Microglia-mediated synaptic pruning in neural circuit remodeling: multidimensional control in homeostasis and neuropathology

Abstract

Background

Microglia, the resident immune cells of the central nervous system, are key regulators of synaptic plasticity and neural circuit homeostasis.

Main body

This review summarizes the mechanisms by which microglia shape synaptic structure and function, including dynamic synaptic interactions, selective pruning, epigenetic regulation, extracellular matrix remodeling, metabolic adaptation, and communication with other glial cells. Under physiological conditions, these processes support circuit refinement, synaptic stability, and cognition, which are modulated by circadian rhythms and the microbiota-gut-brain axis. In Alzheimer’s disease, schizophrenia, and related disorders, microglial dysfunction can shift synaptic pruning from a controlled homeostatic process to pathological synapse loss. Excessive complement-mediated pruning, disrupted excitation-inhibition balance, neuroinflammation, and metabolic dysregulation may jointly impair synaptic integrity and circuit function.

Conclusion

This review highlights microglial heterogeneity, state transitions, and targeted modulation as important directions for understanding synaptic remodeling and developing therapeutic strategies for neurological diseases.

Background

More than a century ago, Río-Hortega first proposed the concept of microglial diversity. On the basis of their transcriptomic and functional profiles, microglia can be categorized primarily into the following states: homeostatic (surveillant) state, reactive (disease-associated) state, developmental and aging-associated states, and region- and sex-specific states [1]. Under physiological conditions, they are uniformly distributed in the central nervous system (CNS), which has a dendritic morphology, and their ability to continuously survey the microenvironment and the activity of neurons and synapses through immune surveillance and neuroprotective support. Upon activation, they undergo a “morphology-function” cascade transition (from dendritic to hypertrophic and amoeboid forms), migrate to lesion/injury sites, and release cytokines and inflammatory mediators to participate in inflammatory responses and neural repair [2, 3]. In vivo two-photon imaging in CX3CR1-GFP mice directly demonstrated this dynamic surveillance, revealing microglia as actively patrolling cells that rapidly extend their processes toward sites of local injury (structural evidence, CX3CR1-GFP mouse) [4]. However, as research has progressed further, our understanding of these resident immune cells of the CNS has undergone a profound evolution—their functions extend far beyond the traditionally perceived role of immune surveillance.

Microglia are not “passive bystanders” but “active participants” in the regulation of synaptic plasticity. Seminal work by Beth Stevens and colleagues established the classical complement cascade (complement component 1q-complement component 3) (C1q-C3) as a key “eat‑me” signaling system that tags weak synapses for elimination. Microglia expressing the complement receptor complement receptor 3 (CR3) subsequently engulf these tagged connections, thereby refining neural circuits with remarkable precision (structural and functional evidence, C1q knockout mice and C3 knockout mice) [5]. In adulthood, microglia serve as indispensable “synaptic architects” in learning and memory by precisely shaping the structure and plasticity of neural circuits. They achieve this through maintaining the number and function of glutamatergic synapses in the hippocampal CA1 region, curbing excessive long-term potentiation, promoting synaptic regeneration, and remodeling the extracellular matrix (ECM). This finely tuned regulation relies on the chemokine (C-X3-C motif) ligand 1/chemokine (C-X3-C motif) receptor 1 (CX3CL1/CX3CR1) signaling axis and involves the secretion of trophic factors that help establish a microenvironment conducive to synaptic maturation [6].

In chronic neurological diseases or under sustained immune stimulation, microglia become persistently hyperactivated, releasing proinflammatory and neurotoxic mediators that exacerbate neuroinflammation and neurodegeneration [7]. Under disease conditions, the synaptic regulatory function of microglia changes [8]. Using a series of knockout and antibody blockade experiments in an Alzheimer’s disease (AD) mouse model, Soyon Hong and colleagues reported that Aβ oligomers aberrantly activate the complement system, triggering erroneous synaptic engulfment by microglia (microglial phagocytosis is the process by which these cells recognize targets—such as synapses, neuronal debris, and protein aggregates—through “eat-me” signals displayed on the surface of those targets. Following recognition via surface receptors, microglia then engulf and clear these materials, leading to synapse loss and cognitive impairment even before plaque formation. The core pathological consequence of this microglia-mediated disruption of synaptic homeostasis is a breakdown of the excitatory/inhibitory (E/I) balance, a central pathogenic mechanism in multiple neurological disorders, including Alzheimer’s disease, multiple sclerosis, autism spectrum disorder, and schizophrenia [9, 10].

Previous reviews (e.g., Cornell et al, 2022 [2]; Andoh & Koyama, 2021 [11]) have focused mainly on specific mechanisms during development or adulthood, such as complement‑mediated pruning, often presenting regulatory modules in parallel without causal ordering or evidence grading. They also paid less attention to metabolism, epigenetics, ECM remodeling, circadian rhythms, and the gut-brain axis. In contrast, the present review introduces a driver-context-outcome framework that organizes eight regulatory modules—morphology, apoptosis, epigenetics, metabolism, ECM, glial crosstalk, circadian rhythm, and the gut-brain axis—into a causal hierarchy of execution → regulation → coordination → integration. For each mechanism, we explicitly assign evidence tiers (perturbational in vivo, ex vivo/in vitro, human association/omics, conceptual extrapolation), and we systematically incorporate the emerging areas mentioned above. This shift from a parallel catalog of mechanisms toward a testable synthesis provides a unified conceptual framework for understanding the dual role of microglia in maintaining synaptic homeostasis versus driving pathological synapse loss.

Literature search and evidence prioritization

The PubMed, Web of Science, and Google Scholar databases were searched for literature published between January 2012 and December 2025 using a strategy that combined MeSH terms and keywords across three thematic domains: microglia (“Microglia” [Mesh] OR “microglia”), synaptic regulation (“Synaptic Pruning” [Mesh] OR “Synaptic Plasticity” [Mesh] OR “Neuronal Plasticity” [Mesh]), and disease contexts (“Neurodegenerative Diseases” [Mesh] OR “Alzheimer Disease” [Mesh] OR “Parkinson Disease” [Mesh] OR “Huntington Disease” [Mesh] OR “Schizophrenia” [Mesh] OR “Autism Spectrum Disorder” [Mesh]), with additional free-text terms including “Extracellular Matrix”, “Epigenetics”, “Metabolic Reprogramming”, “Circadian Rhythm” and “Gut-Brain Axis”. To align claim strength with evidence, studies were prioritized according to the following hierarchy: in vivo perturbational studies (genetic knockout, pharmacological intervention) for mechanistic claims; human-derived evidence (postmortem, genetics, and iPSC models) for translational statements; and seminal pre‑2012 studies incorporated by expert selection as a foundational background and recent literature Titles and abstracts were screened by the authors, with full-text review conducted for studies meeting these prioritization criteria.

Main text

Microglial regulation of synaptic plasticity under physiological and pathological conditions

During early brain development, microglia depend on the colony-stimulating factor 1 receptor (CSF1R) signaling pathway, the transforming growth factor-β (TGF-β)-dependent signaling pathway, and the regulation of transcription factors such as PU.1 and interferon regulatory factor 8 (IRF8), as well as miR-124, for proper maturation (genetic evidence, mouse) [12, 13]. These signals not only shape the molecular identity of microglia but also determine their functional states. Microglia originate from yolk sac-derived progenitors, migrate into the developing nervous system, and maintain population homeostasis through lifelong self-renewal, laying the foundation for their later role in regulating neural circuit plasticity [1]. The distribution and functional states of microglia are regulated by multiple factors, including developmental stage, brain region specificity, and the local microenvironment, leading to significant heterogeneity that is particularly prominent during development and in humans [14, 15].

During development, microglia display distinct transcriptional profiles that support their role in synaptic pruning (synaptic pruning refers to the selective elimination of unused or excess synapses that occurs during brain development, following a phase of overproduction). In humans, primary visual cortex pruning is completed between the 4th and 6th years of life, whereas in areas involved in complex cognitive functions, such as the prefrontal cortex, synaptic pruning often continues until the end of adolescence [16]. In adulthood and during aging, microglia retain a transcriptional state that enables continuous surveillance and modulation of the synaptic environment [17]. Transcriptomic analysis has revealed pronounced regional heterogeneity in adult mouse microglia across brain areas, including the cerebellum, hippocampus, cortex, and striatum, and has revealed differential sensitivity to the aging process (transcriptome analysis, but no functional validation has been conducted; mouse) [18]. Their developmental program also enables them to sense neuronal activity, allowing them to dynamically regulate the synaptic stability and plasticity that underlie advanced cognitive functions [13].

Microglia participate in processes related to postnatal brain development and neuronal plasticity. Although previous studies have shown that microglia promote the refinement of neural circuits and the maintenance of homeostasis by phagocytosing apoptotic neurons, the direct and consistent causal role of microglia in different neural system regions remains unclear and requires further investigation [19]. Using high-resolution techniques such as light-sheet fluorescence live imaging and focused ion beam scanning electron microscopy (FIB-SEM), Weinhard et al [20] provided the first direct visual and quantitative evidence that microglia remodel synapses via two distinct mechanisms: “synaptic precontact” and the induction of “head-shrinking filopodia”. This work has reshaped our understanding of microglial functions in neural circuit development and plasticity (structural and functional evidence, organotypic hippocampal slice culture, triple transgenic mouse). Emerging evidence challenges the long-held dogma that microglia are indispensable for neural development. In microglia-deficient mice from conception (e.g., Csf1rΔFIRE/ΔFIRE mice), despite certain embryonic brain abnormalities, postnatal neural development and fundamental network functions (e.g., seizure susceptibility and hippocampal-prefrontal synchrony) progress normally. These findings indicate that basic synaptic refinement can occur independently of microglia (genetic evidence, Csf1r∆FIRE/∆FIRE mice) [21–23]. Notably, these findings underscore their essential protective role under pathological conditions. Under homeostatic conditions, microglial populations are primarily maintained through stochastic self‑renewal. However, upon acute neural injury or other pathological insults, they rapidly shift toward clonal expansion, resulting in the formation of discrete cellular clusters—a core manifestation of their specific, adaptive response to damage [24]. Although microglia contribute to synaptic pruning in certain systems, they appear to be dispensable for the dendritic refinement of mitral cells in the olfactory bulb and layer 4 cortical neurons, where no significant changes in the number, density, or size of excitatory synapses are observed [25]. These findings indicate that the role of microglia varies across neural circuits and developmental stages and that dendritic and synaptic refinement in some regions may rely principally on neuron‑intrinsic programs or other cellular regulators rather than on microglial phagocytosis. Using mice with microglia‑specific SIRPα knockout and Alzheimer’s disease models, the study revealed that microglia suppress synaptic pruning via the SIRPα-CD47 “don’t eat me” signal. Loss of this signal leads to excessive microglial phagocytosis of synapses, resulting in reduced synaptic density and cognitive impairment [26] (genetic evidence from mice). During critical-period plasticity and under pathological conditions, microglia protect synapses by recognizing this signal. SIRPα deficiency also disrupts the excitatory/inhibitory (E/I) balance, as reflected by a decrease in the miniature excitatory postsynaptic current (mEPSC) frequency [26]. Together, these findings highlight the essential role of microglia in maintaining synaptic homeostasis.

The fine regulation of synaptic plasticity is achieved through the coordination of multiple mechanisms:

1. Direct physical contact: By combining in vivo two-photon imaging with 3D electron microscopy reconstruction, researchers have shown that microglial processes establish highly dynamic, direct physical contact with presynaptic terminals and postsynaptic dendritic spines. Under inflammatory or neurodegenerative conditions, these contacts are markedly prolonged, leading directly to spine remodeling (e.g., filopodium formation), physical pulling, and ultimately synapse elimination (in vivo imaging, transgenic mouse models, and pharmacological intervention) [27]. Moreover, molecules such as Rac1 also participate in modulating the frequency of microglia-synapse interactions [14] (genetic evidence from mice).

2. Secretory signal modulation: Through the release of signaling molecules such as chemokines (e.g., chemokine (C-X3-C motif) ligand 1), cytokines (e.g., TNF-α, IL-1β, and IL-33), and neurotrophic factors (e.g., brain-derived neurotrophic factor), microglia modulate synaptic function and plasticity in a paracrine or exosome-mediated manner. These secreted factors can directly act on neuronal receptors to regulate synaptic strength, spine dynamics, and long-term potentiation/depression (LTP/LTD) or indirectly shape the synaptic microenvironment by influencing astrocyte function and extracellular matrix composition [28–30].

3. ECM remodeling: Microglia actively reshape the extracellular matrix (ECM), particularly perineuronal nets (PNNs), by secreting proteases such as matrix metalloproteinases (MMPs). This proteolytic activity not only removes physical barriers to synaptic growth but also liberates sequestered growth factors and alters the biochemical landscape of the synaptic cleft. The degradation of aggrecan, a key PNN component, reduces structural inhibition and reopens critical periods of plasticity, thereby facilitating experience-dependent synaptic remodeling and memory consolidation [31].

4. Multiomics cross-talk: Recent multiomics studies have revealed extensive crosstalk between microglia-dependent synaptic phosphoproteomes and local transcriptomes, resulting in the formation of a complex intercellular signaling network.

In summary, through the integrated actions of direct physical contact, secretory signaling, and ECM remodeling, microglia not only initiate plasticity but also function as dynamic modulators. They precisely calibrate synaptic connection strength, circuit stability, and network adaptability, playing indispensable roles in both developmental refinement and the ongoing homeostatic maintenance of functional neural circuits.

Role of microglial morphological characteristics in regulating synaptic plasticity: structural execution

The morphological plasticity of microglia is a fundamental determinant of their functional state and a critical interface for their regulation of synaptic connectivity [32]. Research has shown that upon inflammatory stimulation, microglial morphology reversibly shifts from a ramified state to an amoeboid state. This transition depends on molecules such as Ras-related C3 botulinum toxin substrate 1 (Rac1) and Ras gene family member A (RhoA) and is accompanied by the relocalization of F-actin from cellular processes to the soma. Moreover, morphological changes directly determine phagocytic efficiency, confirming that dynamic and reversible shifts in microglial shape reflect a functional switch from environmental surveillance to phagocytic clearance (in vivo imaging, transgenic mouse models, and pharmacological intervention) [33]. This section describes the principal morphological states of microglia, examines how dynamic morphological changes underpin physiological synaptic regulation, and explores how pathological transformations contribute to synaptic dysfunction.

Morphological states and their functional correlates

As demonstrated by in vivo two-photon imaging [4], microglial processes constantly extend and retract, making brief contacts with nearby neurons, astrocytes, and blood vessels, thereby acting as dynamic sentinels that maintain homeostasis in the brain (in vivo) [34]. Microglia display diverse morphological responses under different pathological conditions. Their observed transcriptional heterogeneity reflects functional specialization, often linked to specific structural changes. However, the translation of transcriptional profiles into functional outcomes requires downstream execution through processes such as protein synthesis and cytoskeletal reorganization, which are dynamically regulated by local environmental cues [1]. Moreover, the structural complexity of microglial processes, particularly their length and degree of branching, is strongly correlated with their ability to monitor the synaptic environment. When this complexity declines, surveillance function is often compromised, potentially limiting their ability to fine‑tune the synaptic microenvironment [35, 36].

This structure-function relationship may be embodied in specific microglial subpopulations, such as disease‑associated microglia (DAMs). DAMs exhibit a distinct transcriptional profile and are thought to influence selective synapse elimination by modulating phagocytic activity and metabolic state. However, the precise causal role of these transcriptional changes in driving synaptic pruning remains to be directly established (transcriptome analysis, but has not yet been directly confirmed to have a causal role in the process of synaptic elimination; mouse) [1, 9, 37]. CYFIP1 deficiency markedly impaired synaptic phagocytosis and migration speed and increased travel distance in human microglia while also shifting their morphology from an amoeboid state to a more ramified, resting‑like state (extracellular cells) [38]. Together with the known disease associations—excessive synaptic pruning in schizophrenia versus insufficient pruning in autism—these observations suggest that microglia may exhibit context‑dependent, bidirectional functional regulation during development and in disease. Notably, although specific morphologies (e.g., ramified or amoeboid) provide a physical basis and functional bias for distinct roles, such as immune surveillance or phagocytosis, the relationship between form and function does not constitute a strict one‑to‑one instruction [38]. As summarized in Table 1, the morphological features of microglia and their functional implications should be interpreted as a summary of morphology-function correlations under specific conditions rather than as static or deterministic relationships.

Table 1.

Morphological characterization and functional significance of microglia

Morphological Type	Location	Morphological Features	Function	Mechanisms of synaptic plasticity	Quantitative results and their influence	Significance	Evidence Tags
Surveillant microglia	Cerebral cortex; Hippocampus (especially the dentate gyrus); Cerebellum; Corpus callosum (white matter area); Lateral ventricle area; Dense synaptic area (whole brain); Brain vascular network; Retina (during development) [234].	Cell body; Primary/secondary branches; Filamentous pseudopods [234].	During the developmental stage, precise neural circuits are shaped (such as regulating cell death and synaptic pruning), while in adulthood, the network is maintained in a stable state and damage is prevented [234].	Through complement-dependent synaptic labeling, chemokine-mediated recognition, dynamic contact associated with activity, and ultimately phagocytic clearance (pruning via phagocytosis) [234].	By conducting synaptic contacts that last for 4-5 min each time and occur approximately once per hour, synaptic pruning and neuronal excitability are regulated. Dysfunction of this process can lead to a significant decline in cognitive flexibility as well as social and motor behavioral impairments [234].	Plays a key supportive role in the normal development and function of microglia-dependent brain regions under most physiological conditions; however, certain aspects of postnatal synaptic development can proceed independently of microglia (e.g., in Csf1rΔFIRE/ΔFIRE mice) [234].	Genetic methods;Genetic methods; Pharmacological methods; In vivo studies; Human genetics research; In vitro studies; Omics research [234].
Hyperramified microglia	Prefrontal cortex; Hippocampus [235].	The number of branches has significantly increased; the number of branch points (nodes) has also increased; the length of the branches may have increased; the shape of the cell body remains relatively unchanged [235].	Stress response and regulation; Regulation of synaptic plasticity; Promotion of pathological conditions; Enhanced potential monitoring function [235].	Indirect signaling pathway: Release of soluble mediators (such as cytokines/cytokines); Potential direct structural intervention: Through physical contact (such as enhanced synaptic monitoring; participation in synaptic pruning or stabilization) [235].	It is closely related to reduced synaptic density, increased colocalization of synaptic proteins (indicating excessive pruning), imbalance of the dopamine pathway, and schizophrenia-related behaviors (PPI, LI impairment, amphetamine hypersensitivity) [236].	The “bridge” between neuroimmune abnormalities and synaptic dysfunction [235].	Intracellular research; Pharmacological research; Genetic/Geneticist research; Study on the Relationship between Behavior and Form [235].
Hypertrophic microglia	In the brain regions where neurodegenerative disorders, pathological protein aggregation, chronic inflammation or tissue damage occur, their location is dynamic and dependent on the pathology [237].	The cell body has significantly enlarged. The branch has changed, but it has not completely disappeared [237].	By enhancing immune surveillance and maintaining moderate phagocytic activity in response to pathological stimuli, it is thought to maintain an active state. However, this persistent activation state may also lead to synaptic dysfunction and exacerbate neuroinflammation [237].	Through pathways such as the complement system, abnormal synaptic phagocytosis occurs; pro-inflammatory mediators that damage synaptic function are continuously released; at the same time, the neurotrophic support that is crucial for synaptic maintenance is reduced [237].	In elderly rhesus monkeys with cortical injury, EV treatment increased the proportion of C1q-positive large-sized microglia in the M1 area from 10-15% to 25-30%, and its density was significantly negatively correlated with the number of days for motor function recovery [238].	The early and dynamic visualization of the imbalance of immune homeostasis in the brain under aging or pathological stress represents both an adaptive response marker for the nervous system’s attempt to maintain function and a turning point where there is a potential shift toward chronic inflammation and nerve damage [237].	Ultrastructural studies using electron microscopes [237].
Amoeboid microglia	Mainly focused on the corpus callosum during the early developmental stage [239].	Lack of complex branches; large and round cell bodies; pseudopod-like structures [239].	Maintaining developmental homeostasis is coordinated with potential immune surveillance and synaptic pruning, jointly shaping neural circuits to engulf oligodendrocyte precursor cells to regulate myelin formation [239].	Regulating myelin sheaths indirectly is associated with synaptic pruning (via phagocytosis of oligodendrocyte precursor cells) to coordinate white matter maturation and gray matter circuit refinement [239].	After H7N7 virus infection, the proportion of amoeboid microglia in the hippocampal cultures of male mice significantly increased, and they preferentially phagocytosed postsynaptic components. In females, there was an increase in dendritic spine density and functional synaptic density, but none of these were associated with behavioral outcomes [240].	Regulating myelin sheaths, by inducing the programmed phagocytosis of healthy oligodendrocyte progenitor cells, precisely regulates myelin formation, thereby establishing an efficient electrical signal transmission foundation for neural networks. This is the key “development engineer” for synchronously coordinating the construction of white matter and the optimization of gray matter circuits [239].	Genetic methods; Pharmacological methods; In vivo studies; In vitro studies [239].
Rod-shaped microglia	Cerebral cortex; Hippocampus; There have also been reports of its presence in regions such as cerebellar cortex and brainstem, especially in pathological conditions (such as multiple sclerosis, encephalitis, etc.) [241].	Long and slender rod-shaped; with protrusions at both ends; No increase in length [241].	As a specialized state of microglia cells that are specifically designed to “manage” the damaged neural microenvironment in the context of brain injury, aging or disease [241].	Perhaps the microglia play a highly specialized and spatially precise role in the execution of synaptic plasticity. The mechanism is likely to integrate various approaches such as physical isolation, precise phagocytosis, and local secretion of signal transduction [241].	In the early stages of TDP-43-related neurodegenerative diseases (such as ALS), approximately 70% of the rod-shaped microglia cells come into contact with the dendrites of neurons, especially with the apical dendrites of the pyramidal neurons in the fifth layer, and approximately 50% of the rod-shaped microglia cells interact with these dendrites, playing a protective role in inhibiting excessive cortical activity [242].	Understanding the key windows of the neural-immune dialogue, inflammatory regulation, and repair and remodeling mechanisms in the brain under conditions of injury, aging, and disease [241].	Histological and morphological observations; In vivo study; In vitro study; Human brain tissue study [241].
Degenerative/Dark microglia	Hippocampus; Cerebral cortex; Striatum Retina; White matter area [243].	The cytoplasm and nucleus are highly electron-dense and overall appear “dark” [243].	During the developmental stage, the function of dark microglia is characterized by “highly stressed phagocytosis”, which is the price paid for performing necessary but arduous tasks; in diseases, however, it may become out of control, transforming from the “gardener” of the neural network to a “destroyer”, becoming a key link in driving the progression of the disease [243].	Through “structural synaptic pruning”, neural circuits are restructured. This pruning is physiological during the developmental stage, but may become dysregulated in pathological conditions and lead to the loss of synapses [243].	By activating the ISR of microglia through chemical genetics, the density of dark microglia increased approximately fourfold (approximately 67 cells/mm), accompanied by a significant increase in synaptic loss and deterioration of tau pathology; inhibition of the ISR almost eliminated the dark microglia (approximately 3 cells/mm) and restored the number of synapses [243].	Establish the intrinsic stress pathways within cells as the core driving factors for synaptic loss and cognitive impairment in neurodegenerative diseases [243].	Ultrastructural research (core identification method); Intracellular research; Genetic model; Human genetics and pathology [243].

Footnote: This table summarizes the predominant morphological types of microglia observed across various physiological and pathological contexts, along with their commonly associated functional features. The morphology-function relationship is dynamic; the same morphological state may participate in distinct functional programs depending on the local microenvironment, and the precise functional output (e.g., the specificity of pruning) relies on finely regulated molecular signals

Morphological dynamics in physiological synaptic regulation

Confocal and electron microscopy have revealed that microglial processes can engage in trogocytosis—the selective nibbling of small fragments of presynaptic axons containing synaptic vesicles—through rapid, localized events. This mode of action is distinct from canonical phagocytosis, which typically involves engulfment of larger or intact synaptic structures and highlights the capacity of microglia for precise, graded regulation of synaptic components (structural evidence from mice). As detailed in the Introduction and Section State transitions and functional polarization of microglia in neurodegenerative diseases, microglia utilize the classical complement pathway (C1q-C3-CR3) and “eat-me” signals such as phosphatidylserine (PS) to tag and eliminate dysfunctional synapses, whereas “don’t-eat-me” signals such as CD47-SIRPα (Cluster of Differentiation 47-Signal Regulatory Protein Alpha) protect active synapses from erroneous removal [39, 40]. Compared with microglia in the dorsal hippocampus, microglia in the ventral hippocampus exhibited more elaborate processes; greater numbers of immature lysosomes; higher expression of Cx3cr1, Mertk, CD68, and IL-6; lower levels of IL-1β and TNF-α; and greater outward potassium currents. Conversely, the opposite profile was observed for dorsal microglia. Moreover, microglia from the two poles oppositely modulate LTP [41]. These findings reveal regional heterogeneity in microglial process properties, where differences in gene expression, protein composition, and functional activity shape immune-related functions, motility, and phagocytic capacity, thereby directly influencing process dynamics and behavior [40].

Morphological transformation in pathology and synaptic dysfunction

Under pathological conditions such as neural injury, inflammation, or neurodegenerative disease, microglia transition into a reactive or disease-associated state, often resulting in the development of an amoeboid or hypertrophic morphology. This shift enhances their phagocytic and inflammatory capabilities, yet its functional consequences are highly context dependent. While moderate activation can promote debris clearance and may initiate repair mechanisms, sustained or excessive activation is frequently associated with synaptic dysfunction [37], which is reflected in two principal ways: first, it may lead to excessive, nonselective synaptic pruning, thereby diminishing functional connectivity (structural and functional evidence, CSDS mouse model, C3 knockout mice, two-photon imaging) [42]; second, the release of high levels of proinflammatory cytokines such as IL‑1β and TNF‑α can directly or indirectly affect synaptic plasticity and neuronal health, potentially disrupting the functional stability of neural networks (structural and functional evidence, 5xFAD mouse, electrophysiology) [43]. In a P301S tauopathy mouse model, pathological tau triggered complement-dependent synapse tagging, leading to excessive engulfment by microglia and astrocytes and reducing synapse density by half (animal experiments) [44]. These changes drive microglia into a reactive, altered state, where they engage in overactive synapse pruning. The result is damaged synapses and compromised plasticity. Thus, pathological morphological alterations shift microglial function from precise homeostatic regulation toward dysregulated synaptic clearance, a process closely linked to synaptic loss and cognitive decline (Table 2). A deeper dissection of the relationships between microglial morphology and function, as well as their underlying regulatory mechanisms, is therefore critical for unraveling the complex regulatory networks of the central nervous system and for identifying novel therapeutic strategies for neurological diseases.

Table 2.

Roles and significance of molecules regulating synaptic pruning

Regulatory Molecule	Effect on Synaptic Pruning	Significance	Relationship with microglia	Evidence Tags
C1q	Enhancement [244]	Converting neural activities into immune signals; the basis for achieving selective synaptic pruning via complement-mediated phagocytosis; Maintaining the balance between excitation and inhibition [244].	The relationship between C1q and microglia is a typical cooperative partnership between a “signal reporter” and an “executioner of clearance”, jointly forming the core engine of immune-mediated synaptic pruning [244].	Genetic research; In vivo study; In vitro study [5].
C3	Enhancement [244]	Signal conversion and amplification; The foundation for achieving efficient and selective clearance; The key link that connects innate immunity with neural development [244].	The direct “ligand-receptor” relationship is the final transmission of the clearance instruction [244].	Genetic research; In vivo study [5].
CX3CL1/CX3CR1	Enhancement [244]	Provide nonphagocytic regulatory interfaces; realize the “active summoning” and “bidirectional communication” of neurons with microglia. Separation of functions from the classical immune functions [244].	Highly specific ligand-receptor pairs; regulating the “working state” of microglia cells; Regional and task-specificity [244].	Genetic research; In vivo study [245].
CD47	Inhibition [244]	The essential elements for achieving “selective” removal, preventing excessive trimming and accidental injury; The “molecular balance” that maintains the stability of neural circuits [244].	It is the “identity confirmation” and “self-protection” signals actively transmitted by neurons/synapses to microglia [244].	Genetics, in vivo experiments and in vitro experiments [246].
SRPX2	Inhibition [244]	Provide the “local fine-tuning knob” for synaptic pruning; maintain the dynamic balance between “promoting pruning” and “anti-pruning” signals. It demonstrates the active control that neurons have over their own destinies [244].	Indirect “remote control” achieved by regulating the microenvironment [244].	Genetics, in vivo experiments and in vitro experiments, human genetic evidence [247].
MHC I	Regulatory, rather than direct, clearance markers [244]	Transform the immune logic into the neural development logic; Connect functional plasticity with structural pruning; Provide the autonomous pruning regulatory ability of neurons [244].	It is indirect and not essential. There may be synergy among them, but they are not the main pathways of action [244].	Genetics, in vivo experiments, human genetics and histology research, in vitro experiments, pharmacology, transcriptional Regulation and Epigenetics Research [248].
GPR56	Enhancement [244]	Provided an “alternative” or “synergistic” clearance pathway beyond complement; Linked local “apoptotic-like” signals directly to synaptic pruning; Achieved higher temporal and spatial precision in the trimming regulation [244].	Microglia act as “molecular antennas” that sense “danger” or “removal” signals in the local microenvironment [244].	Genetic methods; In vivo research methods; Cell type-specific analysis [249].
Caspase 3	Enhancement [244]	Establish a molecular bridge connecting functional weakening and structural elimination;Achieve the precision guarantee for “targeted elimination”. It is the common downstream executor of multiple upstream signals [244].	It is the upstream-downstream relationship between “signal preparers” and “signal responders” [244].	Genetic methods [250].
TREM 2	Enhancement [244]	As a “multifunctional sensor”, it integrates complex signals; connects the metabolism and phagocytic function of microglia; and acts as a bridge between physiological and pathological pruning [244].	Highly specific expression; Functional master switch; The interaction hubs with astrocytes [244].	Genetic methods; In vivo studies [251].
MEGF10/MERTK	Enhancement [244]	Established the active pruning function of astrocytes; Revealed the “cooperative division of labor” pruning pattern; Provides redundancy and robustness of the clearing mechanism [244].	Independent but parallel pathways; competition or coordination among cells; differentiated phagocytic patterns [244].	Genetic methods; In vivo studies [252].
Phosphatidylserine	Enhancement [244]	As a convergence point for multiple upstream pathways; achieving “local death” without harming the whole. Provide a universal language across cell types [244].	It is the direct molecular communication between “key ligands” and “main effectors”, which is one of the most direct bases for microglia to recognize and eliminate synapses [244].	In vivo experiments; Ex vivo experiments; Human genetics and histology [58].
IL-33	Inhibition [244]	Provide examples of bidirectional glial cell communication; act as a “buffer” for the stability of neural circuits; “train” the functions of cytokines into stable regulatory mechanisms [244].	It is the relationship between “inhibitory regulators” and “functional executors”, which is a top-down regulatory instruction [244].	Genetic methods; Pharmacological intervention; In vivo experiments; Ex vivo experiments; transcriptome and single-cell RNA sequencing [253].

The relationship between microglial morphology and synaptic pruning is supported by multiple lines of experimental evidence, including in vivo genetic perturbation (e.g., Rac1 knockout [14]), in vivo observational studies (two‑photon imaging in live mice capturing real‑time microglial dynamics [4]), in vitro observational and perturbation experiments (real‑time imaging combined with genetic labeling in acute brain slice cultures to visualize dynamic microglia-synapse interactions, followed by knockout or pharmacological interventions for validation [20]), and conceptual inference (integrating structural evidence from mouse electron microscopy with transcriptomic and regional heterogeneity analyses of human postmortem tissue to derive overarching conclusions about experience‑dependent brain development [1, 16])

Microglial regulation of apoptotic pathways in synaptic plasticity: clearance execution

Microglia function beyond mere scavengers; they recognize “apoptotic/apoptosis‑like signals”, trigger phagocytosis and cellular reprogramming, and thereby selectively eliminate or modulate synapses, influencing synaptic plasticity and network stability from both structural and functional perspectives. During neural development and homeostasis maintenance, microglia achieve selective synaptic pruning and plasticity regulation through the coordinated action of local apoptosis‑like signaling and the complement system, providing a critical molecular basis for understanding synaptic pathology in neurodevelopmental and related disorders.

Regulation of apoptosis under physiological conditions and precise synaptic pruning

Overall neuronal apoptosis clearance and circuit remodeling Wholesale neuronal apoptosis refers to the selective elimination of redundant or aberrant neurons via programmed cell death mechanisms during nervous system development and maintenance. This process is particularly critical during embryonic and postnatal neurodevelopment, ensuring that only neurons of appropriate size, normal morphology, and proper axonal/synaptic connections survive, thereby establishing and preserving normal neural network architecture (structural and functional evidence, mouse knockout models) [45]. In the adult dentate gyrus (DG), ramified microglia clear apoptotic newborn granule cells efficiently. Each microglial cell removes 8–20 dying cells per day, completing the process from initial contact to full degradation within 3–6 h. This rapid clearance reflects microglial responsiveness to apoptosis and supports their role in the refinement of neural circuits (animal in vivo experiments) [46]. Programmed neuronal apoptosis releases nucleotide signals such as adenosine triphosphate (ATP), which recruit microglia to the brain. This nucleotide-mediated chemotaxis not only regulates microglial number and spatial distribution but also provides positional guidance for their localization within the brain (structural and functional evidence, zebrafish, pharmacological and genetic inhibition) [47]. Such a mechanism reflects a functional‑demand‑based mode of cellular organization during development, akin to how cells in other tissues respond and localize according to functional cues. In support of this concept, two molecularly and functionally distinct microglial subpopulations have been identified in the zebrafish brain: “synapse‑associated microglia” residing in synaptic‑rich regions and “neurogenesis‑associated microglia” located in neurogenic zones. SAMs are enriched in the hindbrain and highly express synaptic regulatory genes, including those of the complement system, indicating a specialized role in synaptic pruning and remodeling. In contrast, NAMs display a spherical morphology and exhibit active caspase activity, which is primarily responsible for clearing apoptotic neurons. Functionally, both subpopulations possess phagocytic capacity; however, SAMs mediate synaptic dynamics by modulating the extracellular matrix and complement system, whereas NAMs rely on an active lysosomal system to efficiently remove apoptotic cells (structural and functional evidence, zebrafish, single-cell transcriptomics, and in situ imaging) [48].

Identification and precise trimming of local “apoptotic-like” events at synapses Localized neuronal apoptosis refers to programmed cell death occurring within specific subcellular compartments of a neuron. This spatially restricted regulation of apoptosis plays a key role in nervous system development, synaptic pruning, and responses to pathological states [49].

Molecular and electrophysiological studies have shown that hippocampal LTD and spine pruning require local nonapoptotic caspase-3 activation. Blocking this enzyme in Parkinson’s disease models reverses both spine loss and LTD deficits. Reaction-diffusion modeling and microfluidic experiments further revealed that the narrow neck of dendritic spines physically blocks apoptotic trigger waves, keeping them localized (Theoretical/computational model; Ex vivo experiment) [50]. The clearance mechanism relies heavily on the complement system: once a synapse initiates an apoptosis‑like program—such as caspase‑3 activation and phosphatidylserine exposure—the complement initiator protein C1q binds to the presynaptic membrane, forming a phagocytic tag. Microglia then recognize these labeled synapses via complement receptors on their surface and initiate phagocytosis, thereby achieving activity-dependent selective elimination (structural and functional evidence from mice and humans) [51, 52]. Furthermore, apoptotic dendrites release “find‑me” signals (e.g., ATP, glutamate, and calcium overload) and “eat‑me” signals (e.g., phosphatidylserine, MHC class I proteins, and complement components C1q and C3), which attract microglia and promote phagocytosis. Conversely, “don’t‑eat‑me” signals such as CD47 protect synapses from excessive pruning (structural and functional evidence, mouse, in vitro) [53].

The intrinsic apoptotic balance of microglia Microglia themselves undergo programmed cell death, a critical mechanism for maintaining their population homeostasis and preventing excessive activation and neuroinflammation. Neuronal CX3CL1-CX3CR1 signaling and TREM2 signaling coordinately regulate microglial survival and apoptosis, fine-tuning their functional state (structural and functional evidence from mice and humans) [54]. The polarized state of microglia profoundly influences their regulation of apoptosis and synaptic plasticity. A neuroprotective/homeostatic phenotype promotes neuronal protection and repair, improves neuronal survival, inhibits apoptosis, and enhances synaptic remodeling [55]. For instance, the cytokine IL‑33 can drive microglia toward this neuroprotective/homeostatic phenotype, thereby increasing the clearance of neurotoxic substances such as Aβ and reducing neuronal apoptosis (structural and functional evidence from mice in vitro) [56]. The sequential activation of caspase-8 and its downstream effector caspase-3 is regulated via a PKC-δ-dependent pathway, which modulates the activation state of microglia. Excessive activation of this signaling cascade is closely associated with neurotoxicity. Under nonapoptotic conditions, caspase-3 is involved in the regulation of synaptic function and plasticity. Further studies have demonstrated that enhanced nonapoptotic activity of caspase-3 is significantly correlated with early synaptic dysfunction in mouse models of Alzheimer’s disease. These findings suggest that caspase-8 may influence synaptic plasticity mechanisms underlying learning and memory through its regulation of caspase-3 activity (structural and functional evidence from mice) [57].

Dysregulation of the apoptotic pathway under pathological conditions and excessive loss of synapses

Under pathological conditions, this regulatory network shifts from precise trimming to excessive clearing, driving extensive synaptic loss and cognitive impairment.

Aberrant “apoptosis-coupled” synaptic clearance: Under pathological conditions such as Aβ oligomers, abnormal tau, or chronic inflammation, excessive or aberrant activation of apoptosis‑like signaling can occur at neuronal synapses. Although the initial trigger lies within intrinsic neuronal alterations, microglial responses to these signals often become dysregulated. As shown in a real-time imaging study [58], Aβ oligomers induce PtdSer exposure on dendritic spines, resulting in microglial engulfment (73% of phagocytosed synapses are PtdSer positive), indicating that exposure to apoptotic “eat‑me” signals serves as a key marker of selective microglial synaptic engulfment. Enhanced recognition and response of microglial surface receptors (e.g., CR3) to C1q-tagged signals lead to indiscriminate phagocytosis of synapses that remain functional or are only mildly impaired,, resulting in excessive synaptic loss, disrupting the physiological balance between clearance and preservation. Consequently, widespread impairment of neural network connectivity and cognitive decline ensue (genetic evidence, C1q ^−/− and CR3^−/− mice; pharmacological evidence)[52]. This apoptotic regulatory network not only participates in developmental synaptic pruning but also becomes aberrantly activated in neurodegenerative diseases, driving excessive synaptic loss and cognitive impairment (Fig. 1).

Fig. 1.

Microglial apoptotic pathways regulate synaptic plasticity. The extrinsic and intrinsic apoptotic pathways converge on caspase-3 activation. The extrinsic pathway is initiated by death ligand-receptor binding, which activates caspase-8. The intrinsic pathway is triggered by intracellular signals such as DNA damage, which modulate the expression of Bcl-2 family proteins, leading to mitochondrial cytochrome c release and caspase-9 activation. Both pathways converge to activate caspase-3, initiating apoptosis. Activated microglia release proinflammatory factors (e.g., IL-1β) and induce endoplasmic reticulum stress, which disrupts calcium homeostasis, alters Bcl-2 family regulation, increases mitochondrial membrane permeability, promotes cytochrome c release, elevates reactive oxygen species, and activates initiator caspases, ultimately driving neuronal apoptosis and synaptic dysfunction. IL-33 modulates immune cell polarization and inflammatory status, influencing the expression of Bcl-2 proteins and caspases and thereby regulating neuronal apoptosis and synaptic plasticity. In the intrinsic pathway, caspase-3 activates phospholipid scramblase and inhibits flippase, leading to the externalization of phosphatidylserine (PS), a key “eat-me” signal. Microglia secrete MFG-E8, which links exposed PS with integrin ανβ3 receptors on the microglial surface, facilitating the phagocytic clearance of apoptotic synapses

Disorder of autophagic apoptosis in microglia: In chronic pathological conditions, the CX3CL1‑CX3CR1 axis serves as a central pathway for neuron-microglia communication. It plays a key role in preserving microglial homeostasis and restraining excessive inflammation. In several disease models—including stroke, Parkinson’s disease, and Amyotrophic Lateral Sclerosis (ALS)—this axis exhibits context-dependent dual functions, with both neuroprotective and neurotoxic effects (mechanistic evidence, mouse) [59]. Moreover, dysfunction of Caspase‑8—a key regulator of microglial inflammatory responses and synaptic remodeling—can lead to unchecked microglia-mediated neurotoxicity and disrupt normal activity-dependent synaptic elimination. This failure likely exacerbates synaptic dysfunction from the early stages of disease [57]. Under pathological conditions, microglia predominantly polarize toward a proinflammatory, neurotoxic phenotype. These activated microglia release inflammatory factors (e.g., TNF‑α and IL‑1β) that directly induce neuronal apoptosis and synaptic damage while simultaneously suppressing the neuroprotective/homeostatic phenotype (mechanistic evidence, mouse/in vitro)[55]. This shift establishes a microenvironment that undermines neuronal survival and synaptic maintenance. Although studies indicate that agents such as cordycepin can promote the conversion from a proinflammatory/neurotoxic to a neuroprotective/homeostatic phenotype—activating cAMP response element-binding protein (CREB), upregulating nerve growth factor (NGF), and thereby inhibiting neuronal apoptosis while enhancing synaptic protein expression [60]—such protective phenotypic switching often remains insufficient during disease progression (mechanistic evidence, mouse/in vitro). Under pathological conditions, apoptotic signaling pathways become dysregulated in both microglia and neurons. Aberrant “apoptosis‑like” tagging of synapses, coupled with uncontrolled phagocytic responses by microglia, leads to the erroneous elimination of a large number of synapses. Concurrently, impairments in microglial homeostasis, polarization balance, and the function of key regulatory molecules such as caspase‑8 further exacerbate neuroinflammation and synaptic damage.

Supported by diverse lines of evidence. These include in vivo experiments—such as imaging of microglial subsets in zebrafish [48] and studies using C1q knockout or caspase‑3 inhibitors in mice [52]; in vitro approaches, including acute slice imaging with PSVue to label phosphatidylserine‑exposed synapses [58]; human association and omics data from postmortem studies linking dysregulated apoptotic pathways to neurodegenerative disease pathology; and conceptual inference, exemplified by the proposed “apoptosis‑coupled phagocytosis” framework that integrates local apoptotic signals with complement‑mediated clearance [45, 52].

Epigenetic regulation of microglial function and synaptic plasticity: transcriptional regulation

Emerging evidence indicates that microglia can act as effectors of synaptic plasticity through epigenetic mechanisms (mechanistic evidence from mice) [61]. By responding to prenatal and postnatal environmental cues, microglia modulate their gene expression profiles, enabling them to adapt to changing conditions flexibly. During neurodevelopment, this epigenetic regulation allows microglia to influence the formation and pruning of neural connections, thereby shaping the structural and functional integrity of neural networks.

Epigenetic mechanisms in physiological synaptic regulation

Under physiological conditions, epigenetic mechanisms serve as dynamic regulators of microglial function, enabling these cells to fine‑tune their transcriptional programs in response to developmental signals, regional brain demands, and neuronal activity—thereby supporting synaptic homeostasis and plasticity [62]. Through DNA methylation, histone modifications, and non‑coding RNAs, microglia establish a flexible chromatin architecture that dynamically adapts to environmental changes. DNA methylation modulates microglial genes involved in cytokine production, receptor signaling, and cytoskeletal remodeling—which are critical for synaptic recognition and phagocytosis (in vivo evidence from mice) [63]. Histone acetylation/methylation alters chromatin accessibility, allowing rapid responses to environmental signals and neural activity. miRNAs regulate gene expression posttranscriptionally by binding to 3′UTRs and controlling microglial activation, polarization, and phagocytosis. With respect to miR‑34a, the results of luciferase assays confirmed the direct targeting of the 3′UTR of TREM2. Modulation of miR‑34a levels in cells inversely affects TREM2 expression and Aβ phagocytosis, and inhibition of miR‑34a in AD animal models increases TREM2 expression and reduces Aβ deposition. Thus, miR‑34a negatively regulates TREM2 to influence Aβ clearance (case-control study, human; in vitro, human cells) [64]. This epigenetic-mediated synaptic plasticity in microglia enables them to maintain an optimized monitoring state and perform precise, activity-dependent synaptic remodeling.

Lifespan and region-specific adaptations represent critical features of their physiological regulation. As aging progresses, microglia undergo dynamic epigenetic changes—including DNA methylation [65] and histone modifications [66]—which remodel their gene expression profiles and drive a transition from a homeostatic, neuroprotective phenotype toward reactive or disease-associated states. Sex differences, mediated by sex hormones and sex‑chromosome‑linked genes, further shape the functional diversity of microglia (structural evidence from primates) [35]. In human spinal cord microglia, all genes whose expression is sexually dimorphic are located on sex chromosomes. Notably, the expression of genes such as XIST and NLGN4Y differed markedly between males and females. This sex-specific gene expression profile likely influences chromatin architecture and transcriptional activity, which may underlie functional differences in synaptic regulation and contribute to sex-specific patterns of neural circuit development and disease susceptibility (structural evidence from mice) [67]. However, direct experimental confirmation at the level of protein activity and cellular behavior is still needed to define sex-specific functional outputs (transcriptomic evidence from mice and humans) [35]. Furthermore, distinct brain regions exhibit a region‑specific epigenetic landscape, enabling microglial function to adapt to local circuit demands (structural evidence, mouse) [68].

In a rat model, knockdown of DNMT1 was shown to reduce the expression of the proinflammatory cytokines TNF-α and IL-1β, with evidence indicating that DNMT1 modulates microglial polarization through the phosphatidylinositol 3-kinase/protein kinase B (PI3K/Akt) signaling pathway (genetic evidence from rats) [69]. In addition, TNF‑α and IL‑1β can induce rapid, redox‑dependent reorganization of the actin cytoskeleton in neuronal cells (e.g., SH‑SY5Y neuroblastoma cells), manifested by the formation of lamellipodia and membrane ruffles (structural evidence, cell line) [70] (Fig. 2A). This cytokine‑induced actin reorganization does not reflect genuine, beneficial synaptic plasticity enhancement but rather represents a pathological early stress response that ultimately culminates in cytoskeletal collapse and synaptic dysfunction. In contrast, epigenetic regulation supports synaptic plasticity by modulating neurotrophic signaling in microglia. Studies have shown that interventions such as histone deacetylase inhibition or electroacupuncture increase histone acetylation at the PI3K promoter (structural evidence, mouse) [71, 72]. This epigenetic modification upregulates PI3K in microglia, which in turn promotes the phosphorylation of downstream AKT and CREB. Consequently, microglia increase their secretion of brain‑derived neurotrophic factor (BDNF), supporting learning and memory processes. These findings illustrate how physiological epigenetic dynamics can directly link microglial function to cognitive output (Fig. 2B).

Fig. 2.

Epigenetic mechanisms mediate microglial regulation of synaptic plasticity. A DNA methylation regulates the morphology and synaptic plasticity of neurons by influencing the production of cytokines (such as TNF-α and IL-1β) and modulating the actin rearrangement mediated by Rho GTPases in neurons. B H3K9 acetylation and SUMO1 modification activate the PI3K-AKT signaling pathway, promote CREB phosphorylation, regulate BDNF expression and secretion, and act not only on TrkB on the postsynaptic membrane but also enhance synaptic function through the “synaptic tagging and capture” mechanism, promoting the release of neurotransmitters from the presynaptic membrane and thereby maintaining and enhancing LTP. C The H3.3G34R mutation reduces H3K36me2 modification at the tail of histone H3.3 and hinders the recruitment and chromatin redistribution of the DNA methyltransferase DNMT3A, leading to the loss of non-CG (CH) methylation, causing the continuous expression of the complement system and other innate immune genes, promotes neuroinflammation and the accumulation of activated microglia, and thereby affects synaptic pruning and synaptic plasticity. D Epigenetic changes promote the transformation of microglia to an inflammatory phenotype. The release of IL-1β by microglia through exosomes directly causes the loss of neurons and synapses, further disrupting synaptic plasticity. Activation of the NLRP3 inflammasome promotes caspase-1-dependent IL-1β and IL-18 maturation and cell pyroptosis, further exacerbating the inflammatory response and affecting synaptic plasticity

Epigenetic dysregulation in pathology and synaptic dysfunction

Under pathological conditions, sustained aberrant signaling drives maladaptive reprogramming of the microglial epigenome, resulting in marked alterations in gene expression profiles. These changes affect key microglial functions—including phagocytic activity, inflammatory cytokine secretion, and the balance between neuroprotection and neurotoxicity—which are closely linked to synapse formation, pruning, and remodeling. Ultimately, such functional shifts may contribute to the loss of synaptic integrity (structural evidence, mouse) [73].

Environmental factors—such as early-life stress and infection—act as key epigenetic triggers that shape microglial gene expression, thereby influencing how these cells sense and respond to the neural milieu. In mouse models, maternal immune activation (MIA), a specific and potent form of early‑life stress, has been shown to induce an immunologically blunted state in microglia. This state is associated with an impaired ability to regulate synaptic function (genetic/behavioral evidence, mouse), as evidenced by a reduced presynaptic vesicle release probability at synapses of dopamine receptor D2‑expressing medium spiny neurons (D2R‑MSNs) in the striatum, along with abnormal synaptic transmission. Notably, replacing the altered microglia in MIA‑exposed offspring with healthy microglia partially restored immune responsiveness, improved the probability of presynaptic release, and rescued circuit function. These findings further substantiate a causal link between the epigenetic state of microglia and the integrity of synaptic signaling [74]. Studies have indicated that microglia exposed to early-life stress (ELS) undergo persistent alterations in gene expression in adulthood, affecting pathways such as TNF signaling and protein ubiquitination (structural evidence from mice) [75]. Such epigenetic reprogramming can render microglia either “trained” or “desensitized” upon subsequent immune challenges, reflecting long‑term functional adaptation. Notably, these microglia exhibit reduced phagocytic capacity toward synaptic structures, a deficit that becomes more pronounced in adult animals. In a mouse model of HSV-1 infection, recurrent viral reactivation was observed to activate microglia and astrocytes, accompanied by a significant increase in the expression of the proinflammatory cytokine interleukin-1β (IL-1β). Increased IL-1β expression is associated with the translocation of the histone deacetylase HDAC4 from the cytoplasm to the nucleus (pharmacological evidence, mouse), where it enhances the SUMOylation of methyl‑CpG‑binding protein 2 (MeCP2). This post‑translational modification strengthens the transcriptional repressor activity of MeCP2. By acting as an epigenetic silencer, MeCP2 binds to and suppresses a broad set of neural‑related genes—particularly those encoding synaptic proteins—leading to the downregulation of synaptic gene expression and, consequently, to synaptic structural and functional deficits [76].

Such pathological epigenetic cascades are exemplified in neurodegenerative contexts. In models of Alzheimer’s disease, epigenetic mechanisms act as sustained switches that promote the expression of inflammatory genes. Epigenetic modifications can drive the transcription of IL‑1β and IL‑18, thereby activating the NOD-like receptor family pyrin domain-containing 3 (NLRP3) inflammasome. This cascade leads to the release of neurotoxic factors via microglial exosomes, directly contributing to synaptic loss (structural evidence, mouse) [77] (Fig. 2D). Research has demonstrated that the H3.3G34R mutation triggers widespread redistribution of the DNA methyltransferase DNMT3A, resulting in the failure of complement genes to be epigenetically silenced (in vivo mouse model) [78] (Fig. 2C). Consequently, these genes remain persistently active, pathologically reactivating synaptic pruning by microglia and accelerating the degeneration of neural circuits. Direct evidence from human microglia supports this epigenetic link to disease. ATAC‑seq and H3K4me2/H3K27ac ChIP‑seq analyses of microglia isolated from human surgical brain specimens have mapped chromatin accessibility and enhancer landscapes, revealing that Alzheimer’s disease risk SNPs—such as rs6733839 upstream of BIN1—are enriched in microglial enhancers. The implicated genes, including BIN1, CX3CR1, and TREM2, are involved in synaptic signaling (associative evidence; chromatin profiling from human microglia; functional validation linking enhancer variants to disease mechanisms not yet established) [79]. Therefore, targeting epigenetic dysregulation holds considerable therapeutic promise. Moreover, interventions such as histone deacetylase inhibitors or ascorbate—a cofactor in TET‑mediated DNA demethylation—can restore homeostatic microglial functions and improve synaptic integrity, demonstrating the reversibility of these pathological signatures and underscoring the central role of epigenetic dysregulation in synaptic dysfunction (in 5xFAD mice and in vitro) [80].

In conclusion, epigenetic mechanisms bridge external stimuli and changes in synaptic plasticity through multilevel regulation of microglial proinflammatory gene expression and intercellular communication (Fig. 2).

Multiple lines of experimental evidence support epigenetic regulation of microglial synaptic function. These include in vivo studies (e.g., histone deacetylase inhibition in rats [71]; maternal immune activation models in mice [74]), in vitro experiments (e.g., histone deacetylase inhibitor treatment in microglial cultures [71]), human association and omics data (genetic studies linking epigenetic regulators to neurodevelopmental disorders [79]), and conceptual inference (the “training/desensitization” framework proposed for microglial epigenetic memory [75]). Causal links between specific epigenetic modifications and microglial functional outcomes come primarily from genetic and pharmacological models in rodents, whereas human tissue studies provide correlational evidence for epigenetic dysregulation in disease contexts.

ECM proteins mediate microglial regulation of synaptic plasticity: microenvironmental remodeling

Synapses are not isolated. Instead, they are embedded in the complex ECM and form a functionally integrated unit—the “tetrapartite synapse”—together with presynaptic/postsynaptic terminals and glial cells (electrophysiological and morphological studies of animal models) [81]. The ECM not only physically connects presynaptic and postsynaptic structures but also mediates bidirectional intercellular signaling through interactions with cell adhesion molecules (CAMs), thereby regulating synaptic structural stability and functional states. Notably, microglia can precisely degrade perisynaptic ECM components by secreting specific proteases to promote synaptic structural remodeling and plasticity (in 5xFAD mice; in vitro only) [82].

Stable state regulation mediated by the ECM under physiological conditions

Structural and Functional Basis: Neurons are enveloped by a specialized type of ECM structure—the neuronal perineuronal network (PNN), especially inhibitory gamma-aminobutyric acid-ergic (GABAergic) parvalbumin-positive (PV) interneurons. Perineuronal nets (PNNs) provide crucial structural support to neurons, yet their inherent stability also imposes physiological constraints on excessive structural plasticity in regions such as the adult cerebral cortex and hippocampus. This restriction contributes to memory consolidation and helps maintain neural circuit stability [83]. The synaptic extracellular matrix (ECM) is composed primarily of chondroitin sulfate proteoglycans (CSPGs). CSPGs play dual roles: they provide critical anchoring sites for the complement initiator protein C1q [84], and themselves constitute a modifiable barrier that regulates structural plasticity (genetic evidence, mouse). For example, CSPGs restrict dendritic spine plasticity by inhibiting β1-integrin activity. Enzymatic digestion of CSPGs (e.g., with chondroitinase ABC, ChABC) relieves this inhibition in an integrin‑dependent manner, thereby promoting experience‑dependent structural remodeling [85].

Active regulation of microglia: Research in mouse models has indicated that the stability of perineuronal nets (PNNs) is influenced by microglia. Microglial depletion or deficiency results in the accumulation of PNNs and diffuse ECM components (pharmacological/genetic evidence, mouse)—such as aggrecan, hyaluronic acid, and brevican—in the brain, accompanied by a reduction in PNN pore size. These observations suggest that under basal conditions, microglia continuously engage in low-level degradation and clearance of ECM elements, thereby maintaining appropriate PNN density and pore architecture (in 5xFAD mice) [86, 87]. During experience‑dependent neural activity, IL-33 secreted by neurons recruits microglia to phagocytose PNN components, thereby creating space for synaptic remodeling [49, 88]. In the developing visual thalamus, knockout of CR3 or C3 reduces microglial synapse engulfment by approximately 50%, leading to a 100–130% increase in eye‑specific projection overlap and a 1.3- to 1.8-fold increase in adult synapse density. These findings indicate that developing microglia rely on the CR3‑C3 pathway to engage in precise neural circuit pruning actively; loss of this pathway results in pruning deficits, defective eye‑specific segregation, and excess synapses in adulthood (animal in vivo experiments) [89], highlighting the critical role of microglia in establishing accurate neural connection maps. Beyond the classical “tag-and-uptake” model, emerging evidence suggests more specific pathways, such as “synaptic cutting”—where microglia actively migrate and sever neuronal membrane connections—and “synaptic cleanup”, in which neurons autonomously initiate synaptic detachment and microglia clear the resulting debris (mouse models and in vitro experiments) [90]. Importantly, all of these processes critically depend on the dynamic interplay between microglia and the extracellular matrix (ECM).

Dysregulation of the ECM regulatory network under pathological conditions

Under pathological conditions such as neuroinflammation or neurodegenerative disease, this finely tuned ECM-microglia regulatory network becomes significantly disrupted. The aberrant activation of microglia and their dysregulated interactions with the ECM collectively drive synaptic dysfunction and loss.

Inflammation-driven ECM and dysfunction of microglia: In response to pathological stimulation (e.g., with poly(I:C)), microglia become overactivated and secrete a broad range of effector molecules, including the inflammatory cytokines IL-6 and TNF-α, the chemokines CCL5 and CXCL1, and the matrix metalloproteinases MMP-2 and MMP-9. These factors degrade the extracellular matrix—particularly perineuronal nets composed of aggrecan—and directly disrupt synaptic structure and function (in vitro study) [91]. Furthermore, during acute or chronic inflammatory conditions, microglia can directly and rapidly alter neuronal excitability and synaptic transmission through the overproduction of factors such as TNF‑α and IL‑1β—a process that is often accompanied by pathological changes in the ECM [92].

Disruption of ECM integrity accelerates the neurodegenerative process: In chronic pathological processes such as Alzheimer’s disease (AD), aberrant microglial activation is often accompanied by progressive loss of perineuronal nets (PNNs), particularly those surrounding parvalbumin‑positive interneurons [93]. In contrast to the activity-dependent, localized, and reversible remodeling of PNNs observed under physiological conditions, widespread pathological PNN loss impairs synaptic input onto inhibitory neurons. This disruption of the excitatory/inhibitory (E/I) balance contributes directly to cognitive deficits and circuit dysfunction (in 5xFAD mice; human tissue immunohistochemical and transcriptomic analysis) [94]. This disruption creates a vicious cycle: pathological protein aggregates (e.g., Aβ) activate microglia → excessive release of proteases and inflammatory factors drives the overdegradation of PNNs → PNN loss weakens inhibitory control, exacerbating network hyperexcitability and excitotoxicity → further aggravating neurodegeneration.

ECM disruption impairs complement function: The structural integrity of the extracellular matrix (ECM) is essential for the proper localization and function of the complement system. When this integrity is compromised by pathological factors—such as excessive degradation of ECM components in disease conditions—C1q fails to anchor effectively to chondroitin sulfate proteoglycans (CSPGs). This disruption leads to aberrant local expression of C1q, C3, and their corresponding receptors (in 5xFAD mice) [84]. Under pathological conditions, compromised ECM scaffolding disrupts microglia-mediated synaptic pruning—either by hindering the clearance of tagged synapses or by promoting erroneous attacks on healthy synapses—thereby exacerbating synaptic loss.

In summary, microglia secrete multiple ECM components (Table 3) that regulate synaptic plasticity and cognitive function. Through precise regulation of these targets, microglia exert bidirectional control over the local microenvironment—creating space for synaptogenesis and circuit reorganization under physiological conditions—while ECM dysregulation exacerbates synaptic dysfunction.

Table 3.

Influence of ECM proteins on synaptic plasticity and cognitive function

ECM Protein	Cellular Source	Microglia-Mediated Mechanism of Synaptic Plasticity	Effect on Synaptic Plasticity	Impact on Cognitive Function	Evidence Tags
MMP-9	Multiple cell types (e.g., neurons, microglia, pericytes, infiltrating leukocytes) [254].	Inflammatory activation is associated with microglial MMP-9 release, which correlates with BBB basement membrane disruption, alterations in the synaptic microenvironment, and reduced neuroplasticity [254].	Correlates with dendritic spine expansion and arbor growth; is associated with enhanced LTP; and with reduced PNN integrity (which restricts neuroplasticity) [254].	Is implicated in amyloid-beta degradation and is associated with reduced AD-related pathological burden [254].	Ex vivo study; In vivo study; Pharmacological study; Behavioral correlation [254].
tPA	Striatal GABAergic neurons [255].	Interacts with NMDAR1 to trigger neuroinflammation, activate microglia, and upregulate pro-inflammatory factors, impacting synaptic function and neuronal survival [255].	Enhances synaptic transmission and interneuronal signaling; excessive activity induces neuroinflammation and synaptic dysfunction [255].	tPA deficiency is associated with protection of dopaminergic neurons from α-synuclein-induced degeneration and with alleviated behavioral deficits [255].	Genetic methods; Ex vivo study; In vivo study; Pharmacological method; Molecular biology method [256].
Aggrecan	Glial precursor cells (primary source) [257].	Secretes matrix metalloproteinases to degrade aggrecan, regulating PNN composition/structure and ECM homeostasis, thereby modulating synaptic plasticity [86].	Core PNN protein associated with restriction of synaptic plasticity and stabilization of circuits; degradation correlates with neuronal maturation, synaptic remodeling, and neuroplasticity [86].	Loss/reduction is associated with disrupted PNN integrity, impaired inhibitory neuronal function and synaptic stability, and correlates with cognitive impairment in cerebral small vessel disease [258].	Genetic methods; Ex vivo study; In vivo study [257].
PNNs	Neurons + surrounding glial cells (cosecreted/assembled) [259].	Interacts with PNNs to mediate remodeling, regulating GABAergic inhibitory synapses within PNN meshwork and promoting synaptic remodeling/functional recovery [260].	Is involved in critical period opening/closing; is associated with restriction of adult synaptic plasticity [259].	Stable PNNs are associated with normal cognition; moderate structural relaxation correlates with neuroplasticity/functional recovery, while complete elimination impairs cognition [260].	Genetic research; Pharmacological research; In vivo study; Ex vivo study; Nanotechnology and gene therapy; Human genetics research [259].

The regulatory role of microglia in the extracellular matrix and perineuronal nets is supported by in vivo experimental evidence (microglial depletion models [86, 87]; IL‑33 knockout and overexpression studies [88]), human association and omics data (postmortem studies linking perineuronal net loss to cognitive decline in Alzheimer’s disease [94]), and conceptual inference (the “tetrapartite synapse” framework integrating extracellular matrix remodeling with complement‑mediated pruning [81]; the supportive role of neuronal Clq‑like proteins in C1q function [84]). Causal relationships between microglial protease secretion and extracellular matrix degradation are primarily derived from pharmacological and genetic models in mice, whereas human tissue studies provide correlational evidence for extracellular matrix disruption in neurodegenerative diseases.

Microglial energy metabolism regulates synaptic plasticity: metabolic support

Microglia support synaptic plasticity and cognitive function by modulating brain energy metabolism, involving regulating local glucose utilization, supporting mitochondrial homeostasis, and supplying metabolic substrates for neuronal activity. While this role has been clearly demonstrated in the hippocampus (in vivo mouse model and in vitro experiments) [95], evidence suggests that metabolic regulation by microglia constitutes a fundamental mechanism for maintaining neural circuit function across brain regions. In the pathological context of diabetic encephalopathy, alterations in microglial energy metabolism extend beyond the hippocampus and are observed throughout the central nervous system (CNS). The energy expenditure of microglia during immune surveillance is closely linked to their activation state. Furthermore, microglia express various transporter proteins for energy substrates, enabling them to flexibly utilize multiple substrates, such as glucose, fatty acids, and glutamine (in 5xFAD mice) [96].

Glucose metabolism-mediated regulation of synaptic plasticity by microglia

The processing of energy substrates directly restricts the quality of synaptic pruning, and metabolism is the prerequisite for the regulation of synaptic plasticity by microglia. Using microglia-specific monocarboxylate transporter 4 (MCT4) knockout mice, we found that disrupting lactate uptake impaired lysosomal function and synaptic pruning, leading to hippocampal hyperexcitability, seizure susceptibility, and anxiety-like behavior (in vivo animal models and in vitro experiments) [97]. These results demonstrate that microglial lactate utilization is essential for developmental pruning and excitation/inhibition balance, revealing a causal link between energy substrate handling and synaptic plasticity. Therefore, the reprogramming of glucose metabolism may serve as a crucial link in connecting the synaptic pruning of microglia.

Glucose metabolism under physiological conditions and under synaptic support Under homeostatic/surveillance conditions, microglia continuously extend and retract their processes, contact synapses, phagocytose excess synaptic material and ECM components, and secrete plasticity-related factors. These activities are powered by finely regulated glucose metabolism and complementary substrate utilization. Oxidative phosphorylation in microglia supplies the energy required for sustained immune surveillance and homeostatic maintenance. This efficient metabolic mode underpins their neuroprotective functions and serves as a foundation for maintaining normal synaptic plasticity (acute brain slice model and 5xFAD mice) [98–100]. In the HMC3 human microglial cell line, an in vitro model, basal energy metabolism predominantly relies on mitochondrial oxidative phosphorylation (OXPHOS), providing initial insight into the metabolic capacity of microglia (an in vitro cell line). This metabolic phenotype is characterized by a high ATP/ADP ratio (approximately 13) and substantial energy charge potential, reflecting a robust ATP reserve and efficient oxidative metabolic capacity [101].

Glycolytic flux in microglia directly supports key physiological functions. For instance, efficient glucose metabolism is required for the phagocytic activity and secretion of neurotrophic factors, which in turn facilitate precise synaptic refinement during development and adulthood, as well as the remodeling of perineuronal nets (PNNs). Through these mechanisms, microglia help regulate synaptic plasticity in neural circuits, including those of the hypothalamus, thereby contributing to the establishment of effective systemic glucose homeostasis [102]. Under glucose-sufficient conditions, microglial function is supported by glucose oxidative metabolism. In contrast, during acute glucose deprivation, microglia rapidly shift to glutamine-driven mitochondrial metabolism (glutaminolysis) to sustain motility and injury sensing (in vivo imaging of CX3CR1EGFP/+ mice and in vitro imaging of primary microglia cultured in vitro) [103]. This metabolic flexibility—enabling persistent surveillance and phagocytic capacity even under low‑glucose conditions—is thought to be crucial in safeguarding synaptic and vascular microenvironments, preventing transient energy crises from escalating into widespread synaptic loss.

Metabolic reprogramming under pathological conditions and under conditions of synaptic dysfunction In inflammatory or neurodegenerative pathological environments, microglia shift from a “monitoring” state to a “response” state, and their glucose metabolism is profoundly reprogrammed. This change is typically the core driving factor of synaptic dysfunction.

Inflammation drives metabolic transformation: Multiple in vitro and in vivo studies have demonstrated that upon lipopolysaccharide (LPS)-induced TLR4 activation, microglia rapidly upregulate aerobic glycolysis and, over a period ranging from several hours to 24 h, progressively suppress oxidative phosphorylation (in vitro and in AD cell models) [104–106]. This process is precisely orchestrated by the Toll-like receptor-mechanistic target of rapamycin-hypoxia‑inducible factor 1‑alpha (TLR-mTOR-HIF‑1α) axis. Activation of microglia by LPS via TLRs upregulates mTOR signaling, whereas pharmacological inhibition of mTOR—for example, with rapamycin—completely suppresses the ensuing increase in glycolysis[107]. In mouse models of Alzheimer’s disease, microglia carrying loss-of-function mutations in TREM2 exhibit impaired mTOR-HIF-1α signaling; this impairment is associated with reduced activation of the glycolytic program and is accompanied by diminished key protective functions such as phagocytic capacity (genetic evidence in mice) [108]. Restoring activity in this pathway can partially rescue these functional deficits [109, 110] (Fig. 3). The metabolic shift of microglia toward glycolysis provides rapid energy to support their high‑energy‑demand phagocytic functions. However, this enhanced phagocytic capacity depends not only on increased energy supply but also on the transcriptomic reprogramming and metabolic activation induced by NHE1 deficiency. This includes upregulation of the expression of phagocytosis‑related genes, activation of the LXR/RXR signaling pathway, and a general increase in energy metabolism. Together, these changes synergistically increase the ability of microglia to recognize and clear injury‑associated debris—such as damaged synapses and myelin fragments—thereby promoting tissue repair and limited neural circuit reorganization (in vivo experiments) [111]; however, overall, this represents an inefficient metabolic pattern, which is often accompanied by and exacerbates neuroinflammation, ultimately leading to synaptic dysfunction [96].

Fig. 3.

Metabolic reprogramming of microglia under inflammatory conditions impairs synaptic plasticity. Inflammatory stimulation leads to the upregulation of mTOR and HIF-1α expression, triggering the metabolic reprogramming of microglia, manifested as an increase in glycolytic flux, the accumulation of some metabolites in the tricarboxylic acid cycle, decreased mitochondrial respiration, and reduced ATP production. This metabolic transformation promotes the production of proinflammatory cytokines, ultimately enhancing the neuroinflammatory response and inhibiting the LTP of synapses, thereby impairing synaptic plasticity and affecting the function and learning-memory processes of neural networks. Furthermore, microglia are activated by inflammatory stimuli (such as LPS) and under hypoxic conditions, which subsequently trigger the activation of NADPH oxidase, leading to the production of reactive oxygen species (ROS) and ultimately inhibiting the long-term potentiation (LTP) of synapses

In neurodegenerative conditions such as Alzheimer’s disease (AD), this metabolic reprogramming results in pathological dysregulation. Early Aβ accumulation activates microglia and induces pronounced metabolic alterations, notably in lactate metabolism (in 5xFAD mice) [112]. In an LPS-induced neuroinflammation model, NOX4 promotes ROS production and glycolytic reprogramming, driving M1 polarization. Treatment with a ROS scavenger or NOX4 knockout significantly decreases the expression of inflammatory markers [113]; in contrast, enhancing NOX4 activity amplifies neuroinflammation and triggers synaptic damage, thereby suppressing hippocampal LTP and impairing related learning and memory (pharmacological/genetic evidence, rats) [114]. Mice deficient in progranulin (PGRN) exhibit impaired glucose uptake and utilization, accompanied by synaptic loss and elevated neurofilament light chain levels, exacerbating neurodegenerative damage [115]. These findings suggest that metabolic dysregulation may represent a core pathological component that is independent of excessive inflammation. Targeting this pathological metabolic reprogramming has therapeutic potential. For example, sodium rutin (NaR) can reprogram microglial metabolism from anaerobic glycolysis back toward mitochondrial oxidative phosphorylation, increase ATP production, and thereby enhance Aβ clearance, promote a shift to a neuroprotective/homeostatic phenotype, and mitigate neuroinflammation (in a 5xFAD mouse model and in vitro experiments) [116]. Additionally, activated microglia indirectly influence neuronal synaptic function and plasticity through cytokine secretion and modulation of astrocyte metabolism. Therefore, improving the metabolic state and polarization profile of microglia is crucial for maintaining neural homeostasis and supporting neuronal health and synaptic function.

Lipid metabolism-mediated regulation of synaptic plasticity by microglia

Beyond glucose metabolism, microglial lipid sensing, synthesis, and β‑oxidation pathways are key nodes that regulate migration, phagocytosis, and inflammatory signaling (transcriptomic/lipidomic association, mouse) [117].

Lipid metabolism under physiological conditions and the maintenance of synaptic homeostasis As sentinels of the neural microenvironment, microglia depend on endogenous lipid synthesis and remodeling to sustain their dynamic structure and function. Given that cholesterol and fatty acids in the brain are largely synthesized by glial cells and cannot freely cross the blood-brain barrier, lipid metabolism in microglia is essential for preserving local neural homeostasis (animal models and in vitro experiments)[118] and lies at the core of the regulation of their activation, migration, phagocytosis, and inflammatory signaling [119].

Lipid metabolism regulates synaptic structure and function: Under homeostatic conditions, microglia continuously sense and utilize surrounding fatty acids, cholesterol, and phospholipids, which serve as both energy substrates and signaling molecules. Their lipid metabolic profile is closely aligned with their homeostatic/surveillance state (in vivo and human postmortem evidence) [120]. These processes enable the selective engulfment of redundant or weakened synapses—a key mechanism of synaptic pruning that contributes to the refinement of mature neural circuits [121]. ω-3 Polyunsaturated fatty acids (PUFAs) (e.g., DHA) promote anti-inflammatory phenotypes and phagocytic clearance, whereas ω-6 derivatives (e.g., PGE₂) drive proinflammatory responses, highlighting the context-dependent role of PUFA metabolism (extracellular cell experiments and animal models) [122]. Lipid rafts—cholesterol‑ and sphingolipid‑enriched microdomains—are critical for synapse formation and function. By regulating AMPA receptor trafficking and localization and modulating the clustering and activity of signaling molecules, lipid rafts participate in the fine‑tuning of synaptic nanostructures, representing an emerging mechanism of synaptic plasticity (extracellular cell model) [123]. These rafts modulate neurotransmitter release and receptor responsiveness by regulating receptor localization and the assembly of signaling complexes. Through their membrane lipid composition—particularly the dynamics of lipid rafts—microglia fine‑tune their interactions with synapses.

Furthermore, by regulating lipid metabolism, microglia influence inflammatory and apoptotic signaling, thereby affecting synaptic function and network stability [124]. These finely tuned lipid metabolic processes are profoundly shaped by the systemic lipid balance. Under physiological conditions, lipid metabolic pathways operate in concert. However, upon dietary or pathological challenges, the preferential engagement of specific pathways allows saturated fatty acids (e.g., palmitic acid) and unsaturated fatty acids (e.g., docosahexaenoic acid and eicosapentaenoic acid) to directly reshape distinct functional phenotypes of microglia (in vitro and in vivo evidence) [28] (Fig. 4).

Fig. 4.

Lipid metabolism in microglia regulates synaptic plasticity. The SFA pathway tends to induce proinflammatory responses in microglia, and the metabolic state related to SFAs is usually associated with decreased oxidative phosphorylation activity, which may promote inflammation and neurodegenerative processes. The PUFA pathway promotes a microglial activation state akin to the immunomodulatory neuroprotective/homeostasis maintenance phenotype, marked by the upregulation of alternative activation markers such as Arg1 and Ym1/Chil3 (in mice) and is mediated through receptors such as GPR120, leading to downstream suppression of the NF-κB pathway

Lipid metabolism imbalance under pathological conditions and under conditions of synaptic damage Under conditions of dietary imbalance or neurological disease, lipid homeostasis is disrupted. The preferential engagement of specific metabolic pathways directly reshapes the functional phenotype of microglia, driving neuroinflammation and synaptic dysfunction [28].

Inflammation and the vicious cycle of lipid metabolism: In response to stimulation with LPS and other inflammatory signals, activation of the TLR4 pathway drives metabolic reprogramming in microglia, involving a shift from mitochondrial oxidative phosphorylation to aerobic glycolysis, accompanied by increased fatty acid synthesis, lipid droplet formation, and mitochondrial fission. Together, these alterations promote a proinflammatory phenotype and functional state in microglia (extracellular cell culture experiments and in vivo animal models) [125]. Inflammatory stimuli activate microglia, promoting the synthesis and secretion of 25‑hydroxycholesterol (25HC) in a Ch25h‑dependent manner. 25HC subsequently induces negative synaptic plasticity in the hippocampal CA1 region via an NMDA receptor‑dependent mechanism, thereby inhibiting LTP and impairing learning and memory (in vitro and in vivo mouse models) [126]. Under pathological conditions, proinflammatory lipid pathways often dominate. Microglia‑driven network hyperexcitability can enhance neuronal lipid droplet metabolism, and microvesicles released by inflamed microglia may stimulate neuronal sphingolipid metabolism, paradoxically leading to excessive synaptic potentiation (in vivo and in vitro studies, the mechanism is partially unresolved) [127]; this dysregulated crosstalk likely disrupts overall network balance.

Lipid droplet accumulation and impaired lipid clearance compromise microglial phagocytic function and synaptic pruning, disrupting network homeostasis (in 5xFAD mice and patient samples) [128, 129]. SHIP1 (Src homology 2 domain‑containing inositol‑5‑phosphatase 1) is a lipid phosphatase that employs its SH2 domain to bind phosphorylated tyrosine residues on receptors, thereby blocking the recruitment of downstream kinases such as SYK. This negative regulation of immune receptor signaling affects microglial migration and phagocytic function. Within microglia, SHIP1 restrains complement‑mediated synaptic pruning by suppressing signals from TREM2 and the complement receptor CR3, thus supporting normal neural circuit development (genetic evidence from mice) [130].

Under pathological conditions, dysregulated cholesterol metabolism compromises the homeostasis and composition of lipid rafts. These altered rafts fail to properly organize synaptic receptors and signaling molecules, thereby impairing synaptic transmission and plasticity [131]. Furthermore, pathological compositional alterations in the membrane lipids of microglia can disrupt their interactions with synapses and their capacity to regulate inflammation, which may amplify synaptic damage [132].

Microglial modulation of glutamatergic signaling and synaptic plasticity

Glutamate is the primary excitatory neurotransmitter in the mammalian brain. The majority of excitatory synapses are glutamatergic and mediate the bulk of information transfer and network activity integration. AMPA receptors underlie fast excitatory synaptic currents and largely determine moment‑to‑moment synaptic efficacy, whereas NMDA receptors—with their voltage‑dependence and high Ca²⁺ permeability—act as key triggers for synaptic plasticity (e.g., LTP/LTD) and synaptic computation (in vitro and in vivo models) [133]. As key regulators of glutamatergic synaptic plasticity, microglia act as gatekeeping and homeostatic hubs that fine‑tune neuronal structure and function. By modulating glutamate receptor composition, release probability, synapse number, and the cytokine milieu, microglia help ensure proper neural network development, functional equilibrium, and adaptive plasticity—processes essential for maintaining central nervous system health and stability [21].

The maintenance of glutamate homeostasis under physiological conditions and under conditions of synaptic support Beyond receptor dynamics, precise glutamate metabolism—encompassing uptake, recycling, and homeostasis—is fundamental for maintaining this plasticity. Microglia actively participate in metabolic regulation. They express excitatory amino acid transporters (EAATs), primarily EAAT1 (GLAST) and EAAT2 (3 × Tg-AD mice, 5 × FAD mice, and in vitro cell experiments) [134, 135]. These processes are crucial for clearing excessive glutamate in synapses, thereby preventing excitotoxicity and enhancing synaptic efficacy.

Beyond their established roles, microglia participate in the glutamate-glutamine shuttle—a process traditionally ascribed to astrocytes. Emerging evidence indicates that microglia also contribute to glutamate conversion and nitrogen metabolism. In vitro, microglia express GLT-1 and glutamine synthetase, enabling glutamate uptake and conversion to glutamine, thereby supporting the neuron-glia glutamine-glutamate cycle and neurochemical balance [136]; notably, microglial activation often precedes that of astrocytes [137]. By integrating glutamatergic and GABAergic receptor signals [138] along with “do not‑eat‑me” cues such as CD47 [139], microglia coordinate synaptic pruning and preservation. This integrated signaling helps maintain the excitatory/inhibitory (E/I) balance during critical developmental windows, thereby preventing aberrant dendritic pruning or apoptosis triggered by excessive glutamate release [53]. In microglia, the enzyme glutaminase (GLS) catalyzes the conversion of glutamine (Gln) to glutamate (Glu), which can subsequently enter the tricarboxylic acid (TCA) cycle as α‑ketoglutarate. This pathway supplies carbon and nitrogen backbones that support both energy metabolism and proinflammatory metabolism in microglia, thereby promoting their proinflammatory phenotype (in vitro only) [140]. The balance of Gln/Glu nitrogen flux represents a key metabolic mechanism that influences whether microglia adopt a proinflammatory or neuroprotective state—a shift that critically determines whether synaptic plasticity is pathologically heightened or preserved (in vitro and gene knockout animal models) [141].

Microglia mediate positive pruning and negative feedback inhibition of glutamatergic synapses: Microglia support the structural and functional maturation of glutamatergic networks. Studies in mice lacking the microglia‑specific receptor CX3CR1 (Cx3cr1^-/-) revealed transient increases in dendritic spine density in the developing hippocampus, along with defects in glutamatergic synaptic maturation. These include altered amplitude ratios of spontaneous to miniature excitatory postsynaptic currents (sEPSC/mEPSC), impaired induction of long‑term depression (LTD), and heightened seizure susceptibility. Together, these observations indicate that through cytokine-mediated signaling, microglia help to fine-tune both the structure and function of developing glutamatergic synapses (in vivo experiments) [142]. These electrophysiological deficits arise primarily from impaired presynaptic function, leading to decreased reliability of synaptic transmission. Microglia dynamically control glutamatergic transmission through TNFα. Low TNFα enhances glutamate signaling by increasing the expression of GluA1‑containing AMPA receptors at glutamatergic synapses, whereas activated microglia reverse this potentiation within 24 h, restoring baseline excitatory drive and concurrently increasing inhibition (in vitro only) [143]. This feedback loop enables microglia to stabilize glutamate-mediated plasticity and prevent runaway excitation.

Pathological conditions lead to imbalances in glutamate metabolism and dysfunction of synapses Under pathological conditions, the regulatory role of microglia in glutamate metabolism becomes disrupted, shifting from homeostatic guardians to dysfunctional drivers of synaptic impairment.

Glutamate homeostasis imbalance and excitotoxicity: During the early stages of neurodevelopmental or neurodegenerative disorders, an imbalance between excitatory and inhibitory neurotransmission can lead to excessive glutamate release in localized regions [53]. Under these conditions, the glutamate‑clearing capacity of microglia may become compromised or dysregulated because of their altered activation state—such as a shift toward a proinflammatory phenotype—rendering them unable to buffer synaptic glutamate levels adequately. This failure exacerbates excitotoxicity, directly damaging neurons and driving excessive dendritic pruning and apoptosis—an early event in synaptic loss [53]. During neuroinflammation and neurodegeneration, the activity of the glutamate transport system in glial cells significantly changes. The expression of system xc⁻ (composed of xCT/SLC7A11) is upregulated, increasing glutamate efflux, whereas the expression of the glutamate transporter EAAT1/2 is downregulated, reducing glutamate reuptake [135]. This dual dysregulation elevates extracellular glutamate levels, leading to excitotoxicity and impaired synaptic plasticity.

The interaction damage between inflammatory signals and glutamatergic synapses: CR3 signaling provides another pathway that links neuroinflammation to glutamatergic dysfunction. Upon activation, CR3 triggers oxidative stress via NADPH oxidase, leading to PP2A‑mediated internalization of GluA2‑containing AMPA receptors. This cascade induces long‑term depression (LTD) and weakens synaptic efficacy, a process that may precede and promote structural pruning (in vitro only) [144]. At low concentrations, TNF‑α promotes the synaptic accumulation of GluA1‑only AMPA receptors, thereby strengthening excitatory synaptic transmission. At higher concentrations, TNF‑α enhances both excitatory and inhibitory transmission. When proinflammatory activated microglia are present, this TNF‑α‑driven potentiation of excitatory transmission emerges within approximately 3 h but returns to baseline within 24 h—a negative feedback response that limits further synaptic enhancement, resembling a “ceiling” effect. These observations suggest that proinflammatory microglia contribute to synaptic homeostasis by providing negative feedback regulation, thereby preventing excessive excitation [143].

The metabolic mechanisms by which microglia regulate synaptic plasticity are supported by in vivo perturbation studies (SHIP1 knockout mice [130]; NOX4 knockout or antioxidant treatment in rats [113]), in vitro/ex vivo perturbation experiments (exogenous lactate stimulation of primary microglia [97]), evidence from human iPSC‑derived microglia (the TREM2 R47H mutation impairs glycolytic capacity [109]), human association and omics data (postmortem metabolomic studies linking lipid metabolism dysregulation to neurodegenerative diseases [124]), and conceptual inference (the hypothesis that metabolic reprogramming serves as a core pathological component independent of overt inflammation [115]). Causal links between specific metabolic pathways and microglial functional outcomes are derived primarily from genetic and pharmacological models in rodents, whereas human iPSC-derived microglia offer mechanistic insights relevant to human biology.

Microglia-astrocyte crosstalk regulates synaptic plasticity: interglial coordination

Synaptic plasticity arises from tripartite communication among neurons, astrocytes, and microglia, in which bidirectional microglia-astrocyte crosstalk plays a central role. Astrocytes and microglia interact through secreted factors—such as growth factors, neurotransmitters, cytokines, and chemokines—as well as extracellular vesicles. This interglial dialogue critically shapes neuronal function and synaptic plasticity [145]. By maintaining neural environmental stability, modulating immune responses, and releasing signaling molecules, this coordinated interplay promotes dynamic synaptic regulation and optimizes network function—processes that are critical for both normal brain operation and neuroprotection under pathological conditions.

Synergistic effects under physiological conditions and the maintenance of synaptic homeostasis

During neural circuit development and adult plasticity, microglia and astrocytes play complementary and synergistic roles, and their coordinated crosstalk is essential for synaptic homeostasis. However, most current research on microglia-astrocyte interactions has focused on injury, trauma, or degenerative disease. The mechanisms by which these cells functionally integrate and physically interact to modulate synaptic plasticity under physiological conditions remain poorly characterized.

Complementary functions and synergy formation: Microglia help shape mature neural networks primarily through activity‑dependent phagocytic clearance of apoptotic neurons and redundant synapses [146], whereas astrocytes serve as pivotal coordinators that regulate the balance of neuronal excitation and inhibition through glial transmitter release [147] (in vitro only and ex vivo). Microglia mediate the structural pruning of astrocytic processes, thereby reducing astrocytic ensheathment of neuronal synapses. In a rat model fed a high-salt diet (HSD), activated microglia were observed surrounding hypothalamic vasopressin neurons, where they phagocytosed astrocytic processes; this was associated with reduced astrocytic coverage (in vivo imaging evidence from rats) [148]. This microglia‑mediated “de‑ensheathment” of astrocytes represents a dynamic mechanism for modulating the synaptic microenvironment. It not only fine‑tunes the spatiotemporal specificity of glutamate signaling but also contributes to the regulation of neuronal excitability balance, thereby preventing excessive excitation or inhibition and supporting the stability of neural circuits.

Molecular signal communication mechanism: Travis E. Faust et al [149] delineated a Wingless-related integration site (Wnt) signaling pathway initiated by neuronal activity, relayed by microglia, and ultimately modulating astrocyte morphology. Neuronal activity triggers microglial release of Wnt proteins via CX3CL1-CX3CR1 signaling, leading to astrocytic cytoskeletal reorganization and reduced astrocyte-synaptic contact, thereby creating space for synaptic remodeling (intracellular neonatal mouse model and in vitro cell experiments) (Fig. 5A). Astrocyte-derived IL-33 enhances microglial synaptic phagocytosis via the IL1RL1 receptor, a pathway critical for synaptic homeostasis during development, and in the adult hippocampus (an intracellular mouse model) [150, 151] (see also Sections Autism Spectrum Disorder and 3.4.1). Microglia further support normal neurodevelopment and synaptic function by secreting factors such as IL‑6 and LIF, which activate the Janus kinase/Signal Transducer and Activator of Transcription (JAK/STAT) and Mitogen-Activated Protein Kinase (MAPK) pathways in astrocytes, promoting their differentiation and maturation (in vivo and in vitro models) [152] (Fig. 5B). Furthermore, microglia and astrocytes exhibit a precise spatiotemporal division of labor in clearing cellular debris. Upon neuronal death, a single microglial cell typically migrates to and phagocytoses the soma. In contrast, astrocytes utilize extensive processes to remotely clear distal neurite fragments. This complementary spatial strategy enables the efficient and orderly removal of neuronal debris, which minimizes the inflammatory burden and preserves brain homeostasis (in vivo) [153].

Fig. 5.

Microglia-astrocyte crosstalk regulates synaptic plasticity. A Changes in neuronal activity regulate the release of CX3CL1 (possibly through a membrane protease cleavage mechanism), activating the expression of the CX3CR1 receptor in microglia and subsequently promoting the secretion of Wnt proteins through the Ca²⁺ signal mediated by Gi/Gz proteins, resulting in the regulation of signal transmission and synaptic remodeling among neurons, microglia, and astrocytes. B microglia activated by Aβ secrete inflammatory factors, activating astrocytes, enhancing their glycolytic metabolism, and generating excessive L-lactic acid, thereby disrupting synaptic plasticity and promoting Aβ aggregation. During normal neural development, microglia secrete IL-6 and LIF, which activate the JAK/STAT and MAPK pathways in astrocytes, thereby promoting their differentiation and maturation

Physical Contact and Functional Integration: Research has indicated that microglial and astrocytic processes establish physical contact via integrin β1 mechanoreceptors. This direct interaction not only promotes cytoskeletal remodeling in microglia, enhancing their migratory and phagocytic capacity, but also facilitates efficient intercellular signaling, thereby jointly modulating the astrocyte phenotype and functional state [154]. Inducible astrocyte‑specific knockout of gap‑junction proteins (connexin 30 and connexin 43) revealed that decoupling of the astrocytic network triggers widespread microglial activation, even in the absence of overt pathology (in vivo) [155]. These findings suggest that the integrity of astrocytic gap junctions regulates microglial functional status, reflecting their integrated role in maintaining neural network homeostasis.

Interactions disorder and synaptic damage under pathological conditions

In neuroinflammatory or neurodegenerative conditions, the physiological crosstalk between microglia and astrocytes is disrupted, shifting toward a mutually reinforcing pathogenic cycle that collectively drives synaptic dysfunction and loss.

The amplification of inflammatory signals and the transformation of A1-type astrocytes: Under pathological conditions, hyperactivated microglia release proinflammatory factors that drive astrocytes toward a neurotoxic A1 phenotype (in vitro experiments) [156]. A1 astrocytes lose their normal synaptic support functions and instead release neurotoxic factors that directly damage neurons and synapses, thereby exacerbating neurodegenerative processes [156]. This process is well documented in Alzheimer’s disease (AD) models: Aβ-activated microglia release inflammatory factors that induce astrocytes to adopt an A1 phenotype. This transition enhances astrocytic glycolysis and lactate release, impairing LTP and cognition [112] (see also Section 3.5.1.2). In addition to signals originating from microglia, intracellular pathways such as NF‑κB within astrocytes themselves contribute to the coordination of their pathological polarization [157]. Notably, protective glial crosstalk also occurs under disease conditions. For instance, astrocyte‑derived interleukin‑3 (IL‑3) acts via the IL‑3Rα receptor on microglia, enhancing their chemotaxis, recruitment, and capacity to clear Aβ and tau aggregates, thereby exerting a protective effect that limits disease progression in Alzheimer’s disease models (in 5xFAD mice) [158]. This observation underscores the therapeutic potential of augmenting such specific protective interactions.

Pathological physical contact and hyperfunction: Systematic analyses across multiple animal models—including the TauP301S transgenic, TauPS2APP, and Trem2 knockout models—reveal a division of labor between astrocytes and microglia in terms of synaptic clearance. During complement C1q‑mediated synaptic elimination, astrocytes and microglia selectively target excitatory and inhibitory synapses, respectively. Moreover, astrocytic phagocytosis can compensate for impaired microglial function (in vivo experiments) [44]. The interplay and hyperactivity between these cell types are closely linked to alterations in synaptic plasticity. In addition to the secretion of proinflammatory factors such as TNF‑α, IL‑1α, and C1q to polarize astrocytes, as noted above, direct physical contact via integrin β1 further enhances microglial migration and phagocytosis [115]. While ostensibly reparative, such intensified interactions may instead drive pathological cytoskeletal reorganization in microglia, leading to dysregulated migration and phagocytosis that ultimately results in excessive synaptic elimination.

Thus, microglia-astrocyte crosstalk bidirectionally modulates synaptic plasticity, with protective versus detrimental outcomes depending on context. The interplay between microglia and astrocytes and their regulation of synaptic plasticity is supported by in vivo perturbation studies (microglial engulfment of astrocytic processes in a high-salt diet rat model [148]; CX3CL1-Wnt signaling in synaptic remodeling [149]), in vitro/ex vivo perturbation experiments (the IL‑33/Interleukin‑1 receptor‑like 1 axis modulating microglial phagocytosis [150]), human association and omics data (postmortem tissue studies in Alzheimer’s disease [44]), and conceptual inference (the framework of A1-type astrocyte conversion and neurotoxicity [156]). Causal links in glial signaling pathways come mainly from genetic and pharmacological models in rodents, whereas human tissue studies provide correlational evidence for the pathological role of glial crosstalk in neurodegenerative diseases.

Disruption of circadian rhythm mediates microglial regulation of synaptic plasticity: temporal integration

Microglia do not operate uniformly to regulate synapses; rather, their activity is co‑driven by intrinsic molecular clocks and sleep‑wake states. These cells possess a functional molecular circadian machinery, with core clock genes such as Per1, Per2, Rev‑erbα, and Bmal1 exhibiting robust oscillatory expression throughout the 24-h cycle. This rhythmic expression plays a pivotal role in modulating microglial function, enabling them to switch between “surveillance‑downregulation‑pruning” programs across different circadian phases. Consequently, microglia contribute macroscopically to the maintenance of network homeostasis and, at a microscopic level, provide temporally defined windows of synaptic plasticity that support learning and memory.

Homeostatic integration: circadian rhythms as a core driver of microglial synaptic support

Molecular and Cellular Foundations: Microglia express core circadian clock genes (such as Bmal1, Per, and REV-ERBα), forming a functional transcription-translation feedback loop, thereby endowing them with an intrinsic biological clock timing mechanism [159] (Table 4). This molecular clock acts as a primary driver, dynamically shaping the morphodynamics and secretory activities of microglia. For instance, the production and release of inflammatory mediators such as TNF‑α are not constant; instead, they are tightly regulated by core clock genes such as Bmal1 and Rev‑erbα, which exhibit clear circadian fluctuations [160, 161]. Furthermore, the expression of the core clock gene Per1 in microglia oscillates in a circadian manner and is modulated by purinergic receptor signaling, which further shapes microglial functional states and their regulatory effects on synapses.

Table 4.

Molecular clock genes: mechanisms and characteristics of synaptic regulation

Molecular clock gene	Core negative feedback transcription-translation loop	Synaptic plasticity regulation	Impact on nervous system diseases	Evidence Tags
Bmal1	BMAL1/CLOCK → Activates Per2, Rev-erbα/β, Rora transcription → REV-ERB proteins (repressor) and RORα protein (activator) competitively bind to the RRE on Bmal1 promoter → Rhythmic fine-tuning of Bmal1 expression  → Drives stable ~24-hour molecular oscillation  → Regulates output genes (e.g., Cathepsin S) [159].	The core circadian clock circuit of microglia → Periodic oscillations in expression of effector molecules such as Cathepsin S → Periodic secretion of Cathepsin S → Associated with diurnal fluctuations of synaptic functions (such as mEPSC) and structures (such as dendritic spine density) → Correlating with synaptic plasticity and neural circuit function [159].	The influence of Bmal1 on neurological diseases is systemic and bidirectional [159].	Genetic methods; Pharmacological methods; Intracellular study [261].
Per1	BMAL1/CLOCK heterodimer → Activates Per/CRY transcription → Cytoplasmic PER/CRY complex formation → Nuclear translocation → Inhibits BMAL1/CLOCK activity → Per/Cry transcription declines → PER/CRY degradation via ubiquitin-proteasome system → Inhibition released → BMAL1/CLOCK activity restored → Cycle repeats (~24 h period) [161].	The core circadian clock circuit of microglia (with rhythmic expression of genes such as Per1) → Periodic regulation of the expression of effector molecules (such as cathepsin S) → Periodic secretion of cathepsin S → Regulation of the diurnal fluctuations in synaptic functions (such as miniature excitatory postsynaptic currents/mEPSC) and structures (such as dendritic spine density) → Influencing synaptic plasticity and neural circuit function [161].	The impact on the nervous system is not isolated; rather, it occurs through its rhythmic expression [161].	Genetic methods; Ex vivo study [161].
Per2	BMAL1/CLOCK Heterodimer Formation → Per/Cry Gene Transcription → PER/CRY Complex Formation and Nuclear Translocation → Inhibition of BMAL1/CLOCK Activity → PER/CRY Degradation → Inhibition Release → Cycle Restarts [161].	Microglial Circadian Clock (Per2 rhythm) → Rhythmic phagocytosis/cytokine secretion → Synaptic pruning & microenvironment regulation → Diurnal synaptic function/structure modulation → Altered synaptic plasticity & neural circuit activity → Regulation of learning, memory & higher-order brain functions [161].	Same as Per1 [161].	Genetic methods; Ex vivo study; In vivo study; Human genetic association study [161].
Rev-erbα	BMAL1/CLOCK → Activates Per, Cry, Nr1d1 transcription → Formation of PER/CRY complex & synthesis of Rev-erbα protein → PER/CRY complex enters the nucleus → Directly inhibits the activity of BMAL1/CLOCK → Rev-erbα enters the nucleus → Bmal1 promoter is bound → Inhibits its transcription → PER/CRY and Rev-erbα are degraded → Inhibition is relieved → BMAL1/CLOCK regains activity → Cycle restarts (≈24 h) [161].	The rhythmic expression of Rev-erbα → Is associated with rhythmic production and release of inflammatory factors in microglia → Inflammatory factors (such as TNFα, IL-1β) act on neurons → Are associated with synaptic strength, pruning or stability → Affects synaptic plasticity [161].	Is associated with the occurrence, progression and repair of neurological diseases [161].	Pharmacology Genetics; In vivo study; Ex vivo study [161].
Cathepsin S	Per1/Per2 expression ↑ → Cathepsin S transcription ↑ → Cathepsin S protein synthesis ↑ → Cathepsin S protein activity ↑ → Per1/Per2 expression ↓ → Per1/Per2 expression decreases → Cathepsin S transcription ↓ → Cathepsin S protein synthesis ↓ → Cathepsin S protein activity ↓ → Removal of inhibition on Per1/Per2 → Circulation continues [159].	Neuronal activity/ATP/Inflammatory signals → Activation of microglia → Upregulation of Cathepsin S expression and secretion → Hydrolysis of two key substrates (neuronal surface signaling protein CX3CL1 and extracellular matrix around synapses) → Promotion of synaptic pruning and regulation of transmission efficiency [161].	The common pathological pathway connecting neuroinflammation, extracellular matrix degradation, synapse loss and dysfunction of neural networks [161].	Genetics; In vivo study; Ex vivo study; Molecular and cellular biology methods; Pharmacological research [169].

Temporal and spatial coordination of synaptic function: Against this rhythmic backdrop, microglia serve as key intermediaries that translate circadian signals into precise synaptic regulation. The circadian system modulates neuronal excitability within synaptic circuits, thereby shaping cellular physiology and cognitive function. Notably, microglia mediate daily time‑dependent synaptic plasticity through their intrinsic circadian clock and dynamic interactions with synapses. Under normal physiological conditions, microglia facilitate synaptic weakening during the light phase (the sleep period) via synaptic phagocytosis or the release of regulatory factors, whereas during the dark phase (the wake period), they allow synaptic strengthening, thereby generating a circadian rhythm in terms of synaptic transmission strength. Following microglial depletion, this rhythmic regulation is disrupted, leading to a failure of normal synaptic weakening during the sleep period (pharmacological evidence, mouse) [162]. Moreover, the circadian system acts as an internal timing mechanism that dynamically regulates transport processes and barrier permeability at the blood-brain interface [163]. This time-controlled mechanism influences the trafficking of inflammatory mediators, metabolites, and neurotransmitters across the barrier, thereby altering the homeostasis of neurons and their surrounding microenvironment. In doing so, it further supports the maintenance of neural network stability and facilitates adaptive remodeling [163].

Pathological desynchronization: disrupted rhythms as drivers of microglial dysfunction and synaptic loss

Disruption of circadian rhythms—due to factors such as shift work, trans‑meridian travel, aging, or chronic sleep deprivation—can impair this precise temporal coordination, leading to microglial dysfunction that in turn drives neuroinflammation and synaptic impairment.

Molecular clock dysregulation and functional disorders: Under conditions of circadian disruption, aging, or sleep deprivation, the expression of core clock genes (e.g., Per1 and Per2) becomes dysregulated in microglia (in 5xFAD mice) [164]. In models of circadian disruption, dysregulation of the molecular clock is associated with impaired downstream regulatory networks and is accompanied by excessive activation of microglial inflammatory responses and aberrant, non-rhythmic overproduction of inflammatory mediators such as TNF-α (associative evidence from mice) [164]. In aged mice with microglia‑specific BMAL1 deletion, the synaptic density increased, whereas complement C1q deposition and microglial phagocytic markers decreased, indicating impaired synaptic tagging and clearance despite microglial activation (in vivo) [165].

Sleep deprivation and environmental disruption. In a mouse model of 72-hour sleep deprivation, sleep loss downregulated the expression of the CX3CR1 signaling pathway in the hippocampus, leading to microglial activation accompanied by impaired synaptic pruning, as reflected by reduced internalization of PSD95, accumulation of synaptic proteins, and increased dendritic spine density. Ultimately, these changes result in deficits in spatial memory and cognitive performance (in vivo) [166]. Deficiency in microglia-specific TNFα impairs sleep-dependent memory consolidation by disrupting P2RX7-mediated ATP signaling and altering GABA-A receptor dynamics at inhibitory synapses (in vivo microglia-specific TNFα knockout mice; pharmacological P2RX7 blockade) [167].

Synaptic dysfunction and self‑perpetuating mechanisms. Circadian disruption impairs the temporal coordination of microglial function. This breakdown of spatiotemporal regulation directly compromises synaptic function and undermines the stability of neural networks (in vivo mouse models and molecular/electrophysiological evidence) [168]. In Alzheimer’s disease, where circadian abnormalities are already present, further rhythm disruption exacerbates microglial pro-inflammatory responses and aberrant phagocytosis, accelerating synaptic loss [164] (see also Section 4.1.1). Moreover, the age-related attenuation of circadian rhythm amplitude is linked to the emergence of a “senescence‑associated phenotype” in microglia, which collectively contributes to increased neuroinflammation and cognitive decline [161].

Thus, circadian disruption transforms microglia from homeostatic coordinators into synaptic disruptors. Therefore, restoring or stabilizing circadian rhythms may represent a novel therapeutic strategy for treating related neurological disorders through the modulation of microglial function.

Multiple lines of evidence support the regulation of microglial synaptic plasticity by circadian rhythms. These include in vivo perturbation studies (microglia‑specific BMAL1 deletion leading to increased synaptic density and impaired phagocytosis [165]; sleep deprivation downregulating the CX3CR1 signaling pathway and causing synaptic pruning deficits [166]; microglia‑specific TNFα deletion altering sleep slow-wave activity and impairing memory consolidation [167]), in vitro/ex vivo perturbation experiments (rhythmic expression profiling of core clock genes in microglia [161]; cathepsin S as a downstream effector of Bmal1 [169]), human association and omics data (links between circadian disruption and microglial dysfunction in Alzheimer’s disease [164]), and conceptual inference (the pathological amplification hypothesis that circadian disruption and microglial dysfunction form a self-perpetuating cycle [170]). Causal relationships between core clock genes and microglial functional outcomes are derived primarily from rodent genetic and pharmacological models, whereas human studies provide correlational evidence for the pathological role of circadian disruption in neurodegenerative diseases.

Regulation of synaptic plasticity by the microbiota-gut-brain axis and microglia: systemic integration

The gut‑brain axis enables bidirectional communication between the gut and the brain through multiple signaling pathways, including vagal nerve signaling, endocrine regulation, immune responses, and the actions of microbial metabolites. Together, these pathways form a complex network that modulates the function and homeostasis of the central nervous system. Within this axis, the gut microbiota serves as a crucial peripheral signaling source, producing a diverse array of metabolites and molecular signals that profoundly influence microglial maturation, activation state, and the capacity to regulate synapses (animal model; clinical observation) [171].

Homeostatic signaling: the microbiota as a core driver of microglial support for synaptic plasticity

Within this intricate network, microglia serve as a central relay—their function and maturation state are finely tuned by microbial signals, thereby profoundly shaping synaptic plasticity and neurodevelopment (in germ-free mice and animal models) [172, 173].

Microbial metabolites as drivers. Metabolites produced by the gut microbiota—especially short‑chain fatty acids such as acetate, propionate, and butyrate—are major factors that shape microglial phenotype and function [174]. In germ-free (GF) mice, the Prinz team demonstrated that gut microbes are required for shaping the microglial phenotype. GF mice exhibit immature microglial transcriptomic profiles and abnormal morphology; oral administration of short-chain fatty acids (SCFAs) rescues these deficits, indicating that microbial metabolites provide necessary signals for normal microglial maturation (genetic/pharmacological evidence from mice) [175]. Short‑chain fatty acids act on microglia by binding to specific receptors—such as FFAR2/3 and G‑protein‑coupled receptors, including GPR41 and GPR43—thereby modulating downstream signaling [176, 177].

Microglia as key intermediaries. Within this overarching framework, microglia serve as core mediators that translate peripheral microbial signals into precise synaptic regulation. Using patch‑clamp recordings in acute brain slices combined with molecular approaches, it was found that short‑term disruption of the gut microbiota by antibiotics in the adult hippocampus leads to increased microglial density, reduced surveillance capacity, altered synaptic arrangement, and decreased spontaneous glutamatergic currents as well as functional connectivity in the CA3-CA1 pathway without changing dendritic density. These effects were absent in CX3CR1‑deficient mice (in vivo adult mouse model; molecular and imaging analyses) [178]. These findings suggest that the microbiota modulates the functional coupling between microglia and synapses via the CX3CL1-CX3CR1 axis. In a mouse model of postoperative cognitive dysfunction, treatment with SCFAs in hippocampal tissue and primary microglia increased histone H3 acetylation levels, suppressed neuroinflammation, and modulated the complement C1q pathway, thereby significantly reversing excessive synaptic elimination in cognitively impaired mice (in vitro; in vivo) [179].

Synaptic support and neural circuit health. The functional outcome of this microbiota-driven microglial regulation is the maintenance of synaptic homeostasis and plasticity. SCFA-mediated reprogramming of microglia reduces the secretion of proinflammatory cytokines while increasing the release of neurotrophic factors, thereby supporting neuronal growth, synapse formation, and homeostatic maintenance—effects that have been associated with enhanced cognitive performance in preclinical models [176]. Mice housed in an enriched environment have significantly higher levels of SCFAs in their feces, and this effect is abolished by antibiotic-induced depletion of the gut microbiota. Oral administration of an SCFA mixture (butyrate, propionate, or acetate) mimics the effects of environmental enrichment, promotes visual cortex plasticity in standard‑housed mice, and reshapes microglial morphology (in vivo experiment) [180]. In addition, neurotransmitters or their precursors synthesized by the gut microbiota can reach the brain via vagal afferents or systemic circulation, directly modulating synaptic strength and neuronal excitability and influencing plasticity processes such as LTP and LTD [177] (Table 5).

Table 5.

Regulatory effects of intestinal microbial metabolites on the functions of microglia and synaptic plasticity

Regulatory signal	Source	Generation process	Mechanisms in the Gut-Brain Axis	The effect on microglia cells	Synaptic plasticity regulation	Evidence Tags
Acetic acid	Specific bacteria (such as Bacteroides, Prevotella, Bifidobacterium, etc) [176].	The intestinal bacteria produce through a fermentation pathway centered on acetyl coenzyme A [176].	Immune pathway; Nervous pathway; Endocrine pathway; Circulatory pathway [176].	Maintain the homeostasis of microglia; Provide energy and signals; Inhibit harmful activation [176].	Create a favorable microenvironment: synaptic pruning, release of neurotrophic factors, removal of cellular debris; inhibition of neuroinflammation; Repair the damage [176].	In vivo study; In vitro study; Pharmacology Clinical/Translational Research [176].
Propionic acid	The Bacteroides genus is the most important producer of propionic acid in the human intestinal tract. Other genera include Prevotella, Veillonella, and some bacteria from the Clostridium order [176].	Through the succinate-acetate pathway or acrylate pathway of intestinal bacteria, anaerobic fermentation is carried out to produce [176].	Circulatory and direct signaling pathways; receptor-mediated pathways; epigenetic regulatory pathways; Immune and endocrine pathways [176].	Bidirectional regulatory effect: promoting homeostasis, inhibiting activation, and protecting the blood-brain barrier; Activate the C/EBPβ/AEP inflammatory signaling pathway, induce the activation of microglia and exacerbate Aβ and tau pathologies, etc [176].	Create an anti-inflammatory neural environment; Support the synaptic support function of microglia; Collaborative effect with other SCFAs [176].	In vivo study; In vitro study; Human epidemiology [176].
Butyric acid	The bacteria of the Firmicutes phylum are the main producers, especially several clusters of the Clostridiales (such as Clostridium clusters IV and XIVa), including genera such as Clostridium, Roseburia, Eubacterium, and Butyricicoccus [176].	The acetyl-CoA is the most important pathway [176].	Local and systemic anti-inflammatory effects; Epigenetic regulation; Receptor signal transduction; Permeability of the blood-brain barrier and direct action; Endocrine indirect pathway [176].	Inhibit excessive activation of microglia, promote phagocytic clearance function, maintain stable phenotype; may activate the C/EBPβ/AEP inflammatory signaling pathway, induce microglial activation and aggravate Aβ and tau pathologies [176].	Eliminating inflammation inhibition, creating a neurotrophic environment, and performing epigenetic reprogramming, thereby inhibiting microglial cell activation and promoting synaptic plasticity [176].	Intracellular research; Pharmacology and epigenetics; Experiential research [176].
Indole derivatives	A variety of intestinal bacteria, such as Escherichia coli, etc [176].	The main precursor is tryptophan, which is converted through the microbial tryptophanase pathway [176].	Direct entry into the brain pathway; Core receptor pathway (dominant mechanism). The indirect pathway of intestinal initiation [176].	Inhibiting microglial cell-mediated neuroinflammation: Inhibiting the NF-κB pathway, inhibiting the NLRP3 inflammasome, promoting an anti-inflammatory phenotype; and indolizine sulfate has a neurotoxic effect in both AD and Parkinson’s Disease (PD) [176].	Inhibiting inflammatory responses that eliminate synaptic plasticity; Provide a supportive environment for synaptic maintenance [176].	In vivo study; Pharmacology [176].
UDCA	The liver first synthesizes primary bile acids, mainly CDCA [176].	The CDCA secreted into the intestine is biotransformed by intestinal microorganisms (especially those with 7α-dehydrogenase activity) to generate UDCA [176].	Direct pathway: UDCA can pass through the blood-brain barrier and enter the brain. Within the central nervous system, it acts as a signaling molecule and activates the membrane receptor FXR and the nuclear receptor TGR5. Indirect pathway: At the intestinal level, UDCA activates TGR5 and FXR on intestinal cells [176].	UDCA is an effective inhibitor of microglial cell activation and neuroinflammation [176].	Remove the inflammatory barrier that inhibits synaptic plasticity; Protect the integrity of synaptic structures [176].	Ex vivo study; In vivo study; Clinical research [176].
TUDCA	TUDCA is the taurine-bound form of ursodeoxycholic acid and belongs to the secondary bile acids [176].	Intestinal microbiota transformation (core step); Host liver conjugation [176].	Direct pathway (core): TUDCA can effectively pass through the blood-brain barrier and enter the brain. In the brain, it directly activates nuclear receptor FXR and membrane receptor TGR5; indirect pathway (amplification by intestinal hormones) [176].	TUDCA is an efficient inhibitor of neuroinflammation mediated by microglia [176].	Remove the inflammatory barrier, create a plastic environment; protect the integrity of synaptic structures; restore the efficacy and plasticity of synaptic connections [176].	In vivo study [262].
ω-3 fatty acids	Dietary intake	The intestinal microbiota can convert a portion of polyunsaturated fatty acids into bioactive metabolites [176].	Directly enters the brain and achieves structural integration; precursor of signaling molecules; systemic anti-inflammatory effect. The indirect pathway is regulated by the microbiota [176].	Promote phenotypic transformation; Enhance phagocytic ability; Inhibit neuroinflammation [176].	Indirect approach: By creating a “plasticity-friendly” environment through microglia. Direct and collaborative pathways: Direct protection and support for neurons [176].	In vivo study; Ex vivo study; Human observational studies [122].

Pathological dysbiosis: disrupted microbiota as a driver of microglial dysfunction and synaptic damage

When the homeostatic microbial driver is compromised—resulting in dysbiosis of the gut microbiota—finely tuned communication along the microbiota‑gut‑brain axis is disrupted, transforming microglia from synaptic supporters into pathological mediators of neuroinflammation and synaptic loss.

Loss of beneficial signals. Under pathological conditions, gut dysbiosis leads to reduced production of protective microbial metabolites such as short‑chain fatty acids. This loss of beneficial signals deprives microglia of crucial homeostatic inputs, thereby weakening their neuroprotective functions. The loss of SCFAs deprives microglia of this homeostatic input, weakening their neuroprotective functions (in 5xFAD mice) [176]. This protective mechanism is supported by intervention studies: FMT from young to aged mice reversed age-related changes in key genes related to microglial “sensing” in the hippocampus. For example, it restored the expression of the complement component C1qb—a gene involved in synaptic pruning whose reduction attenuates cognitive decline—and modulated other genes linked to inflammation and cognition, such as Gpr84 and Fcgr2b (in vivo) [181].

Emergence of harmful signals. Moreover, gut dysbiosis is associated with elevated levels of detrimental microbial drivers such as lipopolysaccharides. Once LPS enters the bloodstream, it activates the Toll‑like receptor 4 (TLR4) pathway on microglia, triggering the release of proinflammatory cytokines (e.g., IL‑1β and TNF‑α) that directly damage neurons and synapses [180]. This neuroimmune crosstalk, shaped by the microbiota, can drive microglia-mediated neuroinflammation and synaptic remodeling [182].

Synaptic dysfunction and self‑perpetuating mechanisms. The pathological outcome of these damaging factors is disruption of synaptic homeostasis. Under conditions such as chronic stress, this process becomes exacerbated and leads to measurable damage. Mice subjected to chronic unpredictable mild stress (CUMS) exhibit gut microbiota dysbiosis (e.g., increased abundance of Proteobacteria), elevated circulating levels of lipopolysaccharide, and activation of the complement component C3 both peripherally and centrally. These changes promote aberrant microglial phagocytosis of synapses in the prefrontal cortex via the C3/CR3 signaling pathway, resulting in the loss of synaptic proteins (e.g., SYN and PSD95) and ultimately leading to depressive‑like behaviors and cognitive impairment (in vivo) [183].

The regulation of synaptic plasticity by the gut-brain axis through microglia is supported by multiple lines of evidence. These include in vivo perturbation studies (in which depletion of the microbiota in germ‑free mice leads to immature microglia, a defect reversible by short‑chain fatty acid supplementation [175]; antibiotic‑induced disruption of the microbiota reduces microglial surveillance and alters synaptic functional connectivity [178]), in in vitro/ex vivo perturbation experiments (in which treatment of primary microglia with SCFAs increases histone H3 acetylation and suppresses neuroinflammation [179]), human association and omics data (postmortem tissue and genetic studies linking dysbiosis to neurodegenerative diseases[176]), and conceptual inference (a framework of dual imbalance—loss of beneficial signals coupled with the emergence of harmful signals—driving a pathological cycle [176]). Causal links between microbial metabolites and microglial function are derived primarily from rodent genetic and pharmacological models, whereas human studies provide correlational evidence for the pathological role of the gut-brain axis in neurological disorders.

Cross-module interactions: a networked view of synaptic regulation in microglia

The eight regulatory modules described above are not independent parallel pathways but rather form an integrated network in which each module interacts with the others through specific molecular and cellular mechanisms. These interactions give rise to a functional architecture that follows a driver-context-outcome logic, where upstream signals converge onto the microglial state to determine synaptic outcomes.

Functional stratification

These modules can be categorized according to their position within the regulatory network. Morphology (see Section Microglia and Synaptic Plasticity in NDDs) and apoptosis (see section State transitions and functional polarization of microglia in neurodegenerative diseases) constitute the execution core, directly governing synaptic remodeling—the former provides the structural interface for contact and engulfment, whereas the latter supplies molecular “eat‑me” signals such as those of phosphatidylserine that tag synapses for removal. These execution mechanisms are built upon an upstream regulatory layer. Epigenetics (see Section Microglia and Synaptic Plasticity in Psychiatric Disorders) establishes transcriptional competence, shaping how microglia respond to environmental stimuli. Metabolism (Section 3.5) supplies the energy and biosynthetic capacity needed for processes such as motility, phagocytosis, and cytokine secretion. The extracellular matrix (ECM; Section 3.4) defines the physical substrate in which synapses are embedded and modulates the accessibility of complement proteins such as C1q. Building on these regulatory layers, glial-glial interactions (Section 3.6) coordinate microglial activity with astrocytic networks. Finally, circadian rhythms (Section 3.7) and the gut-brain axis (Section 3.8) act as temporal and systemic integrators, respectively—they impose daily rhythms on module activities and link peripheral signals to central microglial function.

Interaction mechanisms

The specific interactions among modules reveal the network nature of microglial regulation. Metabolism-morphology coupling is evident because cell motility depends on glycolytic flux, and the Rac1 signaling pathway—which controls the frequency of synaptic contacts—requires energy derived from glucose metabolism (see Section Microglia and Synaptic Plasticity in NDDs). An apoptosis-ECM interaction is reflected in the requirement of intact perineuronal nets for proper C1q localization; disruption of the ECM impairs the accurate tagging of apoptotic synapses (see sections State transitions and functional polarization of microglia in neurodegenerative diseases, 3.4). Glial-epigenetic coordination occurs through the action of astrocyte‑derived IL‑33 on the Interleukin‑1 receptor‑like 1 (ST2) receptor on microglia, enhancing the phagocytosis of excitatory synapses while also modulating microglial transcriptional programs (see Sections Microglia and Synaptic Plasticity in Psychiatric Disorders, 3.6). Integration of circadian rhythms with ECM remodeling may occur via cathepsin S, a downstream effector of Bmal1 (Sections 3.4, 3.7). Gut-brain-immune interactions are exemplified by short‑chain fatty acids, which promote oxidative phosphorylation over glycolysis, modulate histone acetylation in microglia, and increase the precision of phagocytosis (see Sections Microglia and Synaptic Plasticity in Psychiatric Disorders, 3.5, 3.8) (Fig. 6).

Fig. 6.

Multidimensional causal network of microglia-mediated synaptic plasticity. This figure illustrates the multidimensional causal regulatory network underlying microglia-mediated synaptic plasticity. Macroscopic upstream signals—circadian rhythms, astrocytic communication, and the gut-brain axis—reprogram microglial internal modules, including epigenetics, metabolism, and the extracellular matrix (ECM). This reprogramming drives morphological changes and activates apoptotic clearance mechanisms such as the C1q/C3 pathway. Under physiological conditions, this cascade leads to precise synaptic pruning and optimization of neural circuits

Network logic

Under normal physiological conditions, the interactions among modules remain tightly coordinated. Circadian rhythms control the time window for metabolic activation, which in turn supports precise synaptic pruning, whereas the epigenetic state ensures that responses are always tailored to specific environmental conditions. Under pathological conditions, however, a disturbance in one node of the network can propagate through the entire system, amplifying dysregulation across multiple modules. Gut dysbiosis impairs microglial metabolism and epigenetics, triggering aberrant complement activation (see Section 3.8). Similarly, circadian disruption weakens the rhythmic expression of phagocytic receptors, exacerbates metabolic dysregulation, and amplifies glial inflammatory responses. Therefore, understanding cross‑module interactions is essential for developing therapeutic strategies that target the regulatory network as a whole rather than isolated pathways.

Role of microglial regulation of synaptic plasticity in neuropathological states

While microglia-mediated synaptic pruning is essential for circuit refinement under physiological conditions, its dysregulation in neurodegenerative and psychiatric disorders drives pathological synapse loss and cognitive decline [121].

Microglia and synaptic plasticity in NDDs

NDDs affect the CNS (brain, retina, and spinal cord), encompassing disorders ranging from childhood congenital leukodystrophy to age-related AD, PD, and Huntington’s disease [184]. Their core features include [185] (1) pathological protein aggregation; (2) synaptic and neural network dysfunction; (3) abnormalities in proteostasis, the cytoskeleton, and energy metabolism; (4) DNA/RNA defects; (5) inflammatory responses; and (6) neuronal death. These features interact synergistically to drive disease initiation and progression.

In NDDs, microglia are tightly linked to synaptic dysfunction, and their role is particularly prominent in regulating synaptic plasticity. They act not only as “scavengers” for amyloid clearance but also as key “regulators” of neuronal function and homeostasis [186]. NDD-associated protein aggregates—such as β-amyloid (Aβ), tau, α-synuclein (α-syn), prion protein (PrP) fibrils, and mutant superoxide dismutase 1 (mutant SOD1)—can induce microglial activation [187]. Activated microglia play dual roles: they limit the spread of pathological proteins (e.g., Aβ and tau) via phagocytosis; conversely, overactivation may accelerate protein propagation and promote neurodegeneration [54], reflecting their complex role in NDDs. However, there are disease-dependent differences in the activation state of microglia, metabolic reprogramming, and the way in which synapses are regulated.

Alzheimer’s disease

Alzheimer’s disease (AD), the most prevalent neurodegenerative disorder, is characterized by progressive cognitive decline. Although β-amyloid (Aβ) plaques and neurofibrillary tangles (TAs) are central to AD pathogenesis, recent research highlights synaptic dysfunction and loss as the pathological events most closely correlated with cognitive impairment [39].

Recent advances in imaging, biomarker detection, and multi‑omics analyses have improved our understanding of the synaptic pathology underlying AD. Synaptic injury in AD arises from a complex interplay of multiple factors, including pathological protein aggregation, excitotoxicity, axonal injury, and neuroimmune inflammation (imaging technology; detection of liquid biomarkers; multimodal analysis) [188–190]. Among these, microglia-mediated, complement-dependent aberrant synaptic pruning has emerged as a pivotal mechanism linking upstream pathology to downstream synaptic loss [39]. In mouse models of tauopathy, abnormal tau aggregation not only disrupts neuronal structure but is also associated with the activation of microglial immune responses; this activation is correlated with excessive synaptic elimination (genetic evidence from mice) [191]. For example, reduced expression of the synaptic protein synaptogyrin‑3 alleviates tau‑induced memory deficits and synaptic loss, a process accompanied by microglial activation (in Tau P301S mice) [191]. These observations underscore the critical role of immune cells in AD synaptic pathology. As detailed in Section Parkinson’s disease (PD), the classical complement pathway (C1q-C3-CR3) drives aberrant microglial synaptic phagocytosis in AD, contributing to synapse loss (in 5xFAD mice; in Tau P301S mice) [39, 192] (Fig. 7).

Fig. 7.

The classical complement pathway mediates synaptic clearance and excitation/inhibition imbalance. C1q acts as the initiating factor of the classical pathway and binds and labels synapses, promoting the activation of the complement cascade reaction, which subsequently leads to the cleavage and deposition of C3. The activated products of C3 further facilitate the recognition and phagocytosis of labeled synapses by microglia, resulting in synaptic clearance. Moreover, it promotes the release of proinflammatory cytokines and chemokines by microglia, increases their phagocytic activity and migration ability, and contributes to neuroinflammation and an imbalance of excitatory and inhibitory neurotransmitters. Additionally, the neuronal pentamer proteins NPTX2, TREM2, and APOE can bind to C1q to regulate complement activity.

In AD, a subset of TREM2-expressing microglia aggregates around Aβ plaques, and in mouse models, this aggregation is thought to form a protective barrier that limits plaque expansion and neuritic damage (in 5xFAD and CRND8 mice) [193]. However, the precise role of TREM2 in AD remains debated. Evidence from mouse models indicates that TREM2 signaling helps maintain microglial survival, phagocytic capacity, and metabolic homeostasis. However, under chronic pathological conditions, treatment with a chronic TREM2 agonistic antibody in Aβ mice enhances peri‑plaque microglial activation but exacerbates tau‑related neuritic degeneration and neuronal injury (in 5xFAD mice; AD‑tau injection; in vitro BMDM) [194]. Thus, on the basis of current preclinical studies, TREM2 exerts context-dependent dual effects, highlighting the considerable complexity of targeting it therapeutically. Nevertheless, its clinical translational value still requires further validation in human systems.

However, most of the mechanistic evidence on TREM2 comes from mouse models; direct causal evidence in humans remains limited, and postmortem analyses are correlational (in 5xFAD mice; in tauopathy mouse models; GWAS association; mechanism unresolved) [195, 196] (see Section Necessity of microglia in synaptic plasticity: Controversy and boundary conditions for methodological considerations). Moreover, analyses of postmortem tissue are correlational by nature, making it difficult to determine causality—that is, whether microglial dysfunction is a driver of synapse loss or a secondary response to neuronal injury [37]. Therefore, therapeutic strategies for AD should not simply involve broad suppression or activation of microglia; instead, they should aim for precise reshaping of their functional state. However, such approaches are still in the preclinical exploration stage.

Multiple levels of evidence support the central role of microglia in synaptic damage in Alzheimer’s disease. In vivo perturbation studies—such as those using C1q, C3, or CR3 knockout mouse models—show reduced synaptic phagocytosis and improved cognitive function [8]; antibodies targeting C1q in tauopathy mice alleviate synapse loss [192]. In vitro and ex vivo mechanistic studies have demonstrated that Aβ oligomers directly induce microglial preferential phagocytosis of PS‑exposed synapses [58]. Human genetic and omics data reveal that AD risk genes such as TREM2 and BIN1 are enriched in microglial enhancer regions and are linked to synaptic signaling pathways [79]. Conceptual frameworks propose complement‑mediated aberrant synaptic pruning as a core link connecting upstream pathology to downstream cognitive decline [39]. Causal relationships have been well validated in animal models, while human genetic and pathological studies further support the role of microglia as key drivers of synapse loss in AD.

Parkinson’s disease (PD)

Parkinson’s disease (PD) is a common neurodegenerative disorder characterized by motor symptoms (e.g., tremor, rigidity, and bradykinesia) and non‑motor manifestations (e.g., cognitive impairment and depression). Its core pathological features include progressive loss of dopaminergic neurons in the substantia nigra pars compacta and the formation of Lewy bodies, which are primarily composed of misfolded α‑synuclein aggregates [39]. In recent years, the research paradigm has shifted toward recognizing synaptic dysfunction as among the earliest pathological events in PD, potentially preceding overt neuronal loss [197]. Simultaneously, microglia—key effectors of innate immunity in the central nervous system—play a central role in driving synaptic damage and disease progression through their aberrant activation.

The core pathological alterations in PD—including the prion‑like spread of α‑synuclein (animal models; in vitro experiments; insufficient direct clinical evidence) [198], oxidative stress secondary to mitochondrial dysfunction [199], and the resulting imbalance in excitation/inhibition in neurons [200]—are associated not only with direct synaptic damage but also with sustained activation signals that may influence microglial responses (intracellular and extracellular electrophysiology, molecular, and genetic manipulation techniques). Consequently, microglial activity has been proposed as a critical hub linking upstream molecular pathology to downstream synaptic loss and neural circuit dysfunction. Recent studies suggest that microglia can modulate the functional and structural integrity of synapses in PD through multiple mechanisms. As detailed in Section Parkinson’s disease (PD), aberrantly activated microglia overexpress complement components, leading to excessive synaptic pruning in PD models (in animal models and in vitro studies) [201]. Neuroinflammation‑mediated synaptic toxicity, driven by cytokines such as TNF‑α and IL‑1β, also contributes to PD pathology (see Section Parkinson’s disease (PD)) [82]. Pathological α-synuclein aggregates, a hallmark of PD, also closely interact with microglial function. Studies using in vitro and in vivo models have demonstrated that these misfolded aggregates can bind to pattern‑recognition receptors on microglia, such as TLRs and TREM2, with evidence suggesting that this interaction promotes inflammatory responses and enhances microglial phagocytic activity [202]. Although such activation may facilitate aggregate clearance, it has also been implicated in collateral damage to adjacent synapses, potentially contributing to the loss of neural function.

On the basis of findings from animal models, microglia are considered potential hubs in Parkinson’s disease pathogenesis. They may contribute to linking upstream events—such as α‑synuclein aggregation, mitochondrial damage and oxidative stress—to downstream synaptic loss and neural circuit dysfunction, possibly through mechanisms including synaptic pruning and the release of inflammatory cytokines. To date, direct intervention studies targeting microglia (e.g., selective depletion or pathway‑specific modulation) remain largely confined to animal models, and clinical evidence is scarce. Therefore, although microglia are biologically relevant to Parkinson’s disease pathogenesis, their clinical development as therapeutic targets still faces considerable challenges.

Several lines of evidence support the involvement of microglia in synaptic damage in Parkinson’s disease. In vivo perturbation studies have shown that in α‑synuclein-overexpressing mice, activation of the microglial complement pathway mediates synapse loss, and inhibition of microglia alleviates synaptic injury [203]. In vitro mechanistic studies have demonstrated that α‑synuclein directly activates microglia via TLR2/TLR4, inducing the release of inflammatory factors such as TNF‑α that impair synaptic function [202]. Conceptual inference positions microglia as key nodes linking protein aggregation, mitochondrial oxidative stress, and synaptic dysfunction [204]. It should be noted, however, that current causal evidence comes mainly from rodent models, and direct evidence from human Parkinson’s tissue still needs to be accumulated. Whether microglia constitute a causal hub in human PD requires further investigation.

Huntington’s disease

Huntington’s disease (HD) is an inherited neurodegenerative disorder caused by a mutation in the huntingtin gene. Its core pathological features include extensive loss of striatal projection neurons and neuronal dysfunction in multiple brain regions, such as the cortex and thalamus (animal models; human pathology) [205]. Growing evidence indicates that abnormalities in synaptic structure and function play critical roles in the onset and progression of HD. Like Alzheimer’s disease and Parkinson’s disease, HD is characterized by a reduction in the number of synapses and impairment of synaptic function, which strongly correlate with the progressive decline in cognitive and motor performance [185].

The synaptic pathology observed in Huntington’s disease (HD) is manifested primarily by the breakdown of excitatory synaptic connections and dysfunction in synaptic transmission. Notably, within excitatory projection pathways such as the cortico‑striatal and thalamo‑striatal circuits, synaptic inputs are markedly diminished, leading to functional deficits in neural circuitry (in vivo and in vitro electrophysiological and morphological studies of the genetic HD mouse model) [206]. Microglia play a central role in this process, which can be understood through three interrelated aspects. First, mutant huntingtin (mHtt) directly activates microglia. Immunohistochemical studies in human postmortem HD brains have shown that mutant huntingtin (mHtt) is also expressed within microglia, suggesting a potential cell-autonomous activation mechanism (human postmortem, associative) [197]. Once activated, microglia transition into a proinflammatory state, releasing abundant neurotoxic mediators such as cytokines, chemokines, and reactive oxygen species. This process not only directly exacerbates the toxic microenvironment surrounding neurons but also illustrates how pathology in HD propagates via non‑cell‑autonomous mechanisms—spreading from compromised neurons to the immune system of the brain—thereby amplifying damage to neural circuits, particularly synaptic function [197]. Second, as detailed in Sections Parkinson’s disease (PD) and 4.1.1, aberrant microglial synaptic pruning—driven by chronic inflammation—likely contributes to synaptic loss in HD, although direct evidence remains limited [207, 208]. Finally, synaptic dysfunction and protein aggregation act as danger signals that sustain microglial activation, resulting in a self‑amplifying cycle of neuroinflammation and synaptic damage (in 5xFAD mice; in Tau-P301S mice) [185] (see Section 4.1.1).

The potential role of microglia in synaptic pathology in Huntington’s disease is supported primarily by in vivo evidence (the expression of mutant huntingtin protein in microglia drives their autonomous activation; microglial activation occurs before overt neuronal loss in HD animal models [197]), in vitro mechanistic studies (mutant huntingtin induces a proinflammatory phenotype in microglia, leading to the release of IL‑1β, TNF‑α, and other inflammatory factors that can damage synapses [197]), and conceptual inference (microglia may serve as an intermediary linking protein toxicity to synapse loss through aberrant synaptic pruning and inflammatory amplification [209]). At present, direct causal evidence from human sources is lacking in this field.

State transitions and functional polarization of microglia in neurodegenerative diseases

During disease progression, microglia undergo phenotypic transitions, resulting in distinct subsets with specific transcriptional signatures, such as disease‑associated microglia (DAMs) or microglia in neurodegeneration (MgND). However, current evidence for these subsets derives primarily from mouse models (associative evidence from human postmortem tissue; functional validation in human disease and causal relationships to synaptic outcomes not yet established) [185, 210]. The upregulation of genes involved in cytokine production, phagocytosis, and reactive oxygen species suggests that these subsets may possess enhanced clearance capacity or altered metabolic profiles, ultimately influencing synaptic function and neuronal survival.

Under the persistent stimulation of neurodegeneration, microglia not only undergo phenotypic polarization but also initiate substantial local proliferation, resulting in the formation of dense cellular networks. For example, in prion disease models, the increase in microglial numbers primarily stems from their clonal expansion rather than infiltration from peripheral sources—a process tightly regulated by signaling axes involving CSF1R and transcription factors such as PU.1 and C/EBPα. These mechanisms are similarly active in human diseases such as Alzheimer’s disease (in prion disease mouse models and human AD and vCJD brains) [211]. Inhibiting pathological microglial proliferation can mitigate neuronal damage, suggesting that modulating microglial network expansion represents a potential therapeutic strategy.

Microglia and synaptic plasticity in psychiatric disorders

The conceptualization of psychiatric disorders has undergone a paradigm shift—moving from the traditional “chemical imbalance” framework to a focus on three interrelated pillars: genetic risk, altered molecular/cellular developmental functions, and brain circuit dysregulation. Psychiatric disorders are characterized by abnormalities in key neural circuits, including the default mode network (DMN), negative affect circuit, reward circuit, and cognitive control circuit. These circuits exhibit functionally correlated activation patterns during resting states or emotional/cognitive task performance (GWAS association; in vitro; human neuroimaging; correlational evidence) [212]. Notably, the high heterogeneity and complexity of these disorders arise from dysfunction in distinct brain networks, which directly correlate with alterations in mood, cognition, and behavior.

Aberrant microglia-mediated pruning has been linked to disrupted neural circuit function, directly contributing to core symptoms of neurodevelopmental and psychiatric conditions such as ASD and schizophrenia (in vivo; in vitro; genetic association study) [213].

Autism spectrum disorder

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by significant heterogeneity. Its core features include impairments in social communication, difficulties in verbal and nonverbal interaction, and the presence of restricted, repetitive patterns of behavior. The etiology of ASD is multifactorial and involves a complex interplay of genetic, environmental, immune, epigenetic, and neural‑circuit abnormalities, all of which contribute to its onset and progression.

Autism spectrum disorder (ASD) arises from a complex interplay of genetic susceptibility, environmental influences, and neuro‑immune dysregulation, with its core pathology rooted in disrupted neural circuit formation and synaptic function during early brain development (transgenic mice) [214]. Genetic studies have indicated that ASD risk genes are predominantly involved in synaptic function and chromatin remodeling. These genes regulate the expression of multiple downstream targets and strongly affect neurodevelopment, especially during fetal stages [215]. Concurrently, environmental factors—notably maternal immune activation (MIA)—can trigger neuroinflammation during critical developmental windows, leading to persistent activation of immune cells, including microglia (in vivo) [216]. As detailed in Sections Parkinson’s disease (PD) and State transitions and functional polarization of microglia in neurodegenerative diseases, in ASD, microglia engage in aberrant complement-mediated synaptic pruning. Additionally, the expression of ASD-associated risk genes in microglia (e.g., TREM2, CX3CR1, and ADNP) leads to cell-autonomous deficits in phagocytosis and neuron-microglia communication (in vivo mouse model; in vitro human brain organoid model) [217, 218]. This convergence of genetic, immune, and synaptic dysregulation ultimately results in aberrant neural connectivity and functional network development in key brain regions (e.g., the cortex and cerebellum), forming the neurobiological basis for the core symptoms of ASD, including social deficits and repetitive behaviors.

The role of microglia in ASD-related synaptic abnormalities is complex and supported by multiple lines of evidence. In vivo perturbation studies have shown that in maternal immune activation mouse models, altered microglial function is associated with aberrant synaptic pruning [219]; mice deficient in ASD‑related genes such as CX3CR1 or TREM2 exhibit synaptic developmental abnormalities[218]. Human genetic and omics data indicate that ASD risk genes are enriched in genes related to synaptic function and chromatin remodeling, with some risk genes highly expressed in microglia [215]. In vitro human cell models have demonstrated that in SCN2A‑deficient human brain organoids, microglia excessively prune synapses [217]. Taken together, these findings suggest that microglia in ASD are better viewed as “important modulators” that integrate genetic susceptibility and environmental risk signals, influencing synaptic refinement during critical developmental windows, rather than serving as a single causal hub.

Schizophrenia

Schizophrenia is a prevalent and severe psychiatric disorder characterized by a constellation of clinical features, including cognitive, positive, and negative symptoms. The disorder typically arises from the interplay of early‑environmental and genetic risk factors that alter neurodevelopmental trajectories, ultimately contributing to disease onset.

The pathogenesis of schizophrenia involves a multistage process encompassing neurodevelopmental disturbances and multisystem dysfunction. Genetic risk factors, including polygenic variations and specific copy‑number variants, interact with early‑environmental exposures (e.g., obstetric complications and childhood trauma) to alter typical brain developmental trajectories, particularly by disrupting normal synaptic formation and pruning. Later, psychosocial stressors—such as urban environments or migrant status—can then activate stress‑responsive brain circuits (e.g., the amygdala‑prefrontal pathway), leading to dysregulation of the mesolimbic dopamine system and the emergence of positive symptoms such as hallucinations and delusions (in vivo evidence) [220].

Findings from human genetic studies (genome‑wide association studies, postmortem analyses) and human iPSC‑derived microglia-neuron coculture models have revealed that increased copy number and elevated expression of the C4A gene are associated with an increased risk of SCZ. This has been interpreted as a genetic basis for excessive neurite pruning [221]. This, in turn, could disrupt the balance between cortical excitatory (glutamatergic) and inhibitory (GABAergic) transmission, impair gamma oscillations and neural network synchronization, and ultimately manifest as cognitive deficits and negative symptoms. As detailed in Section Parkinson’s disease (PD), the “over‑pruning” hypothesis posits that C4 activates the complement cascade (C1q-C3-CR3), leading to excessive synaptic elimination and reduced cortical connectivity in schizophrenia (animal model; GWAS association; mechanism analysis part) [222, 223]. However, recent studies have revealed a more refined regulatory landscape: C4‑mediated synaptic loss can also occur through a novel intracellular pathway that operates independently of both CR3 and microglial phagocytosis. Specifically, the neuronally overexpressed C4 protein interacts with the endosomal sorting protein SNX27, disrupting the recycling of the AMPA receptor subunit GluR1 and causing its aberrant accumulation in lysosomal compartments. This process directly impairs synaptic structure and function from within the neuron (in vivo) [224]. Thus, microglia may act as secondary responders, clearing synapses already rendered vulnerable by C4-driven intracellular pathology. Together, these layers constitute the multilevel immunopathological basis of the role of synaptic dysregulation in the neurodevelopmental disruption characteristic of schizophrenia.

The involvement of microglia in synaptic dysregulation in schizophrenia is supported by several lines of evidence: human genetic studies (strong association between C4 gene copy number variation and disease risk, implicating the complement-microglia pathway in excessive synaptic pruning [222]), human iPSC‑derived models (enhanced synapse loss in neuron-microglia cocultures overexpressing C4 [225]), and a conceptual framework (the “excessive pruning hypothesis” linking genetic risk to reduced cortical connectivity [222]). It is worth emphasizing, however, that recent studies have shown that C4 can also directly damage synapses through microglia‑independent, intracellular pathways, suggesting that microglia may act as secondary responders rather than the sole executors. Therefore, in schizophrenia, microglia should be viewed as key nodes within a multilayered pathological network rather than as isolated causal hubs.

In the pathological context of psychiatric disorders, microglial function is perturbed by dysregulated epigenetic and transcriptional programs, which manifests as 1) the overexpression of proinflammatory genes and 2) the suppression of homeostatic genes. Such transcriptional dysregulation impairs the capacity of microglia to modulate synapses, potentially leading to abnormal synaptic loss or functional impairment—ultimately compromising neural network stability and information processing efficiency. Epigenomic studies further revealed that microglial gene expression is finely tuned by a repertoire of regulatory elements and epigenetic marks (e.g., DNA methylation and histone modifications). Disruption of these regulatory mechanisms in psychiatric disorders directly underlies impaired microglial control of synaptic plasticity.

Beyond transcriptional regulation, microglial morphology and dynamic behavior are governed by glutamatergic and GABAergic neurotransmitter signaling—dysregulation of this interaction can disrupt neurotransmitter homeostasis. Additionally, microglia promote synaptic plasticity by remodeling the ECM; dysregulation of this remodeling disrupts synaptic structure and function, precipitating cognitive and behavioral deficits [226].

Methodological considerations and standardization suggestions

Single‑cell RNA sequencing, qRT‑PCR, and digital PCR have become essential for studying cellular heterogeneity and gene regulation, but their complex and variable workflows hinder their accuracy and reproducibility. In microglia‑synapse research, methodological pitfalls—including pseudoreplication, non‑blinded region‑of‑interest selection, and thresholding/segmentation variability—can exaggerate observed pruning or plasticity effects. Given the field’s reliance on imaging, electrophysiology, and behavioral quantification, ensuring reliability and biological relevance demands a rigorous distinction between biological and technical replicates, alongside standardized experimental and analytical workflows (see Table 6 for common metrics, biases, and recommended practices).

Table 6.

Quantitative bias analysis and standardized practice recommendations for studies on microglial cell synaptic plasticity

Research Field	Common quantitative indicators	Common methodological biases	best practices	Definition of ROI	Partitioning method	Statistical independence	Consistency of scoring
Morphodynamic analysis	Cell body size; Complexity of branches; Fractal dimension; Morphological classification and scoring; Color intensity/Expression level; Highly complex branches; Three-dimensional morphological analysis [263].	Errors introduced by imaging and measurement techniques; deviations in marker selection and interpretation; errors due to the lack of analytical methods and standardization; deviations caused by uncontrolled biological variables; deviations in sample processing and preparation; deviations between static thinking and dynamic processes [263].	Method transparency; multidimensional verification; carefully designed control; selection of appropriate technology [263].	Tumor area (TUM); Interface area (INT); Peritumoral brain area (PBZ); Contralateral control area (CL) [264].	Automated morphometric quantification; Cluster analysis [264].	Biological replicates (individual animals or human donors) serve as independent statistical units. Single cells, fields of view, and tissue sections are nested subsamples. Mixed-effects models (e.g., outcome ~ condition + (1|animal) + (1|slide)) or pseudobulk approaches should be used to avoid pseudoreplication [264].	By combining multiple morphological parameters with different analytical methods (such as principal component analysis (PCA), cluster analysis, etc.) to achieve [264].
Synaptic pruning event	Synaptic density; content of synaptic proteins in microglia; frequency and duration of synaptic contacts; frequency and amplitude of mEPSC (microexcitatory postsynaptic current) and the current ratio of AMPAR/NMDAR [11].	Observation and detection deviations; Model and intervention method deviations; Causal inference deviations; Technical and system deviations; Quantitative analysis deviations [11].	Treat animals/donors as independent units. For nested data, apply a mixed‑effects model with random intercepts for each animal and each batch/slice. Example: response ~ condition + factor + (1|animal) + (1|slice). Report the numbers of biological and technical replicates separately [265].	The CA1 area of the hippocampus and the dorsal lateral nucleus of the optic system (dLGN) [265].	Open-source software Ilastik (an interactive image analysis based on machine learning) and CellProfiler are used for image processing and object segmentation [265].	Biological replicates (individual animals) serve as independent statistical units, whereas single cells or images are nested within them. Mixed-effects models or pseudobulk approaches should be used to account for this nested structure and avoid pseudoreplication. Technical replicates (e.g., multiple images from the same animal) must not be treated as independent samples [265].	Blindness analysis [265].
Synaptic contact event	Contact area ratio and frequency; Contact duration; Number of new contacts formed per unit time [11, 266].	Define and detect bias; causal inference bias; temporal and dynamic bias [11, 266].	Clearly, define high-resolution imaging; combine dynamic in vivo imaging with functional recording; and must link the subsequent fate and functional outcome [11, 266].	The radiating layer of the CA1 layer of the hippocampus [20].	Manual segmentation combined with IMOD/Amira software was used to conduct three-dimensional reconstruction of microglia and synaptic structures [20].	Same as above: biological replicates (animals/donors) are independent units; cells/images nested within them. Mixed-effects models or pseudobulk approaches required [20].	Follow a unified and predefined objective standard [20].
Image formation and segmentation threshold	Image quality parameters (such as signal-to-noise ratio and contrast), segmentation accuracy indicators (such as accuracy and F1 score), as well as specific operational thresholds set for cell-synapse interactions (such as colocalization area ≥ 20%) and adaptive threshold segmentation methods [27].	Biological disturbances caused by phototoxicity, false positive contact determinations resulting from subjective threshold setting, reduced data comparability due to inconsistent imaging parameters, and sampling deviations (such as only observing specific brain regions or synaptic types) [27].	While ensuring the physiological state of the living sample (such as minimizing phototoxicity), standardized imaging parameters are adopted, and combined with automated and repeatable segmentation algorithms (such as adaptive threshold method) for quantification. At the same time, multimodal verification (such as confocal microscopy, electron microscopy) is conducted to ensure the accuracy, consistency and biological relevance of the results [27].	By using multiphoton imaging to set the ROI, performing image denoising and alignment, and applying segmentation-related methods such as visual contrast adjustment and motion thresholding in structural annotation and motion tracking [267].	Multilevel threshold segmentation; Binaryization and seed point generation; Morphological reconstruction and quantification [268].	-----	-----

A key consideration in microglial research is distinguishing biological replicates from technical replicates. Biological replicates—such as individual animals or independent iPSC lines—are the appropriate units for statistical inference. Technical replicates—for example, multiple images from the same animal or repeated measurements from the same culture—reflect measurement precision but do not provide independent statistical information. Treating multiple cells or images from the same biological replicate as independent units constitutes pseudoreplication, which inflates degrees of freedom and increases the false‑positive rate [227]. To avoid pseudoreplication, we recommend the following: (1) treat animals/donors as independent units—when making statistical comparisons between experimental groups, use biological replicates as the unit of analysis; (2) apply mixed‑effects models—for nested data structures (e.g., multiple cells per animal, multiple slices per brain), linear mixed‑effects models with random intercepts appropriately account for within‑animal correlation; (3) use pseudo‑aggregation methods—when mixed‑effects models are not feasible, aggregating single‑cell measurements to the animal level (e.g., the mean per animal) provides a conservative but effective analytical strategy; (4) transparently report the replication structure—manuscripts should clearly distinguish biological from technical replicates and specify the unit of analysis for each statistical test.

Methodological pitfalls in published studies. To illustrate the importance of proper statistical independence, we examined five studies cited in this review that reflect common analytical practices: (1) The Rac1 knockout study (Socodato et al, 2023) used animals as independent units and applied mixed‑effects models to analyze multiple synaptic contacts per animal—methodologically sound. (2) The microglia-synapse protective signaling study (Lehrman et al, 2018) used pseudo‑bulk aggregation with animals as the statistical unit—also correct. (3) The microglia-aging and neuromodulation study (Qu & Li, 2020) used mixed‑effects models with animal random intercepts when I/O curves were analyzed—correct. (4) The microglia-synapse imaging study (Winter et al, 2020) did not clearly specify whether animals were treated as independent units or how nested structures were handled—this ambiguity highlights the need for transparent reporting. (5) The human microglia transcriptomics study (Gosselin et al, 2017) treated donors as independent units and used a fixed‑effects model (limma‑voom); a preferable approach would have been a linear mixed‑effects model with donors as random effects. These examples underscore that rigorous analysis requires a clear specification of the unit of analysis and proper handling of nested data structures. Failure to do so inflates type I error rates and undermines the validity of mechanistic conclusions.

In microglial research utilizing multi‑omics technologies (e.g., transcriptomics, proteomics, and metabolomics), strict adherence to reporting standards and principles of statistical inference is essential for ensuring the reliability of conclusions and the validity of biological interpretations. For single‑cell or spatial transcriptomics studies, reports should include quality control thresholds (number of genes per cell, mitochondrial/ribosomal percentages), methods for doublet detection and ambient RNA correction, batch correction algorithms, differential expression analysis (using pseudobulk or mixed‑effects models to nest single‑cell data within biological replicates and avoid pseudoreplication), and gene set enrichment methods, along with a clear description of the multiple comparison correction strategy (e.g., FDR) [228]. Proteomics studies should report the acquisition type (DDA/DIA), identification and quantification workflows (software, parameters, and databases), false discovery rate thresholds at each level, missing value handling strategies (distinguishing MNAR from MAR), and key quality control metrics (number of identifications, median CV, and data completeness) [229]. The chosen statistical model should match the experimental design; for nested or longitudinal data, mixed‑effects models or generalized estimating equations are recommended. Transparency in data processing is critical: the nature of missing data (random vs. non‑random) should be clarified, and situations where imputation is not appropriate should be justified. Batch effects need systematic correction. Reporting effect sizes along with their confidence intervals helps assess biological relevance. For pathway and gene set enrichment analyses, the testing method, correction scope, and direction of effect should be explicitly stated, rather than relying solely on significance thresholds. Mechanistic claims based on omics data should be interpreted as correlational evidence; causal inference requires validation through perturbation experiments or time series data. Multilayer analyses integrating transcriptomic, proteomic, and morphological data are encouraged.

In summary, as high-throughput and highly sensitive technologies become increasingly prevalent, strict adherence to current consensus guidelines and standardized operating procedures can effectively minimize technical errors and biases, substantially increase data reliability and interstudy comparability, and thereby provide a robust foundation for scientific inquiry.

Necessity of microglia in synaptic plasticity: controversy and boundary conditions

Microglia are widely recognized as key regulators of synaptic plasticity, yet several fundamental controversies remain regarding their necessity, mechanistic redundancy, and causal role in disease.

Pathway necessity and redundancy

P2Y1 receptors are critical for microglial motility and experience-dependent plasticity [11]. In contrast, the classical complement pathway (C1q-C3-CR3), while essential for retinogeniculate pruning, is dispensable for ocular dominance plasticity in the visual cortex [230]. These differences indicate that pruning mechanisms are context-dependent and may operate through parallel, redundant pathways.

Duality of function: physiological pruning versus pathological over-pruning

The same molecular mechanisms (e.g., complement and CX3CR1) can be protective during development but aberrantly activated in disease, leading to excessive synapse loss [231]. What distinguishes adaptive pruning from pathological destruction remains unclear.

Cause or consequence?

In some neuropsychiatric conditions, microglial dysfunction may be a reactive response to primary neuronal deficits rather than the initial trigger (see Sections 4.1.3 and 4.3.2) [40].

Heterogeneity as a confounder

Microglial heterogeneity across age, region, and context (see Section Alzheimer’s disease) complicates cross-study comparisons and challenges the notion of a single “microglial function” in synaptic plasticity [232].

Resolving these controversies will require standardized experimental paradigms, cell-type-specific manipulations and functional readouts that directly link microglial activity to synaptic and behavioral outcomes.

Conclusions

Human genetics and single-cell transcriptomics have confirmed that NDD risk genes are enriched in microglia, supporting a key role for immune signaling in disease [39]. However, multiscale discrepancies between microscale synaptic changes and macroscale network behavior remain a major unresolved challenge [233].

Microglia play a dual role in NDDs: protective synaptic pruning under physiological conditions versus pathological overactivation that drives excessive synapse loss and cognitive decline. This duality calls for therapeutic strategies that precisely modulate microglial functional states, judged by functional recovery rather than molecular markers alone. Nevertheless, current research is limited by microglial spatiotemporal heterogeneity, translational limitations of animal models, and technical hurdles in capturing dynamic synaptic events (see Section Necessity of microglia in synaptic plasticity: Controversy and boundary conditions) [17, 54].

Future efforts should prioritize four areas. First, multi‑omics data must be integrated with real‑time imaging to link microglial molecular states to actual synaptic outcomes. Second, tools that simultaneously monitor morphology, signaling, and synapse fate in vivo are needed to decipher causal morphology-function relationships. Third, the functional relevance of disease‑associated microglia (DAMs) and region‑restricted subsets requires validation in human tissue. Fourth, iPSC‑derived microglia and humanized mouse models should be benchmarked against primary human cells to validate causal mechanisms for risk genes such as TREM2 and C4. The goal should be to reprogram microglial function rather than ablate it, guided by early biomarkers of synaptic recovery.

Abbreviations

Ac

Acetylation

Acetyl-CoA

Acetyl-Coenzyme A

AKT

AKT serine/threonine kinase

AMPAR

α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor

APOE

apolipoprotein E

BAK

BCL-2 Antagonist/Killer

BAX

BCL-2-Associated X Protein

BBB

Blood-Brain Barrier

BCL-2

B-cell lymphoma 2

BCL-xL

B-cell lymphoma-extra large

BID

BH3-interacting domain death agonist

CDC42

Cell division cycle protein 42

CDCA

Chenodeoxycholic Acid

c-fos

FBJ Osteosarcoma Sarcoma virus oncogene

cIAPs

Cellular Inhibitor of Apoptosis Proteins

DISC

death-inducing signaling complex

ERK

Extracellular signal-regulated kinase

FADD

Fas-Associated protein with death domain

FasL

Fas ligand

FASR

Fas Receptor

FXR

Farnesoid X Receptor

GDP

Guanosine diphosphate

GEF

GTPase-activating protein

Gluts

Glucose Transporters

GPR56

G protein-coupled receptor 56

GPR120

G-Protein Coupled Receptor 120

gp130

Glycoprotein 130

GTP

Guanosine triphosphate

HIF-1α

Hypoxia-Inducible Factor 1-alpha

HK

Hexokinase

IFN-γ

Interferon-gamma

IKKα

Inhibitor of Nuclear Factor Kappa-B Kinase subunit Alpha

IKKβ

Inhibitor of Nuclear Factor kappa-B kinase subunit beta

IKKγ

Inhibitor of Nuclear Factor Kappa-B Kinase subunit Gamma

IMOD

Image Modeling and Display

ISR

Integrated Stress Response

IκBα

Inhibitor of Kappa B Alpha

Lac

Lactate

LI

Latent Inhibition

LIF

Leukemia Inhibitory Factor

MAC

Membrane Attack Complex

MCL-1

Myeloid cell leukemia 1

MEK

MAPK/ERK kinase

MEGF10

Multiple EGF-like domains 10

MERTK

MER tyrosine kinase

MFG-E8

Milk Fat Globule-EGF Factor 8

MHC I

Major histocompatibility complex class I molecules

MYD88

Myeloid Differentiation Primary Response 88

Me

Methylation

NF-κB

Nuclear factor kappa-light-chain-enhancer of activated B cells

NMDAR

N-methyl-D-aspartate receptor

NOX

NADPH oxidase

NPTX2

Neuronal Pentraxin 2

p-AKT

Phosphorylated AKT

PFKFB3

6-Phosphofructo-2-Kinase/Fructose-2,6-Bisphosphatase 3

Pi

Phosphate molecule

p-IκB

Phosphorylated IκB

p-mTOR

Phosphorylated mTOR

PPI

Prepulse Inhibition

p-STAT

Phosphorylated signal transducer and activator of transcription

Pyr

Pyruvate

Raf

Rapidly accelerated fibrosarcoma

Ras

Rat sarcoma virus

RIP

Receptor-Interacting Protein

ROS

Reactive oxygen species

SFAs

saturated fatty acids

SRPX2

Sushi Repeat Containing Protein, X-Linked 2

STAT

signal transducer and activator of transcription

TAK1

Transforming Growth Factor-β-Activated Kinase 1

tBID

Truncated BID

TGR5

Takeda G-protein-coupled receptor 5

TNFR

Tumor necrosis factor receptor

tPA

Tissue-type plasminogen activator

TRADD

TNFR1-associated death domain protein

TRAF2

TNF Receptor-Associated Factor 2

Trk

Tropomyosin receptor kinase B

TUDCA

Tauroursodeoxycholic Acid

UDCA

Ursodeoxycholic Acid

α-KG

Alpha-ketoglutarate

αnβ3 receptor

Integrin alpha-V beta-3

β-Arrestin2

Beta-arrestin

Author contributions

PX, HH, and ZX jointly undertook the conception and structural design of this review. ZL, AL, WA, SG, and JW carried out the systematic literature search, study selection, and data extraction. ZL prepared the initial draft under the supervision of PX, HH, and ZX. CY conducted a critical review of the manuscript. All the authors contributed to the subsequent revisions and participated in the discussion and confirmation of the key academic content of the final version. PX, HH, and ZX supervised the entire project. All the authors read and approved the final manuscript.

Funding

This research was funded by the Guizhou Basic and Clinical Research Scientific and Technological Innovation Talent Team Project (No. CXTD[2022]013); funding from the Guizhou Provincial “Hundred” Level Innovative Talents Fund (No: GCC-2022-038-1); and platform support from the Key Laboratory of Brain Function and Brain Disease Prevention and Treatment of Guizhou Province (Qiankehe Foundation - ZSYS(2025)030).

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

N/A.

Consent for publication

N/A.

Competing interests

The authors declare no competing interests.