Astrocytic NF-κB Signaling: Compensatory Functions and Neurotoxic Reactivity

Astrocytic NF-κB signaling illustrates particularly well why this pathway cannot be classified as uniformly protective or harmful in AD. Astrocytes regulate extracellular glutamate, participate in Aβ handling and plaque organization, communicate with microglia, contribute to extracellular protein degradation, and release mediators that influence neuronal function. NF-κB affects these processes differently. In some settings, astrocytic NF-κB appears to support compensatory responses to established pathology; in others, it suppresses Aβ degradation or drives inflammatory and neurotoxic signaling. Consequently, changes in plaque burden, astrogliosis, or inflammatory markers alone are insufficient to determine whether an astrocytic response is beneficial.

Early studies already suggested that reducing astrocytic or glial inflammation does not necessarily improve amyloid pathology. In APPswe/PS1dE9 mice, systemic treatment with the NF-κB inhibitor pyrrolidine dithiocarbamate reduced cyclooxygenase−2 (COX−2), tumor necrosis factor-α (TNF-α), and astrogliosis but increased cerebral Aβ42 (Zhang et al., 2009). Although the intervention was not astrocyte specific and PDTC has additional antioxidant and metal-chelating effects, the finding suggested that glial inflammatory responses may also contain compensatory functions relevant to Aβ handling.

More direct evidence for a compensatory astrocytic response came from selective activation of IKK2/NF-κB in astrocytes of APP23 mice. As expected, astrocytic pathway activation produced marked astrogliosis and induced prominent microglial reactivity. Surprisingly, however, both the number and size of Congo-red-positive amyloid plaques decreased in the cortex and hippocampus. Nearby microglia also acquired a transcriptional profile compatible with increased plaque-associated Aβ handling, whereas direct IKK2/NF-κB activation in microglia produced a more conventionally inflammatory phenotype (Yang et al., 2021b). These findings suggest that astrocytic NF-κB can modify amyloid pathology indirectly by changing astrocyte–microglia communication and possibly neuronal APP processing.

The reduction in compact plaques, however, should not be equated with global neuroprotection. Congo-red staining predominantly detects fibrillar amyloid deposits and does not describe the abundance of soluble oligomeric Aβ species, which can exert substantial synaptic toxicity. Nor did the study demonstrate comparable improvements across neuronal survival, synaptic physiology, and cognition. The supported interpretation is therefore narrower: astrocytic NF-κB can organize a plaque-associated glial response that alters deposited amyloid, but the experiment does not establish that chronic pathway activation improves overall brain function.

A later study provided particularly useful evidence that the effect of astrocytic NF-κB depends on the disease context. Huat et al. combined human AD transcriptomic and postmortem analyses with selective astrocytic NF-κB manipulation in mice. In otherwise healthy mice, sustained activation of astrocytic NF-κB disrupted the brain proteomic environment, reducing mitochondrial-associated proteins while increasing inflammatory proteins. Chronic activation also induced microglial reactivity and was associated with accumulation of the senescence-related protein p16INK4A in neurons. Thus, persistent astrocytic NF-κB activation imposed on an otherwise healthy brain was clearly not homeostatic (Jong Huat et al., 2024). The result was different when NF-κB was inhibited in an established AD model. In aged 3xTgAD mice, the investigators overexpressed the NF-κB inhibitory protein A20 predominantly in hippocampal GFAP-positive astrocytes. Astrocytic p65 activation decreased, yet Aβ plaques, soluble and insoluble Aβ, and phosphorylated tau increased. APP and its cleavage products were not altered, arguing against increased Aβ production as the explanation. Astrocytic NF-κB inhibition also reduced aquaporin−4 (AQP4), a water-channel protein important for astrocyte polarization and glymphatic solute handling (Jong Huat et al., 2024). These findings suggest that once AD-like pathology is established, at least some NF-κB-dependent astrocytic responses contribute to the handling or containment of pathological proteins.

These two results are not necessarily contradictory. Sustained NF-κB activity can disrupt a previously healthy astrocyte, while disease-associated astrocytic NF-κB may simultaneously induce compensatory functions that become useful once Aβ and tau are already present. Importantly, the study did not follow the same astrocytic manipulation continuously from a healthy state through successive stages of AD. It therefore demonstrates context dependence more convincingly than a defined temporal switch from “protective” to “harmful.”

A distinct mechanism involving extracellular Aβ degradation further demonstrates why the term “amyloid clearance” is too broad. Sudo et al. investigated kallikrein-related peptidase 7 (KLK7), a secreted protease capable of degrading extracellular Aβ. In cultured cells, 24-hour treatment with the NF-κB inhibitors IKK−16 at 2 µM or JSH−23 at 30 µM increased KLK7 expression and enhanced degradation of extracellular Aβ40. Predominantly astrocytic primary mouse glial cultures showed the same direction of effect. Importantly, the reduction in extracellular Aβ could not be explained by increased cellular uptake or loss of cell viability, supporting increased proteolytic degradation as the relevant mechanism. The investigators then tested this mechanism in vivo. IKK−16 was infused directly into one hippocampus at 4 µL of 100 µM solution, delivered at 0.3 µL/min, while the contralateral hippocampus received vehicle. Klk7 expression increased after 24 h in wild-type mice. In 14–16-month-old AppNL-G-F knock-in mice, 48 h of local NF-κB inhibition reduced soluble and SDS-insoluble Aβ and decreased plaque size, although plaque number did not change. Most importantly, the Aβ-lowering effect disappeared in AppNL-G-F mice lacking Klk7. This knockout experiment strongly supports the sequence NF-κB inhibition → KLK7 upregulation → enhanced extracellular Aβ degradation, rather than KLK7 merely changing in parallel with Aβ (Sudo et al., 2026).

The Sudo et al., study (Sudo et al., 2026) appears at first to conflict with the APP23 experiment, in which astrocytic NF-κB activation reduced compact plaques. In fact, the two studies examine different components of Aβ biology. Astrocytic NF-κB may facilitate glial organization around established plaques while simultaneously suppressing KLK7-dependent proteolysis of extracellular Aβ. These functions can occur in parallel. A smaller plaque after one intervention and greater extracellular Aβ degradation after another therefore do not imply that NF-κB has a single bidirectional “clearance switch.” Rather, they show that plaque organization, cellular uptake, and extracellular proteolysis are mechanistically distinct processes.

The therapeutic implications of the KLK7 mechanism also require caution. IKK−16 was delivered locally to the hippocampus rather than selectively to astrocytes, and KLK7 is not expressed exclusively by astrocytes. The in-vivo observation period was only 24–48 h, which is insufficient to determine chronic effects on cognition, synaptic function, host defence, or other physiological NF-κB programs. The experiment therefore provides strong mechanistic evidence that NF-κB can restrain KLK7-dependent Aβ degradation, but it does not establish that sustained global NF-κB inhibition would be beneficial in AD.

Evidence for detrimental astrocytic NF-κB is stronger when pathway activation is linked to defined secreted mediators and subsequent neuronal injury. Blanco et al. showed that Aβ1–42 activated NF-κB-dependent COX−2 transcription in primary human astrocytes and astrocytoma cells, increasing prostaglandin E2 (PGE2) production (Blanco et al., 2010b). Deletion of κB-responsive promoter elements reduced COX−2 transcription, providing direct evidence of NF-κB-dependent regulation. Conditioned medium from Aβ-activated astrocytes was toxic to neuronal cells, supporting a pathway in which Aβ-induced astrocytic NF-κB activity contributes to a harmful secretory response.

The functional consequence was demonstrated using astrocyte-conditioned medium. Medium collected from Aβ-activated astrocytes transferred toxicity to neuronal cells, indicating that the harmful effect did not require direct astrocyte–neuron contact. The resulting sequence—Aβ → astrocytic NF-κB → COX−2/PGE2-associated secretion → neuronal injury—provides substantially stronger evidence for a detrimental astrocytic program than the observation of reactive astrocyte markers alone. The use of astrocytoma-derived cells and acute Aβ exposure limits direct extrapolation to aged human astrocytes, but the promoter manipulation and conditioned-medium experiment provide a coherent mechanism.

An even more completely resolved astrocyte-to-neuron pathway involves complement component 3 (C3). Lian et al. demonstrated that Aβ activates astrocytic NF-κB and increases C3 production and release. Astrocyte-derived C3 then acts on neuronal C3a receptors, altering intracellular calcium, dendritic structure, AMPA-receptor trafficking, and neuronal network function (Lian et al., 2015b). Importantly, C3aR antagonism reduced these downstream abnormalities and rescued cognitive impairment in APP-transgenic mice. This rescue experiment provides strong evidence for an astrocytic NF-κB → C3 → neuronal C3aR pathway contributing to synaptic and cognitive dysfunction.

Cell-specific genetic studies support the broader principle that astrocytic NF-κB can amplify neuronal proteotoxicity. In Drosophila models, neuronal expression of proteotoxic proteins activated the NF-κB homolog Relish in astrocyte-like glia. Selective reduction of Relish in these glial cells decreased inflammatory gene expression, delayed neurodegeneration, prolonged survival, and also extended lifespan in flies expressing neuronal human Aβ42 (Li et al., 2018). Although fly glia differ from mammalian astrocytes, the cell-restricted genetic design demonstrates a non-cell-autonomous loop in which neuronal proteotoxic stress activates glial NF-κB, which then contributes to further neuronal damage.

Genetic and metabolic background can also determine how readily astrocytes enter an inflammatory NF-κB state. In isogenic human iPSC-derived astrocytes, APOE4 reduced TAGLN3 expression and produced low-grade inflammation with exaggerated NF-κB responses to subsequent stimulation. TAGLN3 supplementation attenuated this phenotype, and interaction between TAGLN3 and IκBα supported a role for TAGLN3 in restraining NF-κB activity (Arnaud et al., 2022b). These findings suggest that APOE4 alters the baseline regulatory state of astrocytes, making them more susceptible to excessive inflammatory signaling.

Altered lipid handling may create a similar predisposition. FABP7 was increased in Aβ-exposed astrocytes, plaque-associated astrocytes in APP/PS1 mice, and human AD tissue. In human iPSC-derived astrocytes, FABP7 overexpression induced a broad inflammatory transcriptional program and activated NF-κB, whereas a ligand-binding-deficient FABP7 mutant failed to reproduce the same response (Hamilton et al., 2024a). This indicates that FABP7-mediated lipid binding is required for the inflammatory phenotype and links astrocytic lipid metabolism to NF-κB-dependent reprogramming. Together, the APOE4 and FABP7 studies suggest that astrocytes exposed to similar pathological stimuli may respond differently according to their genetic and metabolic state.

Among the strongest mechanistic evidence for detrimental astrocytic NF-κB signaling comes from studies of histone deacetylase 7 (HDAC7). Ye et al. found that HDAC7 was selectively increased in astrocytes near amyloid plaques in APP/PS1 mice (Ye et al., 2026). Mechanistically, Aβ-induced HDAC7 interacted with and deacetylated IKKα and IKKβ, promoting IKK activation, NF-κB nuclear translocation, and expression of a neurotoxic astrocytic program. Astrocyte-specific HDAC7 overexpression was sufficient to induce neurotoxic transcription, neuronal loss, and cognitive impairment, including in wild-type and relatively young disease-model mice. Conditioned-medium experiments further showed that the neuronal toxicity was mediated by astrocyte-secreted factors rather than requiring direct cell contact. The reverse manipulations strengthened the causal interpretation. Astrocyte-specific HDAC7 knockdown reduced the neurotoxic phenotype, while IKK inhibition prevented the effects of HDAC7 overexpression, placing IKK–NF-κB downstream of HDAC7. Pharmacological inhibition with TMP195 likewise attenuated IKK–NF-κB signaling, reduced neurotoxic reactive astrocytes, limited neuronal and synaptic loss, and improved behavioral deficits in APP/PS1 mice. The resulting sequence—Aβ → astrocytic HDAC7 → IKKα/IKKβ regulation → NF-κB activation → neurotoxic secretome → synaptic and neuronal loss → cognitive impairment—is among the most complete mechanistic demonstrations of detrimental astrocytic NF-κB signaling in the current AD literature (Ye et al., 2026).

Taken together, these studies show that the apparent dual role of astrocytic NF-κB does not arise because the same biological process inexplicably becomes both beneficial and harmful. Rather, NF-κB controls different astrocytic operations, some of which may be compensatory while others are clearly pathogenic. Astrocytic NF-κB can influence glial organization around established amyloid deposits and may support disease-associated functions such as AQP4 expression. At the same time, it can suppress KLK7-dependent extracellular Aβ degradation and activate COX−2/PGE2, complement C3, and HDAC7-dependent secretory programs that adversely affect neurons. APOE4-dependent TAGLN3 loss and FABP7-dependent lipid signaling further determine how strongly astrocytes enter these inflammatory states.

This distinction also explains why apparently conflicting amyloid findings should not be interpreted from plaque burden alone. A reduction in plaque size could reflect altered glial organization or plaque remodeling, whereas a reduction in soluble and insoluble Aβ after NF-κB inhibition may result from increased extracellular KLK7-mediated proteolysis. Conversely, suppression of astrocytic NF-κB during established disease can remove compensatory processes and accelerate Aβ and tau accumulation. Plaque compaction, extracellular Aβ degradation, soluble Aβ abundance, glial coordination, and neuronal toxicity are distinct outcomes and can move in different directions after the same pathway is manipulated.

The available evidence consequently supports a more selective therapeutic interpretation. Astrocytic NF-κB should not be globally suppressed, because some NF-κB-dependent responses may help the diseased brain manage existing pathology. However, defined neurotoxic outputs have increasingly strong causal support, particularly the COX−2/PGE2, C3/C3aR, and HDAC7–IKK–NF-κB pathways. Therapeutic strategies should therefore aim to disrupt these pathological astrocytic programs while preserving homeostatic and compensatory functions required for protein handling, neurovascular support, and neuronal maintenance.