Immunometabolism Reframes Alzheimer’s Disease: From Systemic Dysmetabolism to Glial Rewiring

Abstract

Alzheimer’s disease (AD) is increasingly recognized as a disorder of dysregulated immunometabolism at the neurovascular–glia–neuron interface. Systemic metabolic stressors such as insulin resistance, dyslipidaemia, and obesity converge on brain innate immune cells to reprogram energy pathways and sustain maladaptive inflammation. In microglia, metabolic rewiring across glycolysis–oxidative phosphorylation balance, glutaminolysis, and lipid handling governs trained-immunity programs that dictate amyloid and tau clearance, synaptic maintenance, and neurotoxicity. These processes converge on druggable nodes including AMPK–mTOR signaling, HIF-1α, and tricarboxylic-acid intermediates. Neurovascular fuel delivery is likewise impaired: endothelial GLUT1 loss and mitochondrial stress at the blood–brain barrier accelerate amyloid accumulation and neuronal injury. Lipid metabolism bridges metabolism and inflammation, as APOE4-driven microglial lipid droplets link genetic risk to inflammatory phenotypes. NLRP3 integrates metabolic danger signals into chronic neuroinflammation. Translational momentum now builds around metabolic interventions particularly GLP-1 receptor agonists and SGLT2 inhibitors that modulate glial metabolism, systemic inflammation, and barrier integrity. Converging metabolomic, lipidomic, and extracellular-vesicle biomarkers enable tracking of these pathways in humans, defining an immunometabolic axis of AD and supporting precision-medicine strategies to reprogram metabolism for disease modification.

Graphical Abstract

Systemic metabolic stress primes neurovascular and glial metabolism, driving pro-inflammatory microglia that can be therapeutically rewired toward homeostatic, neuroprotective states.

Introduction

Despite unprecedented investment in anti-amyloid therapeutics, Alzheimer’s disease (AD) remains only partially tractable. Phase 3 monoclonal antibodies targeting Aβ lecanemab and donanemab demonstrate modest clinical slowing (~ 25–35% over 18 months), validating amyloid as a therapeutic entry point but confirming that amyloid clearance alone does not halt neurodegeneration (van Dyck et al. 2023; Sims et al. 2023). The persistence of cognitive decline despite biomarker normalization highlights deeper disease layers involving neuroinflammation, metabolic stress, and vascular dysfunction.

Mounting genetic, epidemiological, and mechanistic evidence now converges on metabolism-immunity crosstalk as a unifying explanation for this therapeutic gap. Large-scale cohort analyses and Mendelian-randomization studies consistently link type 2 diabetes, dyslipidaemia, obesity, and mid-life hypertension to elevated dementia risk (Biessels and Despa 2018; Proitsi et al. 2014; Cho et al. 2022). In aggregate, it has been estimated that metabolic and vascular risk factors account for up to 40% of global dementia burden (Livingston et al. 2020). These data implicate peripheral metabolic dysfunction not merely as a comorbidity but as a causal driver of AD pathogenesis.

At the mechanistic level, systemic metabolic stressors re-programme brain innate immunity. Microglia, the resident macrophages of the central nervous system (CNS), undergo metabolic switching between glycolysis and oxidative phosphorylation that determines inflammatory tone. Activation of mTOR and inhibition of AMPK promote glycolytic, pro-inflammatory states; reciprocal activation of AMPK enhances oxidative metabolism and resolution phenotypes (Saxton and Sabatini 2017). Tricarboxylic-acid intermediates such as succinate, fumarate, and itaconate further encode metabolic “danger signals” that licensing NLRP3 inflammasome activation and sustaining chronic cytokine release (Masters et al. 2010; Tannahill et al. 2013). This immunometabolic licensing underlies trained-immunity-like behaviour observed in AD microglia, providing a mechanistic framework through which systemic energy imbalance may contribute to maladaptive neuroinflammation (Mills et al. 2018a).

Astrocytes likewise integrate systemic and local metabolic cues. Beyond their canonical roles in glutamate detoxification and lactate shuttling, astrocytes contribute to neuronal lipid and cholesterol supply (Valenza et al. 2015; Yuan et al. 2025). ApoE isoforms particularly ApoE4 alter cholesterol efflux and lipid-droplet dynamics, predisposing astrocytes and microglia to lipid accumulation and oxidative stress (Yamazaki et al. 2019b; Mallick et al. 2024). Recent single-cell transcriptomic and spatial-omics studies reveal ApoE4-driven astrocytic and microglial states characterized by lipid-storage pathology and inflammatory gene expression (Yamazaki et al. 2019b). These findings position lipid metabolism as a mechanistic bridge between genetic risk and innate-immune activation.

Metabolic dysregulation also compromises the neurovascular unit. Endothelial GLUT1 down-regulation at the blood-brain barrier (BBB) reduces glucose transport, producing focal energy failure that precedes neuronal loss in both mouse models and human PET imaging (Winkler et al. 2015; Benzinger et al. 2013). Vascular oxidative stress, pericyte dropout, and mitochondrial dysfunction further impair nutrient delivery and amplify microglial activation (Sweeney et al. 2018a). Collectively, these processes define an energy-vascular bottleneck that couples systemic dysmetabolism to local immune dysregulation. BBB accessibility represents a critical translational constraint for immunometabolic therapies. Several interventions discussed here may exert central effects indirectly, by improving systemic metabolism, endothelial function, and substrate delivery to the brain, even without robust penetration of an intact BBB. Others target brain-accessible nodes, either through small molecules with demonstrable CNS exposure or by exploiting disease-associated BBB permeability. Importantly, BBB integrity itself evolves across Alzheimer’s disease progression, suggesting that disease stage and neurovascular status may critically shape therapeutic responsiveness. Explicit consideration of BBB accessibility therefore complements immunometabolic target selection and reinforces the need for stage-aware translational strategies (Sweeney et al. 2018b; Montagne et al. 2020).

This multidimensional evidence has catalyzed a conceptual shift: AD should be viewed as a metabolic-immune convergence disorder rather than a purely proteinopathic neurodegeneration. Within this framework, cellular metabolism is not a by-product of inflammation but its upstream regulator and potential therapeutic lever. Indeed, several metabolic agents already in clinical use for diabetes GLP-1 receptor agonists, SGLT2 inhibitors, and metformin demonstrate anti-inflammatory, neuroprotective, and blood–brain-barrier-stabilizing properties independent of glycaemic control (Zheng et al. 2024; Lee et al. 2025a; DiBona et al. 2021). GLP-1 agonists are now being tested in phase 3 trials (EVOKE and EVOKE +) for early AD, signaling a translational pivot toward metabolic neurotherapeutics (Cummings et al. 2025b).

Parallel advances in biomarker technology further enable this reframing. Multi-omic platforms integrate plasma and cerebrospinal-fluid metabolomics, lipidomics, and extracellular-vesicle profiling with imaging readouts such as FDG-PET and emerging microglial tracers. These approaches permit endotype-based stratification that transcends the classical AT(N) scheme by incorporating metabolic and inflammatory dimensions (Netea et al. 2016; Sweeney et al. 2018a; Cummings et al. 2025b). Such stratification could enhance trial sensitivity and guide precision interventions targeting discrete immune-metabolic nodes.

In this Review, we synthesize recent mechanistic and translational progress to define an immune-metabolic axis of AD. We first outline how systemic dysmetabolism primes glial and neurovascular metabolism, then dissect the molecular switches (mTOR-AMPK, NLRP3, lipid signaling) that integrate energy sensing with inflammation. Finally, we outline diagnostic and therapeutic implications, highlighting how metabolic interventions could reshape the clinical landscape of AD.

This narrative review was conducted to synthesize emerging evidence linking systemic metabolic dysfunction with glial immunometabolic reprogramming in AD. Literature searches were performed using PubMed/MEDLINE, Web of Science, and Google Scholar to identify relevant peer-reviewed articles published primarily between January 2000 and April 2025, with an emphasis on studies from the past 10 years. Search terms included combinations of “Alzheimer’s disease,” “immunometabolism,” “microglia,” “astrocytes,” “systemic metabolism,” “insulin resistance,” “lipid metabolism,” “neuroinflammation,” and “blood–brain barrier.”

Studies were selected based on their relevance to immunometabolic mechanisms in AD, with priority given to human studies, human-anchored experimental models, and translational investigations that interrogated metabolic–immune interactions in glial cells or the neurovascular unit. Both observational and interventional studies were included where they provided mechanistic insight. Review articles were consulted to contextualize key concepts but were not used as primary evidence for mechanistic claims.

To minimize selection bias, titles and abstracts were screened for relevance, and full texts were evaluated to ensure conceptual and methodological alignment with the scope of immunometabolism rather than general neuroinflammation. Where available, findings were cross-referenced across independent studies and experimental systems. This review does not represent a formal systematic review or meta-analysis, and therefore some degree of selection bias is inherent; however, care was taken to integrate convergent evidence and to critically discuss discrepancies and limitations within the existing literature.

Reframing AD Through Immunometabolism

For three decades, AD has been framed largely through amyloid and tau. Anti-amyloid antibodies now demonstrate statistically significant but modest slowing of clinical decline in early AD, validating amyloid involvement yet underscoring that amyloid-centric therapy alone does not fully arrest disease progression (~ 27–35% slowing over 18 months; pivotal phase-3 data) (van Dyck et al. 2023; Sims et al. 2023). In parallel, the 2018 NIA-AA framework recast AD as a biological construct defined by AT(N) biomarkers rather than symptoms, a shift that clarifies pathology staging but also exposes gaps namely, how systemic physiology shapes brain inflammation and neurodegeneration across this continuum (Jack et al. 2018a, b; Jack et al. Jr 2018a, b).

We propose an immunometabolic reframing in which systemic dysmetabolism (e.g., insulin resistance, dyslipidaemia) conditions the central nervous system (CNS) by reprogramming innate immune and neurovascular metabolism, thereby sustaining maladaptive inflammation and synaptic failure. Epidemiologically, the 2024 Lancet Commission integrates new evidence that several metabolic and vascular risk factors (including high LDL cholesterol) contribute substantially to global dementia burden, strengthening the rationale to treat metabolism as causal terrain rather than background noise (Livingston et al. 2024). Mechanistically, convergent human and experimental data position innate immunity particularly microglia as a disease-promoting node when chronically activated, with metabolism acting as the throttle.

Two lines of evidence anchor this reframing. First, neurovascular fuel delivery measurably fails in AD: endothelial (Endothelial Glucose Transporter 1) GLUT1 reduction at the blood-brain barrier (BBB) impairs glucose transport and accelerates Aβ pathology and neurodegeneration in vivo, linking systemic/vascular metabolism to amyloid handling and neuronal injury (Winkler et al. 2015). Second, glial lipid metabolism couples genetic risk with inflammatory tone: in human brain and models, APOE4 biases microglia toward lipid-droplet-rich, neurotoxic states that promote tau-related toxicity directly tying lipid biology to innate immune dysfunction (Marschallinger et al. 2020; Haney et al. 2024a).

Within this axis, immunometabolic switches (AMPK–mTOR balance, glycolysis↔OXPHOS, TCA-derived signals) intersect with inflammasome pathways; notably, NLRP3 integrates metabolic danger cues and is repeatedly implicated as a driver of chronic neuroinflammation in AD, with a growing inhibitor toolbox (Xu et al. 2025; Coll et al. 2015). This mechanistic lattice broadens the therapeutic and biomarker landscape: it explains why amyloid lowering yields partial benefit, identifies metabolic-immune nodes as drug targets, and motivates endotype-based stratification that blends AT(N) with metabolic and inflammatory readouts. The result is a unifying framework immune-metabolic AD that connects systemic risk, glial and vascular metabolism, and clinical outcomes, and that is testable with existing human tools.

Systemic Dysmetabolism Primes the Central Nervous System

AD does not arise in isolation within the brain; it unfolds within a metabolic ecosystem that extends from the liver and adipose tissue to the cerebral microvasculature. A growing convergence of epidemiological and mechanistic data now indicates that systemic dysmetabolism comprising insulin resistance, dyslipidaemia, and obesity acts as an upstream modulator of brain vulnerability. Rather than being mere comorbidities, these metabolic states reprogram energy distribution, endothelial function, and immune tone, effectively “pre-inflaming” the brain decades before the onset of cognitive symptoms (Livingston et al. 2024; Gudala et al. 2013; Zhang et al. 2022a).

Large-scale syntheses, including the 2024 Lancet Commission on dementia prevention, quantify the impact of metabolic and vascular risk factors as among the most potent drivers of dementia incidence. Type 2 diabetes increases the risk of developing dementia by approximately 50%, and dyslipidaemia particularly high LDL cholesterol and altered triglyceride-to-HDL ratios further amplifies this effect (Livingston et al. 2024; Gudala et al. 2013; Cholerton et al. 2016). These observations are not simply statistical correlations: longitudinal cohorts and Mendelian randomization analyses increasingly support a causal chain linking peripheral energy imbalance to neuronal injury. The traditional narrative that “what happens in the body stays in the periphery” has been decisively overturned; systemic metabolic failure seeps through the BBB both physically and biochemically.

The BBB, once viewed as a passive gatekeeper, is now recognized as a dynamic metabolic sensor. GLUT1 is downregulated in AD, a defect that impairs cerebral glucose uptake and accelerates amyloid deposition in vivo (Sweeney et al. 2018b; Winkler et al. 2015). Reduced endothelial metabolism also limits lactate and ketone delivery to astrocytes and neurons, compounding energetic stress via monocarboxylate transporter-1 (MCT1) dependent fuel shuttling. This vascular bottleneck creates a chronic energy deficit that not only weakens neuronal resilience but also transforms glia into “energy sentinels” that interpret metabolic scarcity as danger. The result is a smoldering state of immune readiness that renders the CNS hypersensitive to inflammatory cues (Chasseigneaux et al. 2024; Vijay and Morris 2014).

Insulin resistance compounds this vulnerability. While systemic hyperinsulinaemia promotes peripheral inflammation, within the brain it blunts insulin receptor signaling on neurons and glia (Watson et al. 2022; van der Heide et al. 2006), reducing synaptic plasticity and altering microglial activation thresholds. The concept of “type 3 diabetes” remains controversial, yet the molecular echoes of impaired insulin signaling such as downregulated IRS-1 and Akt phosphorylation are consistently observed in post-mortem and PET-imaging studies. Importantly, these metabolic signatures emerge early, long before overt tauopathy, implying that insulin resistance is not a consequence but a conditioning state for neurodegeneration (Talbot et al. 2012; Kim and Feldman 2015; Ferreira et al. 2018).

The inflammatory conduit between periphery and brain involves both cytokines and lipids. Chronic adipose inflammation releases IL-6, TNF, and C-reactive protein, which can cross or signal across the BBB to prime resident immune cells. Experimental models reveal that peripheral endotoxaemia induces NLRP3-dependent BBB permeability and triggers microglial inflammasome activation in vivo (Yoon et al. 2025; Banks et al. 2015). This pathway may underlie the clinical observation that metabolic syndrome and chronic infection accelerate cognitive decline even in amyloid-negative individuals. The BBB, therefore, functions as a bidirectional relay translating metabolic distress into neuroinflammatory tone.

Lipids add another dimension to this priming. Circulating ceramides and sphingomyelin species elevated in metabolic syndrome and diabetes are not inert passengers; they can modulate membrane order and receptor clustering in microglia, enhancing responsiveness to secondary insults. Prospective human studies show that plasma ceramide signatures predict future dementia risk (Mielke et al. 2012b; McGrath et al. 2020). Mechanistically, ceramide accumulation activates TLR4 and NLRP3 signaling, reinforcing a feed-forward loop between lipid metabolism and innate immunity (Scheiblich et al. 2017; De Nardo and Latz 2011).

As shown in Fig. 1, systemic dysmetabolism including insulin resistance, dyslipidaemia, and obesity imposes metabolic stress on the blood-brain barrier, facilitating the passage of inflammatory and metabolic danger signals that reprogram resident glia toward a primed, pro-inflammatory state.

Fig. 1.

The system–central axis of Alzheimer’s Disease. Systemic metabolic dysregulation compromises the blood–brain barrier (BBB), allowing the influx of inflammatory and metabolic danger signals that prime resident immune and glial cells, driving chronic neuroinflammation. Systemic metabolic stressors such as obesity, type 2 diabetes, and dyslipidaemia induce endothelial GLUT1 downregulation and promote the translocation of inflammatory mediators (IL-6, TNF-α, CRP) and metabolic danger signals (ceramides, succinate) across the BBB, initiating microglial and astrocytic priming

Integrating these findings reveals a self-amplifying cycle. Systemic dysmetabolism diminishes brain fuel supply and disrupts BBB integrity, priming microglia toward pro-inflammatory phenotypes. Activated microglia and astrocytes, in turn, secrete cytokines and reactive oxygen species that further compromise endothelial function, establishing a chronic energy-inflammation loop (Sweeney et al. 2018b; Winkler et al. 2015; Xu et al. 2025). This bidirectional circuit offers a compelling explanation for the limited efficacy of amyloid-centric therapies: even when amyloid burden is reduced, the metabolic context continues to sustain a neuroinflammatory milieu that drives progression. A deeper implication is that AD may, in part, be an organ-to-brain communication disorder. The metabolic signatures of liver, adipose tissue, and gut microbiota reverberate through vascular and immunological channels, shaping neural outcomes. Targeting this systemic central dialogue through metabolic drugs, dietary modulation, or BBB-protective strategies could therefore transform therapeutic logic from damage control to upstream prevention.

Systemic metabolic stressors such as obesity, type 2 diabetes, and dyslipidaemia induce endothelial GLUT1 downregulation and promote the translocation of inflammatory mediators (IL-6, TNF-α, CRP) and metabolic danger signals (ceramides, succinate) across the BBB, initiating microglial and astrocytic priming.

Microglial Immune-Metabolic Switches

Microglia translate metabolic context into inflammatory decisions. Their basal surveillance state leans on oxidative phosphorylation (OXPHOS), whereas danger sensing rapidly pivots them toward aerobic glycolysis, a switch that prioritizes speed over efficiency and reprograms transcription through HIF-1α, NF-κB and epigenetic writers (Wang et al. 2021; Vlassenko and Raichle 2015; Zheng et al. 2021). This is not a generic “Warburg” reflex: it is pathway-specific, node-addressable, and disease-relevant. Emerging human-anchored reviews now frame microglial phenotype as the downstream readout of integrated fuel handling (glucose, glutamine, fatty acids), mitochondrial function, and the cytosolic licensing of inflammasomes (Sadeghdoust et al. 2024).

Two levers dominate this switchboard. First, the mTOR-AMPK axis sets the metabolic tone: mTOR activation favors glycolytic, cytokine-competent states, while AMPK activation restrains glycolysis, supports mitochondrial quality control, and promotes resolution programs (Ulland et al. 2017; Garza-Lombó et al. 2018). Second, chronic microglial complex I activity via reverse electron transport (RET) drives mitochondrial ROS and sustains neuroinflammation; limiting complex I/II-RET signaling protects neural circuits (Peruzzotti-Jametti et al. 2024a). Elegant recent work shows that blocking complex I-RET in pro-inflammatory microglia reduces chronic inflammatory damage and preserves neural function, nominating respiratory control as a therapeutic entry point rather than a housekeeping detail (Peruzzotti-Jametti et al. 2024a).

Krebs-cycle-linked metabolites act as fast signaling currencies that hard-wire inflammatory bias. Transient spikes in succinate stabilize HIF-1α and drive IL-1β production, providing a metabolic short circuit from mitochondrial flux to cytokine output (Tannahill et al. 2013). Conversely, itaconate, generated by IRG1, exerts brake-like control by alkylating KEAP1, inhibiting succinate dehydrogenase, and directly blunting NLRP3 activation (Mills et al. 2018b). These metabolites also imprint longer-lived “trained-immunity” or tolerance programs through chromatin remodeling microglia remember metabolic danger and adjust future responses accordingly (Netea et al. 2020).

Beyond mTOR–AMPK signaling, mitochondrial ROS, and TCA-derived metabolites, glycolytic rewiring can also lock inflammatory programs through metabolite-coupled chromatin regulation. Lactate is not only a terminal product of aerobic glycolysis but can serve as a substrate for histone lactylation, directly linking carbon flux to transcriptional control. In an AD context, histone lactylation is increased in the brains of 5XFAD mice and in human AD samples, with elevated H4K12 lactylation (H4K12la) detected in microglia; H4K12la is enriched at promoters of glycolytic genes and supports transcriptional upregulation, forming a positive feedback loop between glycolysis, H4K12la, and PKM2 that exacerbates microglial activation and neuroinflammation. Pharmacologic inhibition of PKM2 and microglia-specific genetic ablation of Pkm2 attenuate microglial activation, reduce amyloid burden, and improve learning/memory in 5XFAD mice, nominating lactylation-reinforced glycolysis as a druggable amplification motif rather than a passive by-product of inflammation (Pan et al. 2022). A second emerging layer is the non-canonical immune regulatory role of metabolic enzymes. Hexokinase 2 (HK2) is selectively enriched in microglia relative to other brain cell types and functions as a metabolic checkpoint that gates glycolytic flux while also shaping mitochondrial activity. Genetic ablation of HK2 reduces microglial glycolytic flux/energy production and dampens surveillance and damage-triggered migration; however, under immune challenge and in an ischaemic stroke model, HK2 loss can aggravate neuroinflammation, which is associated with mitochondrial dysfunction and increased mitochondrial ROS, and can be attenuated by mitochondria-targeted antioxidant rescue. Mechanistically, HK2’s mitochondrial association is implicated in regulating mitochondrial permeability/ROS signaling and can intersect with inflammasome activation pathways, emphasizing that enzyme expression and localization can couple glucose handling to inflammatory thresholds (Hu et al. 2022). Inflammasome licensing sits at the convergence of these metabolic cues. NLRP3 requires priming (NF-κB-driven transcription) and activation (ion flux, mitochondrial stress, cardiolipin/extracellular ATP signals). Metabolic stress tilts both steps: glycolytic bias, complex I-RET-derived mtROS, and accumulated TCA intermediates lower the activation threshold (Peruzzotti-Jametti et al. 2024a; Tannahill et al. 2013). Contemporary syntheses map a dense regulatory network NEK7, ubiquitin ligases, and ion channels that explains how microglia become “hair-triggered” under chronic metabolic strain (Cheng and Zhao 2024; Li et al. 2023; He et al. 2016). Crucially, this is druggable: tool compounds such as MCC950 and electrophile metabolites like itaconate derivatives demonstrate that NLRP3 can be shut down upstream of catastrophic pyroptosis, providing a rational adjunct to amyloid- or tau-directed therapy.

The metabolic logic of innate immunity is exemplified by microglia, which translate energy availability into inflammatory tone through reciprocal mTOR-AMPK signaling and metabolite-driven inflammasome licensing (Fig. 2). Although microglia are the clearest exemplar of immune-metabolic switching, astrocytes are an equally decisive metabolic control layer in AD: they gate synaptic fuel availability (lactate/ketone handling), constrain excitotoxic risk via glutamate clearance and glutamine recycling, and buffer activity-induced lipid stress via FA trafficking and β-oxidation. Together, microglia and astrocytes implement a coupled “fuel–inflammation” circuit in which immune activation and metabolic support cannot be analytically separated (Shen et al. 2023; Suzuki et al. 2011; Ioannou et al. 2019b).

Fig. 2.

Microglial immune-metabolic switching. Microglia integrate metabolic and inflammatory cues to toggle between homeostatic and pro-inflammatory states. In the homeostatic (M0/M2-like) state, oxidative phosphorylation (OXPHOS) predominates within mitochondria, sustained by AMPK activation, autophagy, and mitochondrial quality control to support waste clearance and synaptic maintenance. In the homeostatic state, microglia exhibit a highly ramified morphology with elongated processes reflecting continuous environmental surveillance, in contrast to the amoeboid morphology associated with pro-inflammatory activation. Upon pattern-recognition or metabolic-danger signaling (PAMPs, DAMPs, succinate), microglia shift toward aerobic glycolysis with HIF-1α stabilization and mTOR activation, driving lipid synthesis, NLRP3 inflammasome licensing, and IL-1β/IL-18 release. This metabolic reprogramming links nutrient sensing to chronic neuroinflammation in Alzheimer’s disease and highlights AMPK-mTOR balance as a therapeutic axis to restore resolution phenotypes

Genetics hard-codes part of this immunometabolic bias. APOE4, the strongest common genetic risk factor for late-onset AD, does more than alter lipid trafficking it reshapes microglial bioenergetics and organelle physiology. Human brain-anchored work shows APOE4/4 microglia accumulate lipid droplets, adopt a neurotoxic secretome that propagates tau pathology, and exhibit metabolic features consistent with a glycolysis-favored, inflammasome-competent state (Haney et al. 2024b; Marschallinger et al. 2020).

A practical synthesis emerges. Microglial inflammatory output is the algebra of fuel availability (BBB glucose and alternative substrates), organelle fitness (complex I/RET control and mitochondrial ROS), pathway toggles (mTOR vs. AMPK), and metabolite messaging (succinate vs. itaconate) filtered through genetic and aging contexts (Tannahill et al. 2013; Mills et al. 2018a; Garza-Lombó et al. 2018). This algebra explains clinical paradoxes: why partial amyloid removal yields modest benefits if microglia remain metabolically pre-licensed; why vascular and metabolic endotypes show exaggerated neuroinflammatory readouts; and why interventions that appear peripheral such as GLP-1 receptor agonists or SGLT2 inhibitors plausibly shift CNS immunity by rebalancing systemic fuels, dampening inflammasome tone, and improving endothelial substrate delivery (Li et al. 2021). In our view, the most testable near-term strategy is combinatorial: limit microglial complex I/RET-driven ROS, bias the mTOR-AMPK dial toward resolution, and raise the NLRP3 activation threshold, while addressing upstream systemic dysmetabolism that continuously re-primes the brain.

Temporal Dynamics of Immunometabolic States Across Alzheimer’s Disease Progression

AD unfolds over decades, and accumulating evidence suggests that immunometabolic mechanisms are not static across this trajectory but instead evolve from adaptive to maladaptive states (Jack et al. 2018a, b; Livingston et al. 2024). In early or preclinical stages, glial metabolic reprogramming may serve compensatory functions. Microglial shifts toward glycolysis, enhanced phagocytic activity, and transient inflammasome priming can facilitate amyloid clearance and tissue surveillance (Keren-Shaul et al. 2017; Hickman et al. 2013; Ulland et al. 2017), while astrocytic upregulation of lactate shuttling, glutamate–glutamine cycling, and lipid buffering supports synaptic resilience under conditions of fluctuating energy supply (Suzuki et al. 2011; Pellerin and Magistretti 1994; Vandenberg and Ryan 2013; Ioannou et al. 2019a). These early responses are consistent with a model in which metabolic stress is initially interpreted as a reversible perturbation, triggering protective rewiring rather than overt neurotoxicity [2,4].

With disease progression, however, persistent systemic dysmetabolism, vascular insufficiency, and genetic risk factors such as APOE4 appear to lock these same pathways into chronically activated states (Jack et al. 2018a, b; Sweeney et al. 2018b; Marschallinger et al. 2020; Tcw et al. 2022). Sustained glycolytic bias, impaired mitochondrial quality control, lipid-droplet accumulation, and lowered inflammasome activation thresholds progressively erode the distinction between adaptive immune surveillance and maladaptive inflammation (Tcw et al. 2022; Marschallinger et al. 2020; Heneka et al. 2015a; Peruzzotti-Jametti et al. 2024b; Ulland et al. 2017). What may initially function as a short-term metabolic buffer or danger response becomes a self-reinforcing loop of energetic inefficiency, oxidative stress, and cytokine release, ultimately accelerating synaptic failure and neuronal loss.

This temporal reframing has important therapeutic implications. Interventions targeting immunometabolism are likely to be stage-dependent: early modulation may aim to preserve or fine-tune adaptive metabolic programs, whereas later interventions may require active suppression or rewiring of entrenched inflammatory states. Explicitly recognizing this shift from adaptive to maladaptive immunometabolic logic helps reconcile divergent findings in the literature and underscores why timing, disease stage, and metabolic endotype are likely to determine therapeutic responsiveness in AD (Cummings et al. 2023; Livingston et al. 2024).

Astrocyte Metabolism and Neuron Support

Astrocytes sit at the metabolic crossroads of the brain, converting systemic cues into local substrate allocation, lipid traffic, and inflammatory tone. In physiological conditions, they buffer glutamate, supply neurons with glycolytically derived metabolites such as lactate and serine, and export cholesterol packaged on ApoE lipoproteins that are taken up by neuronal LDL-receptor-family members to build synapses and maintain membrane excitability (Suzuki et al. 2011; Sotelo-Hitschfeld et al. 2015; Mauch et al. 2001; Kanekiyo et al. 2014; Karasinska et al. 2009). This outsourcing of lipid and energy management to astrocytes is not ornamental; it is a design principle of the adult brain in which neurons prioritize signal transmission while glia orchestrate fuel routing and membrane biogenesis. When this division of labor is stressed by systemic dysmetabolism or genetic risk, astrocytes do not merely fail, they actively reshape the neuronal milieu in ways that bias the network toward degeneration (Lee et al. 2021; Tcw et al. 2022).

Three mechanistic channels are central. First, the astrocyte-neuron lactate axis tunes synaptic computation and plasticity. Contemporary work refines the classic Pellerin-Magistretti model: activity- and K⁺-dependent astrocyte depolarization elevates glycolysis and lactate release, and neuronal uptake of lactate via MCT2 is required for long-term memory formation (Suzuki et al. 2011; Sotelo-Hitschfeld et al. 2015; Ruminot et al. 2011; Alberini et al. 2018). Disruptions of astrocytic lactate provision or neuronal lactate import alter memory circuits and higher-order behavior, underscoring that lactate is a bona fide signaling substrate acting via the lactate receptor HCAR1/GPR81 rather than a mere overflow metabolite (Suzuki et al. 2011; Morland et al. 2017; Briquet et al. 2022; Kennedy et al. 2022). In AD, chronic vascular fuel limitation and insulin resistance plausibly throttle this astroglial buffer, lowering the safety margin for synaptic activity and thereby increasing the cost of inflammation (Vinuesa et al. 2021). Together with lactate shuttling, glutamate–glutamine cycling defines an astrocyte-dominant metabolic buffering system that is highly effective under physiological conditions but energetically fragile under systemic dysmetabolism. Glutamate–glutamine cycling constitutes a core astrocytic metabolic axis that couples fuel availability to synaptic stability (Vandenberg and Ryan 2013; Danbolt 2001; Zhang et al. 2022b). Astrocytes clear the majority of extracellular glutamate through high-capacity transporters (EAAT1/GLAST and especially EAAT2/GLT-1), maintaining low extrasynaptic glutamate and limiting excitotoxic signaling (Fontana 2015). Once internalized, glutamate is converted to glutamine by astrocytic glutamine synthetase via an ATP-dependent amidation reaction; glutamine is then supplied back to neurons to sustain neurotransmitter pools, linking neurotransmission to brain energy metabolism (Rose et al. 2013; Andersen et al. 2021; Murphy-Royal et al. 2017). Disruption of this astrocyte-dependent conversion step has been demonstrated in an AD-relevant model (3xTg-AD), where astrocytic glutamine synthetase declines despite preserved GLT-1/EAAT2 expression, consistent with selective vulnerability of the recycling step (Kulijewicz-Nawrot et al. 2013). Because glutamate uptake is tightly coupled to ionic gradient maintenance (Na⁺/K⁺-ATPase activity), impairments in energy availability can plausibly narrow the energetic margin for glutamate clearance/recycling and increase susceptibility to excitotoxic–inflammatory amplification (McKenna 2013; Massucci et al. 2013).

Second, cholesterol economics link astrocyte biology to amyloid and tau ecosystems. Astrocytes are the principal source of brain lipoproteins: ABCA1 lipidates nascent ApoE, which ferries cholesterol and phospholipids to neurons via LDLR-family receptors (LDLR, LRP1, VLDLR, ApoER2). This supply is rate-limiting for synaptogenesis and receptor compartmentalization (Haney et al. 2024a; Karasinska et al. 2009; Belaidi et al. 2025b). In AD-relevant states, ApoE genotype and astrocyte lipidation status shift neuronal membrane order and lipid-raft composition, conditions that favor amyloidogenic APP processing; human iPSC systems show APOE4 astrocytes oversupply cholesterol and expand neuronal lipid rafts, increasing Aβ generation, while APOE4 also drives lysosomal cholesterol sequestration and de novo cholesterol synthesis programs in astrocytes (Karasinska et al. 2009; Tcw et al. 2022). A third astrocyte-specific pathway with immunometabolic consequences is fatty-acid handling and mitochondrial β-oxidation. Astrocytes buffer activity-induced lipid stress by accepting excess fatty acids generated during neuronal hyperactivity and sequestering them into lipid droplets, thereby preventing lipotoxic damage to neuronal membranes (Ioannou et al. 2019a; Qi et al. 2021). These lipids are subsequently processed through mitochondrial β-oxidation as part of an astrocyte-mediated detoxification program that preserves circuit integrity, linking lipid buffering to astrocytic bioenergetics and redox control (Ioannou et al. 2019a; Morant-Ferrando et al. 2023). This neuron–astrocyte lipid coupling is dynamically engaged by neuronal activity, and experimental impairment of key FAO nodes, including CPT1A-dependent mitochondrial entry, disrupts mitochondrial organization and is associated with cognitive deficits in vivo, establishing functional relevance beyond lipid storage (Morant-Ferrando et al. 2023; Qi et al. 2021). In AD-relevant contexts, APOE4 is associated with disruption of neuron–astrocyte lipid coupling, astrocytic lipid-droplet accumulation, and metabolic stress, linking genetic risk to maladaptive astrocyte programs and heightened susceptibility to neuroinflammatory amplification (Tcw et al. 2022; Marschallinger et al. 2020; Yamazaki et al. 2019a).

Third, APOE genotype rewires astrocyte organelle physiology and innate-immune crosstalk. Human cellular and in vivo work converge on an APOE4-biased phenotype characterized by lipid-droplet accrual, endo-lysosomal cholesterol stress, and impaired neuron–astrocyte coupling of fatty-acid metabolism, shifting secretory and inflammatory profiles toward neurotoxic states that potentiate proteopathic spread (Lee et al. 2021; Tcw et al. 2022; Qi et al. 2021). The practical corollary is that interventions improving ApoE lipidation (e.g., via LXR/RXR→ABCA1) or normalizing astrocytic sterol flux may raise the threshold for inflammatory activation and synaptic failure in APOE4 carriers (Yamazaki et al. 2019b; Belaidi et al. 2025a; Karasinska et al. 2009).

Microglia complete this circuit by dictating astrocyte fate under stress. Classically activated microglia release IL-1α, TNF and C1q, a cytokine triad sufficient to induce A1-type reactive astrocytes that lose trophic functions and acquire neurotoxic programs linking microglial priming to astrocyte loss-of-function/gain-of-toxicity and accelerated neuronal injury (Liddelow et al. 2017).

Two observations sharpen the translational edge. First, astrocyte metabolism is exquisitely plastic: adenosine and K⁺ signaling, circuit activity, and substrate availability rapidly retune glycolysis, lactate release, and substrate export implying that even modest improvements in vascular fuel delivery or systemic insulin sensitivity could re-expand the astrocytic energy buffer in humans (Suzuki et al. 2011; Sotelo-Hitschfeld et al. 2015; Ruminot et al. 2011; Sweeney et al. 2018b; Vinuesa et al. 2021). Second, lipid-centric levers LXR/RXR agonism to enhance ABCA1-dependent ApoE lipidation, ApoE-mimetic approaches, or strategies that clear pathologic lipid droplets should be evaluated not only for plaque/tau endpoints but for synaptic physiology and glial inflammatory thresholds, where effects may be larger and faster to detect (Yamazaki et al. 2019b; Belaidi et al. 2025a; Karasinska et al. 2009). In short, astrocytes are not passive victims of AD pathology; they are programmable agents whose metabolic state determines whether neurons face adversity with resilience or collapse.

This concept is exemplified in Fig. 3, which summarizes how APOE4 reprograms astrocytic metabolism impairing lactate and cholesterol trafficking, promoting lipid droplet accumulation, and triggering inflammatory crosstalk that undermines neuronal resilience.

Fig. 3.

Astrocyte dysfunction and the APOE4 link. Healthy APOE3 astrocytes sustain neuronal function through efficient cholesterol trafficking and lactate shuttling. Under APOE4 genotype or metabolic stress, astrocytes accumulate lipid droplets, secrete inflammatory mediators, and lose neurotrophic support, leading to impaired neuron–astrocyte coupling and neurotoxicity

Neurovascular and Blood–Brain Barrier Metabolism: The Failing Interface of Fuel and Immunity

AD progressively erodes the brain’s vascular-metabolic synchrony. The neurovascular unit (NVU) a tightly coupled consortium of endothelial cells, pericytes, astrocytic endfeet, and neurons acts as the brain’s metabolic switchboard, matching substrate delivery to neuronal activity while restraining immune traffic (Iadecola 2017; Sweeney et al. 2018b). In the healthy brain, endothelial mitochondria sustain barrier integrity, pericytes regulate capillary tone, and astrocytic endfeet buffer ions and lactate (Iadecola 2017; Sweeney et al. 2018b; Divecha et al. 2025). With aging and metabolic stress, this choreography collapses: glucose and oxygen delivery falter, endothelial cells adopt pro-inflammatory phenotypes, and the BBB becomes both a metabolic bottleneck and an inflammatory amplifier.

Fuel Delivery Failure as an Initiating Lesion

High-resolution PET studies show that cerebral glucose hypometabolism appears years before cognitive symptoms (Mosconi et al. 2010) and can predate or progress independently of classical proteinopathy in at-risk individuals. Mechanistically, endothelial GLUT1 downregulation curtails glucose entry; and while BBB lactate transport can be altered in models, consistent human evidence for reduced BBB MCT1 is lacking (Winkler et al. 2015; Montagne et al. 2017; Knox et al. 2022). Experimental, endothelial-specific GLUT1 reduction is sufficient to trigger neurovascular uncoupling, BBB breakdown, Aβ accumulation, and synaptic/neuronal degeneration evidence that metabolic insufficiency can drive pathology (Winkler et al. 2015).

Pericytes, which act as local metabolic rheostats, are equally vulnerable. They rely on glycolysis/FAO for contractility and redox balance; pericyte loss or metabolic stress destabilizes capillary perfusion and exposes neurons to fluctuating substrate delivery. In humans and models, pericyte injury (including APOE4-CypA-MMP9 signaling) correlates with BBB leakiness and cognitive decline (Bell et al. 2012; Nation et al. 2019; Montagne et al. 2021).

Endothelial Metabolism and Immune Licensing

Endothelial mitochondria are more than power generators they help define barrier phenotype. Quiescent brain endothelium leans on fatty-acid β-oxidation and restrained glycolysis to support tight-junction maintenance; inflammatory or metabolic stress pushes ECs toward glycolysis/ROS, upregulating VCAM-1/ICAM-1 and licensing leukocyte adhesion and transendothelial migration (Kalucka et al. 2018; Leung and Shi 2022; Wang et al. 2023).

BBB permeability is also controlled by lipid composition. Cholesterol- and sphingolipid-rich microdomains support tight-junction protein organization; disturbed lipid metabolism (e.g., with APOE4 or systemic dyslipidemia) reduces membrane order and disrupts claudin/occludin scaffolding, while excess ceramide signaling promotes endothelial injury and cell death, yielding a structurally and biochemically leaky barrier (Sweeney et al. 2018b; Jernigan et al. 2015; Yuan et al. 2023; Samsonov et al. 2001).

The Hypoperfusion–Inflammation Feedback Loop

Neurovascular decoupling amplifies neuroinflammation in a feed-forward loop: hypoperfusion/hypoxia stabilizes HIF-1α in glia, biases glycolysis, and augments IL-1β production; endothelial cells shed inflammatory mediators that recruit perivascular macrophages, and microglial ROS further damage endothelial mitochondria driving basement-membrane remodeling and arteriole stiffening characteristic of cerebral small-vessel disease increasingly recognized within the AD spectrum (Sweeney et al. 2018b; Folco et al. 2014; Huang et al. 2024; Wardlaw et al. 2019).

Translational Outlook

The vascular-metabolic axis is therapeutically approachable. Agents that restore endothelial metabolism AMPK/SIRT1 pathway modulators and GLP-1 receptor agonists show BBB-protective and perfusion/glucose-transfer benefits in preclinical and early clinical studies (Zlokovic 2011; Winkler et al. 2015; Gejl et al. 2017). Lifestyle and pharmacologic strategies that improve systemic insulin sensitivity and lipids indirectly re-establish NVU homeostasis (Sweeney et al. 2018b). Imaging advances such as ASL-MRI for perfusion and O-15 PET (or hybrid PET/MR methods) for oxygen/glucose flux now support real-time mapping of these processes in humans (Bernetti et al. 2025; Fan et al. 2020). Conceptually, trials should integrate vascular-metabolic endpoints (perfusion, BBB permeability, pericyte/EC biomarkers) alongside amyloid/tau to capture disease modification.

Cumulatively, the NVU/BBB are not collateral victims but a metabolic front line. Their dysfunction converts systemic metabolic noise into neuroinflammatory signal. Re-energizing the vascular interface by protecting endothelial mitochondria, preserving pericyte fitness, and stabilizing lipid microdomains may be as fundamental to slowing AD as removing plaques or tangles.

Lipid Biology as the Bridge Between Metabolism and Inflammation

Among the molecular threads connecting systemic dysmetabolism to Alzheimer’s pathology, lipids form the most versatile and least appreciated bridge. The brain is the body’s most lipid-dense organ, with over half of its dry mass composed of complex lipids that structure membranes, anchor signaling proteins, and govern vesicular traffic. Lipid metabolism therefore defines not only energy homeostasis but also the immunological and communicative architecture of neurons and glia. In AD, perturbations in lipid composition, trafficking, and oxidation collectively convert membranes from signaling platforms into inflammatory scaffolds.

Lipid Compartmentalization and Metabolic Signaling

Under physiological conditions, neuronal and glial membranes are enriched in cholesterol and sphingomyelin, organized into ordered microdomains that regulate receptor clustering and synaptic transmission (Pantelopulos et al. 2024). The biosynthesis and turnover of these lipids depend on coordinated activity between astrocytes, neurons, and microglia. Cholesterol is synthesized primarily by astrocytes, exported via ApoE particles, and then recycled by neurons; sphingolipids are generated through the ceramide–sphingomyelin cycle that tunes membrane order and intracellular signaling (Tcw et al. 2022; Pantelopulos et al. 2024). In metabolic stress, these tightly balanced cycles become asymmetric: reduced cholesterol efflux from ApoE4 astrocytes and increased sphingomyelinase activity in microglia elevate ceramide and dihydroceramide levels, leading to membrane stiffening, altered vesicle curvature, and facilitation of innate immune complex assembly (Pantelopulos et al. 2024; Tcw et al. 2022).

Ceramides and other sphingolipid intermediates serve as bioactive messengers that intersect with cytokine networks. High ceramide levels amplify TLR4 and NLRP3-inflammasome signaling, whereas their depletion or blockade blunts IL-1β release (Summers et al. 2019; Heneka et al. 2015b; Maceyka and Spiegel 2014). This immune-metabolic feedback loop is increasingly recognized as a defining feature of glial pathology: once activated, NLRP3 further stimulates sphingomyelinase, generating more ceramide and perpetuating inflammation.

Lipid Droplets: Metabolic Sentinels or Inflammatory Fuses?

The recent discovery of lipid-droplet–accumulating microglia (LDAMs) reframes our understanding of how lipids participate in neuroinflammation. Single-cell and spatial transcriptomics reveal that LDAMs arise in both aging and AD brains and are enriched for genes controlling lipid catabolism, ROS detoxification, and cytokine release (Marschallinger et al. 2020). Functionally, these microglia exhibit impaired phagocytosis, excessive ROS production, and heightened inflammasome readiness. APOE4/4 microglia show exaggerated droplet formation and pro-inflammatory secretomes, linking genotype to bioenergetic and immunological dysfunction (Haney et al. 2024b). Lipid droplets, far from being inert stores, are metabolically active organelles that concentrate polyunsaturated fatty acids prone to oxidation and act as hubs for innate immune sensors such as NLRP3 and interface with cGAS-STING signaling in a context-dependent manne (Thiam et al. 2013).

Astrocytes display a parallel but potentially reparative form of lipid droplet biogenesis. Under oxidative or excitotoxic stress, they sequester peroxidized fatty acids into droplets, shielding neurons from lipotoxic damage (Thiam et al. 2013; Ioannou et al. 2019b). However, chronic overload or impaired lipid export, as seen with ApoE4, turns this defense into a liability astrocytes become pro-inflammatory, release ceramide-rich extracellular vesicles, and lose their neuroprotective buffering capacity (Tcw et al. 2022). The duality of lipid droplets thus encapsulates the EV paradox of the AD brain: what begins as metabolic rescue ends as inflammatory feed-forward signaling.

Lipid Rafts and Proteopathic Propagation

Lipid reorganization also modifies how amyloid and tau assemble and spread. Cholesterol- and sphingolipid-enriched rafts facilitate amyloid precursor protein (APP) cleavage by β- and γ-secretases; disrupting raft integrity through cholesterol depletion or SREBP2 modulation reduces amyloidogenic processing in models (Pfrieger 2003; Pantelopulos et al. 2024). Similarly, tau propagation occurs preferentially via exosomes and membrane microdomains; microglia-dependent exosome pathways promote tau spread, and ceramide-dependent exosome biogenesis has been linked to Aβ trafficking together suggesting that lipid microenvironment controls proteopathic contagion (Asai et al. 2015; Yuyama et al. 2012).

Therapeutic Implications

Targeting lipid metabolism offers a broad entry point into AD modification. Statins, while inconsistent in cognitive outcomes, improve endothelial function and reduce neuroinflammatory markers in selected metabolic endotypes. Sphingolipid-directed therapies acid sphingomyelinase inhibitors, ceramide synthase modulators, and S1P receptor agonists attenuate neuroinflammation and restore glial phagocytic function in preclinical studies (Brinkmann et al. 2010; Summers et al. 2019; Maceyka and Spiegel 2014). More nuanced strategies involve enhancing ApoE lipidation via LXR/ABCA1 activation or stabilizing lipid raft architecture to restore synaptic signaling fidelity (Tcw et al. 2022; Pantelopulos et al. 2024). Because lipid pathways intersect with mTOR, AMPK, and inflammasome circuits, these interventions can be conceptualized as metabolic anti-inflammatories rather than mere cholesterol modulators.

In essence, lipid biology provides the missing biophysical substrate for the immune-metabolic model of AD. Altered lipid composition reorganizes membranes into pro-inflammatory landscapes, converts glial lipid storage into chronic stress signaling, and shapes the microenvironments where amyloid and tau pathology take root. The next frontier is to move beyond correlative lipidomics and quantify lipid flux and spatial organization in living human brains. This will require integrating high-resolution lipid imaging, stable-isotope tracing, and plasma-CSF lipid exchange studies. Only then can lipid metabolism be repositioned from the periphery of Alzheimer’s biology to its mechanistic core.

Biomarkers and Imaging Readouts: Capturing the Immuno-Metabolic Axis in Humans

A major barrier to translating the immune-metabolic model of AD has been the absence of integrative biomarkers that can quantify metabolic and inflammatory processes simultaneously in living humans. Recent advances across metabolomics, lipidomics, extracellular vesicle (EV) profiling, and multimodal neuroimaging now make this convergence measurable. Together, these tools transform immunometabolism from a conceptual framework into an experimentally and clinically testable domain.

Plasma and CSF Metabolomics: Decoding Systemic Signatures

Untargeted and targeted metabolomic analyses from large-scale cohorts such as ADNI, EPAD, and UK Biobank have uncovered reproducible metabolic fingerprints preceding cognitive decline (Horgusluoglu et al. 2022; Mapstone et al. 2014). Reduced levels of branched-chain amino acids, ketone bodies, and tricarboxylic-acid intermediates coincide with elevated acylcarnitines and sphingolipid derivatives, suggesting impaired mitochondrial β-oxidation and lipid turnover (Mapstone et al. 2014; Horgusluoglu et al. 2022). These alterations are not epiphenomena: flux studies show they correlate with insulin resistance and systemic inflammation, the same features mechanistically linked to microglial and astrocytic dysfunction.

Cerebrospinal fluid (CSF) metabolomics provides complementary information. Shifts in glucose and lactate ratios, decreased citrate, and accumulation of kynurenine-pathway metabolites signal both energy failure and neuroimmune activation (Horgusluoglu et al. 2022). Importantly, several metabolic clusters (e.g., ceramide–sphingomyelin, kynurenine–tryptophan) outperform amyloid or tau alone in predicting cognitive decline when combined with ApoE genotype and vascular risk scores (Mapstone et al. 2014). The next frontier is dynamic assessment longitudinal metabolomic trajectories that capture how systemic therapies reshape the metabolic milieu in parallel with cognition.

Lipidomics and Extracellular Vesicles: Systemic Windows into Glial Metabolism

Plasma and CSF lipidomics bridge peripheral and central compartments. Distinct ceramide, lysophosphatidylcholine, and acylcarnitine patterns stratify early AD from healthy aging and correlate with hippocampal atrophy (Mapstone et al. 2014; Mielke et al. 2012a). These lipidomic endotypes often mirror glial lipid-droplet phenotypes identified in post-mortem tissue, implying that circulating lipids can serve as proxy reporters of microglial and astrocytic metabolism.

EVs Further Enrich This Landscape

Microglia- and astrocyte-derived EVs isolated from plasma or CSF carry metabolic enzymes, sphingolipids, and cytokine transcripts reflective of cell-specific stress states. For example, increased neuronal/astroglial EV cargo related to lysosomal-autophagic pathways and inflammation distinguishes AD from controls and tracks longitudinally with disease progression (Goetzl et al. 2015, 2016). The analytical challenge remains cellular specificity: improved immunocapture and lipidomic profiling now enable discrimination between neuron-, astrocyte-, and microglia-origin EVs, though standardization across platforms is still needed.

Imaging the Immune-Metabolic Brain

Positron emission tomography (PET) and magnetic resonance imaging (MRI) have entered a new era of metabolic resolution. ^18F-FDG PET remains the workhorse for assessing cerebral glucose utilization, but newer tracers such as ^11 C-acetate enable visualization of astrocytic oxidative metabolism with human readouts; ^18F astrocyte-targeted tracers are investigational rather than established in humans (Nam et al. 2023). In parallel, high-field MR spectroscopy detects brain lactate, glutamate, and lipid resonances, directly quantifying metabolic imbalance.

Microglial imaging is also evolving beyond the limitations of TSPO. Next-generation PET ligands targeting CSF1R have first-in-human data, while P2RY12 and NLRP3 tracers are in preclinical/early translation; subtype-specific quantification in humans is promising for CSF1R but remains to be established for P2RY12 and NLRP3 (Coughlin et al. 2022a, b; van der Wildt et al. 2021; Xu et al. 2024). Early human studies show regional overlap between CSF1R-PET signal and hypometabolic regions, reinforcing immune-metabolic coupling; TSPO biology itself warrants caution because it is not microglia-exclusive (Nam et al. 2023; Notter et al. 2021). Combined metabolic-inflammatory PET dual-tracer approaches capturing both glucose flux and immune activation will be central to future intervention trials.

Toward Precision Stratification

Integration of multi-omic and imaging datasets is already refining disease taxonomy. Machine-learning analyses identify reproducible metabolic endotypes insulin-resistant, dyslipidaemic, mitochondrial-oxidative, and vascular-hypoperfused subgroups with distinct trajectories and potential drug responsiveness; combining genetics with metabolomics improves prognostication over single-modality AT(N) markers alone (Liu et al. 2025; Horgusluoglu et al. 2022). A critical step forward will be harmonizing systemic and CNS biomarker platforms combining plasma metabolomics, CSF lipidomics, and metabolic PET to generate multidimensional profiles that can serve both as diagnostic classifiers and pharmacodynamic readouts.

Translational Outlook

The convergence of omics and imaging defines a new stage of Alzheimer’s biomarker science: metabolism is no longer invisible. Longitudinal tracking of immune-metabolic signatures can now test whether interventions such as GLP-1 receptor agonists, SGLT2 inhibitors, ketogenic strategies, or lifestyle modulation produce measurable shifts in CNS metabolism and neuroinflammation (Horgusluoglu et al. 2022; Liu et al. 2025). In effect, biomarker integration provides the experimental infrastructure to verify whether reprogramming metabolism translates into disease modification a decisive step toward precision prevention and treatment of AD.

Therapeutic Landscape: Reprogramming Immunometabolism in Alzheimer’s Disease

The slow erosion of neuronal circuits in AD reflects an imbalance between energy availability, lipid homeostasis, and inflammatory control. Pharmacologic attempts to dismantle amyloid or tau have yielded partial benefit; the next frontier is to reprogram the metabolic context that sustains neuroinflammation. Metabolic drugs long validated in cardiometabolic disease offer a rare combination of mechanistic plausibility, clinical accessibility, and translational immediacy.

Repurposed Metabolic Agents Entering the Alzheimer’s Pipeline

GLP-1 receptor agonists represent the most clinically advanced metabolic therapeutics in AD. Originally developed for type 2 diabetes, these agents modulate insulin sensitivity, enhance mitochondrial biogenesis, and exert direct anti-inflammatory effects in glia. Preclinical models show that liraglutide and semaglutide improve cerebral blood flow, normalize microglial morphology, and reduce amyloid and tau burden independently of glycaemia. The phase 3 EVOKE and EVOKE + trials now test semaglutide in early AD, while secondary analyses from large population cohorts report reduced incidence of dementia among long-term GLP-1RA users. Mechanistically, GLP-1R signaling activates cAMP-PKA-CREB pathways in neurons and glia, enhances autophagy, and suppresses NF-κB and NLRP3 activation essentially biasing the brain toward an anti-inflammatory metabolic set-point (Lin et al. 2025; Wang et al. 2024; Zheng et al. 2024; Cummings et al. 2025a).

SGLT2 inhibitors (empagliflozin, dapagliflozin) offer complementary benefits by promoting glycosuria and mild systemic ketosis, improving endothelial function, and lowering circulating ceramide levels. In rodent models of AD and cerebral hypoperfusion, SGLT2 blockade restores BBB integrity and reduces microglial priming, partly through AMPK activation. Although human cognitive data remain preliminary, SGLT2i use is associated with improved cerebrovascular compliance and reduced risk of incident dementia in registry analyses (Kim et al. 2024; Tu et al. 2024).

Metformin, one of the oldest metabolic drugs, exemplifies dual-system reprogramming. By activating AMPK and inhibiting complex I of the respiratory chain, it reduces systemic insulin resistance while restraining glial glycolysis and inflammasome activation. Epidemiologic evidence for metformin’s neuroprotection is mixed confounded by disease stage and comorbidities but mechanistic trials show improvements in brain glucose uptake and functional connectivity. Its low cost, safety, and penetrance make metformin a strategic component of combination trials targeting metabolic and neuroinflammatory nodes (Luchsinger et al. 2016; Rosell-Díaz and Fernández-Real 2023).

Lipid-Centric and Mitochondrial Interventions

Lipid modulation extends beyond statins. LXR and PPAR agonists enhance ApoE lipidation and ABCA1-dependent cholesterol efflux, improving glial lipid trafficking and reducing amyloidogenic processing (Yamazaki et al. 2019b; Maceyka and Spiegel 2014; Summers et al. 2019). While first-generation agonists failed because of hepatic steatosis, brain-selective or partial agonists under development aim to decouple CNS benefit from peripheral toxicity. Parallel efforts focus on ceramide synthase inhibition and acid sphingomyelinase modulation, which in models normalize microglial metabolism and attenuate neuroinflammation.

Mitochondrial-targeted antioxidants such as MitoQ and SS-31 (elamipretide), and NAD⁺-repletion strategies using nicotinamide riboside or NRH, restore oxidative balance and improve synaptic resilience (Peruzzotti-Jametti et al. 2024b). These agents act upstream of inflammation by maintaining the mitochondrial membrane potential that constrains NLRP3 activation. Combining them with metabolic re-balancers (GLP-1RAs or AMPK activators) offers a coherent path to sustain microglial respiratory competence (Saxton and Sabatini 2017; Karaa et al. 2023; Tannahill et al. 2013).

Dietary and Lifestyle Interventions: Physiology as Pharmacology

Precision nutrition now intersects mechanistic neurobiology. Ketogenic and fasting-mimicking regimens provide alternative fuels (β-hydroxybutyrate) that engage G-protein–coupled receptors and HDAC inhibition, collectively lowering NLRP3 tone. Controlled trials reveal improvements in cognitive scores and network connectivity over six to twelve months in mild cognitive impairment, though adherence remains a challenge. Exercise, likewise, functions as a metabolic drug: it enhances brain-derived neurotrophic factor (BDNF), stimulates lactate signaling, and restores vascular reactivity. The magnitude of effect rivals single-agent drugs, especially when combined with GLP-1RA therapy (Fortier et al. 2021; Baker et al. 2025).

Emerging Immune-Metabolic Targets

Several intracellular nodes link energy metabolism and inflammation and are now pharmacologically tractable. AMPK activators (e.g., MK-8722 derivatives) reduce tau phosphorylation and microglial activation. mTOR inhibitors (rapalogs) restore autophagic flux and normalize lipid handling, though systemic immunosuppression limits chronic use (Saxton and Sabatini 2017). NLRP3 inhibitors (e.g., dapansutrile, MCC950 analogs) have advanced into early clinical testing for neurodegenerative indications, showing brain penetration and cytokine reduction without major toxicity (Coll et al. 2015; Mangan et al. 2018; Klück et al. 2020). Collectively, these compounds inaugurate a new class of “metabolic anti-inflammatories” that operate upstream of cell death and synaptic loss. BBB accessibility represents a critical translational constraint for immunometabolic therapies. Several interventions discussed here may exert central effects indirectly, by improving systemic metabolism, endothelial function, and substrate delivery to the brain, even without robust penetration of an intact BBB. Others target brain-accessible nodes, either through small molecules with demonstrable CNS exposure or by exploiting disease-associated BBB permeability. Importantly, BBB integrity itself evolves across Alzheimer’s disease progression, suggesting that disease stage and neurovascular status may critically shape therapeutic responsiveness. Explicit consideration of BBB accessibility therefore complements immunometabolic target selection and reinforces the need for stage-aware translational strategies.

Precision Endotyping and Trial Design

The immune-metabolic framework compels a redesign of AD clinical trials. Rather than recruiting broad, pathology-defined populations, next-generation trials should stratify by metabolic endotypes insulin-resistant, dyslipidaemic, or vascular-hypoperfused identified through metabolomic and imaging signatures. Endpoints must include not only cognitive scales but metabolic and inflammatory biomarkers (FDG-PET, TSPO- or CSF1R-PET, plasma ceramide panels, metabolomic trajectories). Such composite endpoints will capture drug effects on the processes that sustain neurodegeneration rather than its terminal histopathology.

Conceptual Synthesis

Immuno-metabolic therapy reframes AD from an untreatable neurodegeneration into a reversible systems disorder. The conceptual power lies in its scalability: pathways tuned by GLP-1RAs, AMPK activators, or dietary ketosis are already druggable and safe. The therapeutic goal is not merely to suppress inflammation but to restore the energy logic of the brain balancing nutrient sensing, mitochondrial fitness, and lipid signaling to shift the system from chronic activation to metabolic homeostasis. Success here would redefine disease modification, aligning prevention and therapy within the same biochemical grammar.

Precision Medicine Roadmap: Rebalancing the Immuno-Metabolic Axis

AD emerges not from a single lesion but from the failure of metabolic governance across interconnected cellular and systemic networks. The immune-metabolic model reframes the disease as a multiorgan disorder in which systemic dysmetabolism, endothelial dysfunction, and glial metabolic maladaptation coalesce into chronic neuroinflammation. This reframing transforms prevention and therapy from late-stage proteinopathy management to early metabolic regulation. The challenge and opportunity are to convert this conceptual synthesis into a precision medicine strategy with measurable endpoints and actionable interventions.

From Descriptive Biomarkers To Mechanistic Stratification

Current AT(N) biomarkers amyloid, tau, neurodegeneration accurately stage pathology but offer limited predictive value for therapeutic response. Integrating metabolic and inflammatory biomarkers adds mechanistic resolution. Plasma and CSF metabolomics, lipidomics, and EV profiles can define metabolic endotypes insulin-resistant, dyslipidaemic, mitochondrial-stressed, or vascular-hypoperfused each representing a distinct therapeutic target class. Such endotyping is already transforming oncology and cardiometabolic fields; applying the same logic to AD could align metabolic interventions with biological subtypes rather than clinical severity (Jack et al. 2018a, b, 2024; Liu et al. 2025).

Multi-level Intervention Design

Precision medicine in AD must operate across three metabolic tiers. At the systemic level, early identification and correction of dyslipidaemia, insulin resistance, and hypertension are foundational; pharmacologic and lifestyle interventions at midlife yield the largest risk reductions (Livingston et al. 2024; Feng and Gao 2024; Barloese et al. 2022). At the vascular–endothelial level, strategies that restore BBB metabolism, maintain pericyte fitness, and modulate endothelial mitochondria through AMPK activators, GLP-1 receptor agonists, or SGLT2 inhibitors are projected to amplify cognitive benefit by improving substrate delivery and neurovascular coupling (Zheng et al. 2024; Lee et al. 2025b; Fisslthaler and Fleming 2009). At the glial–neuronal level, direct modulation of microglial and astrocytic metabolism via mTOR–AMPK tuning, NLRP3 inhibition, or ApoE lipidation enhancement represents the most mechanistically specific approach to extinguish maladaptive inflammation while preserving tissue repair. Each tier requires tailored biomarker feedback loops to monitor target engagement in real time.

Digital and Multi-omics Integration

Machine-learning pipelines can integrate multi-omic and imaging data to create individualized immune-metabolic fingerprints. Combined with wearable biosensors that track sleep, activity, and glycaemic variability, such systems can continuously map the metabolic tone of patients and predict transitions toward decompensation. Longitudinal modeling of these data layers will redefine clinical trial design: rather than static before–after comparisons, trials can adapt therapy dynamically based on metabolic and inflammatory trajectories. This data-rich design aligns with the trajectory-based precision frameworks emerging in oncology and rare metabolic disorders (Butler et al. 2025; Bracher-Smith et al. 2025; Qi et al. 2025; Fowler et al. 2025).

Translational Priorities

Three immediate priorities arise from this synthesis.

First, develop integrative biomarkers that link systemic interventions to CNS metabolic shifts metabolomic and PET surrogates must move from observational to regulatory-grade validation (Jack et al. 2018a, b, 2024). Second, establish cross-disciplinary trial networks combining metabolic clinics, neurology centers, and imaging cores to test repurposed drugs in biomarker-defined subgroups (Barloese et al. 2022; Feng and Gao 2024). Third, invest in model systems that replicate systemic CNS metabolic coupling: organ-on-chip platforms, humanized BBB micro-physiological models, and induced-pluripotent-stem-cell co-cultures with glial and endothelial compartments (Ingber 2022; Hajal et al. 2022). Only with such models can candidate drugs be screened for effects on energy allocation, lipid flux, and inflammasome thresholds before human translation.

Conceptual Closure: Metabolism as the Grammar of Neuroinflammation

If amyloid and tau are the nouns of AD, metabolism is its grammar the structural rule set that determines how these pathologies interact, persist, and propagate. Immuno-metabolic therapy thus aims not merely to edit the nouns but to rewrite the grammar. By targeting energy sensing, lipid orchestration, and metabolic–immune crosstalk, it becomes possible to alter the trajectory of disease long before neuronal death is irreversible. The practical endgame is an adaptive, mechanism-guided prevention-therapy continuum in which systemic metabolic health, endothelial performance, and glial immunometabolism are monitored and modulated as one network.

The intellectual return is equally transformative: AD ceases to be a neurocentric enigma and becomes a testbed for systems medicine a model for how organ-level metabolism and immunity can be tuned to preserve cognitive resilience across the lifespan.

Outlook and Roadmap

The immune-metabolic framework recasts AD as the product of systemic, vascular and glial energy failure rather than an isolated neuronal disorder. The challenge for the coming decade is to translate this conceptual shift into concrete, testable hypotheses and clinically actionable tools that can be validated across models, biomarkers and human trials.

Testable Predictions

Metabolic alterations are likely to emerge before classical pathology. Longitudinal metabolomics and neuroimaging should demonstrate that plasma and CSF lipid-metabolite signatures especially ceramide–sphingomyelin and acylcarnitine ratios shift before amyloid positivity, establishing metabolic dysfunction as an initiating lesion. The degree of microglial respiratory reprogramming will probably determine response to therapy: patients with glycolytic, inflammasome-biased microglia may respond better to metabolic interventions than to anti-amyloid antibodies. Restoration of endothelial and BBB metabolism is expected to improve both perfusion and potentially cognition, while correcting ApoE4-linked lipid-droplet accumulation should re-establish glial metabolic flexibility.

Experimental Priorities

New model systems are essential. Humanized organ-on-chip and iPSC co-culture platforms that couple neurons, glia and endothelium should allow real-time tracking of fuel flux, inflammasome activation and barrier permeability features that rodent models cannot recapitulate. Stable-isotope tracing and spatial metabolomics will move the field from static profiles to dynamic flux analysis, revealing how systemic fuels are redistributed within the brain. Cross-disciplinary clinical networks linking cardiometabolic, endocrine and neurology clinics will be needed to harmonize biomarkers and enable mechanism-based trial design. Computational models that integrate metabolomic, lipidomic and imaging data can infer patient-specific immune-metabolic states and guide adaptive dosing strategies.

Clinical Translation Milestones

In the near term, composite metabolic-inflammatory biomarkers combining plasma ceramides, FDG-PET, and EV transcripts should process toward regulatory qualification as early efficacy endpoints. Within five years, a platform trial combining a GLP-1 receptor agonist, SGLT2 inhibitor and metformin in biomarker-defined prodromal AD could establish proof-of-concept for metabolic disease modification. Parallel cross-talk with cardiovascular medicine will accelerate repositioning of vascular-protective drugs, while mid-life immune-metabolic screening fasting ceramides, ApoE genotype, insulin resistance and vascular imaging may evolve into standard dementia-risk assessment. Ultimately, computational “digital twins” of Alzheimer’s metabolism, integrating personalized vascular and glial parameters, could simulate trajectories and de-risk therapeutic combinations before human testing.

Strategic Outlook

The next phase of Alzheimer’s research will hinge on viewing metabolism as the grammar of neuroinflammation the rule set that dictates how amyloid and tau interact and propagate. Mechanistic interdependence across microglial bioenergetics, astrocytic lipid export and endothelial mitochondrial health defines a single regulatory system that can be tuned rather than merely suppressed. Achieving this requires a precision-medicine continuum that monitors metabolic health, vascular function and innate immune tone as one network. The pharmacological armamentarium already exists; what remains is integration, guided by biomarkers and patient-specific metabolic maps.

When realized, this strategy will not only redefine therapeutic success in Alzheimer’s disease but also establish a prototype for systems medicine, where re-engineering metabolism becomes the principal route to restoring cognitive resilience.

Box 1 | Immunometabolism Essentials for Neurologists

mTOR (Mechanistic Target of Rapamycin)

A nutrient-sensing kinase complex (mTORC1/2) that couples amino acids, growth factors, and cellular energy to protein/lipid synthesis and autophagy.

AD relevance: mTORC1 biases microglia and astrocytes toward glycolytic, pro-inflammatory states and suppresses autophagy, lowering clearance of Aβ/tau and other aggregate-prone cargo, including stress granules.

AMPK (AMP-Activated Protein Kinase)

The cell’s energy gauge; activated by rising AMP/ADP to restore ATP via catabolism and to inhibit anabolic programs (including mTORC1).

AD relevance: AMPK activation promotes mitochondrial quality control and resolution phenotypes in glia, improves endothelial/BBB metabolism, and counterbalances mTOR-driven inflammation.

Itaconate (IRG1/ACOD1-Derived Metabolite)

A TCA-cycle-related macrophage/microglial metabolite that alkylates cysteines (e.g., KEAP1), boosts NRF2 antioxidant programs, and restrains succinate dehydrogenase/NLRP3.

AD relevance: Endogenous “metabolic brake” that raises the inflammasome activation threshold and limits chronic IL-1β signaling in the brain.

Trained Immunity (Innate Immune Memory)

Long-lived epigenetic/metabolic reprogramming of innate cells (including microglia) after a first hit, resulting in heightened or tolerized future responses.

AD relevance: Systemic dysmetabolism and prior inflammatory cues can “pre-license” microglia toward exaggerated glycolytic, inflammasome-ready states that accelerate neurodegeneration.

Glycolysis ↔ OXPHOS Switch

Rapid reallocation between aerobic glycolysis (fast ATP, pro-inflammatory bias via HIF-1α) and mitochondrial oxidative phosphorylation (restorative/resolution programs).

AD relevance: Chronic tilt toward glycolysis sustains cytokines and impairs phagocytosis; preserving microglial respiratory capacity is protective.

NLRP3 Inflammasome

A danger-sensor complex (NLRP3–ASC–caspase-1) that matures IL-1β/IL-18 and induces pyroptosis; licensed by NF-κB “priming” and activated by ion flux or mitochondrial stress.

AD relevance: Integrates metabolic danger signals (succinate, ROS, lipid stress) into neuroinflammation; a druggable upstream node.

Succinate/Fumarate (TCA “signals”)

Succinate stabilizes HIF-1α and fuels IL-1β; fumarate modifies proteins (succination).

AD relevance: Elevated during hypoperfusion and mitochondrial stress, tending to lower inflammasome thresholds.

NRF2–KEAP1 Axis

Antioxidant transcriptional program restrained by KEAP1 and released by electrophiles (e.g., itaconate).

AD relevance: Restores redox balance in glia and endothelium and dampens inflammatory tone.

Glutaminolysis

Conversion of glutamine to TCA intermediates to support biosynthesis and immunity when glucose is limited.

AD relevance: Microglia exhibit glutamine-fueled flexibility in vivo; greater reliance can support persistent inflammatory readiness during BBB fuel deficits.

Lactate Shuttle (astrocyte→neuron)

Activity-linked export of lactate from astrocytes to neurons as a signaling fuel.

AD relevance: Vascular or insulin resistance can reduce the effectiveness of this buffer via impaired substrate delivery and astrocytic metabolism, increasing the metabolic cost of inflammation and synaptic failure.

GLUT1 (Endothelial Glucose Transporter)

Gatekeeper for brain glucose entry at the BBB.

AD relevance: Reduced GLUT1 is observed in AD microvessels and, in models, can precede hypometabolism and thereby amplify glial priming.

Lipid Droplets (LDs)

Neutral-lipid organelles that sequester fatty acids and oxidized lipids; hubs for innate signaling.

AD relevance: Lipid-droplet-accumulating microglia (LDAM) show impaired phagocytosis, excess ROS, and high inflammasome competence, especially in APOE4 contexts.

ApoE Lipidation/ABCA1–ABCG1

Cholesterol-efflux machinery that lipidates ApoE particles for neuronal and glial trafficking.

AD relevance: Poor lipidation (ApoE4-biased) disturbs membrane microdomains, APP processing, and glial lipid stress, feeding neuroinflammation.

Autophagy/Mitophagy

Lysosome-directed clearance of proteins and organelles (including damaged mitochondria).

AD relevance: mTORC1 overactivity and metabolic stress impede autophagy, limiting clearance of proteopathic species and damaged mitochondria in neurons and glia.

β-Hydroxybutyrate (Ketone Body)

Alternative fuel and signaling metabolite (HDAC inhibition; NLRP3 restraint).

AD relevance: Supports brain energetics during hypometabolism and can lower inflammasome tone.

Box 2 | Outstanding Questions: Testable Predictions, Risks, and Falsifiers

Is the Immune-Metabolic Axis the Primary Driver or a Late Amplifier of AD Pathology?

Prediction: Early microglial metabolic rewiring (mTOR↑, AMPK↓, glycolysis bias) precedes overt amyloid and tau deposition.

Test: Longitudinal ^13 C-glucose or ^13 C-acetate fluxomics and single-cell metabolite imaging across Braak stages, aligned with amyloid/tau PET timelines.

Falsifier: Metabolic inflection detected only after plaque or tangle burden stabilizes, indicating immunometabolism is secondary.

Can Restoring Mitochondrial Flexibility (AMPK–PGC1α–OXPHOS Axis) Reset Microglial Tone In vivo?

Prediction: AMPK agonists, mTORC1 inhibitors, or metabolic training (ketones, exercise) reprogram microglia toward oxidative, reparative states and enhance Aβ/tau clearance.

Test: Combine microglial PET (TSPO or CSF1R) with metabolic-flux imaging (FDG or ^11 C-acetate PET) and behavioral outcomes in mouse-to-human translation.

Falsifier: Energetic rebalancing achieves target engagement but fails to alter cytokine or clearance profiles.

Does Peripheral Dysmetabolism “Train” Brain Innate Immunity Through Circulating metabolites?

Prediction: Obesity, diabetes, or hyperlipidemia imprint pro-inflammatory chromatin marks (H3K4me3, H3K27ac) in microglia through succinate, itaconate, and cholesterol signaling.

Test: Compare microglial epigenomes in metabolic-syndrome versus lean cohorts and validate causality with bone-marrow chimeras.

Falsifier: Absence of metabolic-memory marks or no phenotype transfer after marrow replacement.

Are Extracellular Vesicles Metabolic Messengers or Mere Debris?

Prediction: Glia-derived EVs carry bioactive lipids, metabolic enzymes, and small RNAs that modulate neuronal redox and astrocytic AMPK signaling.

Test: Perform stable-isotope tracing in EV fractions and selectively block EV biogenesis or receptor uptake in vivo.

Falsifier: EV depletion or receptor blockade fails to change metabolic or inflammatory readouts.

Is Trained Immunity Reversible in the Aging Brain?

Prediction: Epigenetic or metabolic-reset interventions (such as NAD⁺ boosters or fasting-mimicking diets) can erase maladaptive microglial training and restore homeostatic surveillance.

Test: Longitudinal single-cell ATAC-seq with metabolic-flux profiling after intervention in aged mice or early-AD subjects.

Falsifier: Persistence of enhancer marks and cytokine memory despite metabolic restoration.

Can Multi-modal Metabolic Imaging Serve as a Clinical Biomarker?

Prediction: Combined metabolic-flux MRI/MRS with labeled substrates and microglial PET (TSPO or CSF1R) predicts cognitive decline better than amyloid or tau PET alone.

Test: Conduct prospective multi-tracer PET/MR studies with concurrent plasma and CSF metabolomics validation.

Falsifier: No added predictive value beyond standard amyloid and tau metrics.

What are the Risks of Metabolic Intervention?

Altering mTOR, AMPK, or inflammasome activity may compromise immune defense or induce maladaptive glial tolerance, while TSPO- or EV-targeting strategies could disrupt astro-neurovascular signaling or lipid clearance.

Abbreviations

Aβ

Amyloid-β

ABCA1

ATP-binding cassette transporter A1

ABCG1

ATP-binding cassette transporter G1

ACOD1 (IRG1)

Aconitate decarboxylase 1 (Immune-responsive gene 1)

AD

Alzheimer’s disease

AMPK

AMP-activated protein kinase

APP

Amyloid precursor protein

ASC

Apoptosis-associated speck-like protein containing a CARD

ASL-MRI

Arterial spin-labeling magnetic resonance imaging

AT(N)

Amyloid/tau/neurodegeneration biomarker framework

ATAC-seq

Assay for transposase-accessible chromatin using sequencing

ApoE

Apolipoprotein E

APOE4

Apolipoprotein E ε4 allele

BBB

Blood–brain barrier

BCAA

Branched-chain amino acid

BDNF

Brain-derived neurotrophic factor

cAMP

Cyclic adenosine monophosphate

cGAS–STING

Cyclic GMP-AMP synthase–stimulator of interferon genes pathway

CNS

Central nervous system

CSF

Cerebrospinal fluid

CSF1R

Colony-stimulating factor 1 receptor

CSVD

Cerebral small-vessel disease

EC

Endothelial cell

EV

Extracellular vesicle

FAO

Fatty-acid oxidation

FDG-PET

2-[^18F]fluoro-2-deoxy-D-glucose positron emission tomography

GLP-1RA

Glucagon-like peptide-1 receptor agonist

GLUT1

Glucose transporter type 1

HDAC

Histone deacetylase

HIF-1α

Hypoxia-inducible factor-1α

iPSC

Induced pluripotent stem cell

IL-1β

Interleukin-1β

KEAP1

Kelch-like ECH-associated protein 1

kMCT

Ketogenic medium-chain triglyceride

LD

Lipid droplet

LDAM

Lipid-droplet-accumulating microglia

LDLR

Low-density lipoprotein receptor

LXR/RXR

Liver X/retinoid X receptor

MCT1/2

Monocarboxylate transporter 1/2

MRI

Magnetic resonance imaging

MRS

Magnetic resonance spectroscopy

mTOR

Mechanistic target of rapamycin

mTORC1/2

mTOR complex 1 / complex 2

NAD⁺

Nicotinamide adenine dinucleotide (oxidized form)

NF-κB

Nuclear factor κB

NIA-AA

National Institute on Aging–Alzheimer’s Association

NLRP3

NOD-, LRR- and pyrin-domain-containing protein 3 (inflammasome)

NRF2

Nuclear factor erythroid 2–related factor 2

NVU

Neurovascular unit

OXPHOS

Oxidative phosphorylation

PET

Positron emission tomography

PGC-1α

Peroxisome proliferator-activated receptor-γ coactivator 1-α

PPAR

Peroxisome proliferator-activated receptor

P2RY12

Purinergic receptor P2Y12

ROS

Reactive oxygen species

S1P

Sphingosine-1-phosphate

SGLT2

Sodium–glucose cotransporter 2

SIRT1

Sirtuin 1

SREBP2

Sterol regulatory element-binding protein 2

TCA

Tricarboxylic acid (cycle)

TLR4

Toll-like receptor 4

TNF

Tumor necrosis factor

TSPO

18-kDa translocator protein

Author Contributions

D.-H.B. and T.L.N wrote the manuscript.

Funding

No funding for this project.

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Competing Interests

The authors declare no competing interests.

Ethical Approval

Not applicable.

Consent to Participate

This review synthesizes previously published studies and did not involve new studies with human participants or animals.

Consent for Publication

Not applicable.

Disclosure

D.-H.B utilized ChatGPT and Gemini to assist in the structural outlining of the manuscript and to generate initial summaries of selected literature. D.-H.B manually selected all reference materials, verified the accuracy of AI-generated summaries against the original texts, and revised the final manuscript to ensure intellectual integrity. The author(s) take full responsibility for the final content.