The NVU is the core structural platform for realizing NVC functions. Rather than being a simple cell collection, it is a highly integrated, multi-cellular functional module whose integrity is essential for maintaining brain homeostasis, regulating BBB permeability and accurately matching local CBF with neuronal metabolic demands [11]. Core cellular members of the NVU include neurons, astrocytes, endothelial cells, pericytes and vascular smooth muscle cells, as well as microglial cells, the basement membrane and the extracellular matrix that encases these cells (Fig. 1). The precise cellular composition, structural arrangement and complex intercellular interactions together form the physiological foundation for NVC. Structural or functional abnormalities in any NVU component can disrupt overall coordination, leading to NVC dysfunction and BBB integrity disruption.

Fig. 1.

Schematic illustration of the neurovascular unit (NVU) and its cellular components. The NVU is a multicellular functional unit that maintains brain homeostasis and regulates the blood-brain barrier. The magnified view displays the core cellular components, including endothelial cells, pericytes, astrocytes, microglia, neurons, and oligodendrocytes, highlighting their spatial organization around a cerebral blood vessel. Created with BioRender (https://www.biorender.com/).

Within the NVU, different cells play unique and coordinated roles (Table 1, Ref. [16, 17, 18, 19, 20, 21, 22, 23, 24]). Neurons, as initiators of NVC, release neurotransmitters such as glutamate, directly activating neuronal-type nitric oxide synthase (nNOS) to generate NO, or by indirectly acting on astrocytes via metabotropic glutamate receptors to trigger vasodilation. Neurons also release adenosine triphosphate (ATP) and prostaglandins (PGE) (e.g., PGE₂), which further regulate vascular tone [16]. Specific gamma-aminobutyric acid (GABA) interneurons also release NO or vasoactive peptides to assist in blood flow regulation [25]. Astrocytes, as integrators, surround blood vessels with their vascular end-feet. They respond to glutamate released by neurons, raising intracellular calcium (Ca²⁺) levels and releasing vasoactive substances such as potassium ions (K⁺), arachidonic acid (AA) metabolites and ATP, finely regulating vascular contraction and dilation. Astrocytic signal integration is influenced by local oxygen concentrations and vascular tone, allowing dynamic vascular responses according to the microenvironment [26].

Brain microvascular endothelial cells not only form the BBB but also actively participate in NVC by producing NO through endothelial nitric oxide synthase (eNOS), synthesizing prostacyclin 2 (PGI₂), and mediating vasodilation through the endothelium-derived hyperpolarizing factor (EDHF) pathway [17]. Additionally, endothelial cells produce the vasoconstrictor endothelin-1 (ET-1), with its release forming a dynamic balance with vasodilatory substances [27]. Endothelial cell Ca²⁺ signal integration comes from neuronal and astrocytic signals that responds to various stimuli through transient receptor potential (TRP) channels, coordinating and propagating vasodilation signals upstream along the vascular tree [28]. Pericytes envelop capillaries and respond to signals such as glutamate, ATP, norepinephrine and AA metabolites with their contractile proteins, directly regulating capillary diameter, potentially reacting before small arteries in NVC [29]. Microglial cells, as immune sentinels, sense ATP/adenosine diphosphate (ADP) through their purinergic receptor Y12 (P2Y12) receptors and hydrolyze ATP/ADP to adenosine via Cluster of Differentiation 39 (CD39) ectonucleotidase, thus regulating local CBF and NVC, partially independent of the NO pathway [18]. Oligodendrocytes mainly promote NVC indirectly by forming myelin and providing metabolic support to axons, maintaining neuronal function and the normal demands of NVC [19]. Finally, vascular smooth muscle cells (VSMCs), as the primary effector cells of small artery walls, integrate various vasodilation and vasoconstriction signals from the NVU components, precisely regulating CBF by changing their contraction state and vascular diameter, with their reactivity significantly influenced by pre-existing vascular tension [20].

The NVU’s structure shows high specificity along the vascular tree (Fig. 1). Small arterial walls consist of endothelial cells, a basement membrane, multi-layered VSMCs and astrocyte end-feet, with the possibility of a perivascular space. By contrast, the capillary wall is simpler, composed of endothelial cells, a shared basement membrane, embedded pericytes and tightly surrounding astrocyte end-feet, without the VSMC layer [30]. This precise structural arrangement facilitates efficient signal transmission within the NVU. The BBB is a critical manifestation of the NVU’s structure and function, with the core being the physical barrier formed by tight junctions between brain microvascular endothelial cells. Pericytes and astrocytes induce and maintain the endothelial cell barrier through signaling molecules or direct contact [31]. BBB integrity is crucial for maintaining CNS homeostasis, protecting neurons from harmful substances, and ensuring proper NVC function [32].

NVC is initiated by neuronal electrical activity and synaptic transmission. Neurons release a variety of neurotransmitters and neuromodulators, which trigger a complex signaling cascade that regulates local CBF [33]. Glutamate, released by excitatory neurons, is one of the primary triggers for NVC. Glutamate acts on N-methyl-D-aspartate receptors on postsynaptic neurons, leading to Ca²⁺ influx. This activation subsequently triggers nNOS to produce the vasodilator NO, or catalyzes the synthesis of PGE₂ via cyclooxygenase-2, both of which promote vasodilation [34]. Inhibitory interneurons also play a complex role in NVC regulation, mediating both vasodilation and vasoconstriction. For instance, interneurons expressing somatostatin release neuropeptide Y (NPY), which induces vasoconstriction via Y1 receptors, while neurons expressing vasoactive intestinal peptide (VIP) or nNOS promote vasodilation [35]. Recent studies have identified a specific type of cortical GABAergic neuron, co-expressing nNOS and tachykinin receptor 1, that plays a critical role in mediating CBF increases through feedforward excitatory pathways and the release of signaling molecules such as NO. Their axons project both locally and over long distances, regulating pericyte relaxation [36]. Additionally, neurons with subcortical regions (such as the basal ganglia and locus coeruleus) directly project to cortical microvessels, influencing local blood flow through the release of acetylcholine (ACh), norepinephrine and serotonin (5-HT) [37]. Recent research has further demonstrated that in deeper brain regions, glutamatergic axons cross the gaps in astrocytic end-feet, forming tight “synaptic-like” structures with arteriolar smooth muscle cells, mediating direct and rapid neuronal control over blood vessels, thereby providing a new direct pathway for NVC [10]. Neurons utilize various vasoactive molecules through different signaling pathways to ensure the robustness and precision of the NVC response [38].

In addition to classical neurotransmitter signaling, ion concentration changes induced by neuronal activity are also key signals in initiating NVC. During neuronal firing, intracellular K⁺ rapidly effluxes through delayed rectifier potassium channels, causing a rapid increase in the K⁺ concentration in the synaptic cleft. This elevated K⁺ concentration can be directly detected by nearby vascular smooth muscle cells and inward rectifier potassium (Kir) channels 2.1 on pericyte membranes. The activation of Kir channels leads to hyperpolarization of the vascular cell membrane, closing voltage-dependent Ca²⁺ channels, which ultimately results in vasodilation. This “neuron-K⁺-Kir channel” pathway bypasses complex intermediate signaling cascades and is considered one of the fastest feedforward mechanisms for initiating functional hyperemia [39]. However, the efficiency of this signaling pathway is tightly regulated by the precise microstructural organization of the neurovascular unit. Computational model studies suggest that the dense, sheath-like structure formed by the terminal astrocytic end-feet around blood vessels acts as a physical barrier, restricting the diffusion of K⁺ from the synapse to the blood vessel, thereby limiting the propagation of the K⁺ signal. Consequently, the ability of the K⁺ signal to effectively trigger vasodilation may depend on the geometric configuration of the astrocytic end-feet or their dynamic changes during physiological processes [40].

Astrocytes are key integrators in the transmission of signals between neurons and blood vessels in NVC. They regulate vascular responses through complex Ca²⁺ signal transduction and the release of various vasoactive substances. Neuronal activity, particularly glutamatergic signaling, activates metabotropic glutamate receptors on astrocytic membranes, promoting the release of Ca²⁺ from intracellular stores via the inositol trisphosphate (InsP3) pathway, leading to an increase in intracellular Ca²⁺ levels. These Ca²⁺ signals spread within the cell and potentially are transferred to neighboring cells, ultimately reaching the astrocytic end-feet surrounding blood vessels [41]. Elevated intracellular Ca²⁺ levels in astrocytes initiate the synthesis and/or release of various vasoactive substances, including PGE₂, epoxyeicosatrienoic acids (EETs) and K⁺ efflux, all of which induce vasodilation. PGE₂ and EETs act on corresponding receptors or channels on VSMCs, while increased extracellular K⁺ concentration induces vasodilation by activating Kir channels on VSMCs. Alternatively, astrocyte-derived 20-hydroxyeicosatetraenoic acid (20-HETE) and released ATP cause vasoconstriction [42]. The net effect of astrocyte-derived signals (vasodilation or vasoconstriction) is dynamically regulated by current vascular tone and the local microenvironment, positioning astrocytes as “gatekeepers” [16].

The ultimate effectors of NVC are vascular cells — endothelial cells, pericytes and VSMCs — which integrate upstream signals and regulate vessel diameter to control CBF [38]. Endothelial cells are located on the inner surface of blood vessels, serving as a critical interface between blood flow and brain tissue. These cells play a central role in sensing, integrating, and transmitting NVC signals to smooth muscle cells. They produce and release various vasoactive factors, promoting signal propagation along the vessel wall. Endothelial-derived vasodilators include NO, prostacyclin (PGI₂) and EDHFs, while ET-1 primarily acts as a vasoconstrictor [43]. Pericytes, which closely encase the outer walls of capillaries between endothelial cells and astrocyte end-feet, actively regulate capillary diameter, thus precisely modulating microcirculatory blood flow [44]. The relative contribution of pericytes and upstream small artery smooth muscle in regulating local CBF remains a core controversy in the field of NVC. One perspective proposes that pericytes serve as the primary regulatory site in NVC. Pioneering research from Attwell’s lab [23] has demonstrated that pericytes possess potent contractile abilities, with their diameter changes sufficiently altering the vascular resistance of the entire capillary network, thereby controlling local CBF. In contrast, an alternative view challenges this notion, suggesting that the primary regulation of blood flow occurs in the upstream small arteries, which possess stronger contractile capacity. Research from Grutzendler’s lab [45] provides opposing evidence, arguing that under physiological conditions, the contractile force of pericytes is insufficient to significantly alter overall blood flow, while the dilation and constriction of small artery smooth muscle are the key determinants of regional blood flow responses. Pericytes express a range of receptors and ion channels, enabling them to respond to signals from neurons and astrocytes, and regulate their contraction through Ca²⁺ signaling [46]. Recent research has highlighted the critical role of thin-strand pericytes in NVC. These cells detect increased neuronal activity and changes in local metabolic states, generating electrical signals via ATP-sensitive potassium (KATP) channels. These signals propagate retrogradely to upstream arterioles, causing vasodilation and increasing local blood supply [46, 47]. Optogenetic activation of individual thin-strand pericytes is sufficient to regulate arteriolar diameter remotely, confirming their importance in signal transmission [46]. VSMCs, which surround the walls of arterioles, are the final effector cells in the NVC pathway. They regulate local cerebral blood flow by altering vessel diameter through contraction and relaxation [11]. VSMCs integrate a variety of vasodilatory signals, including NO, PGI₂, EDHFs/hyperpolarization signals from endothelial cells and signals from neurons or astrocytes (e.g., PGE₂, K⁺, adenosine). Potassium channels play a key role in VSMC relaxation. Their opening leads to potassium efflux, membrane hyperpolarization and inhibition of voltage-dependent Ca²⁺ channel opening, reducing intracellular Ca²⁺ concentrations and resulting in smooth muscle relaxation [48].

NVC is a dynamically integrated multi-loop control system, whose regulation depends on the precise coordination of feedforward and feedback mechanisms. The feedforward loop, primarily driven by neuronal activity, involves rapid signal transmission processes, including glutamate-mediated neurotransmitter release, calcium wave propagation in astrocytes, and the swift generation of vasoactive molecules such as K⁺, NO, and PGE₂. These processes enable the early dilation of the local microvascular bed before there is a significant increase in metabolic demand, resulting in “predictive” perfusion. This can even lead to a temporary, excessive increase in regional cerebral blood flow that exceeds metabolic needs [49]. In contrast, the feedback loop operates at a slower pace and includes not only traditional metabolic by-products such as CO₂, O₂, lactate, and adenosine, but more critically, a series of active negative feedback mechanisms. These mechanisms fine-tune vascular tone in response to changes in local metabolic load, ensuring that CBF is restored to baseline levels with precision [16, 49].

In these negative feedback mechanisms, endothelial cells detect shear stress induced by congestion through mechanosensitive ion channels, activating vasoconstriction signals to actively modulate the magnitude and duration of the congestion response, thereby preventing excessive shear stress from disrupting the microenvironment. Concurrently, vascular smooth muscle cells restore basal vascular tone through contraction, protecting downstream microvessels from damage due to pressure fluctuations [50]. Furthermore, several key molecules regulate the balance between vasodilation and vasoconstriction: when neuronal activity decreases and NO levels decline, the inhibition of 20-HETE synthesis is lifted, resulting in an increase in its concentration, which actively drives vasoconstriction to restore blood flow [51]. Notably, functional congestion can lead to local CO₂ depletion, causing transient hypocapnia, which subsequently triggers rebound vasoconstriction. Metabolites, such as adenosine, also play a role in regulating vascular tone, thereby maintaining tissue homeostasis [52]. Hence, the dynamic integration and balance of feedforward and feedback mechanisms collectively constitute the “fail-safe” system of NVC, ensuring rapid response and precise matching.

Additionally, NVC characteristics are significantly influenced by the brain’s overall functional state and diffuse neuromodulatory systems. Different types of anesthetics and depths of anesthesia profoundly impact the baseline, response amplitude, and spatiotemporal dynamics of NVC through multiple pathways, including alterations in neuronal excitability, ion channel function, receptor sensitivity and systemic physiological parameters [53, 54]. Research has shown that during “sleep inertia” following sleep deprivation, electroencephalography-functional magnetic resonance imaging (EEG-fMRI) coupling patterns, used as an alternative NVC index, exhibit dynamic changes, suggesting that NVC may not remain constant across varying states of alertness [55]. Additionally, ascending neuromodulatory systems originating from the brainstem or basal forebrain, including ACh, Norepinephrine, 5-HT, and dopamine, project widely to the cerebral cortex and other regions, modulating NVU components to regulate their signal pathways, ion channel activities and receptor sensitivities [56]. This higher-level regulation endows NVC with enhanced adaptability, allowing it to dynamically adjust based on the brain’s varying physiological and behavioral demands.

Blood-oxygen-level-dependent functional magnetic resonance imaging (BOLD-fMRI) is one of the most widely used non-invasive techniques for brain functional imaging. It provides an indirect reflection of the NVC process by detecting dynamic changes in local brain oxygenation levels driven by neuronal activity [57]. The underlying principle is that neural activity induces functional hyperemia, which lowers the local concentration of deoxygenated hemoglobin, leading to an increase in MRI signals (T2 signals) [58]. The key advantages of BOLD-fMRI include its high spatial resolution (down to millimeters), its ability to cover the entire brain, and its completely non-invasive nature [59]. However, BOLD signals do not directly measure neuronal firing and are susceptible to various sources of interference, such as head motion, physiological noise, and the effects of large blood vessels [60]. In particular, in the presence of brain diseases, pathological changes in the NVU result in a phenomenon known as “neurovascular uncoupling”, where BOLD signals no longer accurately reflect neural activity [61]. Consequently, BOLD-fMRI results often need to be validated using other techniques that directly measure neural activity or more directly assess CBF.

BOLD-fMRI remains a cornerstone in cognitive neuroscience research, extensively applied to assess brain functional activation patterns and NVC characteristics in both healthy individuals and those with various diseases. For example, in migraine research, fMRI has not only highlighted dynamic blood oxygen level changes in the visual cortex at different stages of the migraine cycle but also revealed that even during the interictal phase without prodromal symptoms, patients with migraines may exhibit persistent abnormalities or dysfunction in baseline cortical activity, particularly in the visual cortex, which may be linked to migraine susceptibility [62]. Recently, the evaluation of cerebrovascular reactivity (CVR) using BOLD-fMRI has attracted increasing attention [63]. CVR is typically assessed by inducing global vasodilation through techniques such as inhaling carbon dioxide or breath-holding and then measuring changes in the BOLD signal. Since both CVR and task-induced NVC depend on the vasomotor function of blood vessels, CVR assessment provides critical insight into the degree to which observed BOLD signal changes during tasks are due to variations in neural activity versus the vascular functional state, thus offering valuable information for a more accurate interpretation of BOLD-fMRI results.

Transcranial Doppler ultrasonography (TCD) is a non-invasive, portable diagnostic tool that uses a low-frequency ultrasound probe (usually $\le$2 MHz), placed on specific acoustic windows of the skull (such as the temporal, occipital, or ocular windows), to emit ultrasound waves and receive Doppler-shifted signals reflected from red blood cells flowing through intracranial vessels. Based on the Doppler effect, TCD allows for real-time, continuous measurement of blood flow velocity in the intracranial large arteries, indirectly reflecting changes in CBF [64]. Functional TCD is the main method for assessing NVC, by recording dynamic changes in the blood flow velocity of target arteries during tasks or stimuli (e.g., the maximum percentage increase in blood flow velocity after stimulation or changes in specific frequency components observed in spectral analysis) [65]. TCD’s primary advantages include its non-invasive nature, real-time monitoring, portability and low cost, making it suitable for bedside use with a high temporal resolution during detection of rapid hemodynamic changes [66]. However, its accuracy depends on the operator’s experience and about 10–15% of individuals may not achieve satisfactory signals due to suboptimal acoustic windows. Additionally, TCD primarily measures the blood flow velocity in the main arteries at the skull base, which indirectly reflects the status of downstream microcirculation and lacks precise spatial localization. Furthermore, blood flow velocity is influenced by various systemic physiological changes [67].

Despite these limitations, TCD is widely used in both clinical and research settings to assess NVC function, cerebral vascular reactivity, and cerebral autoregulation in healthy individuals and patients with cerebrovascular diseases [68, 69]. For example, functional TCD studies have shown hemodynamic abnormalities and impaired NVC in the posterior cerebral artery in migraine patients during both the aura and attack phases [70]. TCD reflects the integrative response of upstream large arteries to downstream microcirculation NVC, making it valuable for assessing overall cerebrovascular function. To explore the mechanisms of NVC at the microvascular level, however, it is necessary to combine TCD with other techniques capable of directly detecting microcirculation or neural activity [71].

Near-infrared spectroscopy (NIRS) is a non-invasive optical imaging technique that uses near-infrared light at specific wavelengths to penetrate biological tissues, exploiting the differing absorption properties of oxygenated and deoxygenated hemoglobin. This allows for the estimation of relative changes in the concentrations of these two hemoglobin types in local cortical tissue, indirectly reflecting NVC [72]. The primary advantage of NIRS in assessing NVC is its non-invasive nature, portability and lower cost compared to larger devices like fMRI, as well as its greater tolerance to motion artifacts, making it ideal for use in natural environments or with specific populations (such as infants, children, the elderly, and patients with movement disorders) [60]. NIRS has high temporal resolution, theoretically down to the millisecond level, though practical resolution is limited by the inherent slow hemodynamic response (in the seconds range) [73]. Additionally, functional NIRS can be synchronized with other techniques, enabling multimodal NVC research [74].

Despite its limitations, such as lower spatial resolution (typically at the centimeter level), limited depth penetration (mainly reflecting superficial cortical regions) and potential interference from blood flow changes in extra-cranial tissues [62], NIRS offers significant advantages in terms of portability, cost-effectiveness and suitability for specific populations [73]. It has been widely applied in NVC research related to cognitive functions, motor control, emotional regulation and language and its use in studying NVC in various neurological conditions such as stroke, neurodegenerative diseases, psychiatric disorders and pediatric developmental disorders has been increasing [10, 75]. For example, one study using synchronous NIRS and EEG recordings has found that in AD patients, the coupling between neural activity in certain brainwave frequencies and corresponding cortical blood oxygen signals is significantly weaker than in healthy controls, indicating that impaired NVC may be a key factor in the pathophysiological process of AD [75]. This capability to assess cortical NVC in settings where fMRI is not feasible constitutes NIRS’s core application value.

The NVC process is inherently complex, involving various components such as neuronal activity, glial cell signaling and vascular responses. Due to this complexity, no single technique fully captures the scope of NVC. Multimodal evaluation strategies combine two or more complementary methods, such as electrophysiology with hemodynamics, or macro with micro techniques, to collect data from different levels, either synchronously or in a close sequential manner. The approach enables a more comprehensive and accurate understanding of the underlying mechanisms of NVC and its dynamic alterations in both healthy and diseased brain states [76].

EEG records the cumulative synaptic potentials of numerous neurons through electrodes placed on the scalp. It offers direct insights into cortical neural activity and provides excellent temporal resolution (in the millisecond range) to capture fast dynamic changes in neural function. However, its spatial resolution is relatively low, making it difficult to precisely pinpoint the subcortical origins of the signals [77]. To overcome the limitations of single-modality approaches, combining EEG with hemodynamic techniques that offer higher spatial resolution, such as fMRI, TCD, or NIRS, has emerged as an ideal strategy for NVC research [76].

For instance, the EEG-fMRI integration leverages EEG’s high temporal resolution and fMRI’s high spatial resolution, enabling the study of the region-specific and dynamic properties of NVC by analyzing the correlation between specific EEG events and subsequent BOLD signal changes. This combination has been widely used to investigate NVC in conditions such as epilepsy, sleep disorders, stroke, and AD [78]. The EEG-TCD technique allows for simultaneous monitoring of neural activity and blood flow dynamics. During cognitive tasks, neuronal activity typically precedes blood flow changes, with a slight delay. By combining the high temporal resolution of EEG and TCD, this delay can be accurately measured, providing insight into how blood flow responds to neural activity [71]. Finally, EEG-NIRS, owing to its portability, low cost, and insensitivity to motion, is especially useful in NVC assessments across diverse populations, such as infants, elderly, or patients with mobility limitations, and natural settings, such as bedside monitoring or during rehabilitation [79].

Alterations in cerebral hemodynamics, including reduced resting CBF and impaired functional hyperemia, are core pathological features of AD. These changes often precede clinical symptoms and correlate with the severity of dementia, potentially accelerating disease progression [9]. According to the “vascular hypothesis” of AD, vascular risk factors and dysfunction of the NVU serve as initial triggers, leading to chronic brain hypoxia. This, in turn, decreases the clearance of $\beta$-amyloid (A$\beta$) and increases its production, ultimately exacerbating neurodegeneration [85]. A$\beta$ pathology impairs NVC through various mechanisms, including inhibiting tissue plasminogen activator (tPA) activity via upregulation of plasminogen activator inhibitor-1. This disruption of the tPA-NO signaling pathway weakens N-methyl-D-aspartate (NMDA) receptor-dependent vasodilation [86]. Additionally, A$\beta$ induces vascular oxidative stress, which is toxic to endothelial cells and leads to the deposition of amyloid in blood vessel walls, forming cerebral amyloid angiopathy and irreversibly impairing vascular function [87]. Tau protein pathology also contributes by interfering with the coupling of NMDA receptors and nNOS, reducing NO production, and impairing functional hyperemia. These effects occur before neuronal degeneration and neurofibrillary tangles form. Tau may also inhibit the phosphorylation of nNOS and eNOS [88]. Importantly, both A$\beta$ and Tau independently and synergistically disrupt NVC through interference with NO-mediated vasodilation, suggesting that the NO signaling pathway is a common pathological convergence point.

Furthermore, dysfunction of various NVU components plays an important role in AD pathology. Pericyte degeneration, shedding and impaired metabolic sensing and vasodilation are early events in AD, leading to BBB breakdown and reduced CBF [9]. Loss of pericyte-derived laminin and abnormal capillary pericyte contraction also contribute to insufficient brain perfusion [23]. Endothelial dysfunction is marked by oxidative stress, decreased NO bioavailability, increased BBB permeability and loss of cholinergic innervation, resulting in weakened eNOS-dependent NVC [89]. Astrocytes in AD models exhibit abnormal calcium signaling, particularly in their end-feet and their decoupling from vascular reactivity weakens NVC. This persists even when vascular elasticity remains intact. Dysfunction of their aquaporin-4 may exacerbate A$\beta$ accumulation [90]. This multifaceted NVU damage collectively leads to chronic cerebral hypoperfusion and hypoxia in AD patients, further worsening neuronal dysfunction and cognitive decline. Hypoxic conditions also stimulate A$\beta$ production and inhibit its clearance, creating a damaging cycle.

In ischemic stroke, acute ischemia leads to severe NVC impairment, particularly in the ischemic penumbra, where residual neuronal activity cannot adequately trigger CBF increases, thus exacerbating energy mismatch and tissue damage [80]. Although reperfusion therapy is essential, it also results in reperfusion injury due to incomplete microvascular reperfusion (no-reflow phenomenon) and ongoing NVC dysfunction [91]. Even when large vessels are reperfused, high capillary transit time heterogeneity may lead to “reperfusion failure”, which results in low oxygen extraction efficiency. In some cases, sudden restoration of blood flow may trigger “hyperperfusion syndrome”, accompanied by tissue hypoxia. NVC failure accelerates the transition from the ischemic penumbra to the infarct core [92]. Under ischemic conditions, astrocytes fail to efficiently uptake glutamate, exacerbating excitotoxicity. Other NVU cells, such as oligodendrocytes, endothelial cells, and pericytes, also express glutamate receptors and are susceptible to damage [80]. Ischemia triggers a robust inflammatory response, including activation of microglia and astrocytes, infiltration of peripheral immune cells, and the release of pro-inflammatory cytokines and damage-associated molecules. This leads to BBB breakdown, further exacerbating NVU damage and NVC dysfunction [93].

Post-stroke recovery involves neurogenesis and angiogenesis, two closely linked processes that are critical for functional improvement. Activated endothelial cells and neural stem cells promote each other through growth factor secretion. Neuroinflammation plays a complex role during recovery, being harmful during the acute phase but beneficial in later stages, with certain inflammatory mediators (e.g., short-term interleukin-6 (IL-6)) promoting repair [58]. Recovery of NVC is crucial for the support of the metabolic needs of regenerating tissues and new synapses. Persistent NVC dysfunction outside the infarct zone may result in long-term neurological deficits [94]. Thus, stroke treatment should not only focus on reperfusing large vessels but also aim to improve microvascular dysfunction and restore NVC, ensuring effective tissue reperfusion and supporting recovery.

CSVD refers to a heterogeneous group of disorders affecting small arteries, arterioles, capillaries and small veins in the brain [95]. Common causes include hypertensive vascular disease and cerebral amyloid angiopathy. The primary pathological change is arteriosclerosis, characterized by thickening and hardening of vascular walls and lipid-laden degeneration. These structural changes underlie NVC dysfunction and lead to endothelial dysfunction, impaired vascular reactivity, reduced CBF and disruption of the BBB [96, 97]. Endothelial damage is an early event in CSVD, resulting in an imbalance of vasoactive substances and a reduction in NO bioavailability, directly affecting NVC. Thakore et al. [67] reported that in a CSVD mouse model, depletion of phosphatidylinositol 4,5-bisphosphate in capillary endothelial cells led to the loss of activity in inward-rectifying potassium channel Kir2.1, impairing the dilation signal from capillaries to arterioles and causing functional cerebral hyperperfusion. Furthermore, carotid hemodynamic parameters, such as reduced wall shear stress and increased pulsatility index, were associated with impaired NVC in CSVD patients, suggesting that hemodynamic changes in large arteries directly affect NVC function in brain microvasculature [98]. This “large vessel to microvessel” pathological influence highlights the complexity of CSVD pathophysiology.

A clinical study showed that NVC function declines with the severity of CSVD imaging biomarkers, such as white matter hyperintensities and cerebral microbleeds [70]. Impaired NVC results in chronic cerebral underperfusion, particularly in the deep white matter regions. This underperfusion, together with BBB disruption, exacerbates white matter damage and is closely associated with cognitive decline [99]. A study using resting-state fMRI with local consistency and CBF ratio analysis has revealed early NVC abnormalities in patients with subcortical ischemic vascular disease that correlate with the degree of cognitive impairment [100]. NVC assessment is considered a potential biomarker for monitoring CSVD severity and progression. Current treatments focus primarily on controlling vascular risk factors, especially hypertension, but there is a great need for specific therapies targeting NVU dysfunction [73].

NVC dysfunction in migraine with aura is primarily associated with cortical spreading depression (CSD). CSD is a large-scale, slowly propagating depolarization wave involving neurons and glial cells, followed by neural activity suppression, and is considered the neurophysiological basis of migraine aura [62]. CSD induces significant changes in CBF: an initial brief local increase in blood flow (hyperemia), followed by a prolonged and gradually spreading decrease in blood flow (hypoperfusion). This biphasic blood flow response represents a transient and severe disturbance in NVC, with the hypoperfusion phase lasting for several hours and associated with NVC impairment [81]. During CSD, all components of the NVU participate in the dynamic changes in NVC. Pericytes in capillaries play an active regulatory role, with prolonged vasoconstriction after CSD being associated with sustained increases in calcium concentration in pericytes. After CSD, sensory stimuli no longer induce normal capillary dilation or calcium responses in pericytes, suggesting that pericyte dysfunction contributes to NVC impairment following CSD [101]. Astrocytes also exhibit intense calcium wave activity during CSD and pharmacologically inhibiting astrocyte function blocks CSD-induced meningeal afferent nerve sensitization, indicating that astrocytes may contribute to harmful information processing downstream of CSD via calcium-independent signaling pathways [78]. Calcitonin gene-related peptide (CGRP) release increases during CSD, possibly contributing to the initiation and propagation of CSD and its vascular effects. CGRP receptor antagonists inhibit the occurrence of CSD in vitro [102]. Additionally, abnormalities in KATP channel function are also linked to NVC dysfunction in migraine. Global regulation of KATP channel activity significantly interferes with NVC. Mice with Cantú syndrome show diminished baseline NVC and increased sensitivity to KATP channel openers, suggesting that KATP channel dysfunction may play a key role in the pathogenesis of migraine with aura [103].

Epilepsy is characterized by abnormal synchronous high-frequency discharges of neurons, with NVC dysfunction playing a crucial role in its pathophysiology. Both early-onset and late-onset idiopathic epilepsy patients show lower global NVC coefficients and reduced CBF/“Amplitude of Low-Frequency Fluctuation” (ALFF) ratios in specific brain regions, such as the precentral and postcentral gyri, when compared to healthy controls [104]. During seizures, epileptic foci typically exhibit marked increases in local CBF and metabolic rates, with fMRI revealing heightened BOLD signals at the seizure sites. However, one study has reported transient vasoconstriction or more complex biphasic perfusion patterns at the onset of seizures [105]. During the interictal period, abnormal discharges, such as spikes, also affect NVC. One to five seconds prior to a seizure, vasoconstriction may be observed in the cortical region surrounding an epileptic focus, potentially due to blood shunting in response to imminent high metabolic activity or a reduction in local metabolism [106]. Post-seizure, localized brain tissue often experiences severe and persistent hypoxia, accompanied by reduced CBF and sustained microarterial constriction, which is closely linked to fluctuations in Ca²⁺ levels in vascular smooth muscle cells and astrocytes. Astrocytic Ca²⁺ signaling increases sharply during seizures, possibly triggering both vasodilation and vasoconstriction, leading to complex vascular responses [14].

Components of the NVU contribute to epilepsy-associated NVC dysfunction, with neuronal hyperexcitability being the primary initiating factor. Endothelial cells and the integrity of the BBB may be compromised during epileptic seizures, exacerbating inflammation and NVU dysfunction [107]. The ET-1 signaling pathway plays a significant role, with endothelin receptor type A (ETA) receptor activation causing local ischemia and excitotoxic damage, while endothelin receptor type B (ETB) receptor activation drives non-ischemic seizures through inflammatory pathways. NVC dysfunction leads to energy failure, hypoxia, and neuronal damage in the epileptic focus and surrounding regions, promoting the propagation of epileptic discharges and resulting in neurological deficits and long-term secondary brain damage after a seizure [14]. Notably, persistent post-seizure vasoconstriction and severe local hypoxia represent critical periods for secondary brain damage and may significantly contribute to cognitive and neurological dysfunction associated with epilepsy [14].

Diabetes, particularly Type 2 diabetes, results in metabolic disturbances, such as chronic hyperglycemia, insulin resistance, dyslipidemia, and low-grade chronic inflammation, leading to cerebrovascular pathology and NVC dysfunction, ultimately contributing to cognitive impairment [108]. One study has shown regional reductions in the coupling between neuronal activity and CBF in specific brain areas, such as the hippocampus, fusiform gyrus and amygdala, in diabetic patients, although global gray matter NVC may remain unchanged [109]. The degree of NVC impairment correlates with the severity of cognitive deficits, particularly in executive function, memory, and processing speed. Research also indicates that reduced ALFF-CBF coupling in the left hippocampus and amygdala is associated with poorer executive function in diabetic patients [109].

Several key mechanisms contribute to NVC dysfunction in diabetes. Endothelial dysfunction, induced by hyperglycemia, is a central factor, involving impaired eNOS function, reduced NO production and excessive reactive oxygen species (ROS) generation, which damage endothelial cells, decrease NO clearance and promote inflammation [110]. Advanced glycation end-products (AGEs) accumulate in the brain’s microvasculature, leading to vascular sclerosis and the triggering of oxidative stress and inflammation through interactions with the “Receptor for Advanced Glycation End-products” (RAGE), thereby impairing NVC [108]. Chronic inflammation and insulin resistance further damage endothelial NO production, glucose transport and neuronal function, leading to NVC dysregulation [82]. Diabetes-associated microvascular changes, including basal membrane thickening, pericyte loss/dysfunction, endothelial cell damage, capillary rarefaction and astrocyte activation and dysfunction, all directly impair the vascular response to neuronal activity [82]. Additionally, the dysfunction of specific signaling molecules and ion channels contributes to NVC impairment [111]. These combined mechanisms underscore NVC dysfunction as a key link between diabetes and cognitive decline.

Emerging therapeutic modalities represent cutting-edge strategies for repairing NVC, aiming to utilize advanced biotechnologies for more precise interventions. However, their path toward clinical application is complex, commonly involving significant challenges in safety, efficacy, and targeted delivery. Gene therapy seeks to correct the molecular defects causing NVC dysfunction through the long-term introduction of therapeutic genes, such as those encoding NOS or neuroprotective factors [136]. For instance, targeting cyclin-dependent kinase 5, which is aberrantly activated in AD and disrupts the function of NVU cells, represents a promising gene therapy approach to restore NVU integrity [137]. Despite slow progress in clinical translation, there has been some momentum. The core challenge lies in selecting the ideal viral vector (such as adeno-associated virus, AAV), avoiding immune responses, and ensuring precise and stable gene expression. In the field of Parkinson’s disease, some early clinical trials have shown promise, such as a phase 1b trial delivering adeno-associated virus serotype 2 carrying glial cell line-derived neurotrophic factor to the striatum, which confirmed the safety of the therapy and the effectiveness of targeting, marking a key step in repairing neural circuits [138]. Stem cell-based therapies primarily support neurorepair by leveraging endogenous paracrine effects, such as the release of growth factors and anti-inflammatory molecules, promoting angiogenesis, alleviating inflammation, strengthening the BBB, and enhancing synaptic plasticity, which collectively enhance the NVU microenvironment and restore NVC [139]. Clinically, mesenchymal stem cells (MSCs) have been widely evaluated in clinical trials for diseases such as ischemic stroke and AD due to their immunomodulatory properties. A meta-analysis of stroke indicated that MSC transplantation is safe and may improve functional outcomes. However, there remain core issues to address, such as cell source, survival, functional integration, and risks of immune rejection and tumorigenicity [140]. Nanomedicine, utilizing nanomaterials as targeted delivery systems for drugs or genes, particularly those crossing the BBB, or for direct neuroprotection and the construction of nanofiber scaffolds to support neural regeneration, holds potential to enhance the specificity and efficacy of pharmacological interventions [141]. However, the clinical translation of nanomedicine in neurology is still in its early stages, mainly due to the formidable challenge of efficiently and safely crossing the BBB. Most nanomedicine applications for central nervous system diseases remain at the preclinical stage. However, some concepts are approaching clinical reality, such as the development of nanoparticle-based systems for more effective delivery of antioxidants, anti-inflammatory agents, or thrombolytics to ischemic brain tissue following stroke [142]. Optogenetics, a research tool that precisely modulates the activity of specific neurons using genetic engineering, has advanced our understanding of NVC mechanisms. In stroke animal models, it has shown potential to enhance NVC responses and promote functional recovery by regulating nNOS expression, providing a foundation for the development of new therapies [143]. However, its clinical translation remains distant. The primary barrier is that this technology requires invasive gene delivery via viral vectors (for implanting light-sensitive proteins) and surgical implantation of fiber-optic devices for light delivery. These procedures raise significant safety concerns regarding immunogenicity, potential neurotoxicity, and long-term stability of the equipment, making it more distant for clinical application compared to gene and cell therapies [144].