Introduction
Alzheimer’s disease (AD) is an aging-related neurodegenerative disorder characterized clinically by progressive memory impairment, cognitive decline, and impairment in activities of daily living.1 The incidence of AD increases with advancing age. It is estimated that by 2050, the global number of people living with dementia will reach 152 million, among whom 60–70% will have AD.2 AD has an insidious onset and a continuously progressive course, profoundly compromising patients’ quality of life and imposing a substantial economic burden on families and society. Therefore, AD has become a major challenge in global public health.
The classical pathological hallmarks of AD include senile plaques formed by the abnormal deposition of amyloid-β (Aβ), neurofibrillary tangles resulting from abnormal hyperphosphorylation of Tau protein, and the consequent synaptic dysfunction, neuronal injury, and cognitive decline.3 The Aβ cascade hypothesis and the Tau pathology hypothesis constitute important pathological frameworks for understanding the onset and progression of AD. However, explaining the pathological progression of AD solely from these perspectives remains insufficient. Increasing evidence indicates that neuroinflammation, disruption of cerebral immune homeostasis, metabolic dysfunction, white matter injury, and abnormalities of the neurovascular unit also contribute to the development and progression of AD.4 These findings suggest that AD is not driven by a single cell type or pathological pathway, but rather arises from dysregulation of complex interactions among multiple brain cell populations.
Glial cells are major non-neuronal cell populations in the central nervous system and mainly include microglia, astrocytes, and oligodendrocytes. These cells play essential roles in maintaining brain homeostasis, regulating synaptic function, supporting neuronal metabolism, participating in myelination, and mediating immune and inflammatory responses.5 Recent studies have revealed marked stage-dependent and functional heterogeneity in microglial and astrocytic responses to AD pathology. Depending on disease stage and local microenvironment, these cells may contribute to Aβ clearance and tissue homeostasis, but under sustained pathological stimulation they may also promote neuroinflammation, synaptic dysfunction, and neuronal injury.6,7 In parallel, oligodendrocyte dysfunction and myelin abnormalities are increasingly recognized as active contributors to AD pathophysiology, although they remain less extensively characterized than microglial and astrocytic alterations.8 Importantly, different glial cell populations do not act in isolation.
Instead, they communicate through complex intercellular signaling networks, referred to as glial crosstalk, thereby jointly modulating Aβ pathology, neuroinflammation, and neuronal damage. Recent studies of peri-plaque glial networks have provided direct evidence supporting such coordinated interactions.9 Accordingly, a shift from studying individual glial cell types in isolation toward an integrated framework encompassing interactions among glial cells, neurons, and vascular cells may provide a more comprehensive understanding of AD pathogenesis.
In recent years, advances in single-cell sequencing,10 spatial transcriptomics,11 molecular imaging,12 and biofluid biomarker detection have drawn increasing attention to the active regulatory roles of glial cells in AD.13 Glial cells are now considered key hubs that connect multiple pathological processes in AD, including proteinopathy, metabolic imbalance, inflammatory responses, white matter injury, and neurovascular dysfunction.14 Nevertheless, several important knowledge gaps remain. The temporal sequence and causal relationships underlying transitions among distinct glial states have not been fully defined, the precise contribution of oligodendrocytes to AD remains incompletely understood, and the mechanisms by which different glial populations regulate one another require further clarification. Moreover, much of the existing literature has focused on individual glial cell types, whereas relatively few studies have systematically integrated microglia, astrocytes, oligodendrocytes, and their intercellular interactions within a unified framework.
In addition, glia-related biomarkers provide an important bridge between mechanistic research and clinical translation. Whereas conventional Aβ- and tau-related biomarkers primarily reflect the core proteinopathological features of AD, glia-related biomarkers may provide complementary information on neuroimmune activation, astrocytic responses, myelin injury, and other non-neuronal pathological processes.
Against this background, this review summarizes the functional alterations and intercellular interactions of microglia, astrocytes, and oligodendrocytes in AD and discusses recent advances in glia-related fluid and imaging biomarkers as well as their potential clinical applications. By integrating multiple glial populations, glial crosstalk, and biomarker evidence within a neuron–glia–vascular network framework, this review aims to provide a more comprehensive understanding of AD pathogenesis and to offer insights into early diagnosis and precision therapeutic strategies.
Roles of Microglia in AD
Microglia are the principal resident innate immune cells of the central nervous system. They can recognize and respond to Aβ deposition and Tau-related pathological stimuli, undergo activation and disease-associated phenotypic transitions, and participate in multiple pathological processes, including neuroinflammation, complement-dependent synaptic engulfment, and immunometabolic reprogramming.
Microglial Recognition of and Responses to Aβ and Tau Pathology
In AD, microglia sense Aβ deposits through multiple receptors, including TREM2, Axl, Mertk, Piezo1, Scara1, TMEM119, and a cell-surface receptor complex composed of CD36, CD47, and α6β1-integrin. TREM2 is predominantly expressed in microglia and exhibits high affinity for Aβ oligomers and fibrils. It promotes microglial migration and clustering around amyloid plaques and enhances microglial phagocytosis of Aβ, thereby reducing the toxic effects of Aβ on surrounding neurites and synapses.15 The TAM receptor family members Axl and Mertk can recognize phosphatidylserine co-deposited with amyloid plaques, whereas Piezo1, a mechanosensation-related molecule, senses plaque hardness and stiffness and participates in Aβ phagocytosis.16 Microglia can also recognize and bind Aβ through a cell-surface receptor complex consisting of CD36, CD47, and α6β1-integrin, thereby mediating inflammatory activation.17 Scavenger receptors are important membrane proteins involved in microglial recognition and uptake of abnormally aggregated proteins.
Scara1, expressed on myeloid cells, can bind soluble and partially aggregated forms of Aβ and promote microglial uptake and clearance of Aβ.18 Together, these mechanisms contribute to microglial localization to, encapsulation of, and phagocytosis of Aβ plaques. It should be noted that the strength of evidence supporting the involvement of different Aβ-recognition receptors is not uniform. For example, the causal evidence that Piezo1 and TMEM119 facilitate Aβ processing is derived primarily from in vitro studies and AD mouse models.
A 2025 study demonstrated that TMEM119 can directly bind Aβ and promote its clearance, although this effect was mainly validated in 5×FAD mice.19 In contrast, recent analyses of brain tissue from patients with AD who had received Aβ immunotherapy showed that regions with greater Aβ clearance were associated with enhanced TREM2/APOE- and complement-related microglial programs, providing more direct evidence for the involvement of microglia in Aβ clearance in human disease.20 Overall, the available evidence supports an important role for microglia in Aβ recognition and processing; however, the relative contribution of individual receptors to sporadic AD in humans remains to be clarified. In AD, Tau can form intracellular aggregates and hyperphosphorylated structures, but it is not completely confined within neurons. Tau can be released into the extracellular space and subsequently taken up by microglia, thereby contributing to Tau pathological progression.
Tau can bind to microglial CX3CR1 and promote tau endocytosis through this receptor.21 In addition, microglia can phagocytose Tau aggregates in vivo and may transfer Tau to neurons through extracellular vesicles. Depletion of microglia or inhibition of extracellular vesicle generation reduces Tau propagation, indicating that microglia not only participate in Tau uptake and processing but may also promote the spread of Tau pathology when processing is incomplete or vesicle release is enhanced.22 Similarly, evidence supporting CX3CR1-mediated tau uptake and microglia-facilitated tau propagation via extracellular vesicles has been derived predominantly from cellular and animal models, whereas direct causal evidence in human AD remains limited. Thus, microglia may exert dual and potentially opposing effects on tau pathology by contributing to both tau clearance and intercellular propagation.
The net effect is likely to depend on the molecular species and phosphorylation state of tau, the degradative capacity of microglia, and the stage of disease.
Microglial Activation and Disease-Associated Phenotypic Transitions
Microglial activation in AD cannot be simply classified into the traditional “M1” and “M2” phenotypes. With advances in single-cell transcriptomics, spatial omics, and epigenomics, microglia have been shown to transition gradually from a homeostatic phenotype into diverse disease-associated states, such as disease-associated microglia (DAM), plaque-associated microglia (PAM), inflammatory microglia, and lipid metabolism–dysregulated microglia.23,24 Aβ plaques, Tau aggregates, neuronal injury, lipid debris, and cell death signals can all serve as triggers that induce microglial phenotypic transformation.25,26
DAM represent one of the most characteristic microglial phenotypes in neurodegenerative diseases. During the transition from homeostasis to disease-associated states, DAM typically show upregulation of genes related to disease response, lipid metabolism, and phagocytosis, including ApoE, Trem2, Itgax, Igf1, Lpl, Cst7, Ctsd, and Tyrobp, along with downregulation of homeostatic microglial markers such as P2ry12/P2ry13, Cx3cr1, and Tmem119.27,28 This phenotypic transition has dual effects. On the one hand, it enhances lipid metabolism, lysosomal degradation, and phagocytic capacity, thereby helping to restrict plaque expansion and clear Aβ and cellular debris.
On the other hand, when the phagocytic burden is excessive or lysosomal function is impaired, it can induce chronic inflammation, reactive oxygen species (ROS) production, and NLRP3 inflammasome activation, thereby weakening Aβ clearance and exacerbating synaptic damage, Tau pathology, and neuronal dysfunction.6,29 Accordingly, rather than defining disease-associated microglia simply as an early “protective” state that later becomes “detrimental”, current evidence supports a more continuous spectrum of microglial states that is highly dependent on disease stage and the local tissue microenvironment. Importantly, the canonical DAM signature was largely derived from mouse models, whereas studies of human microglia have revealed additional Aβ-associated states characterized by prominent HLA-related antigen-presentation programs and inflammatory/chemokine responses that are not fully captured by the murine DAM framework.24 A study of ex vivo primary microglia from 189 human brains further demonstrated that distinct clinical and neuropathological AD phenotypes are associated with different transcriptional modules and disease-related microglial states.30 More recently, a 2026 human spatial transcriptomic study identified a transition from an early Aβ-associated inflammatory program toward a tau-associated antigen-presentation program, while individuals with cognitive resilience exhibited distinct microglial trajectories.31 Collectively, these findings suggest that DAM is better regarded as a conceptual framework encompassing a spectrum of disease-associated microglial states rather than as a single cell type with a fixed functional identity.
Microglia-Mediated Neuroinflammation
Neuroinflammation is a critical pathological link connecting Aβ deposition, Tau pathology, synaptic injury, and cognitive decline. Aβ oligomers or fibrils, phosphorylated Tau, neuronal injury debris, pathogen-associated molecular patterns (PAMPs), and damage-associated molecular patterns (DAMPs) can be recognized by microglial surface receptors such as CD14, TLR2, TLR4, TNFR, IL-1R, P2X7R, and RAGE. This subsequently activates signaling pathways involving MyD88, TRIF, NF-κB, IKK, ERK, and MAPK, thereby inducing the expression of inflammation-related genes such as TNF-α, IL-1β, IL-6, CCL2, CXCL10, iNOS, and COX-2.32–35
Extracellular ATP released during cellular injury or stress can activate P2X7R and promote NLRP3 inflammasome activation through mechanisms such as K+ efflux, thereby linking purinergic signaling to microglial inflammatory responses.36 Aβ can activate the NLRP3 inflammasome in microglia and promote caspase-1 and IL-1β activation. ASC forms speck-like aggregates and is released into the extracellular space, where it propagates inflammasome signaling among neighboring cells and forms an inflammatory amplification loop.37 Meanwhile, leakage of cytosolic DNA or mitochondrial DNA can activate the cGAS–STING pathway. Activation of the STING–TBK1 axis induces type I interferon–related transcriptional programs and the expression of interferon-stimulated genes, such as Ifit, Oasl1/Oasl2, Isg15, and Stat2.38
A moderate inflammatory response facilitates the clearance of pathological proteins and cellular debris. However, persistent stimulation leads microglia to continuously release inflammatory mediators such as IL-1β, IL-6, and TNF-α, forming a chronic neuroinflammatory microenvironment that further promotes the spread of Aβ and Tau pathology, synaptic loss, and neuronal injury. Thus, microglia-mediated neuroinflammation has dual properties: transient activation at early stages may be protective, whereas sustained inflammatory amplification is a major driver of AD deterioration.
Microglia-Mediated Complement-Dependent Synaptic Engulfment
Synaptic dysfunction and synapse loss are major pathological features of AD, and complement-dependent synaptic engulfment is an important mechanism underlying early synaptic loss in AD. Under physiological conditions, low-activity or damaged synapses can be recognized by C1q, which subsequently promotes C3 cleavage and deposition of complement fragments such as C3b/iC3b on the synaptic surface. CR3 expressed on microglia recognizes these complement-tagged synapses and mediates their engulfment, thereby participating in synaptic pruning and neural circuit remodeling.39,40 Normally, this process is an important component of physiological synaptic pruning during development and under homeostatic conditions.
In AD, however, Aβ, Tau pathology, aging, and the inflammatory microenvironment can aberrantly activate this pathway, causing some still-functional synapses to be mistakenly labeled for elimination.
Aβ oligomers or early pathological stimuli can induce C1q deposition at synaptic sites, followed by C3 activation and microglial engulfment of complement-tagged synapses. C1q is a key molecule that initiates synaptic tagging through the classical complement pathway and can act as an “eat-me”-like signal on synaptic surfaces, triggering downstream cascades involving C4, C2, and C3.41 After C1q initiates the classical complement cascade, C3 is cleaved into C3b and further processed into fragments such as iC3b. iC3b-tagged synapses can be recognized by microglia expressing CR3/CD11b-CD18, thereby promoting synaptic internalization and phagocytic clearance.42 Studies have shown that C3 is activated in AD and contributes to neurodegenerative injury, suggesting that C3 is an important amplification node connecting the complement cascade, microglial responses, and synapse loss.43 Tau pathology can also induce C1q-related synaptic tagging and microglial engulfment, and C1q blockade alleviates Tau-induced synapse loss, indicating that Aβ and Tau may converge on the complement-dependent synaptic elimination pathway.44
In addition, TREM2 can bind C1q and protect synapses from complement-mediated excessive clearance, whereas insufficient TREM2 function enhances microglial synaptic engulfment and accelerates synapse loss.45 The neuron-derived secreted protein Nptx2 can bind C1q and limit complement activation. Loss of Nptx2 enhances microglia-mediated synaptic engulfment, suggesting that synapses themselves possess anti-complement protective mechanisms.46 The CD47–SIRPα axis functions as a “do-not-eat-me” signal to restrict excessive microglial pruning. When this protective signal is insufficient, complement-tagged synapses are more readily eliminated.47 Therefore, complement-dependent synaptic engulfment is an important bridge connecting Aβ and Tau pathology, synapse loss, and cognitive decline.
Complement-dependent synaptic engulfment therefore represents an important mechanistic link between Aβ and tau pathology, synaptic loss, and cognitive decline. However, microglial synaptic engulfment is not necessarily pathological under all conditions. In Aβ-related mouse models, TREM2-dependent microglia preferentially eliminate abnormally hyperactive synapses exposing phosphatidylserine on the outer leaflet of the plasma membrane, thereby reducing neuronal network hyperexcitability and suggesting that selective synaptic removal may exert an adaptive and protective effect.48 In addition, TREM2 can directly bind C1q and restrain activation of the classical complement cascade; impaired TREM2 function may therefore exacerbate complement-mediated pathological synapse loss.45 Importantly, C1q-dependent synaptic elimination is not restricted to microglia.
In tau-related AD models, astrocytes also participate in synaptic engulfment and may exhibit compensatory increases in phagocytic activity when microglial synapse clearance is impaired.46 These findings highlight the need for future studies to distinguish the physiological removal of damaged or abnormally hyperactive synapses from the pathological over-elimination of otherwise functional synapses, while further defining the relative contributions of microglia and astrocytes to these processes.
Microglial Immunometabolic Reprogramming
Following persistent stimulation by Aβ, Tau, cellular debris, oxidative stress, and inflammatory factors, microglia undergo immunometabolic reprogramming. Homeostatic microglia primarily rely on oxidative phosphorylation to maintain a low-inflammatory and highly surveillant state, whereas under disease conditions they tend to enhance glycolysis, lipid handling, and amino acid metabolism to support migration, phagocytosis, and inflammatory cytokine release.49 Glycolytic reprogramming is a key feature of microglial activation in AD. Studies have shown that PKM2+Iba1+ glycolytic microglia are significantly increased in the hippocampal–entorhinal cortex of patients with AD and are enriched around Aβ plaques, p-tau lesions, and cerebral vessels.
This enhanced glycolytic state is associated with an inflammatory-like phenotype, lipid droplet accumulation, and phagocytic exhaustion, suggesting that excessive glycolysis may drive microglia from a clearance-oriented phenotype toward a chronic inflammatory state.50
Microglial metabolic reprogramming is therapeutically modifiable. Activation of TRPV1 can maintain microglial glycolysis and oxidative phosphorylation through pathways such as AKT/mTOR/HIF-1α, thereby enhancing Aβ uptake and clearance.51 TREM2 is a core receptor that links recognition of AD pathology with metabolic adaptation. Its function helps maintain ATP production, biosynthetic metabolism, autophagy–lysosomal function, and the ability of microglia to cluster around plaques, proliferate, and phagocytose.52 TREM2 agonistic antibodies can enhance energy metabolism and mitochondrial function in microglia in AD models, suggesting that increasing microglial metabolic capacity may promote protective microglial responses.53 Reviews have noted that microglial metabolic reprogramming in AD involves not only glucose metabolism but also fatty acid, lipid, and amino acid metabolism.
These metabolic changes are regulated by signaling pathways such as TREM2, PI3K–AKT–mTOR, HIF-1α, AMPK, PPARs, and LXRs, thereby influencing microglial phagocytic function, mitochondrial activity, glycolytic capacity, and inflammatory responses.54 Thus, immunometabolic reprogramming provides an important entry point for understanding microglial dysfunction and its therapeutic potential. Notably, metabolic alterations in AD microglia cannot be adequately described as a simple and sustained shift from oxidative phosphorylation to glycolysis. A 2025 study of the hippocampal–entorhinal cortex region in patients with AD reported an increase in PKM2-positive glycolytic microglia, which was associated with lipid droplet accumulation and impaired phagocytic capacity.
However, the study included only eight patients with AD and eight controls, and these findings therefore require validation in larger cohorts.50 In contrast, experimental models exposed to prolonged Aβ stimulation have shown that an initial increase in glycolysis may be followed by a metabolically tolerant state characterized by impairment of both glycolysis and oxidative phosphorylation.51 These apparently divergent findings may reflect differences in disease stage, duration of pathological stimulation, and experimental model systems.
Similarly, enhanced TREM2 signaling does not appear to exert uniform effects across all microglial states. A 2026 study showed that TREM2 expression levels are closely associated with oxidative phosphorylation, cholesterol homeostasis, cellular bioenergetics, and phagocytic capacity; however, the efficacy of TREM2 agonist antibodies was confined to a specific expression-dependent “response window”, rather than increasing proportionally with TREM2 abundance.55 These observations suggest that therapeutic modulation of microglial immunometabolism may need to be tailored to cellular state and disease stage, rather than relying on indiscriminate enhancement or suppression of a single metabolic pathway.
Roles of Astrocytes in AD
Astrocytes are abundant glial cells in the central nervous system. During the onset and progression of AD, Aβ deposition and Tau pathology can induce reactive changes in astrocytes, enabling them to participate in multiple pathological processes, including inflammatory responses and complement activation, synaptic dysfunction, metabolic reprogramming, disruption of lipid homeostasis, oxidative stress, and injury to the blood–brain barrier and neurovascular unit.
Aβ- and Tau-Induced Reactive Changes in Astrocytes
Both Aβ and Tau pathology can induce reactive changes in astrocytes in AD. These changes do not represent a simple “A1/A2” dichotomy but instead constitute a spectrum of disease-associated states that dynamically evolve with pathological type and disease stage. They are characterized by upregulation of molecules such as GFAP, SERPINA3, C3, APOE, CHI3L1/YKL-40, VIM, and MAOB, as well as cellular hypertrophy, process remodeling, and functional alterations.56–58 During the early stage of Aβ pathology, astrocytes can recognize and phagocytose Aβ through receptors such as CD36, CD47, and RAGE.59 Astrocyte-derived APOE and cholesterol can influence neuronal Aβ production and plaque formation.
Astrocyte-specific APOE expression is sufficient to restore fibrillar Aβ plaques and peri-plaque glial responses, indicating that astrocytes are not only activated by Aβ but also reciprocally shape the Aβ plaque microenvironment.60,61
Tau pathology can also be recognized by and elicit responses from astrocytes. Extracellular Tau can be taken up by astrocytes. Uptake of monomeric Tau may not be entirely dependent on heparan sulfate proteoglycans (HSPGs), whereas cellular entry of Tau aggregates or fibrils is associated with pathways involving HSPGs, LRP1, and integrin receptors.62–64 Fibrillar recombinant human Tau can directly bind the astrocytic integrin αV/β1 receptor complex and mediate the entry of Tau fibrils into astrocytes.65 AD brain–derived Tau fibrils can accumulate in astrocytes and resist efficient degradation, inducing upregulation of reactive and inflammatory molecules such as GFAP, vimentin, IL-8, and MCP-1.
They can also impair neuronal synaptic and metabolic function through intercellular propagation and conditioned media.66 These findings suggest that when astrocytic clearance or degradation of Tau is insufficient, astrocytes may shift from clearance cells to intermediaries that amplify Tau propagation and neurotoxicity.
In addition, Aβ can first induce microglial activation. Inflammatory signals released by microglia subsequently amplify astrocytic responses and promote cognitive impairment through Tau phosphorylation and aggregation.67 Blocking microglia-induced reactive astrocytes can alleviate neuroinflammation and neuronal injury in AD, indicating that glial crosstalk is an important component of Aβ/Tau-induced astrocytic alterations.68 Aβ- and Tau-induced astrocytic reactivity is stage-dependent and heterogeneous. Moderate early responses may help encapsulate plaques, clear abnormal proteins, and limit the spread of damage.
However, persistent or excessive responses lead to the release of pro-inflammatory factors and complement molecules, disruption of glutamate and ion homeostasis, reduced metabolic support, and enhanced Tau propagation, ultimately promoting synaptic dysfunction, neuronal injury, and cognitive decline.69
Astrocytic Inflammatory Responses and Complement Activation
Aβ oligomers or plaques, p-tau, neuronal injury debris, and vascular-derived inflammatory signals can jointly induce astrocytes to shift from homeostatic supportive cells toward inflammatory and complement-related reactive states. Aβ can directly induce the expression of IL-6, TNF-α, chemokines, and other inflammatory mediators through the astrocytic MyD88/NF-κB pathway.70 Studies have shown enhanced activation of NF-κB p50 and p-p65 in GFAP+ astrocytes in AD. NF-κB can drive the expression of molecules such as GFAP, IL-6, KC/CXCL1, YKL-40, AQP4, and complement C3.
However, its role is bidirectional: excessive activation promotes inflammation and neurotoxicity, whereas complete inhibition may weaken astrocytic adaptive responses to Aβ/Tau and impair their support of clearance functions.71 Aβ-induced astrocytic inflammation is also regulated by AhR. AhR activation can limit Aβ-induced release of pro-inflammatory cytokines and C3 expression, suggesting that AhR may serve as an upstream regulatory node that suppresses astrocytic inflammatory and complement responses.72 In addition, SerpinA3N is an inflammatory molecule highly associated with reactive astrocytes in AD. The interaction between SerpinA3N and APOE can exacerbate astrocytic inflammatory responses and AD progression through NF-κB activation.73 Complement activation is an important downstream mechanism of astrocytic Complement activation is an important downstream mechanism of inflammatory responses in astrocytes.
NF-κB activation can induce astrocytes to release C3, and C3a generated by C3 cleavage can act on C3aR expressed by neurons or glial cells, leading to dendritic abnormalities, synaptic dysfunction, and neuroinflammation.74 Astrocyte-derived C3 can also regulate Aβ dynamics through microglial C3aR. Acute C3/C3a signaling may enhance microglial phagocytosis, whereas chronic C3/C3a exposure impairs phagocytic function and exacerbates Aβ pathology, indicating that astrocytic complement activation is strongly time-dependent.75 The C3/C3aR axis also participates in Tau pathology–related neuroinflammation. In the brains of patients with AD, C3 and C3aR1 correlate with Braak stage and cognitive decline, whereas C3aR deficiency reduces Tau pathology, inflammatory responses, and synaptic deficits.76 Although substantial experimental evidence supports the involvement of NF-κB–C3/C3aR signaling in AD-related neuroinflammation and synaptic injury, the strongest causal evidence to date has been derived primarily from transgenic mouse and cellular models.
More importantly, the effects of complement signaling are highly dependent on cell type and pathological context and therefore should not be interpreted simply as “complement activation equals neurotoxicity”. For example, in tau and Aβ/tau mouse models, C1q-dependent synaptic elimination involves both astrocytes and microglia, with the two glial cell types showing distinct preferences for engulfing excitatory vs inhibitory synapses. Moreover, when microglial phagocytic function is impaired, astrocytes can exhibit compensatory increases in synaptic engulfment.40 These findings indicate that the role of complement in AD should be considered within the broader context of microglia–astrocyte–synapse interactions.
Its cellular sources, stage-specific effects, and optimal therapeutic window in human AD remain to be clarified.
Astrocytic Synaptic Homeostasis and Neurotransmitter Imbalance
Synapse loss and neural network dysfunction are closely associated with cognitive decline in AD. Aβ oligomers can directly interfere with postsynaptic receptor function and can also impair astrocytic regulation of the synaptic microenvironment by inducing reactive changes.77 In AD models, reduced astrocytic Ca2+ responsiveness, decreased endoplasmic reticulum Ca2+ stores, and reduced STIM1 expression weaken astrocytic responses to neuronal activity. Restoration of astrocytic STIM1 improves Ca2+ signaling and long-term synaptic plasticity.78 Astrocytic GLT-1 is responsible for clearing glutamate from the synaptic cleft.
In AD, GLT-1 downregulation can cause insufficient glutamate clearance, abnormal activation of presynaptic mGluRs, and impairment of short-term synaptic plasticity, whereas restoration of GLT-1 function improves glutamate uptake and cognitive impairment in APP/PS1 mice.79,80 In addition, reduced astrocytic L-serine synthesis decreases D-serine availability and diminishes activation of the NMDA receptor co-agonist site, thereby impairing long-term potentiation and spatial memory. Supplementation with L-serine can partially reverse these synaptic plasticity deficits.81
As AD progresses, astrocytic regulation of inhibitory transmitters, neuromodulators, and synaptic structure becomes further imbalanced. Studies have shown that GAD67 and GAT3/4 are upregulated in reactive astrocytes and are associated with enhanced tonic inhibition in the hippocampal CA1 region and dentate gyrus, suggesting that altered astrocytic GABA homeostasis can influence the excitatory–inhibitory balance.82 MAO-B is significantly elevated in reactive astrocytes in AD. On the one hand, it contributes to astrocytic GABA production and alters tonic inhibition; on the other hand, it generates H2O2, thereby aggravating oxidative stress and Tau-related neuronal injury.83 In addition, astrocytic A2A receptors and cAMP signaling can influence Ca2+ activity, glutamate uptake, ATP/adenosine release, and synaptic plasticity.
Therefore, the A2A–cAMP axis may participate in AD-related memory impairment.84 The mechanisms involving STIM1, GLT-1, L-serine, GABA/MAO-B, and A2A receptors collectively provide substantial experimental support for the role of astrocytes in regulating neuronal network function. However, most causal evidence showing that restoration of an individual pathway can improve cognition has been derived from AD animal models. Notably, single-nucleus transcriptomic analyses of the human brain have revealed reduced astrocytic expression of SLC1A2, which encodes EAAT2/GLT-1, in selectively vulnerable brain regions,58 providing human evidence for impaired glutamate homeostasis.
Nevertheless, it remains unclear whether these neurotransmitter alterations represent early drivers of neuronal dysfunction or instead reflect secondary or compensatory responses to pre-existing synaptic and neuronal injury. Therefore, caution is warranted when extrapolating pathway-specific interventions from animal models directly to therapeutic targets in human AD.
Astrocytic Metabolic Reprogramming
Astrocytes are major hubs of brain energy metabolism. Through glucose uptake, glycolysis, glycogenolysis, and lactate transport, they provide energy substrates to neurons via the “astrocyte–neuron lactate shuttle”.85 Reduced cerebral glucose metabolism appears early in AD and coincides with glial activation, suggesting that astrocytic metabolic abnormalities may participate in disease progression before overt neuronal death.86 When Aβ-induced microglial inflammatory signals persistently activate the AKT–mTOR–HIF-1α pathway in astrocytes, glycolysis may shift from compensatory enhancement to pathological hyperactivation, leading to excessive L-lactate production, impaired synaptic plasticity, and increased Aβ burden.87
As the disease progresses, astrocytic glucose metabolic reprogramming becomes increasingly coupled with inflammatory metabolism and mitochondrial dysfunction. Aβ and Tau oligomers can activate astrocytic IDO1, enhancing tryptophan metabolism through the kynurenine pathway. This subsequently suppresses glycolysis and lactate transport to neurons through AhR signaling, resulting in insufficient hippocampal glucose utilization and cognitive decline.88 Meanwhile, Aβ can also induce mitochondrial fission, reduced membrane potential, decreased ATP production, and ROS accumulation in astrocytes, shifting glucose metabolism from efficient oxidative utilization toward inefficient stress-related glycolysis.89 At the redox level, glucose diversion into the pentose phosphate pathway generates NADPH, which is essential for astrocytes to maintain the GSH/GPX4 antioxidant system and suppress lipid peroxidation.
NRF2 deficiency weakens this antioxidant branch of glucose metabolism, promoting astrocytic ferroptosis and amplification of neuroinflammation.90 Current findings on astrocytic metabolic reprogramming in AD are not entirely consistent. Some experimental studies suggest that reactive astrocytes exhibit enhanced glycolysis and excessive lactate production, whereas more recent work has shown that Aβ and tau oligomers can activate the astrocytic IDO1–kynurenine–AhR pathway, thereby suppressing glycolysis and lactate production. Pharmacological inhibition of IDO1 restored astrocyte–neuron metabolic support and improved cognitive performance in both Aβ- and tau-based mouse models.
Supporting evidence was also obtained from human iPSC-derived astrocytes and astrocyte–neuron co-culture systems derived from patients with AD; however, the causal evidence for therapeutic benefit remains largely confined to animal models.88 These apparently divergent findings suggest that astrocytic metabolic states in AD may vary according to the nature and duration of pathological stimuli as well as disease stage, and therefore should not be reduced to a unidirectional model of either “increased glycolysis” or “suppressed glycolysis”.
Astrocytic Lipid Homeostasis and Oxidative Stress
Astrocytes are the main source of cholesterol and ApoE in the brain. Through the ABCA1 pathway, they form lipoprotein particles and supply cholesterol to neurons to maintain membrane structure and synaptic function. However, in the AD pathological environment, abnormal increases or imbalanced distribution of astrocyte-derived cholesterol can promote neuronal lipid raft expansion, facilitating the colocalization of APP with β- and γ-secretases within lipid rafts and thereby enhancing Aβ production.
In addition to cholesterol, excess fatty acids and peroxidized lipids generated by neuronal activity or stress are normally transferred to astrocytes, stored in lipid droplets, and subjected to β-oxidation to reduce neuronal lipotoxicity. APOE4 disrupts this neuron–astrocyte coupling of fatty acid transport and detoxification, leading to reduced fatty acid oxidation, lipid droplet accumulation, and buildup of lipid intermediates, ultimately weakening astrocytic metabolic protection of neurons.91,92
As disruption of lipid homeostasis persists, impaired lipid droplet turnover, mitochondrial dysfunction, and oxidative stress in astrocytes amplify one another, forming a vicious cycle that promotes AD progression. APOE can localize to the surface of astrocytic lipid droplets and regulate lipid droplet size and triglyceride saturation. APOE4-associated lipid droplets are larger, enriched in unsaturated triglycerides, and more susceptible to lipid peroxidation, providing an organelle-level mechanistic explanation for the conversion of lipid droplet homeostatic abnormalities into oxidative injury.93 APOE4 can also cause lysosomal cholesterol accumulation, impair the clearance of damaged mitochondria, and disrupt oxidative phosphorylation, leading to decreased ATP production and increased mitochondrial ROS.
ROS further oxidize polyunsaturated fatty acids and generate reactive lipid products such as 4-HNE, MDA, and lipid hydroperoxides, which in turn damage mitochondrial membranes, lysosomal membranes, and synaptic support functions.94 In terms of antioxidant defense, NRF2 regulates antioxidant and anti-ferroptotic molecules such as GSH-synthesis enzymes, HO-1, NQO1, and GPX4, thereby maintaining astrocytic lipid redox homeostasis. NRF2 deficiency leads to increased ROS, downregulation of GPX4, enhanced lipid peroxidation, and ferroptosis-like astrocytic injury.90 In addition, NOX4-mediated ROS production can further promote iron-dependent lipid peroxidation and astrocytic ferroptosis, transforming oxidative stress from a metabolic byproduct into an active driver of inflammation and neurodegenerative injury.95 Notably, findings regarding the lipid droplet phenotype of APOE4 astrocytes have not been entirely consistent across experimental systems. Earlier studies reported that APOE4 astrocytes contained a greater number of smaller lipid droplets,96 whereas a 2024 study conducted under lipogenic stimulation found that APOE4 astrocytes formed fewer but larger lipid droplets enriched in unsaturated triglycerides, together with impaired lipid droplet turnover and increased susceptibility to lipid peroxidation.93 These discrepancies may reflect differences in cell source, lipid-loading conditions, culture state, and methods used for lipid droplet quantification.
Thus, the more consistent conclusion is not that APOE4 produces a fixed lipid droplet phenotype, but rather that it disrupts lipid droplet turnover, fatty acid oxidation, and resistance to lipid peroxidation, thereby compromising the ability of astrocytes to handle neuron-derived lipid burden.
Astrocyte-Mediated Injury to the Neurovascular Unit
Astrocytic endfeet ensheath cerebral microvessels and, together with endothelial cells, pericytes, the basement membrane, and microglia, form the neurovascular unit (NVU), which participates in blood–brain barrier maintenance, cerebral blood flow regulation, and Aβ efflux. In AD, Aβ deposition, APOE4, inflammatory factors, and oxidative stress can disrupt the structure of astrocytic endfeet and disturb the polarized distribution of molecules such as AQP4 and Kir4.1, leading to impaired water channel function, K+ buffering, and glymphatic clearance.97 Reactive astrocytes can also release IL-1β, TNF-α, CCL2, VEGF-A, MMPs, ROS, and other factors, inducing abnormal expression or localization of endothelial tight junction proteins such as ZO-1, occludin, and claudin-5. This promotes leukocyte adhesion and peripheral immune cell infiltration, causing blood–brain barrier injury to evolve from structural disruption into inflammatory barrier destabilization.98 As the disease progresses, astrocyte-mediated NVU injury further manifests as AQP4 depolarization, neurovascular coupling dysfunction, and failed Aβ clearance.
AQP4 is normally enriched in perivascular astrocytic endfeet and participates in cerebrospinal fluid–interstitial fluid exchange and glymphatic clearance. In AD, AQP4 depolarizes from the endfoot domain and becomes redistributed, reducing glymphatic clearance efficiency, weakening Aβ efflux, and promoting plaque deposition.99 Astrocytes also regulate local cerebral blood flow through Ca2+ signaling, K+ buffering, and arachidonic acid metabolites. Aβ, Tau, ROS, and inflammatory factors disturb endfoot Ca2+ oscillations, Kir4.1/K+ buffering, and the balance of vasoactive factors such as PGE2, EETs, and 20-HETE, resulting in insufficient activity-evoked vasodilation, neurovascular coupling impairment, and cerebral hypoperfusion.100,101 Meanwhile, damaged endothelial cells can transfer ENG to astrocytes via extracellular vesicles, activate the TGFBRI/Smad3 pathway, and enhance astrocytic reactivity.
This suggests a bidirectional pathological loop in AD in which astrocytes damage blood vessels and vascular injury reciprocally activates astrocytes.102 Recent experimental evidence further supports this bidirectional interaction. A 2025 study integrating cerebrovascular transcriptomic data with cerebrospinal fluid proteomic profiles from patients with AD identified increased endothelial cell-derived endoglin (ENG). Subsequent cellular and APP/PS1 mouse experiments showed that brain microvascular endothelial cells can transfer ENG to astrocytes via extracellular vesicles, thereby activating TGFBRI/Smad3 signaling and promoting astrocytic reactivity.102 However, evidence that targeting this pathway improves cognitive outcomes is currently limited to mouse models.
Thus, the causal relevance and therapeutic feasibility of this endothelial cell–astrocyte signaling axis in human AD remain to be established.
Roles of Oligodendrocytes in AD
Oligodendrocytes and their precursor cells are essential for maintaining central nervous system myelin integrity and axonal function. Aβ deposition and Tau pathology can injure the oligodendrocyte lineage, induce disease-associated states and immunometabolic abnormalities, and lead to oligodendrocyte precursor cell (OPC) senescence, impaired differentiation, and remyelination failure. These processes ultimately result in myelin disruption, white matter network abnormalities, and cognitive decline.
Aβ and Tau Pathology and Oligodendrocyte-Lineage Injury
The oligodendrocyte lineage includes OPCs, differentiating premyelinating oligodendrocytes, and mature myelinating oligodendrocytes. Their major functions are to maintain myelination, axonal conduction, neuronal metabolic support, and white matter network integrity. Aβ oligomers and plaques can injure oligodendrocytes through oxidative stress, mitochondrial damage, glutamate excitotoxicity, and inflammatory cytokine release, reducing the expression of myelin-related proteins such as MBP, PLP1, MAG, and MOG and leading to myelin loosening and reduced axonal conduction efficiency.103,104 Aβ can also directly act on OPCs, inducing dystrophy, morphological degeneration, and reduced renewal capacity.
This weakens the compensatory proliferation and differentiation of OPCs in response to myelin injury and thereby diminishes remyelination potential.105 In addition, Aβ can disrupt oligodendrocyte-lineage dynamics through PKC signaling, causing imbalance in OPC proliferation, differentiation, and myelin formation. Aβ can also interfere with hnRNPA2-mediated RNA transport and local translation, resulting in abnormal MBP expression and localization, thereby disrupting myelin protein homeostasis and myelin structural stability.106,107 Meanwhile, specific OPC subpopulations and mature oligodendrocytes themselves can produce Aβ and contribute to the local plaque microenvironment and neuronal dysfunction, indicating that the oligodendrocyte lineage is not only a target of Aβ toxicity but may also be an important participant in Aβ pathology.108,109
Notably, the causal relationship between oligodendrocyte/myelin abnormalities and Aβ pathology remains incompletely resolved and may be bidirectional. On the one hand, several AD mouse models have shown that primary myelin dysfunction can promote the local accumulation of APP/BACE1-related amyloidogenic machinery within axonal swellings, thereby enhancing Aβ deposition; meanwhile, myelin debris may divert microglial phagocytic capacity away from amyloid plaques.104 On the other hand, amyloid pathology itself can induce oligodendrocyte and myelin injury through a CD8+ T cell–IFNγ–microglia axis.110 In addition, recent studies have demonstrated that human oligodendrocytes are capable of producing Aβ, although the causal evidence that selective suppression of oligodendrocyte-derived Aβ reduces plaque burden has thus far been obtained primarily in AD mouse models.109 Collectively, these findings support a bidirectional, mutually reinforcing relationship between myelin dysfunction and Aβ pathology rather than a simple unidirectional cause–consequence model. Compared with Aβ, direct causal evidence that tau pathology drives oligodendrocyte-lineage injury in human AD remains relatively limited.8
Disease-Associated Oligodendrocyte States and Immunometabolic Abnormalities
In AD, Aβ, Tau, APOE4, aging, and chronic inflammation can jointly induce oligodendrocytes to transition from a myelin-maintaining phenotype toward disease-associated states. In this process, oligodendrocytes shift from cells responsible for myelination and axonal metabolic support into pathological participants characterized by stress, immune activation, and metabolic abnormalities. Single-cell transcriptomic studies have identified disease-associated oligodendrocytes (DOLs/DAOs) in AD mouse models and human brains.
These cells are enriched near Aβ plaques and in brain injury regions and show upregulation of genes related to injury, complement, chemokines, and antigen presentation, including SERPINA3N/SERPINA3, Apoe, C4b, Cd74, Ccl3, and Ccl4. These findings suggest that disease-associated oligodendrocyte states not only reflect myelin loss but may also actively participate in inflammatory amplification and pathological progression.111 The inflammatory microenvironment can cause OPCs and oligodendrocytes to acquire “immune-like” phenotypes marked by MHC-I/MHC-II, B2m, and Cd74, thereby conferring potential capacities for antigen processing, antigen presentation, and interactions with T cells. This may transform myelin injury from structural degeneration into immune-mediated persistent damage.112 Aβ can also promote the progressive accumulation of aberrantly interferon-activated microglia through CD8+ T cells, and these microglia exhibit myelin-damaging activity.110 Immunometabolic abnormalities are important features of disease-associated oligodendrocyte states.
Mature oligodendrocytes strongly rely on lipid synthesis, cholesterol metabolism, and glycolysis to maintain myelin and can provide axons with energy substrates such as lactate.113 In AD, chronic inflammation and oxidative stress can disrupt oligodendrocyte mitochondrial function, cholesterol metabolism, lipid synthesis, and myelin membrane renewal, shifting them from a myelin-maintaining and axon-supporting state toward stress and metabolic imbalance.114 At the same time, OPCs and oligodendrocytes can express immune-regulatory molecules. When the inflammatory microenvironment becomes excessive, it inhibits their differentiation and remyelination capacity, thereby aggravating failure of myelin repair.115 Tau pathology and oxidative stress can disturb lipid droplet formation, lipid peroxide handling, and iron homeostasis in glial cells, rendering oligodendrocytes more vulnerable to lipid peroxidation, ferroptosis-like injury, and destabilization of the myelin–axon interface.116,117
OPC Senescence, Differentiation Impairment, and Remyelination Failure
OPCs are the core cell population responsible for maintaining oligodendrocyte renewal and remyelination repair in the adult central nervous system. Their normal functions include sensing axonal activity, proliferating and migrating, differentiating into mature oligodendrocytes, and forming new myelin, thereby maintaining axonal conduction, neural network synchronization, and neuronal metabolic support. Under normal conditions, OPCs can proliferate, migrate, and differentiate into mature oligodendrocytes after demyelination or axonal injury to repair myelin.
In AD, however, Aβ, inflammatory factors, oxidative stress, and the aging microenvironment can induce OPCs to exhibit senescence-associated features such as p16INK4a, p21, and SA-β-gal positivity, reducing their proliferative and differentiation capacities.118 As OPC senescence worsens, OPCs may accumulate around lesions but fail to effectively transform into mature oligodendrocytes, leading to insufficient generation of new oligodendrocytes, reduced synthesis of myelin proteins, and incomplete myelin repair.119 OPC senescence is supported by some human neuropathological evidence, whereas the causal evidence for its therapeutic targeting remains derived largely from animal models. In postmortem brain tissue from patients with AD, OLIG2/NG2-positive OPCs located near Aβ plaques have been reported to express senescence-associated markers such as p16 and p21. By contrast, evidence that senescent cell clearance can reduce Aβ burden, attenuate neuroinflammation, and improve cognitive performance has been obtained primarily from APP/PS1 mice.118 Thus, whether senolytic strategies can enhance remyelination or confer meaningful clinical benefit in human AD remains to be determined.
OPC differentiation impairment and remyelination failure in AD are also jointly regulated by lipid metabolism, inflammatory signaling, and disease-associated oligodendrocyte states. Single-cell studies have identified disease-associated oligodendrocyte states in AD models, accompanied by increased expression of immune and chemokine-related genes such as Cd74, C4b, Apoe, and Ccl3/Ccl4 and abnormal activation of ERK1/2 signaling. These findings suggest a continuous pathological process linking OPC differentiation impairment and transformation into disease-associated oligodendrocyte states.109 Remyelination failure, in turn, exacerbates core AD pathology.
Myelin dysfunction can cause axonal swelling, leading to local accumulation of Aβ-generating molecules such as APP and BACE1. In addition, clearance of myelin debris can divert microglial phagocytic resources, reducing their ability to clear Aβ plaques and thereby promoting increased Aβ deposition.101 Meanwhile, multimodal single-cell atlases of the human brain have shown that specific mature oligodendrocyte subpopulations are reduced at relatively early stages of AD pathology, accompanied by increased OPC differentiation and remyelination-related programs.120 These findings suggest that early OPC alterations in AD may not simply reflect “repair failure”, but may also include compensatory remyelination responses. However, because current human data are derived largely from postmortem samples and pathology-based pseudotemporal analyses, it remains unclear when and why such compensatory responses transition to decompensation.
Myelin Injury, White Matter Abnormalities, and Cognitive Decline
Oligodendrocytes regulate axonal conduction velocity through myelin formation and contribute to activity-dependent myelin plasticity, axonal metabolic support, and information processing. In AD, myelin injury and white matter abnormalities represent early network-level pathological events driven by instability of the oligodendrocyte lineage. Oligodendrocytes can transition from homeostatic myelin-maintaining cells into disease-associated oligodendrocytes, characterized by abnormalities in myelin genes, stress responses, immune regulation, and cell survival pathways, thereby weakening axonal myelination and remyelination capacity.56 APOE4 can disrupt cholesterol synthesis, transport, and localization in oligodendrocytes, leading to endoplasmic reticulum stress and insufficient myelin formation, which makes white matter structure more vulnerable in an AD-susceptible background.
Among the currently proposed oligodendrocyte-related mechanisms, APOE4-associated cholesterol dysregulation is supported by relatively robust cross-model evidence. Single-cell transcriptomic analyses of the human brain, tissue lipid profiling, human iPSC-derived oligodendrocytes, and APOE-targeted replacement mouse models have consistently shown that APOE4 promotes abnormal cholesterol accumulation in oligodendrocytes and is associated with impaired myelination.121 Meanwhile, genes related to glycolysis and ketone body utilization are downregulated in oligodendrocytes in the AD brain, reducing their ability to provide energy substrates such as lactate to axons and thereby causing unstable axonal conduction and delayed axonal degeneration.122 Under increasing metabolic stress, mature oligodendrocytes can activate the inflammasome through the Drp1–HK1–NLRP3 axis, inducing inflammatory injury, demyelination, and axonal degeneration.123 Myelin injury also forms a bidirectional amplification loop with core AD pathological proteins. Myelin dysfunction and demyelinating injury can promote the accumulation of Aβ-generating mechanisms within axonal swellings and increase APP cleavage, while myelin debris diverts microglia and weakens their encapsulation and clearance of Aβ plaques.
At the same time, lipid debris generated by myelin disruption can further activate microglia and astrocytes, amplify neuroinflammation, and promote Aβ deposition.124 AD may also feature abnormalities in paranodal structures, imbalanced myelin–axon signaling, and protein aggregation, causing decreased conduction efficiency even in white matter tracts without widespread demyelination.125 Clinically, white matter hyperintensity burden is associated with Aβ deposition, vascular risk factors, and cognitive status. Higher baseline white matter hyperintensity burden and its progression are associated with poorer cognitive performance and faster memory decline, indicating that white matter abnormalities can impair hippocampal–frontal–parietal network connectivity and lead to declines in memory, executive function, and information processing speed.126,127
Glia-Related Biomarkers and Their Applications
Glia-related biomarkers can reflect pathological processes such as neuroinflammation, myelin injury, white matter abnormalities, and impaired intercellular communication, thereby serving as important complements to traditional Aβ, p-tau, and NfL biomarkers. Glia-related biomarkers more specifically reflect the pathological microenvironment and disease activity in AD and may provide additional information for early screening, disease staging, progression prediction, therapeutic target selection, and treatment monitoring.
Biomarkers of Astrocyte Reactivity
Biofluid biomarkers of astrocyte reactivity include GFAP, YKL-40/CHI3L1, and S100B. GFAP is an intermediate filament protein in astrocytes and is mainly derived from reactive astrocytes. It can be detected in CSF, plasma, or serum.
Across the AD continuum, particularly in Aβ-positive individuals and patients with preclinical AD, blood and CSF GFAP levels are generally elevated, suggesting that astrocytic reactivity occurs before the appearance of typical clinical symptoms.128 Compared with CSF GFAP, plasma GFAP shows relatively high sensitivity for Aβ positivity and is easier to sample, making it a promising blood-based biomarker for early screening, prediction of Aβ positivity, and risk stratification.129 Autopsy and cohort studies further show that elevated GFAP in brain tissue and blood is associated with Aβ burden, cerebral amyloid angiopathy, and cognitive decline, indicating that GFAP not only reflects astrocyte reactivity but may also indicate the activity of AD pathology and prognostic risk.130 YKL-40, also known as CHI3L1, is mainly produced by reactive astrocytes, microglia, and inflammation-associated glial cells. CSF-based studies are more mature, whereas blood YKL-40 shows relatively weaker stability and disease specificity.131 In AD, CSF YKL-40 is elevated in most studies and is associated with p-tau, brain atrophy, inflammatory factors, and cognitive decline. Therefore, it more strongly reflects the intensity of neuroinflammation, Tau-related neurodegenerative progression, and disease activity.132,133 GFAP and YKL-40 do not represent identical astrocytic states.
GFAP is more commonly associated with Aβ-related reactive changes, whereas YKL-40 is more closely linked to Tau pathology, the inflammatory microenvironment, brain atrophy, and longitudinal cognitive decline. Thus, combining these two markers can complement traditional Aβ, p-tau, and NfL measures by capturing both “early Aβ-related glial reactivity” and “disease progression–related inflammatory responses”. S100B is an astrocyte-derived calcium-binding protein detectable in blood and CSF.
Although it may be elevated in AD, it is strongly influenced by brain injury, blood–brain barrier disruption, and peripheral inflammation, resulting in insufficient specificity. Therefore, S100B is more suitable as an auxiliary indicator of glial injury or blood–brain barrier dysfunction and should not be used alone for AD diagnosis.134
Biomarkers of Microglial Activation and Metabolism
Representative microglial biomarkers include sTREM2, TREM2-related genetic variants, TSPO, and certain microglia-derived extracellular vesicle markers. sTREM2 is a soluble product released after cleavage of the extracellular domain of membrane-bound TREM2 by enzymes such as ADAM10/17. It is mainly derived from microglia and can be detected in CSF and blood. CSF sTREM2 is elevated in patients with MCI and AD, and plasma sTREM2 also shows an increasing trend in some AD cohorts, although its ability to distinguish different disease stages is limited.135 Therefore, the main clinical significance of sTREM2 is not as a standalone diagnostic marker for AD, but rather as an indicator of TREM2-related microglial activation, phagocytosis, survival, and lipid metabolic responses.
It can serve as a complementary marker of microglial functional state beyond Aβ and Tau.sTREM2 shows pronounced disease-stage dependence and interindividual variability. Longitudinal ADNI cohort studies indicate that CSF sTREM2 increases with age and is influenced by APOE ε4, sex, and Aβ pathological status. Therefore, age, genetic background, and AD pathological stage should be considered when interpreting sTREM2 levels clinically.136 Combined imaging and CSF studies suggest that sTREM2 has complex relationships with Aβ-PET, Tau-PET, cognitive changes, and A/T/N staging.
In early Aβ-positive stages in which Tau pathology and neurodegeneration are not yet obvious, higher sTREM2 levels or a faster rate of increase may represent a protective microglial plaque response. In later disease stages, however, they may reflect persistent inflammatory activation and metabolic exhaustion.137 Risk variants such as TREM2 R47H weaken microglial clustering around Aβ plaques, phagocytosis, lipid handling, and metabolic adaptation. Therefore, “CSF sTREM2 level + TREM2 genotype + APOE genotype” is more suitable than sTREM2 alone for microglial state stratification, clinical trial enrollment, and monitoring the efficacy of TREM2-targeted therapies.138 In addition, P2RY12, TMEM119, CD11b, Iba1, CD68, complement molecules, and microglia-derived exosomal proteins can reflect the transition from homeostatic microglia to disease-associated microglial states.
However, these markers currently remain largely research-oriented and lack standardized detection procedures and clear clinical thresholds.139
Biomarkers of Oligodendrocytes and Myelin Integrity
AD is not only a disease of gray matter neuronal and synaptic degeneration but is also accompanied by oligodendrocyte dysfunction, white matter injury, disruption of myelin homeostasis, and reduced efficiency of neural network connectivity. MBP is an important structural protein of central myelin and is mainly derived from mature oligodendrocytes and myelin membranes; it is primarily measured in CSF. Elevated CSF MBP generally indicates myelin injury, demyelination, or enhanced myelin remodeling.
Studies have shown that CSF MBP changes dynamically with aging and is higher in Aβ-positive individuals, Aβ-PET-positive individuals, and APOE ε4 carriers. It is also associated with p-tau181, t-tau, and cognitive measures such as MMSE, suggesting that MBP may serve as a candidate biofluid biomarker of impaired myelin integrity and white matter pathology in AD.140 Its clinical value lies in complementing the dimensions of white matter–myelin injury that traditional Aβ, p-tau, and NfL biomarkers cannot fully capture, especially for evaluating network disconnection and white matter pathological burden in cognitive decline. In addition to MBP, MAG, MOG, PLP1, CNPase, and NG2 may reflect the state of the myelin system from different aspects of mature myelin, oligodendrocytes, and OPCs.
These markers can be explored in CSF, blood exosomes, or oligodendrocyte-derived extracellular vesicles, but their direction of change, standardized methods, and large-scale clinical thresholds in AD remain insufficiently established.141,142 Moreover, oligodendrocytes themselves can produce Aβ and cooperate with neurons in plaque formation, while enhancement of myelin renewal can improve cognitive function in AD mice. These findings suggest that myelin-related biofluid markers may be useful not only for assessing white matter injury but also for monitoring the efficacy of remyelination therapies or oligodendrocyte-targeted interventions.143,144 At the imaging level, MRI white matter hyperintensities, DTI diffusion abnormalities, and increased free water reflect integrated effects of myelin, axonal, vascular, and glial injury. Higher white matter hyperintensity burden and its progression are associated with poorer cognitive performance, faster memory decline, and neurodegenerative changes.
Therefore, these imaging markers should be combined with CSF MBP, myelin-related proteins, and cognitive scales to improve assessment of white matter pathology and cognitive prognosis.123,145
Molecular Imaging of Glial Cell Activation
Biofluid biomarkers can reflect the overall level of glial activation but cannot reveal the spatial distribution of glial responses in the brain, their regional selectivity, or their relationship with Aβ/Tau pathology. TSPO-PET is currently the most widely used PET imaging method for neuroinflammation. TSPO mainly reflects activated or phagocytic microglia, but in the human brain it may also be influenced by endothelial cells, vascular-associated cells, and genotype-dependent differences in TSPO binding affinity.
Therefore, TSPO-PET cannot be simply equated with “pure microglial imaging”.146,147 In AD, TSPO-PET uptake is generally increased and is associated with Aβ burden, Tau burden, and MMSE decline. Its main clinical significance lies in visualizing the spatial distribution of neuroinflammation, assisting disease-stage stratification, assessing the relationship between inflammation and protein pathology, and serving as a potential treatment-monitoring tool for anti-inflammatory or microglia-targeted therapies.148 However, TSPO-PET is affected by ligand generation, TSPO rs6971 genotype, insufficient cellular specificity, and signal-to-noise ratio. Therefore, clinical interpretation requires integration with biofluid biomarkers and structural/functional imaging.MAO-B PET has recently emerged as an important approach for imaging reactive astrocytes.
MAO-B is mainly expressed in reactive astrocytes and is upregulated in the AD cortex, white matter, and astrocytes surrounding dense-core plaques, making it a candidate imaging marker for astrocyte reactivity. This modality uses PET imaging, with representative ligands including11 C-deprenyl and18 F-SMBT-1. Increased MAO-B signals in AD may reflect local astrocyte reactivity, activation of fibrous astrocytes in white matter, and risk of tissue injury.
Its main clinical value lies in providing spatial localization information that cannot be obtained from biofluid markers such as GFAP and YKL-40.149 In addition to TSPO and MAO-B, imaging targets such as P2X7R, P2Y12R, CSF1R, COX, CB2R, and S1PR are being explored to characterize different glial inflammatory phenotypes. Most of these targets remain in early-stage or translational research. A more realistic future application will be to integrate PET targets with plasma GFAP, CSF YKL-40, CSF sTREM2, MRI white matter indicators, and Aβ/Tau-PET to construct multimodal maps of glial cell activation.150,151
Applications of Glia-Derived Exosomes in AD
Glia-derived extracellular vesicles, including exosomes and other subtypes, can carry proteins, lipids, mRNAs, miRNAs, and metabolites and can enter the CSF and peripheral blood. They therefore serve as an important bridge connecting glial states in the brain with peripheral liquid biopsy. The advantage of glia-derived exosomes is their ability to enrich cell-of-origin information.
Astrocyte-derived exosomes can be used to detect GFAP, YKL-40, EAAT1/GLAST, complement proteins, and inflammation-related miRNAs. Microglia-derived exosomes can be used to detect TREM2, TMEM119, P2RY12, CD11b, inflammatory proteins, cholesterol, and Tau-related components. Oligodendrocyte-derived exosomes can be used to detect myelin-related proteins such as MBP, MOG, MAG, PLP1, and CNPase.136 In AD and Aβ-positive MCI, plasma extracellular vesicle proteomic profiles may show early molecular alterations and correlate with CSF p-tau181, MRI indicators, and MMSE, suggesting that exosomal biomarkers may be useful for molecular subtyping in preclinical or early AD.Studies of microglia-derived exosomes have shown that in small extracellular vesicles from CD11b-positive microglia in AD brain tissue, homeostatic markers such as P2RY12 and TMEM119 are decreased, whereas TREM2, ferritin, Tau, free cholesterol, and myelin/neuronal debris–related components are increased.
These findings suggest that such vesicles may reflect the transition of microglia from a homeostatic state to a disease-associated state, abnormal lipid handling, and increased phagocytic burden.152 Oligodendrocyte-derived exosomes have been less extensively studied in AD, but the myelin proteins they carry, including MBP, MOG, MAG, and PLP1, may be promising indicators of white matter injury, oligodendrocyte stress, and remyelination status.153 It should be noted that identification of the cellular origin of exosomes remains technically controversial. For example, L1CAM has been widely used to enrich neuron-derived exosomes, but recent studies have shown that its specificity in plasma and CSF is unreliable. Therefore, studies of glia-derived extracellular vesicles also need to employ multi-marker capture, single-vesicle analysis, proteomics, and rigorous positive and negative controls to improve the accuracy of cell-of-origin determination.154,155 Overall, the major clinical significance of glia-derived exosomes lies in early diagnosis, disease-stage stratification, cell type–specific inflammatory monitoring, assessment of responses to anti-Aβ, anti-inflammatory, or pro-remyelination therapies, and potential targeted delivery.
However, current challenges remain, including isolation methods, biomarker standardization, batch-to-batch variability, and insufficient cross-cohort validation.
Conclusions and Outlook
The onset and progression of AD are complex processes jointly driven by multiple factors, including Aβ deposition, Tau pathology, neuroinflammation, metabolic dysfunction, vascular abnormalities, synaptic injury, and white matter degeneration. As key cellular populations involved in maintaining central nervous system homeostasis and mediating pathological responses, glial cells participate in multiple stages of the AD pathological cascade. Microglia can sense Aβ, Tau, myelin debris, and cellular injury signals through pattern-recognition receptors and participate in abnormal protein clearance and tissue repair during early disease stages.
However, under persistent pathological stimulation, microglia can shift toward chronic inflammation, aberrant complement activation, excessive synaptic engulfment, and immunometabolic dysfunction, thereby exacerbating neuronal injury and cognitive decline. Astrocytes are involved in neurotransmitter homeostasis, energy metabolism, lipid transport, and maintenance of the neurovascular unit, but they can also transform into reactive states driven by Aβ, Tau, and microglial inflammatory signals. Reactive astrocytes release inflammatory factors and complement molecules and disrupt the synaptic microenvironment, blood–brain barrier, and neurovascular coupling.
Oligodendrocytes and their precursor cells maintain myelin integrity, axonal conduction efficiency, and white matter network function, thereby supporting cognition. In AD, disease-associated oligodendrocyte states, OPC senescence, and remyelination failure can lead to white matter abnormalities and, in turn, promote Aβ deposition, Tau propagation, and microglial inflammatory activation. Thus, glial cells are not passive bystanders in AD pathology but central hubs linking pathological proteins, inflammatory responses, metabolic imbalance, vascular injury, and cognitive decline.
However, several major challenges remain in translating mechanistic insights into glial biology into clinical applications. First, much of the current mechanistic evidence is derived from transgenic AD mouse models and other experimental systems, which reproduce only selected aspects of AD pathology. Moreover, substantial differences exist between rodent and human glial cells in gene expression profiles, immune responses, and disease-associated cellular states.
For example, the transcriptional response of human microglia to Aβ pathology cannot be fully captured by the canonical disease-associated microglia states defined in mouse models.24 In addition, sporadic AD in humans is influenced by the combined effects of aging, APOE genotype, cerebrovascular pathology, and multiple coexisting pathologies, which are difficult to recapitulate simultaneously in most animal models. Future validation of glia-related therapeutic targets should therefore increasingly incorporate human induced pluripotent stem cell-derived models, brain organoids, human cell xenograft models, patient-derived brain tissues, and longitudinal clinical cohorts. Such approaches will be essential to avoid directly extrapolating “protective” or “pathogenic” glial states identified in animal models to human AD.
Second, single-cell and spatial omics studies have revealed substantial heterogeneity among disease-associated glial states. Distinct microglial and astrocytic subpopulations may exhibit different, and sometimes opposing, responses depending on brain region, Aβ/tau burden, and disease stage, while oligodendrocytes and oligodendrocyte precursor cells may likewise adopt functionally distinct disease-associated states. Accordingly, a key question for future research is no longer simply whether a particular glial cell type is “activated”, but rather which cellular state is involved, in which brain region and disease stage it emerges, and through which intercellular interactions it exerts its effects.
Integrating longitudinal sampling with spatial multi-omics and functional validation will be critical for defining the temporal sequence and causal relationships among glial state transitions and, ultimately, for distinguishing adaptive responses from genuinely pathogenic mechanisms. Glia-related biomarkers are expected to become important complements to precision diagnosis and treatment systems for AD. Biofluid markers such as plasma or CSF GFAP, YKL-40, sTREM2, and MBP; molecular imaging techniques such as TSPO-PET and MAO-B PET; and glia-derived extracellular vesicle biomarkers can complement traditional Aβ, p-tau, and NfL measures by reflecting astrocyte reactivity, microglial activation, myelin injury, and the spatial distribution of neuroinflammation.
Nevertheless, most currently available glia-related fluid biomarkers are neither AD-specific nor strictly specific to a particular glial cell state. For example, circulating GFAP may be elevated in a range of central nervous system injuries and neurodegenerative disorders; peripheral YKL-40 may also originate from macrophages, neutrophils, and other cell types; and TREM2 expression is not restricted to brain-resident microglia. Therefore, changes in peripheral biomarker concentrations cannot be interpreted as direct surrogates for alterations in a specific glial subpopulation within the brain.156 In addition, age, renal function, systemic inflammation, and comorbid conditions may influence the interpretation of some circulating biomarkers.
Glia-related biomarkers are therefore more likely to complement, rather than replace, core AD biomarkers such as Aβ and p-tau. Future multicenter studies with large sample sizes and longitudinal follow-up are needed to define their dynamic trajectories across disease stages. Integration of blood-based measures with CSF biomarkers, PET imaging, and multimarker models may further improve interpretation with respect to brain origin, cellular state, and disease specificity, while standardized analytical procedures, reference ranges, and clinical interpretation frameworks will also be required.
Finally, therapeutic modulation of glial function represents a promising direction for AD treatment, but future interventions should not simply aim to suppress glial activation globally. Instead, therapeutic strategies will need to account for disease stage and cellular state. During earlier disease stages, enhancing the capacity of microglia to clear Aβ and cellular debris while preserving TREM2-dependent metabolic adaptation and phagocytic function may be beneficial.
By contrast, during sustained neuroinflammatory stages, targeting NLRP3 inflammasome signaling, aberrant complement activation, or excessive synaptic engulfment may become more relevant. Importantly, current anti-Aβ immunotherapies have already provided clinical evidence that glial cells can actively participate in the effects of disease-modifying therapies. At the same time, the complex immune and neurovascular responses associated with these treatments underscore the limitations of unidirectional modulation of glial activity.
Thus, the central challenge for future glia-targeted therapy is to correct maladaptive glial states without disrupting essential homeostatic functions, including protein clearance, metabolic support, myelin maintenance, synaptic regulation, and neurovascular support.
Taken together, advances in glial biology are contributing to a conceptual shift in AD from a predominantly protein-deposition-centered model toward a more integrated model of neuron–glia–vascular network dysfunction. However, translating this framework into reliable diagnostic tools and effective therapies will depend on bridging the gap between experimental models and human disease, resolving the spatiotemporal heterogeneity of glial states, improving the specificity and interpretability of glia-related biomarkers, and identifying therapeutic windows that balance pathological modulation with preservation of physiological homeostasis. Addressing these challenges will be essential for translating glial research into clinically meaningful strategies for early detection, individualized patient stratification, and precision treatment of AD.