Cognitive demand as a neurometabolic stress test in diabetes-associated cognitive impairment: from static metabolite pools to dynamic energy responses.
Frontiers in endocrinology · 2026
Abstract originale
BACKGROUND: Diabetes-associated cognitive impairment (DCI) is heterogeneous and cannot be explained by chronic hyperglycaemia or isolated resting metabolic abnormalities. Measurements of transporter abundance, metabolite pools, perfusion, or functional endpoints alone do not determine whether the brain can dynamically recruit energetic support when cognitive demand increases. OBJECTIVE: To conceptualize cognitive demand as a neurometabolic stress test, distinguish static metabolic states from dynamic adaptive responses, and establish a falsifiable framework for activity-dependent neurometabolic failure in DCI. METHODS: This concept-driven critical review used an iterative PubMed/MEDLINE-centred search updated through August 2026, supplemented by backward and forward citation tracking. Evidence was organized along the demand-to-function pathway, encompassing cognitive-task, glucose/insulin, hypoglycaemic or alternative-substrate, and neural/network challenges. Studies were prioritized when they provided information on temporal dynamics, pathway flux, cellular or regional specificity, or causal perturbation. RESULTS: Normal cognitive function requires coordinated recruitment of vascular delivery, substrate mobilization and allocation, oxidative metabolism, phosphotransfer buffering, and ATP turnover, even when metabolite and high-energy phosphate pools remain relatively stable. Evidence from diabetes and insulin resistance demonstrates heterogeneous, challenge-dependent abnormalities that may involve reduced response amplitude, temporal misalignment, altered substrate distribution, or impaired metabolic utilization rather than uniform cerebral hypometabolism. We propose four non-exclusive within-challenge phenotypes of neurometabolic failure: blunted recruitment, mistimed recruitment, substrate misallocation, and malutilization. Neurometabolic reserve denotes recruitable adaptive capacity; its exhaustion is considered a longitudinal contraction of that capacity, not a fifth within-challenge phenotype or a conclusion warranted by a single attenuated response. However, current evidence has not yet established a complete causal chain linking cognitive demand, substrate fate, energetic support, neural computation, and cognitive performance in clinically characterized DCI. CONCLUSIONS: We propose cognitive demand as an experimental neurometabolic stress test for investigating metabolic vulnerabilities that may remain undetected at rest, rather than as a validated clinical diagnostic test. Activity-dependent neurometabolic failure should therefore be considered a testable contributor rather than an established universal mechanism of DCI. Future studies integrating challenge-locked, temporally aligned, and compartment-resolved metabolic measurements with causal perturbation and longitudinal assessment are required to define when endocrine-metabolic dysfunction translates into cognitive vulnerability.
Parole chiave
Leggi il testo integrale originale
Testo estratto dalla pubblicazione: sono esclusi immagini, media, supplementi e parti soggette a diritti distinti.
Cognitive demand as a neurometabolic stress test in diabetes-associated cognitive impairment: from static metabolite pools to dynamic energy responses
Background
Diabetes-associated cognitive impairment (DCI) is heterogeneous and cannot be explained by chronic hyperglycaemia or isolated resting metabolic abnormalities. Measurements of transporter abundance, metabolite pools, perfusion, or functional endpoints alone do not determine whether the brain can dynamically recruit energetic support when cognitive demand increases.
Objective
To conceptualize cognitive demand as a neurometabolic stress test, distinguish static metabolic states from dynamic adaptive responses, and establish a falsifiable framework for activity-dependent neurometabolic failure in DCI.
Methods
This concept-driven critical review used an iterative PubMed/MEDLINE-centred search updated through August 2026, supplemented by backward and forward citation tracking. Evidence was organized along the demand-to-function pathway, encompassing cognitive-task, glucose/insulin, hypoglycaemic or alternative-substrate, and neural/network challenges. Studies were prioritized when they provided information on temporal dynamics, pathway flux, cellular or regional specificity, or causal perturbation.
Results
Normal cognitive function requires coordinated recruitment of vascular delivery, substrate mobilization and allocation, oxidative metabolism, phosphotransfer buffering, and ATP turnover, even when metabolite and high-energy phosphate pools remain relatively stable. Evidence from diabetes and insulin resistance demonstrates heterogeneous, challenge-dependent abnormalities that may involve reduced response amplitude, temporal misalignment, altered substrate distribution, or impaired metabolic utilization rather than uniform cerebral hypometabolism. We propose four non-exclusive within-challenge phenotypes of neurometabolic failure: blunted recruitment, mistimed recruitment, substrate misallocation, and malutilization. Neurometabolic reserve denotes recruitable adaptive capacity; its exhaustion is considered a longitudinal contraction of that capacity, not a fifth within-challenge phenotype or a conclusion warranted by a single attenuated response. However, current evidence has not yet established a complete causal chain linking cognitive demand, substrate fate, energetic support, neural computation, and cognitive performance in clinically characterized DCI.
Conclusions
We propose cognitive demand as an experimental neurometabolic stress test for investigating metabolic vulnerabilities that may remain undetected at rest, rather than as a validated clinical diagnostic test. Activity-dependent neurometabolic failure should therefore be considered a testable contributor rather than an established universal mechanism of DCI. Future studies integrating challenge-locked, temporally aligned, and compartment-resolved metabolic measurements with causal perturbation and longitudinal assessment are required to define when endocrine-metabolic dysfunction translates into cognitive vulnerability.
Graphical Abstract
Infographic explains brain metabolism during cognitive stress, showing a static resting snapshot with key substrates—glucose, lactate, glycogen, ATP, and transporters—then illustrating astrocyte-neuron interactions under challenge with neural, vascular, and metabolic processes. Right panel depicts four dynamic failure patterns: blunted response, mistimed response, misallocated fuel, and malutilized substrate. Bottom icons indicate consequences including impaired synaptic plasticity, abnormal oscillations, and memory vulnerability. Color-coded graphics and callouts clarify each step.
1Introduction
Diabetes is associated with a continuum of cognitive outcomes, ranging from subtle, domain-specific decrements to mild cognitive impairment and dementia. Across prospective studies, diabetes is consistently associated with a higher risk of cognitive impairment and dementia, although the magnitude of that association varies with diabetes type, age, disease duration, comorbidity and outcome definition (1, 2). Longitudinal cohort data further indicate that diabetes in midlife is associated with faster cognitive decline over subsequent decades (3). The phenotype is not confined to memory: executive function, processing speed, attention and psychomotor performance can also be affected, with generally small-to-moderate effects at the group level (4). Accordingly, diabetes-associated cognitive impairment (DCI) is best used as an umbrella term rather than a uniform diagnosis. Its clinical expression probably reflects different combinations of vascular injury, neurodegenerative pathology, metabolic disturbance and reduced cognitive reserve (5, 6). Glycaemic indices are relevant, but they explain only a limited proportion of the observed cognitive variance (7). The central mechanistic problem is therefore not simply whether the diabetic brain has been exposed to abnormal glucose, but why particular circuits become unable to sustain cognition.
Current mechanistic accounts have identified important contributors to DCI, including brain insulin resistance, microvascular dysfunction, blood-brain barrier injury, glial reactivity, neuroinflammation and mitochondrial stress (5, 6, 8). Much of this evidence, however, is derived from measurements obtained at rest or from isolated molecular and biochemical endpoints. Transporter abundance and enzyme expression describe potential capacity; concentrations of glucose, lactate, glycogen, ATP or phosphocreatine describe the size of a pool at a particular time. Neither, by itself, establishes how rapidly substrate is delivered, which cells acquire it, where the carbon is routed, how much is oxidized, or whether ATP production increases when neural work rises. A larger pool may reflect increased production, impaired clearance, reduced uptake or diminished downstream utilization. This ambiguity is not theoretical: during hypoglycaemia, individuals with type 1 diabetes showed markedly increased brain lactate without a corresponding increase in lactate-derived oxidative labelling (9). Static abundance and realized pathway use can therefore diverge. If metabolic vulnerability is expressed mainly when a circuit is challenged, resting snapshots may obscure the biologically relevant phenotype.
Cognitive activity provides a physiologically meaningful perturbation with which to expose that phenotype. In rodents, a hippocampus-dependent spatial task produces a rapid, regionally specific fall in extracellular glucose that scales with task complexity (10). At presynaptic terminals, electrical activity recruits both glycolytic and mitochondrial ATP synthesis, and even brief interruption of this recruitment disrupts vesicle recycling despite a substantial resting ATP pool (11). Learning also mobilizes astrocytic glycogen and lactate transport, while the relative contribution of glucose and lactate varies with the computational and attentional load imposed by the task (12, 13). These cellular responses are embedded within activity-dependent vascular recruitment and rapid neuronal glycolysis (14, 15). Moreover, high-energy phosphate concentrations can remain comparatively stable while ATP-synthesis-related flux changes markedly across levels of brain activity (16). Cognitive demand can therefore be treated as an experimental neurometabolic stress test-not as a validated clinical diagnostic test, but as a controlled means of probing metabolic reserve. In this framework, activity-dependent neurometabolic adaptation denotes the coordinated recruitment of vascular delivery, cellular fuel mobilization and routing, oxidation and phosphotransfer buffering needed to preserve synaptic, network and cognitive function as demand rises.
A challenge-centred view is particularly relevant to diabetes because several studies suggest that basal adequacy can coexist with an abnormal evoked response. In healthy adults, acute lipid infusion used to induce insulin resistance altered the hippocampal phosphocreatine-to-ATP response during sustained cognitive activity (17). This study provides a human proof of principle, but its small sample, acute intervention and ratio-based endpoint do not establish chronic DCI, glucose flux or neuronal ATP deficiency. In rats, high-fat-diet-induced insulin resistance attenuated both the cognitive and local metabolic responses to hippocampal insulin (18). Diet-induced diabetes was also associated with lower hippocampal glycogen and weaker glycogen mobilization after local insulin administration (19). At the vascular level, a mouse model of type 1 diabetes showed delayed stimulus-evoked neurovascular recruitment despite comparatively preserved resting flow and peak responses (20). These observations identify distinct challenge-sensitive abnormalities, but they do not yet form a single causal pathway. They differ in species, diabetes type, disease stage, brain region and experimental challenge; vascular kinetics do not measure substrate flux, and a change in glycogen or high-energy phosphate balance does not by itself establish functional energy failure. No study has yet closed the full chain from cognitive demand to cellular carbon flux, energetic state, neural computation and cognitive performance in clinically characterized DCI.
This Critical Review advances the bounded proposition that impaired activity-dependent neurometabolic adaptation is a testable contributor to DCI. It distinguishes molecular capacity and metabolite pools from task-evoked proxies, pathway flux, energetic state and functional sufficiency. Cognitive, glucose-insulin, hypoglycaemic-substrate and neural-network perturbations are treated as challenge paradigms, not failure phenotypes. The four non-exclusive phenotypes describe how inadequate support may be expressed within a challenge: blunted recruitment, mistimed recruitment, substrate misallocation and malutilization. They are not proposed as new disease mechanisms or discrete DCI subtypes. Their added value is a cross-mechanism, response-based layer of interpretation. Established accounts, including neurovascular dysfunction, brain insulin resistance and mitochondrial impairment, identify candidate lesion classes; the present framework asks how a candidate lesion manifests under demand, at which measured node the abnormality is directly observed, and what additional observation would distinguish competing explanations. It thereby converts broad mechanistic labels into node-specific, falsifiable predictions about response magnitude, timing, carbon allocation and downstream utilisation. Practically, the framework is intended to prevent overinterpretation of a single readout and to specify the next discriminating measurement required to distinguish competing explanations. Reserve exhaustion remains separate and denotes longitudinal contraction of the recruitable response range rather than a fifth acute phenotype. The framework does not assume universal cerebral energy deficiency or one obligatory substrate pathway.
Two examples illustrate how this framework changes interpretation and experimental design. First, a peak-focused analysis can miss temporal dysfunction. In diabetic mice, peak and integrated neurovascular responses can remain comparatively preserved despite delayed recruitment (20), while human studies in type 2 diabetes show altered response rise, peak timing and recovery (21, 22). The framework therefore predicts that a normal or near-normal peak does not exclude a candidate mistimed phenotype and changes the experimental decision from comparing peak amplitude alone to analysing the full response curve in relation to neural timing, using response models flexible enough to capture latency and recovery differences (23). Second, higher insulin-stimulated cerebral FDG uptake reported in insulin-resistant states (24–30) should not be equated with greater oxidative metabolism or successful compensation. FDG primarily reflects delivery, transport and phosphorylation-dependent trapping and does not establish downstream oxidation or ATP support (31–36). Kinetic modelling further suggests that an increased FDG signal can reflect altered inward glucose transfer rather than a proportional increase in downstream metabolism (37). The framework therefore changes the next experimental decision: rather than labelling greater FDG uptake as compensation, the subsequent measurement should target the downstream step required by the specific claim, such as oxidative metabolism or energetic support aligned with neural or cognitive function.
2Literature search and review approach
This concept-driven critical review used iterative PubMed searches, supplemented by backward and forward citation tracking, to identify evidence on activity-dependent neurometabolic adaptation in DCI. Searches combined Medical Subject Headings and title/abstract terms for diabetes or insulin resistance, cognition and hippocampal function, neural activity, neurovascular coupling, substrate metabolism and transport, mitochondrial function, and energetic responses. Complementary searches addressed activity-related diabetes studies, dynamic metabolic measurements, network energetics, human task imaging, substrate transport and reserve capacity, neurovascular and blood-brain barrier function, metabolic interventions, and related reviews. Full search strategies, retrieval counts and evidence-selection criteria are provided in Supplementary Methods S1.
Studies were selected for their relevance to the relationships among neural demand, vascular recruitment, substrate delivery and allocation, oxidative metabolism, energetic state, and neural or cognitive function. Priority was given to primary research with clearly described populations or models, brain regions, challenges and measured outcomes, particularly studies providing temporal, pathway-specific, cellular or regional information, or mechanistic perturbation. Countervailing findings and potentially compensatory responses were retained. Physiological and methodological studies established reference mechanisms and measurement limitations; findings outside clinically characterised DCI were interpreted according to their population and experimental context. Static molecular, metabolite-pool or ex vivo capacity measurements informed candidate mechanisms but were not treated as direct evidence of task-evoked metabolic failure.
Evidence was synthesised qualitatively rather than through exhaustive systematic screening. Retrieval limits were applied to three broad searches, which may have restricted literature coverage. Representative primary studies were further evaluated according to their clinical relevance, measurement directness and mechanistic support, with explicit consideration of whether cognitive impairment was characterised in participants with diabetes.
3Cognitive demand as a neurometabolic stress test: the physiological basis
Cognitive demand creates a rapid and spatially localized energetic burden that must be matched within the same challenge window by neurovascular recruitment, fuel mobilization and allocation, oxidative and phosphotransfer responses, and increased ATP turnover (Figure 1). Physiological adequacy is defined not by depletion of static metabolite or ATP pools, but by whether energetic support is recruited with sufficient amplitude, timing and spatial precision to preserve synaptic, network and memory function.
Operationally, we use “cognitive neurometabolic stress test” to denote a conceptual and experimental framework, not a validated clinical diagnostic test. Within this framework, a controlled, preferably graded cognitive challenge is paired with measurement of a prespecified vascular or metabolic response from baseline through task performance and recovery. The response should be interpreted alongside achieved performance, systemic physiology and, where feasible, independently quantified neural engagement. The relevant comparison is the response to effective neural demand rather than task exposure alone. The cognitive domain, target circuit, manipulated load and sampling window must therefore be specified. Nominal difficulty does not constitute an equivalent energetic dose across tasks or individuals.
3.1Cognitive demand creates a rapid and spatially localized energy burden
Cognitive demand increases ATP-consuming electrical work within task-engaged circuits. Its magnitude and location depend on which circuit elements carry the computation. Action potentials and postsynaptic currents dissipate Na+ and K+ gradients that the Na+/K+-ATPase must restore. Quantitative energy budgets place much of signalling expenditure in postsynaptic conductances and spike propagation, although estimates depend on circuit architecture and modelling assumptions (38, 39). Direct measurements show that action-potential cost varies with neuronal type, membrane state and subcellular location (40, 41). Energetic burden therefore means increased work for local supply and buffering systems, not prior ATP depletion.
Synaptic terminals concentrate this demand. Sustained firing recruits local ATP synthesis, and inhibiting glycolytic or mitochondrial production lowers presynaptic ATP and disrupts vesicle recycling (11). Oxidative ATP is particularly important for endocytosis, although ATP can diffuse along axons to boutons lacking resident mitochondria (42). Vesicular proton leak imposes a substantial resting maintenance cost (43). Mitochondrial pyruvate transport also helps determine presynaptic ATP availability and vesicle-cycle timing (44). The relevant reserve therefore encompasses ongoing maintenance as well as evoked production. These experiments establish a local energetic bottleneck under imposed activity, although they do not directly quantify cognitive-task metabolism in vivo.
Excitatory transmission distributes work beyond the firing neuron. Na+-coupled glutamate uptake increases ionic pumping and glycolytic demand in astrocytes (45). Neuronal stimulation can simultaneously accelerate neuronal glucose consumption and glycolysis (15). This distributed workload supports transmitter clearance and ionic homeostasis around active synapses. Active circuits therefore recruit parallel cellular responses whose relative contributions vary with preparation, stimulus and time scale. Opposing local changes may be obscured in bulk tissue averages.
Memory formation adds biosynthetic demand. Protein synthesis and structural remodelling extend that demand beyond the initial electrical event. Stable dendritic mitochondrial compartments support local translation and spine growth, and their disruption impairs plasticity within nearby dendritic segments (46). Learning also depends on hippocampal monocarboxylate metabolism, protein synthesis and persistent plasticity in intervention-rescue models (12, 47). These findings locate the relevant metabolic burden near modified synapses without establishing a single obligatory route of substrate transfer.
Cognitive tasks show that this burden is graded and regionally restricted. Task difficulty can therefore operate as a graded metabolic challenge. During maze learning, extracellular hippocampal glucose declined more during the complex spatial condition and remained unchanged at a non-hippocampal control site (10). The concentration change reflects delivery and removal jointly, so it cannot identify uptake or oxidation. Chronic neuronal loss of glucose uptake or glycolysis also compromises presynaptic ATP and age-dependent learning and memory (48). Developmental compensation, however, limits task-locked causal interpretation.
At the systems level, isotope tracing links graded glutamatergic activity to oxidative glucose metabolism (49). Magnetization-transfer studies likewise show activity-dependent changes in cerebral ATP-synthesis-related flux despite comparatively stable high-energy-phosphate concentrations (16). Although neither experiment used a cognitive task and both require model-based interpretation, they define the physiological problem. Demand must recruit delivery, fuel handling and ATP turnover at the relevant time and place without necessarily depleting ATP pools. The defining feature is dynamic recruitment, rather than a large change in stored energy.
3.2Neurovascular recruitment matches substrate and oxygen delivery to neural activity
Neurovascular coupling (NVC) converts neural activity into regional vascular responses, principally functional hyperaemia (14, 50). By increasing local perfusion, it improves access to oxygen and blood-borne substrates within task-engaged tissue. This distributed, time-varying response changes exchange conditions at the neurovascular interface. It thereby establishes supply capacity before cells select and metabolise fuels. It does not imply fixed proportionality among cerebral blood flow (CBF), substrate transport, oxygen consumption and ATP production.
Functional hyperaemia arises from convergent cellular signals rather than a single neuron-astrocyte-vessel relay. Excitatory, inhibitory and nitric oxide synthase-expressing neurons produce distinct vascular effects (51, 52). Sensory dilation can occur without canonical IP3R2-dependent astrocytic calcium signalling (53). Fine-process astrocytic calcium can nevertheless precede olfactory hyperaemia (54). Calcium clamping preferentially attenuates the late response to prolonged stimulation (55). Astrocytic effects may also differ between capillaries and arterioles (56). Their contribution is therefore vessel-, region- and duration-dependent. This heterogeneity becomes especially important as the duration of cognitive demand changes.
The vascular tree further distributes recruitment across endothelial and mural-cell mechanisms. Capillary endothelial K+ sensing can propagate hyperpolarisation towards upstream arterioles (57). Endothelial disruption interrupts stimulus-evoked dilation (58). Arteriolar endothelial caveolae support dilation and capillary red-cell acceleration in awake mice (59). Reports differ on whether early capillary dilation reflects pericyte relaxation (60). Vessel-zone mapping instead assigns the fastest diameter changes to arterioles and ensheathing mural cells (61). Thin-strand pericytes can alter local resistance more slowly (62). Whole-arbor imaging confirms segment-specific diameter and velocity responses (63). Fast upstream control can therefore coexist with slower capillary resistance adjustments. Functional hyperaemia is a coordinated vascular-tree response rather than a unitary event.
Flow recruitment improves delivery but does not specify metabolic use. Somatosensory activation can increase CBF much more than cerebral oxygen consumption (64). Visual stimulation similarly produces larger flow and glucose-utilisation responses than oxygen-metabolic responses (65). Across graded visual stimuli, CBF and CMRO2 may covary linearly, although fractional flow changes remain larger and paradigm-specific (66). CBF is accordingly a supply-side readout rather than a direct measure of use. Delivery, extraction, transport, uptake and oxidation are consecutive but non-equivalent processes.
Timing is equally important. Haemodynamic responses begin within the parenchyma and propagate along selected arterioles with compartment-specific onset and recovery (67). Much vessel-resolved evidence comes from sensory preparations, leaving hippocampal cognitive recruitment less defined. BOLD responses track local field potentials more closely than spiking in visual cortex, with site-specific transfer functions (68). BOLD therefore reports vascular and oxygenation consequences of local processing rather than neural activity or metabolic flux directly (69). Temporal alignment must be evaluated against the neural event that generated the response.
Neurovascular recruitment is thus necessary but insufficient for matching demand. Preserved CBF or BOLD may coexist with impaired barrier transport, cellular allocation, oxidation or ATP support. Modest haemodynamic changes may instead be offset by extraction or local buffering. A normal peak alone does not establish timely supply. Adequacy depends on amplitude, latency, duration and spatial reach within the relevant activity window. The next question is how delivered substrates are allocated to active cells and compartments.
3.3Cellular and subcellular fuel mobilization supports memory processing
Once substrates reach the neurovascular interface, allocation becomes decisive. Delivery alone does not determine cellular uptake, glycogen mobilisation or carbon transfer to active boutons, dendrites and mitochondria. Current evidence supports concurrent neuronal glucose use, astrocytic carbohydrate metabolism and monocarboxylate exchange, with contributions varying by region, workload and memory phase (70). Fuel mobilisation is therefore a kinetic and compartmental response rather than a change in bulk metabolite abundance.
Neurons can rapidly recruit glucose. Stimulation increases neuronal uptake of a 2-deoxyglucose analogue in awake mice (71). Genetically encoded sensors identify an early neuronal glycolytic response that does not require lactate uptake within the sampled interval (15). Conditional loss of neuronal GLUT3 or PKM1 impairs hippocampal energetic state and learning despite metabolic remodelling (48). At boutons, AMPK-dependent GLUT4 insertion sustains vesicle recycling during prolonged firing (72). These models establish direct neuronal recruitment without excluding parallel astrocytic or later monocarboxylate support.
Astrocytes provide complementary support while meeting their own activity-dependent workload. Glutamate uptake stimulates astrocytic glycolysis and lactate production (45). Hippocampal activation recruits glycolytic and oxidative responses in both astrocytes and neurons (73). Astroglial networks can redistribute glucose-related metabolites when extracellular glucose is limited (74). Glycogen supplies a rapid reserve. Blocking glycogenolysis during memory tasks impairs performance that an oxidisable substrate can rescue (75). Disrupting glycogen breakdown or monocarboxylate transport also impairs long-term memory and persistent potentiation (12). Deleting brain glycogen synthase weakens learning-associated CA3-CA1 plasticity (76). These interventions establish functional importance without identifying the final recipient or oxidative fate of glycogen-derived carbon.
Lactate is therefore a conditional fuel and signalling route, not an obligatory endpoint. Cell-resolved sensing reveals an in vivo gradient compatible with astrocyte-to-neuron transport (77). Dynamic 13C magnetic resonance spectroscopy shows that circulating lactate can contribute carbon to human brain oxidation when plasma availability rises (78). Lactate also modifies neuronal redox state and NMDA receptor-dependent plasticity-gene expression (79). Lactate, pyruvate and beta-hydroxybutyrate can each rescue selected translation or memory deficits (47). Such flexibility, together with bidirectional monocarboxylate transport, prevents transporter abundance from specifying carbon-flow direction.
Allocation extends below the cellular level. Activity-dependent GLUT4 trafficking positions glucose entry at presynaptic boutons (72). Stable dendritic mitochondrial domains support translation and spine enlargement within restricted segments (46). Their local disruption can impair nearby plasticity despite apparently preserved cell-wide metabolism. Transporter localisation and mitochondrial proximity therefore define potential access to fuel and ATP support, but neither alone identifies the oxidised substrate or the energy supplied to an engaged synapse.
Activity-dependent fuel mobilisation is organised by substrate source, recipient, timing and metabolic fate. Neuronal glucose use and astrocytic glycogen-monocarboxylate support may contribute concurrently and shift across acquisition, consolidation and sustained network activity. Assigning carbon flow to specific cells and subcellular domains in the behaving brain remains difficult (80). Mobilisation is physiologically adequate only when routed carbon supports the ATP response required by the relevant computation.
3.4Dynamic ATP production preserves energetic and functional homeostasis
During neural activity, energetic homeostasis can be preserved without increasing the ATP pool. In vivo magnetization transfer shows that ATP-synthesis-related and creatine-kinase exchange fluxes change while high-energy-phosphate concentrations remain comparatively stable (16). At boutons, stimulation recruits glycolytic and mitochondrial ATP production that limits presynaptic ATP decline (11). Cytosolic ATP and ADP may remain stable across physiological workloads (81). Larger sodium loads can instead produce delayed somatodendritic ATP falls (82). Physiological-like firing may also cause small decreases at specialised terminals (83). These observations define a load-, time- and compartment-dependent energetic margin.
ATP support draws on coordinated pathways rather than a fixed substrate hierarchy. White-matter ATP dynamics track electrical conduction and local metabolic support (84). Oxidative phosphorylation can rapidly power hippocampal presynaptic and postsynaptic processes (85). Glycolysis or respiration may temporarily sustain basal transmission, whereas recovery from intense vesicle depletion depends more strongly on oxidative metabolism (86). Neurons also accelerate glycolysis during stimulation (15). Mitochondrial calcium uptake promotes NADH generation and tricarboxylic-acid-cycle recruitment (87). MICU3 increases axonal mitochondrial calcium sensitivity, enabling activity-dependent ATP synthesis from oxidative fuels (88). Pathway contributions consequently vary by workload, substrate environment and compartment.
Buffering and spatial organisation extend this margin. Axonal ATP diffusion can support boutons without resident mitochondria, although endocytosis becomes vulnerable when oxidative supply is constrained (42). Brain-type creatine-kinase deletion preserves ATP and phosphocreatine concentrations but reduces phosphocreatine-ATP exchange and slows spatial learning (89). Syntaphilin-dependent mitochondrial docking regulates presynaptic calcium dynamics and short-term facilitation (90). Stable dendritic mitochondrial domains support local translation and spine plasticity (46). These effects cannot be assigned solely to ATP because mitochondria also regulate calcium and redox signalling.
Functional impairment may emerge before global ATP collapse. Local interruption of ATP production rapidly disrupts vesicle endocytosis, ion-gradient restoration and sustained transmitter cycling (11, 42). Long-term potentiation recruits dendritic mitochondrial fission and matrix calcium, and blocking this response impairs structural and electrophysiological plasticity (91). Hippocampal gamma oscillations require strong complex-I performance (92). Oxygen use differs among gamma activity, sharp-wave ripples and lower-demand states (93). Reducing the mitochondrial calcium uniporter disrupts gamma synchrony, sharp-wave ripples and activity-dependent pyruvate-dehydrogenase regulation (94). Together, these studies link mitochondrial recruitment to demanding network states without attributing every functional change solely to ATP loss.
At larger scales, increased ATP-synthesis-related flux can coexist with buffered ATP and phosphocreatine pools (16). A local ATP decline instead indicates that consumption has temporarily exceeded synthesis, redistribution or buffering (82, 83). Neither pattern is inherently pathological. Interpretation depends on challenge intensity, duration, compartment and sampling time. Redox, oxygen-consumption, phosphocreatine-exchange and ATP signals describe different response components, while preserved output may reflect compensation or a lower effective workload.
Normal neurometabolic adaptation therefore requires glycolytic, mitochondrial and phosphotransfer processes to increase turnover with sufficient amplitude, timing and spatial precision to keep the active circuit within its energetic operating range. Adequacy is established by preservation of the relevant synaptic or network function, not by a single ATP concentration or substrate measurement. This demand-matched response provides the physiological benchmark against which diabetes-related abnormalities should be judged.
4From static pools to dynamic responses: a measurement-to-inference framework
Neurometabolic evidence spans molecular capacity, metabolite pools, task-evoked proxies, tracer- or model-derived rates, energetic state and functional outcomes (Figure 2). These domains are complementary but not interchangeable. Table 1 summarises what each major readout measures or estimates, what it cannot establish alone and its principal interpretive limitations. Cross-level claims require measurements aligned in time, compartment and challenge context.
Table 1What common neurometabolic measurements establish and their principal interpretive limitations.
Measurement/readout What it measures or estimates What it does not establish on its own Principal interpretive limitations
Molecular abundance and ex vivo enzyme capacity Transporter, enzyme, transcript or protein abundance indicates potential molecular capacity, while ex vivo activity assays define capacity under the imposed assay conditions (80). These measurements do not establish realised in vivo transport, pathway throughput, task-evoked flux or adequate ATP support (80, 95, 96). Interpretation depends on cellular and membrane localisation, substrate and cofactor supply, saturation, regulation, compartmentation and assay conditions (80, 95, 96).
Metabolite concentrations and pool dynamics(tissue assays, microdialysis, ¹H MRS) They quantify metabolite abundance or concentration in a sampled tissue, voxel or extracellular compartment; serial measurements describe changes in pool size over time (10, 80, 97, 98). Concentration alone does not separate delivery, production, consumption or export, and does not establish absolute uptake, oxidation or ATP production (80, 95, 96). The same pool size can arise from different turnover rates; spatial averaging and sampling interval can obscure compartment-specific or rapid changes (80, 95, 96).
FDG-PET/FDG-fPET FDG signals integrate tracer delivery, transport and hexokinase-dependent phosphorylation/trapping; with appropriate kinetic modelling, net influx or glucose utilisation can be estimated (31–33). FDG does not directly establish downstream glycolytic flux, TCA-cycle oxidation, ATP synthesis or neuron-specific glucose use (31, 34). Interpretation depends on the reported metric, input function, plasma glucose, lumped constant, acquisition window and temporal filtering; astroglial processes can contribute to the signal (32–36).
CBF/perfusion measurements(ASL, H2¹5O PET; TCD for arterial velocity) ASL and H2¹5O PET estimate regional perfusion, whereas transcranial Doppler measures arterial blood-flow velocity rather than tissue perfusion (64–66, 99, 100). Perfusion alone does not establish substrate extraction, transendothelial transport, cellular uptake, oxidation or adequate ATP support (64–66, 99). Flow can dissociate from glucose and oxygen metabolism, and interpretation is affected by vascular tone, systemic physiology and the measurement modality (64–66, 99, 100).
Conventional BOLD-fMRI BOLD reflects deoxyhaemoglobin-sensitive haemodynamic changes associated with local neural processing (68, 69). BOLD is not an independent quantitative measure of neural firing, CBF, CMRO2, glucose metabolism or ATP production (68, 69). Flow, blood volume, oxygen extraction, baseline vascular physiology and neurovascular coupling jointly shape BOLD amplitude and kinetics (68, 69, 101).
Stable-isotope tracing and dynamic ¹³C MRS These approaches track label incorporation and substrate contribution; with precursor input and kinetic modelling, specified pathway or compartmental rates can be estimated (95, 96, 102–106). Isotopic enrichment alone does not establish absolute net flux, direct cell-to-cell transfer or ATP output (95, 96, 102–107). Precursor enrichment, pool size, unlabelled dilution, reversible exchange, competing pathways, acquisition duration and model structure determine the inference (95, 96, 102–106).
Hyperpolarised ¹³C MRI/MRS Hyperpolarised tracers provide rapid regional substrate-conversion or label-exchange signals and can yield kinetic or redox-related information when the tracer and model permit (108). A conversion signal alone does not establish net glycolytic or oxidative flux (108). Delivery, transport, reversible exchange, redox state, polarisation decay and model assumptions contribute to the observed signal (108).
CMRO2(¹5O-gas PET or calibrated MR estimates) CMRO2 quantifies or estimates cerebral oxygen consumption within an explicit physiological model (109–112). CMRO2 does not identify the oxidised substrate or consuming cell type and does not directly measure total ATP production (110–112). Flow/oxygenation calibration and modelling remain part of the estimate; conversion to ATP output assumes coupling, an effective P/O ratio and limited non-mitochondrial oxygen use (110–112).
Respirometry and extracellular-flux assays These assays quantify oxygen-consumption and acidification responses and can derive ATP-linked respiration, proton leak, maximal respiration, spare capacity and corrected ATP-production estimates (113–115). They do not by themselves establish intact-brain ATP support during cognition or longitudinal neurometabolic reserve (113–115). Uncoupling, non-mitochondrial oxygen use, respiratory CO2, substrate conditions and stoichiometric corrections influence derived quantities; ex vivo capacity is context dependent (113–115).
High-energy-phosphate pools and ratios(ATP, PCr, Pi; PCr/ATP by ³¹P MRS) Conventional ³¹P MRS measures high-energy-phosphate abundance, intracellular pH and relative pool balance; PCr/ATP is a pool ratio (16, 17). Static pools or ratios do not establish ATP synthesis or turnover rate, preserved neurometabolic reserve, or adequate ATP support of active synapses (11, 16, 17). Buffering and spatial averaging can maintain bulk ATP or PCr despite altered local synthesis, consumption, redistribution or turnover (11, 16).
³¹P magnetisation transfer Magnetisation-transfer approaches provide model-dependent estimates of unidirectional phosphate-exchange fluxes within the ATP-PCr-Pi network (16, 116). The measured exchange does not necessarily equal net oxidative ATP synthesis unless the relevant biochemical and kinetic assumptions are validated (116, 117). Reversible near-equilibrium exchange and model assumptions can separate measured exchange from net ATP production; interpretation is tissue- and protocol-specific (116, 117).
ATP/ADP and redox biosensors Genetically encoded sensors can resolve local or subcellular ATP-to-ADP or NADH-to-NAD+ dynamics with high temporal resolution (118–120). These signals do not by themselves quantify absolute pathway flux or identify the provenance of metabolised carbon (118–120). Affinity, saturation, pH, expression, calibration and targeting affect signals that integrate production, consumption, transport and buffering (118–120).
ASL, arterial spin labelling; ATP, adenosine triphosphate; BOLD, blood oxygen level-dependent; CBF, cerebral blood flow; CMRO2, cerebral metabolic rate of oxygen; FDG, fluorodeoxyglucose; fPET, functional PET; MRS, magnetic resonance spectroscopy; PCr, phosphocreatine; PET, positron emission tomography; Pi, inorganic phosphate; TCA, tricarboxylic acid cycle; TCD, transcranial Doppler.
4.1Molecular capacity and static metabolite pools
Transporter or enzyme abundance and metabolite concentrations define molecular capacity and pool state rather than realised metabolic throughput (Table 1). In vivo pathway activity additionally depends on cellular localisation, substrate and cofactor supply, compartmentation and downstream use (80, 95, 96). Diabetes illustrates this pool-flux dissociation: GLUT3 abundance can increase despite lower glucose utilisation (121), brain glycogen can remain relatively preserved despite slower 13C incorporation (122), and a larger lactate pool during hypoglycaemia need not be accompanied by greater downstream oxidative labelling (9). High-energy-phosphate pools are similarly buffered, because ATP and phosphocreatine may remain comparatively stable while activity-related exchange or synthesis-related flux changes (11, 16). Static abundance should therefore be treated as a constraint or hypothesis-generating signal rather than direct evidence of impaired delivery, oxidation or ATP support.
4.2Task-evoked concentration, uptake, and haemodynamic responses
Task-evoked concentration, uptake and haemodynamic responses can reveal abnormalities hidden at rest, but each remains pathway-specific (Table 1). Microdialysis and 1H MRS report concentration balance rather than uptake or oxidation, FDG integrates delivery, transport and phosphorylation-dependent trapping, and CBF or BOLD characterises vascular or haemodynamic responses rather than downstream ATP production (10, 31–36, 64, 65, 68, 69, 97–101). Multimodal dissociation is therefore informative rather than contradictory: FDG and BOLD responses can diverge within the same task (35, 36).
In diabetes, memory-task FDG responses may be spatially diffuse (123), cognitive-task cerebrovascular recruitment can be reduced despite similar resting flow velocity (100), hypoglycaemia can redistribute perfusion (124), and adjustment for cerebrovascular reactivity can alter apparent BOLD associations (101). Preclinical models likewise show smaller or delayed evoked vascular responses (20, 125). These observations identify uptake or vascular-response abnormalities but do not establish the complete pathway from cognitive demand to oxidative ATP support.
4.3Tracer-derived flux, oxidative metabolism, and cellular energetic state
Tracer and energetic methods resolve deeper steps of the pathway but remain model-dependent (Table 1). Isotopic enrichment is not equivalent to flux unless precursor input, pool size, sampling kinetics and an explicit model are incorporated. Hyperpolarised 13C improves temporal sensitivity but remains influenced by delivery, exchange, redox state and polarisation decay (108) (97). CMRO2 more directly addresses oxidative demand but does not identify the oxidised substrate, consuming cell or ATP output (109–112). Respirometry and extracellular-flux assays quantify capacity under assay conditions rather than intact-brain recruitment during cognition (113–115).
High-energy-phosphate measurements likewise separate state from turnover. Conventional 31P MRS measures ATP, phosphocreatine and inorganic-phosphate pools, whereas magnetisation transfer provides model-dependent exchange estimates (16, 116, 117). Genetically encoded ATP/ADP and redox sensors add temporal and subcellular resolution but still integrate production, consumption, transport and buffering (11, 15, 118–120). No single readout establishes functional sufficiency; inference should therefore stop at the last measured link.
4.4Mapping measurements to permissible claims
Interpretation should remain closest to the measured domain (Table 1). Molecular capacity, static pools, task-evoked proxies, model-derived rates, oxidative or energetic state and neural or cognitive outcomes form an inferential map, not a quality ranking or obligatory biological sequence (16, 36, 64, 80, 95, 103). Cross-level causal claims require temporal alignment, appropriate spatial and cellular resolution, and perturbation or rescue linking the proposed steps (11, 80, 107). A neural or cognitive deficit establishes a functional phenotype, and an unchanged mean does not establish intact adaptation. Inference should stop at the last process, timescale and compartment directly resolved.
5Neurometabolic responses to cognitive and metabolic challenges in diabetes
The four challenge categories are grouped according to the primary experimental input being manipulated, rather than according to an assumed disease mechanism. Cognitive-task challenges increase endogenous neural demand; glucose or insulin challenges alter hormonal or substrate availability; hypoglycaemic or alternative-substrate challenges stress fuel availability and compensatory flexibility; and controlled neural or network challenges impose more spatially or temporally defined circuit activity (Figure 3). These categories are pragmatic perturbation classes rather than mutually exclusive biological mechanisms, and some experiments may span more than one category. Importantly, challenge paradigms are not failure phenotypes: any challenge may reveal more than one phenotype, and the same phenotype may be elicited by different challenges. Together, these paradigms resolve selected links extending from vascular recruitment and substrate entry to oxidative flux, energetic state and neural function, but the demand-to-cognition chain remains untested end-to-end in clinically characterised DCI.
5.1Cognitive-task challenges: demand-evoked neurometabolic vulnerability
Cognitive testing is a physiological perturbation that raises regional energy demand. In healthy preparations, demanding spatial tasks lower hippocampal extracellular glucose, astrocytic glycogen-derived lactate supports long-term memory, and sustained firing requires ongoing glycolytic and mitochondrial ATP supply (10–12). Contextual learning also increases neuronal glucose uptake and surface GLUT3 recruitment; disrupting these responses lowers event-related glucose and ATP signals and impairs memory acquisition (126, 127). The relevant reference state is therefore the capacity to sustain energetic support as demand rises, rather than transporter or metabolite abundance at rest.
Human studies of metabolism during cognition remain scarce. Under acute lipid loading, sustained testing in healthy adults was accompanied by a lower hippocampal PCr/ATP ratio without poorer overall performance (17). This result identifies a demand-sensitive change in high-energy-phosphate balance, not impaired ATP synthesis: acquisition was not tightly task-locked, no formal lipid-by-task interaction was reported, and the ratio cannot distinguish greater phosphocreatine use from deficient ATP regeneration. More disease-proximal evidence comes from older adults with prediabetes or early type 2 diabetes. During memory encoding, they showed a more diffuse pattern of FDG-derived cerebral glucose metabolism and poorer delayed recall than normoglycaemic controls (123). Because uptake was integrated over a prolonged period and the control group was small, the study supports altered metabolic recruitment but not a temporally resolved failure of glucose flux.
Task-evoked neural phenotypes in established type 2 diabetes are heterogeneous. Memory-task fMRI has shown reduced dorsolateral prefrontal recruitment and incomplete default-mode deactivation despite preserved performance (128). In graded n-back tasks, frontal recruitment weakened progressively as load increased in some cohorts (129). By contrast, newly diagnosed, untreated type 2 diabetes was associated with greater frontoparietal recruitment only at the highest load (130). Acute hyperglycaemia also abolished the normal increment in frontal and anterior cingulate BOLD response from intermediate to high working-memory load in participants with type 2 diabetes, but not in controls (131). Thus, greater recruitment may support performance early in disease, whereas higher demand or an adverse glycaemic state can expose attenuation; there is no universal hypoactivation signature.
Temporal and vascular measurements further qualify these findings. In type 2 diabetes, magnetoencephalography linked poorer glycaemic control to phase-specific abnormalities during working-memory encoding and maintenance (132). In healthy adults, simultaneous FDG-fPET and BOLD fMRI showed that glucose-related and haemodynamic responses can dissociate across networks (35). In diabetes, adjustment for cerebrovascular reactivity changed the spatial pattern of working-memory BOLD associations with HbA1c (101). BOLD abnormalities therefore establish altered task-evoked haemodynamics, but do not locate the limitation in neural activity, vascular delivery, substrate use or oxidative ATP support.
Cognitive demand is therefore an informative neurometabolic stressor, but activity-dependent energy failure is not yet an established mechanism of diabetes-associated cognitive impairment. Disease-proximal metabolic evidence remains temporally coarse, hippocampal high-energy-phosphate evidence comes from acute insulin resistance in healthy volunteers, and chronic type 2 diabetes studies largely measure haemodynamic or electrophysiological responses. No clinically characterised study has combined graded cognitive demand with time-resolved substrate flux or ATP-related measures and a neural or behavioural endpoint.
5.2Glucose and insulin challenges: substrate responsiveness and reserve mobilization
Glucose and insulin challenges test cerebral responsiveness to controlled changes in hormonal or substrate availability. A hyperinsulinaemic-euglycaemic clamp varies insulin while maintaining glycaemia, whereas a hyperglycaemic clamp tests the response to increased circulating glucose. In healthy adults, physiological hyperinsulinaemia produced little change in net whole-brain glucose transport or metabolic rate (133). Marked hyperinsulinaemia likewise did not change estimated cerebral glucose utilisation in an early FDG-PET study of type 2 diabetes (134). These whole-brain null responses do not establish cerebral insulin insensitivity, because central insulin effects may be regional, neural or vascular rather than expressed as a large global change in glucose metabolism.
Later studies identified an altered FDG phenotype in systemic insulin resistance. During paired fasting and clamp experiments, insulin increased FDG-derived cerebral glucose metabolism in impaired glucose tolerance and severe obesity, with little increase in lean controls (24, 25). This obesity-associated response diminished after bariatric surgery as peripheral insulin sensitivity improved (25). Whole-body PET/MRI studies and a large pooled cohort also associated lower peripheral insulin sensitivity with higher cerebral FDG uptake during hyperinsulinaemia (26–29). Several reports analysed related or partially overlapping cohorts and should be viewed as complementary characterisation rather than independent replication. Meta-analysis nevertheless supported a state-dependent reversal, with insulin resistance tending to accompany lower fasting metabolism but higher insulin-stimulated values (30). The direction changed under another endocrine design: after suppression of endogenous insulin, physiological replacement elicited an attenuated global and frontostriatal FDG response in insulin-resistant participants (135). Dynamic PET modelling in obese miniature pigs attributed the higher insulin-stimulated signal mainly to increased inward glucose transfer, without corresponding changes in phosphorylation or outward transport (37). FDG trapping therefore indicates altered delivery, transport and phosphorylation, not downstream oxidation or ATP support.
Hyperglycaemia reveals a complementary disturbance. During controlled exposure, adults with obesity or poorly controlled type 2 diabetes showed a smaller rise in occipital brain glucose concentration than lean controls (97). Repeating the protocol after improved glycaemic management produced a larger cerebral response in the same patients (98). These concentration changes integrate influx, utilisation and redistribution, and the intervention was multifactorial. In younger adults with obesity, [1-13C]glucose MRS showed a lower model-derived ratio of blood-to-brain transport capacity to glucose utilisation but comparable estimated tricarboxylic-acid-cycle flux (136). Higher FDG trapping during hyperinsulinaemia and weaker glucose-pool accumulation during hyperglycaemia are therefore compatible: they describe different input-output relationships, neither of which isolates oxidative ATP support.
Central insulin responsiveness can also change without mirroring FDG uptake. Insulin produced smaller changes in stimulus-evoked cortical activity in overweight adults (137). In high-fat-fed rats, it failed to recruit hippocampal microvascular perfusion or Akt phosphorylation, alongside separately measured hippocampus-dependent cognitive impairment (138). Intranasal insulin altered regional perfusion and vasoreactivity in a small crossover study of older adults with type 2 diabetes, but cognitive effects were limited (139). These findings implicate neural and vascular responsiveness without identifying cerebral glucose flux as the mediator. Regional animal studies more directly suggest impaired reserve mobilisation. Intrahippocampal insulin enhanced spatial memory and altered extracellular glucose and lactate in lean rats, whereas diet-induced obesity attenuated both responses in parallel experiments (18). Diet-induced diabetes was also associated with lower hippocampal glycogen and little additional net glycogen change after local insulin (19). In Goto-Kakizaki rats, incorporation of glucose-derived carbon into brain glycogen was impaired despite relatively preserved glycogen content, particularly in the hippocampus (122). Together, these data support reduced regional insulin responsiveness and less readily recruitable stored substrate, but not metabolic mediation of memory failure because behaviour and metabolism were not aligned within one challenge window.
Glucose and insulin challenges therefore reveal a state-dependent reorganisation rather than a single diabetic phenotype: responses may be exaggerated, attenuated or preserved according to stimulus, hormonal background, region and model. Hippocampal stored fuel may be less readily mobilised, but whether this limits oxidative ATP support during cognition remains unknown. No study in clinically characterised diabetes-associated cognitive impairment has aligned a standardised metabolic challenge with pathway-specific flux, energetic state and cognitive performance.
5.3Hypoglycaemic and alternative-substrate challenges: conditional adaptation and functional trade-offs
Acute hypoglycaemia lowers the dominant circulating fuel while activating counter-regulatory, cerebrovascular and neural responses and increasing reliance on endogenous reserves or alternative substrates (140, 141). The phenotype therefore depends on both the current glucose nadir and metabolic history. Recurrent exposure can lower the thresholds for autonomic warning and counter-regulation, so preserved cerebral function may coexist with reduced awareness and greater clinical risk (140). Controlled hypoglycaemia tests adaptive range, but most evidence comes from insulin-treated type 1 diabetes and cannot by itself explain chronic cognitive decline in diabetes-associated cognitive impairment.
Recurrent hypoglycaemia illustrates why response direction alone cannot distinguish adaptation from vulnerability. In rats, it abolished the usual task-associated decline in hippocampal extracellular glucose and improved spatial alternation during euglycaemia, yet accelerated the decline and worsened performance during a later hypoglycaemic episode (142). Compartmental 13C modelling showed a corresponding shift: neuronal glucose use increased at euglycaemia, whereas neuronal pyruvate-dehydrogenase flux and total tricarboxylic-acid-cycle activity fell during subsequent hypoglycaemia (143). In healthy adults, recurrent exposure shifted the thresholds for declining brain glucose uptake and cognition while delaying symptoms and hormonal defence (144). Evidence for a simple increase in transport is weak. Model-derived hippocampal transport-to-utilisation ratios did not differ among type 1 diabetes groups with impaired or normal awareness and healthy controls (145). Three hypoglycaemic exposures likewise produced no detectable change in corresponding prefrontal or hypothalamic kinetics during later hyperglycaemic clamps (146). These findings support a changed operating range, while leaving regional, chronic and substrate-specific adaptation unresolved.
Human studies similarly reveal preservation coupled to redistribution or functional cost. Whole-brain glucose uptake can remain preserved during moderate hypoglycaemia in intensively treated type 1 diabetes despite weakened symptoms and counter-regulation (147). Awareness-dependent global and prefrontal metabolic responses have also been detected with 3-O-methyl-D-glucose PET (148). During working memory, adults with type 1 diabetes maintained performance through greater task-network activation and less default-mode suppression (149). Yet standardised testing at comparable glucose levels found cognitive impairment across type 1 diabetes, insulin-treated type 2 diabetes and controls, largely irrespective of awareness status (150). Recent real-world exposure was also unrelated to the magnitude of cognitive decline during a later clamp (151). Hypoglycaemia nevertheless produced a modest working-memory cost and regional cerebral-blood-flow redistribution in a crossover study (124). Arterial-spin labelling subsequently revealed disrupted vascular oscillatory regulation in type 1 diabetes and weaker regional perfusion recruitment with impaired awareness (152). Preserved performance, uptake or flow therefore need not represent an unchanged cerebral response.
Metabolic measurements also reveal substrate-specific adaptation. The occipital PCr/ATP ratio remained stable during moderate hypoglycaemia but was lower throughout in type 1 diabetes than in controls (153). This ratio does not measure ATP synthesis. Model-derived tricarboxylic-acid-cycle flux was preserved during [1-13C]glucose clamps in healthy adults (154). It was also maintained within type 1 diabetes and exceeded a historical healthy estimate during hypoglycaemia, although the sample was small, incompletely paired and compared with noncontemporaneous controls (155). Acetate transport and oxidation were higher during hypoglycaemia in well-controlled type 1 diabetes (156). Absolute cortical acetate metabolism was also higher with impaired awareness than with normal awareness or in healthy controls (157). This cannot be generalised to other fuels. Frequent hypoglycaemia was associated with a larger calculated brain lactate pool without greater lactate-derived glutamate or glutamine enrichment (9). Exercise-induced hyperlactataemia produced awareness-dependent brain lactate kinetics (158). In recurrently hypoglycaemic rats, enhanced lactate transport contributed little directly to oxidation, yet lactate preserved glucose metabolism, tricarboxylic-acid-cycle activity and neural function (159). Alternative fuels may thus reshape glucose use rather than simply replace it.
Interventions show that alternative fuels can support function when glucose is scarce. Oral medium-chain triglycerides protected several cognitive measures during hypoglycaemia in adults with type 1 diabetes, and beta-hydroxybutyrate sustained synaptic transmission in rat hippocampal slices exposed to low glucose (160). The human study did not trace cerebral fuel use, and the cross-species observations do not form one causal chain. Complementary 2026 work found a greater relative contribution of labelled beta-hydroxybutyrate to cerebral oxidation in type 1 diabetes because total oxidative flux was lower, not because absolute ketone oxidation was higher. In a separate cohort, medium-chain triglycerides improved working memory and regional BOLD activation during hypoglycaemia without altering counter-regulatory hormones (161). The metabolic and functional findings came from different cohorts, and BOLD does not measure ATP production. Brain slices from db/db mice likewise showed lower glucose metabolism but greater acetate and beta-hydroxybutyrate use, while isolated mitochondria had higher oxygen-consumption and ATP-synthesis capacity (162). These ex vivo capacities demonstrate residual flexibility, not restored in vivo task energetics.
Responses to low glucose depend on exposure depth and duration, prior hypoglycaemia, awareness status, brain region and the selected endpoint. Preserved oxidative flux can coexist with impaired autonomic warning, and preserved accuracy with broader network recruitment; larger substrate pools or functional rescue do not establish greater endogenous oxidation. No study in clinically characterised diabetes-associated cognitive impairment has combined a cognitive task, controlled glucose restriction, absolute alternative-substrate flux, energetic state and performance. Recurrent hypoglycaemia therefore defines a conditional boundary between compensation and failure rather than an established mechanism of chronic cognitive impairment.
5.4Neural and network challenges: stimulus-evoked neurovascular, oscillatory, and synaptic responses
Controlled neural challenges impose defined inputs and resolve physiological outputs more precisely than whole cognitive tasks. Sensory stimulation, oscillations, sharp-wave ripples and long-term potentiation probe vascular, network or synaptic responsiveness under load. Functional hyperaemia arises from coordinated signalling among neurons, astrocytes, mural cells and endothelium (14, 50). It measures vascular recruitment rather than glucose use or ATP synthesis, and blood flow can diverge from glucose uptake in activated cortex (163). Conversely, abnormal oscillations or potentiation establish circuit dysfunction without identifying the limiting metabolic step.
Non-diabetic experiments define the metabolic relevance of these challenges. Electrical activity rapidly recruits glycolytic and mitochondrial ATP synthesis at presynaptic terminals, and interrupting that recruitment impairs vesicle cycling despite a substantial resting ATP pool (11). Loss of the mitochondrial calcium uniporter reduces hippocampal gamma synchrony, alters sharp-wave ripples and prevents activity-related pyruvate-dehydrogenase dephosphorylation (94). In hippocampal slices, lactate alone preserved lower-demand sharp-wave ripples but disrupted gamma and theta-gamma activity despite increased oxygen consumption (164). Conversely, gamma oscillations remained stable during moderate neuroinflammation through greater mitochondrial ATP production and oxygen use despite a changed tissue lactate-to-glucose ratio (165). Network support therefore depends on demand, substrate mixture and the pathway of ATP supply, not substrate availability alone.
Diabetes studies show heterogeneous stimulus-evoked vascular responses. In streptozotocin-diabetic rats, sensory stimulation produced smaller pial arteriolar dilations and impaired BKCa- and Kir-dependent reactivity; PKC inhibition restored these responses, and later islet transplantation prevented them (166, 167). In young Goto-Kakizaki rats, glutamatergic activation evoked briefer nitric-oxide transients and smaller, shorter hippocampal microvascular responses alongside impaired spatial memory (125). Other abnormalities are mainly temporal. Diabetic mice had largely preserved resting perfusion and overall response magnitude but delayed peak recruitment; glyoxalase-1 overexpression normalised the delay and limited visuospatial impairment (20). Luseogliflozin improved functional hyperaemia, vascular integrity and cognition in aged diabetic rats, although systemic glucose-lowering and vascular effects preclude assignment to one neural-metabolic step (168). Akita mice instead retained cortical neurovascular coupling across disease progression (169). Their dilation became more prolonged with age rather than uniformly smaller. Model, region, age, anaesthesia, stimulus and response metric therefore shape the apparent phenotype, and none of these studies measured challenge-locked metabolic flux or ATP turnover.
Oscillatory challenges reveal circuit abnormalities that resting measurements miss. Kainate-evoked hippocampal gamma activity in Goto-Kakizaki rats shifted towards faster components and became less variable, with altered parvalbumin-interneuron markers (170). In type 2 diabetic mice, oscillations slowed, theta-gamma coupling and cortico-hippocampal synchrony weakened, and sharp-wave ripples became longer with greater ripple-associated gamma power (171). These patterns indicate temporal reorganisation but do not identify substrate use or energetic state. In streptozotocin-diabetic mice, exogenous L-lactate improved long-term object recognition and ex vivo sharp-wave-ripple amplitude in separate experiments (172). This rescue demonstrates substrate-responsive capacity, not endogenous lactate deficiency, oxidation or metabolic mediation of memory recovery. Across network states, lactate is not uniformly sufficient (164).
Long-term potentiation links metabolic support to durable synaptic change, but its impairment is mechanistically non-specific. In diabetic rat hippocampal slices, pyruvate or insulin restored potentiation, whereas additional glucose did not (173). Acute GLUT4 blockade disrupted CA3-CA1 transmission, synchronisation and potentiation, with several deficits rescued by D-β-hydroxybutyrate (174). These interventions show substrate flexibility but do not reproduce chronic diabetes. Mitochondrial rescue provides a stronger disease link. In db/db mice, GSK3β-Drp1-dependent fragmentation, impaired complex I activity and lower ATP production accompanied defective potentiation; pathway modulation restored plasticity (175). In streptozotocin-diabetic mice, cyclophilin-D deletion protected ATP-synthase function, respiration, potentiation and cognition (176). Because metabolic assays and synaptic challenges were not aligned temporally, neither study identifies the energetic step failing during potentiation. Hippocampal insulin-receptor knockdown also impaired potentiation and learning without systemic dysglycaemia, showing that receptor signalling can produce a similar endpoint (177). High-fat feeding was associated with lower brain glucose uptake, altered neuronal and astrocytic carbon metabolism and weaker potentiation (178). Lipoic acid improved these domains, but measurements were neither synchronous nor pathway-specific. Presynaptic pyruvate transport, astrocytic glycogen and lactate transfer, neuronal translation and activity-triggered GLUT4 insertion can each support synaptic function (12, 44, 47, 72, 76). Their roles during a diabetes-relevant challenge remain unresolved.
Diabetes can alter the amplitude, timing or duration of evoked vascular responses and reorganise oscillatory and synaptic activity under load, but preserved coupling and successful substrate rescue show that residual capacity can coexist with vulnerability. Reduced gamma activity, abnormal sharp-wave ripples and impaired potentiation should therefore be treated as challenge-sensitive functional endpoints, not metabolic diagnoses. The literature has not yet located the first limiting energetic step connecting these abnormalities to diabetes-associated cognitive impairment.
6An integrated framework for activity-dependent neurometabolic failure in DCI
Activity-dependent neurometabolic failure concerns inadequate matching of metabolic support to neural demand within a defined challenge rather than a uniform reduction in brain metabolism. Established mechanisms and the proposed phenotypes answer different questions. Neurovascular dysfunction, brain insulin resistance and mitochondrial impairment identify candidate lesion classes, whereas the four phenotypes describe how a failure is expressed at a measured support node during a defined challenge and which observations could distinguish it from competing explanations (Figure 4). The phenotypes are non-exclusive response descriptors rather than mutually exclusive causal diagnoses. A downstream utilisation defect may secondarily alter upstream recruitment through feedback, while an upstream delivery defect may propagate to downstream energetic failure. A single lesion may therefore produce different phenotypes across workload, disease stage or compensatory state, while the same phenotype may arise from different lesions. Phenotype assignment should remain tied to the response dimension and node actually measured rather than inferred across unmeasured links. Neurometabolic reserve denotes the additional support recruitable above baseline. Reserve exhaustion requires repeated within-person evidence that this range progressively contracts under comparable demand and physiological conditions and that the contraction precedes and predicts neural or cognitive failure.
The four proposed phenotypes have different evidential bases and should not be interpreted as equally established mechanisms of DCI. Table 2 summarises the study-level findings, their evidential roles and the remaining inferential boundaries. Challenge-based observations support abnormalities at selected vascular and substrate-pool readouts, whereas cellular misallocation and downstream utilisation failure rely more heavily on preclinical modelling or perturbation studies. Support for an individual component does not establish the complete demand-to-function pathway in clinically characterised DCI.
Table 2Evidence proximity to clinically characterised DCI, study-level support and inferential boundaries of the four proposed neurometabolic failure phenotypes.
Reference Population/model and clinical relevance Challenge and readout Reported finding Evidence role and directness Inferential boundary
Blunted recruitmentProposed signature. A smaller challenge-evoked response or input–output slope at a specified support node, after accounting for baseline state, systemic input and neural engagement.Evidence status. Challenge-based support at selected vascular and substrate-pool readouts; deficient cognitive-task-evoked oxidative or ATP recruitment remains unconfirmed.
Nealon et al. (100) Adults with T2D and matched controls.Other human diabetes evidence. Cognition assessed; clinical DCI is not established by lower group-average performance. Cognitive tasks; transcranial Doppler measurement of middle-cerebral-artery flow velocity. Resting flow velocity was similar, but task-evoked cerebrovascular responses were smaller and cognitive performance was poorer in T2D. Human task-evoked vascular observation. Arterial velocity is not regional substrate delivery. Without matched neural drive, the finding does not identify a metabolic bottleneck or metabolic mediation.
Hwang et al. (97) Adults with obesity or poorly controlled T2D; lean controls.Mixed human evidence. T2D: other human diabetes evidence; obesity: other human evidence. Relevance is subgroup-specific. Matched hyperglycaemic clamp; occipital brain-glucose concentrations measured by MRS. Brain-glucose concentration rose less in obesity and T2D during matched hyperglycaemia. Human dynamic substrate-pool observation. No cognitive task. A smaller concentration increment does not distinguish reduced input from increased use, or establish deficient oxidation or ATP production.
Sanchez-Rangel et al. (98) Eight adults with T2D; within-person pre–post study.Other human diabetes evidence. No cognitive endpoint; clinical DCI applicability remains undetermined. Hyperglycaemic clamp and brain-glucose MRS before and after 12 weeks of intensified glycaemic management. The brain-glucose concentration response increased after improved glycaemic control. Human reversibility evidence for a pool response. Small, uncontrolled, multifactorial intervention; no cognitive endpoint or pathway-specific flux measurement.
Emmanuel et al. (17) Healthy adults with acute lipid-induced insulin resistance.Other human evidence. Acute experimental proof of principle, not a clinical DCI cohort. Sustained cognitive testing; hippocampal PCr/ATP measured by ³¹P MRS. PCr/ATP fell during cognitive testing with lipid infusion, but not in the no-lipid condition; overall performance did not deteriorate. Human proof of principle for demand-associated energetic vulnerability; not phenotype-specific validation. Acute, small, non-DCI model; no formal lipid-by-task interaction reported. PCr/ATP does not distinguish increased expenditure or buffering from deficient ATP synthesis.
McNay et al. (18) High-fat-fed, insulin-resistant rats and chow-fed controls.Preclinical diabetes/insulin-resistance evidence. Local hippocampal insulin; memory testing and extracellular glucose/lactate microdialysis in parallel experiments. Both mnemonic and extracellular glucose/lactate responses to hippocampal insulin were attenuated by high-fat feeding. Parallel behavioural and metabolic evidence of impaired insulin responsiveness. Exogenous insulin and separate experiments do not establish metabolic mediation of memory. Microdialysis measures concentration, not carbon flux.
McNay (19) Diet-induced diabetic rats and controls.Preclinical diabetes/insulin-resistance evidence. Local hippocampal insulin; one post-challenge glycogen-pool endpoint. Hippocampal glycogen was lower; no additional net decrease after insulin was detected in diabetic rats. Animal substrate-pool response to insulin. No cognitive endpoint. A single net pool change cannot distinguish glycogen synthesis, breakdown or a baseline floor effect.
Mistimed recruitmentProposed signature. An altered onset, peak, duration or recovery that is misaligned with the neural event requiring support; a normal peak or integral does not exclude such misalignment.Evidence status. Human and animal evidence for altered haemodynamic timing; extension to cellular substrate flux, oxidation or ATP timing remains inferential.
Duarte et al. (21) Adults with T2D and matched controls.Other human diabetes evidence. Performance-matched task; a clinical DCI phenotype is not established by this comparison. Performance-matched visual discrimination; event-related fMRI haemodynamic response profiles. Haemodynamic response profiles differed in T2D despite matched task performance. Human task-evoked haemodynamic observation. BOLD combines neural and vascular influences; it does not isolate substrate metabolism or establish delayed cellular energy supply.
Duarte et al. (22) Expanded T2D/control cohort, including participants from (21).Other human diabetes evidence. Overlapping cohort; clinical DCI applicability remains undetermined. Event-related fMRI; response rise, peak, amplitude and post-stimulus recovery across regions. T2D was associated with slower response evolution, later peaks and altered amplitude/recovery. Expanded human analysis of haemodynamic kinetics. Not an independent replication of (21). Metabolic response timing and its relation to independent neural timing were not resolved.
Berends et al. (20) STZ-diabetic mice with or without Glo1 overexpression.Preclinical diabetes/insulin-resistance evidence. Whisker-evoked cortical blood-flow kinetics; separate later visuospatial testing. Recruitment was delayed despite preserved peak/integrated flow; Glo1 overexpression shortened the delay and limited later learning impairment. Animal vascular perturbation evidence; also limits universal blunting. Systemic overexpression and separate vascular/behavioural measurements do not demonstrate task-aligned metabolic mediation.
Gonçalves et al. (125) Young Goto–Kakizaki rats and controls.Preclinical diabetes/insulin-resistance evidence. Local glutamatergic stimulation; hippocampal nitric-oxide/blood-flow kinetics; separate spatial-memory testing. Nitric-oxide transients and microvascular responses were smaller and shorter, alongside poorer spatial memory. Animal vascular amplitude–duration evidence; relevant to blunting and mistiming. Stimulation and memory testing were not one task-locked metabolic experiment; nitric-oxide/flow kinetics do not measure substrate or ATP flux.
Chiang et al. (169) Akita mice followed longitudinally.Preclinical diabetes/insulin-resistance evidence. Two-photon imaging of stimulus-evoked cortical neurovascular coupling. Peak coupling remained preserved across disease progression; older diabetic mice showed more prolonged dilation. Temporal heterogeneity and evidence against universal vascular blunting. Prolonged dilation alone is not proof of harmful mistiming. No cognitive endpoint or metabolic-flux measurement.
Ugidos et al. (179) STZ-treated mice exposed to recurrent non-severe hypoglycaemia.Preclinical diabetes/insulin-resistance evidence. Awake two-photon imaging of stimulus-evoked penetrating-arteriole dilation. Recurrent hypoglycaemia produced a transient delay without reducing peak or integrated dilation; sustained hyperglycaemia alone did not reproduce the delay. Exposure-dependent vascular kinetic evidence. A sensory vascular response, not a cognitive-task metabolic response; does not identify delayed substrate entry, oxidation or ATP regeneration.
Candidate substrate misallocationProposed signature. An input-controlled shift in substrate delivery, retention or fate away from the task-relevant cell or compartment that is associated with inadequate energetic or functional support. Redistribution without demonstrated functional cost remains descriptive.Evidence status. Predominantly preclinical, model-derived and cell-associated mechanistic support; demand-specific pathological allocation remains a hypothesis.
Girault et al. (180) Goto–Kakizaki rats and controls.Preclinical diabetes/insulin-resistance evidence. Dynamic in vivo [1, 6-¹³C]glucose MRS and neuronal–astrocytic compartmental modelling. Models estimated lower neuronal TCA activity, higher astrocytic TCA activity and reduced glutamate–glutamine cycling. Model-derived compartmental metabolic evidence. Compartmental rates depend on precursor input, exchange and model structure; no task-locked demonstration of carbon diversion from active neurons.
Zheng et al. (181) db/db mice with cognitive decline and controls.Preclinical diabetes/insulin-resistance evidence. Acetate- and lactate-derived ¹³C labelling in brain metabolites. Acetate-derived labelling decreased whereas lactate-derived labelling increased in selected metabolites. Substrate-dependent metabolic reorganisation. Labelling depends on input, dilution and pool size. It does not directly establish cell-to-cell transfer or pathological allocation during cognition.
Chen et al. (182) db/db mice with endothelial-enriched Hif1a knockdown.Preclinical diabetes/insulin-resistance evidence. Animal intervention component. Endothelial glycolytic features, bulk hippocampal lactate, memory and neurogenesis. Knockdown reduced glycolytic features and lactate accumulation while improving memory and neurogenesis, without correcting systemic glycaemia. Endothelial-associated perturbation evidence. Incomplete lineage restriction, bulk lactate and HIF-1α pleiotropy limit source/fate attribution. Neuronal glucose deprivation was not demonstrated.
Li et al. (183) db/db mice with hippocampal AAV-shRNA Khk knockdown; complementary microglial experiments.Preclinical diabetes/insulin-resistance evidence. Fructose metabolism; mitochondrial, synaptic, electrophysiological and cognitive readouts. Microglia-associated KHK metabolism was implicated; hippocampal knockdown improved mitochondrial, synaptic and cognitive phenotypes. Hippocampal perturbation with microglial mechanistic support. The in vivo intervention was not microglia-specific conditional deletion. It did not show task-locked substrate capture at the expense of neurons.
MalutilizationProposed signature. Relatively preserved substrate entry or availability accompanied by disproportionately impaired downstream oxidation, energetic support or function during the same challenge.Evidence status. Preclinical intervention-based support for mitochondrial constraints, with indirect human pool–use evidence; the complete within-challenge phenotype remains unconfirmed.
De Feyter et al. (9) Adults with T1D and controls during hypoglycaemia.Other human diabetes evidence. Acute hypoglycaemic adaptation; a chronic DCI phenotype is not established. [3-¹³C]lactate infusion and brain MRS; calculated lactate pool and downstream fractional enrichment. The calculated brain-lactate pool was larger in T1D, without greater lactate-derived glutamate/glutamine enrichment. Human substrate-pool–use dissociation. No cognitive challenge. Similar fractional enrichment does not establish reduced absolute lactate oxidation or an ATP deficit.
Yan et al. (176) STZ-diabetic mice with or without cyclophilin-D deficiency.Preclinical diabetes/insulin-resistance evidence. Mitochondrial permeability, ATP-synthase activity, respiration, oxidative stress, LTP and memory. Cyclophilin-D deficiency protected ATP-synthase activity and respiration and alleviated plasticity and memory deficits. Genetic perturbation linking mitochondrial and functional outcomes. No aligned measurement of preserved substrate entry and deficient energetic support during cognition; separate readouts do not establish metabolic mediation.
Huang et al. (175) db/db mice and high-glucose neuronal models.Preclinical diabetes/insulin-resistance evidence. Drp1/GSK3β modulation; mitochondrial dynamics, complex-I activity, ATP-related measures and LTP. Pathway modulation mitigated mitochondrial fragmentation, complex-I/ATP abnormalities and impaired LTP. Perturbation-based downstream mitochondrial support. No paired substrate-entry and task-evoked oxidation measurement; LTP improvement does not identify the energetic mediator.
Chandrasekaran et al. (184) STZ-diabetic rats receiving NMN from early diabetes.Preclinical diabetes/insulin-resistance evidence. Hippocampal metabolites/NAD+; isolated-mitochondrial respiration and spare capacity; CA1 neurons and memory. Glucose accumulation coexisted with lower NAD+ and respiration; NMN preserved respiration/spare capacity, CA1 neurons and memory without correcting systemic glycaemia. Preventive intervention linking respiratory and functional protection. Respiratory assays were ex vivo, not task-evoked flux measurements. Early treatment does not demonstrate reversal of established cognitive impairment.
Malutilization
Ruegsegger et al. (185) High-fat-fed mice receiving exercise or metformin.Preclinical diabetes/insulin-resistance evidence. ATP-production capacity in isolated cerebral mitochondria and oxidative-enzyme activity. High-fat feeding reduced mitochondrial ATP-production capacity and oxidative-enzyme activity; interventions improved these measures. Mitochondrial-capacity and reversibility evidence. No cognitive endpoint or in vivo demand-evoked metabolism. Assay capacity is not equivalent to ATP support at engaged synapses.
Ruegsegger et al. (186) STZ-diabetic mice after insulin withdrawal; separate non-diabetic mice given intranasal insulin.Mixed preclinical evidence. Diabetes-related withdrawal experiment; separate non-diabetic reference experiment. Cerebral mitochondrial ATP production, oxidative enzymes and monocarboxylate-related measures. Withdrawal reduced ATP production. Intranasal insulin increased ATP production per tissue mass, but not per mitochondrial protein, in non-diabetic mice. Separate insulin-deficiency and insulin-administration experiments. Not a diabetic intranasal-insulin rescue experiment. Normalisation matters; no cognition or task-locked substrate-to-ATP measurement.
Zhao et al. (187) Diabetic mice treated with FGF21.Preclinical diabetes/insulin-resistance evidence. Hippocampal lactate, MCT2, LDH-B activity, NAD+, ATP and memory. FGF21 lowered lactate and increased MCT2, LDH-B activity, NAD+ and ATP, alongside improved memory. Intervention linking metabolic markers with functional improvement. Expression, enzyme activity and pool measures do not quantify lactate uptake or oxidation flux; systemic and signalling effects remain possible.
Mason et al. (156) Adults with well-controlled T1D and controls.Other human diabetes evidence. Acute hypoglycaemic challenge; a chronic DCI phenotype is not established. Hypoglycaemic clamp; [2-¹³C]acetate MRS and metabolic modelling. Brain acetate concentration and model-derived acetate utilisation relative to TCA cycling were higher in T1D. Substrate-specific limiting evidence against universal malutilization. Model-dependent and challenge-specific; no cognitive task. Enhanced use of one alternative fuel does not establish intact overall energetic support.
Andersen et al. (162) Brain slices and isolated mitochondria from db/db mice.Preclinical diabetes/insulin-resistance evidence. Glucose, acetate and β-hydroxybutyrate tracing; mitochondrial respiration and ATP-production assays. Lower glucose-derived labelling coexisted with increased use of selected alternative fuels and higher mitochondrial respiratory/ATP-producing capacity. Preclinical limiting evidence for residual metabolic flexibility. Ex vivo substrates and capacity assays do not establish adequate in vivo ATP delivery during cognition or refute a conditional utilisation deficit.
Clinical relevance. Source labels apply to the population or experimental component supporting each listed finding, not every component of a report. Direct clinical DCI evidence requires explicitly characterised cognitive impairment in participants with diabetes; diabetes status, lower group-average performance or acute challenge effects alone do not establish it. Where cognitive status is insufficiently characterised, clinical DCI applicability remains undetermined; normal cognition is not assumed.
Source categories. Other human diabetes evidence concerns T1D or T2D without an established clinical DCI phenotype for the reported comparison. Other human evidence concerns non-diabetic metabolic-risk or experimentally challenged participants. Preclinical diabetes/insulin-resistance evidence comprises the specified animal, cellular or ex vivo models; separate non-diabetic reference experiments remain distinct. Mixed sources retain their component-specific relevance. These categories are not quality or certainty grades.
Interpretation and scope. Clinical relevance is assessed separately from measurement directness and mechanistic support. Proposed signatures are experimental criteria for testing mechanistic hypotheses, not clinical diagnostic criteria or findings established by every entry. The four phenotypes are non-exclusive candidates, not validated DCI subtypes. These phenotypes describe responses within a challenge; reserve exhaustion requires longitudinal evidence of a progressively contracting recruitable response range and cannot be inferred from a single attenuated response. The matrix includes 25 distinct reports (21); and (22) have overlapping participants. No included study resolves the complete demand–substrate-fate–energetics–function pathway in clinically characterised DCI. Reference numbers follow the main manuscript bibliography.
AAV, adeno-associated virus; ATP, adenosine triphosphate; BOLD, blood oxygen level-dependent; DCI, diabetes-associated cognitive impairment; Drp1, dynamin-related protein 1; FGF21, fibroblast growth factor 21; fMRI, functional magnetic resonance imaging; Glo1, glyoxalase 1; GSK3β, glycogen synthase kinase 3β; HIF-1α, hypoxia-inducible factor 1α; KHK, ketohexokinase; LDH-B, lactate dehydrogenase B; LTP, long-term potentiation; MCT2, monocarboxylate transporter 2; MRS, magnetic resonance spectroscopy; NAD+, nicotinamide adenine dinucleotide; NMN, nicotinamide mononucleotide; PCr, phosphocreatine; shRNA, short hairpin RNA; STZ, streptozotocin; T1D and T2D, type 1 and type 2 diabetes; TCA, tricarboxylic acid cycle.
6.1Blunted recruitment: insufficient scaling of metabolic support with demand
Blunted recruitment is a defect of response gain relative to effective neural demand, not a synonym for resting cerebral hypometabolism or low task activation. In cognitive challenges, it denotes attenuated scaling of a prespecified energy-support response with effective demand, relative to a task- and readout-specific reference. This interpretation requires accounting for baseline state, imposed load, achieved performance, task-relevant neural engagement where available and systemic input. Without adequate characterisation of neural engagement, a smaller response remains descriptive and cannot distinguish deficient metabolic adaptation from reduced effective demand. The expected direction must also be specified for each readout, because a smaller signal is not uniformly pathological. A normal baseline may therefore conceal restricted dynamic reserve, whereas a low resting value may coexist with preserved scaling. In healthy preparations, hippocampal extracellular glucose responses increase with spatial-task demand, and active synaptic terminals recruit glycolysis and mitochondrial ATP synthesis to maintain presynaptic ATP (10, 11). Blunting should consequently be localised to vascular recruitment, substrate entry or mobilisation, oxidative metabolism, or energetic support rather than assigned to “brain metabolism” as a whole.
Human evidence is suggestive but incomplete. In healthy adults exposed to acute lipid loading, sustained cognitive testing was associated with a lower hippocampal PCr/ATP ratio, whereas a small resting cohort showed no comparable change (17). This study demonstrates that metabolic strain can emerge under demand, but it does not model chronic DCI or measure ATP-synthesis or glucose flux. In prediabetes or early type 2 diabetes, memory encoding elicited a more spatially diffuse FDG-derived metabolic response rather than a uniform decrement (123). Established type 2 diabetes has also been associated with weaker cognitive-task-evoked cerebrovascular responsiveness despite similar resting flow velocity (100)and less expansion of frontal recruitment with increasing working-memory load (129). Memory-task fMRI shows reduced dorsolateral prefrontal recruitment with incomplete default-mode deactivation (128).Acute hyperglycaemia can also abolish the normal BOLD increment from intermediate to high working-memory load (131). These vascular and neural findings are consistent with reduced recruitment but are not direct measurements of metabolic flux.
Non-cognitive metabolic challenges reveal a similar loss of responsiveness. Hyperinsulinaemia produces a smaller cerebrocortical response in overweight adults (137), while controlled hyperglycaemia elicits a smaller rise in occipital brain glucose in obesity and poorly controlled type 2 diabetes (97). That response increases after improved glycaemic management (98), indicating that recruitment can remain modifiable rather than irreversibly lost.
Animal studies provide mechanistic bridges at the level of stored fuel and insulin-responsive substrate handling. Intrahippocampal insulin improves memory and changes extracellular glucose and lactate in lean rats, whereas high-fat feeding attenuates both behavioural and metabolic responses (18). In diet-induced diabetes, basal hippocampal glycogen is reduced and local insulin no longer produces the net glycogen decrease observed in controls (19). Because this is a single post-challenge endpoint, altered synthesis, breakdown, and a floor effect cannot be distinguished. Complementary 13C-MRS in Goto-Kakizaki rats showed slower incorporation of glucose-derived carbon into glycogen despite relatively preserved glycogen abundance, together with lower model-derived glucose-transport capacity (122). These findings suggest altered glycogen turnover and insulin-responsive substrate mobilisation, but do not establish reserve exhaustion or trace glycogen-derived carbon into neuronal oxidation during learning.
The delivery arm can also be attenuated. High-fat feeding prevents insulin-mediated recruitment of hippocampal microvascular perfusion and vascular Akt signalling, alongside cognitive impairment assessed separately (138). Sensory stimulation evokes smaller pial arteriolar dilations in streptozotocin-diabetic rats through a PKC-sensitive vascular mechanism (166). Likewise, hippocampal glutamate challenge in young Goto–Kakizaki rats produces briefer nitric-oxide transients and smaller, shorter blood-flow responses in the same small cohort with poorer spatial-memory performance (125). These experiments establish activity-evoked vascular abnormalities, but not reduced glucose or oxygen delivery in the absence of matched neural and substrate measurements.
Blunting is nevertheless not universal. Insulin-stimulated cerebral FDG uptake may increase in systemic insulin resistance (27), and newly diagnosed type 2 diabetes can show greater frontoparietal recruitment at high working-memory load (130). In streptozotocin-diabetic mice, peak and integrated sensory-evoked blood-flow responses can remain near normal despite delayed recruitment (20). Longitudinal imaging in Akita mice likewise found preserved cortical neurovascular coupling across disease progression (169). Such increases or preservation should not automatically be labelled compensation. Response gain varies with disease stage, region, challenge intensity, hormonal state, and the measured step, and adequate upstream recruitment does not guarantee downstream energy support.
Because blood flow and glucose uptake can dissociate (163) and neurovascular coupling resolves only the vascular response (14), blunted recruitment should be assigned to a measured node. Challenge-based abnormalities support selected vascular and substrate-pool phenotypes, whereas deficient oxidative or ATP recruitment during cognition and its causal contribution to clinically characterised DCI remain unresolved (Table 2).
6.2Mistimed recruitment: temporal misalignment between neural demand and metabolic support
Neurometabolic adequacy depends on timing as well as magnitude. Mistimed recruitment denotes a support response that is misaligned with the neural event generating energetic demand. It may begin too late, terminate before activity subsides, or persist into recovery despite reaching a normal peak. Functional hyperaemia itself emerges from interactions among neurons, astrocytes, mural cells, and endothelium (50). At active synapses, neuronal glycolysis and local ATP synthesis are recruited rapidly enough to sustain neurotransmission (11, 15). Haemodynamic signals, however, integrate blood flow, blood volume, and oxygenation over distinct kinetics (69). Timing abnormalities must therefore be assigned to the measured step rather than described generically as delayed energy delivery.
The clearest clinical evidence in type 2 diabetes concerns haemodynamic kinetics. During event-related visual discrimination, participants with diabetes showed an altered haemodynamic response despite performance matched to controls (21). A subsequent analysis of the expanded cohort identified a slower rise, later peak, and abnormal post-stimulus recovery, together with amplitude changes (22). These reports are progressive analyses rather than independent replications. Transcranial Doppler studies provide weaker corroboration: visually evoked posterior cerebral artery responses were slower in hypertension and more abnormal when diabetes was also present (188). However, hypertension, the absence of a non-hypertensive diabetic group, and arterial velocity as the outcome prevent attribution to diabetes alone. None of these studies measured substrate flux, oxygen consumption, or high-energy-phosphate turnover.
Animal models more directly show that preserved magnitude can conceal altered kinetics. In streptozotocin-diabetic mice, resting cortical perfusion and peak and integrated whisker-evoked responses remained near normal, but the rise to half-maximum and maximum flow was delayed (20). Glyoxalase 1 overexpression shortened this delay and improved later spatial learning, although systemic intervention and separate vascular and behavioural measurements preclude mediation claims. Awake two-photon imaging subsequently showed that sustained hyperglycaemia could leave penetrating-arteriole kinetics intact, whereas recurrent non-severe hypoglycaemia produced a transient delay without reducing peak or integrated dilation (179). Other models show distinct distortions. Goto–Kakizaki rats generate briefer hippocampal nitric-oxide and blood-flow responses after glutamate challenge (125). Older Akita mice instead show prolonged arteriolar dilation despite preserved peak coupling (169). Mistiming is therefore conditional and is not synonymous with delayed onset. The diversity of trajectories is central to the phenotype. Latency, duration, and recovery are separable response dimensions, and each may interact differently with a brief cognitive operation.
Such distortions may matter because encoding, maintenance, retrieval, and consolidation impose energy costs over different time windows. Diabetic mice show slower hippocampal rhythms, impaired theta–gamma coordination, and prolonged sharp-wave ripples (171), together with abnormal delay-period hippocampal–cortical coupling during working memory (189). These findings establish altered neural timing, not metabolic phase delay. Recurrent hypoglycaemia also changes task-associated hippocampal extracellular glucose trajectories, with an adaptation that benefits performance during euglycaemia but becomes detrimental during acute hypoglycaemia (142). In adults with type 1 diabetes, acute hypoglycaemia slows working-memory performance while redistributing task-related regional blood flow (124). The functional effect of a dynamic response thus depends on both its timing and the prevailing substrate environment. A response adequate during sustained activation may still fail during rapid encoding, whereas prolonged support may be costly or interfere with recovery.
Potential delays may occur in neural-to-vascular signalling, substrate entry or downstream metabolism, but the measured signal localises only the observed step (14, 36, 63, 190, 191). A delayed BOLD or flow response therefore establishes altered haemodynamic recruitment, not delayed transport, oxidation or ATP regeneration. Current evidence supports mistimed recruitment as an emerging haemodynamic phenotype in diabetes. It does not yet establish delayed glucose or lactate flux, oxidation or ATP regeneration during cognition in clinically characterised DCI (Table 2). Testing requires task-locked measurements, flexible response models and control of systemic physiology (23, 192).
6.3Substrate redistribution and candidate misallocation: relevance to task-dependent support
Metabolic support may be adequate in magnitude and timing yet fail to reach the compartment carrying the energetic cost. Misallocation denotes a mismatch between where carbon is delivered, retained or transformed and where neural computation occurs. It differs from global substrate scarcity and from inefficient oxidation after delivery. Normal brain metabolism already uses parallel, context-dependent routes. Sensory activation can preferentially increase neuronal glucose uptake (71), whereas hippocampal memory formation can depend on astrocytic glycogenolysis and monocarboxylate transport (12). Cell-resolved imaging has identified an astrocyte-to-neuron lactate gradient (77), and behavioural studies suggest that glucose and monocarboxylate dependence varies across regions and memory phases (47, 126). Neuronal activity can also recruit astrocytic glucose metabolism through adenosine signalling (193). Diabetes may therefore reweight several coexisting routes rather than abolish a single canonical shuttle. A claim of misallocation must specify the substrate, source, recipient, anatomical scale and demand window.
Metabolic redistribution or compartmental reorganisation alone should not be termed misallocation. In this framework, misallocation requires evidence that substrate remains available at an aggregate level but that task-relevant support to the cell, subcellular compartment or biochemical fate bearing the computational load is reduced. Evidence of where the displaced carbon is redirected would strengthen this inference. Without demonstrated compromise of task-relevant energetic or functional support, the more appropriate description is redistribution or metabolic reorganisation.
Available disease evidence is most consistent with reciprocal compartmental remodelling, not literal competition for a fixed glucose pool. Serial in vivo [1, 6-13C]glucose labelling in Goto–Kakizaki rats yielded model-based estimates of reduced neuronal tricarboxylic-acid-cycle activity, increased astrocytic activity and more strongly depressed glutamate–glutamine cycling (180). In db/db mice, acetate-derived labelling of TCA- and neurotransmitter-related metabolites decreased, whereas lactate-derived labelling increased (181). Ex vivo db/db brain slices also showed less glucose-derived labelling but greater hippocampal use of acetate and beta-hydroxybutyrate (162). Preserved glycogen abundance can similarly coexist with slower incorporation of glucose-derived carbon (122). Hyperpolarised pyruvate studies further indicate stage- and region-dependent pyruvate-to-lactate label exchange (194, 195). These signals do not directly measure competitive, cell-specific carbon flow. Their interpretation depends on tracer delivery, precursor enrichment, pool size, exchange kinetics and model structure.
The brain endothelium is a plausible redistribution site because it controls substrate entry while consuming glucose. Endothelial HIF-1α supports GLUT1 expression and cerebral glucose uptake (196), while endothelial MCT1-dependent transport contributes to hippocampal lactate homeostasis (197). In db/db mice, endothelial-enriched Hif1a knockdown reduced glycolytic features and hippocampal lactate accumulation while improving memory and neurogenesis without correcting systemic glycaemia (182). This supports a pathogenic contribution of endothelial-associated lactate handling, but not neuronal glucose deprivation. The intervention was not absolutely lineage restricted, lactate remained a bulk-tissue pool, and HIF-1α has pleiotropic vascular and inflammatory actions.
Microglia provide another candidate route for pathological redistribution. Suppressing microglial ketohexokinase-dependent fructose metabolism in db/db mice improved mitochondrial homeostasis, synaptic structure, CA1 physiology and cognition (183). Diabetes-associated microglial lactate accumulation has also been linked to H4K12 lactylation and FOXO1-PGC-1α-dependent oxidative stress, implicating lactate-linked signalling in an immunometabolic and epigenetic programme (198). In non-diabetic neuroinflammation, microglial reprogramming altered the local lactate-to-glucose environment while oxidative compensation maintained neuronal gamma activity (165). Conversely, physiological microglial signalling can promote vascular GLUT1 expression and neuronal energy support during activity (199). Microglia are therefore neither constitutive competitors nor passive bystanders. Direct evidence that diabetic microglia capture substrate at the expense of neuronal carbon entry remains absent.
Lactate further shows why abundance cannot identify destination. Chronic streptozotocin diabetes produces progressive hippocampal lactate accumulation and partial behavioural improvement after LDH-sensitive intervention (200), whereas acute exogenous lactate can improve recognition memory and sharp-wave-ripple amplitude in diabetic mice (172). During hypoglycaemia in type 1 diabetes, a much larger brain lactate pool was not accompanied by greater lactate-derived glutamate or glutamine labelling (9). These findings are compatible with chronic accumulation and acute network rescue coexisting. They do not identify the endogenous source, recipient cell, oxidative fate or ATP contribution of lactate.
Compensation and subcellular organisation further weaken inference from transporter abundance or regional averages. Partial astrocytic GLUT1 reduction can induce reprogramming that preserves energetics and cognition (201). Local ATP synthesis sustains active synapses (11), while stable dendritic mitochondrial compartments support nearby translation during plasticity (46). Neuronal GLUT3 deletion can reduce terminal glucose and ATP sensor signals despite comparatively preserved somatic measures and bulk FDG uptake (48). Outside diabetes, inflammatory-cell glucose consumption can also mask parenchymal deficits in regional FDG signals (202). Allocation must therefore be resolved at the relevant cellular and subcellular scale.
Current evidence supports substrate-, region- and cell-programme-dependent reorganisation of cerebral carbon handling in diabetes. Tracer models provide the strongest evidence for reciprocal neuronal and astrocytic remodelling, while endothelial and microglial interventions provide cell-associated causal bridges. However, no study aligns cognitive demand, cell-specific carbon flow, energetic state, neural computation and behaviour. Misallocation therefore remains a testable mechanism rather than an established universal or zero-sum phenotype.
6.4Malutilization: substrate-oxidation-function uncoupling
Malutilization begins where delivery ends. It denotes failure to convert an available substrate into useful oxidative and functional energy support, rather than reduced abundance or uptake alone. A molecule may enter brain tissue or exchange rapidly without reaching the mitochondrial matrix, engaging pyruvate dehydrogenase and the tricarboxylic acid cycle, driving coupled phosphorylation, or supplying ATP locally. The phenotype therefore differs from blunted recruitment, mistiming and misallocation, although all four may coexist. A defensible claim must identify the impaired conversion step and connect it to an energetic or functional consequence.
Normal activity recruits glycolysis, oxidative phosphorylation, local ATP synthesis and mitochondrial calcium-dependent regulation across workloads and compartments (11, 15, 72, 85, 86, 94). Static ATP or PCr cannot establish preserved turnover or delivery because high-energy-phosphate pools may remain stable while ATP-related flux changes (16).
Human studies reveal several dissociations, although none closes the pathway in clinically characterised DCI. Acute lipid-induced insulin resistance was associated with a lower hippocampal PCr/ATP ratio during sustained cognitive activity without measuring glucose oxidation or ATP synthesis (17). In type 1 diabetes, transient insulin deprivation creates systemic hyperglycaemia and ketosis while cerebral high-energy-phosphate measures, connectivity and selected cognitive functions deteriorate (203). Conversely, during hypoglycaemia with [3-13C]lactate infusion, a markedly larger calculated brain lactate pool was not accompanied by greater lactate-derived glutamate or glutamine enrichment (9). This supports a pool-use dissociation, not reduced absolute lactate oxidation. Other 13C studies found increased acetate transport and oxidation or preserved model-derived tricarboxylic-acid-cycle activity during hypoglycaemia (155, 156). Beta-hydroxybutyrate also contributed more directly to brain metabolism in type 1 diabetes, while medium-chain triglycerides supported working memory during clamped hypoglycaemia (161). Cerebral fuel use is therefore substrate and challenge specific.
Diabetic animal models provide stronger evidence for downstream energetic constraints. In chronically diabetic rats, hippocampal glucose accumulation coexisted with lower NAD+ availability, impaired mitochondrial respiration and reduced respiratory reserve. Nicotinamide mononucleotide improved respiration, neuronal survival and memory without substantially correcting systemic metabolic abnormalities (184). In streptozotocin-diabetic mice, cyclophilin D deletion protected mitochondrial respiration, ATP synthesis, long-term potentiation and memory (176). Manipulating GSK3β–Drp1 signalling likewise ameliorated complex-I, ATP and synaptic defects (175). High-fat-diet-induced insulin resistance reduced mitochondrial ATP-production capacity and oxidative-enzyme activity, which improved with exercise or metformin (185). Insulin withdrawal also reduced cerebral mitochondrial ATP production despite increased circulating monocarboxylates and MCT transcripts, with partial reversal after intranasal insulin (186). These findings support mitochondrial constraints downstream of substrate availability, but isolated-organelle capacity and behavioural rescue do not reconstruct an unmeasured task-evoked flux pathway.
Tracer studies indicate regional and cell-specific reconfiguration rather than uniform loss of use. Goto-Kakizaki rats show lower modelled neuronal tricarboxylic-acid-cycle activity with greater astrocytic activity (180), whereas db/db mice show substrate-specific labelling patterns (181). Hyperpolarised pyruvate signals vary by stage and region (194, 195), and experiments targeting pyruvate entry show that carbon reaching pyruvate or lactate need not enter mitochondrial oxidation or sustain presynaptic function (44, 204).
Lactate highlights the same inferential problem. Chronic streptozotocin diabetes enlarges hippocampal lactate pools and partially responds to LDH-sensitive intervention (200). Fibroblast growth factor 21 reduces lactate accumulation while increasing MCT2, LDHB activity, NAD+, ATP and memory, but did not directly measure lactate uptake or oxidation flux (187). Exogenous lactate can improve sharp-wave-ripple amplitude and recognition memory in diabetic mice (172). Recurrent-hypoglycaemia studies further show that increased lactate transport contributes little additional oxidation even when lactate preserves glucose metabolism and neuronal function (159). Lactate also modulates NMDA signalling and plasticity-gene expression through redox-dependent mechanisms (79). Rescue under these conditions does not identify an endogenous substrate deficit or prove ATP-mediated rescue.
Countervailing findings preclude a unitary failure model. Despite reduced glucose-derived labelling, db/db brain slices increase acetate and beta-hydroxybutyrate use, while isolated mitochondria show greater oxygen consumption and ATP-synthesis capacity (162). These adaptations may preserve energetics, but they do not establish adequate ATP delivery during cognition. Lactate can increase oxygen consumption while high-demand gamma activity and synaptic transmission deteriorate (164). During moderate neuroinflammation, altered local substrate conditions were instead compensated by greater oxidative metabolism that maintained gamma oscillations (165). Oxidation may therefore be insufficient, compensatory or effective according to location, timing, coupling efficiency and workload.
The evidence supports several possible breaks between substrate availability and function, but not a uniform failure of cerebral fuel use. Malutilization remains a substrate-, compartment-, stage- and demand-dependent phenotype until task-locked carbon fate, energetic state and function are linked within one design (Table 2).
6.5Disease stage, metabolic history, and reserve exhaustion
The stress-test framework is intended as a common experimental scaffold rather than a prediction that diabetes types, cognitive phenotypes or disease stages share the same challenge-response pattern. In type 1 diabetes, the available human evidence is weighted towards insulin treatment, recurrent hypoglycaemia, impaired awareness and alternative-substrate adaptation, whereas type 2 diabetes evidence more often reflects insulin resistance, obesity, vascular dysfunction and altered task- or insulin-evoked recruitment. These contexts may shift the operating point, the substrate available for recruitment and the node at which failure becomes detectable. Cognitive phenotype is likewise relevant because memory, executive, attentional and processing-speed deficits engage partially different circuits and may therefore expose different regional or temporal vulnerabilities. Disease stage may further alter the direction of the observed response: preserved or increased recruitment can coexist with vulnerability at one stage, whereas blunted, mistimed or inefficient responses may emerge under greater demand or accumulated metabolic burden. These possibilities should not be interpreted as fixed sequential stages or diabetes-type-specific signatures; the framework instead requires challenge-response patterns to be defined within the relevant diabetes type, cognitive domain, metabolic history and disease context.
Neurometabolic phenotypes in diabetes are contingent rather than fixed. They reflect disease duration, cumulative glycaemic exposure, treatment, antecedent hypoglycaemia, vascular burden, substrate availability and imposed demand. Metabolic reserve is operationally the additional delivery, pathway flux and energetic output recruitable above baseline within a physiologically relevant window, together with post-challenge recovery. Reserve exhaustion would require progressive contraction of this increment that predicts neural or cognitive failure. No study has demonstrated this trajectory end to end in clinically characterised DCI. The concept remains an interpretive model, not a synonym for low resting metabolism or an established natural history.
Early diabetes does not shift cerebral metabolism uniformly. During euglycaemic hyperinsulinaemia, lower peripheral insulin sensitivity was associated with greater brain FDG uptake (27), whereas adults with prediabetes or early type 2 diabetes showed lower resting metabolism in Alzheimer-vulnerable regions, diffuse encoding responses and poorer delayed recall (123). These distinct operating conditions preclude a simple hypermetabolism-versus-hypometabolism contrast. After shorter high-fat-diet exposure, increased hyperpolarised pyruvate-to-lactate exchange preceded differences in the selected water-maze outcomes; with longer exposure, it coexisted with reduced 13C delivery, regional pyruvate-dehydrogenase suppression and memory impairment (194). Streptozotocin diabetes also produced stage-dependent glucose- and acetate-labelling patterns (205), whereas an intermediate stage showed enlarged high-energy-phosphate pools with largely preserved cognition (206). Each pattern is compatible with reconfiguration, but none quantifies spare capacity.
Longitudinal evidence supports asynchronous, partly reversible change rather than a single decline. Serial MRS studies show stage- and region-specific metabolite trajectories in streptozotocin and db/db models (207, 208), including progressive hippocampal lactate accumulation (200, 209). Other staged observations include loss of insulin-mediated microvascular recruitment after prolonged high-fat feeding (138) and an association between two-year perfusion decline and functional deterioration (210). These studies establish time dependence, but most did not repeat the same demand-evoked flux across time.
Late-stage findings nevertheless identify plausible reserve limits. Diet-induced diabetes reduced hippocampal glycogen and abolished its net response to local insulin, although one endpoint cannot resolve synthesis, breakdown or task-related mobilisation (19). High-fat-diet-induced insulin resistance lowered mitochondrial ATP-production capacity and oxidative-enzyme activity in insulin-responsive regions (185), whereas chronic experimental diabetes can reduce NAD+ availability, respiratory reserve and memory (184). Modelled 13C data in Goto–Kakizaki rats also indicated lower neuronal tricarboxylic-acid-cycle activity and glutamate–glutamine cycling (180). Counterevidence includes db/db brain slices that increased acetate and ketone use while isolated mitochondria showed greater oxygen consumption and ATP-generating capacity (162). Brain glucose responses to matched hyperglycaemia improved with better glycaemic control (98), and Akita mice retained, and sometimes prolonged, cortical neurovascular responses during progression (169). Reserve exhaustion is therefore neither universal nor necessarily irreversible; it depends on substrate, compartment, challenge and stage.
Current glycaemia cannot fully represent disease state. Over 32 years in DCCT/EDIC, cumulative glycaemic exposure, severe hypoglycaemia and vascular risk independently contributed to later cognitive decline in type 1 diabetes (211). Longer type 2 diabetes duration and poorer control predicted faster decline in Whitehall II (212), whereas midlife diabetes was associated with greater decline over two decades in ARIC (3). Decade-long time-weighted HbA1c was associated with later executive decline and brain-volume loss (213), and visit-to-visit glucose and HbA1c variability predicted Alzheimer disease risk beyond mean exposure (214). These associations do not establish a persistent neuronal metabolic programme and may reflect vascular injury, comorbidity, treatment or reverse causation. A four-year cohort found moderate but not accelerated decline (215), while intensive glucose lowering in ACCORD MIND altered brain-volume loss without improving its primary cognitive outcome (216). Glycaemic history shapes risk and response but is not a deterministic measure of cerebral reserve.
Antecedent hypoglycaemia provides strong evidence that metabolic history can reset subsequent responses. Recent exposure attenuates autonomic, symptomatic and glucose-defence responses, while sustained avoidance can partly restore awareness (217–219). Cerebral glucose responses vary with awareness and prior control, although repeated clamps do not consistently alter modelled transport kinetics (146, 147). Across rodent and human studies, recurrent hypoglycaemia alters task-associated glucose dynamics and context-dependent use of acetate, lactate and glycogen (9, 142, 155, 156, 159, 220–222). These adaptations demonstrate history-dependent resetting, not longitudinal exhaustion of cognitive neurometabolic reserve.
Adaptive recruitment may preserve function at one workload while narrowing the margin at another. The most defensible trajectory is therefore non-monotonic: diabetes may shift the operating point, redistribute activity or recruit alternative substrates before overt failure, followed by stabilisation, further contraction or partial recovery (149, 161, 223). Only a progressive loss of challenge-evoked energetic increment that precedes and predicts functional failure should be termed reserve exhaustion.
7Testing activity-dependent neurometabolic failure in DCI: a falsifiable research agenda
Testing activity-dependent neurometabolic failure requires a staged strategy that moves from phenotype definition to temporally aligned measurement, source–recipient–fate resolution, causal perturbation and longitudinal validation (Figure 5). Mechanistic support requires verified engagement of the predicted metabolic node within the relevant challenge window and evidence that downstream function depends on that node, whereas reserve exhaustion can be established only through repeated within-person assessment of dynamic response trajectories.
7.1Operationalizing the four phenotypes as discriminating predictions
A useful mechanistic phenotype must distinguish the proposed abnormality from plausible alternatives. The labels are assigned to the response dimension and support node directly measured; they are not mutually exclusive causal diagnoses. A challenge may support more than one phenotype at different nodes, and a downstream defect may secondarily alter an upstream response. Causal ordering therefore requires temporally aligned upstream and downstream measurements or targeted perturbation. Each challenge should prespecify demand, neural engagement, baseline state, the relevant compartment and the response dimension being tested. Because transporter abundance, metabolite concentration, pathway rate, high-energy-phosphate turnover and neural function occupy distinct inferential levels (80, 95, 96, 107), every phenotype requires a defined node, a positive criterion and an observation that would refute it.
Blunted recruitment denotes reduced response gain at a specified node after demand has been matched. Cognitive demand produces graded hippocampal glucose responses, while neural activity recruits local ATP synthesis and alters cerebral ATP metabolic rate (10, 11, 16). A low resting value cannot establish blunting. Preserved challenge-evoked amplitude and input-output slope, after accounting for neural engagement, systemic substrate supply and baseline floor or ceiling effects, argue against blunting at the measured node, although failure may remain downstream.
Mistimed recruitment is a kinetic misalignment. Delayed onset, a displaced peak, premature termination or incomplete recovery may preserve peak amplitude or the integrated response yet miss the neural event requiring support. Neurovascular signals evolve across vascular compartments and have intrinsic temporal transfer characteristics (50, 63, 69), while rigid haemodynamic models can convert latency differences into apparent amplitude effects (23). Testing therefore requires a common event clock, adequate sampling and independent neural timing. Equivalent onset, peak latency, duration, recovery and phase relation argue against mistiming.
Candidate substrate misallocation requires three observations: adequate aggregate substrate input or availability, reduced allocation to the task-relevant cell, subcellular compartment or biochemical fate, and compromised energetic or functional support at that site. Cell-resolved uptake and carbon-fate measurements can distinguish parallel routes (15, 48, 71, 77), but current DCI studies do not satisfy all three conditions. Evidence of a reciprocal destination for displaced carbon would strengthen the inference. Redistribution without functional compromise should not be classified as misallocation. Preserved input-corrected allocation, or an apparent shift explained by precursor dilution, pool size, cell composition or regional averaging, argues against misallocation.
Malutilization occurs after substrate delivery: entry is relatively preserved, but oxidation, ATP support or neural output is disproportionately impaired. Lactate accumulation without greater oxidation illustrates why availability cannot be equated with productive use (9), whereas perturbing mitochondrial pyruvate transport links a downstream bottleneck to oxidative ATP failure and impaired presynaptic function (44). Testing must pair entry with oxidation or energetic state and an independent neural endpoint. A deficit explained by insufficient upstream supply should not be assigned to malutilization. These four phenotypes may coexist at identified nodes; reserve exhaustion instead denotes longitudinal contraction of recruitable capacity. Table 3 links these discriminating criteria to current direct observations, indirect support, major competing explanations and the minimum measurements needed to distinguish each candidate phenotype.
Table 3Discriminating criteria linking candidate neurometabolic failure phenotypes to current evidence and competing explanations.
Candidate phenotype Direct observations Indirect support and limiting evidence Major competing explanations and overlap Minimum measurements needed to discriminate
Blunted recruitment Reduced challenge-evoked response gain at a specified support node relative to effective neural demand. Smaller task-evoked cerebrovascular responses and smaller hyperglycaemic brain-glucose pool responses have been observed at selected human nodes (97, 98, 100). Demand-associated energetic and regional insulin-response abnormalities provide indirect support (17–19, 137, 138). Preserved peak flow, increased insulin-stimulated FDG uptake and greater task activation argue against a universal blunted phenotype (20, 24–30, 130, 169). Reduced neural engagement; altered systemic substrate or physiology; baseline floor or ceiling effects; vascular limitation; mistiming; or downstream malutilization that secondarily reduces recruitment. Graded challenge; achieved performance; independent neural-engagement measure; systemic substrate and physiology; baseline-referenced, time-resolved response at the claimed node; downstream energetic or functional readout when causal sufficiency is claimed.
Mistimed recruitment Altered onset, peak latency, duration or recovery relative to the neural event requiring support. Human and animal studies show delayed, abbreviated or prolonged haemodynamic responses, sometimes with preserved peak or integrated magnitude (20–22, 125, 169, 179). Phase-specific neural and network abnormalities are compatible with temporal vulnerability but do not establish metabolic mistiming (132, 171, 189). Altered neural timing; modality-specific delay; vascular transit or reactivity; rigid response modelling; anaesthesia, stimulus or systemic physiological differences. Shared event clock; independent neural timing; time-resolved vascular or metabolic readout; flexible response modelling; modality-specific delay correction; onset, peak, duration and recovery sampling.
Candidate substrate misallocation No study currently demonstrates task-locked carbon diversion that both preserves aggregate availability and compromises support of the active compartment in clinically characterised DCI. Tracer modelling, ex vivo substrate studies and cell-associated perturbations support compartmental metabolic reorganisation (162, 180–183, 194, 195). Altered delivery; precursor enrichment; dilution; pool-size differences; cell composition; regional averaging; or adaptive redistribution without functional cost. Redistribution alone is not misallocation. Controlled precursor input; cell- and compartment-resolved carbon fate; task-relevant neural, energetic or functional readout; preferably a reciprocal destination or carbon-balance-consistent account of displaced carbon.
Malutilization Relatively preserved substrate entry or availability with impaired downstream oxidation, energetic state or neural function during the same challenge. A larger lactate pool without greater lactate-derived labelling is a direct pool–use dissociation, but not a complete cognitive-challenge phenotype (9). Mitochondrial and functional perturbation studies support downstream constraints (44, 175, 176, 184–187), whereas preserved or increased alternative-fuel use and mitochondrial capacity limit universal claims (156, 162). Upstream delivery failure; lower effective neural demand; substrate switching; compensatory oxidation; assay-capacity effects; or pleiotropic intervention effects. Input-corrected entry or uptake; downstream oxidation or ATP/redox dynamics; independent neural or cognitive endpoint during the same challenge; perturbation or rescue when causality is claimed.
7.2Aligning cognitive demand, metabolic flux, energetic state, and neural function
Activity-dependent neurometabolic failure is relational: metabolic support must be judged against neural demand during the same biological episode. Resting values, task responses obtained on another day and terminal assays may reveal coexisting abnormalities, but cannot establish demand-energy uncoupling. The strongest designs align demand and neural engagement, a defined metabolic response, energetic state, and neural or behavioural function within one challenge window (80, 107, 224). Simultaneous acquisition is preferable, although matched within-person sessions remain informative when technical constraints preclude it.
The framework is modular rather than all-or-none. A study need not resolve the complete demand-to-function pathway to make a valid node-specific claim. Technical feasibility may legitimately constrain the number of modalities, provided that inference is restricted to the measured node and that major competing explanations are either measured or explicitly acknowledged. The minimum measurements specified in Table 3 should therefore be interpreted as claim-specific requirements, not as a requirement that every study acquire the entire multimodal chain.
Imposed task load should be specified, achieved performance quantified, and task-relevant neural engagement measured independently where feasible. Together, these measures characterise effective demand but do not constitute a validated measure of energetic cost. Load-dependent task-evoked responses in diabetes vary across cohorts and glycaemic conditions (129–131), so a shared task label does not establish a shared biological demand. A practical design should retain a prespecified fixed-load series and, where feasible, add individually calibrated conditions. Fixed loads test a common external requirement, whereas calibration can express load relative to estimated capacity or target a prespecified performance range. The manipulated parameter, such as set size or retention interval, and sensory-motor requirements should be specified. Evidence outside diabetes shows that expressing load relative to individual working-memory capacity can materially alter apparent group differences in activation (225). Both the absolute workload and achieved performance should therefore be reported. Accuracy alone is insufficient; reaction time and its variability, errors, omissions, completed trials, strategy and fatigue should be documented across the challenge. Performance-matched or correct-trial analyses should complement, rather than replace, the complete load-performance curve. Behavioural and neural summaries should cover the same acquisition window as block-integrated metabolic measurements. Importantly, trial-level behaviour alone does not confer trial-level metabolic resolution.
Neural engagement should be assessed independently of the metabolic or vascular endpoint whenever feasible. Task-relevant EEG or MEG can be used in humans, whereas electrophysiology can provide aligned neural readouts in experimental models (68, 132). Each measure should target the same circuit, task phase and acquisition window as the metabolic endpoint. Conventional BOLD provides complementary spatial information but should not serve as the sole measure of both neural demand and the vascular response under evaluation (68). Cerebrovascular reactivity should also be characterised, because adjustment can alter task-activation associations in type 2 diabetes (101). Blunting should therefore be evaluated from baseline-referenced response curves and response gain across loads, conditional on systemic input and measured neural engagement. A smaller response is a candidate blunting phenotype only when it is disproportionate to effective demand. Energetic and functional readouts are needed to assess whether altered recruitment is associated with insufficient support at the tested workload; preserved performance alone establishes neither intact reserve nor compensation. Conversely, concurrent reductions in neural and metabolic signals remain directionally ambiguous. Reduced neural engagement can lower metabolic demand, whereas inadequate metabolic support can itself constrain neural recruitment (80, 107, 224). Performance and neural activity should therefore be analysed as both demand indicators and possible downstream outcomes, rather than automatically adjusted away.
The metabolic link must also be specified. CBF indexes perfusion, whereas delivery additionally depends on arterial input and transvascular exchange. FDG-fPET integrates delivery, transport and phosphorylation-dependent trapping; tracer kinetics can estimate defined pathway rates; and CMRO2 addresses oxidative metabolism. Simultaneous BOLD-fMRI and FDG-fPET can map haemodynamic and glucose-related responses during a shared task (36, 226), yet BOLD overactivation may occur without greater regional glucose metabolism (227). Glucose utilisation, oxygen consumption and CBF can also change by different fractions, even where CBF and CMRO2 are approximately coupled under defined conditions (65, 66). Multimodal discordance therefore localises an unresolved link rather than allowing one readout to represent the whole pathway.
Temporal alignment must accommodate each modality’s response function. Electrical activity changes within milliseconds, haemodynamic signals over seconds, and FDG or many spectroscopic readouts over longer integration periods. Calibrated fMRI can estimate task-evoked CMRO2 under explicit physiological and modelling assumptions (109, 228–230), whereas inflexible haemodynamic models may misclassify altered latency as altered amplitude (23). Baseline, onset, sustained-demand and recovery windows should be prespecified, with modality-specific delays modelled and both magnitude and kinetics retained.
A feasible minimum human design can be illustrated for blunted cerebrovascular recruitment, for which challenge-based human evidence is comparatively direct (100). Adults with T2D and objectively characterised DCI could be compared with cognitively unimpaired adults with T2D during a prespecified graded working-memory challenge. Essential measurements would include accuracy and reaction time, task-relevant neural engagement assessed by EEG (68, 132), task-evoked middle-cerebral-artery flow-velocity responses measured by transcranial Doppler (100), and concurrent glucose, blood pressure and end-tidal CO2, with cerebrovascular reactivity characterised separately (101, 192, 231). The primary test would be the gain of the vascular response across increasing effective demand. Simultaneous acquisition should be used where feasible; tightly matched within-person sessions remain acceptable when technical constraints preclude it, provided that the loss of temporal alignment is acknowledged.
Blunted recruitment would be supported by a reduced vascular response slope despite comparable or explicitly modelled neural engagement, achieved performance, systemic input and cerebrovascular reactivity. The hypothesis would be weakened if the difference were explained by lower neural engagement, baseline or cerebrovascular-reactivity differences, systemic physiology, or a latency shift rather than reduced response gain. Because transcranial Doppler measures arterial velocity rather than regional tissue perfusion, this design would support only node-specific cerebrovascular blunting. Preserved vascular response gain would argue against blunting at this node but would not exclude downstream substrate-delivery, oxidative or energetic failure. The design would not establish causal mediation, reserve exhaustion or therapeutic benefit.
7.3Resolving cell- and compartment-specific carbon allocation
Regional glucose uptake, lactate concentration or ATP content cannot identify cellular recipients or carbon fate. Allocation therefore requires a source-recipient-fate account. Isotopic enrichment informs flux only when precursor input, pool size, sampling kinetics and a suitable model are incorporated (80, 95, 96, 107). Dynamic ¹³C MRS and isotopologue analysis can trace labelled carbon through glycolysis, pyruvate carboxylation, tricarboxylic-acid-cycle turnover and neurotransmitter exchange, but quantitative neuronal and astroglial rates remain model-dependent (232, 233). Misallocation requires reduced task-relevant allocation plus a reciprocal destination or carbon-balance-consistent loss; reduced enrichment alone remains compatible with altered delivery, dilution, efflux or cell composition.
Normal-brain studies caution against imposing one obligatory route on every activity state. Activation can recruit rapid neuronal glucose uptake (71), while cell-targeted measurements support an astrocyte-neuron lactate gradient whose magnitude depends on calibration and intracellular metabolism (77). Other experiments detect neuronal glycolysis without increased lactate uptake (15), and selective interference with neuronal glucose transport or glycolysis impairs network function in vivo (48). A DCI-associated shift must therefore be defined relative to a challenge, circuit and carbon fate; transporter abundance in one cell class cannot show carbon capture at another cell’s expense.
Cell-targeted sensors provide temporal resolution that bulk tracing lacks. Glucose sensors track intra- or extracellular dynamics (234, 235), and lactate sensors report rapid cellular changes (236), and across neuronal and astrocytic pools (77). Their signals depend on affinity, pH, intracellular conversion, calibration and cell volume (237) and report local state rather than carbon provenance or net flux. Combining them with labelled precursors and isotopologue analysis is therefore more informative. Subcellular ATP sensors (238), presynaptic measures of local ATP synthesis (11)and mitochondrial pyruvate transport (44), spatial isotope imaging (239)and single-cell neurometabolomics (240) can further resolve heterogeneity. Their platform-specific limitations, including sensor calibration, sampling delay, cell purity and ex vivo turnover, require appropriate controls and orthogonal validation.
Current diabetes studies resolve only fragments of this chain. Compartmental ¹³C NMR in db/db mice suggests altered neuron-astrocyte cooperation (181), while ex vivo analyses show reconfigured energetics and ketone metabolism (162). In humans, the brain-glucose response to hyperglycaemia is blunted in obesity and type 2 diabetes (97), and lactate can accumulate during hypoglycaemia without greater oxidation of blood-derived lactate (9). These findings support altered substrate handling, not task-specific cellular misallocation. Deuterium metabolic imaging and hyperpolarised ¹³C MRI provide regional substrate-metabolism signals (241, 242). position-specific hyperpolarised tracing adds exchange, redox and downstream labelling information but remains sensitive to delivery and modelling (108). Dynamic FDG-fPET is task-sensitive but cannot resolve later carbon fate or cellular identity (32). Cross-species designs should apply the same challenge, using human imaging for regional timing and animal experiments for cell-targeted fate mapping without borrowing cellular resolution for the human signal.
A minimum test of misallocation therefore requires a quantified precursor input, cellular and relevant subcellular attribution, downstream carbon fate, independent circuit engagement, and a destination or loss route accounting for missing carbon. Preserved input-corrected allocation within adequate sensitivity and model identifiability would refute misallocation despite altered regional pools or transporter expression.
7.4Establishing causality through perturbation, rescue, and mechanistic mediation
A metabolic abnormality accompanying cognitive impairment need not cause it. Diabetes alters vascular reactivity, insulin signalling, inflammation, substrate handling and mitochondrial function in parallel (5). A causal claim requires a directional model linking an exposure or intervention to a prespecified metabolic node, then to energetic or neural function and cognition, while alternative paths remain explicit (243–245). The candidate defect must precede functional failure within the relevant challenge interval, and perturbations should be restricted by cell type, circuit, compartment and time where feasible. Target engagement must be shown at the proposed node, with systemic glucose, insulin, perfusion, arousal and motor performance treated as competing routes (107).
Necessity and sufficiency answer different causal questions. Necessity is supported when selective disruption of a candidate process prevents the normal challenge-evoked energetic response and impairs the corresponding neural function. Activity-resolved ATP imaging has shown that local ATP synthesis is required to sustain presynaptic vesicle cycling (11), and activity-dependent GLUT4 mobilisation supports active synapses (72). Neuronal GLUT3 or pyruvate-kinase deletion tests glucose uptake and glycolysis in vivo (48), hippocampal GLUT3 knockdown links glucose uptake to memory acquisition (126), and mitochondrial pyruvate-entry manipulation locates a defect downstream of cytosolic availability (44). These are causal precedents, not evidence for the same lesion in DCI.
Pathway-positioned rescue can order successive steps. Brain glycogen-synthase deletion impairs plasticity and long-term memory (76). Lactate restores memory after astrocytic monocarboxylate-transport disruption but not neuronal MCT2 knockdown (12), while pyruvate or β-hydroxybutyrate can bypass inhibited glycogenolysis or MCT1/4 loss but not deficient neuronal import (47). This epistasis supports downstream energetic sufficiency rather than an irreplaceable role for lactate. Cell-restricted deletion of astrocytic adenosine A2B receptors further illustrates circuit-specific testing of demand sensing and metabolic support (193). Rescue remains ambiguous unless target engagement, energetic recovery and neural output are verified.
Mechanistic mediation cannot be reduced to one statistical indirect effect. Even after treatment randomisation, the mediator is not randomised, and estimates depend on mediator-outcome confounding, exposure-mediator interaction, post-treatment confounders and model specification (243–245). Biological mediation instead requires temporal and experimental convergence: the intervention changes the metabolic node in the relevant window; proximal energetic and neural responses follow; selective blockade attenuates benefit; and a correctly positioned rescue restores it. Perfusion, inflammation and mitochondrial function should remain explicit competing or co-mediating paths.
Diabetes models provide an uneven spectrum of causal resolution. Hippocampal insulin-receptor knockdown impairs plasticity and memory without establishing glucose-flux or ATP mediation (177), whereas myeloid-derived VEGF preserves blood-brain barrier glucose transport and cognition during obesity, consistent with a compensatory response (246). Partial Drp1 normalisation improves mitochondrial and synaptic phenotypes (175), but complete neuronal ablation worsens injury (247), making dose and timing decisive. Cyclophilin-D deletion connects mitochondrial permeability, ATP, plasticity and memory more directly (176). PDK1 augmentation is protective without direct evidence for altered neuronal pyruvate-oxidation flux (248). LDHA inhibition limits hippocampal lactate accumulation and memory loss (200), microglial ketohexokinase deletion restores mitochondrial, synaptic and cognitive phenotypes (183); and brain-endothelial HIF-1α suppression counteracts glycolysis-driven lactate production and DCI (182). These studies move beyond association, but differences in developmental compensation, systemic exposure, challenge timing and direct flux measurement still limit how narrowly the rescued pathway can be assigned.
Broad interventions and human studies offer translational relevance but weaker mediation evidence. Exercise and metformin improve brain mitochondrial function without measuring cognition; insulin manipulation changes ATP-related phenotypes (185). Insulin deficiency and intranasal insulin alter brain mitochondrial ATP-related phenotypes, but lactate and ketone use may represent compensation rather than the primary defect (186), and FGF21 improves lactate-use markers and cognition without testing pathway necessity (187). Metabolic management can normalise brain glucose kinetics without proving cognitive mediation (98), while transient insulin deprivation alters energetics, connectivity and cognition amid systemic changes (203). Intranasal insulin trials report parallel changes in cerebral blood flow and selected functional outcomes (249), while systematic evaluation leaves the perfusion-to-cognition relation unresolved (250). These findings establish manipulability and translational relevance, not a completed metabolic mediation chain.
A minimum falsifiable design combines disease-by-perturbation with rescue or occlusion during the same challenge, verifies the target node, and measures an adjacent energetic response, neural function and behaviour. The model is weakened when energetic correction fails to restore function, benefit survives mediator blockade, or functional recovery precedes metabolic recovery. Accordingly, normalisation of a challenge-response abnormality should not be interpreted as cognitive rescue unless the intervention also improves the relevant cognitive outcome and evidence supports a causal contribution of the targeted node to that benefit (243–245).
7.5Longitudinal stress testing in clinically characterised DCI
The longitudinal approach outlined here is intended to evaluate neurometabolic reserve within this experimental framework; any future diagnostic application would require separate clinical validation. Reserve exhaustion is not a fifth phenotype within one challenge. It denotes longitudinal contraction of the response range recruited above baseline: under matched demand and neural engagement, the same individual shows a smaller response, reaches a ceiling at a lower workload or cannot sustain function. A low response at one visit may instead reflect age, engagement, vascular limitation, treatment, acute metabolic state or measurement error. This distinction is essential in heterogeneous DCI (5) and because resting measurements cannot define activation capacity (80).
Longitudinal epidemiology establishes the importance of stage without demonstrating reserve loss. Midlife diabetes has been associated with greater cognitive decline over two decades (3), and established disease is associated with less favourable trajectories than incident diabetes (251). Dysglycaemia is also linked to cognitive decline, white-matter lesions and brain-volume loss (252, 253), with earlier onset associated with greater dementia risk (254). Metabolic history should therefore include mean exposure, glucose variability (255, 256), recurrent severe hypoglycaemia (257–260), disease duration and treatment. These factors may shift the operating point and adaptive range but do not define one causal route; cognitive impairment can itself increase hypoglycaemia risk (258).
A reserve measure requires a response curve rather than one peak. At each visit, calibrated low- and high-demand conditions should align neural engagement, metabolic recruitment, energetic state and performance. Gain, latency, integrated response, recovery and highest sustainable workload can then be estimated. Challenge-sensitive hippocampal energetic changes after acute insulin resistance establish proof of principle (17), while reduced hippocampal glycogen and insulin-evoked mobilisation in diet-induced diabetes nominate stored substrate as one component of reserve (19), consistent with blunted hippocampal metabolic and cognitive responses to insulin (18). None demonstrates longitudinal contraction in clinically characterised DCI.
Clinical characterisation should include diabetes type, age at onset, duration, treatment, vascular and renal disease, obesity, blood pressure, depression, sleep, relevant medication, domain-level cognition and functional status.Challenge selection and interpretation should therefore be stratified by diabetes type and cognitive phenotype rather than assuming a common reference response across heterogeneous DCI. Memory-dominant, executive-attentional or more global cognitive phenotypes may require different task circuits and readouts, while stage comparisons should distinguish altered operating points from true longitudinal contraction of recruitable capacity. Structural MRI can quantify vascular injury and atrophy, with Alzheimer biomarkers added when mixed pathology is likely. Follow-up task difficulty must be recalibrated to maintain comparable neural demand. Continuous glucose monitoring with ecological cognitive testing already resolves person-specific glucose-performance relationships in type 1 diabetes (261, 262), challenge-locked brain metabolism and neural activity are needed to extend this approach to reserve.
Repeated response-curve testing may help distinguish an inaccessible reserve from a progressively contracting one. Improved glycaemic control was accompanied by restoration of the brain-glucose response in a small within-person type 2 diabetes study (98), yet intensive glucose lowering did not improve cognition in ACCORD MIND (216). Intranasal insulin altered perfusion and selected functional outcomes (249), diet reversal normalised regional metabolites and memory in mice (208), and a DCI-enriched multidomain trial had a neutral primary cognitive result (263). These findings are consistent with plasticity or co-recovery, not expansion of demand-evoked capacity. Failure to recover cannot prove irreversible exhaustion without adequate target engagement, treatment duration and consideration of alternative constraints.
Repeated testing can mimic decline or recovery. Practice effects vary with interval and test form (264–266), and strategy, fatigue or habituation can alter activation independently of capacity. Task-fMRI also has limited individual-level reliability (267). Alternate forms, familiarisation, adaptive difficulty, repeated baselines and reliable-change methods are therefore needed (264, 268), together with standardisation of systemic physiology. Reliability must be established for the actual task and platform: resting ASL flow may be reproducible when task-evoked change is less stable (269), static proton-MRS reproducibility does not guarantee reliable dynamic responses (270), and task-specific FDG quantification requires acquisition and kinetic analysis validated for the intended temporal contrast (33). Reserve exhaustion is supported only when within-person contraction exceeds these errors and persists after matching demand and systemic state.
Three potential uses of the stress-test framework should be distinguished. Mechanistic investigation asks whether a challenge-defined abnormality identifies a causal limiting node. It therefore requires targeted perturbation or rescue with verified target engagement. Prognostic assessment asks whether a baseline response phenotype or its longitudinal trajectory predicts subsequent cognitive decline beyond established clinical risk factors. It therefore requires prospective validation and reproducible measurement. Treatment selection is a predictive-biomarker question and requires a stronger standard. It requires prospective evidence that the baseline stress-test phenotype modifies the effect of an intervention targeting the implicated pathway, ideally through a prespecified treatment-by-phenotype interaction in a randomised trial. None of these clinical applications is currently validated. The present framework should therefore be regarded primarily as a mechanistic and experimental research tool; prognostic and treatment-selection applications remain unvalidated possibilities for future study.
8Conclusion
DCI is commonly interpreted through chronic abnormalities in glucose availability, insulin signalling, vascular function or mitochondrial metabolism. Resting measurements of these processes, however, do not establish whether the brain can coordinate energy supply when neural activity increases. Within this conceptual and experimental framework, cognitive demand is proposed as a neurometabolic stress test to investigate adaptive limitations that may not be apparent from resting perfusion, metabolite concentrations or high-energy-phosphate pools. The relevant question is not simply how much substrate is present, but whether vascular recruitment, substrate delivery, cellular allocation, oxidation, phosphotransfer and neural function are matched in magnitude, timing and location within the same challenge window.
The available evidence supports individual components of this model but does not yet establish an uninterrupted causal chain from cognitive demand to metabolic flux, energetic state, neural computation and cognitive performance in clinically characterised DCI. Current findings are compatible with four non-exclusive response phenotypes: blunted recruitment, mistimed recruitment, substrate misallocation and malutilization. These phenotypes should not be regarded as universal features or discrete disease subtypes. Their expression may vary across brain regions, disease stages and metabolic histories, and several may coexist within the same individual. These within-challenge patterns do not establish reserve exhaustion, which requires repeated within-person evidence of a progressively contracting response range under comparable demand and physiological conditions; one attenuated response is insufficient. Likewise, preserved resting measures cannot exclude reduced neurometabolic reserve, whereas altered metabolite pools cannot, by themselves, identify the underlying transport, turnover or oxidative process.
Resolving these uncertainties will require challenge-locked experimental designs that align cognitive demand with neural activity, vascular recruitment, substrate flux, energetic support and functional output. Cell- and compartment-resolved measurements, combined with targeted perturbation, rescue and longitudinal testing, will be essential for distinguishing causal failure from compensation or downstream consequence. This approach shifts the field from describing static metabolic abnormalities towards testing whether the diabetic brain can mount an appropriately timed and spatially directed energetic response. Activity-dependent neurometabolic failure remains a testable contributor to DCI rather than an established universal mechanism. The proposed neurometabolic stress test provides a conceptual and experimental framework for testing this proposition, not a validated diagnostic, prognostic or treatment-selection tool. Its central question is no longer simply how much fuel is present in the diabetic brain, but whether that fuel can be mobilised, routed and used quickly enough, in the right place, to sustain cognition.
References
1Xue M Xu W Ou YN Cao XP Tan MS Tan L .
Diabetes mellitus and risks of cognitive impairment and dementia: a systematic review and meta-analysis of 144 prospective studies . Ageing Res Rev . (2019 ) 55 :100944. doi: 10.1016/j.arr.2019.100944 31430566
2Chatterjee S Peters SA Woodward M Mejia Arango S Batty GD Beckett N .
Type 2 diabetes as a risk factor for dementia in women compared with men: a pooled analysis of 2.3 million people comprising more than 100,000 cases of dementia . Diabetes Care . (2016 ) 39 :300–7.
10.2337/dc15-1588 PMC4722942 26681727
3Rawlings AM Sharrett AR Schneider AL Coresh J Albert M Couper D .
Diabetes in midlife and cognitive change over 20 years: a cohort study . Ann Intern Med . (2014 ) 161 :785–93. doi: 10.7326/m14-0737 25437406
PMC4432464
4Palta P Schneider ALC Biessels GJ Touradji P Hill-Briggs F .
Magnitude of cognitive dysfunction in adults with type 2 diabetes: a meta-analysis of six cognitive domains and the most frequently reported neuropsychological tests within domains . J Int Neuropsychol Soc . (2014 ) 20 :278–91. doi: 10.1017/s1355617713001483 24555960
PMC4132660
5Biessels GJ Despa F .
Cognitive decline and dementia in diabetes mellitus: mechanisms and clinical implications . Nat Rev Endocrinol . (2018 ) 14 :591 –604 . doi: 10.1038/s41574-018-0048-7 30022099
PMC6397437
6Biessels GJ Nobili F Teunissen CE Simó R Scheltens P .
Understanding multifactorial brain changes in type 2 diabetes: a biomarker perspective . Lancet Neurol . (2020 ) 19 :699 –710 . doi: 10.1016/s1474-4422(20)30139-3 32445622
7Geijselaers SLC Sep SJS Stehouwer CDA Biessels GJ .
Glucose regulation, cognition, and brain MRI in type 2 diabetes: a systematic review . Lancet Diabetes Endocrinol . (2015 ) 3 :75 –89 . doi: 10.1016/s2213-8587(14)70148-2 25163604
8Arnold SE Arvanitakis Z Macauley-Rambach SL Koenig AM Wang HY Ahima RS .
Brain insulin resistance in type 2 diabetes and Alzheimer disease: concepts and conundrums . Nat Rev Neurol . (2018 ) 14 :168–81. doi: 10.1038/nrneurol.2017.185 29377010
PMC6098968
9De Feyter HM Mason GF Shulman GI Rothman DL Petersen KF .
Increased brain lactate concentrations without increased lactate oxidation during hypoglycemia in type 1 diabetic individuals . Diabetes . (2013 ) 62 :3075–80. doi: 10.2337/db13-0313 23715622
PMC3749358
10McNay EC Fries TM Gold PE .
Decreases in rat extracellular hippocampal glucose concentration associated with cognitive demand during a spatial task . Proc Natl Acad Sci USA . (2000 ) 97 :2881–5. doi: 10.1073/pnas.050583697 10706633
PMC16024
11Rangaraju V Calloway N Ryan TA .
Activity-driven local ATP synthesis is required for synaptic function . Cell . (2014 ) 156 :825–35. doi: 10.1016/j.cell.2013.12.042 24529383
PMC3955179
12Suzuki A Stern SA Bozdagi O Huntley GW Walker RH Magistretti PJ .
Astrocyte-neuron lactate transport is required for long-term memory formation . Cell . (2011 ) 144 :810–23. doi: 10.1016/j.cell.2011.02.018 21376239
PMC3073831
13Dembitskaya Y Piette C Perez S Berry H Magistretti PJ Venance L .
Lactate supply overtakes glucose when neural computational and cognitive loads scale up . Proc Natl Acad Sci USA . (2022 ) 119 :e2212004119. doi: 10.1073/pnas.2212004119 36375086
PMC9704697
14Iadecola C .
The neurovascular unit coming of age: a journey through neurovascular coupling in health and disease . Neuron . (2017 ) 96 :17 –42 . doi: 10.1016/j.neuron.2017.07.030 28957666
PMC5657612
15Díaz-García CM Mongeon R Lahmann C Koveal D Zucker H Yellen G .
Neuronal stimulation triggers neuronal glycolysis and not lactate uptake . Cell Metab . (2017 ) 26 :361 –374.e364 .
10.1016/j.cmet.2017.06.021 PMC5559896 28768175
16Du F Zhu XH Zhang Y Friedman M Zhang N Ugurbil K .
Tightly coupled brain activity and cerebral ATP metabolic rate . Proc Natl Acad Sci USA . (2008 ) 105 :6409–14. doi: 10.1073/pnas.0710766105 18443293
PMC2359810
17Emmanuel Y Cochlin LE Tyler DJ de Jager CA Smith AD Clarke K .
Human hippocampal energy metabolism is impaired during cognitive activity in a lipid infusion model of insulin resistance . Brain Behav . (2013 ) 3 :134–44. doi: 10.1002/brb3.124 23533158
PMC3607154
18McNay EC Ong CT McCrimmon RJ Cresswell J Bogan JS Sherwin RS .
Hippocampal memory processes are modulated by insulin and high-fat-induced insulin resistance . Neurobiol Learn Mem . (2010 ) 93 :546–53. doi: 10.1016/j.nlm.2010.02.002 20176121
PMC2878207
19McNay EC .
Diet-induced diabetes is associated with lower hippocampal glycogen and reduced glycogenolysis following local exogenous insulin . J Neurochem . (2024 ) 168 :760–4.
10.1111/jnc.16001 PMC11045660 37885343
20Berends E Pencheva MG van de Waarenburg MPH Scheijen J Hermes D Wouters K .
Glyoxalase 1 overexpression improves neurovascular coupling and limits development of mild cognitive impairment in a mouse model of type 1 diabetes . J Physiol . (2024 ) 602 :6209–23. doi: 10.1113/jp286723 39316027
21Duarte JV Pereira JM Quendera B Raimundo M Moreno C Gomes L .
Early disrupted neurovascular coupling and changed event level hemodynamic response function in type 2 diabetes: an fMRI study . J Cereb Blood Flow Metab . (2015 ) 35 :1671–80. doi: 10.1038/jcbfm.2015.106 26058698
PMC4640307
22Duarte JV Guerra C Moreno C Gomes L Castelo-Branco M .
Changes in hemodynamic response function components reveal specific changes in neurovascular coupling in type 2 diabetes . Front Physiol . (2022 ) 13 :1101470 .
10.3389/fphys.2022.1101470 PMC9872943 36703928
23Lindquist MA Meng Loh J Atlas LY Wager TD .
Modeling the hemodynamic response function in fMRI: efficiency, bias and mis-modeling . Neuroimage . (2009 ) 45 :S187–198. doi: 10.1016/j.neuroimage.2008.10.065 19084070
PMC3318970
24Hirvonen J Virtanen KA Nummenmaa L Hannukainen JC Honka MJ Bucci M .
Effects of insulin on brain glucose metabolism in impaired glucose tolerance . Diabetes . (2011 ) 60 :443–7. doi: 10.2337/db10-0940 21270256
PMC3028343
25Tuulari JJ Karlsson HK Hirvonen J Hannukainen JC Bucci M Helmiö M .
Weight loss after bariatric surgery reverses insulin-induced increases in brain glucose metabolism of the morbidly obese . Diabetes . (2013 ) 62 :2747–51. doi: 10.2337/db12-1460 23493575
PMC3717871
26Boersma GJ Johansson E Pereira MJ Heurling K Skrtic S Lau J .
Altered glucose uptake in muscle, visceral adipose tissue, and brain predict whole-body insulin resistance and may contribute to the development of type 2 diabetes: a combined PET/MR study . Horm Metab Res . (2018 ) 50 :627–39. doi: 10.2337/db18-1790-p
30001566
27Rebelos E Bucci M Karjalainen T Oikonen V Bertoldo A Hannukainen JC .
Insulin resistance is associated with enhanced brain glucose uptake during euglycemic hyperinsulinemia: a large-scale PET cohort . Diabetes Care . (2021 ) 44 :788–94. doi: 10.2337/dc20-1549 33446523
PMC7896252
28Eriksson JW Visvanathar R Kullberg J Strand R Skrtic S Ekström S .
Tissue-specific glucose partitioning and fat content in prediabetes and type 2 diabetes: whole-body PET/MRI during hyperinsulinemia . Eur J Endocrinol . (2021 ) 184 :879–89. doi: 10.1530/eje-20-1359 33852422
29Pekkarinen L Kantonen T Rebelos E Latva-Rasku A Dadson P Karjalainen T .
Obesity risk is associated with brain glucose uptake and insulin resistance . Eur J Endocrinol . (2022 ) 187 :917–28. doi: 10.1530/eje-22-0509 36288097
PMC9782452
30Jensen NJ Porse AJ Wodschow HZ Speyer H Krogh J Marner L .
Relation of insulin resistance to brain glucose metabolism in fasting and hyperinsulinemic states: a systematic review and meta-analysis . J Clin Endocrinol Metab . (2025 ) 110 :e525–37.
10.1210/clinem/dgae570 PMC11747694 39185744
31Sokoloff L Reivich M Kennedy C Des Rosiers MH Patlak CS Pettigrew KD .
The [14C]deoxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure, and normal values in the conscious and anesthetized albino rat . J Neurochem . (1977 ) 28 :897 –916 . doi: 10.1111/j.1471-4159.1977.tb10649.x 864466
32Villien M Wey HY Mandeville JB Catana C Polimeni JR Sander CY .
Dynamic functional imaging of brain glucose utilization using fPET-FDG . Neuroimage . (2014 ) 100 :192–9. doi: 10.1016/j.neuroimage.2014.06.025 24936683
PMC4224310
33Hahn A Gryglewski G Nics L Hienert M Rischka L Vraka C .
Quantification of task-specific glucose metabolism with constant infusion of 18F-FDG . J Nucl Med . (2016 ) 57 :1933–40. doi: 10.2967/jnumed.116.176156 27390156
34Zimmer ER Parent MJ Souza DG Leuzy A Lecrux C Kim HI .
(18)F]FDG PET signal is driven by astroglial glutamate transport . Nat Neurosci . (2017 ) 20 :393–5. doi: 10.1038/nn.4492 28135241
PMC5378483
35Stiernman LJ Grill F Hahn A Rischka L Lanzenberger R Panes Lundmark V .
Dissociations between glucose metabolism and blood oxygenation in the human default mode network revealed by simultaneous PET-fMRI . Proc Natl Acad Sci USA . (2021 ) 118 . doi: 10.1073/pnas.2021913118 34193521
PMC8271663
36Hahn A Reed MB Vraka C Godbersen GM Klug S Komorowski A .
High-temporal resolution functional PET/MRI reveals coupling between human metabolic and hemodynamic brain response . Eur J Nucl Med Mol Imaging . (2024 ) 51 :1310–22.
10.1007/s00259-023-06542-4 PMC11399190 38052927
37Bahri S Horowitz M Malbert CH .
Inward glucose transfer accounts for insulin-dependent increase in brain glucose metabolism associated with diet-induced obesity . Obes (Silver Spring) . (2018 ) 26 :1322–31. doi: 10.1002/oby.22243 29956494
38Attwell D Laughlin SB .
An energy budget for signaling in the grey matter of the brain . J Cereb Blood Flow Metab . (2001 ) 21 :1133–45. doi: 10.1097/00004647-200110000-00001 11598490
39Harris JJ Jolivet R Attwell D .
Synaptic energy use and supply . Neuron . (2012 ) 75 :762–77. doi: 10.1016/j.neuron.2012.08.019 22958818
40Alle H Roth A Geiger JR .
Energy-efficient action potentials in hippocampal mossy fibers . Science . (2009 ) 325 :1405–8. doi: 10.1126/science.1174331 19745156
41Hallermann S de Kock CP Stuart GJ Kole MH .
State and location dependence of action potential metabolic cost in cortical pyramidal neurons . Nat Neurosci . (2012 ) 15 :1007–14. doi: 10.1038/nn.3132 22660478
42Pathak D Shields LY Mendelsohn BA Haddad D Lin W Gerencser AA .
The role of mitochondrially derived ATP in synaptic vesicle recycling . J Biol Chem . (2015 ) 290 :22325–36. doi: 10.1074/jbc.m115.656405 26126824
PMC4566209
43Pulido C Ryan TA .
Synaptic vesicle pools are a major hidden resting metabolic burden of nerve terminals . Sci Adv . (2021 ) 7 :eabi9027. doi: 10.1126/sciadv.abi9027 34860552
PMC8641928
44Tiwari A Myeong J Hashemiaghdam A Stunault MI Zhang H Niu X .
Mitochondrial pyruvate transport regulates presynaptic metabolism and neurotransmission . Sci Adv . (2024 ) 10 :eadp7423. doi: 10.1126/sciadv.adp7423 39546604
PMC11567002
45Pellerin L Magistretti PJ .
Glutamate uptake into astrocytes stimulates aerobic glycolysis: a mechanism coupling neuronal activity to glucose utilization . Proc Natl Acad Sci USA . (1994 ) 91 :10625–9. doi: 10.1073/pnas.91.22.10625 7938003
PMC45074
46Rangaraju V Lauterbach M Schuman EM .
Spatially stable mitochondrial compartments fuel local translation during plasticity . Cell . (2019 ) 176 :73 –84.e15 . doi: 10.1016/j.cell.2018.12.013 30612742
47Descalzi G Gao V Steinman MQ Suzuki A Alberini CM .
Lactate from astrocytes fuels learning-induced mRNA translation in excitatory and inhibitory neurons . Commun Biol . (2019 ) 2 :247 . doi: 10.1038/s42003-019-0495-2 31286064
PMC6606643
48Li H Guglielmetti C Sei YJ Zilberter M Le Page LM Shields L .
Neurons require glucose uptake and glycolysis in vivo . Cell Rep . (2023 ) 42 :112335. doi: 10.1016/j.celrep.2023.112335 37027294
PMC10556202
49Sibson NR Dhankhar A Mason GF Rothman DL Behar KL Shulman RG .
Stoichiometric coupling of brain glucose metabolism and glutamatergic neuronal activity . Proc Natl Acad Sci USA . (1998 ) 95 :316–21. doi: 10.1073/pnas.95.1.316 9419373
PMC18211
50Attwell D Buchan AM Charpak S Lauritzen M Macvicar BA Newman EA .
Glial and neuronal control of brain blood flow . Nature . (2010 ) 468 :232–50. doi: 10.1038/nature09613 21068832
PMC3206737
51Uhlirova H Kılıç K Tian P Thunemann M Desjardins M Saisan PA .
Cell type specificity of neurovascular coupling in cerebral cortex . Elife . (2016 ) 5 . doi: 10.7554/elife.14315 27244241
PMC4933561
52Echagarruga CT Gheres KW Norwood JN Drew PJ .
nNOS-expressing interneurons control basal and behaviorally evoked arterial dilation in somatosensory cortex of mice . Elife . (2020 ) 9 . doi: 10.7554/elife.60533 33016877
PMC7556878
53Nizar K Uhlirova H Tian P Saisan PA Cheng Q Reznichenko L .
In vivo stimulus-induced vasodilation occurs without IP3 receptor activation and may precede astrocytic calcium increase . J Neurosci . (2013 ) 33 :8411–22. doi: 10.1523/jneurosci.3285-12.2013 23658179
PMC3712855
54Otsu Y Couchman K Lyons DG Collot M Agarwal A Mallet JM .
Calcium dynamics in astrocyte processes during neurovascular coupling . Nat Neurosci . (2015 ) 18 :210–8.
10.1038/nn.3906 PMC4651918 25531572
55Institoris A Vandal M Peringod G Catalano C Tran CH Yu X .
Astrocytes amplify neurovascular coupling to sustained activation of neocortex in awake mice . Nat Commun . (2022 ) 13 :7872 . doi: 10.1038/s41467-022-35383-2 36550102
PMC9780254
56Mishra A Reynolds JP Chen Y Gourine AV Rusakov DA Attwell D .
Astrocytes mediate neurovascular signaling to capillary pericytes but not to arterioles . Nat Neurosci . (2016 ) 19 :1619–27. doi: 10.1038/nn.4428 27775719
PMC5131849
57Longden TA Dabertrand F Koide M Gonzales AL Tykocki NR Brayden JE .
Capillary K(+)-sensing initiates retrograde hyperpolarization to increase local cerebral blood flow . Nat Neurosci . (2017 ) 20 :717–26. doi: 10.1038/nn.4533 28319610
PMC5404963
58Chen BR Kozberg MG Bouchard MB Shaik MA Hillman EM .
A critical role for the vascular endothelium in functional neurovascular coupling in the brain . J Am Heart Assoc . (2014 ) 3 :e000787. doi: 10.1161/jaha.114.000787 24926076
PMC4309064
59Chow BW Nuñez V Kaplan L Granger AJ Bistrong K Zucker HL .
Caveolae in CNS arterioles mediate neurovascular coupling . Nature . (2020 ) 579 :106–10. doi: 10.1038/s41586-020-2026-1 32076269
PMC7060132
60Hall CN Reynell C Gesslein B Hamilton NB Mishra A Sutherland BA .
Capillary pericytes regulate cerebral blood flow in health and disease . Nature . (2014 ) 508 :55 –60 . doi: 10.1038/nature13165 24670647
PMC3976267
61Hill RA Tong L Yuan P Murikinati S Gupta S Grutzendler J .
Regional blood flow in the normal and ischemic brain is controlled by arteriolar smooth muscle cell contractility and not by capillary pericytes . Neuron . (2015 ) 87 :95 –110 . doi: 10.1016/j.neuron.2015.06.001 26119027
PMC4487786
62Hartmann DA Berthiaume AA Grant RI Harrill SA Koski T Tieu T .
Brain capillary pericytes exert a substantial but slow influence on blood flow . Nat Neurosci . (2021 ) 24 :633–45. doi: 10.1038/s41593-020-00793-2 33603231
PMC8102366
63Rungta RL Chaigneau E Osmanski BF Charpak S .
Vascular compartmentalization of functional hyperemia from the synapse to the pia . Neuron . (2018 ) 99 :362 –375.e364 . doi: 10.1016/j.neuron.2018.06.012 29937277
PMC6069674
64Fox PT Raichle ME .
Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects . Proc Natl Acad Sci USA . (1986 ) 83 :1140–4. doi: 10.1073/pnas.83.4.1140 3485282
PMC323027
65Fox PT Raichle ME Mintun MA Dence C .
Nonoxidative glucose consumption during focal physiologic neural activity . Science . (1988 ) 241 :462–5. doi: 10.1126/science.3260686 3260686
66Hoge RD Atkinson J Gill B Crelier GR Marrett S Pike GB .
Linear coupling between cerebral blood flow and oxygen consumption in activated human cortex . Proc Natl Acad Sci USA . (1999 ) 96 :9403–8. doi: 10.1073/pnas.96.16.9403 10430955
PMC17795
67Chen BR Bouchard MB McCaslin AF Burgess SA Hillman EM .
High-speed vascular dynamics of the hemodynamic response . Neuroimage . (2011 ) 54 :1021–30. doi: 10.1016/j.neuroimage.2010.09.036 20858545
PMC3018836
68Logothetis NK Pauls J Augath M Trinath T Oeltermann A .
Neurophysiological investigation of the basis of the fMRI signal . Nature . (2001 ) 412 :150–8. doi: 10.1038/35084005 11449264
69Hillman EM .
Coupling mechanism and significance of the BOLD signal: a status report . Annu Rev Neurosci . (2014 ) 37 :161–89. doi: 10.1146/annurev-neuro-071013-014111 25032494
PMC4147398
70Magistretti PJ Allaman I .
A cellular perspective on brain energy metabolism and functional imaging . Neuron . (2015 ) 86 :883 –901 . doi: 10.1016/j.neuron.2015.03.035 25996133
71Lundgaard I Li B Xie L Kang H Sanggaard S Haswell JD .
Direct neuronal glucose uptake heralds activity-dependent increases in cerebral metabolism . Nat Commun . (2015 ) 6 :6807 . doi: 10.1038/ncomms7807 25904018
PMC4410436
72Ashrafi G Wu Z Farrell RJ Ryan TA .
GLUT4 mobilization supports energetic demands of active synapses . Neuron . (2017 ) 93 :606 –615.e603 . doi: 10.1016/j.neuron.2016.12.020 28111082
PMC5330257
73Ivanov AI Malkov AE Waseem T Mukhtarov M Buldakova S Gubkina O .
Glycolysis and oxidative phosphorylation in neurons and astrocytes during network activity in hippocampal slices . J Cereb Blood Flow Metab . (2014 ) 34 :397 –407 .
10.1038/jcbfm.2013.222 PMC3948126 24326389
74Rouach N Koulakoff A Abudara V Willecke K Giaume C .
Astroglial metabolic networks sustain hippocampal synaptic transmission . Science . (2008 ) 322 :1551–5. doi: 10.1126/science.1164022 19056987
75Newman LA Korol DL Gold PE .
Lactate produced by glycogenolysis in astrocytes regulates memory processing . PloS One . (2011 ) 6 :e28427. doi: 10.1371/journal.pone.0028427 22180782
PMC3236748
76Duran J Saez I Gruart A Guinovart JJ Delgado-García JM .
Impairment in long-term memory formation and learning-dependent synaptic plasticity in mice lacking glycogen synthase in the brain . J Cereb Blood Flow Metab . (2013 ) 33 :550–6. doi: 10.1038/jcbfm.2012.200 23281428
PMC3618391
77Mächler P Wyss MT Elsayed M Stobart J Gutierrez R von Faber-Castell A .
In vivo evidence for a lactate gradient from astrocytes to neurons . Cell Metab . (2016 ) 23 :94 –102 .
10.1016/j.cmet.2015.10.010 26698914
78Boumezbeur F Petersen KF Cline GW Mason GF Behar KL Shulman GI .
The contribution of blood lactate to brain energy metabolism in humans measured by dynamic 13C nuclear magnetic resonance spectroscopy . J Neurosci . (2010 ) 30 :13983–91. doi: 10.1523/jneurosci.2040-10.2010 20962220
PMC2996729
79Yang J Ruchti E Petit JM Jourdain P Grenningloh G Allaman I .
Lactate promotes plasticity gene expression by potentiating NMDA signaling in neurons . Proc Natl Acad Sci USA . (2014 ) 111 :12228–33. doi: 10.1073/pnas.1322912111 25071212
PMC4143009
80Rae CD Baur JA Borges K Dienel G Díaz-García CM Douglass SR .
Brain energy metabolism: A roadmap for future research . J Neurochem . (2024 ) 168 :910–54. doi: 10.1111/jnc.16032 38183680
PMC11102343
81Baeza-Lehnert F Saab AS Gutiérrez R Larenas V Díaz E Horn M .
Non-canonical control of neuronal energy status by the Na(+) pump . Cell Metab . (2019 ) 29 :668 –680.e664 . doi: 10.1016/j.cmet.2018.11.005 30527744
82Gerkau NJ Lerchundi R Nelson JSE Lantermann M Meyer J Hirrlinger J .
Relation between activity-induced intracellular sodium transients and ATP dynamics in mouse hippocampal neurons . J Physiol . (2019 ) 597 :5687–705. doi: 10.1113/jp278658 31549401
83Straub I Kunstmann L Baeza-Lehnert F Chowdhry S Renden RB Gonzalez-Aragón G .
Presynaptic ATP decreases during physiological-like activity in neurons tuned for high-frequency transmission . J Neurochem . (2025 ) 169 :e70212. doi: 10.1111/jnc.70212 40916622
PMC12415541
84Trevisiol A Saab AS Winkler U Marx G Imamura H Möbius W .
Monitoring ATP dynamics in electrically active white matter tracts . Elife . (2017 ) 6 . doi: 10.7554/elife.24241 28414271
PMC5415357
85Hall CN Klein-Flügge MC Howarth C Attwell D .
Oxidative phosphorylation, not glycolysis, powers presynaptic and postsynaptic mechanisms underlying brain information processing . J Neurosci . (2012 ) 32 :8940–51. doi: 10.1523/jneurosci.0026-12.2012 22745494
PMC3390246
86Sobieski C Fitzpatrick MJ Mennerick SJ .
Differential presynaptic ATP supply for basal and high-demand transmission . J Neurosci . (2017 ) 37 :1888–99. doi: 10.1523/jneurosci.2712-16.2017 28093477
PMC5320616
87Díaz-García CM Meyer DJ Nathwani N Rahman M Martínez-François JR Yellen G .
The distinct roles of calcium in rapid control of neuronal glycolysis and the tricarboxylic acid cycle . Elife . (2021 ) 10 .
10.7554/eLife.64821 PMC7870136 33555254
88Ashrafi G de Juan-Sanz J Farrell RJ Ryan TA .
Molecular tuning of the axonal mitochondrial Ca(2+) uniporter ensures metabolic flexibility of neurotransmission . Neuron . (2020 ) 105 :678 –687.e675 . doi: 10.1016/j.neuron.2019.11.020 31862210
PMC7035162
89Jost CR Van Der Zee CE In 't Zandt HJ Oerlemans F Verheij M Streijger F .
Creatine kinase B-driven energy transfer in the brain is important for habituation and spatial learning behaviour, mossy fibre field size and determination of seizure susceptibility . Eur J Neurosci . (2002 ) 15 :1692–706. doi: 10.1046/j.1460-9568.2002.02001.x 12059977
90Kang JS Tian JH Pan PY Zald P Li C Deng C .
Docking of axonal mitochondria by syntaphilin controls their mobility and affects short-term facilitation . Cell . (2008 ) 132 :137–48. doi: 10.1016/j.cell.2007.11.024 18191227
PMC2259239
91Divakaruni SS Van Dyke AM Chandra R LeGates TA Contreras M Dharmasri PA .
Long-term potentiation requires a rapid burst of dendritic mitochondrial fission during induction . Neuron . (2018 ) 100 :860 –875.e867 . doi: 10.1016/j.neuron.2018.09.025 30318410
PMC6483400
92Kann O Huchzermeyer C Kovács R Wirtz S Schuelke M .
Gamma oscillations in the hippocampus require high complex I gene expression and strong functional performance of mitochondria . Brain . (2011 ) 134 :345–58.
10.1093/brain/awq333 21183487
93Schneider J Berndt N Papageorgiou IE Maurer J Bulik S Both M .
Local oxygen homeostasis during various neuronal network activity states in the mouse hippocampus . J Cereb Blood Flow Metab . (2019 ) 39 :859–73.
10.1177/0271678X17740091 PMC6501513 29099662
94Bas-Orth C Schneider J Lewen A McQueen J Hasenpusch-Theil K Theil T .
The mitochondrial calcium uniporter is crucial for the generation of fast cortical network rhythms . J Cereb Blood Flow Metab . (2020 ) 40 :2225–39.
10.1177/0271678X19887777 PMC7585921 31722597
95Jang C Chen L Rabinowitz JD .
Metabolomics and isotope tracing . Cell . (2018 ) 173 :822–37. doi: 10.1016/j.cell.2018.03.055 29727671
PMC6034115
96Kim IY Park S Kim Y Kim HJ Wolfe RR .
Tracing metabolic flux in vivo: basic model structures of tracer methodology . Exp Mol Med . (2022 ) 54 :1311–22. doi: 10.1038/s12276-022-00814-z 36075950
PMC9534847
97Hwang JJ Jiang L Hamza M Sanchez Rangel E Dai F Belfort-DeAguiar R .
Blunted rise in brain glucose levels during hyperglycemia in adults with obesity and T2DM . JCI Insight . (2017 ) 2 . doi: 10.1172/jci.insight.95913 29046482
PMC5846903
98Sanchez-Rangel E Gunawan F Jiang L Savoye M Dai F Coppoli A .
Reversibility of brain glucose kinetics in type 2 diabetes mellitus . Diabetologia . (2022 ) 65 :895 –905 . doi: 10.1007/s00125-022-05664-y 35247067
PMC8960594
99Alsop DC Detre JA Golay X Günther M Hendrikse J Hernandez-Garcia L .
Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia . Magn Reson Med . (2015 ) 73 :102–16.
10.1002/mrm.25197 PMC4190138 24715426
100Nealon RS Howe PR Jansen L Garg M Wong RH .
Impaired cerebrovascular responsiveness and cognitive performance in adults with type 2 diabetes . J Diabetes Complications . (2017 ) 31 :462–7. doi: 10.1016/j.jdiacomp.2016.06.025 27431891
101Oliel Y Ravona-Springer R Harel M Azuri J Moshe CB Tanne D .
The role of cerebrovascular reactivity on brain activation during a working memory task in type 2 diabetes . Alzheimers Dement (Amst) . (2025 ) 17 :e70045. doi: 10.1002/dad2.70045 40078378
PMC11899760
102Sauer U .
Metabolic networks in motion: 13C-based flux analysis . Mol Syst Biol . (2006 ) 2 :62 . doi: 10.1038/msb4100109 17102807
PMC1682028
103Buescher JM Antoniewicz MR Boros LG Burgess SC Brunengraber H Clish CB .
A roadmap for interpreting (13)C metabolite labeling patterns from cells . Curr Opin Biotechnol . (2015 ) 34 :189 –201 . doi: 10.1016/j.copbio.2015.02.003 25731751
PMC4552607
104de Graaf RA Rothman DL Behar KL .
State of the art direct 13C and indirect 1H-[13C] NMR spectroscopy in vivo. A practical guide . NMR BioMed . (2011 ) 24 :958–72.
10.1002/nbm.1761 PMC3694136 21919099
105Rothman DL De Feyter HM de Graaf RA Mason GF Behar KL .
13C MRS studies of neuroenergetics and neurotransmitter cycling in humans . NMR BioMed . (2011 ) 24 :943–57. doi: 10.1002/nbm.1772 21882281
PMC3651027
106Mason GF Pan JW Chu WJ Newcomer BR Zhang Y Orr R .
Measurement of the tricarboxylic acid cycle rate in human grey and white matter in vivo by 1H-[13C] magnetic resonance spectroscopy at 4.1T . J Cereb Blood Flow Metab . (1999 ) 19 :1179–88. doi: 10.1097/00004647-199911000-00002 10566964
107Barros LF Bolaños JP Bonvento G Bouzier-Sore AK Brown A Hirrlinger J .
Current technical approaches to brain energy metabolism . Glia . (2018 ) 66 :1138–59.
10.1002/glia.23248 PMC5903992 29110344
108Patel S Porcari P Coffee E Kim N Berishaj M Peyear T .
Simultaneous noninvasive quantification of redox and downstream glycolytic fluxes reveals compartmentalized brain metabolism . Sci Adv . (2024 ) 10 :eadr2058. doi: 10.1126/sciadv.adr2058 39705365
PMC11661454
109Davis TL Kwong KK Weisskoff RM Rosen BR .
Calibrated functional MRI: mapping the dynamics of oxidative metabolism . Proc Natl Acad Sci USA . (1998 ) 95 :1834–9. doi: 10.1073/pnas.95.4.1834 9465103
PMC19199
110Rodgers ZB Detre JA Wehrli FW .
MRI-based methods for quantification of the cerebral metabolic rate of oxygen . J Cereb Blood Flow Metab . (2016 ) 36 :1165–85. doi: 10.1177/0271678x16643090 27089912
PMC4929705
111Fan AP An H Moradi F Rosenberg J Ishii Y Nariai T .
Quantification of brain oxygen extraction and metabolism with [(15)O]-gas PET: A technical review in the era of PET/MRI . Neuroimage . (2020 ) 220 :117136. doi: 10.1016/j.neuroimage.2020.117136 32634594
PMC7592419
112Blockley NP Griffeth VE Buxton RB .
A general analysis of calibrated BOLD methodology for measuring CMRO2 responses: comparison of a new approach with existing methods . Neuroimage . (2012 ) 60 :279–89. doi: 10.1016/j.neuroimage.2011.11.081 22155329
PMC3288960
113Brand MD Nicholls DG .
Assessing mitochondrial dysfunction in cells . Biochem J . (2011 ) 435 :297 –312 . doi: 10.1042/bj4370575u
PMC3076726 21726199
114Schmidt CA Fisher-Wellman KH Neufer PD .
From OCR and ECAR to energy: Perspectives on the design and interpretation of bioenergetics studies . J Biol Chem . (2021 ) 297 :101140. doi: 10.1016/j.jbc.2021.101140 34461088
PMC8479256
115Mookerjee SA Gerencser AA Nicholls DG Brand MD .
Quantifying intracellular rates of glycolytic and oxidative ATP production and consumption using extracellular flux measurements . J Biol Chem . (2017 ) 292 :7189–207. doi: 10.1074/jbc.m116.774471 28270511
PMC5409486
116Du F Zhu XH Qiao H Zhang X Chen W .
Efficient in vivo 31P magnetization transfer approach for noninvasively determining multiple kinetic parameters and metabolic fluxes of ATP metabolism in the human brain . Magn Reson Med . (2007 ) 57 :103–14.
10.1002/mrm.21107 17191226
117From AH Ugurbil K .
Standard magnetic resonance-based measurements of the Pi→ATP rate do not index the rate of oxidative phosphorylation in cardiac and skeletal muscles . Am J Physiol Cell Physiol . (2011 ) 301 :C1 –11 . doi: 10.1152/ajpcell.00345.2010 21368294
PMC3129822
118Tantama M Martínez-François JR Mongeon R Yellen G .
Imaging energy status in live cells with a fluorescent biosensor of the intracellular ATP-to-ADP ratio . Nat Commun . (2013 ) 4 :2550 . doi: 10.1038/ncomms3550 24096541
PMC3852917
119Hung YP Albeck JG Tantama M Yellen G .
Imaging cytosolic NADH-NAD(+) redox state with a genetically encoded fluorescent biosensor . Cell Metab . (2011 ) 14 :545–54. doi: 10.1016/j.cmet.2011.08.012 21982714
PMC3190165
120Koveal D Díaz-García CM Yellen G .
Fluorescent biosensors for neuronal metabolism and the challenges of quantitation . Curr Opin Neurobiol . (2020 ) 63 :111–21. doi: 10.1016/j.conb.2020.02.011 32559637
PMC7646541
121Duelli R Staudt R Duembgen L Kuschinsky W .
Increase in glucose transporter densities of Glut3 and decrease of glucose utilization in rat brain after one week of hypoglycemia . Brain Res . (1999 ) 831 :254–62. doi: 10.1016/s0006-8993(99)01463-8 10412004
122Soares AF Nissen JD Garcia-Serrano AM Nussbaum SS Waagepetersen HS Duarte JMN .
Glycogen metabolism is impaired in the brain of male type 2 diabetic Goto-Kakizaki rats . J Neurosci Res . (2019 ) 97 :1004–17. doi: 10.1002/jnr.24437 31044444
123Baker LD Cross DJ Minoshima S Belongia D Watson GS Craft S .
Insulin resistance and Alzheimer-like reductions in regional cerebral glucose metabolism for cognitively normal adults with prediabetes or early type 2 diabetes . Arch Neurol . (2011 ) 68 :51–7. doi: 10.1001/archneurol.2010.225 20837822
PMC3023149
124Gejl M Gjedde A Brock B Møller A van Duinkerken E Haahr HL .
Effects of hypoglycaemia on working memory and regional cerebral blood flow in type 1 diabetes: a randomised, crossover trial . Diabetologia . (2018 ) 61 :551–61.
10.1007/s00125-017-4502-1 PMC6448973 29188338
125Gonçalves JS Seiça RM Laranjinha J Lourenço CF .
Impairment of neurovascular coupling in the hippocampus due to decreased nitric oxide bioavailability supports early cognitive dysfunction in type 2 diabetic rats . Free Radic Biol Med . (2022 ) 193 :669–75.
10.1016/j.freeradbiomed.2022.11.009 36372286
126Kong L Zhao Y Zhou WJ Yu H Teng SW Guo Q .
Direct neuronal glucose uptake is required for contextual fear acquisition in the dorsal hippocampus . Front Mol Neurosci . (2017 ) 10 :388. doi: 10.3389/fnmol.2017.00388 29209168
PMC5702440
127Wei XY Zou ZM Yao ZX Teng SW Wei XY Tang WJ .
Neuronal activity-induced GLUT3 plasma translocation supports energy demands for memory acquisition . Commun Biol . (2025 ) 8 :1716 . doi: 10.1038/s42003-025-09119-z 41315759
PMC12663466
128Marder TJ Flores VL Bolo NR Hoogenboom WS Simonson DC Jacobson AM .
Task-induced brain activity patterns in type 2 diabetes: a potential biomarker for cognitive decline . Diabetes . (2014 ) 63 :3112–9. doi: 10.2337/db13-1783 24705405
PMC4141362
129Chen Y Liu Z Zhang J Xu K Zhang S Wei D .
Altered brain activation patterns under different working memory loads in patients with type 2 diabetes . Diabetes Care . (2014 ) 37 :3157–63. doi: 10.2337/dc14-1683 25404661
130He XS Wang ZX Zhu YZ Wang N Hu X Zhang DR .
Hyperactivation of working memory-related brain circuits in newly diagnosed middle-aged type 2 diabetics . Acta Diabetol . (2015 ) 52 :133–42.
10.1007/s00592-014-0618-7 PMC4416650 24993663
131Backeström A Papadopoulos K Eriksson S Olsson T Andersson M Blennow K .
Acute hyperglycaemia leads to altered frontal lobe brain activity and reduced working memory in type 2 diabetes . PloS One . (2021 ) 16 :e0247753.
10.1371/journal.pone.0247753 PMC7978337 33739980
132Embury CM Lord GH Drincic AT Desouza CV Wilson TW .
Glycemic control level alters working memory neural dynamics in adults with type 2 diabetes . Cereb Cortex . (2023 ) 33 :8333–41. doi: 10.1093/cercor/bhad119 37005060
PMC10321117
133Hasselbalch SG Knudsen GM Videbaek C Pinborg LH Schmidt JF Holm S .
No effect of insulin on glucose blood-brain barrier transport and cerebral metabolism in humans . Diabetes . (1999 ) 48 :1915–21. doi: 10.2337/diabetes.48.10.1915 10512354
134Eastman RC Carson RE Gordon MR Berg GW Lillioja S Larson SM .
Brain glucose metabolism in noninsulin-dependent diabetes mellitus: a study in Pima Indians using positron emission tomography during hyperinsulinemia with euglycemic glucose clamp . J Clin Endocrinol Metab . (1990 ) 71 :1602–10. doi: 10.1210/jcem-71-6-1602 2229317
135Anthony K Reed LJ Dunn JT Bingham E Hopkins D Marsden PK .
Attenuation of insulin-evoked responses in brain networks controlling appetite and reward in insulin resistance: the cerebral basis for impaired control of food intake in metabolic syndrome ? Diabetes . (2006 ) 55 :2986–92. doi: 10.2337/db06-0376 17065334
136Gunawan F Matson BC Coppoli A Jiang L Ding Y Perry R .
Deficits in brain glucose transport among younger adults with obesity . Obes (Silver Spring) . (2024 ) 32 :1329–38. doi: 10.1002/oby.24034 38764181
PMC11966602
137Tschritter O Preissl H Hennige AM Stumvoll M Porubska K Frost R .
The cerebrocortical response to hyperinsulinemia is reduced in overweight humans: a magnetoencephalographic study . Proc Natl Acad Sci USA . (2006 ) 103 :12103–8. doi: 10.1073/pnas.0604404103 16877540
PMC1567704
138Fu Z Wu J Nesil T Li MD Aylor KW Liu Z .
Long-term high-fat diet induces hippocampal microvascular insulin resistance and cognitive dysfunction . Am J Physiol Endocrinol Metab . (2017 ) 312 :E89 –e97 . doi: 10.1152/ajpendo.00297.2016 27899343
PMC5336564
139Novak V Milberg W Hao Y Munshi M Novak P Galica A .
Enhancement of vasoreactivity and cognition by intranasal insulin in type 2 diabetes . Diabetes Care . (2014 ) 37 :751–9. doi: 10.2337/dc13-1672 24101698
PMC3931384
140Cryer PE .
Mechanisms of hypoglycemia-associated autonomic failure and its component syndromes in diabetes . Diabetes . (2005 ) 54 :3592–601. doi: 10.2337/diabetes.54.12.3592 16306382
141Rooijackers HM Wiegers EC Tack CJ van der Graaf M de Galan BE .
Brain glucose metabolism during hypoglycemia in type 1 diabetes: insights from functional and metabolic neuroimaging studies . Cell Mol Life Sci . (2016 ) 73 :705–22.
10.1007/s00018-015-2079-8 PMC4735263 26521082
142McNay EC Sherwin RS .
Effect of recurrent hypoglycemia on spatial cognition and cognitive metabolism in normal and diabetic rats . Diabetes . (2004 ) 53 :418–25. doi: 10.2337/diabetes.53.2.418 14747293
143Jiang L Herzog RI Mason GF de Graaf RA Rothman DL Sherwin RS .
Recurrent antecedent hypoglycemia alters neuronal oxidative metabolism in vivo . Diabetes . (2009 ) 58 :1266–74. doi: 10.2337/db08-1664 19276443
PMC2682668
144Boyle PJ Nagy RJ O'Connor AM Kempers SF Yeo RA Qualls C .
Adaptation in brain glucose uptake following recurrent hypoglycemia . Proc Natl Acad Sci USA . (1994 ) 91 :9352–6. doi: 10.1073/pnas.91.20.9352 7937768
PMC44810
145Bednařík P Henry PG Khowaja A Rubin N Kumar A Deelchand D .
Hippocampal neurochemical profile and glucose transport kinetics in patients with type 1 diabetes . J Clin Endocrinol Metab . (2020 ) 105 :479–91.
10.1210/clinem/dgz062 PMC7046023 31637440
146Seaquist ER Park YW Kumar A Alvear A Zhang L Strong B .
Impact of recurrent hypoglycemia on brain glucose transport kinetics in type 1 diabetes . J Clin Endocrinol Metab . (2026 ). doi: 10.1210/clinem/dgag205 42128812
PMC13671781
147Boyle PJ Kempers SF O'Connor AM Nagy RJ .
Brain glucose uptake and unawareness of hypoglycemia in patients with insulin-dependent diabetes mellitus . N Engl J Med . (1995 ) 333 :1726–31. doi: 10.1056/nejm199512283332602 7491135
148Bingham EM Dunn JT Smith D Sutcliffe-Goulden J Reed LJ Marsden PK .
Differential changes in brain glucose metabolism during hypoglycaemia accompany loss of hypoglycaemia awareness in men with type 1 diabetes mellitus. An [11C]-3-O-methyl-D-glucose PET study . Diabetologia . (2005 ) 48 :2080–9. doi: 10.1007/s00125-005-1900-6 16143864
149Bolo NR Musen G Jacobson AM Weinger K McCartney RL Flores V .
Brain activation during working memory is altered in patients with type 1 diabetes during hypoglycemia . Diabetes . (2011 ) 60 :3256–64. doi: 10.2337/db11-0506 21984582
PMC3219930
150Verhulst CEM Fabricius TW Nefs G Kessels RPC Pouwer F Teerenstra S .
Consistent effects of hypoglycemia on cognitive function in people with or without diabetes . Diabetes Care . (2022 ) 45 :2103–10. doi: 10.2337/dc21-2502 35876660
PMC9472511
151Svensson CH Fabricius TW Verhulst C Tack CJ Kristensen PL de Galan BE .
Association between recent real-life exposure to hypoglycaemia and cognitive function during hypoglycaemia in people with type 1 diabetes: Findings from the Hypo-RESOLVE clamp study . Diabetes Obes Metab . (2025 ) 27 :5455–63. doi: 10.1111/dom.16587 40642831
PMC12409216
152Filip P Canna A Grohn H Moheet AA Kumar AF Li X .
Disrupted neurovascular-endocrine coupling in type 1 diabetes with impaired awareness of hypoglycemia . J Clin Invest . (2026 ) 136 . doi: 10.1172/jci199725 41712301
PMC13078881
153Bischof MG Mlynarik V Brehm A Bernroider E Krssak M Bauer E .
Brain energy metabolism during hypoglycaemia in healthy and type 1 diabetic subjects . Diabetologia . (2004 ) 47 :648–51.
10.1007/s00125-004-1362-2 15298341
154van de Ven KC de Galan BE van der Graaf M Shestov AA Henry PG Tack CJ .
Effect of acute hypoglycemia on human cerebral glucose metabolism measured by ¹³C magnetic resonance spectroscopy . Diabetes . (2011 ) 60 :1467–73. doi: 10.2337/db10-1592 21464446
PMC3292319
155van de Ven KC Tack CJ Heerschap A van der Graaf M de Galan BE .
Patients with type 1 diabetes exhibit altered cerebral metabolism during hypoglycemia . J Clin Invest . (2013 ) 123 :623–9. doi: 10.1172/jci62742 23298837
PMC3561817
156Mason GF Petersen KF Lebon V Rothman DL Shulman GI .
Increased brain monocarboxylic acid transport and utilization in type 1 diabetes . Diabetes . (2006 ) 55 :929–34. doi: 10.2337/diabetes.55.04.06.db05-1325 16567513
PMC2995526
157Gulanski BI De Feyter HM Page KA Belfort-DeAguiar R Mason GF Rothman DL .
Increased brain transport and metabolism of acetate in hypoglycemia unawareness . J Clin Endocrinol Metab . (2013 ) 98 :3811–20. doi: 10.1210/jc.2013-1701 23796565
PMC4425818
158Wiegers EC Rooijackers HM Tack CJ Groenewoud H Heerschap A de Galan BE .
Effect of exercise-induced lactate elevation on brain lactate levels during hypoglycemia in patients with type 1 diabetes and impaired awareness of hypoglycemia . Diabetes . (2017 ) 66 :3105–10. doi: 10.2337/db17-0794 28935628
159Herzog RI Jiang L Herman P Zhao C Sanganahalli BG Mason GF .
Lactate preserves neuronal metabolism and function following antecedent recurrent hypoglycemia . J Clin Invest . (2013 ) 123 :1988–98. doi: 10.1172/jci65105 23543056
PMC3638906
160Page KA Williamson A Yu N McNay EC Dzuira J McCrimmon RJ .
Medium-chain fatty acids improve cognitive function in intensively treated type 1 diabetic patients and support in vitro synaptic transmission during acute hypoglycemia . Diabetes . (2009 ) 58 :1237–44. doi: 10.2337/db08-1557 19223595
PMC2671041
161Tricò D Jiang L Spicer D Page K Belfort de Aguiar R Naik S .
Ketone bodies derived from medium-chain triglycerides support brain metabolism and function under hypoglycemia in type 1 diabetes mellitus . Diabetes . (2026 ) 75 :1644–56.
10.2337/db26-0293 PMC13493211 42384016
162Andersen JV Christensen SK Nissen JD Waagepetersen HS .
Improved cerebral energetics and ketone body metabolism in db/db mice . J Cereb Blood Flow Metab . (2017 ) 37 :1137–47.
10.1177/0271678X16684154 PMC5363491 28058963
163Devor A Hillman EM Tian P Waeber C Teng IC Ruvinskaya L .
Stimulus-induced changes in blood flow and 2-deoxyglucose uptake dissociate in ipsilateral somatosensory cortex . J Neurosci . (2008 ) 28 :14347–57. doi: 10.1523/jneurosci.4307-08.2008 19118167
PMC2655308
164Hollnagel JO Cesetti T Schneider J Vazetdinova A Valiullina-Rakhmatullina F Lewen A .
Lactate attenuates synaptic transmission and affects brain rhythms featuring high energy expenditure . iScience . (2020 ) 23 :101316. doi: 10.1016/j.isci.2020.101316 32653807
PMC7350153
165Chausse B Malorny N Lewen A Poschet G Berndt N Kann O .
Metabolic flexibility ensures proper neuronal network function in moderate neuroinflammation . Sci Rep . (2024 ) 14 :14405 . doi: 10.1038/s41598-024-64872-1 38909138
PMC11193723
166Vetri F Xu H Paisansathan C Pelligrino DA .
Impairment of neurovascular coupling in type 1 diabetes mellitus in rats is linked to PKC modulation of BK(Ca) and Kir channels . Am J Physiol Heart Circ Physiol . (2012 ) 302 :H1274–1284. doi: 10.1152/ajpheart.01067.2011 22268114
PMC3311480
167Vetri F Qi M Xu H Oberholzer J Paisansathan C .
Impairment of neurovascular coupling in type 1 diabetes mellitus in rats is prevented by pancreatic islet transplantation and reversed by a semi-selective PKC inhibitor . Brain Res . (2017 ) 1655 :48 –54 . doi: 10.1016/j.brainres.2016.11.012 27865779
PMC5195876
168Wang S Jiao F Border JJ Fang X Crumpler RF Liu Y .
Luseogliflozin, a sodium-glucose cotransporter-2 inhibitor, reverses cerebrovascular dysfunction and cognitive impairments in 18-mo-old diabetic animals . Am J Physiol Heart Circ Physiol . (2022 ) 322 :H246 –h259 . doi: 10.1152/ajpheart.00438.2021 34951541
PMC8759958
169Chiang PP Nippert AR Newman EA .
Neurovascular coupling is preserved in the cerebral cortex of diabetic mice: A longitudinal study . J Cereb Blood Flow Metab . (2026 ) 46 :753–63.
10.1177/0271678X251377575 PMC12463920 40995874
170Lietzau G Darsalia V Pintana H Östenson CG Nyström T Fisahn A .
Type 2 diabetes alters hippocampal gamma oscillations: A potential mechanism behind impaired cognition . Psychoneuroendocrinology . (2017 ) 82 :46 –50 . doi: 10.1016/j.psyneuen.2017.04.012 28501550
171Rabiller G Ip Z Zarrabian S Zhang H Sato Y Yazdan-Shahmorad A .
Type-2 diabetes alters hippocampal neural oscillations and disrupts synchrony between the hippocampus and cortex . Aging Dis . (2024 ) 15 :2255–70.
10.14336/AD.2023.1106 PMC11346393 38029397
172Kobayashi R Maruoka J Norimoto H Ikegaya Y Kume K Ohsawa M .
Involvement of l-lactate in hippocampal dysfunction of type I diabetes . J Pharmacol Sci . (2019 ) 141 :79 –82 . doi: 10.1016/j.jphs.2019.09.004 31586517
173Izumi Y Yamada KA Matsukawa M Zorumski CF .
Effects of insulin on long-term potentiation in hippocampal slices from diabetic rats . Diabetologia . (2003 ) 46 :1007–12. doi: 10.1007/s00125-003-1144-2 12827244
174Kula B Antal B Weistuch C Gackière F Barre A Velado V .
D-β-hydroxybutyrate stabilizes hippocampal CA3-CA1 circuit during acute insulin resistance . PNAS Nexus . (2024 ) 3 :pgae196.
10.1093/pnasnexus/pgae196 PMC11138115 38818236
175Huang S Wang Y Gan X Fang D Zhong C Wu L .
Drp1-mediated mitochondrial abnormalities link to synaptic injury in diabetes model . Diabetes . (2015 ) 64 :1728–42.
10.2337/db14-0758 PMC4407851 25412623
176Yan S Du F Wu L Zhang Z Zhong C Yu Q .
F1F0 ATP synthase-cyclophilin D interaction contributes to diabetes-induced synaptic dysfunction and cognitive decline . Diabetes . (2016 ) 65 :3482–94. doi: 10.2337/db16-0556 27554467
PMC5079631
177Grillo CA Piroli GG Lawrence RC Wrighten SA Green AJ Wilson SP .
Hippocampal insulin resistance impairs spatial learning and synaptic plasticity . Diabetes . (2015 ) 64 :3927–36. doi: 10.2337/db15-0596 26216852
PMC4613975
178Liu Z Patil I Sancheti H Yin F Cadenas E .
Effects of lipoic acid on high-fat diet-induced alteration of synaptic plasticity and brain glucose metabolism: A PET/CT and (13)C-NMR study . Sci Rep . (2017 ) 7 :5391 . doi: 10.1038/s41598-017-05217-z 28710347
PMC5511189
179Ugidos IF Iglesias JC Sendall H Calero-Hernandez VM Dugas CM Zsombok A .
Transient neurovascular coupling impairment in brain penetrating arterioles of streptozotocin treated mice post recurrent nonsevere hypoglycemia . iScience . (2026 ) 29 :114371. doi: 10.1016/j.isci.2025.114371 41675061
PMC12886521
180Girault FM Sonnay S Gruetter R Duarte JMN .
Alterations of brain energy metabolism in type 2 diabetic Goto-Kakizaki rats measured in vivo by (13)C magnetic resonance spectroscopy . Neurotox Res . (2019 ) 36 :268–78.
10.1007/s12640-017-9821-y 28971314
181Zheng H Zheng Y Wang D Cai A Lin Q Zhao L .
Analysis of neuron-astrocyte metabolic cooperation in the brain of db/db mice with cognitive decline using 13C NMR spectroscopy . J Cereb Blood Flow Metab . (2017 ) 37 :332–43.
10.1177/0271678X15626154 PMC5363750 26762505
182Chen J Lin R Jiang C Chen F Li W Wang L .
Brain endothelial HIF-1α exacerbates diabetes-associated cognitive impairment by accelerating glycolysis-driven lactate production . Acta Pharm Sin B . (2025 ) 15 :5772–88. doi: 10.1016/j.apsb.2025.09.028 41311407
PMC12648004
183Li Y Jiang T Du M He S Huang N Cheng B .
Ketohexokinase-dependent metabolism of cerebral endogenous fructose in microglia drives diabetes-associated cognitive dysfunction . Exp Mol Med . (2023 ) 55 :2417–32. doi: 10.1038/s12276-023-01112-y 37907746
PMC10689812
184Chandrasekaran K Choi J Arvas MI Salimian M Singh S Xu S .
Nicotinamide mononucleotide administration prevents experimental diabetes-induced cognitive impairment and loss of hippocampal neurons . Int J Mol Sci . (2020 ) 21 . doi: 10.3390/ijms21113756 32466541
PMC7313029
185Ruegsegger GN Vanderboom PM Dasari S Klaus KA Kabiraj P McCarthy CB .
Exercise and metformin counteract altered mitochondrial function in the insulin-resistant brain . JCI Insight . (2019 ) 4 . doi: 10.1172/jci.insight.130681 31534057
PMC6795285
186Ruegsegger GN Manjunatha S Summer P Gopala S Zabeilski P Dasari S .
Insulin deficiency and intranasal insulin alter brain mitochondrial function: A potential factor for dementia in diabetes . FASEB J . (2019 ) 33 :4458–72. doi: 10.1096/fj.201802043r 30676773
PMC6404590
187Zhao L Jiang H Xie J Shen D Yi Q Yan J .
Effects of fibroblast growth factor 21 on lactate uptake and usage in mice with diabetes-associated cognitive decline . Mol Neurobiol . (2022 ) 59 :5656–72. doi: 10.1007/s12035-022-02926-z 35761156
188Monteiro A Castro P Pereira G Ferreira C Sorond F Milstead A .
Neurovascular coupling is impaired in hypertensive and diabetic subjects without symptomatic cerebrovascular disease . Front Aging Neurosci . (2021 ) 13 :728007. doi: 10.3389/fnagi.2021.728007 34690741
PMC8526560
189Wirt RA Crew LA Ortiz AA McNeela AM Flores E Kinney JW .
Altered theta rhythm and hippocampal-cortical interactions underlie working memory deficits in a hyperglycemia risk factor model of Alzheimer's disease . Commun Biol . (2021 ) 4 :1036 . doi: 10.1038/s42003-021-02558-4 34480097
PMC8417282
190Phillips AA Chan FH Zheng MM Krassioukov AV Ainslie PN .
Neurovascular coupling in humans: Physiology, methodological advances and clinical implications . J Cereb Blood Flow Metab . (2016 ) 36 :647–64.
10.1177/0271678X15617954 PMC4821024 26661243
191Li B Freeman RD .
Neurometabolic coupling between neural activity, glucose, and lactate in activated visual cortex . J Neurochem . (2015 ) 135 :742–54. doi: 10.1111/jnc.13143 25930947
PMC4627899
192Burma JS Bailey DM Johnson NE Griffiths JK Burkart JJ Soligon CA .
Physiological influences on neurovascular coupling: A systematic review of multimodal imaging approaches and recommendations for future study designs . Exp Physiol . (2025 ) 110 :23 –41 .
10.1113/EP092060 PMC11689421 39392865
193Theparambil SM Kopach O Braga A Nizari S Hosford PS Sagi-Kiss V .
Adenosine signalling to astrocytes coordinates brain metabolism and function . Nature . (2024 ) 632 :139–46. doi: 10.1038/s41586-024-07611-w 38961289
PMC11291286
194Choi YS Kang S Ko SY Lee S Kim JY Lee H .
Hyperpolarized [1-13C] pyruvate MR spectroscopy detect altered glycolysis in the brain of a cognitively impaired mouse model fed high-fat diet . Mol Brain . (2018 ) 11 :74 . doi: 10.1186/s13041-018-0415-2 30563553
PMC6299662
195Choi YS Song JE Lee JE Kim E Kim CH Kim DH .
Hyperpolarized [1-13C]lactate flux increased in the hippocampal region in diabetic mice . Mol Brain . (2019 ) 12 :88 . doi: 10.1186/s13041-019-0505-9 31675964
PMC6824044
196Huang Y Lei L Liu D Jovin I Russell R Johnson RS .
Normal glucose uptake in the brain and heart requires an endothelial cell-specific HIF-1α-dependent function . Proc Natl Acad Sci USA . (2012 ) 109 :17478–83. doi: 10.1073/pnas.1209281109 23047702
PMC3491491
197Wang J Cui Y Yu Z Wang W Cheng X Ji W .
Brain endothelial cells maintain lactate homeostasis and control adult hippocampal neurogenesis . Cell Stem Cell . (2019 ) 25 :754 –767.e759 . doi: 10.1016/j.stem.2019.09.009 31761722
198Yang Y Song L Yu L Zhang J Zhang B .
H4K12 lactylation potentiates mitochondrial oxidative stress via the Foxo1 pathway in diabetes-induced cognitive impairment . J Adv Res . (2025 ) 78 :391 –407 . doi: 10.1016/j.jare.2025.02.020 39965729
PMC12684910
199Adler D Martín-Ávila A Cheng E Oliveira MM Zhang M Evans HT .
Activity-dependent protein synthesis in neurons requires microglial-metabolic coupling . Cell Metab . (2026 ) 38 :1650 –1667.e1610 . doi: 10.1016/j.cmet.2026.05.006 42242219
PMC13245367
200Zhao L Dong M Ren M Li C Zheng H Gao H .
Metabolomic analysis identifies lactate as an important pathogenic factor in diabetes-associated cognitive decline rats . Mol Cell Proteomics . (2018 ) 17 :2335–46. doi: 10.1074/mcp.ra118.000690 30171160
PMC6283288
201Ardanaz CG de la Cruz A Minhas PS Hernández-Martín N Pozo M Valdecantos MP .
Astrocytic GLUT1 reduction paradoxically improves central and peripheral glucose homeostasis . Sci Adv . (2024 ) 10 :eadp1115. doi: 10.1126/sciadv.adp1115 39423276
PMC11488540
202Backes H Walberer M Ladwig A Rueger MA Neumaier B Endepols H .
Glucose consumption of inflammatory cells masks metabolic deficits in the brain . Neuroimage . (2016 ) 128 :54 –62 . doi: 10.1016/j.neuroimage.2015.12.044 26747749
PMC4767221
203Creo AL Cortes TM Jo HJ Huebner AR Dasari S Tillema JM .
Brain functions and cognition on transient insulin deprivation in type 1 diabetes . JCI Insight . (2021 ) 6 . doi: 10.1172/jci.insight.144014 33561011
PMC8021100
204Zaidi M Ma J Thomas BP Peña S Harrison CE Chen J .
Functional activation of pyruvate dehydrogenase in human brain using hyperpolarized [1-(13) C]pyruvate . Magn Reson Med . (2024 ) 91 :1822–33.
10.1002/mrm.30015 PMC10950523 38265104
205Wang N Zhao LC Zheng YQ Dong MJ Su Y Chen WJ .
Alteration of interaction between astrocytes and neurons in different stages of diabetes: a nuclear magnetic resonance study using [1-(13)C]glucose and [2-(13)C]acetate . Mol Neurobiol . (2015 ) 51 :843–52.
10.1007/s12035-014-8808-4 25048983
206Dong M Ren M Li C Zhang X Yang C Zhao L .
Analysis of metabolic alterations related to pathogenic process of diabetic encephalopathy rats . Front Cell Neurosci . (2018 ) 12 :527 .
10.3389/fncel.2018.00527 PMC6339875 30692917
207Wang WT Lee P Yeh HW Smirnova IV Choi IY .
Effects of acute and chronic hyperglycemia on the neurochemical profiles in the rat brain with streptozotocin-induced diabetes detected using in vivo ¹H MR spectroscopy at 9.4 T . J Neurochem . (2012 ) 121 :407–17. doi: 10.1111/j.1471-4159.2012.07698.x 22353009
PMC3323743
208Garcia-Serrano AM Mohr AA Philippe J Skoug C Spégel P Duarte JMN .
Cognitive impairment and metabolite profile alterations in the hippocampus and cortex of male and female mice exposed to a fat and sugar-rich diet are normalized by diet reversal . Aging Dis . (2022 ) 13 :267–83.
10.14336/AD.2021.0720 PMC8782561 35111373
209Choi IY Wang WT Kim B Hur J Robbins DC Jang DG .
Non-invasive in vivo measurements of metabolic alterations in the type 2 diabetic brain by (1)H magnetic resonance spectroscopy . J Neurochem . (2024 ) 168 :765–80.
10.1111/jnc.15996 PMC11093888 37965761
210Dai W Duan W Alfaro FJ Gavrieli A Kourtelidis F Novak V .
The resting perfusion pattern associates with functional decline in type 2 diabetes . Neurobiol Aging . (2017 ) 60 :192 –202 . doi: 10.1016/j.neurobiolaging.2017.09.004 28992987
PMC5687828
211Jacobson AM Ryan CM Braffett BH Gubitosi-Klug RA Lorenzi GM Luchsinger JA .
Cognitive performance declines in older adults with type 1 diabetes: results from 32 years of follow-up in the DCCT and EDIC Study . Lancet Diabetes Endocrinol . (2021 ) 9 :436–45. doi: 10.1016/s2213-8587(21)00086-3 34051936
PMC8583716
212Tuligenga RH Dugravot A Tabák AG Elbaz A Brunner EJ Kivimäki M .
Midlife type 2 diabetes and poor glycaemic control as risk factors for cognitive decline in early old age: a post-hoc analysis of the Whitehall II cohort study . Lancet Diabetes Endocrinol . (2014 ) 2 :228–35. doi: 10.1016/s2213-8587(13)70192-x 24622753
PMC4274502
213Yu JH Kim REY Park SY Lee DY Cho HJ Kim NH .
Association of long-term hyperglycaemia and insulin resistance with brain atrophy and cognitive decline: A longitudinal cohort study . Diabetes Obes Metab . (2023 ) 25 :1091–100. doi: 10.1111/dom.14958 36564910
214Li TC Yang CP Tseng ST Li CI Liu CS Lin WY .
Visit-to-visit variations in fasting plasma glucose and HbA(1c) associated with an increased risk of Alzheimer disease: Taiwan Diabetes Study . Diabetes Care . (2017 ) 40 :1210–7. doi: 10.2337/dc16-2238 28705834
215van den Berg E Reijmer YD de Bresser J Kessels RP Kappelle LJ Biessels GJ .
A 4 year follow-up study of cognitive functioning in patients with type 2 diabetes mellitus . Diabetologia . (2010 ) 53 :58 –65 .
10.1007/s00125-009-1571-9 PMC2789935 19882137
216Launer LJ Miller ME Williamson JD Lazar RM Gerstein HC Murray AM .
Effects of intensive glucose lowering on brain structure and function in people with type 2 diabetes (ACCORD MIND): a randomised open-label substudy . Lancet Neurol . (2011 ) 10 :969–77. doi: 10.1016/s1474-4422(11)70188-0 21958949
PMC3333485
217Heller SR Cryer PE .
Reduced neuroendocrine and symptomatic responses to subsequent hypoglycemia after 1 episode of hypoglycemia in nondiabetic humans . Diabetes . (1991 ) 40 :223–6. doi: 10.2337/diabetes.40.2.223
1991573
218Dagogo-Jack SE Craft S Cryer PE .
Hypoglycemia-associated autonomic failure in insulin-dependent diabetes mellitus. Recent antecedent hypoglycemia reduces autonomic responses to, symptoms of, and defense against subsequent hypoglycemia . J Clin Invest . (1993 ) 91 :819–28. doi: 10.1172/jci116302 8450063
PMC288033
219Cranston I Lomas J Maran A Macdonald I Amiel SA .
Restoration of hypoglycaemia awareness in patients with long-duration insulin-dependent diabetes . Lancet . (1994 ) 344 :283–7. doi: 10.1016/s0140-6736(94)91336-6 7914259
220Oz G Kumar A Rao JP Kodl CT Chow L Eberly LE .
Human brain glycogen metabolism during and after hypoglycemia . Diabetes . (2009 ) 58 :1978–85.
10.2337/db09-0226 PMC2731528 19502412
221Öz G DiNuzzo M Kumar A Moheet A Khowaja A Kubisiak K .
Cerebral glycogen in humans following acute and recurrent hypoglycemia: Implications on a role in hypoglycemia unawareness . J Cereb Blood Flow Metab . (2017 ) 37 :2883–93.
10.1177/0271678X16678240 PMC5536796 27834283
222Su G Farhat R Laxman AK Chapman-Natewa K Nelson IE Chan O .
Astrocyte glycogen is a major source of hypothalamic lactate in rats with recurrent hypoglycemia . Diabetes . (2023 ) 72 :1154–60. doi: 10.2337/db22-0902 37216640
PMC10382650
223He C Li Q Cui Y Gao P Shu W Zhou Q .
Recurrent moderate hypoglycemia accelerates the progression of Alzheimer's disease through impairment of the TRPC6/GLUT3 pathway . JCI Insight . (2022 ) 7 . doi: 10.1172/jci.insight.154595 35077394
PMC8983129
224Raichle ME Mintun MA .
Brain work and brain imaging . Annu Rev Neurosci . (2006 ) 29 :449–76. doi: 10.1146/annurev.neuro.29.051605.112819 16776593
225Schneider-Garces NJ Gordon BA Brumback-Peltz CR Shin E Lee Y Sutton BP .
Span, CRUNCH, and beyond: working memory capacity and the aging brain . J Cognit Neurosci . (2010 ) 22 :655–69. doi: 10.1162/jocn.2009.21230 19320550
PMC3666347
226Jamadar SD Ward PG Li S Sforazzini F Baran J Chen Z .
Simultaneous task-based BOLD-fMRI and [18-F] FDG functional PET for measurement of neuronal metabolism in the human visual cortex . Neuroimage . (2019 ) 189 :258–66. doi: 10.1016/j.neuroimage.2019.01.003 30615952
227Stiernman L Grill F McNulty C Bahrd P Panes Lundmark V Axelsson J .
Widespread fMRI BOLD signal overactivations during cognitive control in older adults are not matched by corresponding increases in fPET glucose metabolism . J Neurosci . (2023 ) 43 :2527–36. doi: 10.1523/jneurosci.1331-22.2023 36868855
PMC10082451
228Blockley NP Griffeth VE Simon AB Buxton RB .
A review of calibrated blood oxygenation level-dependent (BOLD) methods for the measurement of task-induced changes in brain oxygen metabolism . NMR BioMed . (2013 ) 26 :987 –1003 .
10.1002/nbm.2847 PMC3639302 22945365
229Restom K Perthen JE Liu TT .
Calibrated fMRI in the medial temporal lobe during a memory-encoding task . Neuroimage . (2008 ) 40 :1495–502. doi: 10.1016/j.neuroimage.2008.01.038 18329291
PMC2430147
230Goodwin JA Vidyasagar R Balanos GM Bulte D Parkes LM .
Quantitative fMRI using hyperoxia calibration: reproducibility during a cognitive Stroop task . Neuroimage . (2009 ) 47 :573–80. doi: 10.1016/j.neuroimage.2009.04.064 19398018
231Willie CK Tzeng YC Fisher JA Ainslie PN .
Integrative regulation of human brain blood flow . J Physiol . (2014 ) 592 :841–59. doi: 10.1113/jphysiol.2013.268953 24396059
PMC3948549
232Rodrigues TB Valette J Bouzier-Sore AK .
(13)C NMR spectroscopy applications to brain energy metabolism . Front Neuroenergetics . (2013 ) 5 :9. doi: 10.3389/fnene.2013.00009 24367329
PMC3856424
233Lai M Lanz B Poitry-Yamate C Romero JF Berset CM Cudalbu C .
In vivo (13)C MRS in the mouse brain at 14.1 Tesla and metabolic flux quantification under infusion of [1,6-(13)C(2)]glucose . J Cereb Blood Flow Metab . (2018 ) 38 :1701–14.
10.1177/0271678X17734101 PMC6168901 29047296
234Díaz-García CM Lahmann C Martínez-François JR Li B Koveal D Nathwani N .
Quantitative in vivo imaging of neuronal glucose concentrations with a genetically encoded fluorescence lifetime sensor . J Neurosci Res . (2019 ) 97 :946–60.
10.1002/jnr.24433 PMC6565483 31106909
235Marvin JS Mächler P Meng C Ates T Patel RH Adhikari R .
iGlucoSnFR2: A genetically encoded fluorescent sensor for measuring intracellular or extracellular glucose in vivo in mouse brain . Sci Adv . (2025 ) 11 :eadz3889. doi: 10.1126/sciadv.adz3889 41223288
PMC12609159
236San Martín A Ceballo S Ruminot I Lerchundi R Frommer WB Barros LF .
A genetically encoded FRET lactate sensor and its use to detect the Warburg effect in single cancer cells . PloS One . (2013 ) 8 :e57712.
10.1371/journal.pone.0057712 PMC3582500 23469056
237Dienel GA Rothman DL .
In vivo calibration of genetically encoded metabolite biosensors must account for metabolite metabolism during calibration and cellular volume . J Neurochem . (2024 ) 168 :506–32.
10.1111/jnc.15775 36726217
238Marvin JS Kokotos AC Kumar M Pulido C Tkachuk AN Yao JS .
iATPSnFR2: A high-dynamic-range fluorescent sensor for monitoring intracellular ATP . Proc Natl Acad Sci USA . (2024 ) 121 :e2314604121. doi: 10.1073/pnas.2314604121 38748581
PMC11126915
239Wang L Xing X Zeng X Jackson SR TeSlaa T Al-Dalahmah O .
Spatially resolved isotope tracing reveals tissue metabolic activity . Nat Methods . (2022 ) 19 :223–30. doi: 10.1038/s41592-021-01378-y 35132243
PMC10926149
240Qi M Philip MC Yang N Sweedler JV .
Single cell neurometabolomics . ACS Chem Neurosci . (2018 ) 9 :40 –50 .
10.1021/acschemneuro.7b00304 PMC5800855 28982006
241Kaggie JD Khan AS Matys T Schulte RF Locke MJ Grimmer A .
Deuterium metabolic imaging and hyperpolarized (13)C-MRI of the normal human brain at clinical field strength reveals differential cerebral metabolism . Neuroimage . (2022 ) 257 :119284. doi: 10.1016/j.neuroimage.2022.119284 35533826
242Grist JT McLean MA Riemer F Schulte RF Deen SS Zaccagna F .
Quantifying normal human brain metabolism using hyperpolarized [1-(13)C]pyruvate and magnetic resonance imaging . Neuroimage . (2019 ) 189 :171–9. doi: 10.1016/j.neuroimage.2019.01.027 30639333
PMC6435102
243Pearl J .
Interpretation and identification of causal mediation . Psychol Methods . (2014 ) 19 :459–81. doi: 10.1037/a0036434 24885338
244VanderWeele TJ .
Mediation analysis: A practitioner's guide . Annu Rev Public Health . (2016 ) 37 :17 –32 . doi: 10.1146/annurev-publhealth-032315-021402 26653405
245Imai K Keele L Tingley D .
A general approach to causal mediation analysis . Psychol Methods . (2010 ) 15 :309–34. doi: 10.1037/a0020761 20954780
246Jais A Solas M Backes H Chaurasia B Kleinridders A Theurich S .
Myeloid-cell-derived VEGF maintains brain glucose uptake and limits cognitive impairment in obesity . Cell . (2016 ) 165 :882–95. doi: 10.1016/j.cell.2016.08.010 27133169
247Park G Lee JY Han HM An HS Jin Z Jeong EA .
Ablation of dynamin-related protein 1 promotes diabetes-induced synaptic injury in the hippocampus . Cell Death Dis . (2021 ) 12 :445 . doi: 10.1038/s41419-021-03723-7 33953167
PMC8099876
248Yao Y Shi J Zhang C Gao W Huang N Liu Y .
Pyruvate dehydrogenase kinase 1 protects against neuronal injury and memory loss in mouse models of diabetes . Cell Death Dis . (2023 ) 14 :722 . doi: 10.1038/s41419-023-06249-2 37935660
PMC10630521
249Novak V Mantzoros CS Novak P McGlinchey R Dai W Lioutas V .
MemAID: Memory advancement with intranasal insulin vs. placebo in type 2 diabetes and control participants: a randomized clinical trial . J Neurol . (2022 ) 269 :4817–35. doi: 10.1007/s00415-022-11119-6 35482079
PMC9046533
250Nijssen KMR Mensink RP Joris PJ .
Effects of intranasal insulin administration on cerebral blood flow and cognitive performance in adults: A systematic review of randomized, placebo-controlled intervention studies . Neuroendocrinology . (2023 ) 113 :1 –13 .
10.1159/000526717 36219990
251Spauwen PJ Köhler S Verhey FR Stehouwer CD van Boxtel MP .
Effects of type 2 diabetes on 12-year cognitive change: results from the Maastricht Aging Study . Diabetes Care . (2013 ) 36 :1554–61. doi: 10.2337/dc12-0746 23275366
PMC3661848
252Marseglia A Fratiglioni L Kalpouzos G Wang R Bäckman L Xu W .
Prediabetes and diabetes accelerate cognitive decline and predict microvascular lesions: A population-based cohort study . Alzheimers Dement . (2019 ) 15 :25 –33 .
10.1016/j.jalz.2018.06.3060 30114414
253Samaras K Lutgers HL Kochan NA Crawford JD Campbell LV Wen W .
The impact of glucose disorders on cognition and brain volumes in the elderly: the Sydney Memory and Ageing Study . Age (Dordr) . (2014 ) 36 :977–93. doi: 10.1007/s11357-013-9613-0 24402401
PMC4039246
254Hu J Fang M Pike JR Lutsey PL Sharrett AR Wagenknecht LE .
Prediabetes, intervening diabetes and subsequent risk of dementia: the Atherosclerosis Risk in Communities (ARIC) study . Diabetologia . (2023 ) 66 :1442–9. doi: 10.1007/s00125-023-05930-7 37221246
PMC10467356
255Cui X Abduljalil A Manor BD Peng CK Novak V .
Multi-scale glycemic variability: a link to gray matter atrophy and cognitive decline in type 2 diabetes . PloS One . (2014 ) 9 :e86284. doi: 10.1371/journal.pone.0086284 24475100
PMC3901681
256Lee DY Kim J Park S Park SY Yu JH Seo JA .
Fasting glucose variability and the risk of dementia in individuals with diabetes: A nationwide cohort study . Diabetes Metab J . (2022 ) 46 :923–35. doi: 10.4093/dmj.2021.0346 35609876
PMC9723200
257Whitmer RA Karter AJ Yaffe K Quesenberry CP Jr.Selby JV .
Hypoglycemic episodes and risk of dementia in older patients with type 2 diabetes mellitus . Jama . (2009 ) 301 :1565–72. doi: 10.1001/jama.2009.460 19366776
PMC2782622
258Yaffe K Falvey CM Hamilton N Harris TB Simonsick EM Strotmeyer ES .
Association between hypoglycemia and dementia in a biracial cohort of older adults with diabetes mellitus . JAMA Intern Med . (2013 ) 173 :1300–6. doi: 10.1001/jamainternmed.2013.6176 23753199
PMC4041621
259Lee AK Rawlings AM Lee CJ Gross AL Huang ES Sharrett AR .
Severe hypoglycaemia, mild cognitive impairment, dementia and brain volumes in older adults with type 2 diabetes: the Atherosclerosis Risk in Communities (ARIC) cohort study . Diabetologia . (2018 ) 61 :1956–65. doi: 10.1007/s00125-018-4668-1 29961106
PMC6152822
260Alkabbani W Maxwell CJ Marrie RA Tyas SL Lega IC Gamble JM .
Associations of mid- and late-life severe hypoglycemic episodes with incident dementia among patients with type 2 diabetes: A population-based cohort study . Diabetes Care . (2023 ) 46 :331–40.
10.2337/dc22-1496 36516080
261Hawks ZW Beck ED Jung L Fonseca LM Sliwinski MJ Weinstock RS .
Dynamic associations between glucose and ecological momentary cognition in Type 1 Diabetes . NPJ Digit Med . (2024 ) 7 :59 . doi: 10.1093/geroni/igae098.2212
PMC10948782 38499605
262Hoogendoorn CJ Hernandez R Schneider S Peters A Hawkins M Pyatak EA .
Dynamic relationships among continuous glucose metrics and momentary cognitive performance in diverse adults with type 1 diabetes . Diabetes Care . (2025 ) 48 :799 –806 . doi: 10.2337/dc24-2431 40163370
PMC12034895
263Sugimoto T Araki A Fujita H Fujita K Honda K Inagaki N .
Multidomain Intervention Trial for Preventing Cognitive Decline among Older Adults with Type 2 Diabetes: J-MIND-Diabetes . J Prev Alzheimers Dis . (2024 ) 11 :1604–14. doi: 10.14283/jpad.2024.117 39559873
PMC11573805
264Calamia M Markon K Tranel D .
Scoring higher the second time around: meta-analyses of practice effects in neuropsychological assessment . Clin Neuropsychol . (2012 ) 26 :543–70. doi: 10.1080/13854046.2012.680913 22540222
265Scharfen J Jansen K Holling H .
Retest effects in working memory capacity tests: A meta-analysis . Psychon Bull Rev . (2018 ) 25 :2175–99. doi: 10.3758/s13423-018-1461-6 29907925
266Goldberg TE Harvey PD Wesnes KA Snyder PJ Schneider LS .
Practice effects due to serial cognitive assessment: Implications for preclinical Alzheimer's disease randomized controlled trials . Alzheimers Dement (Amst) . (2015 ) 1 :103–11. doi: 10.1016/j.dadm.2014.11.003 27239497
PMC4876902
267Elliott ML Knodt AR Ireland D Morris ML Poulton R Ramrakha S .
What is the test-retest reliability of common task-functional MRI measures? New empirical evidence and a meta-analysis . Psychol Sci . (2020 ) 31 :792 –806 . doi: 10.1177/0956797620916786 32489141
PMC7370246
268Duff K .
Evidence-based indicators of neuropsychological change in the individual patient: relevant concepts and methods . Arch Clin Neuropsychol . (2012 ) 27 :248–61. doi: 10.1093/arclin/acr120 22382384
PMC3499091
269Yang FN Xu S Spaeth A Galli O Zhao K Fang Z .
Test-retest reliability of cerebral blood flow for assessing brain function at rest and during a vigilance task . Neuroimage . (2019 ) 193 :157–66. doi: 10.1016/j.neuroimage.2019.03.016 30894335
270Terpstra M Cheong I Lyu T Deelchand DK Emir UE Bednařík P .
Test-retest reproducibility of neurochemical profiles with short-echo, single-voxel MR spectroscopy at 3T and 7T . Magn Reson Med . (2016 ) 76 :1083–91.
10.1002/mrm.26022 PMC4846596 26502373
Come citare questo studio
Mingyang Li, Min Ye, Hongyu Xie, Ting Gao, Wenjing Kan, Yiming Chen, Yu Xia, Yonghua Zhao, Aihong Yuan. (2026). Cognitive demand as a neurometabolic stress test in diabetes-associated cognitive impairment: from static metabolite pools to dynamic energy responses.. Frontiers in endocrinology. DOI: 10.3389/fendo.2026.1991245
Consulta la pubblicazione originale ↗