Human stem cell models in cerebral cavernous malformations

Abstract

Cerebral cavernous malformation (CCM) is a rare cerebrovascular disorder characterized by abnormal endothelial architecture and clinically manifests as hemorrhage, epilepsy, and neurological deficits. Two primary factors have constrained the advancement of research in the field of CCM. Firstly, the utilization of animal models, which do not fully recapitulate human neurovascular biology, has been a major impediment. Secondly, the restricted access to patient lesion tissue has hindered progress. These constraints impede mechanistic dissection and therapeutic translation. This review discusses how models derived from human pluripotent stem cells (hPSCs), particularly induced pluripotent stem cells (iPSCs), are advancing research on cardiovascular endothelial cells by enabling human-specific and patient-tailored disease modeling. These stem cell models complement classical mouse models, creating a synergistic approach. The following section summarizes recent advances in three key areas: disease etiology, model development, and translational applications. In particular, the iPSC-derived endothelial cell system has provided mechanistic insights into how mutations in CCM1/2/3 and PIK3CA disrupt endothelial homeostasis in both two-dimensional and three-dimensional contexts, leading to aberrant activation of downstream signaling pathways. The discussion extends to more advanced platforms, including vascular organoids and blood–brain barrier models that more faithfully recapitulate the neurovascular microenvironment and pathological cell–cell interactions. Furthermore, iPSC-based high-throughput drug screening facilitates target validation, drug repurposing, and the development of personalized therapeutic strategies. Although challenges remain regarding model maturity and standardization, stem cell-derived vascular models provide a robust framework for CCM research. This review provides a concise overview of the fundamental iPSC models frequently employed in CCM research and proposes a hierarchical mechanistic framework of “mutation-driven, signal amplification, and lesion evolution.” The advantages of 2D and 3D iPSC models for elucidating early endothelial abnormalities, cell-cell interactions, and tissue-level lesion formation are highlighted, and the applicability of various models for reconstructing the CCM microenvironment is emphasized. In conclusion, a model selection strategy for translational research is proposed: iPSC models should be used to elucidate human-derived mechanisms and for drug screening, while animal and chimeric models should be employed to study long-term disease progression, immune involvement, and in vivo validation.

Introduction

CCM is a rare vascular disease. Traditionally, CCM is divided into two types [1]: scattered and familial. In familial CCM, genetic factors play a leading role. The main pathological feature of the disease is the formation of vascular malformation in the brain or spinal cord. However, these vascular malformations lack normal vascular structure [2], which is caused by the thinning of capillaries and dilation of the lumen. Although CCM is also a vascular disease, it differs significantly from an arteriovenous malformation: CCM lesions lack an arterial blood supply. The unique pathological structure of CCM can lead to recurrent microbleeding and damage to local nerve fibre bundles. The most common clinical manifestations of CCM are epilepsy, followed by focal nerve function defects and acute or chronic headache [3–6]. However, approximately 20% of CCM cases were discovered incidentally without clinical symptoms.

The pathogenesis of CCM is complex (Fig. 1, for the pathogenic pathway). Single-cell analysis reveals that multiple cell types undergo intricate transcriptional reprogramming within CCM lesions, with significant alterations in the expression of genes associated with immune cells and inflammation [7]. In the pathogenesis of CCM, genetic factors occupy a central position. The occurrence of CCM is directly related to mutations in the CCM1/KRIT1, CCM2, and CCM3/PDCD10 genes [8–11]. According to existing research, mutations in these genes will disrupt the intercellular junctional structure of endothelial cells, compromise blood vessel stability, and thus induce pathogenic angiogenesis [12]. Currently, the disease model of CCM primarily utilises a genetically engineered mouse model. Although researchers have made some progress using these models [13], species differences between humans and mice remain an insurmountable obstacle to simulating human-specific blood vessels, blood-brain barrier function, and signal pathway regulation. Additionally, the direct use of patient tissue for research also poses the problem of sample scarcity.

Fig. 1.

Signaling pathways involved in CCM-related endothelial dysfunction. This schematic summarizes established signaling pathways implicated in CCM pathogenesis, integrated with proposed interactions based on current experimental evidence. LPS activates TLR4, leading to disruption of the CCM protein complex and dysregulation of the MEKK3–KLF2/4 signaling cascade, which contributes to transcriptional alterations associated with non-cell-autonomous effects. VEGF stimulates the PI3K–Akt–mTORC1 pathway via VEGFR2 to promote endothelial cell proliferation. TGF-β binds to TβRI/TβRII receptor complexes, inducing Smad2/3–Smad4 complex formation and remodeling of the endothelial phenotype, modulated by adaptor and inhibitory proteins. BMP signaling activates the Smad1/5/9 axis to regulate differentiation, and also engages the Erk1/2 pathway to control proliferation and migration, while Smurf1/2 regulate BMP signaling via ubiquitination. Solid arrows indicate activation or signal transduction; blunt-ended lines indicate inhibition; and dashed lines represent indirect or context-dependent interactions. These pathways collectively regulate endothelial homeostasis and contribute to CCM lesion progression. (Created with BioRender.com)

Given that the risks associated with surgical treatment for CCM are higher than those for other diseases [5], symptomatic drug therapy is the primary intervention strategy for CCM in clinical practice. Although studies have demonstrated that statins can improve endothelial barrier function by inhibiting the RhoA-ROCK signaling pathway [14], their efficacy and safety are limited by various factors, including heterogeneity in CCM genotypes and disease stages, blood-brain barrier permeability, and concerns about long-term safety. These factors collectively result in significant variability in statin response among patients, thereby limiting their widespread clinical use in CCM.

In short, there is an urgent need to establish a comprehensive experimental platform that integrates the patient’s specific genetic background, replicates the disease’s progression under controlled conditions, and enables high-throughput screening of stem cells. Human pluripotent stem cell technology—particularly iPSC technology—offers a highly promising solution to meet this demand, leveraging its unique advantage as a renewable cellular resource [15].

In recent years, stem cell technology has found increasing applications in CCM research. Leveraging their unique advantages of “patient specificity” and “multipotent differentiation potential,” iPSCs have emerged as highly promising research tools and potential therapeutic vehicles in this field [16], providing a systematic solution for establishing a complete research loop from etiological modelling to therapeutic translation (Fig. 2, for a schematic diagram of the closed-loop system).

Fig. 2.

CCM disease modeling and translational applications based on human induced pluripotent stem cell platforms. This schematic illustrates a progressive increase in model complexity, from 2D iPSC-ECs to 3D vascular organoids and BBB-like systems, enabling stepwise investigation of CCM pathophysiology. Patient-derived iPSCs capture genetic heterogeneity, including familial mutations (e.g., CCM1, CCM2, CCM3) as well as somatic mutations and mosaicism. 2D iPSC-EC models recapitulate key cellular phenotypes such as junctional disruption, MEKK3–KLF2/4 activation, and RhoA–ROCK dysregulation, and are primarily assessed using assays including immunofluorescence staining, barrier permeability assays (e.g., TEER), and molecular profiling (e.g., RNA sequencing).3D vascular organoids and BBB-like models exhibit higher-order structural and functional features, including aberrant vascular morphogenesis, barrier leakage, and impaired endothelial–pericyte–astrocyte interactions. These models are typically evaluated using advanced imaging (e.g., confocal microscopy), functional permeability assays, and analyses of multicellular interactions. In translational medicine, high-throughput drug screening is primarily conducted on two-dimensional platforms, while three-dimensional systems are used for functional validation and assessment of physiological relevance. Gene editing (such as CRISPR/Cas9) and the evaluation of patient-specific responses further support precision medicine strategies. (Created with BioRender.com)

The complex pathogenesis of CCM

CCM-related gene mutation

The core pathogenic mechanism of CCM is closely linked to specific gene mutations (Table 1: CCM Classic Pathogenic Genes). The classic causative genes are CCM1, CCM2/MGC4607, and CCM3 [17]. Mutations in these genes are the primary cause of familial and sporadic CCM, with pathogenesis driven by a “double mutation mechanism” [18]. Familial cases predominantly exhibit autosomal-dominant inheritance, with heterozygous CCM1 mutations being the most common [19]. CCM3 mutations carry the poorest prognosis, as they disrupt endothelial-pericellular interactions and cause cytoskeletal abnormalities [20–22]. However, why CCM3 mutations typically result in more aggressive phenotypes than CCM1 or CCM2 variants remains unclear.

Table 1.

CCM classic pathogenic genes

Gene	Pathogenic mechanism	References
CCM1	CCM1 deficiency disrupts endothelial cell-cell junctions. Its mutation induces endothelial barrier disruption, EndMT, oxidative stress, and inflammatory response	[10, 30, 31, 32, 42]
CCM2	CCM2 deficiency impairs the integrity of the CCM protein complex, resulting in endothelial hyperactivation and clonal expansion that drives CCM lesion development	[8, 12, 13, 38, 40, 80]
CCM3	Loss of CCM3 disrupts endothelial cell polarity and vesicular trafficking, promoting abnormal vascular morphogenesis and increased lesion instability	[8, 12, 37, 80]
PIK3CA	Activating mutations in PIK3CA enhance endothelial proliferation and survival, facilitating clonal expansion and accelerated CCM lesion growth	[22, 24]
MAP3K3	Gain-of-function alterations in MAP3K3 cause sustained endothelial activation and dedifferentiation, thereby promoting CCM lesion formation	[7, 12, 23, 35, 40]
mTORC1	Aberrant activation of mTORC1 in endothelial cells enhances abnormal growth and differentiation, contributing to CCM lesion progression	[36]
RhoA / ROCK	Hyperactivation of RhoA–ROCK signaling increases endothelial contractility and junctional instability, leading to vascular leakage and hemorrhage in CCM	[42, 53]
CTCFL	Ectopic activation of CTCFL in endothelial progenitor cells induces vascular malformation–like phenotypes resembling CCM	[28]
STK24/25	Loss of STK24/25 removes inhibitory control over endothelial activation, thereby facilitating the development of CCM-like vascular abnormalities	[43]

In addition, somatic mutations in PIK3CA and MAP3K3 have been detected in patients with CCM and may coexist with, or occur independently of, CCM1/2/3 mutations [23–26]. Mutations in the MAP3K3 gene may cause sporadic CCM but do not cause familial CCM. Researchers have used AAV-BR1 to specifically overexpress Map3k3^I441M^ in brain endothelial cells in a mouse model, resulting in widespread CCM-like lesions, thereby demonstrating that a single somatic Map3k3 mutation is sufficient to drive CCM-like lesions [27]. Developmental venous anomalies (DVAs) frequently occur near sporadic CCMs. Marchuk et al. detected the same PIK3CA mutation in DVAs and their adjacent CCM lesion areas, whereas MAP3K3 mutations were present only within the CCM lesions themselves. This suggests that DVA cells first acquire PIK3CA activation, followed by additional somatic mutations in MAP3K3 or CCM1/2/3 within their subclones, ultimately leading to CCM formation.

In addition to classic pathogenic gene mutations, abnormal expression of the newly identified Ctcfl gene induces cerebral vascular malformations and intracerebral hemorrhage in mice, similar to human CCM, and its expression affects the normal expression of CCM-associated proteins [28]. Studies have shown that postnatal and adult-specific knockout of Ngbr in endothelial cells (ECs) leads to defects in CCM1/2 expression and cerebral vascular pathology (e.g., vasodilation, increased blood-brain barrier permeability, and intracerebral hemorrhage) [29]. Endothelial-mesenchymal transition (EndMT) is also a contributing factor in the pathogenesis of cavernous hemangioma. CCM1, as an effector protein of Rap1, regulates intercellular junctions in endothelial cells [30]. Endothelial cell-specific disruption of the CCM1 gene induces EndMT, thereby promoting the development of vascular malformations. Biallelic loss-of-function mutations in the KRIT1 gene in endothelial cells lead to cerebral cavernous malformations [31]. However, to elucidate the role of germline KRIT1 heterozygous mutations in the pathogenesis of CCM, human preclinical models still need to be established. In addition to mutational mechanisms, inflammation is also widely recognized as a contributing factor to cerebral vascular malformations. CCM1 deficiency enhances the expression and phosphorylation of the redox-sensitive transcription factor c-Jun, thereby inducing the expression of its downstream target gene COX-2 [32, 33].

CCM-related signaling pathways

Initiating pathway

The core signaling pathways involved in CCM pathogenesis primarily include the MEKK3-KLF2/4, PI3K-mTOR, and TGF-β/BMP-β-catenin pathways, with synergistic regulatory interactions among these pathways [34–37]. The core signal axes are not independent of one another; rather, there is significant interaction among them. (Fig. 3, cross-regulation of signaling pathways and therapeutic targets). The MEKK3-KLF2/4 signaling axis is a key driver of this process [34, 38–40]. The MEKK3-KLF2/4 signaling pathway in endothelial cells can be activated by fluid shear stress and inflammatory mediators [35]. This pathway is the primary initiator of downstream disruption of the CCM complex. Zhou et al. [34] used a mouse CCM model to demonstrate that the expression of Mekk3 target genes Klf2 and Klf4 was already elevated in endothelial cells during the early stages of CCM pathology. However, pathways such as EndMT and SMAD/Wnt were not enhanced during the early stages of CCM formation. MEKK3 is an upstream kinase in the MAPK signaling pathway that activates the transcription factors KLF2 and KLF4. Under normal physiological conditions, the CCM protein complex inhibits MEKK3 activation by interacting with the HEG1 receptor, thereby limiting the overexpression of KLF2/4 and maintaining the structural and functional integrity of blood vessels [41]. The MEKK3-KLF2/4 signaling axis plays a central role in CCM lesions (Fig. 4, CCM Signal Transduction Hierarchy).

Fig. 3.

CCM: from anatomical features to molecular mechanisms. The activation mutation of the PIK3CA gene can abnormally activate the PI3K-AKT-mTOR pathway, promote the excessive proliferation of endothelial cells and induce vascular malformations. At the same time, the loss of function of the CCM protein complex will destroy the homeostate of the MEKK3-ERK5-KLF signaling pathway and further change the behavior of endothelial cells. There is significant cross-regulation of these two core pathways, which eventually converge in mTORC1, thus realizing the integrated regulation of genetic signals and mechanical transduction. Targeted mTORC1 interventions (such as rapamycin treatment) are expected to be a potential strategy to improve CCM vascular abnormalities. (Created with BioRender.com)

Fig. 4.

Conceptual model of CCM signaling hierarchy and endothelial responses. CCM mutations initiate disease by activating the MEKK3–KLF2/4 transcriptional program, which drives an endothelial state switch. Downstream, multiple signaling pathways—including PI3K–mTOR, Rho/ROCK, β-catenin, and TGF-β/BMP—amplify endothelial activation and promote pathological outcomes. These convergent signals lead to endothelial-to-mesenchymal transition (EndMT), increased proliferation, and disruption of the endothelial barrier. Microenvironmental factors, including shear stress, immune cells, and Cav1/mTORC1 signaling, further modulate these processes in a context-dependent manner. Endothelial subtype heterogeneity influences responsiveness to CCM signaling, contributing to differential disease susceptibility. Arrows denote activation (solid), integration (bold), crosstalk (bidirectional), or modulation (dashed), while dashed boxes and color/position encode functional grouping and hierarchical organization of signaling, context, and endothelial responses.(Created with BioRender.com)

Microenvironmental regulation and pathological amplification

Alterations in biological processes, including EndMT, immune and inflammatory responses, gut microbiota dysbiosis, hypoxia, and hemodynamic changes, also contribute to the development of CCM. Although these microenvironmental factors do not directly initiate CCM formation, they significantly modulate disease severity.

EndMT underlies the pathogenesis of CCM. Previous investigations found that EndMT markers were expressed in the endothelial layer of the venous sinuses in all cases [42]. Pathological activation occurs through deletion of the CCM protein complex or STK24/25 [43], which upregulates KLF2/4 expression, disrupts endothelial cell junctions, initiates EndMT, exacerbates local inflammation, and modulates downstream PI3K pathways. Enhanced KLF2/4 function triggers EndMT by disrupting endothelial cell junctions. Furthermore, KLF2/4 can upregulate inflammation-associated molecules (such as IL-6, VCAM-1, and ICAM-1), exacerbating local inflammatory responses, which further disrupt vascular homeostasis and expand the lesion [44–47]. MEKK3 and its downstream transcription factors, KLF2 and KLF4, have been identified as key pathways by which endothelial cells transduce hemodynamic shear stress and inflammatory signals. In vitro studies have demonstrated that fluid shear stress strongly stimulates the expression of KLF2 and KLF4 [23–25], while in vivo studies have associated their expression with high-shear-stress regions and established their functional role in shear-dependent cardiovascular remodeling.

Hypoxia also plays a significant role in the formation of cerebral cavernous malformation lesions [48]; persistent mild hypoxia modulates cell-specific neuroinflammatory interactions via the CX3CR1-CX3CL1 signaling pathway, leading to heterogeneity in CCM severity [49]. A study by Sadegh H et al. revealed that gut microbiota dysbiosis and its associated innate immune responses play a key role in the experimental pathogenesis of CCM, highlighting the importance of the gut-brain axis in this disease [50]. CCM3 deficiency also drives abnormal proliferation of cerebral endothelial progenitor cells. It promotes the development of CCM lesions by activating mTORC1 signaling through aberrations in the Cav1/caveolae-mediated intracellular vesicle-lysosome localization mechanism [36]. This further underscores the complexity of the underlying pathophysiological mechanisms of CCM.

The absence of the CCM protein complex or STK24/240 upregulates KLF2/4 expression, leading to pathological activation and modulating the downstream PI3K pathway. Sustained activation of the PI3K-mTOR pathway promotes abnormal proliferation of endothelial cells and enhances the expression of angiogenic factors, thereby causing vascular structural damage [27]. mTOR inhibitors can alleviate the pathology [25, 37, 51]. Within the same CCM lesion, mutations in CCM1, CCM2, CCM3, and MAP3K3 are mutually exclusive but can coexist with PIK3CA. This study supports the notion that PIK3CA acts as a universal “amplifier” that can be superimposed on the loss of different CCM complexes or on the activation of MAP3K3 [26].

The TGF-β/BMP and β-catenin pathways promote disease progression by regulating EndMT [52, 53]. CCM3 deficiency preferentially activates the β-catenin pathway to drive endothelial cell dedifferentiation, subsequently activating the TGF-β/BMP pathway, which inhibits β-catenin activity through a feedback mechanism; together, these two pathways disrupt vascular homeostasis [42]. The physical interaction between CCM1and CCM2 proteins is critical for the localization of intercellular junctions in endothelial cells. The CCM1–CCM2 interaction regulates vascular barrier function by inhibiting the Rho/ROCK signaling pathway, which is dysregulated in human CCM endothelial cells [34, 54]. These signaling pathways act as amplifiers, exacerbating endothelial dysfunction, proliferation, and the transition of endothelial cells to mesenchymal cells.

Recent studies based on mouse CCM models and transcriptomic analyses have introduced the concept of “CALMN (CCM endothelial cells, astrocytes, leukocytes, monocytes/microglia, and neutrophils) interactions,” highlighting that astrocytes and CCM endothelial cells drive the recruitment of microglia and peripheral immune cells to the lesion site via chemokines such as Cxcl1 and Ccl2 [48] (Table 2 The main signaling pathways involved in CCM lesions).

Table 2.

CCM-related signalling pathways

Pathway classification	Signalling pathways	Pathogenic mechanism	References
Core cellular signalling pathways associated with pathogenicity	MEKK3–KLF2/4 Signaling Axis	Loss of CCM proteins relieves inhibition of MAP3K3 (MEKK3), leading to persistent activation of downstream KLF2 and KLF4, transcription factors that drive a pathogenic endothelial program. It integrates hemodynamic and inflammatory signals and promotes cellular proliferation, migration, and lesion formation.	[8, 13, 34, 35, 40]
	RhoA–ROCK Signalling Dysregulation	The CCM protein complex normally inhibits the RhoA–ROCK pathway. Loss leads to ROCK overactivation, increased actomyosin contractility, cytoskeletal stress, and breakdown of endothelial junctions.	[42, 53]
	PI3K-mTOR	PI3K–AKT → mTOR activation promotes endothelial proliferation, survival, and lesion growth, and can synergise with MAP3K3 activation	[25, 35, 36, 41]
	MAP3K3–mTOR	MAP3K3–mTOR signalling constitutes a growth-permissive axis that converts endothelial genetic lesions into progressive CCM pathology.	[33, 36, 43]
	TGF-β/BMP/β-catenin	β-catenin proteins concentrate in the nucleus to drive the expression of dedifferentiation-related proteins, including stem cell/endMT markers, and to activate TGF-β/BMP signalling, promoting vascular pathology.	[42, 51, 52]
	Oxidative Stress Pathways	KRIT1 loss produces oxidative imbalance, and ROS overproduction may sensitize the endothelium to dysfunction	[31]
	Hypoxia-CX3CR1-CX3CL1	Persistent mild hypoxia influences cell-specific neuroinflammatory interactions through the CX3CR1-CX3CL1 signalling pathway, leading to CCM severity heterogeneity	[48]

Endothelial lineage-specific responses to pathogenic signaling

The same signaling pathway can lead to different fates depending on the endothelial state. For example, the role of MEKK3 is not universal among endothelial cells but rather stage- and state-specific [35]. Using single-cell RNA sequencing (scRNA-seq), Kahn et al. identified changes in endothelial cell types during the endothelial-to-hematopoietic transition (EHT) in Mekk3-deficient mice. EHT is a classic developmental pathway: arterial endothelial cells (AE) → pre-hematopoietic endothelial (pre-HE) → hematopoietic endothelial (HEC) → intra-arterial hematopoietic cell (IAHC). This study used scRNA-seq to identify distinct endothelial cell subpopulations and found that not all cell types were equally affected by MEKK3 loss; pre-HE cells were most severely impacted, with Notch/Wnt signaling disrupted in this population; MEKK3 deficiency also leads to impaired subsequent differentiation of the AE subpopulation, resulting in cell accumulation. This indicates that MEKK3 activity is critical for specific stages of endothelial differentiation rather than functioning uniformly across all endothelial cell populations. Furthermore, venous endothelial cells are highly sensitive to KLF2 signaling [55].

Stem cell-derived model

In studies of CCM vascular pathology, mesenchymal stem cells (MSCs), endothelial progenitor cells (EPCs), and induced pluripotent stem cells (iPSCs) are currently the most widely used sources of stem/progenitor cells.

MSCs and EPCs primarily regulate endothelial cell proliferation and vascular repair by releasing angiogenic factors, including vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF), via paracrine mechanisms [51–53]. However, the main challenges these two cell types face in clinical translation are high population heterogeneity and functional drift during in vitro expansion.

In contrast, iPSCs offer unique advantages: (1) they preserve the patient-specific genomic background; (2) they support precise CRISPR/Cas9 gene editing; and (3) they can be directed to differentiate into specific subtypes of cerebral vascular endothelial cells. By reprogramming somatic cells from CCM patients into iPSCs and subsequently differentiating them into cerebral microvascular endothelial cells or vascular organoids, researchers can precisely model the functional effects of CCM1/2/3, MAP3K3, or PIK3CA mutations at various stages of endothelial differentiation, providing an ideal platform for elucidating human-specific pathological mechanisms [56, 57].

iPSC-based model construction

The construction of iPSCs models primarily follows two approaches: patient-derived iPSCs generated by reprogramming somatic cells from individuals carrying CCM mutations, and isogenic iPSCs established via CRISPR/Cas9-mediated gene editing. This approach directly preserves the mutated genetic background, serving as the gold standard for studying clinical heterogeneity and discovering novel genotype-phenotype associations. Alternatively, CRISPR/Cas9 technology is used to knock out the CCM1, CCM2, and CCM3 genes in healthy human iPSCs, establishing homozygous or heterozygous mutant lines. Pilz et al. observed significant dysregulation of CCM-related pathway genes after differentiating CRISPR/Cas9-knockout CCM1 human iPSCs into endothelial cells [58]. However, the study indicates that mutations in the CCM gene have minimal impact in an undifferentiated state but result in significant pathological transcriptomic changes following endothelial differentiation. KLF2/4 and their enhancers are endothelial lineage-specific. Studies have clearly demonstrated that KLF2 and KLF4 bind to numerous enhancer regions in endothelial cells across various vascular beds, including the heart and lungs, and that these enhancers show a high degree of overlap with endothelial-specific open chromatin and H3K27ac.CRISPR/Cas9-mediated inactivation of CCM3 in human ECs demonstrated that CCM3 deficiency causes severe defects in cell-cell interactions and actin cytoskeletal organization [59].

By applying CRISPR/Cas9 genome-editing technology to human iPSCs, we confirmed that this deletion results in a loss of CCM1 protein and dysregulation of KLF2, THBS1, NOS3, and HEY2 gene expression in iPSC-derived endothelial cells. Based on these findings, the variant was reclassified as a “likely pathogenic variant” [60]. Hester ME et al. used CRISPR-Cas9-mediated homology-directed repair to correct two pathogenic TSC2 variants in iPSCs derived from Tuberous Sclerosis Complex (TSC) patients and successfully established two homozygous cell lines. CRISPR-mediated knockout of CCM1/2/3 can recapitulate EndMT, barrier disruption, and abnormal vasculature-like structures in both 2D and 3D models [61]. Professor Ute Felbor’s team isolated blood-vessel-derived endothelial cells (BOECs) from patients with multiple CCMs who carried a CCM1 gene mutation. They then used CRISPR/Cas9 gene-editing technology to transfect patient-derived BOECs with a high-fidelity Cas9 variant, thereby expanding patient-specific CCM1 gene-knockout clones [62].

iPSC-based 2D model: iVECs

iPSC-derived brain microvascular endothelial cells (iBMECs) or iPSC-derived venous endothelial cells (iVECs) provide a fundamental platform for studying cellular autonomy defects, including barrier function, cell junctions, proliferation, and apoptosis. Research indicates that CCM-related phenotypes exhibit strong lineage-specificity, fully manifesting only after cells differentiate into the endothelial lineage. Gu et al. [63] demonstrated that iPSC-derived endothelial cells (iPSC-ECs), through in vitro co-culture and in vivo co-transplantation, promote the maturation of hPSC-derived CCM models, enhance graft vascularization, and improve cardiac function. These effects were confirmed by comparing experimental outcomes between single-cell and combined cell transplantation. Pan et al. first differentiated human iPSCs into iVECs. They established a TIE2-mutant hemangioma iVEC model that successfully reproduces the pathogenesis of venous malformations and fills a gap in reliable disease models for CCM [64].

iPSC-based 3D models: vascular organoids and BBB systems

Recent platforms that combine blood-brain barrier models with three-dimensional vascular organoids have advanced exploratory research on CCM by more accurately recreating the complex physiological environment of the human body. These models enable comprehensive simulation and study of the microstructural environment within CCM lesions. Vascular organoids focus on the development, differentiation, and function of blood vessels themselves, and are typically derived from 3D microvascular networks formed by iPSCs or endothelial precursors. Blood-brain barrier (BBB) organoids are typically microvascular-glial barrier structures composed of cerebral microvascular endothelial cells, pericytes, astrocytes, and other cells [65–67]. In comparison, blood-brain barrier organoids place greater emphasis on the barrier structure, highlighting functions such as tight junctions and trans-BBB transport, making them better suited to simulate the selective permeability of the human blood-brain barrier Complex (hBBB) and interactions within neurovascular units [68, 69].

Vascular Organoids: Two studies published in 2019 were the first to report the differentiation of human vascular organoids [70, 71]. Skowronek et al. used a high-throughput 96-well plate system to induce human iPSCs to differentiate into vascular organoids, and subsequently knocked out the CCM1, CCM2, and CCM3 genes [72]. They observed that organoids lacking CCM1 or CCM3 exhibited significantly enlarged volumes and contained cellular cavities, whereas CCM2 knockout showed no discernible differences. Immunofluorescence analysis showed that the close-linking protein (such as ZO-1) in the CCM1/CCM3 knockout group was significantly reduced, indicating that the intercellular adhesion function was impaired. Further single-cell RNA sequencing analysis reveals that knocking out any gene alters the cell composition pattern of organs. For example, CCM1 deletion leads to significant amplification of a specific cell group, while CCM2 deletion results in significant amplification of another specific cell group. This study combines the high-throughput advantage of single-cell RNA sequencing with an organ-like culture system, revealing that different CCM proteins not only share a common function but also play distinct regulatory roles in vascular development. Research confirms that vascular organs are an ideal model for exploring the mechanism of CCM. Although breakthroughs have been achieved in constructing complex organoids, efficiently vascularizing three-dimensional tissues to provide structural support for long-term organoid growth and directed differentiation remains a critical technical challenge that urgently requires resolution. Three-dimensional vascular organoids derived from human iPSCs were implanted into a chip. Similar to BVOs cultured in wells, these organoids self-assembled into a three-dimensional interconnected network of authentic capillaries on the chip. These organoids developed capillaries featuring hollow lumens encased in prototypical basement membranes. This approach bypasses the need for microfluidic platforms, providing a straightforward, adaptable, and reliable method for creating vascularized organoid chips. It effectively addresses key challenges in organoid growth and maturation while bolstering the potential of organoids for studying CCM and other diseases [73].

hBBB Organoids: Dao et al. assembled human brain organoids with vascular organoids to create functional 3D hBBB organoids, introducing patient-derived iPSC-based models of CCM mutations [74]. They observed that hBBB constructs harboring CCM mutations recapitulated the morphological features and BBB disruption seen in patient lesions: transmission electron microscopy revealed impaired endothelial tight junctions and increased blood-brain permeability. Single-cell spatial transcriptomics comparing these constructs with patient surgical specimens identified key CCM pathological marker genes specifically upregulated in vascular cell clusters. This study represents the first in vitro platform to simulate neuro-vascular interactions in CCM, providing an innovative model for exploring its pathogenesis. A survey of drug brain permeability validation highlighted the lack of reliable in vitro strategies to simulate the BBB and cerebral vascular injury in the development of effective central nervous system (CNS) drugs [63]. Using a newly created hBBB model based on iPSCs, the results demonstrated a strong correlation between in vitro and in vivo drug brain permeability (R² = 0.83; P = 0.008). The research team concluded that the iPSCs-hBBB model can be integrated into CNS drug screening workflows and has the potential to investigate species differences in blood-brain barrier permeability.

iPSC models drive advancements in the CCM field

By co-culturing wild-type and mutant endothelial cells or constructing chimeric organoids, the model visually demonstrates how the wild-type environment “nurtures” the abnormal proliferation of mutant cells, providing a mechanistic explanation for the focal growth of sporadic CCM lesions. Pilz RA et al. [58] employed RNA sequencing to investigate differential gene expression in CCM1-knockout iPSCs (CCM1-/- iPSCs), EPCs, and ECs. Deletion of the CCM1 gene will only cause slight transcriptional changes in iPSCs and EPCs. Only when the cell differentiates into an endothelial state will significant abnormalities in gene regulation occur. This phenomenon demonstrates that the inactivation of CCM1 function is dependent on the synergy of the angiogenic microenvironment, which triggers the pathogenic pathway of CCM.

The gene expression profile of iPSCs exhibits stage-specific changes and forms a complex regulatory network with signaling pathways such as the PI3K-mTOR and TGF-β-BMP pathways. This mechanism elucidates the inherent limitations of single-target interventions and offers crucial theoretical justification for developing combination therapy strategies. Recent studies have employed this technique to derive three iPSC lines from a patient with minor multiple cerebral vascular malformations and a heterozygous CCM1 mutation [75]. Upon integration into a standard vascular system, these iPSCs generated abnormal vascular configurations. Transcriptomic analysis revealed high PEG3 expression in iPSC-derived CCM endothelial cells. The iPSCs model provides a direct demonstration of differences in cellular behavior driven by distinct genotypes, offering a cellular explanation for clinical phenotypic variation. CCM is not merely a localized vascular disorder but is accompanied by systemic molecular dysregulation. A recent prospective study revealed significant alterations in plasma levels of a series of inflammation- and angiogenesis-related proteins (such as CD14, THBS1, and CCL5) in patients with CCM. Among these, markers such as ROBO4 and TM can predict future clinical events, aligning closely with the pathological pathways suggested by iPSC models [76].

Sun et al.’s study generated vascularized brain organoids by fusing brain organoids with vascular organoids. Following fusion, brain-specific pericytes and CD31 + blood vessels, surrounded by astrocytes, were observed, and the resulting structures exhibited more mature neurons and tight junction markers, demonstrating that this model more closely resembles a complete neurovascular unit [67].

Accelerate drug discovery and screening

A Peking University team employed iPSC models and organoids for high-throughput drug screening. Using an iPSC-derived vascular model and a deep learning-based drug efficacy prediction system, they identified the Food and Drug Administration (FDA)-approved anti-leukemia drug bosutinib, which effectively reversed pathological phenotypes in an in vitro iPSC-derived vein endothelial cell model [64]. In AVM chip models, experiments have shown that MEK-targeting drugs reduce leakage in mutant endothelial networks, suggesting the MAPK pathway as a potential therapeutic target [77]. Skowronek et al. [72] established high-throughput drug screening platforms (e.g., 96-well plate formats) using iPSC-differentiated endothelial cells or vascular organoids to identify promising therapeutic compounds, validating the feasibility of targeting CCM-like neoplastic features and accelerating therapeutic development (Fig. 5, Core signaling networks and therapeutic targets).

Fig. 5.

Core signaling network in CCM revealed by human iPSC-based models and therapeutic targets. iPSC-based endothelial models demonstrate that loss of the CCM protein complex drives hyperactivation of MEKK3–KLF2/4, RhoA–ROCK, and PI3K–mTOR signaling, integrating mechanical and inflammatory cues to promote junctional instability, EndMT, and hemorrhagic lesion formation. Key druggable nodes validated in human iPSC models are indicated. (Created with BioRender.com)

Rath et al. [78] reported that endothelial cells lacking CCM3 exhibit markedly increased proliferative activity when cultured with wild-type endothelial cells or when incorporated into chimeric human iPSC-derived vascular organoids. Treatment with the anticancer compound NSC59984 effectively suppressed this excessive proliferation in mutant endothelial populations. Further analyses using flow cytometry–based cell sorting, combined with RNA sequencing, demonstrated that co-culture conditions triggered the elevated expression of proangiogenic chemokine-related genes in wild-type endothelial cells. It is worth noting that in the single-layer culture system of CCM3 gene-knockout (CCM3−/−) endothelial cells, when co-cultured with wild-type endothelial cells, the expression levels of multiple genes that were significantly downregulated in CCM3-/- cells returned to normal physiological levels. These experiments demonstrate that wild-type endothelial cells contribute to the formation of a microenvironment that fosters abnormal endothelial cell proliferation. Therefore, the development of targeted treatment strategies based on the cancer-like biological characteristics of CCM lesions is expected to become the key direction of drug research in this field.

Molina SG et al. [79] successfully established healthy control iPSC lines from peripheral blood mononuclear cells of healthy young donors. These cell lines have the potential to differentiate into various neurovascular progenitor cell types, including endothelial cells, astrocytes, and neurons. Healthy-controlled iPSC cell lines can serve not only as an experimental platform for drug development and screening but also as a health reference standard for comparative research to simulate early CCM.

Decision-making framework for selecting CCM models

To guide model selection for specific CCM research questions, we reviewed a large body of literature. We proposed a decision-making framework that matches each experimental focus with an optimal model system (2D iPSC-ECs, iVECs, vascular organoids, blood-brain barrier organoids, or organoid-on-a-chip systems). We also listed key quantitative metrics, including trans-epithelial electrical resistance (TEER), permeability tracers, junctional markers, and single-cell/ spatial transcriptomics (Table 3 Decision Framework for CCM Model Selection) [22, 46, 58, 65–67, 80–85].

Table 3.

Decision framework for CCM model selection

Research question	Recommended model	Key readouts	References
Early signaling (MEKK3-KLF2/4, Rho/ROCK)	2D iPSC-ECs, iVECs	qPCR/RNA-seq (KLF2/4, RhoA), WB (p-MLC2), IF (stress fibers)	[46, 58]
Barrier function & junctional integrity	BBB organoids	TEER(> 200 Ω·cm²), FITC-dextran permeability, IF (CLDN5, ZO-1)	[65, 67, 80]
EndMT & phenotypic switching	Vascular organoids, 2D co-culture	IF/flow cytometry (α-SMA, CD31 loss), qPCR (SNAIL1/2, TGF-β targets)	[81, 82]
Multi-cellular interactions (EC-pericyte-astrocyte)	BBB organoids + vascular organoids	Live imaging (vascular sprouting), scRNA-seq (cell state transitions)	[66, 67]
Microenvironment & flow effects	Organoids-on-chip (µfluidics)	Permeability tracers under shear, spatial transcriptomics	[83, 84]
Clonal expansion & mosaicism	Mixed population organoids + lineage tracing	Barcoding/scRNA-seq (clone size), live imaging (proliferation)	[22, 85]

Discussion

Endothelial-specific CCM pathway activation

Current evidence suggests that CCM mutations are not directly pathogenic but become functionally relevant after endothelial differentiation and under specific regulatory conditions. During differentiation, chromatin accessibility at key loci such as KLF2 and KLF4, along with enhancer activation, gradually establishes itself, thereby facilitating context-specific transcriptional responses that are absent in the undifferentiated state [46]; hemodynamics plays a critical role in activating CCM-related signaling pathways. Shear stress is the primary inducer of KLF2/4 expression in endothelial cells, linking vascular blood flow to transcriptional regulation. This mechanosensitive regulatory program may be absent in immature cells, which may account for the delayed emergence of the CCM phenotype. Endothelial cell-specific enhancer elements may further restrict the activation of CCM-related pathways to differentiated endothelial cells. Regulatory regions governing genes such as KLF2/4 and components of the Rho/ROCK pathway exhibit endothelial cell-specific activity, suggesting that full activation of this pathway requires a lineage-specific transcriptional architecture [86, 87].

Most of these key pathways are endothelial pathways conserved across species. Evidence from zebrafish, rodent, and human samples consistently indicates that CCM protein complexes maintain vascular homeostasis by regulating pathways such as MEKK3-KLF2/4-RhoA/ROCK and PI3K-mTOR; this mechanism is highly conserved across vertebrates and between in vivo models and human iPSC-derived endothelial cells [34, 38, 58, 88]. Although these pathways are highly conserved between mice and humans, single-cell omics studies suggest the presence of specific endothelial subpopulations in human CCM lesions (such as those with high KLF2/4 and EndMT expression), accompanied by human-specific immune and inflammatory expression profiles; these subtle differences are often not fully replicated in mice [89, 90]. However, human iPSC-derived endothelial and organoid models allow for the precise manipulation of genetic mutations within patient-specific or homozygous genetic backgrounds, enabling the differentiation of the phenotypic effects of various mutation types, allelic doses, and combinations (such as bigenic or trigenic mutations)—a feat that is typically difficult to achieve at high throughput in rodent models; iPSCs can be directed to differentiate into human-specific neurovascular units to reconstruct the microenvironment, allowing for the comparison of responses to CCM mutations across different vascular beds. Furthermore, they enable the systematic evaluation of candidate drugs—such as Rho/ROCK inhibitors, mTOR inhibitors, and anti-inflammatory agents—against the MEKK3–KLF2/4 axis and its downstream amplifiers within the human cerebral microvasculature. This feat is difficult to achieve systematically in single-species models.

It is proposed that iPSC-based models should be understood within a hierarchical CCM mechanism framework, progressing from mutation-driven changes to signal amplification, and then to endothelial phenotypic shifts and lesion progression. Based on this framework, 2D and 3D human-derived models are better suited for early-stage mechanism analysis and drug screening, while in vivo and chimeric models are better suited for reconstructing the chronic disease course and complex microenvironment.

CCM mouse models and iPSC-derived models

In traditional studies of CCM mechanisms, mouse models have reproduced some of the typical pathological features of cerebral cavernous malformations; however, significant limitations remain in characterizing specific molecular regulatory networks and genetic variation among individuals. In recent years, experimental models based on hPSCs have made rapid progress in studying organ systems such as the blood-brain barrier and the vascular system. They are increasingly being used to construct CCM disease models. These organoid systems preserve patients’ original genomic background and, in vitro, reproduce various pathological changes closely associated with CCM pathogenesis, including damage to endothelial cell tight junctions, clonal expansion abnormalities, and imbalances in signaling regulatory networks. Against this backdrop, iPSC models provide a feasible and reproducible experimental platform for systematically studying the pathogenesis of CCMs and exploring related targeted therapies. Models derived from iPSCs are characterized by their “human origin” and “patient specificity”; simultaneously, these models address the scarcity and difficulty in obtaining human pathological samples. These models can accurately reproduce endothelial cell abnormalities and signaling pathway imbalances caused by specific genetic mutations.

Long-term clinical follow-up has shown that CCM is a slowly progressive condition prone to recurrent bleeding. Over several years, CCM lesions exhibit fluctuations in size and bleeding risk, and imaging studies reveal that hemorrhagic foci and hemosiderin deposits gradually accumulate as the disease progresses [91, 92]. However, existing iPSC-derived endothelial and organoid models primarily capture early cellular changes over several weeks, making it difficult to replicate clinical events fully. Taken together, clinical cases and mouse models reveal the evolutionary process of CCM lesions over a broader temporal scale. In contrast, iPSC models are better suited for elucidating how CCM-associated genetic and somatic mutations reshape endothelial cell fate and neurovascular unit interactions in the early stages. In the future, by incorporating immune cells and microglia into organoids and extending culture duration, we may be able to approximate the “subchronic” stage of disease progression [48]. The temporal limitations of organoids can complement those of mouse models; the two types of models are not mutually exclusive within the research framework. Currently, due to the complexity and high risk of intracranial surgery, CCM treatment relies primarily on medication. We also look forward to the role of human pluripotent stem cell models in advancing our understanding of the underlying mechanisms and providing insights for therapeutic interventions.

Limitations of iPSC-derived models

Despite these advantages, existing iPSC-based CCM models still have limitations in their research. Current iPSC-based experimental models are unable to replicate the pathophysiological processes of CCM fully. iPSCs are typically engineered to carry CCM1/2/3 or PIK3CA/MAP3K3 mutations using CRISPR or viral methods, yielding homogeneous mutant cells. While this approach is indeed advantageous for elucidating cell-autonomous signaling abnormalities and morphological phenotypes, it does not account for the coexistence of normal and mutant cells or differences in the microenvironment. Consequently, iPSCs have limited ability to recapitulate true mosaic lesions [22, 93, 59].

Compared with adult cerebral microvascular endothelial cells, iPSC-ECs typically exhibit immature transcriptional regulation, incomplete blood-brain barrier function, and abnormal metabolic states. These defects severely limit their ability to simulate the chronic progression of CCM pathology. The pathological progression in CCM patients typically spans several years or even decades, making the immaturity of these cellular model phenotypes a particularly significant concern. This long-term, dynamic disease process remains the primary technical bottleneck in accurately reproducing pathological processes in in vitro models.

Another technical limitation of existing CCM stem cell models is the near-total absence of immune cells. Inflammatory signaling and immune cell infiltration are considered key drivers of CCM growth, instability, and bleeding. Although immune components play a key role in the progression of CCM lesions, they are currently absent from iPSC models. Therefore, we recommend prioritizing two types of co-culture experiments: first, co-culturing iPSC-derived microglia with CCM-mutant vascular organoids to determine whether microglial-derived TGF-β enhances EndMT. Key indicators to be assessed include the α-SMA/CD31 ratio [94–96], a decrease in TEER, and pathological-like vascular branching [48, 67, 94]. Second, under TGF-β/IL-1β inflammatory stimulation, co-culture mutant endothelial cells with iPSC-derived macrophages to observe the “macrophage-TGF-β-endothelial cell-to-mesenchymal transition” axis [95, 96]. The biomarkers for this axis include p-SMAD2/3 and Snail1 expression levels, as well as changes in permeability labeled with fluorescein isothiocyanate (FITC)-dextran. These studies will address the missing immune-vascular interaction axis in current models [95, 97, 98]. Therefore, it would be of great significance to combine endothelial cells or vascular tissue fragments derived from induced pluripotent stem cells with patient-derived macrophages or microglia to simulate key features of vascular-immune cell interactions in CCM pathology in vitro and establish a co-culture system that incorporates immune cells.

In addition, different laboratories have employed a variety of protocols to differentiate iPSCs into endothelial cells and vascular tissues, including variations in growth factor combinations, extracellular matrix selection, and genetic engineering strategies; these factors have significantly influenced endothelial cell phenotype and signaling pathway activity. Factors such as gender and stress responsiveness have further limited the consistency of research findings. This underscores the urgent need to establish a more rigorous standardized quality control framework.

Engineered chimeric models and future directions

To overcome the current limitations of iPSC-based CCM models, it is necessary to integrate biotechnological strategies, immune-related mechanisms, and corresponding in vivo validation to recreate specific vascular microenvironments (e.g., microfluidic chip systems) and establish engineered chimeric models harboring CCM-associated mutations.

Animal experiments have demonstrated that endothelial mosaicism is a key driver of CCM formation. When CCM3 inactivation is induced in a small number of endothelial cells using the Cre recombinase / LoxP site (Cre-LoxP) system, clonal expansion and lesion formation of the mutant cells can be observed, indicating that the competitive advantage of focal mutant cells within a normal background is crucial for lesion morphology [22].

In fact, to overcome the limitations posed by the chimeric phenomenon, researchers have developed several practical experimental strategies to mimic the chimeric phenomenon and multi-target biological mechanisms: (1) Controlled mixing ratios: Mutant cells (CCM1/2/3 knockout) are co-cultured with wild-type iPSC-ECs at predetermined ratios [99], followed by in vivo imaging of clonal expansion rates, scRNA-seq to assess transcriptional differences between mutants and wild-type cells, and flow cytometry analysis of proliferation and apoptosis, among other key indicators [100, 101]. (2) Inducible CRISPR: This technique utilizes doxycycline-inducible Cas9 combined with timed gRNA delivery to induce secondary mutations (MAP3K3GOF, PIK3CA) in pre-established CCM1-knockout organoids [102, 103]. Subsequently, the temporal dynamics of signaling pathways (p-MEKK3, KLF2/4), the decline in barrier function (TEER), and the formation of lesion-like structures are examined. (3) Lineage tracing/barcoding: During iPSC differentiation, we integrate lentiviral barcodes or CRISPR-based lineage tracers (e.g., scGESTALT) [104]. Track clonal dynamics in vascular organoids under stress (hypoxia, shear stress); key readout metrics include changes in barcode abundance (clonal fitness), spatial transcriptomics (lesion core and periphery), and the competitive advantage of mutated clones. These methods enable the quantitative assessment of “mutational fitness advantage” and the synergistic effects of multiple mutations while preserving physiological cellular heterogeneity [22, 105].

In the future, the use of engineered chimeric systems may be more effective at capturing the complete process of mosaicism and multi-hit interactions [15].

Lessons from organoid research in other diseases

Liver organoids have successfully mimicked the metabolic and multicellular pathological features of Wilson’s disease [106]; meanwhile, a model of periventricular leukomalacia (PVL) has demonstrated that brain endothelial cells derived from iPSCs can better reproduce human-specific developmental vascular pathology. These findings indicate that iPSCs can reproduce tissue-specific cellular interactions while preserving the patient’s genetic background, providing important insights for the design of CCM organoid models.

Conclusion

In summary, while the present model, based on human embryonic stem cells, exhibits certain deficiencies, the platform has inaugurated a hitherto unexplored avenue for elucidating the pathogenesis of CCM. However, its unique advantage lies in its ability to elucidate disease pathology within a human-specific research framework systematically, evaluate the effectiveness of candidate treatments, and conduct direct in vitro studies of the genetic characteristics and signalling pathways of human endothelial cells. The subsequent stage of this research should involve promoting methodological innovation in model development and fostering cross-laboratory collaboration to standardise the research system. It is crucial to understand that this is essential for fully leveraging the transformational potential of the CCM disease model based on induced pluripotent stem cells.

Abbreviations

AE

Arterial endothelial cell

BBB

Blood-brain barrier

bFGF

Basic fibroblast growth factor

BOECs

Blood-vessel-derived endothelial cells

BVO cells

brain/vascular endothelial cells

CALMN

CCM endothelial cells, astrocytes, leukocytes, monocytes/microglia, and neutrophils

Cas9

CRISPR-associated protein 9

CCM

Cerebral cavernous malformation

CCM1−/− iPSCs

CCM1 knockout iPSCs

CCM3−/−

CCM3 gene knockout

CNS

Central nervous system

Cre-LoxP

Cre recombinase/LoxP site system

CRISPR

Clustered regularly interspaced short palindromic repeats

DVAs

Developmental venous anomalies

ECs

Endothelial cells

EHT

Endothelial-to-hematopoietic transition

EndMT

Endothelial-to-mesenchymal transition

EPCs

Endothelial progenitor cells

FDA

Food and Drug Administration

FITC

Fluorescein isothiocyanate

hBBB

Human blood-brain barrier complex

hHOs

Human hepatocellular organs

hPSCs

Human pluripotent stem cells

HEC

Hematopoietic endothelial cell

IAHC

Intra-arterial hematopoietic cell

iBMECs

iPSC-derived brain microvascular endothelial cells

iPSC-ECs

iPSC-derived endothelial cells

iPSCs

Induced pluripotent stem cells

iVECs

iPSC-derived venous endothelial cells

MSCs

Mesenchymal stem cells

pre-HE

Pre-hematopoietic endothelial

PVL

Periventricular leukomalacia

scRNA-seq

Single-cell RNA sequencing

TEER

Trans-epithelial electrical resistance

TSC

Tuberous sclerosis complex

VEGF

Vascular endothelial growth factor

Author contributions

W.Q., S.J.J., and Z.X.S. contributed to the conceptualization of the study. W.Q. and S.J.J. were responsible for the methodology, while Z.X.S. and S.J.J. performed the formal analysis. Visualization was carried out by W.Q. W.Q. drafted the original manuscript, and X.X. and Y.T.B. contributed to manuscript review and revision. Supervision was provided by X.X. and Y.T.B. All authors contributed to manuscript preparation and approved the final version.

Funding

This work was supported by the National Natural Science Foundation of China (No.82401552) and Shandong Provincial Natural Science Foundation (No.ZR2024QH590).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.