Literature Identification and Appraisal of Evidence

The literature search was designed to identify mechanistically informative studies rather than every intervention in which NF-κB changed as a secondary outcome. PubMed/MEDLINE, Europe PMC, publisher databases, and backward and forward citation searches were examined iteratively through 3 August 2026. Search terms combined Alzheimer’s disease, mild cognitive impairment, amyloid-β, amyloid precursor protein, amyloid and tau models, and major NF-κB pathway components, including NF-κB, RelA/p65, NFKB1/p50, NFKB2/p52, c-Rel, IκB kinase, IκB, and NF-κB essential modulator (NEMO). Reviews, editorials, comments, patents, corrections, and retracted records were excluded from the primary synthesis.

The final evidence set comprised 71 primary studies selected for their contribution to human disease relevance, direct pathway manipulation, promoter-level or molecular mechanisms, pathway blockade or rescue, cell-specific causality, or clarification of protective versus detrimental effects. Selection was therefore purposive and mechanism centered. The 71 studies should be interpreted as an analytically defined evidence set rather than as a denominator for estimating the prevalence of protective or pathogenic NF-κB effects.

For each study, we extracted the experimental model and cell type, species or human source, age and sex when reported, Aβ or tau species and exposure conditions, pathway component and manipulation, sample size, principal methods, NF-κB readout, disease-relevant outcome, mechanistic direction, causal strength, and major limitations. Numerical details were included only when verifiable in the primary source; unavailable information was recorded as not reported. This was particularly important for Aβ studies, in which peptide species, aggregation state, concentration, exposure duration, and accompanying stimuli can substantially alter NF-κB responses.

Evidence was classified according to the degree of causal inference supported by the study design. Tier A included cell-relevant in-vivo genetic manipulations testing necessity or sufficiency for an organism-level phenotype. Tier B included direct genetic or pharmacological perturbations strengthened by promoter mutation, pathway blockade, epistasis, or rescue. Tier C comprised human association studies and mechanistic experiments demonstrating disease relevance without organism-level causality, whereas Tier D included localization studies and secondary pathway measurements without direct tests of necessity or sufficiency. These tiers reflect causal resolution rather than overall study quality.

The evidence base was dominated by experimental and amyloid-focused studies. Forty-six studies included an in-vitro component, 40 an in-vivo animal, fly, or nematode component, and 17 participants, postmortem tissue, or human-derived cells; categories overlapped. Aβ exposure or injection was used in 49 studies, compared with seven APP/PS1 studies, two 5xFAD studies, and relatively few APP knock-in or dedicated tau models. Neurons were examined in 42 studies, microglia in 15, astrocytes in 13, and endothelial or pericyte compartments in six. Pathway definition was also uneven: 49 studies reported generic NF-κB activity, 21 examined RelA/p65, ten p50, seven IKK/IKKβ, and only one centered on c-Rel. Thus, the literature is rich in acute Aβ- and RelA-associated findings but comparatively sparse for c-Rel, non-canonical signaling, multicellular interactions, vascular biology, aging, and sporadic AD.

Of the 71 studies, 41 were judged predominantly detrimental, ten beneficial or potentially homeostatic, nine mixed, two beneficial at the IKKβ level without demonstrated mediation by canonical NF-κB transcription, and nine directionally unresolved. Eight studies met Tier A criteria, 25 Tier B, 18 Tier C, and 20 Tier D. These counts describe the composition of the selected literature and should not be interpreted as pooled effect estimates or as evidence that pathogenic NF-κB states are biologically more prevalent. Direct genetic, promoter-level, blockade, or rescue experiments were therefore given greater interpretive weight than secondary changes in NF-κB markers after indirect interventions. (Table 1)

Dimension | Distribution | Critical implication
Publication period | 1996–2005: 18 studies; 2006–2015: 26; 2016–2026: 27 | Later studies increasingly incorporated cell-specific genetics, tau pathology, APOE4, and human-derived models.
Experimental design* | 46 in vitro; 40 in vivo; 17 involving participants, postmortem tissue, or human-derived cells | Mechanistic evidence is extensive, but no human interventional study directly tested selective NF-κB modulation.
Disease model | 49 used Aβ exposure or injection; 7 APP/PS1; 2 5xFAD; few APP knock-in or dedicated tau models | Acute amyloid stress dominates the literature.
Cell or tissue compartment* | 42 neuronal; 15 microglial; 13 astrocytic; 6 endothelial or pericytic | Neuronal models predominate; multicellular and vascular interactions are underrepresented.
Causal tier | Tier A: 8; Tier B: 25; Tier C: 18; Tier D: 20 | Causal weight varies substantially across studies.
Biological direction | 41 detrimental; 10 beneficial or potentially homeostatic; 9 mixed; 2 beneficial at the IKKβ level with canonical NF-κB mediation unresolved; 9 unresolved | Directional counts reflect study composition, not biological prevalence.
Pathway definition* | 49 generic NF-κB activity; 21 RelA/p65; 10 p50; 7 IKK/IKKβ; 1 centered on c-Rel | Broad conclusions are often based on RelA/p65 or nonspecific pathway measures.

Several limitations should be acknowledged. This was a narrative rather than systematic review; the protocol was not prospectively registered, screening and data extraction were not independently duplicated, and formal domain-based risk-of-bias tools were not applied. Considerable heterogeneity in experimental models, exposure conditions, NF-κB measurements, and outcomes also precluded quantitative pooling. Publication bias, model selection, and preferential reporting of statistically significant or mechanistically favorable findings may therefore have influenced the available evidence.

Table 2 provides the study-level evidence map, including study design, population or model, exposure conditions, comparator, principal methods and outcomes, key findings, and major limitations. This structured map was used to distinguish direct causal evidence from association and to support the narrative synthesis of broader mechanistic patterns rather than presenting the 71 studies as isolated summaries. (Table 2)

Author(s) / Date | Population/model | Intervention/exposure | Comparator | Study design/setting | Main findings | Main limitation | Ref
(Sudo, Sung et al. 2026) | H4 cells; primary mouse glia predominantly astrocytes; seven-month-old male wild-type mice; 14–16-month-old female AppNL-G-F knock-in and AppNL-G-F; Klk7-null mice; Mayo and MSBB human AD transcriptomic cohorts. | H4 cells: IKK−16 2 µM or JSH−23 30 µM for 24 h with 20 nM Aβ40; primary glia: IKK−16 with 48-h Aβ-degradation assay. Mice: 4 µL of 100 µM IKK−16 at 0.3 µL/min into each hippocampus, with contralateral vehicle; collection at 24 h (Klk7) or 48 h (Aβ). | transgenic/modified vs wild-type; loss-of-function vs matched control; inhibitor/antagonist vs vehicle/untreated; treated hippocampus vs contralateral vehicle. | Human in silico/omics; Human in vitro; Rodent in vitro; Rodent in vivo | Both NF-κB inhibitors increased KLK7 expression and extracellular Aβ degradation without affecting cell viability or Aβ uptake. Local hippocampal IKK−16 reduced soluble and insoluble Aβ and decreased plaque size, but not plaque number. These Aβ-lowering effects disappeared in Klk7-null mice. | 48-h local inhibition cannot establish chronic efficacy, cognition, cell specificity or preservation of other NF-κB functions; KLK7 is not astrocyte-exclusive. | (Sudo et al., 2026)
(Ye, Deng et al. 2026) | APP/PS1 mice across disease ages; Aβ-exposed primary mouse astrocytes, including Hdac7-floxed cultures; astrocyte-neuron conditioned-medium experiments. | Astrocyte-directed AAV-GfaABC1D HDAC7 gain or loss, Hdac7 excision, IKK blockade, and intraperitoneal TMP195. Vector titre, TMP195 dose, group size, sex, and schedule: NR. | HDAC7 gain-of-function, loss-of-function, and IKK blockade vs matched vector, floxed, or vehicle controls. | Rodent in vitro; Rodent in vivo | Aβ increased astrocytic HDAC7, which activated IKK/NF-κB. HDAC7 knockdown, IKK inhibition, or TMP195 reduced astrocyte-mediated neuronal injury and improved cognitive outcomes. | The APP/PS1 setting and incomplete dosing limit translation to sporadic AD. | (Ye et al., 2026)
(Zhang, Wang et al. 2026) | Aβ1–42-treated SH-SY5Y neuronal cells, APP/PS1 mice, and human-cell/clinical corroborative material reported by the authors. | TLR2 knockdown in cells and sh-TLR2 delivery in APP/PS1 mice; NF-κB overexpression and NLRP3 reactivation as reversal tests. Aβ concentration, vector dose, route, and treatment duration: NR. | TLR2 knockdown vs control; NF-κB overexpression or NLRP3 reactivation as reversal conditions. | Human in vitro; Rodent in vivo | TLR2 knockdown reduced inflammation, neuronal pyroptosis, brain pathology, and behavioral impairment. Re-activating NF-κB or NLRP3 partially removed this protection. | Neuronal-line and familial-transgenic models do not model the aged sporadic brain. | (Zhang et al., 2026a)
(Du, Yu et al. 2025) | PINK1-manipulated AD-model mice and Aβ-producing neuronal cell lines. | Neuronal PINK1 gain and loss plus ROS scavenging; doses, treatment duration, animal age, sex, and group size: NR. | PINK1 gain- and loss-of-function vs matched controls; reactive-oxygen-species scavenging as a pathway-interruption condition. | In vitro; Rodent in vivo | PINK1 loss increased mitochondrial reactive oxygen species, p50/p65 phosphorylation, secretase activity, and Aβ production. Scavenging reactive oxygen species prevented these downstream changes. | NF-κB necessity was not directly tested; pathway activation could remain a correlated intermediate. | (Du et al., 2025)
(Liu, He et al. 2025) | Public human AD transcriptomic datasets, Aβ1–42-treated SH-SY5Y cells, LPS-stimulated HMC3 microglia, and APP/PS1 mice. | SLAMF8 overexpression and NINJ2 knockout or suppression; Aβ/LPS exposure and in-vivo manipulation doses and durations: NR. | SLAMF8 overexpression with or without NINJ2 loss vs matched controls. | Human in silico/omics; Human in vitro; Rodent in vivo | SLAMF8 increased inflammatory and oxidative injury by activating TLR4/NF-κB through NINJ2. Removing NINJ2 prevented these effects in the tested cell and mouse models. | The mixed toxin, immortalized-cell, and familial-mouse systems complicate attribution to one disease stage. | (Liu et al., 2025)
(Vazquez-Coto, Perez-Oliveira et al. 2025) | Spanish case-control cohort of 639 late-onset AD cases and 500 cognitively healthy controls. | Genotyping of functional/common NFKB1, NFKBIA, and NFKBIZ variants; no experimental exposure. | AD vs cognitively healthy controls. | Human observational/genetic | The NFKB1 rs7667496 CC genotype and a related haplotype were associated with higher late-onset AD risk. The study did not test how the variant alters NF-κB function. | Population stratification, multiple testing, linkage, and lack of independent functional replication constrain causal interpretation. | (Vazquez-Coto et al., 2025a)
(Asanomi, Kimura et al. 2024) | HEK293 cells with homozygous CRISPR/Cas9 knock-in of SHARPIN G186R; reported knock-in efficiency was below 1%. | Endogenous risk-variant knock-in then TNF-α stimulation; TNF-α dose and duration were NR in the accessible article summary. | SHARPIN G186R knock-in cells vs reference/wild-type cells. | Human in vitro | Introducing the SHARPIN G186R risk variant reduced stimulus-induced NF-κB signaling and accelerated extracellular Aβ secretion. | A single rare clone in non-neural cells and very low editing efficiency raise clonal and generalizability concerns. | (Asanomi et al., 2024)
(Hamilton, Kinscherf et al. 2024) | Primary rat and mouse hippocampal astrocytes, Aβ25–35 exposure, APP/PS1 mouse tissue, human AD tissue, and human iPSC-derived astrocytes. | Wild-type or ligand-binding-deficient FABP7 overexpression; Aβ concentration and exposure duration: NR. | Wild-type FABP7 vs ligand-binding-deficient FABP7; AD-model vs control tissue. | Human in vitro; Rodent in vitro; Rodent in vivo | FABP7 was increased in plaque-associated astrocytes in mouse and human AD tissue. Only ligand-binding-competent FABP7 induced the broad NF-κB-linked inflammatory gene program. | Monoculture activation and acute Aβ25–35 do not establish chronic in-vivo neurotoxicity. | (Hamilton et al., 2024a)
(Jong Huat, Camats-Perna et al. 2024) | Human bulk and cell-resolved AD transcriptomic data and postmortem brain; healthy mice with chronic astrocytic NF-κB activation; an AD mouse model with astrocytic pathway inhibition. | Chronic astrocyte-specific gain of NF-κB in healthy brain and inhibition in diseased brain; construct, dose, age, and duration details: NR. | Healthy vs AD human tissue/data; astrocytic NF-κB activation in healthy mice and inhibition in an AD model vs matched controls. | Human ex vivo/biospecimen; Human in silico/omics; Rodent in vivo | Sustained astrocytic NF-κB activation disrupted protein homeostasis and increased inflammation in healthy mice. In contrast, inhibiting astrocytic NF-κB in an AD model accelerated Aβ and tau accumulation. | Different backgrounds and intervention directions prevent a simple within-model stage comparison. | (Jong Huat et al., 2024)
(Wahl, Risen et al. 2024) | Nineteen-month-old wild-type mice as an aging model and two-month-old rTg4510 tau-transgenic mice. | A brain-penetrant nanoligomer cocktail directed at NF-κB and NLRP3 for four weeks; dose, route, sex balance, and group size: NR. | Nanoligomer-treated vs control mice across aged wild-type and rTg4510 models. | Rodent in vivo | A four-week nanoligomer treatment targeting both NF-κB and NLRP3 reduced neuroinflammation and improved cognitive and tau-related outcomes in aging and tauopathy models. | The cocktail targets two nodes simultaneously and does not isolate NF-κB-specific mediation. | (Wahl et al., 2024)
(Zhao, Liu et al. 2024) | Plasma from patients with AD, APP/PS1 and 5xFAD mice, and EOC20 microglia in homeostatic or LPS/IFN-γ-activated states. | miR−25802 overexpression or inhibition with KLF4 restoration as a rescue; oligonucleotide, LPS/IFN-γ, and in-vivo doses and durations: NR. | miR−25802 gain/loss vs controls; KLF4 restoration as rescue; AD vs control plasma. | Human observational; Rodent in vitro; Rodent in vivo | miR−25802 promoted an inflammatory microglial state, worsened amyloid pathology, and impaired cognition. Restoring KLF4 weakened these effects. | EOC20 cells and two familial amyloid models may overrepresent an activated microglial phenotype. | (Zhao et al., 2024a)
(Schnöder, Quan et al. 2023) | Neuron-specific IKKβ-deficient mice crossed with APP- and tau-transgenic lines, plus SH-SY5Y experiments. | Conditional neuronal Ikbkb deletion and cellular IKKβ manipulation; ages, sex, group sizes, and cell-treatment doses: NR. | Neuron-specific Ikbkb deletion vs floxed/control mice in APP and tau backgrounds. | Human in vitro; Rodent in vivo | Deleting neuronal IKKβ reduced BACE1, Aβ, phosphorylated tau, apoptosis, and inflammation. Cognitive improvement occurred in some, but not all, transgenic models. | Cognition diverged between APP and tau models, arguing against a single downstream mechanism. | (Schnöder et al., 2023)
(Arnaud, Benech et al. 2022) | Isogenic human iPSC lines carrying APOE3/3, APOE4/4, or APOE knockout differentiated into astrocytes, with AD-brain validation. | CRISPR-defined APOE genotypes, pharmacological NF-κB modulation, and TAGLN3 re-expression or supplementation; compound concentrations and durations: NR. | loss-of-function vs matched control; rescue/restoration vs intervention alone; isogenic APOE3/3 vs APOE4/4 vs APOE-null astrocytes. | Human in vitro | APOE4 reduced TAGLN3 and made human iPSC-derived astrocytes persistently more responsive to inflammatory NF-κB signaling. Restoring TAGLN3 or pharmacologically modifying the pathway reversed this state. | iPSC reprogramming attenuates aging and astrocyte monoculture omits plaque, microglial, and vascular feedback. | (Arnaud et al., 2022a)
(Asanomi, Shigemizu et al. 2022) | Whole-genome sequencing in 180 late-onset AD and 184 MCI participants, then association testing in 5043 cases and 11,984 controls. | Identification and functional testing of SHARPIN R274W; No therapeutic exposure. | AD/MCI vs controls; risk variant vs reference/wild-type condition. | Human observational/genetic | The SHARPIN R274W risk variant altered protein localization, weakened inducible NF-κB signaling, and was associated with increased late-onset AD risk. | Modest effect size and uncertain responsible cell type prevent therapeutic extrapolation. | (Asanomi et al., 2022a)
(Wang, Gu et al. 2022) | Aβ-treated SH-SY5Y neuronal cells and APP/PS1 mice. | IKKβ silencing or constitutive activation with autophagy and RIPK1-necroptosis manipulation; doses, vector titres, animal ages, and durations: NR. | IKKβ silencing vs constitutive activation and matched controls. | Human in vitro; Rodent in vivo | IKKβ activation improved autophagic flux, reduced RIPK1-mediated necroptosis and Aβ accumulation, and improved neuronal, pathological, and behavioral outcomes. | Canonical NF-κB transcription was not shown to mediate autophagy/necroptosis effects. | (Wang et al., 2022b)
(Wang, Gu et al. 2022) | APP/PS1 mice with hippocampal CA1 manipulation and Aβ1–42-treated N2a neuronal cells. | miR−155–5p inhibition, IKKβ knockdown, SKP2 manipulation, and adenoviral delivery; vector titre, Aβ dose, treatment duration, animal age, and sex: NR. | miR−155 inhibition, IKKβ knockdown, and SKP2 manipulation vs matched controls. | Rodent in vitro; Rodent in vivo | miR−155–5p stabilized IKKβ by suppressing SKP2. Inhibiting miR−155–5p or reducing IKKβ lowered Aβ deposition, tissue injury, and cognitive impairment. | Direct canonical NF-κB transcriptional mediation and vector-dose dependence were not resolved. | (Wang et al., 2022c)
(Wang, Fan et al. 2022) | Primary mouse microglia; young (three-month-old) and aged (8–11-month-old) PS19 tauopathy mice with Cx3cr1-CreERT2-mediated microglial Ikbkb deletion or constitutive activation; AD- and PSP-derived tau preparations and tau-biosensor cells. | Primary microglia: full-length tau fibrils 2 µg/mL or K18/PL fibrils 2.5 µg/mL for 24 h; AD-tau 1 µg/mL after 2-h pretreatment with TPCA−1 0.1 µM. In vivo: unilateral hippocampal K18/PL, 2 µL at 2.5 µg/µL for loss-of-function or 2 µL at 0.2 µg/µL for gain-of-function, assessed after one month; chronic PS19 outcomes at 8–10 months. | Microglial Ikbkb deletion vs constitutive activation and matched controls; tau-exposure and TPCA−1 conditions. | Rodent in vitro; Rodent in vivo | Microglial NF-κB activation increased the release of tau capable of seeding new aggregates and promoted tau spread and cognitive decline. Microglial IKKβ deletion reduced propagation, improved autophagy and cognition, and normalized microglial gene expression, even though neuronal tau inclusions increased. | PS19 overexpresses mutant P301S tau; long-term consequences of retained inclusions and exact causal export machinery remain unresolved. | (Wang et al., 2022a)
(Chen, Zhang et al. 2021) | Mice receiving intracerebroventricular Aβ1–42 and Aβ-treated HT22 hippocampal cells. | Hippocampal DJ−1 shRNA, 8 × 10^4 viral particles per mouse for seven days, plus DJ−1/IKKβ/pVHL plasmid or shRNA manipulations in HT22 cells. | DJ−1, IKKβ, and pVHL gain/loss conditions vs matched plasmid or shRNA controls. | Rodent in vitro; Rodent in vivo | The DJ−1/pVHL pathway supported IKKβ-dependent autophagy and reduced phosphorylated tau, neuronal injury, and behavioral impairment. | Acute Aβ injection is not progressive AD and the benefit was not proven to require nuclear NF-κB transcription. | (Chen et al., 2021a)
(Lei, Liu et al. 2021) | Serum from 112 patients with AD and 101 healthy controls; commercial human hippocampal neurons exposed to Aβ25–35. | NF-κB knockdown, TIGAR overexpression, and miR−146a−5p manipulation/rescue; Aβ concentration and exposure duration: NR. | AD vs control serum; loss-of-function vs matched control; gain-of-function vs vector/baseline; rescue/restoration vs intervention alone. | Human observational/genetic; Human ex vivo/biospecimen; Human in vitro | NF-κB increased miR−146a−5p, which suppressed TIGAR and increased oxidative stress and neuronal pyroptosis. Restoring TIGAR or reducing the pathway attenuated these effects. | Serum and acute peptide-treated neurons cannot establish the brain source or temporal order in AD. | (Lei et al., 2021)
(Yang, Magnutzki et al. 2021) | APP23 mice with conditional astrocytic IKK2/NF-κB activation, with primary astrocyte and microglial phenotyping. | Astrocyte-specific IKK2 activation; induction timing, animal age, sex, group size, and activation magnitude: NR. | Astrocyte-specific IKK2/NF-κB activation vs nonactivated controls; direct microglial activation as comparator. | Rodent in vitro; Rodent in vivo | Astrocyte-specific IKK2/NF-κB activation increased gliosis but reduced the number and size of compact plaques and shifted nearby microglia toward an Aβ-handling state. | Soluble Aβ species and neuronal function are needed before plaque reduction is called net protection. | (Yang et al., 2021a)
(Asanomi, Shigemizu et al. 2019) | Whole-exome sequencing of 202 late-onset AD cases lacking APOE ε4, followed by 4563 cases and 16,459 controls plus cellular functional assays. | Functional characterization of rare SHARPIN G186R; No therapeutic exposure. | AD/MCI vs controls; risk variant vs reference/wild-type condition. | Human observational/genetic | The rare SHARPIN G186R variant weakened stimulated NF-κB responses and was associated with substantially higher late-onset AD risk. | Rarity, ancestry, and uncertain brain-cell locus limit generalization. | (Asanomi et al., 2019a)
(Yamaguchi, Ayaki et al. 2019) | postmortem hippocampus from 21 autopsies spanning AD, ALS with optineurin pathology, other neurodegenerative diseases, and controls. | No exposure; comparative neuropathological staining. | AD/disease vs control tissue. | Human ex vivo/biospecimen | Phosphorylated p65 accumulated in granulovacuolar degeneration and tau-positive neurites, particularly in AD tissue. | Cannot distinguish active transcription from sequestration or determine whether phospho-p65 precedes tau pathology. | (Yamaguchi et al., 2019a)
(Li, Sibon et al. 2018) | Drosophila proteotoxicity screen initiated in a neuronal SCA3 eye model and tested in flies expressing human neuronal Aβ. | Astrocyte-specific inhibition of the Drosophila NF-κB homolog Relish; genetic driver strength and induction timing: NR. | Astrocyte-like glial Relish inhibition vs genetic-driver controls. | Invertebrate in vivo | Inhibiting Relish/NF-κB specifically in astrocyte-like glia delayed neurodegeneration and extended survival without directly changing neuronal Aβ expression. | Fly innate immunity and astrocyte biology only partially model mammalian AD. | (Li et al., 2018)
(Tanaka, Sabharwal et al. 2018) | Psen1-deficient mouse endothelial cells and cells with PS1 re-expression or overexpression. | PS1 gain/loss, γ-secretase inhibition, and manipulation of BCR/CK2α/p65 signaling; concentrations and durations: NR. | Psen1 loss, restoration, or overexpression with γ-secretase/pathway inhibition vs matched controls. | Rodent in vitro | Presenilin 1 assembled a BCR-CK2α-p65 complex and promoted p65 Ser529 phosphorylation, NF-κB reporter activity, and interleukin−6 expression independently of γ-secretase. | The study did not establish an AD phenotype or in-vivo vascular consequence. | (Tanaka et al., 2018)
(Shi, Hong et al. 2017) | APP/BACE1 promoter cell systems and an AD mouse model receiving hippocampal lentiviral BAG-1M. | BAG-1M overexpression and promoter-complex manipulation; lentiviral titre, animal age, dose, and follow-up: NR. | BAG-1M overexpression/promoter manipulation vs vector or promoter controls. | In vitro; Rodent in vivo | BAG-1M increased NF-κB activity at the BACE1 promoter, raising BACE1 expression, Aβ production, plaque burden, and memory impairment. | Supraphysiological BAG-1M and incomplete dosing information limit effect-size interpretation. | (Shi et al., 2017)
(Liao, Qi et al. 2016) | postmortem hippocampal, temporal, and frontal cortex from AD and control brains; Aβ1–42-treated SH-SY5Y and U87MG cells. | Aβ1–42 exposure and α3 nicotinic-acetylcholine-receptor siRNA; dose, duration, and human sample size: NR. | AD vs control postmortem tissue; α3-nAChR knockdown vs control siRNA in Aβ-exposed cells. | Human ex vivo/biospecimen; Human in vitro | p65 and inflammatory chemokines were increased in AD brain tissue and Aβ-treated neural cells. Reducing α3 nicotinic acetylcholine receptor expression further increased this inflammatory response. | Bulk tissue composition, acute Aβ exposure, and absent causal in-vivo testing restrict inference. | (Liao et al., 2016)
(Yang, Liu et al. 2016) | Primary mouse microglia basal or LPS. | Lrp1 knockdown or receptor-associated protein blockade, with JNK and NF-κB inhibitors; inhibitor and LPS concentrations and durations: NR. | loss-of-function vs matched control; inhibitor/antagonist vs vehicle/untreated. | Rodent in vitro | Loss or blockade of LRP1 increased JNK/NF-κB activity and made primary microglia more responsive to inflammatory stimulation. | The experiment did not include amyloid or an AD animal phenotype. | (Yang et al., 2016)
(Lian, Yang et al. 2015) | Aβ-exposed astrocyte-neuron systems, APP-transgenic mice, and human AD brain. | Astroglial NF-κB manipulation and neuronal C3a-receptor antagonism; Aβ, genetic, and antagonist doses/schedules: NR. | Astroglial NF-κB manipulation and neuronal C3aR antagonism vs matched controls. | Human ex vivo/biospecimen; In vitro; Rodent in vivo | Aβ-activated astrocytic NF-κB increased complement C3 release. C3 signaling through neuronal C3a receptors damaged dendrites, disrupted network function, and impaired cognition; receptor blockade reduced these effects. | Human evidence is cross-sectional and intervention details require full-text verification. | (Lian et al., 2015a)
(Quan, Yue et al. 2015) | APP Swedish-mutant SH-SY5Y cells and CL2006 Caenorhabditis elegans expressing human Aβ. | PRMT5 knockdown, E2F1 codepletion, and NF-κB or GSK3 inhibition; inhibitor concentrations and worm exposure schedules: NR. | PRMT5 knockdown with E2F1 codepletion or NF-κB/GSK3 inhibition vs matched controls. | Human in vitro; Invertebrate in vivo | PRMT5 loss activated an E2F1-NF-κB-GSK3β stress pathway and increased apoptosis. NF-κB or GSK3 inhibition reduced cell injury and age-related paralysis in the nematode model. | Cultured tumor cells and nematodes lack the mammalian glial and vascular context. | (Quan et al., 2015)
(Rangasamy, Corbett et al. 2015) | Five-month-old 5xFAD mice, with wild-type littermate controls. | Wild-type NEMO-binding-domain peptide or inactive mutant peptide, 0.1 mg/kg intranasally every other day for 30 days; a single-dose distribution experiment assessed brain entry at 30 min. | transgenic/modified vs wild-type; active peptide vs inactive mutant peptide. | Rodent in vivo | Intranasal NEMO-binding-domain peptide entered the brain, reduced canonical NF-κB signaling and gliosis, improved amyloid- and tau-related measures, and prevented memory impairment. | Small groups and broad NEMO/IKK inhibition leave cell specificity and preserved homeostatic signaling unresolved. | (Rangasamy et al., 2015a)
(Liu, Liu et al. 2014) | Six-month-old TgCRND8 amyloid mice with LysM-Cre-mediated conditional myeloid Ikbkb deletion; isolated microglia, bone-marrow-derived macrophages plus recruitment/internalization assays. | Genetic myeloid IKKβ deficiency. In vitro, oligomeric Aβ42 was tested at 1 or 10 µM in macrophage internalization assays; no in-vivo drug dose. | Myeloid Ikbkb deletion vs littermate controls; IKKβ-deficient vs control macrophages in Aβ internalization assays. | Rodent in vitro; Rodent in vivo | Myeloid IKKβ deletion reduced TNF and interleukin−1β, lowered selected Aβ measures, preserved synaptic proteins, and improved cognition. Microglial recruitment and Aβ internalization were preserved or increased. | LysM targeting includes peripheral myeloid cells, the responsible NF-κB dimer was not isolated, and greater uptake did not by itself prove complete lysosomal degradation. | (Liu et al., 2014)
(Ohta, Tremblay et al. 2014) | Temporal cortex from 12 MCI, 12 AD, and 12 age-matched control cases; four MCI cases with episodic-memory deficits were analyzed in depth. | No treatment; postmortem immunoprecipitation, immunoblotting, and immunofluorescence. | MCI and AD tissue vs age-matched controls. | Human ex vivo/biospecimen | A nuclear TDP−43-p65 complex and phosphorylated p65 were detected in a small subset of MCI cases with episodic-memory impairment. | The highlighted signal in four MCI cases is exploratory and cross-sectional. | (Ohta et al., 2014)
(Park, Kook et al. 2014) | bEnd.3 mouse brain endothelial cells, astrocyte-endothelial contact/polarity models, and 5xFAD mouse brain. | Aβ1–42 exposure, a RAGE-neutralizing antibody, and NF-κB inhibition; concentrations, duration, animal age, and group size: NR. | Aβ-exposed vs untreated endothelial models; RAGE neutralization or NF-κB inhibition vs control. | Rodent in vitro; Rodent in vivo | Aβ1–42 activated endothelial RAGE-NF-κB signaling and reduced P-glycoprotein. Blocking RAGE or NF-κB prevented transporter loss. | An immortalized endothelial line and familial amyloid model may not capture aged human BBB physiology. | (Park et al., 2014)
(Rolova, Puli et al. 2014) | Nfkb1/p50-knockout primary microglia, intrahippocampal LPS models, and p50-deficient mice crossed with APdE9 amyloid mice. | Constitutive p50 loss with LPS or Aβ challenges; challenge doses, ages, sex, and group sizes: NR. | Nfkb1/p50-null vs wild-type microglia and mice under acute and chronic inflammatory/Aβ conditions. | Rodent in vitro; Rodent in vivo | Loss of p50 increased microglial Aβ uptake and modestly reduced amyloid burden, but it also intensified chronic inflammatory gene expression and leukocyte infiltration. | Genetic deletion reveals a true trade-off—more uptake/slightly less Aβ but worse inflammation—although lifelong p50 loss permits developmental and systemic compensation. | (Rolova et al., 2014)
(Shi, Zhu et al. 2014) | HT22 mouse hippocampal cells and isolated mitochondria exposed to Aβ. | Aβ treatment and subcellular pathway perturbation; peptide species, concentration, exposure duration, and inhibitor doses: NR. | Aβ-exposed vs untreated cells/mitochondria with pathway perturbation. | Rodent in vitro; Subcellular/cell-free | Aβ activated an IκB/NF-κB system within mitochondria and was associated with impaired respiratory-chain function and mitochondrial dysfunction. | Immortalized cells and isolated mitochondria do not establish organism-level neurodegeneration. | (Shi et al., 2014a)
(Ascolani, Balestrieri et al. 2012) | Mitogen-stimulated peripheral-blood mononuclear cells from 12 patients with AD and 12 age-matched controls. | Ex-vivo mitogenic stimulation; mitogen identity/concentration and incubation duration: NR. | AD vs age-matched control PBMCs. | Human ex vivo | Mitogen-stimulated peripheral blood mononuclear cells from patients with AD showed greater NFKB1 expression, p50/p65 DNA binding, and target-gene expression. Several measures correlated with cognitive severity. | The small blood cohort is vulnerable to medication and systemic comorbidity and is not a direct brain readout. | (Ascolani et al., 2012)
(Bell, Winkler et al. 2012) | Human APOE2-, APOE3-, or APOE4-targeted-replacement mice and Apoe-null mice, with pericyte/vascular analyses. | Genetic CypA ablation and pharmacological CypA-pathway inhibition; compound dose and treatment schedule: NR. | APOE2, APOE3, APOE4, and Apoe-null mouse groups; loss-of-function vs matched control; inhibitor/antagonist vs vehicle/untreated. | Rodent in vivo (human APOE targeted-replacement) | APOE4 failed to suppress the pericyte CypA-NF-κB-MMP9 pathway. This caused blood-brain barrier leakage, reduced cerebral blood flow, and preceded neuronal and synaptic dysfunction. | Mice model APOE-dependent barrier injury rather than the complete Alzheimer spectrum. | (Bell et al., 2012)
(Chami, Buggia-Prévot et al. 2012) | HEK293 cells with or without Swedish-mutant APP expression. | Physiological vs supraphysiological Aβ exposure plus canonical and alternative NF-κB kinase modulators; exact concentration ranges and durations: NR. | Physiological vs supraphysiological Aβ conditions in cells with or without Swedish-mutant APP. | Human in vitro | At lower Aβ levels, NF-κB reduced APP and secretase expression. At higher Aβ levels, the direction reversed and NF-κB increased the same amyloid-producing machinery. | Non-neural HEK293 cells and incomplete exposure levels limit physiological extrapolation. | (Chami et al., 2012a)
(Chen, Zhou et al. 2012) | AD and control brain material plus neuronal/promoter systems, including RelA-deficient cells. | p65 expression, TNF stimulation, NSAID/NF-κB modulation, BACE1-promoter κB-site mutation, and RelA loss; doses and durations: NR. | AD/disease vs control tissue; loss-of-function vs matched control; wild-type vs mutated/deleted κB promoter. | Human ex vivo/biospecimen; In vitro | RelA/p65 and BACE1 were increased in AD brain tissue. Mutating the BACE1 κB site or removing RelA prevented promoter activation and reduced BACE1 expression, APP processing, and Aβ production. | Pharmacological effects and bulk human expression do not prove in-vivo neuronal necessity. | (Chen et al., 2012a)
(Guglielmotto, Monteleone et al. 2012) | SH-SY5Y and NT2 neuronal cells exposed to Aβ1–42, with postmortem AD-brain corroboration. | Aβ1–42 and NF-κB inhibition; peptide and inhibitor concentrations and treatment duration: NR. | Aβ1–42-exposed vs control neuronal cells; NF-κB inhibition vs vehicle; AD vs control tissue. | Human ex vivo/biospecimen; Human in vitro | Aβ1–42 activated NF-κB, reduced UCH-L1, and impaired cathepsin D-dependent lysosomal degradation of BACE1, allowing BACE1 to persist longer. | Rests largely on acute tumor-cell models; human tissue cannot establish temporal ordering. | (Guglielmotto et al., 2012a)
(Gonzalez-Velasquez, Reed et al. 2011) | Cultured cerebral microvascular endothelial monolayers exposed to Aβ1–40 in soluble, monomeric, or fibrillar states. | Aggregate-state and concentration comparisons extending to physiologically relevant low concentrations, with NF-κB blockade; exact range and inhibitor dose: NR. | inhibitor/antagonist vs vehicle/untreated; Aβ aggregation-state comparisons; dose/time comparisons. | In vitro | Soluble Aβ aggregates, more than monomeric or fibrillar forms, activated endothelial NF-κB and increased barrier permeability, leukocyte adhesion, and transmigration. | Monoculture lacks pericyte, astrocyte, flow, and chronic-aging influences. | (Gonzalez-Velasquez et al., 2011)
(Kauppinen, Suh et al. 2011) | hAPP-J20 mice crossed with Parp1-null mice, assessed at six months; intracerebral Aβ1–42 injection and primary microglia/neuronal cocultures. | Germline PARP1 loss and Aβ challenge; injected peptide and culture concentrations: NR. | Parp1-null vs wild-type hAPP-J20 mice and microglia; Aβ-challenged vs control cultures. | Rodent in vitro; Rodent in vivo | PARP1 deficiency reduced Aβ-induced microglial activation, neuronal injury, and cognitive impairment without reducing microglial Aβ uptake. | Germline deletion affects multiple cells and developmental processes. | (Kauppinen et al., 2011a)
(Blanco, Álvarez et al. 2010) | Primary human astrocytes and human astrocytoma cell lines exposed to Aβ42; conditioned medium was applied to neurons. | Aβ42, IκB overexpression, p65 overexpression, PDTC, and COX−2-promoter deletion constructs; exact doses and durations: NR. | Aβ-exposed astrocytes with IκB/p65/promoter manipulations vs matched controls; conditioned medium vs control medium. | Human in vitro | Aβ activated NF-κB-dependent COX−2 expression and prostaglandin E2 release in human astrocytes. Conditioned medium from these astrocytes was toxic to neurons. | Acute Aβ and tumor lines may exaggerate reactive signaling. | (Blanco et al., 2010a)
(Cui, Li et al. 2010) | Sixty-six control and AD brain samples plus primary human astroglia subjected to IL-1β, Aβ42, or oxidative stress. | Stress exposure, miR−146a promoter constructs, curcumin, PDTC, CAY10512, and anti-miR−146a; concentrations and durations: NR. | AD vs control brain; stressed vs unstressed astroglia; anti-miR/inhibitor conditions vs controls. | Human ex vivo/biospecimen; Human in vitro | NF-κB-induced miR−146a reduced IRAK1 and increased IRAK2 in stressed astroglia and AD brain tissue, helping maintain inflammatory signaling after the initial stress. | Mixed postmortem tissue and multiple pharmacological agents complicate cell-specific causality. | (Cui et al., 2010)
(Zhang, Luhrs et al. 2009) | APPswe/PS1dE9 mice and wild-type littermates, treated from seven to twelve months of age. | Pyrrolidine dithiocarbamate (PDTC), 50 mg/kg intraperitoneally once daily for five months. | transgenic/modified vs wild-type; inhibitor/antagonist vs vehicle/untreated. | Rodent in vivo | Long-term PDTC treatment reduced astrogliosis and inflammatory mediators but increased cerebral Aβ42, indicating a trade-off between inflammation and amyloid handling. | More Aβ42—yet PDTC is nonspecific and systemic treatment obscures the responsible cell. | (Zhang et al., 2009)
(Kawamoto, Lepsch et al. 2008) | Primary rat cerebellar mixed-cell cultures. | Aβ1–40 at 1 or 2 µM for 6, 12, or 24 h; receptor/pathway inhibitors were added 20 min before Aβ. The clearest activation occurred with 1 µM for 12 h. | Aβ dose/time conditions with receptor/pathway inhibitors vs untreated or vehicle controls. | Rodent in vitro | Aβ activated both p50/p65 and p50/p50 dimers through an NMDA-receptor-sensitive pathway. The study mapped the signaling route but did not determine whether it protected or injured cells. | Mixed cerebellar neonatal cultures are remote from aged hippocampal AD and outcome direction was not established. | (Kawamoto et al., 2008a)
(Lukiw, Zhao et al. 2008) | postmortem AD and control brain plus primary human neural cells stressed with IL-1β, Aβ42, or oxidative injury. | miR−146a promoter manipulation, PDTC, CAY10512, and anti-miR−146a; stressor and inhibitor doses and durations: NR. | AD vs control brain; stressed vs unstressed cells; promoter, inhibitor, and anti-miR rescue conditions. | Human ex vivo/biospecimen; Human in vitro | NF-κB-sensitive miR−146a was increased in AD brain and stressed human neural cells and reduced complement factor H, an inhibitor of complement-mediated inflammation. | Postmortem association and acute stressed cultures do not define its net role over disease time. | (Lukiw et al., 2008)
(Tan, Schedl et al. 2008) | Drosophila expressing human Aβ42 in the eye/neuronal tissue. | Genetic loss- and gain-of-function across Toll, NF-κB, and downstream innate-immune components; driver strength and developmental timing: NR. | loss-of-function vs matched control; gain-of-function vs vector/baseline. | Invertebrate in vivo | Reducing Toll-NF-κB signaling suppressed Aβ42-induced neurodegeneration in flies, whereas increasing the pathway worsened the phenotype. | The fly Toll/Relish system and developmental eye phenotype cannot model mammalian adaptive immunity, microglia, or cognition. | (Tan et al., 2008a)
(Ait-Ghezala, Volmar et al. 2007) | HEK293 cells coexpressing Swedish-mutant APP and CD40. | CD40 ligand, NF-κB inhibitors, and pathway-targeting siRNAs; concentrations and incubation times: NR. | CD40 ligand with NF-κB inhibitors or siRNAs vs untreated/control-transfected cells. | Human in vitro | CD40 ligand increased APP metabolites and Aβ production. NF-κB inhibitors and pathway-targeting siRNAs reduced this response. | Receptor and APP overexpression in non-neural cells may create supraphysiological signaling. | (Ait-Ghezala et al., 2007a)
(Bourne, Ferrari et al. 2007) | Neuronal cultures and resting or activated astrocytes carrying BACE1-promoter reporters. | Aβ exposure, cell activation, and mutation of the BACE1 κB element; peptide concentration, activation stimulus, and timing: NR. | Resting vs Aβ-exposed neurons and resting vs activated astrocytes; wild-type vs mutated κB promoter. | In vitro | The same BACE1 κB element suppressed transcription in resting neurons but activated transcription in Aβ-exposed neurons and activated astrocytes. | Reporter systems do not model chromatin topology or whole-brain APP metabolism. | (Bourne et al., 2007a)
(Paris, Patel et al. 2007) | CHO cells overexpressing APP. | A panel of NF-κB inhibitors; compound identities, concentrations, and exposure durations were incompletely reported. | inhibitor/antagonist vs vehicle/untreated. | Hamster-derived in vitro | Several NF-κB inhibitors reduced Aβ40 and Aβ42 production in APP-overexpressing cells, although the specific pathway responsible was not established. | Non-neural cells and inhibitor off-target effects prevent precise causal assignment. | (Paris et al., 2007a)
(Schapansky, Olson et al. 2007) | Neurons expressing normal or AD-linked mutant presenilin−1. | Aβ and IP3-linked endoplasmic-reticulum Ca2 + release with NF-κB inhibition or activation; concentrations and durations: NR. | Wild-type vs AD-linked presenilin−1 neurons; NF-κB activation/inhibition conditions. | In vitro | Endoplasmic-reticulum calcium release activated NF-κB, reduced the pro-apoptotic protein CHOP, and protected neurons from Aβ toxicity, especially in cells expressing mutant presenilin 1. | The familial-PS1 acute-culture context may not persist in chronic sporadic disease. | (Schapansky et al., 2007a)
(Choi, Kim et al. 2006) | M3-muscarinic-receptor-expressing SH-SY5Y neuronal cells. | Oxotremorine-M with SN50, BAY11–7085, MG132, PDTC, quinazoline, TMB8, gadolinium, or SKF96365; exact concentrations and durations: NR. | Oxotremorine-M-treated vs untreated cells with calcium-entry/NF-κB inhibitors. | Human in vitro | Muscarinic receptor-driven calcium entry activated NF-κB and increased soluble APPα release, favoring non-amyloidogenic APP processing. | The large pharmacological panel has overlapping off-target effects and lacks an in-vivo amyloid endpoint. | (Choi et al., 2006a)
(Chen, Zhou et al. 2005) | Primary microglia-neuron cultures exposed to Aβ. | Nondegradable IκBα super-repressor, SIRT1 overexpression, and resveratrol; Aβ and resveratrol doses and treatment durations: NR. | Aβ-exposed microglia-neuron cocultures with IκBα super-repressor, SIRT1 overexpression, or resveratrol vs controls. | Rodent in vitro | Blocking microglial NF-κB with a nondegradable IκBα construct reduced Aβ-induced neuronal death. SIRT1 overexpression or resveratrol produced similar protection by reducing p65 acetylation. | Acute coculture lacks chronic plaque and aging states. | (Chen et al., 2005a)
(Chong, Li et al. 2005) | Primary hippocampal neurons exposed to Aβ. | Erythropoietin across concentration conditions, with p65 RNA interference and tests of nuclear translocation; exact concentrations and timings: NR. | Erythropoietin-treated vs untreated Aβ-exposed neurons; p65 RNA interference or blocked translocation vs control. | Rodent in vitro | Erythropoietin protected primary hippocampal neurons from early and late Aβ-induced apoptosis. p65 knockdown or prevention of nuclear translocation abolished the protection. | A short-term pure-neuron system cannot show whether chronic glial activation offsets that benefit. | (Chong et al., 2005)
(Huang, Liu et al. 2005) | postmortem AD and control brain tissue. | No exposure; lectin-based enrichment, mass spectrometry, and immunoblot validation. | AD/disease vs control tissue. | Human ex vivo/biospecimen | The NF-κB precursor p105 and inhibitor IκBγ were both increased in AD brain tissue, indicating altered balance and feedback within the pathway. | Sample size and cellular source were NR in the abstract and direction cannot be inferred. | (Huang et al., 2005a)
(Pizzi, Sarnico et al. 2005) | Primary cortical neurons and neuronal cell lines challenged with Aβ. | mGlu5 agonists, c-Rel RNA interference or overexpression, and TAT-Bcl-xL rescue; concentrations and exposure durations: NR. | c-Rel knockdown vs overexpression and matched controls; TAT-Bcl-xL rescue vs no rescue. | Rodent in vitro | c-Rel was necessary and sufficient to induce MnSOD and Bcl-xL and protect neurons from Aβ toxicity. Providing Bcl-xL downstream restored survival when c-Rel signaling was impaired. | Protection was tested only in acute neuronal systems. | (Pizzi et al., 2005)
(Samuelsson, Fisher et al. 2005) | Mixed primary rat glial cultures. | Aβ25–35 and IL-1β alone or together over a concentration range; NF-κB DNA binding was measured at 30 min, 2 h, and 24 h. Exact concentrations: NR. | Aβ25–35 and interleukin−1β alone vs combined exposure and untreated controls across time. | Rodent in vitro | Aβ and interleukin−1β acted together to produce stronger and more persistent NF-κB DNA binding in primary glia than either stimulus alone. | Mixed glia prevent assignment to astrocytes versus microglia and no neuronal or disease outcome was measured. | (Samuelsson et al., 2005)
(Kim, Kim et al. 2003) | Human neuroblastoma cells exposed to Aβ42. | Aβ42 dose-response, PLC-δ1-promoter deletion/mutation, dominant-negative TAK1 or NIK, and NF-κB pathway manipulation; exact doses/times: NR. | Aβ42 dose conditions; wild-type vs deleted/mutated PLC-δ1 κB promoter; dominant-negative kinase conditions. | Human in vitro | Aβ42 activated the PLC-δ1 promoter through a functional κB site. Deleting or mutating this site prevented promoter activation. | The downstream relevance of PLC-δ1 induction to AD injury was not directly tested. | (Kim et al., 2003a)
(Rodriguez-Kern, Gegelashvili et al. 2003) | Differentiated primary rat astrocytes, with neuronal/glutamate-homeostasis relevance. | Subtoxic Aβ1–42 at 1–5 µM or BDNF; exposure duration and inhibitor doses: NR. | Aβ1–42 or BDNF exposure with pathway inhibitors vs untreated controls. | Rodent in vitro | Aβ and brain-derived neurotrophic factor increased astrocytic GLT−1/EAAT2 expression and glutamate uptake through NF-κB-dependent pathways. | Micromolar acute peptide and isolated astrocytes do not establish net protection in an amyloid-bearing brain. | (Rodriguez-Kern et al., 2003)
(Hattori, Arai et al. 2001) | postmortem AD and control brain tissue containing neurofibrillary pathology. | No exposure; Southwestern histochemistry and immunohistochemistry. | AD/disease vs control tissue. | Human ex vivo/biospecimen | Activated NF-κB and increased IκB were found together in neurofibrillary lesions, suggesting pathway activation accompanied by an attempted inhibitory feedback response. | Sample size was NR and static localization cannot identify cause or productive transcription. | (Hattori et al., 2001)
(Tomita, Fujita et al. 2000) | Neuronal cell systems expressing APP, RelA/p65, p50-containing dimers, and the neuron-specific adaptor X11L. | p65 or p50/dimer expression and X11L gain/loss or PDZ-domain disruption; transfection levels and durations: NR. | p65/p50 expression with or without X11L or PDZ-domain mutants vs matched controls. | In vitro | RelA/p65 selectively increased Aβ42 production. The neuronal adaptor X11L reduced this effect by binding p65 through its PDZ domain. | Subunit and binding-domain specificity support a direct mechanism, but overexpression may distort stoichiometry and no aged in-vivo validation was provided. | (Tomita et al., 2000a)
(Kaltschmidt, Uherek et al. 1999) | Primary neuronal cultures. | Preconditioning with 0.1 µM Aβ1–40 then 10 µM Aβ1–40 challenge; TNF-α produced maximal NF-κB activation at 2 ng/mL; dominant-negative IκBα tested pathway dependence. | 0.1 µM Aβ preconditioning followed by 10 µM challenge vs non-preconditioned controls; dominant-negative IκBα conditions. | Rodent in vitro | Pre-exposure to low-dose Aβ1–40 protected neurons from a later high-dose challenge. Blocking NF-κB with dominant-negative IκBα removed this acquired resistance. | The 100-fold contrast establishes acute preconditioning but not chronic disease modification. | (Kaltschmidt et al., 1999a)
(Bales, Du et al. 1998) | Primary fetal-rat cortical neurons and primary rat astroglia exposed to Aβ. | Concentration- and time-dependent Aβ exposure with IκBα antisense manipulation; exact peptide concentrations: NR. | Aβ-exposed vs untreated neuronal and astroglial cultures; IκBα antisense vs control. | Rodent in vitro | During the same Aβ exposure, constitutive neuronal NF-κB declined as neuronal DNA fragmentation began, whereas astroglial NF-κB and inflammatory cytokines increased. | Fetal cells and acute peptide exposure differ substantially from aged multicellular AD. | (Bales et al., 1998)
(Guo, Robinson et al. 1998) | Differentiated PC12 cells expressing AD-linked PS1 L286V. | Secreted APPα or Aβ/stress challenges, κB-decoy blockade, oxidative stress, thapsigargin, and calcium perturbation; concentrations and durations: NR. | sAPPα rescue vs control in mutant-presenilin cells; κB DNA-decoy vs control decoy. | Rodent in vitro | Soluble APPα required NF-κB signaling to stabilize intracellular calcium and mitochondrial function and protect presenilin−1-mutant cells from apoptosis. | ΚB blockade supports mediation of sAPPα rescue, but PC12 cells and mutant-PS1 overexpression represent an acute familial-stress model. | (Guo et al., 1998a)
(Lukiw and Bazan 1998) | postmortem superior temporal neocortex from control and CERAD-classified sporadic AD cases; nuclear extracts from 20 cases, ages 60–82 years, postmortem interval 0.5–6.5 h. | No treatment; biochemical tissue comparison plus COX−2-promoter deletion constructs. | AD severity groups vs controls; intact vs deleted κB element in the COX−2 promoter. | Human ex vivo/biospecimen | NF-κB DNA binding was associated with κB-dependent COX−2 transcription in aging and sporadic AD neocortex. | Dimer-binding and promoter evidence strengthen biological plausibility, but cross-sectional bulk tissue remains vulnerable to cell-composition and postmortem effects. | (Lukiw and Bazan, 1998a)
(Hirsch, Agid et al. 1997) | Nucleus basalis of Meynert from four patients with AD and four controls, focusing on cholinergic neurons. | No treatment; cell-resolved postmortem immunohistochemistry. | AD vs control postmortem tissue. | Human observational/genetic; Human ex vivo/biospecimen | Nuclear NF-κB was more frequent in basal-forebrain cholinergic neurons from AD brains than controls. | Selective-vulnerability localization is valuable, but n = 4 per group and cross-sectional survivor bias preclude assigning activation as protective or pathogenic. | (Hirsch et al., 1997)
(Bonaiuto, McDonald et al. 1997) | N9 mouse microglia, primary rat microglia, and human monocytes. | Suboptimal Aβ25–35 combined with interferon-γ at 100 U/mL for at least 120 min; higher Aβ alone also activated cells. Exact suboptimal/high peptide concentrations: NR. | Aβ and interferon-γ alone vs combined exposure and untreated controls. | Human ex vivo/biospecimen; Rodent in vitro | Aβ and interferon-γ cooperated to activate RelA/p50 DNA binding in microglia and monocytes. Higher Aβ exposure could activate the pathway without cytokine costimulation. | Dimer and cross-species confirmation support inflammatory amplification, but transformed N9 cells and acute IFN-γ costimulation may exaggerate plaque-associated signaling. | (Bonaiuto et al., 1997a)
(Kaltschmidt, Uherek et al. 1997) | Primary neurons exposed to Aβ1–40 or Aβ25–35 and postmortem human neurons surrounding early plaques. | Maximal activation with 0.1 µM Aβ1–40 or 0.1 µM Aβ25–35; antioxidant blockade tested reactive-oxygen-intermediate dependence. | Aβ dose conditions with antioxidant or NF-κB inhibition vs controls; human plaque-adjacent vs control neurons. | Rodent in vitro; Human ex vivo | Low-concentration Aβ activated neuronal NF-κB through reactive oxygen intermediates. Activated p65 was also detected near early plaques in human brain tissue, but the functional outcome was not determined. | The experiment did not show whether activation itself improved or worsened neuronal outcome. | (Kaltschmidt et al., 1997a)
(Kitamura, Shimohama et al. 1997) | postmortem temporal cortex from AD and neurologically normal controls. | No treatment; subcellular fractionation and antibody-based protein analysis. | AD/disease vs control tissue. | Human ex vivo/biospecimen | AD temporal cortex showed redistribution and enrichment of p65 and STAT1 in particulate and nuclear fractions, consistent with an inflammatory transcriptional state. | Subcellular redistribution supports pathway engagement, but sample size and postmortem details were NR in the abstract and causality is unresolved. | (Kitamura et al., 1997a)
(Terai, Matsuo et al. 1996) | postmortem hippocampal formation, entorhinal and temporal cortex, and visual cortex from AD and control brains. | No treatment; regional p65 immunohistochemistry. | AD/disease vs control tissue. | Human ex vivo/biospecimen | p65 immunoreactivity was enriched in vulnerable hippocampal and cortical regions, neurofibrillary tangles, and dystrophic neurites in AD brain tissue. | Sample size was NR in the abstract and immunoreactivity cannot establish activation or direction. | (Terai et al., 1996a)