The concept of reserve was proposed to explain individual differences in susceptibility to age-related brain changes and pathological alterations, such as those occurring in AD. Reserve can act as a regulator between pathology and clinical outcomes, thereby accounting for this discontinuity (Stern, 2012). BM, on the other hand, is a determinant of cognitive preservation in the elderly and plays a role at different stages of the life cycle. Therefore, understanding the framework of cognitive reserve, brain maintenance, and brain reserve can help us analyze and intervene in those factors that are crucial for maintaining brain health.

Conceptually, cognitive reserve defines an individual property of the brain to maintain cognitive abilities above those expected given age or disease-induced damage. Fundamentally, cognitive reserve is the outcome of the history of each brain’s interaction with environmental factors throughout life, often referred to as the exposome, leading to either beneficial (lifelong learning, plasticity) or harmful (accumulation of pathological changes) alterations. These interactions also depend on the genes encoding an individual’s brain. These modifications are both neural tissue-based and systemic, with the closest interrelationship existing between the two.

The cognitive reserve model postulates the flexibility and adaptability of cognitive/brain networks, enabling the brain to actively resist the effects of age- or disease-related changes (Stern, 2002). Focusing on function, plasticity, and adaptability, cognitive reserve can be regarded as the “software” for computations in the brain, which is influenced by all aspects of life experiences (Stern, 2009) (Table 1).

Evidence supporting cognitive reserve includes epidemiological data on lifestyle and clinical outcomes. Individuals with higher pre-morbid IQ, educational or occupational achievements, or those who engage in late-life leisure activities have a reduced risk of dementia and may experience a slower rate of age-related cognitive decline. A meta-analysis involving 29,000 people showed that compared with those with low cognitive reserve, individuals with high cognitive reserve had a 46% lower risk of developing dementia (Valenzuela and Sachdev, 2006).

Moreover, the theory of cognitive reserve acknowledges individual differences, which can enable some individuals to cope better with brain pathological changes than others. Thus, even in the presence of neuropathology, cognitive decline can be delayed (Valenzuela and Sachdev, 2006). Consequently, individuals with higher cognitive reserve are better able to handle brain changes and maintain higher levels of function despite brain deterioration (Stern, 2012; Stern et al., 2019).

BR can be used to reflect the neurobiological state of the brain (the number of neurons, synapses, etc.) at any point in time. Unlike CR, BR does not involve the active adaptation of functional cognitive processes in the presence of damage or disease.

The brain reserve model defines reserve as a physical characteristic: some people have larger brains with more neurons and synapses, which may enable their brains to absorb more damage before cognitive function is affected. Therefore, brain reserve can be regarded as an individual difference in the brain’s “hardware” (Satz, 1993; Stern, 2012; Stern et al., 2019). Brain reserve can be considered a passive entity. It only manifests obvious signs of damage when an individual’s capabilities fall below a critical threshold of brain substrate loss (Stern et al., 2019). Those who initially have a higher BR can tolerate more depletion before symptoms appear (Stern et al., 2023) (Table 1).

The BM refers to the relative stability of neural resources or the slow progression of neuropathological changes over time, which is a determining factor for cognitive preservation in the elderly. BM is influenced by various factors such as genes, gender, early-life experiences, or differential exposures. These factors can slow down or prevent brain changes associated with aging and diseases. The focus is on the changes occurring over time. Thus, BM can potentially serve as an indicator of minimal changes in brain markers related to aging or diseases while maintaining cognitive function (Stern et al., 2023).

Related studies suggest that BM is not only influenced by genetic factors. It can also be regarded as an active protective mechanism, representing the ability to proactively maintain the integrity of brain structure and function and delay the accumulation of age-related damage. Various cell types in the brain, such as astrocytes, actively and dynamically regulate the ionic composition of the interstitial fluid in the central nervous system through a series of specialized pumps and transporters to maintain ionic homeostasis (Li et al., 2024; Verkhratsky and Zorec, 2024) (Table 1).

The CR is mainly manifested by a more prominent expression of the task-active network, specifically demonstrated by the inactivation of the DMN and the activation of the inferior temporal regions including the fusiform gyrus (van Loenhoud et al., 2020; Stern et al., 2021). The DMN typically consists of a set of core brain regions and a set of supplementary brain regions. Its core brain regions include the medial prefrontal cortex (mPFC), the posterior cingulate cortex or precuneus, and the bilateral inferior parietal cortices; the supplementary brain regions include the medial temporal lobe and the temporal pole, etc. (Mars et al., 2012; Stern et al., 2018; Vockert et al., 2024) (Figure 1). Among them, the key sub-regions of the DMN, such as the left precuneus, the left posterior cingulate gyrus, the precuneus-cingulate gyrus conjunction area, and the medial frontal gyrus, have been proven to be closely related to CR (Cole et al., 2012; Franzmeier et al., 2017).

Related studies have found that the key nodes of the DMN exchange information through white matter fiber bundles (such as the cingulum bundle and the superior longitudinal fasciculus), thus forming an efficient network with small-world characteristics (Cole et al., 2012; Franzmeier et al., 2017). The extensively distributed nervous systems in the human brain support complex and flexible cognitive processes. These nervous systems are connected by white matter fiber bundles, which serve as supporting structures to enable information transmission and network integration between brain regions. Dynamic functional connectivity analysis shows that there is a hierarchical information processing pattern within the DMN: the anterior nodes are mainly responsible for leading self-related thinking, the posterior system focuses on scenario simulation, and the angular gyrus acts as a bridge for cross-modal information integration (Stern et al., 2005; Vockert et al., 2024).

In addition, the neural activity in the left prefrontal cortex (regardless of whether it belongs to the functional scope of the fronto-parietal network) and the global connectivity of the left frontal cortex are both significantly associated with CR. It is worth noting that existing evidence also suggests that the anterior cingulate cortex (ACC) may be involved in the regulation of the neural mechanism of cognitive reserve (Stern et al., 2005; Vockert et al., 2024).

Kahali et al. (2021) used quantitative Gradient-Recalled Echo (qGRE) MRI to map the cellular composition of the human brain and employed blood–oxygen level-dependent (BOLD) MRI to delve into the intrinsic relationship between brain cell components and the DMN.

The research findings indicated that within a single functionally defined unit of the network, the synchronization of connections between cell circuits defined by the BOLD signal mainly depends on the regional neuronal density. Moreover, the connection strength between functional units is also influenced by the glial and synaptic components in brain tissue cells. These mechanisms contribute to the widespread characteristics of the DMN.

Specifically, the visual network exhibits unique features. It has the highest neuronal density (while the glial cell and synaptic density are at the lowest levels), demonstrating the strongest BOLD signal consistency and the tightest internal network connections. In contrast, the DMN lies at the other end of the spectrum. Although it has relatively lower BOLD signal consistency, its cellular contents show a remarkable balance. This balance enables the DMN to play a highly prominent role in the overall organizational structure of the brain and the hierarchical functional network.

In conclusion, the association between neurons and glial cells plays a crucial role in the strength of neural network connections. In the network, the synchronous connections between cell circuits belonging to different functional units are not only closely related to the association between neurons in different functional units but also involve the interactions between neurons and synapses, as well as between neurons and glial cells.

Brain reserve, initially defined as neural or functional reserve (Katzman, 1988), is regarded from an anatomical perspective as a resource determined by the brain size and/or structural characteristics when damaged. Initially, brain reserve was associated with brain volume. Interestingly, people with larger brains have a lower probability of developing dementia (Katzman et al., 1988; Schofield et al., 1997) or the quantity of neurons and synapses. The larger these values and the total brain volume are, the more capable an individual’s brain is of withstanding damage and demonstrating a higher reserve capacity (Katzman, 1988; Satz et al., 1993). Recent studies have shown that Stern et al. (2023) defined the Brain Reserve (BR) as a concept determined by a large number of neurons and the inter – neuronal connections formed by axonal projections (connectome) and synapses. It serves to reflect the neurobiological state of the brain at any given timepoint (the quantity of neurons, synapses, etc.). All of these are regulated by glial cells. Embryonic neurogenesis is a function of radial glial cells, while adult neurogenesis is carried out by radial glial stem cells, also known as neural stem cells, which exhibit all the major characteristics of astrocytes (Kriegstein and Alvarez-Buylla, 2009; Götz et al., 2016; Yeh et al., 2023) (Table 2).

The maintenance mechanisms of the brain keep all substances in a stable state close to equilibrium. This state was defined by Claude Bernard as the stability of the internal environment (Duchesneau, 2024). The homeostasis of the central nervous system (as well as all other tissues, organs, and the entire organism) is not static but is achieved through continuous adjustments in response to environmental challenges, which is referred to as allostasis (Verkhratsky and Zorec, 2024).

Microglia are resident immune cells in the brain that maintain homeostasis. In a healthy brain, they perform homeostatic functions, which are characterized by the expression of transforming growth factor-β (TGF-β), transmembrane protein 119 (TMEM119), and P2RY12 (Butovsky et al., 2014; Krasemann et al., 2017). However, in response to brain injury or neurodegenerative diseases, reactive microglia lose their homeostatic molecular characteristics and functions (Paolicelli et al., 2022; Wang et al., 2024). Previous studies have shown that microglia play a role in physically monitoring the microenvironment through the interaction between ligands and receptors (Favuzzi et al., 2021; Huo et al., 2024). Research indicates that microglia express a variety of neurotransmitter receptors, including metabotropic and ionotropic glutamate receptors, gamma – aminobutyric acid (GABA) receptors, dopamine receptors, norepinephrine receptors, cannabinoid receptors, acetylcholine receptors, as well as some neuropeptide receptors and neuromodulator receptors such as adrenergic receptors and purinergic receptors (Liu et al., 2016; Huo et al., 2024) (Table 2).

Among them, glutamate, an important neurotransmitter in the brain, can be released not only from synaptic vesicles but also from glial cells. Excessive release of glutamate can stimulate the glutamatergic receptors on microglia, leading to the activation of microglia, the production of reactive oxygen species, and the release of ATP and cytokines. Consequently, this has corresponding effects on the reduction of dendritic spines and synaptic loss (Huo et al., 2024).

Astrocytes are the major homeostatic cells in the central nervous system and play a dominant role in brain maintenance. Astrocytes dynamically regulate the ionic composition of the interstitial fluid in the central nervous system through a series of specialized pumps and transporters, which is known as ionic homeostasis (Untiet et al., 2023; Rose and Verkhratsky, 2024). They control all major biological ions (such as sodium, calcium, potassium, and chloride) as well as trace elements that are crucial for the function of the central nervous system (MacAulay, 2020; Li et al., 2024; Verkhratsky and Zorec, 2024).

Astrocytes regulate ion concentrations by recycling neurotransmitters. High-affinity glutamate transporters uptake the neurotransmitter glutamate from the synaptic cleft for recycling (Anderson and Swanson, 2000; Sun et al., 2018). Astrocytes undertake the major task of glutamate transport. Glutamate is transported into the cytoplasm of astrocytes by binding to glutamate transporters, which are co-transported with Na⁺ into astrocytes. Na⁺ is then transported out of the cell by the Na⁺-K⁺-ATPase. Glutamate reacts with glutamine synthetase (GS) to form glutamine. The ATP consumed in this process is presumably supplied by the ATP generated through glycolysis. Notably, GS is only expressed in astrocytes (Parpura et al., 2017; Ma et al., 2022). Therefore, the glutamate-glutamine cycle plays a key role in maintaining the glutamate level in the central nervous system.

The generated glutamine is released into neurons. Then, glutamine enters neurons through the SLC1A5 receptor. The glutamine absorbed by neurons is converted into glutamate under the action of glutaminase (GLS), and it can also produce GABA through the action of glutamate decarboxylase (Ma et al., 2022).

Another important role of glutamate in astrocytes is the synthesis of glutathione (GSH), which is crucial in the cell’s antioxidant defense and detoxification processes, as shown in Figure 3. Under high-glucose conditions, the amino acid metabolism of astrocytes is disrupted, the uptake of glutamate and the activity of GS are altered, and the level of GSH is reduced, which may induce cognitive dysfunction (Ma et al., 2022).

Therefore, as a crucial component of neurotransmission, astrocytes play multiple roles. They clear neurotransmitters (astrocytes express transporters to remove glutamate, γ-aminobutyric acid, catecholamines, and adenosine), carry out neurotransmitter catabolism (astrocytes convert glutamate to glutamine and degrade catecholamines and adenosine), and supply essential neurotransmitter precursors such as glutamine or L-serine to neurons (Verkhratsky et al., 2016; Verkhratsky and Zorec, 2024). Astrocytes store glycogen, and the intracellular breakdown of glycogen can produce L-lactate. L-lactate can provide energy support for neurons and is an important fuel source for them (Yellen, 2018; Fink et al., 2021; Rae et al., 2024). Similarly, oligodendrocytes support axons (Simons and Nave, 2016), while oligodendrocyte precursor cells (OPCs) provide lifelong myelin regeneration and activity-dependent formation of new myelin sheaths, also contributing to the overall homeostasis of the central nervous system (Yi et al., 2023). Finally, microglia are involved in debris clearance, continuous repair, and maintaining the immune homeostasis of neural tissue (Stratoulias et al., 2019).

In AD, the brain regions most frequently involved in alterations of functional connectivity are those within the DMN, which are considered to be selectively vulnerable to AD neuropathology (Warren et al., 2013). From a neurobiological perspective, poor DMN connectivity is thought to be the result of activity-dependent tau propagation, from the medial temporal lobe to densely connected cortical hubs (Sperling et al., 2009; Huijbers et al., 2019; Maass et al., 2019; Vockert et al., 2024). Vockert et al. (2024) found that changes in DMN effective connectivity mediated the association between social isolation and the incidence of dementia. When examining the connectivity differences between cases and controls, the researchers found an increased inhibitory tone from the prefrontal and parietal cortices to the medial temporal lobe, as well as a decreased inhibitory tone from the medial temporal lobe to the prefrontal cortex. The increased inhibitory tone may potentially reflect homeostatic compensation to maintain the excitatory-inhibitory balance within the network. Due to the neuronal silencing effect of tau accumulation in the medial temporal lobe in the early stages of the disease, the inhibitory connections from the parahippocampal formation to the prefrontal cortex are weakened.

Therefore, an important direction for future research is to collect longitudinal imaging data as well as data on subjective and cognitive impairments, in order to understand the clinical significance of different connections in the connection patterns of the brain’s DMN. DMN dysconnectivity may be better understood in the context of a broader dynamics involving other long-range networks. In high-risk populations for AD, both the salience network and the frontal parietal lobe control network are altered, and these changes are associated with future cognitive decline. Incorporating these networks into the development of future effective connectivity-based prediction models is of great significance.

The BR reflects the brain’s capacity to withstand risks in terms of its volume and structure. A larger brain volume indicates a greater number of neurons and synaptic connections, as well as a stronger ability to resist cognitive decline. The interaction between neurons and glial cells promotes the formation and remodeling of synapses. In the absence of glial cells, neurons form weaker synapses (Pannasch et al., 2011; Kahali et al., 2021). The close relationship between neurons and glial cells plays a crucial role in establishing strong network connections. This is consistent with the role of glial cells in supporting neurons by providing metabolic and regulatory functions (Poskanzer and Yuste, 2011; Suzuki et al., 2011; Magistretti and Allaman, 2015; Kahali et al., 2021), as well as in supporting neurite growth and neuronal guidance (Ullian et al., 2001; Pannasch et al., 2011). As people age, astrocytes and microglia also show significant atrophy in the elderly brain, which leads to a decline in homeostatic support and defensive capabilities (Streit et al., 2004; Popov et al., 2023; Verkhratsky and Zorec, 2024). Aging is also associated with tau astrogliopathy, a disease that serves as the basis for various neurodegenerative pathologies related to cognitive impairment (Schultz et al., 2000; Kovacs et al., 2016; Verkhratsky and Zorec, 2024).

Astrocytes take up the neurotransmitter glutamate from the synaptic cleft using high – affinity glutamate transporters for recycling (Duan et al., 1999; Anderson and Swanson, 2000; Lee et al., 2022). When the glutamate transport function of astrocytes is impaired, neurons will over – release glutamate, leading to excitotoxicity and subsequent neuronal death (Schousboe et al., 1997) (Figure 3). This process has long been regarded as a key part of the pathogenesis of neurodegenerative diseases (Doble, 1999; Pitt et al., 2000; Maragakis and Rothstein, 2004; Dong et al., 2009; Lee et al., 2022).

In addition, an increase in the intracellular Ca²⁺ level of astrocytes can stimulate the release of glutamate. The accumulation of Aβ in the brain inhibits glutamate transporters, affecting the uptake of glutamate by astrocytes. This causes excessive activation of glutamate receptors expressed on neurons, leading to an increase in intracellular calcium ion concentration, which affects synaptic plasticity and ultimately results in neurological dysfunction (Lee et al., 2022).

As the main target cells of norepinephrine, astrocytes have been found to exhibit abnormal Ca²⁺ signaling in mouse models of AD due to reduced norepinephrine levels (Åbjørsbråten et al., 2022). Using reversible or irreversible MAO-B inhibitors may be an effective strategy to enhance CR (Park et al., 2019). tDCS can improve memory, promote motor function recovery, relieve depression, and delay the progression of cognitive impairment in AD patients (Kuo et al., 2014). In addition, tDCS presents a promising avenue for activating norepinephrine – releasing neurons. Alternatively, the transplantation of norepinephrine – releasing neurons also stands out as a feasible strategy (Kuo et al., 2014).

Its mechanism of action is achieved through astrocytes and their adrenergic receptors. Exposure to tDCS triggers a large number of Ca²⁺ signals in astrocytes, and these signals can be blocked by ablation of noradrenergic neurons or pharmacological inhibition of α1-adrenergic receptors (Monai et al., 2016; Monai and Hirase, 2018).

In the amyloidosis model of APP/PS1 mice (6–8 months old), Adult oligodendrocyte precursor cells (aOPCs), near Aβ plaques exhibit increased proliferation and differentiation (Behrendt et al., 2012; Yi et al., 2023). In both AD patients and APP/PS1 mice, the immunoreactivity of aOPCs and NG2 near amyloid plaques is enhanced (Nielsen et al., 2013; Sun et al., 2018; Yi et al., 2023). In the postmortem human AD cortex, the total number of oligodendrocytes is reduced (Behrendt et al., 2012; Yi et al., 2023). In both AD patients and mouse models, the total amount of myelin decreases over time. It can be speculated that compensatory myelin regeneration may occur during the progression of AD. In this context, pharmacologically enhancing aOPC differentiation and myelin regeneration may emerge as a potential strategy to alleviate the cognitive impairment caused by AD. Relevant studies have shown that among various neurotransmitter receptors expressed by microglia, the activation of type III subtypes of metabotropic glutamate receptors (mGluR) has a neuroprotective effect, and their stimulation helps inhibit neuronal death. In contrast, type II subtypes promote the activation of neurotoxic microglia. Therefore, interventions targeting microglial receptors may provide new therapeutic strategies for delaying the progression of AD (Ghimire et al., 2025).

Both astrocytes and oligodendrocytes (including those in fetal and adult stages) are vulnerable to oxidative stress, as well as glutamate and ATP excitotoxicity (Káradóttir et al., 2005; Matute et al., 2006; Giacci et al., 2018; Yi et al., 2023). Relevant studies have shown that the dysfunction of astrocytic glutamate transporters can lead to excessive glutamate release from neurons (Schousboe et al., 1997). This promotes neuronal death through excitotoxicity, a process that has long been associated with the occurrence of neurodegenerative diseases (Pitt et al., 2000; Maragakis and Rothstein, 2004; Dong et al., 2009; Soni et al., 2014). The research team led by Ma et al. (2022) found that the expression of glutamate transporters in the cerebral cortex of AD patients is decreased, and the decrease in vesicular glutamate transporter 1 (VGLUT1) is more pronounced. This may result in the failure to efficiently clear glutamate in the synaptic cleft, thereby triggering excitotoxicity. In fact, the pro-inflammatory cytokine tumor necrosis factor (TNF) increases glutamate production, thus triggering excitotoxicity and Ca²⁺ signaling. An increase in Ca²⁺ levels in glial cells leads to the release of glutamate, ATP, and gamma-aminobutyric acid (GABA). Additionally, the release of GABA and ATP by reactive astrocytes has been shown to be associated with the disease in Alzheimer’s disease models (Santello and Volterra, 2012). Dysfunctional glial-neuronal metabolic interactions are characteristic features of various neurodegenerative diseases (Sun et al., 2018).

In addition, glutamate accumulates in the synaptic clefts of the cerebral cortex in AD patients. In the presence of excessive glutamate and/or ATP, aOPCs rapidly die, which can lead to white matter damage and neurological symptoms (Deng et al., 2004; Salter and Fern, 2005; Matute, 2011). The death of aOPCs has also been observed in brain injuries associated with inflammation and neurodegeneration (Yi et al., 2023). In the human Alzheimer’s disease brain, pexin-B3-positive aOPCs are associated with extracellular Aβ_1–42 deposits, suggesting that pexin-B3-positive aOPCs may be the cells that produce Aβ (Nihonmatsu-Kikuchi et al., 2021; Yi et al., 2023). Single-cell RNA sequencing shows that genes encoding GABA receptors (GABRA2 and GABRB1), glutamate receptors (GRIA2 and GRID2), and neurexins (NRXN1 and NRXN3) in aOPCs are downregulated. These genes are all related to behavior, cognition, and synaptic organization (Grubman et al., 2019). In aOPCs from postmortem Alzheimer’s disease brains, the presence of the apolipoprotein ε4 allele (APOE4, a genetic risk factor for Alzheimer’s disease) is associated with impaired pro-inflammatory and immune-related pathways, altered DNA damage responses, abnormal lipid storage, and decreased myelination ability (Yi et al., 2023; Blanchard et al., 2024). Selective deletion of APOE4 in astrocytes can rescue tau-mediated neurodegeneration (Blanchard et al., 2021; Tcw et al., 2022) (Figure 2). APOE4 also contributes to reducing synaptic plasticity and causing blood–brain barrier dysfunction (Bovolental et al., 1997; Tcw et al., 2022) (Table 2).

The latest research has discovered a disease-associated microglia (DAM) signature, namely a significant reduction of microglial homeostatic markers in DAM. In a mouse model of Alzheimer’s disease (AD), TMEM119, a microglia – specific marker, is significantly reduced during amyloid plaque formation and is closely related to the spatial distribution of microglia near Aβ plaques (Liu et al., 2025). Therefore, the underlying mechanisms driving the changes in microglial TMEM119 and its significance in the pathogenesis of AD need further investigation.