nonneuronal cells. Agents that prevent activation of NF-κB and microinjection of IκBα or a p65 antibody induced apoptosis in lymphoma cells, whereas overexpression of p65 prevented apoptosis (Wuet al., 1996). Whereas wild-type fibroblasts and macrophages were not killed by TNF, p65-deficient cells were killed (Beg and Baltimore, 1996). Reintroduction of p65 into the p65-deficient cells restored their resistance to TNFα-induced apoptosis. Inhibition of NF-κB nuclear translocation by overexpression of the superrepressor IκBα enhanced cell killing by chemotherapeutic agents in fibrosarcoma cells, and overexpression of the p50 and p65 subunits of NF-κB conferred resistance to killing (Wang et al., 1996). Collectively, these data suggest a widespread antiapoptotic role for NF-κB.

A well-known means of preventing apoptosis in various cell types is to treat them with the protein synthesis inhibitor cycloheximide. The ability of cycloheximide to prevent cell death has been interpreted to indicate that the cell death process requires synthesis of “killer” proteins. It is interesting that levels of cycloheximide that cause only a small impairment of protein synthesis can also prevent apoptosis by a mechanism involving induction of neuroprotective gene products, including the antiapoptotic protein Bcl-2 and the antioxidant enzyme MnSOD (Furukawa et al., 1997). Treatment of cultures with Bcl-2 antisense oligonucleotide reduced the neuro-protective action of cyloheximide, suggesting that increased Bcl-2 expression was mechanistically involved in the antiapoptotic actions of cycloheximide. Gel-shift analysis showed that cycloheximide is a very strong inducer of NF-κB DNA-binding activity (Fig. 4). These findings suggest potential cooperative antiapoptotic mechanisms involving NF-κB and Bcl-2 because data indicate that NF-κB may play a role in the antiapoptotic actions of Bcl-2 (Ivanov et al., 1995).

Figure 4.

The antiapoptotic compound cycloheximide (CHX) activates NF-κB, whereas the proapoptotic agent 4-hydroxy-2,3-nonenal (HNE) inhibits NF-κB, in cultured hippocampal neurons. Cultured rat hippocampal neurons were incubated in Locke’s buffer and exposed to vehicle (Control), 10 μM HNE, or 100 μg/ml CHX for 6 h. Cell extracts (10 μg of protein) were analyzed for κB-binding factors by electrophoretic mobility shift assay. Note that levels of κB DNA-binding activity were decreased in cultures exposed to HNE and increased in cultures exposed to CHX.

FIG. 4.

Oxidative stress, particularly increases in hydrogen peroxide levels, induces NF-κB activation in many cell types, including neurons (for review, see Schreck et al., 1991). Such stress-induced activation of NF-κB appears to represent a cytoprotective response because treatment of neurons with κB decoy DNA increases their vulnerability to oxidative stress-induced apoptosis (Mattson et al., 1997). It is interesting that the lipid peroxidation product 4-hydroxynonenal, which may mediate oxidative stress-induced neuronal apoptosis in many paradigms (Kruman et al., 1997), strongly inhibits NF-κB activation (Fig. 4). The mechanisms underlying positive and negative modulation of NF-κB activity in neurons subjected to oxidative stress have not yet been determined. However, there are data suggesting that direct oxidative modification of IκB may result in its dissociation from NF-κB (Piette et al., 1997). On the other hand, direct covalent modification of NF-κB subunits or upstream kinases by 4-hydroxynonenal may inhibit NF-κB activation (S.C. and M.P.M., unpublished data). Thus, differential modulation of NF-κB activity by various oxyradicals and products of lipid peroxidation may have a major influence on whether or not a neuron lives or dies in a particular pathophysiological condition.


Increased oxidative stress and perturbed cellular calcium homeostasis appear to be convergence points in the neuronal cell death pathway activated in many different physiological and pathological settings, including developmental cell death and acute and chronic neurodegenerative disorders (for review, see Mattson, 1996). Several different κB-responsive genes have been identified that likely play important roles in increasing cellular resistance to apoptosis. Wong et al. (1989) observed that tumor cells resistant to TNF-induced apoptosis exhibit increased expression of the antioxidant enzyme MnSOD following TNF treatment, whereas cells vulnerable to TNF-induced apoptosis do not increase their levels of MnSOD. Recent studies of the effects of TNF on different types of brain cells further support a widespread and important role for MnSOD production in the cytoprotective action of TNF (Bruce-Keller et al., 1998). Increased expression of MnSOD contributes to the neuroprotective action of TNF because κB decoy DNA suppresses induction of MnSOD and attenuates the neuroprotective effect of TNF (Mattson et al., 1997). Moreover, overexpression of MnSOD in cultured PC12 cells protects them against apoptosis induced by nitric oxide donors and amyloid β-peptide (Keller et al., 1998). The latter study also showed that neurons in the brains of transgenic mice overexpressing MnSOD exhibit increased resistance to ischemic damage following middle cerebral artery occlusion (a model of stroke). Overexpression of MnSOD in cultured fibrosarcoma cells protects them against apoptosis induced by alkaline conditions (Majima et al., 1998). The latter study showed that MnSOD stabilizes levels of intracellular reactive oxygen species and calcium and restores mitochondrial transmembrane potential, indicating that mitochondria are a primary target for alkaline-induced cell death. It was recently reported that TNF induces increases in levels of the antiapoptotic proteins Bcl-2 and Bcl-x in cultured hippocampal neurons and that this increase is blocked by expression of a dominantnegative IκB (Tamatani et al., 1999), which may be another mechanism whereby NF-κB activation suppresses oxidative stress.

In addition to enhancement of antioxidant defenses, NF-κB may protect cells by modulating the expression of proteins involved in the regulation of cellular calcium homeostasis. Levels of the calcium-binding protein calbindin D28k are increased in embryonic hippocampal neurons (Cheng et al., 1994) and in astrocytes (Mattson et al., 1995) following treatment with TNF. Stable overexpression of calbindin D28k in PC12 cells increased their resistance to apoptosis induced by a calcium ionophore, trophic factor withdrawal, and an apoptosis-enhancing mutation in presenilin-1 (Guo et al., 1998b; Wernyj et al., 1999). Calbindin D28k appears to interrupt the apoptotic cascade in the very early stages before increases in superoxide production and mitochondrial membrane depolarization. Overactivation of glutamate receptors can induce neuronal apoptosis by a mechanism involving calcium overload and oxyradical production (Ankarcrona et al., 1995; Chan et al., 1999a). Recent electrophysiological studies of cultured hippocampal neurons suggest that NF-κB can modulate neuronal excitability by altering whole-cell currents through voltagedependent calcium channels and ionotropic glutamate receptor channels (Furukawa and Mattson, 1998). Treatment of the cultured neurons with TNF for 24-48 h resulted in increased voltage-dependent calcium channel current density and decreased glutamate-induced currents. C2-ceramide mimicked the effect of TNF on voltage-dependent calcium channel current and glutamate-induced currents, whereas κB decoy DNA treatment blocked the effects of TNF (Furukawa and Mattson, 1998). Modulation of expression of proteins involved in regulation of cellular calcium homeostasis may therefore be an important mechanisms whereby NF-κB protects cells against excitotoxicity and apoptosis.

Inhibitors of apoptosis (IAPs) were originally discovered in baculoviruses, where they play a role in suppressing cell death response to viral infection. IAP proteins contain a conserved baculoviral IAP repeat domain that contains repeats of cysteine and histidine residues, and mammalian IAPs also contain a caspase recruitment domain (CARD) (for review, see Deveraux and Reed, 1999). Different IAP family members are differentially expressed among tissues, with XIAP and NAIP appearing to play particularly important roles in the nervous system (Roy et al., 1995; Liston et al., 1996). Indeed, the latter studies showed that the NAIP gene is deleted in patients with spinal muscular atrophy. NAIP may play an important role in protecting neurons against ischemic brain injury because its levels increase in neurons resistant to ischemic brain injury and overexpression of NAIP increases resistance of neurons to ischemic injury in vivo (Xu et al., 1997).

Overexpression of several different IAPs, including XIAP, human IAP1, and NAIP, in cultured cerebellar granule neurons resulted in resistance of those cells to apoptosis induced by reducing the potassium concentration in the culture medium (Simons et al., 1999). The mechanisms whereby IAPs suppress apoptosis are beginning to be revealed. Several different IAPs bind to caspases and inhibit their activity, with inhibition of caspase-3 being particularly relevant to neuronal apoptosis (Deveraux et al., 1997). It is interesting that interactions of IAPs with NF-κB signaling appear to play important roles in increasing cellular resistance to apoptosis. Thus, several IAPs, e.g., c-IAP2, have been shown to act upstream of NF-κB in the TNF signaling pathway, and IAP genes are induced by NF-κB (Chu et al., 1997), suggesting a role for IAPs in amplifying NF-κB signaling.

Additional mechanisms whereby NF-κB may suppress apoptosis are being elucidated. For example, activation of NF-κB in tumor cells was shown to block activation of caspase-8 by a mechanism involving induction of expression of TNF receptor-associated factors 1 and 2 and the inhibitor of apoptosis proteins c-IAP1 and c-IAP2 (Wang et al., 1998). IEX-1L is an NF-κB target that appears to play an important role in prevention of apoptosis in tumor cells (Wu et al., 1998). It is interesting that well-known modulators of apoptosis are being linked to NF-κB signaling. For example, it was recently shown that levels of NF-κB activity are increased in cultured myocytes that overexpress the antiapoptotic gene Bcl-2 by a mechanism involving enhanced degradation of IκB (de Moissac et al., 1998). Activation of NF-κB may also indirectly protect neurons by inducing the expression of neurotrophic factors and cytokines. NF-κB induces production of TNF by glial cells and neurons. NF-κB is an important mediator of TNF production in vivo because the normal increase in TNF levels in neurons and glial cells following severe epileptic seizures is attenuated in mice lacking p50 (Yu et al., 1999). In addition to inducing TNFα expression, NF-κB stimulates expression of interleukin-6, which, in turn, induces interleukin-1β expression (Collins et al., 1995).

The extensive literature on the involvement of NF-κB in regulating cytokine cascades suggests that although NF-κB activation can prevent apoptosis of the cell in which it is activated, it can indirectly lead to apoptosis of other cells by promoting production of cytotoxic agents such as nitric oxide. Indeed, the gene encoding inducible nitric oxide synthase contains κB binding sites in its enhancer region (Taylor et al., 1998). Cytokine-mediated activation of microglia may explain the ability of inhibitors of NF-κB to protect against cell damage in certain experimental paradigms (Qin et al., 1998). Another example of cell death-enhancing actions of NF-κB comes from a recent study in which it was shown that NF-κB is an essential mediator of Fas ligand expression in response to DNA-damaging agents and that NF-κB thereby contributes to stress-induced apoptosis (Kasibhatla et al., 1998).

We recently discovered that regulation of NF-κB activity is perturbed in neural cells expressing a mutated form of presenilin-1 linked to early-onset inherited Alzheimer’s disease (Guo et al., 1998a). Presenilin-1 mutations had been shown to increase neuronal vulnerability to apoptosis by a mechanism involving perturbed calcium regulation in the endoplasmic reticulum and increased levels of oxidative stress (Guo et al., 1998b). Cells expressing presenilin-1 mutations exhibited an aberrant pattern of NF-κB activation following exposure to apoptotic insults characterized by enhanced early activation with a subsequent prolonged depression of NF-κB activity that was associated with sensitivity to apoptosis (Guo et al., 1998a). Related studies have elucidated a feed-forward neuroprotective pathway involving NF-κB that is activated by the secreted form of amyloid precursor protein (APP), which is known to exhibit potent antiexcitotoxic and antiapoptotic effects in neurons (Furukawa et al., 1996). The secreted form of APP activates NF-κB in cultured hippocampal neurons by a cyclic GMP-mediated mechanism (Barger and Mattson, 1996). Pretreatment of cells expressing mutant presenilin-1 with the secreted form of APP restored the normal pattern of activation of NF-κB following exposure to apoptotic insults and prevented cell death (Guo et al., 1998a). It is interesting that Grilli et al., (1996b) reported that the enhancer region 5′ to the gene encoding βAPP contains κB-binding sites and that NF-κB induces transcription of APP. This suggests a scenario in which NF-κB activation in cells under stress leads to APP (and sAPPα) production, resulting in further activation of NF-κB.

Whereas activation of NF-κB appears to be a powerful antiapoptotic mechanism, inhibition of NF-κB may represent an important mechanism for inducing or enhancing apoptosis. For example, caspase-3 was shown to cleave IκBα, and evidence was provided that the cleavage transforms IκBα into a constitutive inhibitor of NF-κB, thereby promoting apoptosis (Barkett et al., 1997). Glucocorticoids are a potent trigger for apoptosis of T lymphocytes, wherein these steroids suppress NF-κB activity by a mechanism involving induction of IκBβ expression (Ramdas and Harmon, 1998). The contribution of inhibition of NF-κB to the cell death-promoting actions of glucocorticoids in hippocampal neurons (Stein-Behrens et al., 1994; Smith-Swintosky et al., 1996) remains to be established. Recent studies suggest a possible role for suppression of NF-κB activation in the proapoptotic action of Par-4, a novel leucine zipper- and death domain-containing protein recently linked to apoptosis of tumor cells and of neurons in several different neurodegenerative disorders (Sells et al., 1997; Guo et al., 1998c). Par-4 has been shown to inhibit atypical isoforms of protein kinase C, which appear to be positioned upstream of NF-κB activation in antiapoptotic cascades (Berra et al., 1997). We recently found that activation of NF-κB is suppressed and cell death is enhanced in PC12 cells overexpressing Par-4, whereas NF-κB activity is maintained and cell death is prevented in PC12 cells overexpressing a dominant-negative form of Par-4 (Fig. 5) (S.C. and M.P.M., unpublished data).

Figure 5.

The proapoptotic protein Par-4 suppresses NF-κB activity in PC12 cells. PC12 cells stably transfected with empty vector alone (VA), full-length Par-4 (PAR-4), or the dominant-negative Par-4 leucine zipper domain (dnPAR-4) (see Guo et al., 1998c) were deprived of trophic support for the indicated intervals. Control cultures were not deprived of trophic support. Nuclear extracts were prepared, and the NF-κB DNA-binding activity was assessed by electrophoretic mobility shift assay. Note that NF-κB activity is markedly reduced in the cells that overexpress full-length Par-4.

FIG. 5.


NF-κB is present in synaptic terminals located at great distances from the neuronal cell body, and NF-κB activation can occur locally in such synapses (Kaltschmidt et al., 1993; Meberg et al., 1996). Because NF-κB is located in synaptic terminals and is responsive to increases in levels of intracellular calcium and reactive oxygen species, it is poised to mediate long-term changes in synaptic structure and function. Activation of glutamate receptors and voltage-dependent calcium channels results in NF-κB activation in neurons (Guerrini et al., 1995). Glutamate and membrane depolarization induce NF-κB activation in cultured cerebellar granule neurons, and such activation can be blocked by treatment of the neurons with an antioxidant, indicating that free radicals are involved in glutamate receptor- and depolarization-induced activation of NF-κB (Kaltschmidt et al., 1995). The high basal (constitutive) level of NF-κB activity in neurons may be the result of ongoing activity in neuronal circuits (Korner et al., 1989; Kaltschmidt et al., 1993, 1994, 1995). Indeed, the heterogeneity in constitutive NF-κB activity among neuronal populations suggests a relationship to cell function (Kaltschmidt et al., 1994). NF-κB is activated in association with long-term potentiation (LTP) of synaptic transmission, a process believed to be central to learning and memory (Meberg et al., 1996). It is interesting that NF-κB is also activated in response to low-frequency stimulation, in contrast to other transcription factors, e.g., c-Fos and NGFI-A, that are induced by high-frequency (LTP-inducing) stimulation parameters but not by low-frequency stimulation (Cole et al., 1989; Demmer et al., 1993; Worley et al., 1993). Excessive neuronal activity, as occurs during epileptic seizures, results in marked NF-κB activation in cortex and hippocampus of adult rats (Unlap and Jope, 1995). It is believed that synaptic activity during seizures and during LTP may share many features (Baudry, 1986). NF-κB may therefore represent a signaling pathway designed to protect neurons against the potentially damaging effects of excessive neuronal activity.

If NF-κB serves as a regulator of neuronal plasticity, then in addition to being activated by activity in neuronal circuits, NF-κB should modulate the expression of genes that encode proteins involved in regulating neuronal excitability and plasticity, and blockade of NF-κB should alter synaptic plasticity. Data from whole-cell perforated patch clamp recordings in cultured rat hippocampal neurons suggest that activation of NF-κB results in a decrease in levels of N-methyl-D-aspartate (NMDA) and 2-amino-3-hydroxy-5-methylisoxazole-4-propionate (AMPA)/kainate currents, possibly the result of modulation of expression of specific receptor subunits (Furukawa and Mattson, 1998). The latter study also demonstrated that NF-κB activation leads to increased voltage-dependent calcium currents. Modulation of expression of glutamate receptor channel proteins and voltage-dependent calcium channel proteins may therefore be one mechanism whereby NF-κB could modulate synaptic plasticity. Analyses of synaptic plasticity in hippocampal slices from mice lacking TNF receptors and in slices incubated in the presence of κB decoy DNA suggest an important role for NF-κB in the process of long-term depression of synaptic transmission (Albensi and Mattson, 1999). Stimulation of Schaffer collateral axons at a frequency of 1 Hz induced long-term depression in region CA1 in slices from wild-type mice but not in slices from TNF receptor-deficient mice (Fig. 6). Pretreatment of slices from wild-type mice with κB decoy DNA prevented induction of long-term depression and significantly reduced the amplitude of LTP, providing direct evidence for a role for NF-κB in synaptic plasticity.

Figure 6.

Long-term depression of synaptic transmission is impaired at CA3-CA1 synapses in mice lacking TNF receptors. Representative population spike (PS) responses were recorded in the CA1 on stimulation of the Schaffer collateral pathway in hippocampal slices from (A) a wild-type (WT) mouse and (B) a TNF receptor knockout (TNFRKO) mouse. Following baseline recordings, slices were stimulated at 1 Hz. Note depression of the PS amplitude following 1 Hz stimulation in the slice from the WT mouse but not in the slice from the TNFRKO mouse. Modified from Albensi and Mattson (1999).

FIG. 6.

Recent studies have shown that apoptotic death signals can be activated in synaptic terminals (Mattson et al., 1998a,b). The latter studies showed that glutamate and amyloid β-peptide can induce loss of membrane phospholipid asymmetry, caspase activation, and mitochondrial alterations characteristic of apoptosis in synaptosomes and dendrites of cultured hippocampal neurons. More recent findings suggest that apoptotic cascades may modify signaling associated with synaptic plasticity. Following exposure of cultured hippocampal neurons to apoptotic stimuli, AMPA receptor subunits (but not NMDA receptor subunits) are cleaved via a caspasemediated mechanism (Chan et al., 1999b). Whole-cell patch clamp and calcium-imaging analyses indicate that neuronal responsitivity to glutamate is reduced following exposure of neurons to apoptotic stimuli and that caspase inhibitors prevent the reduced responsivity. The ability of NF-κB to modulate gene expression in ways that suppress apoptotic cascades suggests an additional mechanism whereby NF-κB might modulate synaptic plasticity.


Because of its responsiveness to neuronal activity and to injury to the nervous system, NF-κB is likely to play important roles in an array of physiological and pathological processes. The emerging data described above support the latter statement. Concerning neuronal plasticity, the roles of NF-κB are poorly understood, and there are many intriguing questions waiting to be addressed. For example, can activation of NF-κB in a specific synapse lead to modification of the form or function of just that synapse, and if so, how? What are the synaptic signal transduction cascades that activate or suppress NF-κB, and how are they configured? Several proteins present in synaptic terminals are encoded by genes responsive to NF-κB, e.g., glutamate receptor subunits and inducible nitric oxide synthase, and changes in expression of such proteins could mediate long-term effects of NF-κB on synaptic function. Further work is required to elucidate the full array of plasticity-related genes responsive to NF-κB. Concerning neuronal cell death, considerable evidence indicates that NF-κB induces the expression of antiapoptotic gene products. Future studies should therefore be aimed at identifying signaling mechanisms upstream of NF-κB activation, on the one hand, and identifying gene targets of NF-κB that confer resistance to apoptosis, on the other hand. Functional interrelationships between proapoptotic factors such as Par-4, Bax, and caspases and antiapoptotic factors such as Bcl-2, cyclic GMP, and protein kinase Cζ are being elucidated and will surely reveal novel mechanisms controlling the neuronal life—death continuum.

Although the kinds of data presented above demonstrate that activation of NF-κB can prevent neuronal apoptosis in many cases, there are settings wherein activation of NF-κB may be detrimental to neurons. For example, Qin et al. (1998) reported that SN50, a peptide inhibitor of NF-κB, can reduce the extent of quinolinic acid-induced degeneration of rat striatal neurons. In a subsequent study these investigators reported that SN50 also suppressed quinolinic acid-induced expression of c-Myc and p53 (Qin et al., 1999). Because p53 has been linked to some cases of neuronal apoptosis (Johnson et al., 1999), induction of p53 expression might promote neuronal cell death. Recent studies suggest a detrimental role of NF-κB in ischemic injury to neurons in rodent models of stroke. For example, Schneider et al. (1999) reported that cerebral infarct size was reduced in mice lacking the p50 subunit of NF-κB relative to wild-type mice. We have obtained similar results in our laboratory (Z.Y. and M.P.M., unpublished data). Because excitotoxic injury to hippocampal neurons is increased in mice lacking p50 (Yu et al., 1999), the available data suggest the role of NF-κB is condition-specific and complex.

Because NF-κB plays such important roles in cellular response to injury in the nervous system, NF-κB signaling systems are very important targets for therapeutic intervention in an array of acute and chronic neurological disorders. Development of pharmacological and genetic manipulations that activate or suppress NF-κB activity will likely have broad applicability to treatment of various human diseases. For example, agents that inhibit NF-κB activity can promote apoptosis of tumor cells and are therefore potentially useful in various cancers. Conversely, agents that activate NF-κB may prove effective in preventing neuron death in the many neurodegenerative disorders that involve apoptosis. Because NF-κB modulates synaptic plasticity, abnormalities in NF-κB signaling might also contribute to neurological disorders such as schizophrenia (Weinberger, 1999) that do not involve overt neuronal degeneration. Although much remains to be done, it is clear that a better understanding of the physiological and pathophysiological roles of NF-κB will lead to novel approaches to preventing and treating various neurological disorders.


We thank S. W. Barger, A. J. Bruce-Keller, K. Furukawa, D. Gary, and Q. Guo for their valuable contributions to original research in this laboratory. This work was supported by grants from the National Institute on Aging and National Institute of Neurological Disorders and Stroke, National Institutes of Health, and the Kentucky Spinal Cord and Head Injury Research Trust.