Neuronal NF-κB Signaling: From Adaptive Stress Responses to Pathological Signaling

Evidence from neuronal models helps explain why NF-κB has been assigned apparently opposing roles in AD. The discrepancy becomes less paradoxical when the nature of the signaling response is considered. Brief and tightly regulated NF-κB activation can participate in neuronal stress adaptation, trophic signaling, antioxidant defense, calcium homeostasis, and non-amyloidogenic processing of APP. By contrast, persistent or disease-conditioned signaling, particularly through IKKβ and RelA/p65, is more consistently associated with BACE1 expression, Aβ production, mitochondrial dysfunction, impaired protein degradation, inflammatory signaling, and regulated cell death. Thus, the biological consequence of neuronal NF-κB activity depends on the initiating stimulus, duration and intensity of activation, NF-κB dimer, promoter and cofactor environment, and downstream genes engaged.

Some of the strongest evidence for a protective neuronal response comes from low-intensity Aβ exposure. Kaltschmidt et al. showed that relatively low concentrations of Aβ1–40 or Aβ25–35, with maximal responses around 0.1 µM, activated NF-κB in primary neurons through reactive oxygen intermediates; activated p65 was also observed in neurons near early plaques in human AD tissue (Kaltschmidt et al., 1997b). A subsequent preconditioning study demonstrated the functional significance of this response: neurons initially exposed to 0.1 µM Aβ1–40 became more resistant to a later 10 µM challenge, whereas blocking NF-κB with dominant-negative IκBα abolished the protection and increased susceptibility to Aβ toxicity (Kaltschmidt et al., 1999b). TNF-α similarly produced an NF-κB-dependent protective response. These findings support a model in which mild stress can recruit an adaptive neuronal NF-κB program, but they do not imply that Aβ itself is beneficial or that prolonged NF-κB activation would protect the chronically diseased brain.

Other trophic pathways point in the same direction. In differentiated PC12 cells carrying the familial AD-associated presenilin−1 L286V mutation, soluble APPα (sAPPα) protected against oxidative stress, calcium dysregulation, mitochondrial dysfunction, and apoptosis, whereas blockade of NF-κB-dependent transcription abolished this protection (Guo et al., 1998b). Similarly, erythropoietin reduced Aβ-induced apoptotic injury in primary hippocampal neurons, but the effect was substantially weakened by p65 knockdown or inhibition of its nuclear translocation (Chong et al., 2005). These studies show that RelA/p65 is not intrinsically neurotoxic; when transiently recruited downstream of trophic signals, it can participate in neuronal survival.

A more clearly defined protective transcriptional program involves c-Rel. In Aβ-challenged cortical neurons and neuronal cell lines, stimulation of metabotropic glutamate receptor 5 induced c-Rel-dependent expression of manganese superoxide dismutase (MnSOD) and Bcl-xL, supporting antioxidant and anti-apoptotic protection (Pizzi et al., 2005). Gain- and loss-of-function experiments established that c-Rel was both necessary and sufficient for this response, while TAT–Bcl-xL restored survival when c-Rel signaling was impaired. This provides one of the clearest examples of a specific protective NF-κB transcriptional program, although the mechanism was demonstrated mainly under acute experimental conditions.

Calcium-dependent NF-κB signaling may also support neuronal adaptation. In neurons expressing wild-type or AD-associated mutant presenilin−1, endoplasmic-reticulum calcium release activated NF-κB and suppressed Aβ-associated expression of C/EBP homologous protein (CHOP), a mediator of apoptosis during unresolved ER stress (Schapansky et al., 2007b, Choi et al., 2006b). In M3 muscarinic receptor-expressing SH-SY5Y cells, oxotremorine-M similarly promoted calcium entry, p65 nuclear translocation, and sAPPα release, whereas inhibition of calcium entry or NF-κB reduced the response (Choi et al., 2006b). Because α-secretase cleavage generates sAPPα and prevents the same APP molecule from entering the amyloidogenic pathway, these findings suggest that under specific conditions NF-κB can favor non-amyloidogenic APP processing. Overall, however, these protective effects were observed during brief, stimulus-specific signaling over hours or days and therefore do not establish sustained canonical NF-κB activation as beneficial in chronic AD.

The direction of signaling changes when NF-κB becomes coupled to amyloidogenic APP processing. Tomita et al. showed that RelA/p65 overexpression selectively increased Aβ42 secretion, whereas the neuronal adaptor X11-like protein restrained this effect through a PDZ-domain-dependent interaction with p65 (Tomita et al., 2000b). More compelling mechanistic evidence emerged from studies of the BACE1 promoter. Bourne et al. found that NF-κB-associated regulation could restrain BACE1 promoter activity in resting neurons, whereas Aβ exposure in neurons and activation of astrocytes shifted the same κB-responsive region toward increased BACE1 transcription (Bourne et al., 2007b). Chen et al. subsequently demonstrated that TNF-α-induced BACE1 transcription required both RelA and an intact κB-binding sequence; mutation of the κB site or removal of RelA reduced BACE1 expression, APP processing, and Aβ generation (Chen et al., 2012b). These findings show that the same pathway can exert different transcriptional effects depending on cellular state and inflammatory context.

A dose-dependent shift may help explain this transition. Chami et al. reported that lower Aβ exposure was associated with suppression of βAPP and amyloidogenic secretase expression, whereas higher, supraphysiological Aβ concentrations reversed the response and increased βAPP and β- and γ-secretase components (Chami et al., 2012b). Although the use of HEK293 cells limits direct extrapolation to neurons, the bidirectional response supports the possibility that increasing Aβ-associated stress can shift NF-κB from a compensatory to an amyloidogenic state.

Once established, this state can reinforce BACE1 through more than one mechanism. Bcl−2-associated athanogene 1 M (BAG-1M) acted as a coactivator of NF-κB-dependent BACE1 transcription, increasing BACE1 expression and Aβ production, while hippocampal BAG-1M overexpression increased amyloid pathology and memory impairment in an AD mouse model (Shi et al., 2017). In parallel, Aβ1–42-induced NF-κB activation reduced ubiquitin C-terminal hydrolase L1 (UCH-L1) in neuronal cells, impairing cathepsin D-dependent lysosomal degradation of BACE1 and allowing the enzyme to accumulate (Guglielmotto et al., 2012b). Together, these studies show that disease-conditioned NF-κB signaling can increase BACE1 both by promoting its transcription and by limiting its degradation. Pharmacological studies are consistent with this interpretation: NF-κB inhibitors reduced Aβ40 and Aβ42 production in APP-expressing cells, whereas CD40 ligation increased APP metabolites and Aβ generation through partially NF-κB-dependent signaling (Ait-Ghezala et al., 2007b, Paris et al., 2007b), although these approaches are less mechanistically specific.

NF-κB-dependent neuronal transcription also extends beyond BACE1. Aβ42 increased phospholipase C delta 1 (PLC-δ1) expression through a functional κB element in its promoter, and mutation or deletion of this element abolished the response (Kim et al., 2003b). Similarly, experiments in primary rat cerebellar cultures exposed to 1–2 µM Aβ1–40 for 6–24 h showed that the strongest response occurred after approximately 12 h at 1 µM and involved NMDA-receptor-sensitive formation of both p50/RelA and p50/p50 complexes (Kawamoto et al., 2008b). The coexistence of transcriptionally distinct dimers reinforces the point that “NF-κB activation” does not represent a single molecular state, and that dimer composition can influence the resulting transcriptional program.

Mitochondrial dysfunction provides another link between pathological NF-κB signaling and neuronal injury. In HT22 hippocampal cells and isolated mitochondria, Aβ activated an intramitochondrial IκBα/NF-κB pathway associated with reduced cytochrome-c oxidase subunit III expression, impaired respiratory activity, and neuronal injury (Shi et al., 2014b). A complementary mechanism was identified through PTEN-induced kinase 1 (PINK1), a regulator of mitochondrial quality control. Loss or knockdown of PINK1 in Aβ-producing neuronal systems and AD-model mice increased mitochondrial reactive oxygen species, phosphorylation of NFKB1/p50 and RelA/p65, BACE1 and γ-secretase activity, and Aβ accumulation. Conversely, increasing neuronal PINK1 suppressed Aβ-induced NF-κB activation and amyloidogenic processing, while scavenging mitochondrial reactive oxygen species similarly reduced NF-κB activity, normalized APP processing, and limited Aβ accumulation (Du et al., 2025). These findings support a PINK1–mitochondrial ROS–NF-κB–APP-processing axis linking defective mitochondrial quality control to amyloidogenic signaling.

Persistent neuronal NF-κB activity can also promote inflammatory cell-death pathways. Lei et al. found increased NF-κB and miR−146a−5p together with reduced TP53-induced glycolysis and apoptosis regulator (TIGAR) in AD serum samples and Aβ25–35-exposed human hippocampal neurons (Lei et al., 2021). In the neuronal model, miR−146a−5p directly suppressed TIGAR, while NF-κB knockdown or restoration of TIGAR reduced oxidative stress and pyroptosis; reintroduction of miR−146a−5p weakened these protective effects. Because TIGAR supports pentose-phosphate pathway activity and NADPH-dependent antioxidant defense, these findings provide a plausible mechanism linking NF-κB-dependent microRNA signaling to impaired redox control and inflammatory neuronal death. A related pathway was identified by Zhang et al., who found increased TLR2 in individuals with amnestic MCI and AD and examined its function in Aβ1–42-treated SH-SY5Y cells and APP/PS1 mice. TLR2 suppression reduced NF-κB/NLRP3 signaling, pyroptosis, neuronal injury, AD-related pathology, and behavioral impairment, whereas NF-κB overexpression or direct NLRP3 activation partially reversed these benefits (Zhang et al., 2026b). These reversal experiments provide stronger support for a TLR2 → NF-κB → NLRP3 → pyroptosis pathway than would parallel changes in pathway markers alone. Nevertheless, both mechanisms remain supported primarily by experimental models and do not establish that neuronal pyroptosis driven by these pathways is a dominant process in sporadic human AD.

Broader in-vivo evidence nevertheless supports a detrimental role for persistent neuronal signaling in established disease. Schnöder et al. conditionally deleted Ikbkb, encoding IKKβ, specifically in neurons of APP- and tau-transgenic mice and complemented these experiments with neuronal cell models. Neuronal IKKβ deficiency reduced BACE1 expression and Aβ accumulation as well as phosphorylated tau, apoptosis, and inflammatory abnormalities (Schnöder et al., 2023). These findings place chronic neuronal IKKβ activity upstream of several major AD-related pathological processes and provide stronger causal evidence than studies based solely on pathway activation markers. Cognitive and synaptic improvement, however, was not uniform across the transgenic models, indicating that reducing molecular pathology may not be sufficient to restore neural networks once substantial damage has occurred. Moreover, because IKKβ acts on substrates beyond canonical NF-κB signaling, not every consequence of Ikbkb deletion can be attributed specifically to RelA-dependent transcription.

Taken together, the neuronal literature is better understood as evidence for distinct NF-κB signaling states rather than a simple protective-versus-harmful paradox. Brief activation during mild stress or trophic stimulation can support preconditioning, calcium homeostasis, non-amyloidogenic APP processing, antioxidant defense, and resistance to apoptosis, with the c-Rel–MnSOD/Bcl-xL pathway providing the clearest defined protective transcriptional program. In contrast, persistent signaling in an Aβ-rich, inflammatory, oxidatively stressed, or metabolically compromised environment increasingly engages IKKβ/RelA-associated BACE1 transcription, impaired BACE1 degradation, amyloidogenic processing, mitochondrial dysfunction, inflammatory microRNA signaling, NLRP3 activation, and pyroptotic cell death. The distinction is therefore not simply low versus high NF-κB activity, nor can it be reduced entirely to early versus late disease. The biological outcome depends on the signaling configuration and cellular context. Protective experiments explain why complete suppression of neuronal NF-κB could remove important physiological stress responses, whereas promoter-level studies, rescue and reversal experiments, and neuron-specific IKKβ deletion provide stronger causal evidence that sustained, disease-conditioned IKKβ/RelA signaling represents the more consistently pathogenic form of neuronal NF-κB activity in AD.