EndMT in vascular cognitive impairment and dementia: mechanisms, evidence gaps, and therapeutic opportunities

Abstract

Vascular cognitive impairment and dementia (VCID) is increasingly recognized as a consequence of chronic cerebrovascular dysfunction, including cerebral small vessel disease and clinically defined entities such as post-stroke cognitive impairment (PSCI). Clinical and experimental studies indicate that endothelial injury and maladaptive microvascular remodeling can precede cognitive decline through cerebral blood flow dysregulation, impaired neurovascular coupling, and blood–brain barrier (BBB) disruption, with downstream neuroinflammatory amplification. A recurring feature across these settings is redox imbalance: reactive oxygen species (ROS) derived from mitochondria, NADPH oxidases, and uncoupled endothelial nitric oxide synthase act not only as mediators of oxidative injury but also as compartmentalized signaling cues that remodel endothelial transcriptional and epigenetic states. Here, I review how redox-sensitive modules—including TGF-β/Smad, NF-κB, HIF-1α, Wnt/β-catenin, and Notch signaling, together with chromatin regulation and non-coding RNAs—may initiate or stabilize endothelial-to-mesenchymal transition (EndMT) and partial EndMT-like programs during chronic vascular stress. I then discuss how EndMT-like reprogramming could contribute to BBB fragility, extracellular matrix remodeling, vascular stiffening, hypoperfusion, and neurovascular dysfunction, while highlighting priorities for in vivo causal testing and potential reversibility of partial EndMT. Finally, I map key mechanistic nodes to potential therapeutic classes to inform precision strategies targeting putative pathogenic redox–EndMT circuits rather than global oxidative stress.

Introduction

Vascular cognitive impairment and dementia (VCID) is a comprehensive umbrella term encompassing a broad spectrum of cognitive disorders attributable to cerebrovascular pathologies, ranging from mild cognitive impairment to frank vascular dementia (Lennon et al. 2025). This broad categorization explicitly includes Post-stroke Cognitive Impairment (PSCI), which manifests as acute or delayed cognitive decline following a stroke event (Pasi et al. 2012), as well as cognitive deficits driven by cerebral Small Vessel Disease (SVD), a chronic condition characterized by lacunar infarcts, white matter hyperintensities, and microbleeds (Wardlaw et al. 2013). By integrating these diverse etiologies, the VCID framework highlights the critical contribution of both macrovascular and microvascular damage to cognitive decline, independent of Alzheimer’s disease pathology (Iadecola et al. 2019; Bhend et al. 2026). Despite the growing societal burden of VCID, disease-modifying therapies that directly target the underlying cerebrovascular biology remain limited, in part because mechanistic bridges between vascular stressors and durable endothelial remodeling are incompletely established. A convergent body of clinical and experimental work implicates microvascular dysfunction as an early and sustained driver of VCID phenotypes, including impaired cerebral blood flow regulation, neurovascular uncoupling, and blood–brain barrier (BBB) disruption, with downstream amplification of neuroinflammation and neuronal network vulnerability (Kalaria 2016; Chen et al. 2022; Beaufils et al. 2026). Oxidative stress and redox imbalance are repeatedly observed in cerebrovascular disease and VCID-relevant models; importantly, reactive oxygen species (ROS) are not only mediators of oxidative injury but also spatially and temporally regulated signals that can reprogram endothelial transcriptional and epigenetic states (Li et al. 2025; Hase et al. 2024; Vance et al. 2025). Source-specific ROS circuits—arising from mitochondria, nicotinamide adenine dinucleotide phosphate (NADPH) oxidases, and uncoupled endothelial nitric oxide synthase (eNOS)—interact with inflammatory and hypoxic signaling to shape endothelial fate decisions, raising the possibility that redox-regulated endothelial plasticity contributes to sustained microvascular failure in VCID/PSCI (Hartman et al. 2025; Staehlke et al. 2020; Yan et al. 2014; Liu et al. 2022). Emerging evidences indicate that endothelial cells respond to chronic stress through functional impairment and phenotypic plasticity, involving structural and transcriptional reprogramming (Andueza et al. 2020; Queiroz et al. 2025).

A key form of this plasticity is endothelial-to-mesenchymal transition (EndMT), characterized by the loss of endothelial markers such as vascular endothelial (VE)-cadherin and platelet endothelial cell adhesion molecule-1 (CD31) and gain of mesenchymal traits including α-smooth muscle actin (α-SMA) and vimentin (Piera-Velazquez and Jimenez 2019). Initially identified in cardiac development and fibrosis, EndMT is now implicated in various chronic inflammatory and vascular disorders (Gorelova et al. 2021; Gaikwad et al. 2020). In the cerebrovascular system, single-cell transcriptomics and lineage-tracing studies show endothelial heterogeneity and stress-induced phenotypic shifts within brain microvessels, indicating that endothelial plasticity actively contributes to vascular remodeling and fibrosis (Jeong et al. 2022; Crouch et al. 2024; Brandt et al. 2019; Song et al. 2020). EndMT is particularly harmful in the cerebral microvasculature, disrupting barrier integrity, modifying extracellular matrix (ECM) composition, and promoting pro-inflammatory and pro-fibrotic signaling (Vance et al. 2025; Jiang et al. 2025; Park et al. 2025).

However, direct causal evidence linking EndMT to VCID/PSCI phenotypes in humans remains limited, and much of the current support is associative (e.g., marker-based signatures) or derived from selected in vivo models and in vitro mechanistic studies (Cahill-Smith and Li 2014; Kasal et al. 2025). Therefore, the mechanistic relationships discussed in this review should be interpreted as working hypotheses rather than established causal pathways in human cerebrovascular disease. Accordingly, this review (i) synthesizes redox-sensitive signaling circuits that can initiate or stabilize EndMT programs (including NOX/mitochondrial ROS, uncoupled eNOS/NO–cGMP, TGF-β/Smad, NF-κB, HIF-1α, and chromatin regulation), (ii) explicitly distinguishes associative evidence from causal demonstrations using an evidence hierarchy and minimum criteria for EndMT assignment in human cerebrovascular tissue, and (iii) maps mechanistic nodes to actionable therapeutic classes while highlighting priorities for in vivo causal testing and potential reversibility of partial EndMT in VCID/PSCI.

Endothelial-to-mesenchymal transition

Definition and molecular hallmarks of endothelial-to-mesenchymal transition

EndMT is a complex cellular reprogramming process in which endothelial cells gradually lose their endothelial identity and acquire mesenchymal and fibroblast-like traits. This transition was first described in embryonic heart development, where EndMT shapes cardiac cushion formation and valve morphogenesis, and is now recognized as a pathological mechanism in adult tissues (Hall et al. 2025; Islam et al. 2021). At the molecular level, EndMT involves the coordinated downregulation of endothelial-specific markers and junctional proteins, along with the upregulation of mesenchymal markers and cytoskeletal remodeling. Canonical endothelial markers, including VE-cadherin, platelet endothelial CD31, eNOS, and von Willebrand factor, are consistently downregulated during EndMT, reflecting loss of endothelial polarity and barrier integrity (Piera-Velazquez and Jimenez 2019; Giordanengo et al. 2025; Cho et al. 2018). VE-cadherin suppression is particularly critical, as its loss destabilizes adherence junctions and enables transcriptional reprogramming toward a mesenchymal state (Giannotta et al. 2013; Potter et al. 2005). Concomitantly, EndMT involves the upregulation of mesenchymal markers, including α-SMA, vimentin, fibroblast-specific protein-1 (FSP-1/S100A4), fibronectin, and type I collagen. These markers indicate increased contractility, migration, and ECM production, collectively promoting vascular stiffening and fibrosis (Evrard et al. 2016; Zhang et al. 2016; Rossato et al. 2020). EndMT represents a continuum rather than a binary switch, with endothelial cells often retaining partial endothelial traits while expressing mesenchymal genes—a phenomenon termed partial or hybrid EndMT (Jiang et al. 2025; Li et al. 2025). Experimental lineage-tracing studies using endothelial-specific Cre recombinase systems provide definitive evidence that endothelial cells generate mesenchymal cells in vivo during fibrosis and vascular remodeling. These studies establish EndMT as a true cellular fate conversion, rather than an artifact of marker overlap or cell contamination (Helmke et al. 2019; Chen et al. 2016). In human disease, immunohistochemical analyses reveal cells that co-express endothelial and mesenchymal markers within fibrotic and vascular lesions, supporting the clinical relevance of EndMT (Mendoza et al. 2016; Martin et al. 2021; Basha and Hsi 2019). Together with molecular and lineage-tracing studies, these findings highlight EndMT as a dynamic, context-dependent endothelial fate transition with distinct molecular hallmarks, rather than a fixed phenotypic endpoint. This conceptual framework provides a foundation for understanding how pathological cues—redox imbalance, inflammation, and mechanical stress—activate EndMT programs in vascular disease.

Canonical signaling pathways driving endothelial-to-mesenchymal transition

Transforming growth factor-beta (TGF-β) signaling, the most extensively characterized pathway in EndMT, is regarded as the master regulator of endothelial plasticity. TGF-β drives EndMT primarily by activating Smad-dependent transcriptional programs, repressing endothelial genes while inducing mesenchymal transcription factors such as Snail, Slug, and Twist (Piera-Velazquez and Jimenez 2019; Ma et al. 2020; Zhang et al. 2024; Medici et al. 2011). Genetic or pharmacological blockade of TGF-β signaling effectively suppresses EndMT across multiple experimental models, highlighting its central role (Zhang et al. 2018; Ma et al. 2021). Beyond canonical Smad signaling, non-Smad TGF-β pathways—including p38 MAPK, ERK, and PI3K/Akt—synergize to reinforce EndMT, particularly during inflammation or oxidative stress. These pathways regulate cytoskeletal dynamics, cell survival, and transcriptional stability of mesenchymal phenotypes (He et al. 2019; Zhang et al. 2006; Fujimoto et al. 2023). The integration of Smad and non-Smad signaling allows EndMT to adapt dynamically to environmental cues. The Wnt/β-catenin pathway is a key regulator of EndMT, functioning independently and synergistically with TGF-β signaling. Wnt activation stabilizes β-catenin, driving its nuclear translocation and mesenchymal gene transcription. In endothelial cells, β-catenin signaling disrupts adherens junctions and accelerates EndMT, particularly in fibrotic and vascular disease models (DiRenzo et al. 2016). Crosstalk between Wnt and TGF-β pathways amplifies EndMT by targeting shared transcriptional regulators, including Snail and ZEB family proteins (Medici et al. 2011; Medici et al. 2008). Notch signaling is crucial for EndMT, particularly in developmental and vascular contexts. Notch receptor activation induces mesenchymal gene expression and synergizes with TGF-β signaling to sustain EndMT. Endothelial-specific Notch activation induces EndMT-like phenotypes in vivo, while Notch inhibition reduces fibrosis and vascular remodeling (Timmerman et al. 2004; Tang et al. 2010). Notch signaling also integrates mechanical forces, hypoxia, and inflammatory cues, highlighting its relevance in chronic vascular disease.

Physiological vs. pathological endothelial-to-mesenchymal transition

Physiological EndMT is an essential developmental process that occurs in a tightly regulated time- and site-specific manner. During embryogenesis, EndMT allows endothelial cells to generate the mesenchymal cells needed for heart valve formation and septation, after which endothelial identity is restored or stabilized (Hall et al. 2024; Gise and Pu 2012). This form of EndMT is reversible, context-dependent, and does not cause pathological tissue remodeling. In contrast, pathological EndMT arises in adult tissues in response to chronic stressors such as inflammation, hypoxia, metabolic dysregulation, and oxidative stress. In these conditions, EndMT becomes persistent and maladaptive, promoting fibrosis, vascular rarefaction, and organ dysfunction (Dahal et al. 2017; Millar et al. 2024). Unlike physiological EndMT, pathological EndMT is typically incomplete but sustained, leaving endothelial cells that continually produce ECM components and inflammatory mediators. Clinical and experimental studies link pathological EndMT to various diseases, including cardiac fibrosis, pulmonary hypertension, atherosclerosis, chronic kidney disease, and cerebrovascular disorders. In these conditions, EndMT-derived cells substantially expand the pool of activated fibroblasts and myofibroblasts, worsening tissue stiffness and vascular dysfunction (Yun et al. 2020; Xu et al. 2020). Lineage-tracing experiments show that a substantial share of fibroblast-like cells in fibrotic lesions originates from endothelial cells, highlighting the quantitative importance of EndMT (Zhang et al. 2013; Cooley 2014). Pathological EndMT is also closely associated with aging and redox imbalance. Aging-associated oxidative stress sensitizes endothelial cells to TGF-β and inflammatory signals, lowering the threshold for EndMT induction. Experimental models show that elevated ROS enhances EndMT, whereas restoring redox homeostasis suppresses mesenchymal transition and preserves endothelial identity (Sobierajska et al. 2022; Bellanti et al. 2025). These findings highlight EndMT as a redox-sensitive process linking chronic vascular stress to irreversible structural remodeling.

Cerebrovascular endothelium and vascular cognitive impairment and dementia (VCID)

Structural and functional alterations of cerebral microvessels in VCID

VCID is fundamentally characterized by chronic cerebral microvascular changes that impair cerebral perfusion and metabolic support of neuronal networks. Neuropathological and neuroimaging studies consistently show that microvascular rarefaction, arteriolosclerosis, and capillary degeneration prominently affect white matter regions essential for executive function and processing speed in patients with VCID (Kalaria 2016; Brown et al. 2009; Yang et al. 2017). Structural remodeling of cerebral microvessels in VCID involves basement membrane thickening, increased collagen deposition, and pericyte loss, all impairing vessel compliance and microvascular responsiveness. Post-mortem analyses reveal marked alterations in capillary morphology—reduced capillary density and luminal narrowing—correlating with cognitive impairment severity and white matter lesion extent [68: Pantoni, 2003 #109]. These changes reduce oxygen and nutrient delivery, making the brain particularly vulnerable to hypoperfusion-induced injury. In VCID, cerebral endothelial cells show impaired vasoreactivity and disrupted neurovascular coupling. The initiation of this cascade is predominantly driven by luminal (peripheral) stressors, such as altered shear stress from chronic hypoperfusion and systemic circulating inflammatory factors. Experimental models of chronic cerebral hypoperfusion show reduced endothelium-dependent vasodilation, mainly due to luminal nitric oxide (NO) deficiency and oxidative stress-induced eNOS dysfunction (Kolluru et al. 2012; Kim et al. 2018; Sun et al. 2025). This primary luminal endothelial dysfunction precedes neuronal loss, suggesting that vascular pathology initiates rather than follows neurodegeneration. However, as the endothelial barrier is compromised, secondary abluminal (central) signals—such as hypoxia resulting from reduced blood flow and neuroinflammation—emerge and further exacerbate endothelial injury. At the cellular level, endothelial cells are exposed to a biphasic or dual-source stress environment. Initially, luminal stressors like altered shear stress and circulating inflammatory stimuli trigger early endothelial activation. Consequently, reduced capillary perfusion generates chronic hypoxia within the brain parenchyma, which acts as a potent abluminal signal. Together, these sequential luminal and abluminal stimuli force endothelial cells to undergo metabolic and transcriptional changes that impair their homeostatic functions. These maladaptive responses involve increased ROS production, mitochondrial dysfunction, and activation of pro-inflammatory signaling pathways, further worsening microvascular injury (Wang and He 2020; Qu et al. 2022; Minjares et al. 2023; Pandian et al. 2025). Overall, these structural and functional abnormalities highlight endothelial dysfunction as a key driver of chronic microvascular insufficiency in VCID. Hypoperfusion, oxidative stress, and inflammatory signaling create a permissive environment for sustained endothelial injury and phenotypic remodeling, promoting BBB disruption and maladaptive endothelial plasticity.

Blood–brain barrier disruption and endothelial dysfunction in VCID

BBB disruption is a key pathological feature of VCID and a major contributor to cognitive decline. Clinical imaging studies using dynamic contrast-enhanced MRI show increased BBB permeability in patients with VCID and vascular cognitive impairment, particularly in the hippocampus and white matter regions, even without overt infarction (Taheri et al. 2011; Heye et al. 2014; Yu et al. 2025; Nie et al. 2025). Endothelial dysfunction drives BBB breakdown. Loss of tight junction proteins, including claudin-5, occludin, and zonula occludens-1, has been documented in post-mortem VCID brain tissue, indicating compromised endothelial barrier integrity (Sharma et al. 2022; Romanitan et al. 2007). These molecular changes allow plasma proteins, immune cells, and neurotoxic molecules to enter the brain parenchyma, triggering neuroinflammation and synaptic dysfunction. Experimental studies show oxidative stress as a key driver of BBB disruption in VCID. In rodent models of chronic cerebral hypoperfusion, increased ROS production disrupts endothelial tight junctions and cytoskeleton, causing sustained BBB leakage and white matter damage (Schreibelt et al. 2007; Liu et al. 2019; Lochhead et al. 2010). Reducing oxidative stress pharmacologically or genetically preserves BBB integrity and improves cognitive deficits in these models. BBB dysfunction promotes a feed-forward cycle of vascular and neural injury. Fibrinogen and other plasma-derived factors activate microglia and astrocytes, amplifying inflammation and worsening endothelial function (Huang et al. 2023; Takata et al. 2021). This vicious cycle underscores endothelial pathology as a key driver of VCID progression.

Endothelial-to-mesenchymal transition of cerebral vessels in VCID

Direct evidence of EndMT in human cerebral vessels is limited, but accumulating experimental findings suggest that cerebral endothelial cells can adopt mesenchymal-like phenotypic changes under pathological conditions relevant to VCID. In vitro, brain microvascular endothelial cells exposed to hypoxia, inflammatory cytokines, or oxidative stress downregulate endothelial markers and upregulate mesenchymal genes, reflecting partial EndMT (Shoemaker et al. 2020; Iannucci et al. 2022; Derada Troletti et al. 2019; Takada et al. 2017). Animal models support EndMT-like processes in cerebral vasculature. In models of aging and chronic cerebral hypoperfusion, endothelial cells within affected brain regions show increased expression of α-smooth muscle actin, fibronectin, and collagen, with reduced VE-cadherin and CD31 expression (Vance et al. 2025; Edgerton-Fulton and Ergul 2022). Specifically, there is emerging evidence suggesting a mechanistic link between endothelial senescence and EndMT in the cerebral circulation. Senescent brain microvascular endothelial cells adopt a senescence-associated secretory phenotype (SASP), characterized by the robust release of TGF-β, pro-inflammatory cytokines, and ROS. This SASP may alter the local neurovascular microenvironment and could function as a sustained autocrine and paracrine trigger that potentially contributes to partial EndMT programs, thereby exacerbating age-related vascular fragility and BBB disruption. These phenotypic changes are linked to vascular stiffening, reduced perfusion, and cognitive decline. Lineage-tracing studies in non-cerebral vessels show that endothelial cells can become fibroblast-like cells during fibrosis, and similar mechanisms are increasingly proposed to operate within the brain microvasculature (Piera-Velazquez et al. 2016; Lovisa et al. 2020). While definitive lineage-tracing evidence in human VCID is lacking, converging molecular signatures strongly support a role for endothelial plasticity in cerebrovascular remodeling. EndMT in cerebral vessels is likely partial and context-dependent rather than complete. Partial EndMT may suffice to disrupt endothelial barrier function and promote ECM deposition without full fibroblast conversion, thereby contributing to progressive microvascular dysfunction (Li et al. 2025; Derada Troletti et al. 2019). Collectively, these findings suggest that EndMT-like endothelial plasticity may contribute to chronic cerebrovascular remodeling in VCID, even without complete lineage conversion. The partial and stress-dependent nature of cerebral EndMT suggests that local signaling environments—such as redox imbalance and inflammation—critically determine the extent and persistence of this phenotypic shift. In the context of VCID, EndMT should be viewed as a sequentially driven process. It is largely initiated by luminal signals (e.g., altered hemodynamics, circulating systemic inflammation) that cause early endothelial dysfunction and BBB breakdown. This initial injury allows for the accumulation of abluminal stressors (e.g., parenchymal hypoxia, glial-derived cytokines, and potentially amyloid-β), which then act as a secondary, sustained trigger to fully engage and stabilize the EndMT transcriptional program. Furthermore, recent high-resolution transcriptomic studies provide robust evidence for EndMT in cerebrovascular pathologies. Single-cell RNA sequencing (scRNA-seq) of brain endothelial cells from mouse models of stroke and chronic hypoperfusion has identified distinct subpopulations expressing mesenchymal markers such as Col1a1, Fn1, and Serpine1 alongside a loss of classic endothelial identity (Munji et al. 2019; Olayinka et al. 2025; Lebas et al. 2024; Ren et al. 2023).

Comparison with Alzheimer’s disease-associated vascular pathology

Although VCID and AD are traditionally considered as distinct entities, evidence increasingly indicates substantial overlap in their vascular pathologies. Cerebral microvascular dysfunction, BBB disruption, and endothelial injury occur in both conditions, although their relative contributions and timing differ (Reas et al. 2024; Preis et al. 2024). In AD, vascular pathology often involves cerebral amyloid angiopathy, pericyte-mediated capillary constriction, and amyloid-β–induced endothelial toxicity (Nortley et al. 2019). These mechanisms reduce CBF and increase neuronal vulnerability, often secondary to amyloid pathology. In contrast, VCID arises primarily from luminal insults—such as ischemia, hypoperfusion, and luminal oxidative stress—which initiate endothelial dysfunction (Li et al. 2014; Biesbroek et al. 2024). Conversely, in AD, abluminal signals originating from the brain parenchyma, such as hypoxia driven by amyloid-β or tau pathology, often act as the upstream triggers for vascular damage. Recent findings suggest that the distinction between these entities is further blurred by the fact that vascular stressors such as hypoperfusion can induce endothelial Aβ production (Katusic et al. 2024). This suggests that EndMT-like transitions and redox imbalance not only contribute to vascular remodeling but may also actively participate in the amyloidogenic cascade typically associated with AD. Despite these differences in the temporal sequence and spatial origin of the initiating signals (luminal vs. abluminal), both diseases share redox-sensitive pathways of endothelial injury that converge on EndMT. Oxidative stress drives BBB disruption, inflammation, and endothelial phenotypic changes in AD and VCID, suggesting convergent downstream pathways (Kuriakose et al. 2019). EndMT-like processes may underlie chronic vascular remodeling in dementias, although the upstream triggers vary. Comparing VCID and AD highlights dementia as a spectrum of neurovascular, rather than purely neurodegenerative, disorders. Understanding how endothelial plasticity and redox signaling differentially shape vascular pathology in these diseases may reveal shared therapeutic targets while enabling disease-specific interventions (Kalaria 2003).

Redox signaling pathways regulating endothelial-to-mesenchymal transition

Reactive oxygen species as signaling mediators versus oxidative damage

ROS in cerebral endothelial cells are traditionally regarded as harmful metabolic by-products (Wang et al. 2023; Wang et al. 2022). However, emerging evidence indicates that ROS also serve as regulated signaling mediators influencing endothelial phenotype, barrier integrity, and inflammation (Chen et al. 2018). In the cerebral microvasculature, distinct ROS sources, including mitochondria, NADPH oxidases (NOX), and uncoupled eNOS, produce spatially and temporally constrained redox signals (Daiber et al. 2017). Physiological redox signals drive adaptive endothelial responses, whereas chronic vascular stress triggers pathological amplification and sustained endothelial dysfunction (Landmesser et al. 2003). Among these sources, mitochondrial ROS (mtROS) has emerged as the primary regulator of endothelial redox state. While cerebral endothelial cells primarily use glycolysis for ATP generation, mitochondria remain essential for redox buffering and signaling control. In brain microvessels, mitochondrial dysfunction is closely associated with BBB fragility, linking subcellular bioenergetic failure to junctional instability and paracellular leakage (Guo et al. 2018; Dhariwal et al. 2024). Chronic hypoperfusion, inflammation, or metabolic stress triggers endothelial metabolic reprogramming, altering the NADH/NAD⁺ ratio and exhausting antioxidants to sensitize cells to mtROS-driven signaling shifts (Peng et al. 2021; Forrester et al. 2018; Gomes et al. 2013). Experimental studies show that excessive mitochondrial fission elevates mtROS and compromises endothelial barrier integrity, while inhibiting mitochondrial fission preserves tight junctions and BBB function (Haileselassie et al. 2020; Haileselassie et al. 2019) (Table 1). mtROS act as sources of oxidative injury and signaling inputs that reprogram inflammatory and apoptotic pathways into feed-forward loops, driving endothelial dysfunction (Wang et al. 2023; Wang et al. 2022). Simultaneously, NADPH oxidases serve as dedicated ROS-generating systems central to cerebrovascular pathology. Among NOX isoforms, NOX2 and NOX4 predominate within the cerebral endothelium. NOX4-derived ROS drives inflammatory stress and apoptosis in brain endothelial cells, linking this isoform to BBB breakdown and autonomous injury (Basuroy et al. 2009; Basuroy et al. 2011). NOX2 is strongly linked to neurovascular dysfunction and cognitive impairment across multiple disease models; its genetic inactivation prevents oxidative stress, restores cerebrovascular function, and improves behavioral outcomes (Park et al. 2008; Hernandes et al. 2014) (Table 1). These findings establish NOX-derived ROS as drivers of endothelial activation, microvascular inflammation, and white matter vulnerability, rather than mere epiphenomena of disease progression.

Table 1.

Redox signaling mechanisms regulating endothelial-to-mesenchymal transition (EndMT) in cerebral endothelial cells

Redox component/pathway	Molecular redox mechanism	Functional role in EndMT regulation	Study type	References
Mitochondrial ROS (mtROS)	Metabolic reprogramming, mitochondrial fission–dependent oxidant signaling	Initiation of endothelial plasticity, BBB fragility	Animal and in vitro study	Guo et al. 2018; Dhariwal et al. 2024; Haileselassie et al. 2020; Haileselassie et al. 2019)
NADPH oxidases (NOX2/NOX4)	Source-specific ROS production; NOX4–TGF-β feed-forward signaling	Stabilization of Smad signaling and phenotypic switching	Animal and in vitro study	Basuroy et al. 2009; Basuroy et al. 2011; Park et al. 2008; Hernandes et al. 2014; Jiang et al. 2014; Boudreau et al. 2012; Hiraga et al. 2013)
Uncoupled eNOS	BH4 oxidation → superoxide generation instead of NO	Loss of NO–cGMP signaling, pro-EndMT bias	Animal and in vitro study	Chen et al. 2011; Santhanam et al. 2012; Thum et al. 2007; Santhanam et al. 2015; Yang et al. 2009)
ROS crosstalk networks	ROS-induced ROS release between mitochondria, NOX, eNOS	Amplification of endothelial redox signaling loops	Animal and in vitro study	Kim et al. 2017; Zinkevich and Gutterman 2011; Dikalov et al. 2014; Scialo et al. 2017; Ding et al. 2007)
Latent TGF- β activation	Oxidant-mediated ECM-dependent activation	Upstream induction of EndMT programs	Animal and in vitro study	Krstic et al. 2015; Ghafouri-Fard et al. 2024)
Smad3 redox modification	Cysteine sulfenylation (Cys-64) regulating DNA binding	Redox switch controlling transcriptional output	Animal and in vitro study	Huang et al. 2021; Samarakoon et al. 2013)
NO–cGMP antagonism	NO-dependent suppression of Smad transactivation	Anti-EndMT buffering axis	Animal and in vitro study	Saura et al. 2005)
Nrf2 signaling	Antioxidant response element activation; redox buffering	Inhibition of inflammatory and fibrotic EndMT signaling	Animal and in vitro study	Kim et al. 2010; Wang et al. 2019; Ba et al. 2024)
Notch ligand–receptor bias	DLL4–NOTCH1 vs JAG1–NOTCH4 under oxidative stress	Context-dependent endothelial stabilization vs EndMT promotion	Animal and in vitro study	Ali and Yun 2025; Caliceti et al. 2014)
NF-κB signaling	Redox-regulated inflammatory transcription	Amplification of cytokine-driven EndMT	Animal and in vitro study	Lingappan 2018; Mihira et al. 2012; Yoshimatsu et al. 2020; Lee and Kay 2012; Fang et al. 2024)
HIF-1α signaling	Hypoxia–ROS synergy driving Snail/Twist transcription	Hypoxia-induced endothelial reprogramming	Animal and in vitro study	Rey and Semenza 2010; Cannito et al. 2008; Xu et al. 2015; Zhang et al. 2018; Yang and Wu 2008; Yang et al. 2017)
Chromatin remodeling	Redox-sensitive histone modification, HDAC activity	Stabilization of mesenchymal transcriptional memory	Animal and in vitro study	Shen et al. 2025; Hulshoff et al. 2018; Niu et al. 2015; Gu et al. 2013; Rahman 2003; Lewandowski et al. 2015; Lecce et al. 2021; Jyotirmaya et al. 2025)
Protein sulfenylation	Reversible cysteine oxidation regulating signaling kinetics	Fine-tuning of EndMT signaling strength	Animal and in vitro study	Huang et al. 2021; Hanschmann et al. 2013; Beedle et al. 2016)
Non-coding RNAs	circRNA/miRNA control of SLUG and chromatin regulators	Persistence of partial EndMT states	Animal and in vitro study	Li et al. 2025; Saaoud et al. 2021)

NO imbalance further illustrates the transition from redox signaling to pathology. Under physiological conditions, endothelial NO maintains vascular tone, neurovascular coupling, and anti-inflammatory homeostasis. Oxidative stress, however, induces eNOS uncoupling—shifting eNOS from NO production to superoxide generation—thereby converting a vasoprotective factor into a potent ROS source (Chen et al. 2011; Santhanam et al. 2012; Thum et al. 2007) (Table 1). In cerebral microvessels, oxidation of the essential eNOS cofactor tetrahydrobiopterin (BH4) by peroxynitrite (ONOO⁻) drives uncoupling, impairs NO–cGMP signaling, and amplifies oxidative stress (Santhanam et al. 2015; Yang et al. 2009). Restoring BH4 bioavailability rescues endothelial NO signaling and mitigates microvascular injury, confirming eNOS uncoupling as a mechanistically relevant and potentially reversible driver of cerebrovascular dysfunction (Forstermann and Sessa 2012; Santhanam et al. 2014). eNOS uncoupling triggers a dual-hit phenotype characterized by superoxide overproduction and impaired vasodilation, thereby accelerating vascular failure. Cerebral endothelial redox biology is defined by extensive crosstalk between ROS sources. Mitochondrial ROS, NOX enzymes, and uncoupled eNOS engage in interconnected feed-forward loops through ROS-induced ROS release, escalating localized redox perturbations into sustained endothelial dysfunction (Kim et al. 2017; Zinkevich and Gutterman 2011) (Table 1). Mitochondrial ROS activate cytosolic NOX2, while NOX-derived radicals further stimulate mitochondrial ROS, reinforcing oxidant signaling beyond the initial stimulus (Dikalov et al. 2014; Scialo et al. 2017). Concurrent NO depletion and eNOS uncoupling drive sustained superoxide production, intensifying oxidative amplification and vascular dysregulation (Ding et al. 2007). This network-level propagation explains why chronic hypoperfusion or inflammation produces persistent BBB disruption and microvascular failure rather than transient endothelial dysfunction.

Collectively, these findings indicate that ROS in cerebral endothelial cells operate along a continuum—from regulated signaling mediators to drivers of irreversible oxidative damage. The pathological effect of ROS depends less on their overall abundance than on their source specificity, subcellular localization, and integration within redox signaling networks. This distinction has important translational implications because broad antioxidant strategies have consistently failed in neurovascular disease. Instead, effective treatment demands targeting primary ROS-generating nodes or redox-sensitive signaling pathways before multi-source amplification occurs. This conceptual framework is particularly relevant to EndMT, which is increasingly viewed through a “redox rheostat” model. In this view, moderate and compartmentalized oxidant fluxes trigger endothelial plasticity, whereas excessive or sustained oxidant production overwhelms buffering systems, locking endothelial cells into maladaptive, mesenchymal-like phenotypes (Jiang et al. 2017; Panieri and Santoro 2015). Consequently, because “ROS” is an overgeneralized term, mechanistic studies must specify the molecular species, cellular source, and targets, as distinct oxidants vary in chemistry, diffusion ranges, and signaling capacity (Brand 2010; Brandes et al. 2018). In endothelial biology, low-level, and spatially restricted oxidant signaling—particularly mediated by hydrogen peroxide—reversibly modulates protein activity via cysteine oxidation, thereby tuning kinase–phosphatase balance, receptor signaling, and transcriptional programs without causing injury (Matsui et al. 2020). This redox signaling is defined by its reversibility, selectivity, and restoration via endogenous antioxidant systems including peroxiredoxins, thioredoxin, and glutathione-dependent pathways (Ren et al. 2017). In contrast, oxidative stress involves irreversible damage to lipids, proteins, and DNA; this often coincides with mitochondrial failure and antioxidant depletion, stabilizing pro-inflammatory and pro-fibrotic endothelial states (Pagano et al. 2014).

This distinction remains central to understanding EndMT. While signaling-strength oxidants initiate EndMT, the process becomes pathological if oxidant production overwhelms buffering capacity or chronologically remodels chromatin and mesenchymal gene expression (Kreuz and Fischle 2016). Multiple EndMT-inducing stimuli, including inflammatory cytokines and TGF-β, elevate oxidant tone, creating feed-forward loops between redox imbalance and mesenchymal transcriptional networks (Thuan et al. 2018). Consequently, oxidative stress can act upstream of canonical EndMT ligands by inducing endogenous TGF-β signaling and driving endothelial conversion toward myofibroblast-like phenotypes (Montorfano et al. 2014).

Overall, these findings shift the focus from whether oxidants are inherently “beneficial” or “harmful” to identifying which oxidant species, source, and locations drive EndMT in cerebral endothelial cells. This redox-centered perspective provides a mechanistic bridge between endothelial dysfunction, chronic vascular remodeling, and cognitive decline, highlighting the need for precision redox interventions tailored to endothelial plasticity rather than global oxidative stress suppression.

Redox regulation of transforming growth factor -β signaling

As the dominant canonical driver of EndMT, TGF-β signaling is strongly modulated by redox-dependent processes acting upstream and downstream of receptor activation (Zhou et al. 2021) (Fig. 1). Extensive evidence indicates that NADPH oxidase-derived oxidants—particularly NOX4-dependent signaling—serve as downstream effectors of TGF-β in fibrosis, while also regulating the TGF-β/Smad axis via feed-forward loops (Jiang et al. 2014) (Fig. 1). This bidirectionality provides a mechanism for sustained EndMT in chronic inflammatory or hypoxic vascular environments, where TGF-β induction and oxidant generation mutually reinforce each other (Cooley 2014). At the level of endothelial phenotype decisions, NOX4 mediates TGF-β-induced redox signaling to drive endothelial apoptosis and phenotypic switching, establishing NOX4-derived oxidants as active determinants of endothelial fate rather than passive byproducts (Yan et al. 2014). In mechanistic studies, NOX4-derived oxidants —predominantly hydrogen peroxide (H₂O₂)—regulate TGF-β-induced endothelial migration and fibronectin expression, linking TGF-β–NOX4 coupling directly to EndMT-related ECM remodeling (Boudreau et al. 2012) (Table 1). These findings suggest that TGF-β “wires in” a NOX4-dependent pro-oxidant module (to stabilize pro-fibrotic gene expression and promote mesenchymal features (Boudreau et al. 2012; Hiraga et al. 2013). Redox regulation also controls TGF-β bioavailability, thereby modulating upstream activation of the pathway (Krstic et al. 2015). By activating latent extracellular TGF-β, oxidants increase receptor engagement and amplify downstream Smad signaling, creating a redox-sensitive feedback loop (Ghafouri-Fard et al. 2024). In cerebral endothelial cells, oxidative stress induces the secretion of both TGF-β1 and TGF-β2, driving myofibroblast differentiation and establishing an oxidant-TGF-β-EndMT axis (Montorfano et al. 2014) (Fig. 1). While TGF-β1 is the predominant pro-fibrotic cytokine typically released during neuroinflammation and BBB disruption, TGF-β2 is also highly sensitive to oxidative stress and actively contributes to EndMT via autocrine signaling in the cerebral microvasculature. NO signaling provides a further redox layer that antagonizes or reshapes TGF-β output (Park et al. 2021). In endothelial cells, NO inhibits TGF-β/Smad-mediated transactivation via a cGMP-dependent mechanism, indicating that EndMT-associated pro-fibrotic transcription is restrained by NO bioavailability and redox balance (Saura et al. 2005) (Fig. 1). Thus, redox imbalance that impairs NO signaling—via oxidative scavenging or eNOS uncoupling—indirectly biases endothelial cells toward TGF-β/Smad-driven EndMT (Saura et al. 2005) (Fig. 1). Finally, direct redox modifications further refine TGF-β signaling by post-translationally tuning pathway output. Chemoproteomic and in vivo studies show that Smad3 sulfenylation at Cys-64 inhibits its DNA binding activity; conversely, mutating this site attenuates protection against Angiotensin II–induced vascular remodeling, establishing a specific redox control point within Smad transcriptional machinery (Huang et al. 2021). These findings indicate that EndMT-related TGF-β signaling is controlled not only by phosphorylation cascades but also by reversible cysteine oxidation, which determines transcriptional competence (Samarakoon et al. 2013) (Table 1). Collectively, these findings characterize TGF-β signaling as a redox-integrated pathway in which oxidant generation, NO balance, and cysteine-based redox modifications determine the magnitude and persistence of EndMT. This redox sensitivity explains how chronic vascular stress converts transient TGF-β activation into sustained endothelial phenotypic reprogramming. Accordingly, redox-dependent tuning of the TGF-β/Smad axis represents a critical control point for modulating EndMT-driven cerebrovascular remodeling (Fig. 1).

Fig. 1.

Redox regulation of TGF-β/Smad signaling during endothelial-to-mesenchymal transition (EndMT). Detailed schematic illustrating how redox signaling modulates transforming growth factor-β (TGF-β)/Smad–dependent transcriptional programs during endothelial-to-mesenchymal transition (EndMT). ① Oxidative stress generated by reactive oxygen species (ROS), including hydrogen peroxide (H₂O₂) and superoxide (O₂⁻), promotes activation of latent TGF-β through extracellular matrix (ECM) remodeling and enhances receptor signaling via oxidation of redox-sensitive cysteine residues within receptor-associated signaling components. ② NADPH oxidase 4 (NOX4)–derived ROS amplifies TGF-β receptor (TGFBR2/ALK5) signaling, leading to phosphorylation and activation of Smad2/3 and formation of the Smad2/3–Smad4 transcriptional complex. In addition to phosphorylation-dependent activation, cysteine oxidation of Smad3 (e.g., Cys64-SOH) modulates nuclear trafficking and transcriptional selectivity, thereby influencing EndMT gene expression. ③ Crosstalk with mitochondrial ROS (mtROS) production and nitric oxide (NO)–cyclic guanosine monophosphate/protein kinase G (cGMP/PKG) signaling provides additional layers of redox control that can restrain or redirect profibrotic transcriptional outputs. ④ Epigenetic mechanisms, including histone deacetylase–dependent chromatin remodeling and ubiquitination pathways, stabilize transcriptional reprogramming and promote expression of EndMT-associated genes such as ACTA2/α-SMA, FN1, COL1A1, and VIM. ⑤ The Keap1–Nrf2 antioxidant feedback loop counterbalances excessive ROS by inducing antioxidant responses, thereby limiting maladaptive EndMT progression. Together, these redox-sensitive signaling events integrate extracellular stress cues with intracellular transcriptional and epigenetic regulation to drive endothelial plasticity, extracellular matrix remodeling, migration, fibrosis, and partial EndMT phenotypes. EndMT, endothelial-to-mesenchymal transition; TGF-β, transforming growth factor-beta; TGFBR2, transforming growth factor-beta receptor type 2; ALK5 (TGFBR1), activin receptor-like kinase 5; ROS, reactive oxygen species; H₂O₂, hydrogen peroxide; O₂⁻, superoxide anion; NOX4, NADPH oxidase 4; Smad, small mothers against decapentaplegic proteins; ECM, extracellular matrix; mtROS, mitochondrial reactive oxygen species; NO, nitric oxide; cGMP, cyclic guanosine monophosphate; PKG, protein kinase G; Keap1, Kelch-like ECH-associated protein 1; Nrf2, nuclear factor erythroid 2–related factor 2; HDAC, histone deacetylase; ACTA2/α-SMA, alpha-smooth muscle actin; FN1, fibronectin 1; COL1A1, collagen type I alpha 1 chain; VIM, vimentin

Redox-sensitive transcription factors

Transcription factors are critical nodes that translate redox signals into sustained changes in endothelial gene expression and cellular phenotype. In EndMT, redox-sensitive transcriptional regulators integrate oxidative, inflammatory, and hypoxic cues to determine whether endothelial cells maintain homeostasis or transition toward a maladaptive phenotypic (Fig. 2).

Fig. 2.

Redox-sensitive transcription factor networks and epigenetic control of the EndMT gene program. Integrated schematic illustrating how oxidative stress, inflammatory cytokines, and hypoxia converge on redox-sensitive transcription factor networks to regulate endothelial-to-mesenchymal transition (EndMT). Upstream vascular stressors, including aging and hypertension, promote initiator signals such as inflammation and hypoxia, which are ① amplified by mitochondrial reactive oxygen species (mito-ROS) and NADPH oxidase (NOX)-derived ROS. These redox cues activate interconnected transcriptional pathways, including ② the transforming growth factor-β (TGF-β)/Smad pathway, and ③ the nuclear factor kappa-B (NF-κB) and hypoxia-inducible factor-1α (HIF-1α) pathways, with extensive crosstalk between signaling modules. Reactive oxygen species (ROS) regulate cysteine oxidation–dependent signaling events that modulate Smad2/3 activation and downstream transcriptional responses. Concurrently, ④ nitric oxide (NO)–cyclic guanosine monophosphate/protein kinase G (cGMP/PKG) signaling provides an opposing regulatory axis, while epigenetic remodeling processes such as histone modification and ubiquitination stabilize EndMT-associated transcriptional programs. The coordinated action of these pathways drives expression of mesenchymal genes (ACTA2/α-SMA, FN1, COL1A1, VIM), promoting extracellular matrix (ECM) remodeling, migration, fibrosis, and partial EndMT phenotypes. The hierarchical framework depicted at the top highlights the progression from upstream risk factors to initiators, amplifiers, EndMT activation, and downstream outcomes, including vascular dysfunction and cognitive decline. EndMT, endothelial-to-mesenchymal transition; ROS, reactive oxygen species; mito-ROS, mitochondrial reactive oxygen species; NOX, NADPH oxidase; TGF-β, transforming growth factor-beta; Smad, small mothers against decapentaplegic proteins; NF-κB, nuclear factor kappa-B; HIF-1α, hypoxia-inducible factor-1 alpha; NO, nitric oxide; cGMP, cyclic guanosine monophosphate; PKG, protein kinase G; ECM, extracellular matrix; ACTA2/α-SMA, alpha-smooth muscle actin; FN1, fibronectin 1; COL1A1, collagen type I alpha 1 chain; VIM, vimentin

Nuclear factor erythroid 2–related factor 2 (Nrf2), the primary regulator of antioxidant and cytoprotective gene expression, increasingly emerges as a key inhibitor of maladaptive endothelial phenotypic transitions under oxidative stress (Kim et al. 2010) (Fig. 2). Since sustained oxidant and inflammatory signaling drive EndMT, Nrf2 counteracts this transition by restoring glutathione- and thioredoxin-dependent redox buffering, limiting oxidant amplification, and suppressing pro-fibrotic signaling (Wang et al. 2019). Direct experimental evidence confirms that Nrf2-related pathways exert anti-mesenchymal effects during EndMT. Despite tissue differences from cerebral endothelium, the mechanistic logic remains transferable: Nrf2 activation reduces the oxidative and inflammatory stressors that otherwise potentiate EndMT (Ba et al. 2024). Notch signaling exerts context-dependent and redox-sensitive control over EndMT, with distinct functional outcomes depending on receptor–ligand pairing. Under oxidative stress conditions, reactive oxygen species (ROS) can modulate Notch receptor activation by influencing ligand–receptor binding affinity, ADAM metalloprotease cleavage, and γ-secretase–mediated release of the Notch intracellular domain (NICD) (Ali and Yun 2025). DLL4–NOTCH1 signaling is generally associated with arterial identity and pro-inflammatory endothelial activation; increased ROS levels enhance NF-κB and HIF-1α activity, which can synergize with NOTCH1 signaling to promote Snail/Twist transcription and facilitate EndMT-like transcriptional reprogramming. In contrast, JAGGED1–NOTCH4 signaling often exerts a stabilizing effect on endothelial quiescence and barrier integrity, partly through crosstalk with Nrf2-dependent antioxidant responses and suppression of excessive TGF-β/Smad signaling. Redox-dependent cysteine modifications within Notch receptors and regulatory proteins may further bias signaling output by altering NICD stability and nuclear transcriptional complex formation (Panieri and Santoro 2015). Thus, ROS-driven shifts in ligand availability and receptor context can tilt the balance between DLL4–NOTCH1–mediated pro-EndMT signaling and JAGGED1–NOTCH4–associated protective or homeostatic pathways, highlighting the importance of redox state in determining the directionality and magnitude of Notch-dependent endothelial plasticity (Caliceti et al. 2014) (Table 1).

NF-κB is a central inflammatory transcription factor that integrates cytokine signaling with redox cues (Lingappan 2018); its activation consistently drives EndMT within inflammatory microenvironments (Cho et al. 2018). Oxidant signaling activates NF-κB by regulating redox-sensitive upstream kinases and IκB stability, thereby coupling ROS production to pro-inflammatory transcription that cooperates with TGF-β signaling to drive EndMT (Mihira et al. 2012) (Fig. 2). Experimental studies show that NF-κB drives EndMT phenotypes under pathological stress (Yoshimatsu et al. 2020). In corneal endothelial cells, IL-1β–driven EndMT requires NF-κB-dependent transcriptional regulation, establishing a cytokine → NF-κB → mesenchymal transition axis (Lee and Kay 2012). In a rat model of chronic doxorubicin-induced cardiotoxicity, NF-κB pathway activation correlates with EndMT-linked fibrosis, identifying the pathway as a therapeutic target and a key driver of EndMT (Xu et al. 2020). Studies on ischemic disease report NF-κB activation as a regulator of partial EndMT, emphasizing also its role in tuning intermediate transition states relevant to chronic vascular remodeling (Fang et al. 2024). As a primary transcriptional regulator of hypoxic adaptation, hypoxia-inducible factor 1α (HIF-1α) is central to EndMT, given that hypoxia frequently drives vascular disease and hypoperfusion-related brain injury (Rey and Semenza 2010). Importantly, recent evidence indicates that hypoxia and chronic hypoperfusion can promote the production of amyloid-beta (Aβ) peptides directly within endothelial cells, potentially through the HIF-1α-mediated upregulation of β-site amyloid precursor protein cleaving enzyme 1 (BACE1) (Katusic et al. 2024). This endothelial-derived Aβ may function as a potent driver of further vascular injury and neurodegeneration, establishing a critical mechanistic bridge between vascular stress and Alzheimer’s-like pathology.

Hypoxia increases oxidant production via mitochondrial and NOX-dependent pathways; this redox shift synergizes with HIF signaling to drive mesenchymal gene expression and endothelial plasticity (Cannito et al. 2008) (Fig. 2). A mechanistic study in human coronary endothelial cells shows that hypoxia induces EndMT by positioning Snail as a direct transcriptional target of HIF-1α, thereby linking HIF-1α to a master mesenchymal regulator (Xu et al. 2015) (Fig. 2). A study on pulmonary arterial remodeling shows that hypoxia-induced EndMT depends on the HIF-1α/Twist1 axis, indicating that HIF-1α engages multiple mesenchymal transcriptional drivers across different vascular beds (Zhang et al. 2018). Collectively, these findings support a model where hypoxia and redox imbalance converge on HIF-1α to initiate or stabilize EndMT through the activation of Snail/Twist-centered transcriptional networks (Yang and Wu 2008; Yang et al. 2017) (Table 1). Together, these transcriptional regulators indicate how redox imbalance converges on gene regulatory networks to promote or restrain EndMT, regulated by signal strength, duration, and cellular context. Elucidating the interaction and hierarch of redox-sensitive transcription factors may reveal leverage points to selectively disrupt pathological EndMT while preserving adaptive stress responses.

Epigenetic and post-translational redox regulation

In EndMT biology, converting transient redox signaling into stable mesenchymal phenotypes likely requires epigenetic regulation to establish durable transcriptional memory (Shen et al. 2025). EndMT is epigenetically regulated by DNA methylation, histone acetylation/methylation remodeling, and noncoding RNAs, which collectively tune the accessibility of endothelial and mesenchymal gene programs (Hulshoff et al. 2018). Oxidative stress directly modifies epigenetic marks (Niu et al. 2015). In neuronal-like cells, oxidative stress triggers DNA demethylation and histone acetylation to alter gene transcription, suggesting that redox perturbations can reprogram chromatin states in vascular cells under chronic stress (Gu et al. 2013). In immune-relevant cell models, H₂O₂ exposure increases histone acetylation and alters inflammatory gene expression, supporting the principle that oxidants reshape chromatin landscapes during phenotype switching (Rahman 2003). Similarly, EndMT is regulated by disease-relevant chromatin modifiers (Lewandowski et al. 2015). In atherosclerosis, histone deacetylase 9 promotes EndMT and unstable plaque phenotype, establishing a causal link between this epigenetic enzyme and EndMT-driven vascular pathology (Lecce et al. 2021) (Fig. 2). These findings integrate redox signaling with EndMT by showing how redox imbalance modulates the activity and expression of epigenetic enzymes, including HDAC family members and sirtuins, to drive stable transcriptional remodeling (Jyotirmaya et al. 2025). Post-translational redox modifications complement chromatin marks by providing rapid and reversible regulation of EndMT signaling networks (Hanschmann et al. 2013). Protein sulfenylation is a key modification that modulates protein function as a reversible intermediate, thereby integrating redox state with signaling kinetics through thiol redox systems (Beedle et al. 2016) (Table 1). Smad3 sulfenylation at cysteine-64 modulates its DNA binding and subsequent vascular remodeling outcomes in vivo, serving as a direct redox “switch” within the TGF-β transcriptional apparatus (Huang et al. 2021). Redox status also indirectly regulates TGF-β/Smad signaling by modulating the enzymes and stress pathways that control ubiquitin-dependent protein turnover (Zhang et al. 2014). MicroRNAs and circular RNAs provide a further link between redox signaling and EndMT stability by modulating key drivers and chromatin regulators (Saaoud et al. 2021) (Fig. 2, Table 1). In ischemic disease, the circular RNA regulates partial EndMT by controlling SLUG mRNA methylation dynamics, indicating how RNA modifications serve as durable molecular memory for endothelial transitions (Li et al. 2025). Collectively, these epigenetic and post-translational mechanisms suggest that treating redox-driven EndMT requires reducing oxidant burden and blocking the conversion of redox signals into stable chromatin and protein-state changes that perpetuate the mesenchymal phenotype.

Endothelial-to-mesenchymal transition driven pathological consequences in VCID

Fibrosis and vascular stiffening

Fibrosis and vascular stiffening are key drivers of chronic cerebrovascular dysfunction in VCID. Post-mortem and imaging-based studies consistently show small cerebral vessel remodeling in VCID and vascular cognitive impairment, including excessive ECM deposition, basement membrane thickening, and reduced vessel compliance (Reeve et al. 2024; Iulita et al. 2018). These changes impair microvascular adaptability and promote persistent cerebral hypoperfusion, particularly in white matter, which is highly sensitive to subtle reductions in perfusion pressure (Chen et al. 2022). EndMT provides a mechanistic framework link between endothelial injury and fibrotic remodeling of the cerebral microvasculature. During EndMT, endothelial cells adopt fibroblast- and myofibroblast-like traits, including increased production of collagen, fibronectin, and other ECM components that directly stiffen vessels (Cho et al. 2018; Piera-Velazquez et al. 2011). Lineage-tracing studies in experimental models of organ fibrosis show that many activated fibroblasts originate from endothelial cells, establishing EndMT as a significant and meaningful contributor to fibrotic remodeling rather than a rare or incidental occurrence (Zeisberg et al. 2003; Aisagbonhi et al. 2011). Endothelial-derived mesenchymal cells in these models sustain ECM production and contractile gene expression, highlighting their role in tissue stiffening (Alonso-Herranz et al. 2020; Souilhol et al. 2018) (Table 2). Although definitive lineage-tracing evidence in human cerebral vessels is lacking, molecular and histopathological evidence strongly support EndMT-like processes in cerebral small vessel disease. Patients with vascular cognitive impairment exhibit increased mesenchymal markers, fibrillar collagen accumulation, and loss of endothelial junctional proteins in cerebral microvessels (Park et al. 2025; Edgerton-Fulton and Ergul 2022). These molecular signatures mirror findings from experimental models of chronic cerebral hypoperfusion and aging, where endothelial cells adopt partial mesenchymal phenotypes linked to increased vascular stiffness, reduced NO-mediated vasodilation, and impaired autoregulation (Han and Kim 2023; Wu et al. 2025) (Table 2). Experimental studies further show that vascular stiffening has significant functional effects beyond luminal narrowing. Aging is characterized by maladaptive structural remodeling of cerebral arteries, often manifesting as an increase in luminal diameter (ectasia) accompanied by arterial stiffening and wall thinning. This compensatory enlargement fails to maintain hemodynamic efficiency and instead exacerbates pulsatile stress on the microvasculature, further driving EndMT and neurovascular dysfunction in VCID. Increased stiffness limits the ability of small vessels to buffer pulsatile flow and meet metabolic demands, thereby amplifying shear stress heterogeneity and promoting endothelial injury (Cooper et al. 2016; Yates et al. 2025). Importantly, evidence suggests that the cerebral microvascular endothelium is intrinsically more vulnerable to EndMT than the endothelium of large cerebral arteries. Unlike macrovessels, which possess a robust medial layer of smooth muscle cells to buffer mechanical and hemodynamic forces, microvessels rely on sparse pericyte coverage. This leaves microvascular endothelial cells directly exposed to heightened mechanical stretch and localized oxidative stress. Furthermore, the close proximity of the microvasculature to perivascular glia exposes them more directly to localized inflammatory cytokines and SASP factors, lowering the threshold for TGF-β- and redox-induced phenotypic switching. In rodent models, increased cerebral arterial stiffness precedes white matter damage and cognitive decline, supporting a causal link between mechanical remodeling and downstream neurovascular dysfunction (Han et al. 2021). These molecular and structural changes closely mirror findings from experimental models of chronic cerebral hypoperfusion and aging, where endothelial cells adopt partial mesenchymal phenotypes linked to increased vascular stiffness, impaired NO-dependent vasodilation, and reduced autoregulation (Han and Kim 2023). Beyond increased ECM abundance, EndMT-driven remodeling involves qualitative changes in matrix composition and cross-linking, further enhancing vascular rigidity and resistance to deformation (Lin and Davis 2023; Shimshoni et al. 2021). Vascular stiffening has functional effects that extend beyond simple luminal narrowing. Increased vessel rigidity limits the ability of small cerebral vessels to buffer pulsatile flow and meet metabolic demands, thereby amplifying microvascular stress and endothelial injury, particularly in white matter regions, which are highly sensitive to perfusion instability (Cooper et al. 2016; Tsao et al. 2013; Urbano et al. 2017) (Fig. 3). While direct in vivo evidence in the cerebral microvasculature is currently limited, studies from peripheral vascular beds suggest that even partial EndMT can alter ECM organization and mechanical properties of the vessel wall (Islam et al. 2021; Dahal et al. 2017). Based on these peripheral models, we hypothesize that a similar mechanism may operate in VCID: stiffened vascular matrices could enhance endothelial mechanotransduction signaling and activate the TGF-β pathway, potentially creating a feed-forward loop where EndMT is reinforced by its own mechanical effects (Cooley 2014; Song et al. 2019) (Fig. 3, Table 2). However, establishing this mechanical feedback cycle directly within intact brain microvessels remains an important gap for future research.

Table 2.

Clinical and preclinical evidence linking EndMT-associated vascular remodeling to VCID

Pathological domain	Evidence type	Key findings	Current limitations/gaps	Study type	Representative references
Fibrosis and vascular stiffening	Histopathology & Imaging (Human)	Excessive ECM deposition, basement membrane thickening, and reduced vessel compliance	Human lineage tracing unavailable	Human study	Reeve et al. 2024; Iulita et al. 2018)
Impaired CBF & neurovascular coupling	Clinical imaging	Reduced resting CBF and impaired cerebrovascular reactivity (CVR)	Human causality limited	Human study	Iadecola et al. 2023)
White matter vulnerability	Neuroimaging correlations	Microvascular rarefaction and stiffening drive hypoperfusion in white matter	EndMT specificity unresolved	Human study	Chen et al. 2022; Prins et al. 2005; Wang et al. 2018)
BBB disruption & neuroinflammation	Human tissue & MRI	Tight junction loss (claudin-5, occludin) and increased permeability in hippocampus	Mostly marker-based inference	Human study	Taheri et al. 2011; Heye et al. 2014; Yu et al. 2025; Nie et al. 2025; Sharma et al. 2022; Romanitan et al. 2007)
Fibrosis and vascular stiffening	Experimental models (Aging/Hypoperfusion)	EndMT-derived cells contribute to matrix stiffening and wall thinning (ectasia)	Translation to human VCID needs validation	Animal and in vitro	Zeisberg et al. 2003; Aisagbonhi et al. 2011; Alonso-Herranz et al. 2020; Souilhol et al. 2018; Han and Kim 2023; Wu et al. 2025)
Mechanotransduction feed-forward loop	Vascular remodeling studies	Substrate stiffness activates TGF-β and mechanosignaling to reinforce EndMT	Quantitative contribution in vivo unclear	Animal and in vitro	Islam et al. 2021; Dahal et al. 2017; 2014; Song et al. 2019)
Impaired CBF & neurovascular coupling	Hypoperfusion models (Rodents)	Loss of K+ sensing (Kir2.1) and retrograde hyperpolarization signals	Species differences in hemodynamics	Animal and in vitro	Csipo et al. 2021; Lourenco and Laranjinha 2021; Tarantini et al. 2021; Longden et al. 2017; Ali et al. 2022)
BBB disruption & neuroinflammation	Rodent models (Hypoperfusion)	ROS-driven junctional instability; plasma factors (fibrinogen) activate glia	Marker overlap and partial EndMT characterization	Animal and in vitro	Schreibelt et al. 2007; Liu et al. 2019; Lochhead et al. 2010; Huang et al. 2023; Takata et al. 2021)
Endothelium–glia feedback loop	Preclinical neuroinflammation models	Bidirectional crosstalk: glial cytokines (SASP) stabilize mesenchymal programs	Temporal hierarchy unclear	Animal and in vitro	Singh 2022; Burkhart et al. 2024)
Cognitive decline & neuronal vulnerability	Chronic hypoperfusion models	Endothelial phenotypic changes precede synaptic loss and memory deficits	Direct EndMT quantification limited	Animal and in vitro	Mekala and Qiu 2025; Yang et al. 2022; Smith et al. 2018; Guse et al. 2025; Maddaluno et al. 2013)
Oxidative neurotoxicity	Experimental ROS studies	ROS/Cytokines from EndMT cells impair LTP and dendritic spine density	Translational specificity uncertain	Animal and in vitro	Owens et al. 2022; Shavit-Stein et al. 2021; Aliev et al. 2003)

Fig. 3.

Redox signaling–driven endothelial-to-mesenchymal transition (EndMT) links vascular dysfunction to VCID–related neurovascular injury. Conceptual overview illustrating how redox imbalance promotes endothelial-to-mesenchymal transition (EndMT) and contributes to neurovascular dysfunction associated with vascular cognitive impairment and dementia (VCID). Upstream oxidative stress arising from mitochondrial reactive oxygen species (mtROS), NADPH oxidase (NOX) activity, and uncoupled endothelial nitric oxide synthase (eNOS) enhances reactive oxygen species (ROS) signaling, which activates interconnected amplification pathways including transforming growth factor-β (TGF-β)/Smad, nuclear factor kappa-B (NF-κB), Wnt/β-catenin, and hypoxia-inducible factor-1α (HIF-1α). These transcriptional programs drive endothelial phenotypic reprogramming characterized by reduced endothelial markers and increased mesenchymal features (e.g., α-smooth muscle actin [α-SMA], vimentin, and extracellular matrix [ECM] components), accompanied by heightened pro-inflammatory signaling. EndMT-mediated vascular remodeling promotes ECM deposition, vascular stiffening, and chronic hypoperfusion, while blood–brain barrier (BBB) disruption facilitates neuroinflammation, glial activation, immune responses, and neuronal injury. A feed-forward “vicious cycle” between oxidative stress and inflammatory signaling further stabilizes partial EndMT states, leading to impaired neurovascular coupling and progressive cognitive decline. The schematic highlights the hierarchical relationship between upstream redox triggers, transcriptional amplification nodes, vascular remodeling processes, and downstream neurovascular consequences in VCID. EndMT, endothelial-to-mesenchymal transition; VCID, vascular cognitive impairment and dementia; ROS, reactive oxygen species; mtROS, mitochondrial reactive oxygen species; NOX, NADPH oxidase; eNOS, endothelial nitric oxide synthase; TGF-β, transforming growth factor-beta; Smad, small mothers against decapentaplegic proteins; NF-κB, nuclear factor kappa-B; HIF-1α, hypoxia-inducible factor-1 alpha; ECM, extracellular matrix; α-SMA (ACTA2), alpha-smooth muscle actin; BBB, blood–brain barrier

Impaired cerebral blood flow and neurovascular coupling

Precise CBF regulation is essential for maintaining neuronal metabolism and synaptic activity, and its disruption is a hallmark of VCID. Clinical imaging studies consistently show reduced resting CBF and impaired cerebrovascular reactivity in patients with VCID, even without large territorial infarcts, indicating primary dysfunction at the level of the microvasculature rather than secondary effects of neuronal loss (Iadecola et al. 2023) (Table 2). These abnormalities implicate neurovascular unit failure, with endothelial cells as key integrators of metabolic demand, hemodynamic signals, and vasoactive responses (Fig. 3). EndMT provides a mechanistic link between endothelial injury and disrupted neurovascular coupling. During partial mesenchymal reprogramming, endothelial cells show reduced NO production, increased actomyosin contractility, and decreased sensitivity to shear stress and metabolic cues, collectively impairing neuronal activity-dependent vasodilation (functional hyperemia) (Csipo et al. 2021; Lourenco and Laranjinha 2021). Experimental models of chronic cerebral hypoperfusion show that endothelial phenotypic changes precede functional hyperemia deficits, supporting a causal role for endothelial dysfunction rather than neuronal inactivity in the development of impaired neurovascular coupling (Tarantini et al. 2021). Furthermore, the capillary endothelium acts as a critical primary sensor for neurovascular coupling by detecting local increases in extracellular potassium (K⁺) released during neuronal activity (Longden et al. 2017). This sensing, mediated by inward-rectifier K⁺ (Kir2.1) channels, initiates a retrograde hyperpolarization signal that travels to upstream arterioles to trigger vasodilation (Longden et al. 2017). I propose that EndMT-associated remodeling and the accompanying redox imbalance may impair the expression or function of these K⁺ sensors, potentially contributing to a decoupling of neuronal metabolic demand from localized blood flow delivery. These findings are further supported by in vivo imaging studies, which report that endothelial dysfunction alone can uncouple neuronal activity from local blood flow (Ali et al. 2022) (Table 2).

Beyond altered endothelial signaling, EndMT-driven structural remodeling adds mechanical constraints to blood flow regulation. Increased ECM deposition, basement membrane thickening, and reduced endothelial elasticity limit the capacity of arterioles and capillaries to dynamically dilate in response to neuronal activity (Yates et al. 2025; Hansen et al. 2024). Vascular stiffening also hinders vasodilatory signal transmission along the microvascular network, further disrupting coordinated blood flow redistribution (Durante et al. 2024; Wang et al. 2021). These mechanical constraints are especially harmful in white matter regions, where long penetrating arterioles supply metabolically vulnerable axons with limited collateral support, making them highly sensitive to minor flow impairments (Schilling et al. 2025; Pantoni et al. 1996) (Table 2). Redox imbalance further worsens EndMT-related cerebral perfusion deficits. Oxidative stress lowers NO bioavailability through direct scavenging and triggers eNOS uncoupling, turning endothelial cells from vascular tone regulators into sources of superoxide and peroxynitrite (Penna and Pagliaro 2025; Munzel and Daiber 2023). This shift suppresses vasodilatory signaling and reinforces endothelial dysfunction and mesenchymal transition, creating a self-sustaining cycle of impaired blood flow regulation (Andueza et al. 2020). In experimental models, restoring redox balance or maintaining eNOS coupling improves neurovascular responses and partially reverses cognitive deficits, highlighting the functional role of redox–endothelial interactions (Katusic et al. 2023; Gao et al. 2022).

Together, EndMT-driven endothelial reprogramming, microvasculature fibrotic stiffening, and redox-mediated suppression of vasodilatory signaling converge to cause sustained deficits in cerebral perfusion and neurovascular coupling (Fig. 3). These mechanisms explain the progressive, region-specific hypoperfusion characteristic of VCID and highlight endothelial plasticity as a key driver of neurovascular failure.

Neuroinflammation and glial activation

Neuroinflammation is a key pathological feature of VCID closely linked to cerebrovascular dysfunction. Endothelial injury and BBB disruption allow plasma proteins, immune mediators, and danger-associated molecular patterns to enter the brain parenchyma, thereby triggering activation of resident microglia and astrocytes (Takata et al. 2021). This glial response drives synaptic dysfunction, white matter injury, and progressive cognitive decline, even without large ischemic lesions (Gertje et al. 2023). EndMT amplifies neuroinflammatory signaling by transforming endothelial cells from barrier-forming regulators into active producers of inflammatory mediators. Mesenchymal-like endothelial cells show elevated expression and secretion of pro-inflammatory cytokines, chemokines, and ECM components that drive leukocyte recruitment and glial activation (Cho et al. 2018). EndMT-associated BBB dysfunction further amplifies glial activation. Loss of endothelial junctions allows fibrinogen, albumin, and other plasma-derived factors to enter the brain, directly activating microglia and astrocytes via receptor-mediated and innate immune mechanisms (Takata et al. 2021). Fibrinogen deposition specifically induces microglial clustering, oxidative stress, and synaptic pruning, linking vascular leakage to neurodegeneration (Merlini et al. 2019; Davalos et al. 2012). Astrocytes exposed to plasma proteins become reactive, altering metabolic support and glutamate homeostasis, further sensitizing neural circuits to injury (Mahmoud et al. 2019).

Activated glial cells, in turn, worsen endothelial dysfunction and EndMT progression. Cytokines, ROS, and proteases from microglia and astrocyte damage endothelial cells, destabilize junctional complexes, and promote redox imbalance, thereby reinforcing mesenchymal transition programs (Singh 2022; Burkhart et al. 2024). This bidirectional crosstalk creates a feed-forward loop where EndMT drives neuroinflammation, which in turn sustains endothelial plasticity and BBB disruption. These findings suggest that EndMT is not merely a downstream consequence of inflammation but an active driver of neuroimmune dysregulation in VCID (Fig. 3).

Collectively, EndMT-driven endothelial reprogramming links vascular injury with glial activation and inflammatory amplification, positioning endothelial plasticity as a central nexus connecting cerebrovascular dysfunction to chronic neuroinflammation and cognitive decline in VaD.

Cognitive decline and neuronal vulnerability

Ultimately, EndMT drives neuronal dysfunction and cognitive decline. Clinically, VCID manifests as deficits in executive function, attention, and processing speed—cognitive domains especially sensitive to white matter integrity and microvascular health (Prins et al. 2005; Wang et al. 2018). Neurons depend on precise vascular support for metabolism, ion balance, and synaptic activity, and chronic endothelial dysfunction disrupts this support long before noticeable neuronal loss (Nelson et al. 2016). EndMT-driven vascular remodeling disrupts neuronal homeostasis through multiple, converging mechanisms. Progressive CBF limits oxygen and glucose delivery, impairing neuronal energy metabolism and increasing susceptibility to excitotoxic and ischemic injury. Concurrently, BBB disruption exposes neurons to circulating neurotoxic factors, plasma-derived proteins, and inflammatory mediators that are normally excluded from the brain parenchyma (Sun et al. 2022). Experimental evidence links cerebrovascular pathology and cognitive impairment. In animal models of chronic cerebral hypoperfusion, endothelial dysfunction and microvascular remodeling precede synaptic changes and cognitive deficits, suggesting vascular pathology initiates rather than follows neuronal degeneration (Mekala and Qiu 2025; Yang et al. 2022). Interventions that preserve endothelial function or restore microvascular integrity reduce vascular pathology and cognitive decline in these models, underscoring the critical role of endothelial health in cognition (Smith et al. 2018; Guse et al. 2025). Although EndMT is rarely quantified in such studies, the detection of mesenchymal markers and loss of endothelial identity in cerebral endothelial cells strongly implicates EndMT in the observed vascular and cognitive deficits (Maddaluno et al. 2013) (Table 2). Neuronal vulnerability is heightened by neuroinflammation and oxidative stress arising from dysfunctional neurovascular interactions. Activated glial and EndMT-altered endothelial cells produce ROS and pro-inflammatory cytokines that impair synaptic plasticity, disrupt neurotransmitter homeostasis, and drive neuronal injury (Owens et al. 2022). Chronic exposure to these stressors disrupts long-term potentiation and accelerates dendritic and synaptic degeneration, linking vascular inflammation to cognitive decline (Shavit-Stein et al. 2021). Moreover, oxidative stress signaling directly impairs neuronal mitochondrial function, further reducing resilience to hypoperfusion and metabolic stress (Aliev et al. 2003) (Table 2). Collectively, these findings highlight EndMT as a key upstream mechanism linking chronic vascular stress to progressive neuronal dysfunction and cognitive decline in VCID (Fig. 3). By integrating microvascular remodeling, impaired neurovascular coupling, BBB disruption, and neuroinflammation, EndMT-driven endothelial plasticity offers a unifying framework for understanding how cerebrovascular disease leads to selective neuronal vulnerability and dementia.

Conclusion and future perspectives

VCID is increasingly recognized as a disorder of chronic cerebrovascular dysfunction, with microvascular failure preceding and predicting cognitive decline. Across clinical and experimental studies, endothelial injury emerges as an early driver of impaired cerebral perfusion, BBB disruption, neuroinflammation, and white matter vulnerability. This review proposes a unifying framework in which redox imbalance is a mechanistic regulator of endothelial plasticity, driving EndMT rather than a marker of vascular stress. By connecting source-specific ROS signaling (mitochondria, NADPH oxidases, uncoupled eNOS) to canonical EndMT pathways (TGF-β/Smad, Wnt/β-catenin, Notch), redox-sensitive transcription factors, and epigenetic mechanisms, this model explains how transient insults lead to lasting microvascular remodeling and progressive neurovascular failure. Direct evidence for EndMT in human brain tissue remains limited, as definitive lineage-tracing approaches are not feasible in humans and most current data rely on indirect or correlative observations. A key implication is that EndMT links “functional” endothelial dysfunction and “structural” vascular remodeling. Even partial, context-dependent EndMT states—likely common in the brain—may weaken junctions, shift endothelial secretomes toward pro-inflammatory mediators, and increase ECM deposition that stiffens microvessels. These changes drive VCID processes: chronic hypoperfusion with impaired neurovascular coupling, BBB fragility triggering glial activation, and self-reinforcing inflammatory/oxidative loops that stabilize maladaptive endothelial states. This perspective also explains why broad antioxidant strategies show limited clinical benefit: pathological redox signaling is compartmentalized and networked, not a uniform “oxidant burden” that can be globally neutralized. Future progress requires shifting from association to causality and intervention. First, stronger evidence for EndMT in human cerebrovascular disease is needed, using spatially resolved single-cell approaches and rigorous criteria to distinguish it from endothelial activation or marker overlap. Second, mechanistic studies must clarify “redox specificity”—identifying which oxidant species, from which source, and at which subcellular site are necessary and sufficient to trigger EndMT programs in brain endothelium. Third, identifying how redox signals establish durable transcriptional memory—through chromatin modifiers, RNA regulators, and redox-sensitive post-translational switches—will reveal points to interrupt EndMT before it becomes self-sustaining.

Translationally, promising candidate strategies target dominant initiating nodes or amplification hubs rather than indiscriminate ROS scavenging. Potential therapeutic classes include NOX4 inhibitors (e.g., GKT137831/Setanaxib), which have shown preclinical efficacy in reducing oxidative injury and EndMT-like transitions in experimental stroke models. Additionally, Nrf2 activators (e.g., sulforaphane or dimethyl fumarate) and eNOS recouplers (e.g., folic acid or sapropterin/BH4 supplementation) are actionable candidates to restore redox balance and preserve endothelial identity in the cerebral circulation. It should be acknowledged, however, that translating this redox-EndMT framework to clinical practice will require validation in human cerebrovascular tissue and rigorous causal testing in disease-relevant models, neither of which is currently fully established. Coupled with biomarkers of BBB integrity, cerebrovascular reactivity, endothelial activation, and matrix remodeling, this redox-EndMT framework may support patient stratification and mechanism-driven therapies targeting the vascular roots of dementia.

Glossary

VaD

Vascular dementia

BBB

blood–brain barrier

EndMT

endothelial-to-mesenchymal transition

ROS

Reactive oxygen species

eNOS

endothelial nitric oxide synthase

ECM

extracellular matrix

VE

vascular endothelial

α-SMA

α-smooth muscle actin

TGF-β

Transforming growth factor-β

AD

Alzheimer’s disease

mtROS

mitochondrial ROS

Nrf2

Nuclear factor erythroid 2-related factor 2

CBF

Cerebral blood flow

NO

nitric oxide

Authors’ contributions

Juhyun Song: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review, and editing. All authors have read and approved the final manuscript.

Funding

This research was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government [grant numbers RS-2025—02213506, RS-2025—19612989] (Juhyun Song).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interest

The authors declare no competing interests.