TLNRD1-CCM2 interaction maintains endothelial actin networks through a KLF2/4-independent mechanism

Abstract

Actin cytoskeleton dysregulation contributes to vascular anomalies such as cerebral cavernous malformation (CCM). Talin rod domain containing-1 (TLNRD1) has been reported to interact with cerebral cavernous malformations 2 protein (CCM2), yet the downstream signaling remains debated, as previous studies have described opposite directions of Krueppel-like factor 2/4 (KLF2/4) changes after TLNRD1 depletion. Here, we combined biochemical analyses, structural modeling, transcriptomics, and single-cell network perturbation to examine the TLNRD1-CCM2 axis in endothelial cells. Coimmunoprecipitation and mass spectrometry confirmed the association between TLNRD1 and the CCM complex. Furthermore, protein docking predicted a stable TLNRD1-CCM2 interface (ΔG ≈ −50.36 kcal/mol) supported by prominent hydrogen bonds. Bulk RNA sequencing following TLNRD1 knockdown identified 677 differentially expressed genes, which were heavily enriched for actin cytoskeleton organization, with limited support for activation of the canonical MEKK3-KLF2/4 program. To assess KLF2/4 more directly, we analyzed human CCM single-cell RNA sequencing using scTenifoldKnk alongside complementary in vitro perturbations. Across these orthogonal analyses, KLF2 and KLF4 showed little to no consistent transcriptional alterations. Instead, TLNRD1 perturbation prominently altered endothelial F-actin stress fiber formation. Together, these data support a model in which TLNRD1 preferentially modulates endothelial cytoskeletal remodeling largely independent of overt KLF2/4 transcriptional shifts, helping to contextualize previous discrepancies and refining our understanding of its role in vascular biology.

Clinical Relevance

Current management of cerebral cavernous malformations (CCMs) remains limited by an incomplete understanding of the molecular basis of endothelial instability. This study identifies TLNRD1 as a CCM2-associated regulator of endothelial actin organization and cytoskeletal remodeling, while showing limited support for consistent KLF2/4 transcriptional changes. For clinicians, these findings add mechanistic context to a disease with few medical options and may help explain earlier conflicting experimental observations. Although direct clinical application is not immediate, the TLNRD1-CCM2 axis warrants further study as a biologically relevant pathway that may inform future translational research on cerebrovascular lesion progression.

Article Highlights.

• •Type of Research: In vitro study; modeling study
• •Key Findings: Talin rod domain containing-1 (TLNRD1) knockdown in human umbilical vein endothelial cells produced 677 differentially expressed genes enriched for actin/junctional programs. F-actin staining showed increased stress fiber formation after TLNRD1 depletion, whereas in vitro perturbation and human cerebral cavernous malformation single-cell virtual knockout showed limited support for consistent KLF2/4 transcript changes.
• •Take Home Message: TLNRD1 preferentially regulates endothelial cytoskeletal remodeling through its interaction with cerebral cavernous malformation 2, with limited evidence for uniform activation of the canonical KLF2/4 transcriptional axis.

Cerebral cavernous malformations (CCMs) are common vascular abnormalities of the central nervous system, with an estimated prevalence of approximately 0.16% to 0.5% in the general population.¹ Approximately 20% of cases show an autosomal dominant inheritance pattern.² Loss-of-function mutations in CCM1/Krev interaction trapped protein 1 (KRIT1), cerebral cavernous malformations 2 protein (CCM2)/OSM, and CCM3/programmed cell death protein 10 (PDCD10) are strongly implicated in CCM pathogenesis, and their encoded proteins form the CCM complex.³ Disruption of the CCM complex has been linked to endothelial cytoskeletal defects and altered gene expression programs, frequently accompanied by increased activity of the MEKK3-Krueppel-like factor 2/4 (KLF2/4) axis in several experimental settings.4, 5, 6 However, the extent to which KLF2/4 induction represents a uniform downstream response across CCM-relevant perturbations remains incompletely resolved.^7,8

Actin-binding proteins (ABPs) regulate endothelial barrier integrity and permeability through their effects on F-actin organization and cytoskeletal dynamics.⁹ Talin rod domain containing-1 (TLNRD1) is an F-actin-binding protein implicated in embryonic and cardiovascular development,^10,11 and genetic evidence has suggested TLNRD1 as a candidate shared factor across vascular diseases, including coronary artery disease and CCM.¹² TLNRD1 has been reported to interact with CCM2; however, the downstream consequences remain debated because TLNRD1 depletion has been associated with opposite directions of KLF2/4 expression changes in previous studies.^13,14 These discrepancies may reflect differences in endothelial context, perturbation strategy, or timing.

In this study, we integrated biochemical interaction analyses with bulk transcriptomics and single-cell network perturbation (virtual knockout; scTenifoldKnk) applied to human CCM single-cell RNA sequencing (scRNA-seq) data, aiming to assess TLNRD1-associated regulatory effects at both population and cell-resolved network levels, with particular attention to its prominent role in endothelial cytoskeletal remodeling vs the proposed involvement of the KLF2/4 axis.

Methods

Cell culture and transfection

HEK293T and HeLa cells were cultured in Dulbecco’s modified eagle medium supplemented with 10% (vol/vol) fetal bovine serum at 37 °C with 5% CO₂. Human umbilical vein endothelial cells (HUVECs) (American Type Culture Collection) and human cerebral microvascular endothelial cells were maintained in endothelial cell medium supplemented with 5% (vol/vol) fetal bovine serum. Transient transfection was performed using EZ-Trans (Life-ilab) according to the manufacturer’s instructions.

Plasmids and lentiviral transduction

TLNRD1, KRIT1, CCM2, and PDCD10 complementary DNAs (Genecopoeia) were subcloned into pCMV-FLAG or pCMV-Myc vectors. Short hairpin RNAs were cloned into the pLKO.1-eGFP-Puro vector, and TLNRD1/CCM2 overexpression constructs were cloned into the pLVX-FLAG-eGFP-Puro vector. Lentiviruses were produced in HEK293T cells using standard packaging, and supernatants were collected 48 hours post-transfection. HeLa cells and HUVECs were infected in the presence of LV-Enhance (DNABIO) and selected with puromycin (1 μg/mL).

Reverse transcription quantitative polymerase chain reaction

Total RNA was extracted using TRIzol (Thermo). Complementary DNA was synthesized from 1000 ng RNA using ABScript III RT Master Mix for quantitative polymerase chain reaction (qPCR) with gDNA Remover (ABclonal). qPCR was performed using SYBR Green Fast qPCR Mix (ABclonal), and relative gene expression was calculated by the ^ΔΔCt method with glyceraldehyde-3-phosphate dehydrogenase.

Western blotting

Cells were lysed in lysis buffer containing protease/phosphatase inhibitors (Cell Signaling Technology) and clarified by centrifugation at 12,000g for 15 min at 4 °C. For coimmunoprecipitation (co-IP), lysates were incubated with the indicated primary antibodies and protein A/G magnetic beads at 4 °C overnight. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes (Cytiva). The following antibodies were used: TLNRD1 (ARP67656_P050; Aviva), KRIT1 (R24814; Zen-Bio), CCM2 (A6544; ABclonal), PDCD10 (A17094; ABclonal), Actin (AC026; ABclonal), FLAG (F1804; Sigma), and Myc (M192-7; MBL).

Phalloidin staining

Cells cultured in multiwell plates were fixed, optionally washed, and stained with phalloidin-fluorescein isothiocyanate (Acmec) for 15 min at room temperature in the dark, according to the manufacturer's instructions. Images were acquired using a standard fluorescence microscope. F-actin fiber intensity per cell area and the number of fibers per cell were quantified using ImageJ/Fiji software.

FLAG affinity purification and liquid chromatography-tandem mass spectrometry

HeLa cells stably expressing FLAG-TLNRD1 were generated by lentiviral transduction and puromycin selection. Cells were lysed in BC100 buffer (20 mM Tris-Cl pH 7.9, 100 mM NaCl, 0.2 mM ethylenediaminetetraacetic acid, 20% glycerol) containing 0.2% Triton X-100 and protease inhibitors, clarified, and filtered (0.45 μm; Millipore). FLAG-tagged complexes were captured using anti-FLAG magnetic beads (MCE) and eluted with FLAG peptide (Sigma). Eluates were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (4%-20% gradient gel; Bio-Rad), stained with Coomassie (Abbkine), excised, and analyzed by liquid chromatography-tandem mass spectrometry.

Molecular docking and binding-free energy calculations

TLNRD1 (PDB: 6XZ4) and the AlphaFold2-predicted CCM2 structure were prepared and energy-minimized using OPLS_4 (Schrödinger). Protein-protein docking was performed with PIPER (Schrödinger), and the top-ranked clustered pose was selected for analysis. To assess interface hotspots, in silico alanine substitutions of TLNRD1 (Arg208, Glu182, and Arg189) was performed, followed by local conformational optimization. Binding-free energies (ΔG) and mutational impacts (ΔΔG) for wild-type and mutant complexes were calculated using Prime molecular mechanics/generalized born surface area (MM-GBSA) with the variable dielectric surface generalized born (VSGB) implicit solvent model.

scRNA-seq processing (GSE294555) and virtual knockout

scRNA-seq data (GSE294555; three controls and one CCM sample) were processed in Seurat using standard steps including quality control, normalization, highly variable gene selection, dimensionality reduction, clustering, and Uniform Manifold Approximation and Projection visualization; batch correction was performed using Harmony, and clusters were annotated based on canonical marker genes. Differential expression between CCM and control endothelial cells was determined using the FindMarkers function, and functional enrichment analysis was conducted using the clusterProfiler R package. scTenifoldKnk was used to perform virtual knockout of TLNRD1 and to derive network-level perturbation outputs from the scRNA-seq data.

Statistical analysis

Data are presented as mean ± standard deviation. One-way analysis of variance was performed in GraphPad Prism 10; P < .05 was considered statistically significant.

Results

TLNRD1 associates with the CCM complex and preferentially interacts with CCM2

FLAG-TLNRD1 complexes were affinity-purified from HeLa cells stably expressing full-length FLAG-TLNRD1 and analyzed by liquid chromatography-tandem mass spectrometry (Fig 1, A). KRIT1, CCM2, and PDCD10 were detected among the enriched proteins (Fig 1, B). A potential association between TLNRD1 and CCM complex subunits was also indicated in earlier interactome datasets (Fig 1, C).

Fig 1.

Identification of direct protein-protein interaction between Talin rod domain containing-1 (TLNRD1) and cerebral cavernous malformation (CCM2). A, Tandem affinity purification of TLNRD1 protein complexes was performed in HeLa cells stably expressing FLAG-TLNRD1. Associated proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and visualized by Coomassie blue (CB) staining. B, Quantification of the unique number of peptides identified by mass spectrometry analysis. C, Putative molecules interacting with TLNRD1 predicted by the STRING database. D and E, 293T cells were transfected with the indicated plasmids. Whole cell lysates (WCLs) were subjected to coimmunoprecipitation (Co-IP) using an anti-FLAG antibody. Immunoprecipitates were analyzed by Western blot (WB). F and G, WCLs from human umbilical vein endothelial cells (HUVECs) (F) and human cerebral microvascular endothelial cells (HCMECs) (G) were subjected to endogenous Co-IP with immunoglobulin G (IgG) or an anti-CCM2 antibody, followed by WB analysis to confirm the endogenous interaction. H, 293T cells were transfected with increasing amounts of FLAG-TLNRD1, and WCLs were analyzed by WB to assess the steady-state abundance of CCM complex components. I, Protein-protein docking model for the interaction betweenhuman TLNRD1 and CCM2. J, Putative interacting residues and distances outlining hydrogen bonds and salt bridges at the TLNRD1-CCM2 interface. K, In silico alanine substitution of TLNRD1 interfacial residues (Arg208, Glu182, and Arg189) and the corresponding binding-free energy (MM-GBSA) alterations. MW, molecular weight.

In HEK293T cells, FLAG-TLNRD1 immunoprecipitation recovered endogenous KRIT1, CCM2, and PDCD10 (Fig 1, D). In an exogenous coexpression setting, Myc-TLNRD1 coimmunoprecipitated with FLAG-CCM2, whereas co-IP with FLAG-KRIT1 or FLAG-PDCD10 was not detected under the same conditions (Fig 1, E), supporting a preferential association between TLNRD1 and CCM2 in this assay configuration. Importantly, endogenous co-IP using an anti-CCM2 antibody also recovered TLNRD1 in both HUVECs and human cerebral microvascular endothelial cells (Fig 1, F and G), extending this interaction to human endothelial cells, including a brain microvascular endothelial context.

Increasing FLAG-TLNRD1 expression did not measurably change KRIT1, CCM2, or PDCD10 protein abundance (Fig 1, H), suggesting that the observed association was not accompanied by an overt change in the steady-state abundance of core CCM complex components under these conditions. Protein-protein docking using TLNRD1 (PDB: 6XZ4) and the predicted CCM2 structure yielded a model featuring up to 10 hydrogen bonds and two salt bridges at the interface (Fig 1, I and J), with a calculated binding-free energy of ΔG ≈ −50.36 kcal/mol. Notably, TLNRD1 residues Arg208, Glu182, and Arg189 were predicted to participate in prominent interfacial contacts, each contributing hydrogen bond interactions, whereas Arg208 and Glu182 additionally formed salt bridges with CCM2. To further probe this putative interface in silico, we generated an alanine-substituted TLNRD1 mutant model (Arg208Ala/Glu182Ala/Arg189Ala) based on the wild-type docking interface. Compared with wild-type TLNRD1, the mutant model showed a less favorable MM-GBSA binding energy (−41.43 kcal/mol) and a positive ΔΔG (+8.93), consistent with weakened predicted binding to CCM2 (Fig 1, K). Together, these findings support a model in which TLNRD1 associates with the CCM complex and preferentially interacts with CCM2 through a structurally plausible interface.

TLNRD1 knockdown is associated with transcriptomic changes enriched for junction- and actin-related programs

Bulk RNA sequencing was performed in HUVECs with stable TLNRD1 knockdown. Principal component analysis showed separation between TLNRD1 knockdown and control samples (Fig 2, A). A total of 677 differentially expressed genes were identified (340 upregulated and 337 downregulated) (Fig 2, B).

Fig 2.

Talin rod domain containing-1 (TLNRD1) modulates endothelial transcriptomic programs involved in cytoskeletal organization. A, Principal component analysis (PCA) of RNA sequencing data from TLNRD1-knockdown (shTLNRD1-2) and nonspecific control (shNC) HUVECs (n = 3 biological replicates per group). B, Volcano plot revealing differentially expressed genes following TLNRD1 knockdown. C, Gene Set Enrichment Analysis (GSEA) evaluating motor protein-related signatures. D and E, Gene Ontology (GO) (D) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (E) enrichment analyses of the differentially expressed genes. F and G, Relative mRNA expression levels of TLNRD1, KRIT1, CCM2, PDCD10, KLF2, KLF4, ICAM1, CCL2, and OCLN were analyzed by reverse transcription quantitative polymerase chain reaction following TLNRD1 or CCM2 knockdown (F) and overexpression (G). Data are presented as mean ± standard deviation. (n = 4 independent biological replicates per group) ∗P < .05, ∗∗P < .01, ∗∗∗P < .001.

Gene Set Enrichment Analysis showed a nominal enrichment trend for the motor protein gene set (normalized enrichment score, 1.357; P = .031) (Fig 2, C). Consistent with a structural remodeling signature, Gene Ontology enrichment highlighted terms including cellular component biogenesis, cell-cell junction, actin cytoskeleton, regulation of GTPase activity, adherens junction, and tight junction (Fig 2, D). Kyoto Encyclopedia of Genes and Genomes enrichment included lipid and atherosclerosis, Ras signaling pathway, regulation of actin cytoskeleton, tight junction, ubiquitin-mediated proteolysis, and adherens junction (Fig 2, E). These analyses support an association between TLNRD1 depletion and endothelial actin- and junction-related transcriptional remodeling.

KLF2/4-associated downstream transcripts show limited and nonuniform changes upon TLNRD1 perturbation in vitro

Two independent short hairpin RNAs reduced TLNRD1 messenger RNA (mRNA) in HUVECs (Fig 2, F). Under the same conditions, CCM2 knockdown produced a stronger increase in KLF2 and KLF4 expression than TLNRD1 knockdown, whereas TLNRD1 depletion induced only modest or context-dependent changes in these transcripts (Fig 2, F). Similarly, ICAM1 and CCL2 showed limited changes after TLNRD1 knockdown compared with the more pronounced induction observed after CCM2 knockdown, whereas occludin did not show a consistent reduction across perturbation conditions (Fig 2, F).

TLNRD1 or CCM2 overexpression primarily increased expression of the corresponding transgene but did not produce a consistent reciprocal shift in KLF2, KLF4, ICAM1, CCL2, or occludin transcripts (Fig 2, G). These findings support a model in which TLNRD1 mainly influences endothelial actin- and junction-related programs without eliciting a robust and uniform KLF2/4-associated transcriptional response under the conditions tested.

TLNRD1 is elevated in CCM endothelial cells and preferentially regulates cytoskeletal remodeling

Based on canonical marker genes, we reanalyzed the integrated single-cell transcriptomic dataset (GSE294555) derived from human brain CCM lesions and matched controls and annotated the major cell populations present in CCM and control samples (Fig 3, A and B). Within the endothelial lineage, the CCM sample exhibited higher expression of TLNRD1, KLF2, and KLF4 than the control samples, whereas this pattern was less evident across other annotated cell types (Fig 3, C). To examine whether TLNRD1 is functionally coupled to the canonical CCM-associated KLF2/4 transcriptional program, we performed a scTenifoldKnk-based in silico perturbation analysis. Under the preset threshold, KLF2 and KLF4 were not identified among the top perturbed genes following virtual knockout of TLNRD1; instead, DKK2 and ITLN1 represented the most prominently perturbed genes (Fig 3, D and E).

Fig 3.

Single-cell virtual knockout and in vitro assays reveal that Talin rod domain containing-1 (TLNRD1) regulates endothelial cytoskeletal remodeling independently of prominent KLF2/4 shifts. A and B, The integrated single-cell RNA sequencing dataset (GSE294555) visualized by Uniform Manifold Approximation and Projection (UMAP) to show clustering results (A) and annotation of major cell populations (B). C, Cell type-specific violin plots of TLNRD1, KLF2, and KLF4 expression between CCM and control samples. D, Top perturbed genes ranked by scTenifoldKnk following single-cell virtual knockout of TLNRD1. E, Scatter plot illustrating the global network-level transcriptomic perturbation landscape upon TLNRD1 virtual knockout, with top divergent genes highlighted. F, Functional enrichment analysis of differentially expressed genes between CCM and control endothelial cells in the single-cell dataset. G, Representative immunofluorescence images of F-actin stress fibers (phalloidin, green) and nuclei [4′,6-diamidino-2-phenylindole (DAPI), blue] in human umbilical vein endothelialcells (HUVECs) following the indicated short hairpin RNA-mediated knockdowns (top) or overexpression (bottom). Scale bar = 50 μm. H-K, Quantification of F-actin fiber intensity per cell area (H and J) and the number of fibers per cell (I and K) from the corresponding fluorescence images. Data are presented as mean ± standard deviation. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001.

We next performed differential expression analysis between CCM and control endothelial cells in the single-cell dataset, followed by functional enrichment analysis of the resulting differentially expressed genes. Genes upregulated in CCM endothelial cells were enriched in biological processes related to small GTPase-mediated signal transduction, cell-substrate adhesion, actin filament bundle assembly, and cell-cell junction organization, whereas genes downregulated in CCM endothelial cells were enriched in response to mechanical stimulus, endothelial cell migration, actomyosin structure organization, and cell-matrix adhesion (Fig 3, F).

To test whether TLNRD1 directly influences these structural phenotypes in endothelial cells, phalloidin staining was performed and showed enhanced F-actin fiber formation after TLNRD1 knockdown, as reflected by increased fiber intensity per cell area and increased numbers of fibers per cell; a similar phenotype was observed following CCM2 knockdown (Fig 3, G-I). Conversely, TLNRD1 overexpression reduced stress fiber formation, particularly in the number of fibers per cell, whereas CCM2 overexpression produced a broader suppressive effect on F-actin organization (Fig 3, G, J, and K). These data indicate that TLNRD1 is more closely associated with endothelial cytoskeletal remodeling than with strong activation of the canonical CCM/KLF2/4 transcriptional network.

Discussion

Mutations in three autosomal genes, CCM1/KRIT1, CCM2/OSM, and CCM3/PDCD10, have been identified as being highly associated with CCM disease.¹⁵ These three proteins encoded by the CCM family could form a heterotrimer, with CCM2 playing a crucial role by interacting with both KRIT1 and PDCD10.¹⁶ However, due to the complexity of vascular cell homeostasis and endothelial regulation, effective therapy for CCM remains elusive, necessitating a more comprehensive understanding of the molecular signaling pathways and regulatory factors involved. In this study, our results support TLNRD1 as a candidate CCM-associated regulator that can specifically interact with CCM2 and is linked to cytoskeleton- and junction-related endothelial programs, thereby providing a plausible molecular connection between CCM core components and endothelial cytoskeletal homeostasis.

TLNRD1 features a crucial four-helix bundle domain that mediates the formation of symmetric antiparallel dimers, which has been previously reported as indispensable for its physiological functions.¹⁰ Interestingly, the hydrogen bonding and salt bridge binding sites on TLNRD1 displayed by molecular docking are all precisely located within the four-helix bundle domain, suggesting that this may be the core domain for TLNRD1 to bind to CCM2. Indeed, our in silico MM-GBSA calculations further support the energetic importance of these interfacial residues. Although computational analyses are predictive, these findings offer a testable structural hypothesis that can be further examined by interface-directed mutagenesis and biochemical binding assays.

The endothelial barrier of the vascular system is crucial for separating blood flow from underlying tissues, which is formed by tight junctions and adhesion junctions that create intercellular contacts.¹⁷ Tight junctions and adhesion junctions are integral membrane structures that connect to the actin cytoskeleton through various junctional molecules.^18,19 Therefore, the actin cytoskeleton plays a vital role in regulating the stability of endothelial cell contacts and overall structural integrity. The role of ABPs in modulating endothelial stability has long been underestimated, but recent studies have indicated that ABPs could modulate actin remodeling or function through various signaling transduction mechanisms.¹⁰ Consistent with this framework, our transcriptomic analyses indicate that TLNRD1 perturbation is associated with the upregulation of biological processes and molecular functions related to cell adhesion, cell junctions, and integrin binding in endothelial cells, suggesting that TLNRD1 may reshape a gene regulatory network relevant to cytoskeletal architecture and cellular adhesion.

Mechanistically, the CCM complex exerts a strong suppressive effect on MEKK3 under physiological conditions, which also leads to the suppression of KLF2/4 as downstream factors.²⁰ Otten et al²¹ further revealed this mechanism through gene knockout mouse models, demonstrating that the loss of any single CCM member in neonatal mouse brain endothelial cells ultimately leads to the development of CCM, confirming that the constitutive activation of the MEKK3-KLF2/4 signaling pathway is directly associated with the disruption of the CCM complex.

In our study, although knocking down CCM2 relieved this inhibitory effect and upregulated KLF2/4, TLNRD1 knockdown did not produce similar transcriptional shifts. Furthermore, this phenomenon was not observed in TLNRD1 or CCM2 overexpression groups. This implies that although CCM2 exerts a substantial inhibitory constraint on the MEKK3-KLF2/4 pathway under physiological conditions, TLNRD1 operates outside this specific transcriptional axis.

Notably, the in silico depletion of TLNRD1 resulted in minimal global transcriptomic perturbations, with no significant alterations observed in KLF2/4 or major cytoskeletal gene expression. This lack of a broad transcriptional response strongly supports the premise that TLNRD1 functions primarily as a structural or scaffolding protein, modulating actin dynamics directly, as evidenced by our robust F-actin stress fiber phenotypes, rather than a direct transcriptional regulator. Consequently, its role in modulating endothelial junctional dynamics likely occurs at the post-translational level rather than by altering mRNA abundance. However, as our study primarily evaluated steady-state mRNA levels, we cannot entirely rule out temporally restricted influences, cell subtype-specific effects, or post-transcriptional regulation of KLF2/4. Furthermore, our single-cell virtual knockout analysis, based on a restricted dataset, is inherently exploratory and hypothesis-generating.

TLNRD1’s role within or alongside the CCM complex may therefore be largely uncoupled from this specific MEKK3-KLF2/4 transcriptional axis. Notably, two earlier studies reported opposite directions of KLF2/4 changes upon TLNRD1 depletion-one describing KLF2/4 upregulation and the other reporting downregulation.^13,14 In line with our in vitro findings, our single-cell virtual knockout analysis did not reveal a prominent regulatory dependence between TLNRD1 and KLF2/KLF4, supporting a context-dependent or parallel role of TLNRD1 in endothelial regulation rather than a direct upstream driver of KLF2/4.

Conclusions

In conclusion, our data establish TLNRD1 as a CCM2-interacting protein that modulates endothelial cytoskeletal programs without consistent KLF2/4 induction. Recognizing this uncoupled pathway refines the molecular model of CCM, suggesting that cytoskeletal and structural components represent distinct pathogenic branches contributing to vascular anomalies.

Funding

This work was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project (Grant No. 2024ZD0537800) and National Natural Science Foundation of China (Grant Nos. 82270415 and U24A20651).

Disclosures

None.