Human Evidence for Context-Dependent NF-κB Dysregulation in AD

Human studies support a role for NF-κB signaling in AD, but they do not support a simple model in which greater pathway activity is always harmful and reduced activity is necessarily beneficial. Instead, the evidence points to biologically distinct forms of dysregulation. Impaired inducible NF-κB signaling may increase susceptibility to late-onset Alzheimer’s disease (LOAD), whereas persistent or exaggerated RelA/p65-associated activity is more commonly observed in MCI and established AD. These states are not contradictory: failure to mount an appropriate stress response differs fundamentally from prolonged NF-κB activation in cells already exposed to Aβ, tau pathology, or chronic inflammation.

The clearest human evidence for impaired inducible signaling comes from studies of SHANK-associated RH domain-interacting protein (SHARPIN), a component of the linear ubiquitin chain assembly complex (LUBAC). Whole-exome sequencing of 202 APOE ε4-negative individuals with LOAD identified the rare SHARPIN G186R variant, which was subsequently associated with markedly increased LOAD risk in 4563 cases and 16,459 controls (OR: 6.1). Functionally, cells carrying G186R showed reduced NF-κB responsiveness after stimulation (Asanomi et al., 2019b). A second SHARPIN variant, R274W, was identified through whole-genome sequencing of 180 individuals with LOAD and 184 with MCI and was subsequently tested in 5043 LOAD cases and 11,984 controls. Although its association with LOAD was more modest (OR: 1.43), R274W similarly altered SHARPIN localization and reduced stimulus-induced NF-κB signaling (Asanomi et al., 2022b). The convergence of these findings suggests that intact SHARPIN–LUBAC function may be important for mounting an appropriate NF-κB response to cellular stress.

Experimental introduction of G186R using CRISPR/Cas9 provided a further link to Aβ biology. Homozygous G186R introduced into HEK293 cells produced weaker TNF-α-induced NF-κB signaling together with increased extracellular Aβ40 and Aβ42 secretion (Asanomi et al., 2024). This supports a biological connection among SHARPIN dysfunction, impaired inducible NF-κB signaling, and altered Aβ metabolism, but it does not establish that reduced NF-κB activity directly caused the increase in Aβ. SHARPIN may influence APP processing, trafficking, degradation, or secretion through additional mechanisms, and the use of a rare homozygous clone in a non-neural cell line limits extrapolation to heterozygous human carriers. The SHARPIN findings therefore argue that loss of normally regulated NF-κB responsiveness may increase disease vulnerability, not that generalized NF-κB activation would be protective.

Common NF-κB-related genetic variation is more difficult to interpret mechanistically. In a Spanish case-control study of 639 individuals with LOAD and 500 cognitively healthy controls, the CC genotype of NFKB1 rs7667496 was associated with LOAD, whereas the examined NFKBIA and NFKBIZ variants were not (Vazquez-Coto et al., 2025b). Because NFKB1 encodes p105 and its processed product p50, which can participate either in transcriptionally active p50/RelA complexes or in repressive p50/p50 homodimers, the genetic association alone does not indicate whether disease risk reflects increased activation, reduced repression, altered p105 processing, or another linked mechanism (Savinova et al., 2009, Somma et al., 2021).

Postmortem studies provide a different view by demonstrating NF-κB abnormalities directly in vulnerable human brain regions. Early investigations reported enhanced or altered RelA/p65 immunoreactivity in hippocampal and cortical regions, neurofibrillary tangles, dystrophic neurites, cholinergic neurons of the nucleus basalis of Meynert, and temporal cortex (Hirsch et al., 1997, Kitamura et al., 1997b, Terai et al., 1996b). These observations established that NF-κB components are altered in regions affected by AD, but localization or nuclear accumulation of p65 alone cannot determine whether the pathway was driving inflammatory transcription, participating in a compensatory stress response, or accumulating secondary to cellular injury.

Stronger functional evidence came from studies linking NF-κB DNA binding to specific transcriptional outputs. Lukiw et al. found that NF-κB DNA-binding activity in superior temporal neocortex closely paralleled transcription from the κB-responsive COX−2 promoter in sporadic AD brain (Lukiw and Bazan, 1998b). This finding is consistent with a pro-inflammatory NF-κB program in established disease, although the cross-sectional postmortem design cannot determine whether pathway activation preceded or followed neurodegeneration. Other studies further showed that NF-κB dysregulation cannot be described simply as uniform pathway activation. Increased IκB expression was observed alongside NF-κB activation and neurofibrillary pathology, while p105 and IκBγ were also elevated in AD brain (Huang et al., 2005b, Yoshiyama et al., 2001). Because these proteins have inhibitory or regulatory functions, their accumulation may reflect compensatory negative feedback, altered processing or degradation, or coexistence of different signaling states across cell types and brain regions.

NF-κB abnormalities are also associated with tau-related pathology. Ohta et al. identified a nuclear complex containing TDP−43 and p65 in temporal cortex, with the most prominent signal observed in four individuals with MCI and episodic-memory impairment (Ohta et al., 2014). Yamaguchi et al. later detected phosphorylated p65 within granulovacuolar degeneration structures and tau-positive neurites, particularly in AD tissue (Yamaguchi et al., 2019b). Together, these studies support a spatial association between altered NF-κB signaling and tau-related neuronal pathology across the MCI–AD spectrum. However, they do not establish that NF-κB initiates tau aggregation or drives its propagation, because phosphorylated or sequestered p65 within pathological structures could also reflect impaired clearance or a response to existing neuronal injury.

Altered NF-κB signaling in AD is not restricted to the brain. Ascolani et al. found that mitogen-stimulated peripheral blood mononuclear cells from individuals with AD showed increased NFKB1 expression, greater inducible p50/p65 DNA binding, and altered NF-κB target-gene expression, with some measures related to cognitive severity (Ascolani et al., 2012). These findings indicate systemic immune dysregulation, although peripheral NF-κB activity should not be assumed to mirror signaling within the brain. Cholinergic dysfunction may provide another regulatory influence. In AD brain tissue and Aβ1–42-treated neuronal and glial cell models, reduced α3 nicotinic acetylcholine receptor expression was associated with higher p65, CCL2, and CCL3 levels, while α3 nAChR knockdown reproduced part of the inflammatory response (Liao et al., 2016). This suggests that loss of cholinergic restraint may contribute to NF-κB-associated inflammation.

Human induced pluripotent stem cell (iPSC) models have provided more controlled evidence for cell-specific effects of AD risk factors on NF-κB signaling. Arnaud et al. compared isogenic APOE3/3, APOE4/4, and APOE-null iPSC-derived astrocytes and found that APOE4 reduced transgelin 3 (TAGLN3) expression and increased NF-κB responsiveness. Restoring TAGLN3 or pharmacologically modifying the pathway attenuated the inflammatory phenotype (Arnaud et al., 2022b). These findings suggest that APOE4 can shift astrocytes toward a state that is more responsive to NF-κB-mediated inflammatory signaling.

A related mechanism links astrocytic lipid handling to NF-κB activity. Hamilton et al. found increased fatty acid-binding protein 7 (FABP7) in plaque-associated astrocytes across AD-related mouse models, human AD tissue, and human iPSC-derived astrocytes. Overexpression of ligand-binding-competent FABP7 induced a broad NF-κB-associated inflammatory transcriptional program, whereas a ligand-binding-deficient mutant did not (Hamilton et al., 2024b). This supports a mechanistic connection between altered lipid handling and inflammatory NF-κB signaling in astrocytes. However, both APOE4- and FABP7-based models rely partly on isolated or reprogrammed astrocytes, which incompletely reproduce aging and the interactions with neurons, microglia, amyloid, tau, and the neurovasculature present in the human AD brain.

Taken together, the human evidence is most coherent when NF-κB is viewed as a dynamic, context-dependent signaling system rather than as a pathway that is uniformly increased or suppressed throughout AD. SHARPIN genetics suggests that impaired inducible NF-κB responsiveness may increase susceptibility to LOAD, whereas postmortem and human-cell studies more often identify lesion-associated, inflammatory, or exaggerated RelA/p65 signaling once disease pathology is established. APOE4, FABP7, cholinergic dysfunction, and peripheral immune changes further indicate that genetic, metabolic, and systemic factors can modify the threshold and character of NF-κB responses.

The apparent duality of NF-κB therefore reflects different failures of pathway regulation rather than a true contradiction. Genetic studies can identify abnormalities that may precede clinical disease but often cannot specify the responsible cell type or downstream mechanism; postmortem studies reveal pathway abnormalities in vulnerable brain regions but cannot establish temporal causality; peripheral studies demonstrate systemic immune changes but do not directly represent brain signaling; and isogenic human-cell models enable mechanistic investigation but incompletely reproduce aging and multicellular interactions. The available human evidence therefore supports neither global NF-κB activation nor global inhibition. Instead, preservation of appropriately regulated NF-κB responsiveness may support cellular homeostasis, whereas persistent, cell-specific, and RelA/p65-biased signaling appears increasingly associated with inflammatory and neurodegenerative pathology across the symptomatic AD spectrum.