Anisodine hydrobromide targets matk and prevents delayed rtPA thrombolysis-induced vasogenic cerebral edema in ischemic stroke

Abstract

Vasogenic cerebral edema is a severe complication of delayed thrombolysis for ischemic stroke, for which no pharmacological treatment exists. Anisodine hydrobromide (Ani), an alkaloid used clinically in China for vascular disorders, is investigated for its potential to mitigate this condition. Here we show that Ani treatment improves survival and neurological function in a mouse model of delayed rtPA-induced cerebral edema by preserving the integrity of the blood-brain barrier. Utilizing proteomics and microarray screening, we identify megakaryocyte-associated tyrosine kinase (Matk) as a direct target of Ani. We demonstrate that Ani binding stabilizes Matk, preventing its degradation and suppressing the activation of Src kinase. This inhibition consequently blocks the dual paracellular and transcellular leakage pathways that drive vasogenic edema. Our findings reveal the Matk-Src signaling axis as a therapeutic target and support Ani as a promising clinical candidate for preventing post-thrombolytic complications in stroke management.

Ischemic stroke treatment can be complicated by brain edema if thrombolytic drugs are given too late. Here, the authors show that anisodine hydrobromide mitigates this dangerous edema in mice by targeting the Matk-Src pathway to protect blood vessels in the brain

Introduction

Stroke, a cerebrovascular disorder caused by vascular occlusion or rupture, results in disrupted cerebral blood flow and subsequent ischemic or hemorrhagic brain injury¹. As the second leading cause of death worldwide (after ischemic heart disease) and the primary cause of mortality in China, stroke represents a significant global health burden². Ischemic stroke, accounting for 60%–80% of cases, is the predominant subtype³. Recombinant tissue plasminogen activator (rtPA) remains the gold-standard thrombolytic therapy for acute ischemic stroke, with European and American guidelines recommending administration within 4.5 hours of symptom onset^4,5. However, delayed rtPA treatment (beyond this therapeutic window) is common in clinical practice and is associated with increased blood-brain barrier (BBB) disruption, vasogenic cerebral edema (occurring in 10%–40% of patients), and mortality⁶. Emerging strategies aim to mitigate rtPA thrombolysis-induced BBB injury. Sublingual edaravone dexborneol improves functional recovery by reducing oxidative stress and inflammation, while also preserving BBB integrity in rtPA-thrombolyzed acute ischemic stroke patients⁷. Similarly, fingolimod exerts BBB protection through anti-inflammatory and antioxidant mechanisms, thereby attenuating rtPA thrombolysis-related vascular damage⁸. However, further Phase III clinical trials are required to confirm their therapeutic efficacy and safety profiles.

The integrity of the BBB is maintained through complicated interactions among tight junctions, adherens junctions, and the basement membrane of cerebral endothelial cells, which collectively regulate vascular permeability and neurovascular homeostasis⁹. Previous studies have demonstrated that in delayed rtPA-thrombolyzed ischemic stroke mice, the expression of tight junction proteins (Occludin, Claudin-5, JAM-1, and ZO-1) and the adherens junction protein VE-cadherin is significantly downregulated^10,11. This disruption destabilizes the adherens junction complex, widens intercellular gaps in microvascular endothelial cells, and ultimately facilitates paracellular leakage¹². In addition to paracellular leakage, Caveolin-1-mediated transcellular leakage plays a critical role in vasogenic cerebral edema¹³. Delayed rtPA thrombolysis activates Src via phosphorylation at Tyr419, which in turn phosphorylates Caveolin-1 at Tyr14. This cascade promotes the formation and endocytosis of endothelial plasmalemmal vesicles, exacerbating transcellular leakage in microvascular endothelial cells^14,15 Therefore, targeting both paracellular and transcellular leakage pathways represents a promising therapeutic strategy for mitigating delayed rtPA thrombolysis-associated cerebral edema.

Anisodine hydrobromide (Ani), an alkaloid extracted from traditional Chinese herb Anisodus tanguticus, is clinically approved by the China Food and Drug Administration (CFDA; Approval No. H51023360) for treating vascular headaches, retinal vascular spasms, and ischemic optic neuritis¹⁶. Emerging clinical evidence indicates that Ani safely improves outcomes in ischemic stroke patients by reducing neurological deficits, promoting functional recovery, and enhancing collateral circulation in ischemic regions^17,18. Preclinical studies further demonstrate that, Ani could cross the BBB in rats¹⁹ improve cerebral blood flow in chronic cerebral hypoperfusion models²⁰, and enhance neuronal survival during cerebral ischemia through calcium influx modulation and ROS reduction²¹. However, its potential therapeutic effects on delayed rtPA thrombolysis-induced vasogenic cerebral edema and the molecular mechanisms involved remain unexplored.

The present study found that Ani inhibited delayed rtPA thrombolysis-induced vasogenic cerebral edema in ischemic stroke mice. Using proteomics, HuProt™ microarray screening, molecular docking, dynamics simulations, SPR, etc., we found that Ani bound to megakaryocyte-associated tyrosine kinase (Matk) and inhibited its ubiquitin-mediated degradation. Through in vitro (Matk knockdown/overexpression) and in vivo (AAV-BI30-Tie1-mediated endothelial-specific modulation) experiments, we demonstrated that Ani suppressed Src activation and subsequent VE-cadherin/Caveolin-1 phosphorylation via Matk, blocking paracellular and transcellular vascular leakage to protect against delayed rtPA thrombolysis-induced cerebral edema.

Result

Ani increases survival rate, and mitigates neurological deficits and vasogenic cerebral edema in rtPA-thrombolyzed ischemic stroke mice

Figure 1A displays Ani’s chemical structure and molecular formula, while Fig. 1B outlines the experimental design. To determine the optimal treatment window, we first measured the plasma MDA levels, a well-established biomarker of cerebral ischemia-reperfusion injury²² at 4.5, 5.5, 6.5, 8.5, 12.5, and 28.5 h post-thrombosis (corresponding to 0, 1, 2, 4, 8, and 24 h post-thrombolysis). MDA levels peaked at 5.5, 6.5, and 8.5 h post-thrombosis (1, 2, and 4 h post-thrombolysis), indicating a period of significant oxidative damage during this early phase (Supplementary Fig. 1A). In parallel, a pharmacokinetic study of Ani revealed a relatively short half-life (t_1/2 ≈ 15 min) (Supplementary Fig. 1B). Based on the convergence of peak oxidative injury and Ani’s rapid clearance, we selected three time points (5.5, 6.5, and 8.5 h post‑thrombosis) to systematically evaluate the therapeutic window of Ani. Mice were administered with Ani at these time points at low (0.6 mg/kg), medium (1.2 mg/kg), or high (2.4 mg/kg) doses. Survival and neurological behavioral scores were then analyzed across groups to assess treatment efficacy. Compared to Sham group, the Thrombus+rtPA group showed significantly reduced 7-day survival, which was improved in Thrombus+rtPA+5.5 h(L), +5.5 h(M), +6.5 h(M) and +6.5 h(H) groups (Fig. 1C). Neurological evaluation revealed improved NES scores in Thrombus+rtPA+5.5 h(H) and +6.5 h(H) groups (Fig. 1D), and enhanced mNSS scores in Thrombus+rtPA+5.5 h(H), +6.5 h(H), and +8.5 h(H) groups versus Thrombus+rtPA (Fig. 1E). Laser speckle flowmetry demonstrated comparable cerebral perfusion improvements at 24 h post-thrombolysis in all treatment groups versus Thrombus (Fig. 1F–G). Evans blue assays showed reduced vascular leakage in Thrombus+rtPA+5.5 h(M), +5.5 h(H), +6.5 h(M), +6.5 h(H), and +8.5 h(H) groups (Fig. 1H, I). Brain water content significantly decreased in Thrombus+rtPA+5.5 h(H) and +6.5 h(H) groups (Fig. 1J). Notably, no significant differences were observed between the Sham+Ani group and the Sham group in the above parameters. Based on these findings, the Thrombus+rtPA+6.5 h(H) group showed optimal efficacy and was selected for subsequent studies.

Fig. 1. Ani increases survival rate, and mitigates neurological deficits and vasogenic cerebral edema in rtPA-thrombolyzed ischemic stroke mice.

A Ani’s chemical structure and molecular formula. B Experimental workflow and group assignments in this study. C Survival rates of mice in each group (n = 15 biological replicates, log-rank test). D The NES score of mice in each group (n = 10 biological replicates). E The mNSS score of mice in each group (n = 10 biological replicates). F, G Representative images and statistical analysis of CBF at baseline, 4.5 h, and 28.5 h in each group (n = 6 biological replicates, *p < 0.05 vs. Sham group; #p < 0.05 vs. Thrombus group). H, I Representative images and statistical analysis of Evans blue extravasation of brain tissue homogenate (n = 10 biological replicates). J Quantitative analysis of cerebral water content in each group (n = 10 biological replicates). Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. *p < 0.05 vs. Sham group; #p < 0.05 vs. Thrombus+rtPA group. Source data are provided as a Source Data file.

Given the well-documented sex differences in ischemic and thrombolytic responses²³, we also conducted validation experiments in female mice to assess the generalizability of our findings. Results in female mice were similar to those observed in males, demonstrating comparable therapeutic effects of Ani (Supplementary Fig. 1C-J).

Ani inhibits rtPA thrombolysis-induced cerebral microvascular leakage in ischemic stroke mice

We used dynamic visualization technique to monitor FITC-dextran leakage from the cerebral microvessels to reflect the microvascular permeability. We observed that rtPA thrombolysis significantly increased FITC-dextran leakage compared to Sham group, an effect that was effectively inhibited by Ani treatment (Fig. 2A, B). We subsequently investigated its potential to regulate key proteins involved in paracellular and transcellular leakage pathways. In the Thrombus+rtPA group, the marked reduction in ZO-1, JAM-1, and Occludin expression (Fig. 2C and Supplementary Fig. 2A-C) along with fragmented Occludin immunofluorescence (Fig. 2D) were all reversed by Ani. Similarly, while VE-cadherin and α-catenin (but not β-catenin) were significantly downregulated in Thrombus+rtPA group (Fig. 2E and Supplementary Fig. 2E–G), Ani restored their expression and improved VE-cadherin continuity in immunofluorescence (Fig. 2F). Ani significantly reduced the rtPA thrombolysis-induced upregulation of p-Caveolin-1 (Tyr14) and p-Src (Tyr419) without affecting total protein levels (Fig. 2G, H, Supplementary Fig. 2H, I). These findings were further supported by transmission electron microscopy, which showed that while rtPA thrombolysis widened endothelial junction gaps and increased vesicle count in endothelial cells, Ani treatment attenuated these ultrastructural changes (Fig. 2I-K).

Fig. 2. Ani inhibits rtPA thrombolysis-induced cerebral microvascular leakage in ischemic stroke mice.

A, B Representative images and statistical analysis of FITC-dextran leakage through mouse cerebral venules (n = 6 biological replicates). The fluorescence intensity was represented as the ratio of I (interstitial area) to V (venules), scale bar = 50 μm. C Representative western blotting images of ZO-1, JAM-1, Occludin and Claudin-5 protein of mouse brain (n = 6 biological replicates). D Representative immunofluorescence staining of Occludin in mouse brain (n = 3 biological replicates). E Representative western blotting images of VE-cadherin, α-catenin and β-catenin protein of mouse brain (n = 6 biological replicates). F Representative immunofluorescence staining of VE-cadherin in mouse brain (n = 3 biological replicates). G Representative western blotting images of p-Caveolin-1 (Tyr14), Caveolin-1, p-Src (Tyr419) and t-Src protein of mouse brain (n = 6 biological replicates). H Representative immunofluorescence staining of p-Src (Tyr419) and p-Caveolin-1 (Tyr14) in mouse brain (n = 3 biological replicates). I–K Representative electron microscopy images and statistical analysis of cerebral microvessels in each group of mice (n = 3 biological replicates). Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. *p < 0.05 vs. Sham group; #p < 0.05 vs. Thrombus+rtPA group. Source data are provided as a Source Data file.

Integration of proteomic analysis and HuProt™ human proteome microarray identifies Matk as a potential target of Ani

To investigate the molecular mechanism of Ani’s protection against rtPA thrombolysis-induced BBB disruption, we performed proteomic analysis on right hemisphere cortex tissues. Comparative analysis (Fold Change > 1.2, p < 0.05) revealed 105 proteins downregulated in the Thrombus+rtPA group versus Sham that were upregulated by Ani treatment (Fig. 3A and Supplementary Table 1). GO analysis identified 40 of these proteins as enriched in small molecule binding (Fig. 3B and Supplementary Table 2), suggesting potential interactions with ions, metabolites, drugs, or cofactors. Conversely, 120 proteins (Supplementary Table 3) upregulated in Thrombus+rtPA were downregulated by Ani, none of which showed small molecule binding enrichment (Fig. 3C). For direct target identification, biotin-labeled Ani (Ani-biotin, Fig. 3D) was screened against the HuProt™ human proteome microarray (Supplementary Fig. 2J), identifying 729 Ani-binding proteins (Supplementary Table 4), with 239 showing small molecule binding capacity in GO analysis (Fig. 3E and Supplementary Table 5). Intersection analysis between the 40 proteomic candidates and 239 HuProt™ hits yielded 5 potential targets (Fig. 3F), among which Matk demonstrated the highest binding affinity in both the proteome microarray and molecular docking analysis (Fig. 3G-I).

Fig. 3. Proteomic analysis and HuProt™ microarray identify Matk as a potential target of Ani.

A Heat map of differential genes in proteomic analysis. B, C GO enrichment analysis of the 105 and 120 differential genes in proteomic analysis. D Ani-biotin’s chemical structure and molecular formula. E GO enrichment analysis of 729 positive proteins from the HuProt™. F Venn diagram of the 40 proteomic candidates and 239 HuProt™ hits from the small molecule binding category. G Ranking of fluorescence intensity fold change of the five proteins on the HuProt™. H The affinity scoring results of the five proteins. I Fluorescence position of Matk on the microarray. The left image presents a full chip scan of the Ani, while the right image offers a magnified view of Block 1, in which Matk localizes. The red arrow marks the positive control (Biotin-BSA), the pink arrow indicates the negative control (BSA), and the yellow arrow highlights Matk. Source data are provided as a Source Data file.

Ani directly interacts with Matk and inhibits its ubiquitin-mediated degradation

Through integrated computational and biochemical approaches, we validated the specific interaction between Ani and Matk. Molecular docking demonstrated stable hydrogen bonding between Ani and Matk residue Glu220 (Fig. 4A), while 100-ns MD simulations confirmed interaction stabilization after 10 ns (Fig. 4B). Hydrogen bond occupancy analysis was performed to evaluate the stability of interactions between the ligand and the binding site. Ani formed stable hydrogen bonds with residues Trp122 (occupancy: 61%), Ser197 (occupancy: 61%), and Glu194 (occupancy: 65%) over the 100-ns MD trajectory (Fig. 4C). To assess the enthalpic contribution to ligand binding, molecular mechanics generalized Born surface area (MM/GBSA) calculations were performed. The estimated binding enthalpy (ΔH) was -13.44 kcal/mol, indicating a favorable interaction from an enthalpic perspective. The binding enthalpy was predominantly driven by van der Waals (-23.35 kcal/mol) and electrostatic interactions (-31.59 kcal/mol). In contrast, the polar solvation energy (+44.79 kcal/mol) had a destabilizing effect, whereas non-polar solvation (-3.29 kcal/mol) contributed modestly to stabilization (Fig. 4D). To further elucidate the enthalpic determinants of binding, per-residue energy decomposition analysis was conducted. The results revealed that Leu296 exhibited the strongest contribution (-2.19 kcal/mol), followed by Met190 (-1.23 kcal/mol), Ser197 (-0.93 kcal/mol), Trp122 (-0.91 kcal/mol), and His297 (-0.58 kcal/mol) (Fig. 4E). These residues are spatially close to the ligand and likely engage in stabilizing interactions such as hydrophobic contacts and hydrogen bonding. Computational alanine scanning was performed for four residues (Trp122, Met190, Ser197 and Leu296), the results revealed that mutations at residues Trp122, Ser197 and Met190 led to ΔH increases of 1.0–1.5 kcal/mol, indicating their critical roles in ligand binding. In contrast, mutation at residue Leu296 showed a smaller effect (ΔH = 0.4 kcal/mol), suggesting a limited contribution (Fig. 4F). SPR analysis confirmed the direct binding of Ani to Matk, with a dissociation constant (Kd) of 2.12 × 10⁻⁶ M (Fig. 4G).

Fig. 4. Ani directly interacts with Matk and inhibits its ubiquitin-mediated degradation.

A Molecular docking analysis of Ani and Matk. B The results of the 100-ns MD simulations between Ani and Matk. C The outcomes of hydrogen bond occupancy analysis. D The outcomes of molecular mechanics generalized Born surface area (MM/GBSA) calculations. E The outcomes of per-residue energy decomposition analysis. F The outcomes of computational alanine scanning. G SPR analysis of Ani and Matk. H Pull-down analysis of Ani and Matk (n = 1). I CETSA assay (n = 3 biological replicates). J Representative western blotting images and statistical analysis of Matk in mice (n = 6 biological replicates). K Representative western blotting images and statistical analysis of Matk in HCMECs (n = 6 biological replicates). L Representative immunofluorescence staining of Matk in HCMECs (n = 3 biological replicates). M, N Representative western blotting images and statistical analysis of Matk in HCMECs (n = 6 biological replicates). O Result of Matk ubiquitination analysis (n = 1). Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. *p < 0.05 vs. Sham group; #p < 0.05 vs. Thrombus+rtPA group. Source data are provided as a Source Data file.

The interaction between Ani and Matk was further validated in HCMECs. Cellular uptake assays demonstrated that biotinylated Ani efficiently entered HCMECs, as visualized by strong Streptavidin-FITC staining (Supplementary Fig. 2K, L). This uptake was substantially attenuated by pretreatment with either a pharmacological OCT2 inhibitor Cimetidine or OCT2‑specific siRNA, consistent with previous reports identifying Ani as an OCT2 substrate²⁴. Direct molecular interaction was confirmed by pull‑down assay, in which biotinylated Ani selectively precipitated Matk; this interaction was competitively inhibited by an excess of non‑biotinylated Ani (Fig. 4H). Furthermore, CETSA showed that Ani significantly stabilized Matk against thermal degradation (Fig. 4I). In the proteomic analysis, Matk was downregulated in the Thrombus+rtPA group, and was restored by Ani treatment (Fig. 3A). This was further confirmed by western blot (Fig. 4J). Consistent with these findings, OGD/R markedly reduced Matk protein levels in HCMECs, and this reduction was reversed by Ani (Fig. 4K, L). Notably, qPCR analysis revealed no significant changes in Matk mRNA levels across experimental groups in both in vivo and in vitro settings (Supplementary Fig. 2M, N), suggesting post-translational regulation. Mechanistic investigations revealed that OGD/R-induced Matk degradation was selectively inhibited by the proteasome inhibitor MG132 but not by the lysosomal inhibitor chloroquine (CQ) (Fig. 4M, N), implicating the ubiquitin-proteasome pathway in this regulatory process. Importantly, we observed concomitant increase in Matk ubiquitination under OGD/R conditions, which was substantially suppressed by Ani treatment (Fig. 4O).

Matk mediates Ani’s protection of endothelial barrier function via Src signaling regulation in HCMECs

Matk, a CSK family tyrosine kinase, phosphorylates Src at Tyr530, thereby inhibiting Src activation by preventing its phosphorylation at Tyr419²⁵. Src kinase activation induces VE-cadherin phosphorylation at Tyr658, leading to dissociation of VE-cadherin from β-catenin, disruption of adherens junctions, endothelial gap formation, and subsequent paracellular leakage²⁶. Additionally, Src activation phosphorylates Caveolin-1 at Tyr14, promoting plasmalemmal vesicle formation and endocytosis in endothelial cells, which exacerbates transcellular leakage^13,14. Thus, Ani may mitigate both paracellular and transcellular leakage through Matk-mediated Src inhibition.

To test this hypothesis, we established an OGD/R model in HCMECs. Ani treatment significantly reduced FITC-dextran leakage and increased TEER in OGD/R-induced HCMECs (Fig. 5A, B). Subsequently, we analyzed Src activation status by assessing phosphorylation at its Tyr419 (activation site) and Tyr530 (inhibitory site) and measuring its activity. Western blot analysis revealed that OGD/R downregulated p-Src (Tyr530), VE-cadherin, and ZO-1 while upregulating p-Src (Tyr419), p-VE-cadherin (Tyr658), and p-Caveolin-1 (Tyr14) compared to control - changes that were all normalized by Ani (Fig. 5C and Supplementary Fig. 2O–U). Similar results were observed in the mouse brain tissues (Figs. 2E–G, 5D and Supplementary Fig. 2V–W). Immunofluorescence further confirmed Ani’s suppression of p-Src (Tyr419) (Fig. 5E). Critically, Src activity assay in HCMECs revealed that OGD/R significantly increased Src kinase activity in HCMECs, which was inhibited by Ani treatment (Supplementary Fig. 2X). Consistently, rtPA thrombolysis markedly enhanced Src kinase activity in mouse brain tissues, an effect also suppressed by Ani (Supplementary Fig. 2Y). Collectively, these data suggested that Ani preserved vascular integrity by inhibiting Src axis to coordinately suppress both paracellular and transcellular leakage pathways via Matk.

Fig. 5. Matk Mediates Ani’s Protection of Endothelial Barrier Function via Src Signaling Regulation in HCMECs.

A, F, J the impact of Ani on FITC-dextran leakage of HCMECs (n = 5 biological replicates). B, G, K the impact of Ani on TEER of HCMECs (n = 5 biological replicates). C, I, M Representative western blotting images of Matk, t-Src, p-Src (Tyr419), p-Src (Tyr530), VE-cadherin, p-VE-cadherin (Tyr658), p-Caveolin-1 (Tyr14), ZO-1 protein of HCMECs (n = 6 biological replicates). D Representative western blotting images of p-Src (Tyr530) and p-VE-cadherin (Tyr658) in mice (n = 6 biological replicates). E Representative immunofluorescence staining of p-Src (Tyr419) in HCMECs (n = 3 biological replicates). H, L Representative immunofluorescence staining of Matk, p-Src (Tyr419) and VE-cadherin in HCMECs (n = 3 biological replicates). Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. *p < 0.05 vs. Control, Control+NC or Control+siNC group; #p < 0.05 vs. OGD/R, OGD/R + NC or OGD/R+siNC group; †p < 0.05 vs. OGD/R+Ani+siNC group. Source data are provided as a Source Data file.

To further confirm Matk’s role in mediating Ani’s effect, we performed Matk knockdown in HCMECs (Supplementary Fig. 3A). This genetic intervention completely abolished Ani’s protective effects, as evidenced by the persistence of increased FITC-dextran leakage, reduced TEER (Fig. 5F, G). Furthermore, it significantly reversed Ani’s inhibitory effects on Src activation, and the phosphorylation of its downstream targets VE-cadherin (Tyr658) and Caveolin-1 (Tyr14) (Fig. 5H, I and Supplementary Fig. 3B-I). Conversely, Matk overexpression in HCMECs recapitulated Ani’s therapeutic effects, significantly improving barrier integrity (Fig. 5J, K) and inhibiting Src signaling (Fig. 5L, M and Supplementary Fig. 3J-Q). Notably, the combination of Matk overexpression and Ani treatment did not produce additive effects, suggesting they function through convergent molecular pathways.

Endothelial-specific Matk modulation reveals its essential role in Ani-mediated protection against rtPA thrombolysis-induced vasogenic cerebral edema

Endothelial-specific Matk knockdown using AAV-shMatk revealed the critical role of Matk in Ani-mediated protection. Compared to Thrombus+rtPA+AAV-NC group, the Thrombus+rtPA+Ani+AAV-NC group showed significantly reduced FITC-dextran leakage, while Thrombus+rtPA+Ani+AAV-shMatk mice exhibited no such improvement (Fig. 6A, B). Evans blue assays confirmed these findings (Fig. 6C, D).

Fig. 6. Endothelial cell-specific Matk knockdown prevents Ani’s effect on microvascular extravasation after rtPA thrombolysis in ischemic stroke mice.

A, B Representative images and statistical analysis of FITC-dextran leakage through mouse cerebral venules (n = 4 biological replicates), scale bar = 50 μm. C, D Representative images and statistical analysis of Evans blue extravasation (n = 4 biological replicates). E Representative western blotting images of Matk, t-Src, p-Src (Tyr419), p-Src (Tyr530), VE-cadherin, p-VE-cadherin (Tyr658), p-Caveolin-1 (Tyr14), and ZO-1 protein of isolated mouse brain endothelial cells (n = 6 biological replicates). F Representative immunofluorescence staining of Matk, p-Src (Tyr419), p-Caveolin-1 (Tyr14) and VE-cadherin in mouse brain (n = 3 biological replicates). G–I Representative electron microscopy images and statistical analysis of cerebral microvessels in each group of mice (n = 3 biological replicates). Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. *p < 0.05 vs. Sham+AAV-NC group; #p < 0.05 vs. Thrombus+rtPA+AAV-NC group; †p < 0.05 vs. Thrombus+rtPA+Ani+AAV-NC group. Source data are provided as a Source Data file.

To directly validate the proposed mechanism, we isolated brain endothelial cells from the relevant experimental groups and performed Western blot (Supplementary Fig. 3R). The results demonstrated that Thrombus+rtPA+AAV-NC groups displayed decreased Matk, p-Src (Tyr530), VE-cadherin, and ZO-1 alongside increased p-Src (Tyr419), p-VE-cadherin (Tyr658), and p-Caveolin-1 (Tyr14) versus Sham+AAV-NC - changes reversed by Ani treatment unless Matk was knocked down (Fig. 6E and Supplementary Fig. 4A–H). Immunofluorescence verified Matk, p-Src (Tyr419), p-Caveolin-1 (Tyr14), VE-cadherin expression patterns and showed Matk knockdown prevented Ani’s restoration of continuous VE-cadherin expression (Fig. 6F). Electron microscopy confirmed that Matk knockdown abolished Ani’s protective effect on BBB integrity (Fig. 6G–I).

Conversely, endothelial-specific Matk overexpression via AAV-Matk significantly attenuated both FITC-dextran and Evans blue leakage compared to Thrombus+rtPA+AAV-NC group (Fig. 7A–D). Matk overexpression upregulated Matk, VE-cadherin, ZO-1, and p-Src (Tyr530) while downregulating p-Src (Tyr419), p-VE-cadherin (Tyr658), and p-Caveolin-1 (Tyr14) in isolated brain endothelial cells (Fig. 7E and Supplementary Fig. 4I–P). Immunofluorescence analysis further confirmed the expression of Matk, p-Src (Tyr419), p-Caveolin-1 (Tyr14) and VE-cadherin (Fig. 7F). Electron microscopy showed comparable junctional restoration by either Matk overexpression or Ani treatment (Fig. 7G–I). Notably, no synergistic effects were observed when Ani and Matk overexpression were combined.

Fig. 7. Endothelial cell-specific overexpression of Matk reduced microvascular extravasation following rtPA thrombolysis in ischemic stroke mice.

A, B Representative images and statistical analysis of FITC-dextran leakage through mouse cerebral venules (n = 4 biological replicates), scale bar = 50 μm. C, D Representative images and statistical analysis of Evans blue extravasation of brain tissue homogenate (n = 4 biological replicates). E Representative western blotting images of Matk, t-Src, p-Src (Tyr419), p-Src (Tyr530), VE-cadherin, p-VE-cadherin (Tyr658), p-Caveolin-1 (Tyr14), and ZO-1 protein of isolated mouse brain endothelial cells (n = 6 biological replicates). F Representative immunofluorescence staining of Matk, p-Src (Tyr419), p-Caveolin-1 (Tyr14) and VE-cadherin in mouse brain (n = 3 biological replicates). G–I Representative electron microscopy images and statistical analysis of cerebral microvessels in each group of mice (n = 3 biological replicates). Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. *p < 0.05 vs. Sham+AAV-NC group; #p < 0.05 vs. Thrombus+rtPA+AAV-NC group. Source data are provided as a Source Data file.

Discussion

This study demonstrated that Ani effectively ameliorated delayed rtPA thrombolysis-induced vasogenic cerebral edema in stroke mice, improving neurological function, survival rates, and BBB integrity. Mechanistically, Ani maintained junction proteins expression via binding to Matk and inhibiting its proteasomal degradation, and consequently suppressed Src activation and downstream phosphorylation of VE-cadherin and Caveolin-1. Collectively, these results underscore Ani as a promising adjunct therapy to stabilize BBB integrity and reduce post-thrombolytic microvascular hyperpermeability in clinical practice.

Stroke is a leading cause of global mortality, with ischemic stroke comprising 60–80% of cases³. While rtPA remains the standard thrombolytic therapy for acute ischemic stroke, delayed administration beyond the 4.5-h therapeutic window exacerbates vasogenic cerebral edema⁶, highlighting an urgent need for adjunctive therapies that can stabilize the BBB. Ani, already clinically approved for vascular headaches, retinal vascular spasms, and ischemic optic neuritis¹⁷, could reduce neuronal apoptosis, and modulate leukocyte function through its anti-inflammatory properties in cerebral hypoperfusion and ischemic rodent models^20,21. However, its specific role in addressing delayed rtPA thrombolysis-induced cerebral edema had not been explored. Our dose- and time-window experiments revealed that, Ani significantly attenuated delayed rtPA-induced vasogenic cerebral edema while simultaneously improving neurological function and 7-day survival rates in stroke mice. These results provide compelling evidence for Ani’s potential clinical utility in managing thrombolysis-associated cerebral edema. Notably, when administered at a fixed dose, Ani produced largely comparable therapeutic effects across the three time points tested. This plateau-like response likely reflects both the persistent vulnerability of the BBB in the early phase following thrombolysis and the potential saturable effect of Ani at the tested dose. Importantly, this result suggests a degree of clinical flexibility in using adjunct Ani administration.

The disruption of BBB integrity constitutes the fundamental pathological mechanism underlying delayed rtPA thrombolysis-induced vasogenic cerebral edema in ischemic stroke²⁷. On one hand, delayed rtPA thrombolysis down-regulates tight junction protein Occludin and ZO-1, and adherens junction protein VE-cadherin, leading to increased paracellular leakage^28,29 On the other hand, delayed rtPA thrombolysis enhances phosphorylation of Src (Tyr419) and Caveolin-1 (Tyr14), which promotes transcellular vesicular transport³⁰. Our experimental results demonstrated that Ani simultaneously restored the expression of junctional proteins to limit paracellular leakage and suppressed Src/Caveolin-1 phosphorylation to prevent transcytosis in the post-thrombolytic cerebral vasculature. This dual targeting is important because paracellular junction failure and transcellular vesicular transport are often co-activated following ischemia–reperfusion, and addressing only one route may be insufficient to fully stabilize the BBB in vivo.

A key advance of this study is the identification of Matk as the direct molecular target of Ani. Through an integrated approach combining proteomics with HuProt™ human proteome microarray analysis, we identified Matk as the lead candidate and validated the interaction between Ani and Matk through molecular docking, MD simulations, SPR, CETSA and pull-down assay. We further demonstrated that Ani stabilizes Matk by inhibiting its ubiquitin-proteasomal degradation, thereby increasing its cellular abundance. This target identification strategy moves beyond phenotypic observation to establish a firm molecular link between Ani and its effect.

Matk, a CSK family tyrosine kinase containing characteristic SH2, SH3 and catalytic domains³¹, serves as a key regulator of Src family kinase activity with established roles in cell signaling, proliferation and differentiation³². While aberrant Matk expression has been implicated in breast cancer, leukemia and autoimmune disorders³², its involvement in cerebrovascular pathology remains unknown. Based on the evidence that Matk phosphorylates Src at Tyr530 to suppress its activation^25,33 and coupled with Src’s known role in promoting endothelial permeability through VE-cadherin (Tyr658) and Caveolin-1 (Tyr14) phosphorylation³⁴, we hypothesized that Ani preserves endothelial barrier integrity through Matk-mediated Src inhibition. Our data support this hypothesis by showing that Ani reduced OGD/R‑induced hyperpermeability in HCMECs, suppressed Src activity, and lowered phosphorylation of VE‑cadherin (Tyr658) and Caveolin‑1 (Tyr14) while restoring Matk expression. Loss‑ and gain‑of‑function studies in vitro and in vivo showed that Matk knockdown abolished Ani’s protective effects, whereas Matk overexpression mimicked them, with no additive benefit from combined treatment. Thus, Ani maintains vascular integrity by modulating the Matk/Src axis. Given that Src is involved in multiple permeability signals in ischemia–reperfusion, stabilizing Matk to restrain Src may represent a generalizable strategy to limit microvascular leakage that extends beyond targeting individual upstream triggers, offering a broader and potentially more effective approach for controlling microvascular leakage in ischemia-reperfusion injury.

The translational potential and generalizability of Ani are supported by its established clinical safety profile¹⁷. In sham-operated animals, Ani administration did not produce adverse effects on survival, neurological function, or BBB integrity, indicating a favorable safety window within our experimental context. Moreover, to address the important question of sex-specific responses in stroke and thrombolysis, we conducted validation experiments in female mice, which demonstrated that Ani administration yielded comparable therapeutic benefits to those observed in males across key endpoints including survival, neurological scores, blood perfusion, Evans blue extravasation, and brain water content. These safety and efficacy findings across sexes enhance the translational relevance of Ani as a potential adjunct therapy for post‑thrombolytic cerebral edema management.

This study also has several limitations. First, while our study focused on the effect and mechanism of Ani in ameliorating delayed rtPA thrombolysis-induced vascular leakage and edema from the prospective of endothelial cells, future studies are needed to explore the mechanisms underlying its neuroprotective and anti-inflammatory effects. Second, considering the important role of Src in controlling vascular hyperpermeability, the potential of Ani in treating other diseases related to vascular leakage and edema warrants further investigation. Finally, given the short half-life of Ani demonstrated in our pharmacokinetic studies, future work may optimize the dosing regimens or develop extended‑release formulations to enhance its translational potential.

In conclusion, this study demonstrates that Ani prevents delayed rtPA thrombolysis-induced cerebral edema by targeting Matk to suppress Src-mediated phosphorylation of VE-cadherin and Caveolin-1. Given Ani’s clinical safety profile, these findings highlight its potential as a therapeutic adjunct to rtPA thrombolysis, addressing a critical unmet need in stroke management (Fig. 8).

Fig. 8. Anisodine Hydrobromide Targets Matk and Prevents Delayed rtPA Thrombolysis-Induced Vasogenic Cerebral Edema in Ischemic Stroke.

This schematic illustrates how anisodine hydrobromide (Ani) attenuates delayed rtPA‑induced cerebral edema via the Matk–Src pathway. Delayed rtPA promotes Matk degradation, reducing Matk‑mediated Tyr530 phosphorylation of Src and increasing Src activation (Tyr419 phosphorylation). Activated Src enhances vascular leakage by phosphorylating Caveolin‑1 at Tyr14 (transcellular route) and VE‑cadherin at Tyr658 (paracellular route), the latter also downregulating ZO‑1 and VE‑cadherin. Ani binds Matk, prevents its degradation, and restores Matk levels, thereby maintaining Src inactivation via increased Tyr530 and suppressed Tyr419 phosphorylation. This reduces Caveolin‑1 and VE‑cadherin phosphorylation, preserves ZO‑1/VE‑cadherin expression and localization, and ultimately decreases both trans‑ and paracellular permeability, alleviating edema. This figure is created using BioRender with publication license.

Methods

Reagents

Ani (C₁₇H₂₁NO₅·HBr, CFDA Approval No. H51023360) was supplied by Chengdu First Pharmaceutical (Chengdu, Sichuan, China). Recombinant tissue plasminogen activator (rtPA) was purchased from Actilyse (Boehringer Ingelheim, Biberach, Germany). Ferric chloride (FeCl₃) was obtained from Fuchen Chemicals (Tianjin, China).

Animal

Male and female C57BL/6 J mice (8-week-old) were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Animal License No. SCXK 2021-0006). All animals were maintained under controlled environmental conditions (20 ± 2 °C, 40 ± 5% humidity) with a 12-h light/dark cycle, and provided with standard rodent chow and water ad libitum. Experimental protocols were approved by the Committee on the Ethics of Animal Experiments of the Health Science Center of Peking University (NO. LA2021109) and were complied with the Guide for the Care and Use of Laboratory Animals published by National Institutes of Health (8^th edition, 2011).

Animal model and treatment

We established the ischemic stroke model by using FeCl₃-induced thrombosis combined with rtPA thrombolysis. Mice were anesthetized, and the right carotid artery was exposed. A filter paper soaked in 20% FeCl₃ solution was applied to induce thrombosis. The thrombus was then advanced to induce infarction in the right cerebral hemisphere³⁵. Cerebral blood flow (CBF) was monitored 4.5 h post-surgery using moorFLPI-2 laser speckle flowmetry (Moor Instruments, Devon, UK), with only mice showing 60-90% CBF reduction in the ischemic hemisphere compared to contralateral side being included for further experiments. Mice were randomly divided into Sham, Sham+Ani (2.4 mg/kg), Thrombus, Thrombus+rtPA (rtPA thrombolysis at 4.5 h post-thrombosis), and Thrombus+rtPA+Ani groups, where Ani was administered intravenously at 5.5, 6.5 or 8.5 h post-thrombosis (1, 2 or 4 h post-thrombolysis, respectively) at doses of 0.6 mg/kg (low dose, L), 1.2 mg/kg (medium dose, M, clinically equivalent dose) or 2.4 mg/kg (high dose, H), while control animals received normal saline. The three doses were determined based on the body surface area normalization factor of 9.1 for mouse-to-human dose conversion³⁶, while 0.6 mg/kg represents the lower end of the clinical dose range, 1.2 mg/kg is around the upper clinical dose, and 2.4 mg/kg is nearly twice the high clinical dose. Randomization was performed using a randomization sequence generated by computer software (Excel). For all mice received, animals in the sham-operated group were first randomly selected. The remaining mice were then subjected to surgery to induce cerebral infarction. After successful modeling, the animals were randomly assigned to the model group and various treatment groups. Data analysis was conducted by a researcher who was blinded to group allocation.

Plasma malondialdehyde (MDA) measurement

Plasma MDA levels were measured at 4.5, 5.5, 6.5, 8.5, 12.5, and 28.5 h post-thrombosis (corresponding to 0, 1, 2, 4, 8, and 24 h post-thrombolysis) using a commercial MDA ELISA assay kit (M0441; Andy Gene, Beijing, China).

Plasma Ani measurement

Six male C57BL/6 J mice (8-week-old) were prepared and administered Ani via intravenous injection at a dose of 2.4 mg/kg. Plasma samples were collected at 5 min, 15 min, 30 min, 1 h, 2 h, and 4 h via the capillary method from the angular vein. The plasma was mixed with three times its volume of methanol, vortexed, and then centrifuged at 14,000 rpm for 15 min. The supernatant was analyzed using a liquid chromatography-triple quadrupole mass spectrometry system (AB SCIEX API 4000 QTRAP mass spectrometer) for accurate quantification. Quantitative analysis was performed in multiple reaction monitoring (MRM) mode, with the ion pairs for the analyte Ani set to 320.2-156.1/138.1. Declustering potential (DP) and collision energy (CE) were carefully optimized to achieve the highest signal intensity. The analyte was separated using a Waters ACQUITY UPLC BEH C18 column (50 mm × 2.1 mm, 1.7 µm) equipped with a pre-column. Mobile phase A was water (containing 0.1% formic acid), and mobile phase B was acetonitrile. The flow rate was set at 0.3 mL/min, with a total analysis time of 5 min. The gradient elution program was as follows: 0–0.5 min, 2%–50% B; 0.5–1 min, 50%–95% B; 1–2.5 min, 95% B; and finally, 2.5 min of column equilibration with 2% B. The data were analyzed using Analyst software version 1.5.1 (AB SCIEX).

Isolation of mouse brain endothelial cells

Mouse brain endothelial cells were isolated using magnetic‑activated cell sorting (MACS) and the purity was confirmed by flow cytometry. Briefly, brain tissue was dissociated into a single‑cell suspension using the Adult Brain Dissociation Kit (130‑107‑677; Miltenyi, Bergisch Gladbach, Germany) on a gentleMACS™ Octo Dissociator (Miltenyi). The suspension was filtered through a 70‑μm strainer, centrifuged, and resuspended in DPBS. After debris removal and red blood cell lysis, cells were incubated with mouse anti-CD45 microbeads (130-052-301; Miltenyi) and passed through an LD column to deplete CD45⁺ cells. The flow‑through was then incubated with mouse anti-CD31 microbeads (130-097-418; Miltenyi) and positively selected on an MS column to obtain CD31⁺CD45⁻ endothelial cells. Purified cells were used immediately for downstream experiments. For purity assessment, cells were stained with mouse anti-CD31-PE (130-111-540; Miltenyi), anti-CD45-VioGreen (130-110-803; Miltenyi), and 7‑AAD viability dye (130-111-568; Miltenyi), and analyzed by a MACSQuant analyzer 16 (Miltenyi). Endothelial cell purity was defined as the percentage of live, single cells that were CD31⁺CD45⁻.

Survival analysis

Mice survival was monitored at 24-h intervals for 7 days following thrombosis induction, with survival rates analyzed using Kaplan-Meier methodology.

Neurological deficits

Neurological deficits were evaluated at 28.5 h post-thrombosis (24 h post-thrombolysis) by the neurological evaluation scale (NES) and the modified neurological severity scores (mNSS)³⁷. The NES consisted of five sub‑tests, each scored 0–3. Spontaneous activity was rated as: 0 = almost no activity; 1 = minimal activity without contacting cage walls; 2 = contacts 1–2 walls; 3 = contacts ≥ 3 walls. Limb symmetry (assessed when lifting the mouse by the tail) was scored: 0 = complete asymmetry; 1 = near asymmetry; 2 = moderate asymmetry; 3 = near symmetry. Open‑field locomotion was evaluated as: 0 = no movement; 1 = circling in place; 2 = curved trajectory favoring one side; 3 = straight‑line movement. The balance beam sub‑test was scored: 0 = falls off; 1 = hangs on beam > 30 s; 2 = stands with minimal movement; 3 = moves freely along beam. The vibrissae‑touch test assessed sensory response by gently touching vibrissae on both sides with a thin wooden stick: 0 = no response on either side; 1 = no response on affected side; 2 = weak response on affected side; 3 = symmetrical response. The mNSS assessment included motor tests, a balance beam test, and reflex absence tests. Motor tests comprised a tail‑lift test (scored 0–3: 1 point each for forelimb flexion, hindlimb flexion, and head lift > 10° from vertical within 30 s) and a free‑walking test (0 = normal walk; 1 = unable to walk straight; 2 = circling toward paretic side; 3 = falling to paretic side), yielding a maximum motor score of 6. The balance beam test was scored from 0 to 6: 0 = normal balance; 1 = grasps beam side; 2 = one limb slips off; 3 = two limbs slip off; 4 = falls after 20 s; 5 = falls after 10 s; 6 = falls within 10 s. Reflex absence tests (maximum 2 points) awarded 1 point each for loss of the ear‑flick reflex (no head shake upon touching the auditory canal) and loss of the corneal reflex (no blink upon gentle corneal contact with a cotton tip).

CBF monitoring

CBF was monitored at baseline, 4.5 h (before thrombolysis) and 28.5 h post-thrombosis (24 h post-thrombolysis) using moorFLPI-2. The relative reduction in ischemic hemisphere blood flow (%) was calculated using moorFLPI-2 analysis software as follows: (ischemic hemisphere flow / contralateral hemisphere flow) × 100%.

Evans blue leakage

At 25.5 h post-thrombosis (21 h post-thrombolysis), anesthetized mice were intravenously administered 2% Evans blue (4 mL/kg; Sigma-Aldrich, St. Louis, MO, USA) via the femoral vein. 3 h later, animals were transcardially perfused with ice-cold normal saline. Brains were then harvested and sectioned coronally into five 2-mm-thick slices, which were immediately photographed using a stereomicroscope. The right hemisphere was homogenized in 50% trichloroacetic acid, followed by 24 h incubation at 4 °C and subsequent centrifugation. Evans blue extravasation was quantified by measuring the absorbance of the supernatant at 620 nm using Synergy 2 microplate reader (BioTek, Winooski, VT, USA).

Brain water content measurement

Brain water content was evaluated at 28.5 h post-thrombosis (24 h post-thrombolysis). Following euthanasia under anesthesia, whole brains were rapidly collected and weighed to determine the wet weight. Brain tissues were then dehydrated at 60 °C for 72 h to obtain the dry weight. The brain water content (%) was calculated as (wet weight - dry weight) / wet weight × 100%.

FITC-dextran leakage of microvascular

At 28.5 h post-thrombosis (24 h post-thrombolysis), mice received intravenous administration of FITC-dextran (50 mg/kg; Sigma-Aldrich, St. Louis, MO, USA) via the femoral vein. Following a 30-min circulation period, a 2 × 3 mm² thinned-skull cranial window was surgically prepared to visualize cerebral microvasculature. Fluorescence imaging (excitation/emission: 488/520 nm) of cerebral venules was performed using an upright fluorescence microscope. Quantitative analysis of vascular permeability was conducted by measuring fluorescence intensities in both venules (V) and adjacent interstitial regions (I) using ImageJ software, with the I/V ratio serving as an index of FITC-dextran leakage.

Cell culture and OGD/R modeling

Human cerebral microvascular endothelial cell line hCMEC/D3 (HCMECs; BNCC337728, BNCC, Henan, China) were cultured in DMEM (Thermo Fisher, Waltham, MA, USA), supplemented with 10% FBS (Thermo Fisher) and 1% penicillin-streptomycin (100 U/mL penicillin and 100 μg/mL streptomycin) (Thermo Fisher). To validate the hCMEC/D3 cell line, STR genotyping was performed following the workflow: genomic DNA was extracted using TIANGEN’s genome extraction kit, amplified via a 21-STR amplification protocol, and analyzed for STR loci and the gender gene Amelogenin on a Seqstudio genetic analyzer. The validation results showed no multi-allele gene phenomenon, with the matched cell line being hCMEC/D3 from the ExPASy database (with an EV value of 1.00). The in vitro data in this study are based on independent biological replicates. Specifically, cells from different frozen stocks were seeded in multiple individual culture dishes and expanded for at least three days before treatment. Each dish was maintained as an independent culture. On the day of treatment, cells from each dish were processed separately, and were subsequently harvested and analyzed individually. Control groups were defined and processed in the same manner. To establish the oxygen-glucose deprivation/Reoxygenation (OGD/R) model, cells at 70-80% confluence were first subjected to glucose- and serum-free DMEM and maintained in a hypoxic chamber (94% N₂, 1% O₂, 5% CO₂) for 4.5 h. Then, reoxygenation was achieved by replacing the medium with complete DMEM (containing 10% FBS) under normoxic conditions (37 °C, 5% CO₂) for 24 h.

RNA isolation and qPCR

Total RNA was extracted by TRIzol™ reagent (Invitrogen, Waltham, MA, USA), and reversely transcribed to cDNA. qPCR was performed on a AriaMx system (Agilent, CA, USA). The mRNA expression was calculated using the 2^-ΔΔCt method and normalized to GAPDH.

Western blot

Mouse cerebral cortex of the right hemisphere and HCMECs were homogenized in RIPA lysis buffer (supplemented with protease and phosphatase inhibitors; Roche, Basel, Switzerland) on ice, followed by centrifugation at 12,000 × g for 20 min at 4 °C. Protein concentrations were determined using a BCA assay kit (Applygen, Beijing, China). Equal amounts of protein lysates (20–50 µg) were resolved by 8–10% SDS-PAGE and electrotransferred onto PVDF membranes (Millipore, Burlington, MA, USA). Membranes were blocked with 5% non-fat milk or 1% BSA in TBST for 1 h at room temperature, then probed overnight at 4 °C with the following primary antibodies: anti-ZO-1 (13663, 1:1000), -JAM-1 (82196, 1:1000), -Occludin (91131, 1:1000), -Claudin-5 (49564, 1:500), -α-catenin (36611, 1:1000), -Matk (20729, 1:1000) and -β-catenin (9582, 1:1000) (Cell Signaling Technology, Danvers, MA, USA); anti-Src (ab109381, 1:1000), -p-Src (Tyr419) (ab185617, 1:1000), -p-Src (Tyr530) (ab317421, 1:1000), -Caveolin-1 (ab32577, 1:1000), -p-Caveolin-1 (Tyr14) (ab75876, 1:1000), -VE-cadherin (ab205336, 1:1000) and -p-VE-cadherin (Tyr658) (ab119785, 1:1000) (Abcam, Cambridge, UK). After three washes with TBST, membranes were incubated with HRP-conjugated anti-rabbit or anti-mouse IgG (1:5000, Emarbio Science, Beijing, China) for 1 h at room temperature. Protein bands were visualized using an Image Lab system (Bio-Rad, Hercules, CA, USA) with ECL substrate (Applygen, Beijing, China). For protein expression levels quantification, band intensities were measured using ImageJ software (v1.53, NIH, Bethesda, MD, USA).

Immunofluorescence staining

Frozen brain sections (10 μm thickness) were subjected to antigen retrieval using 0.01 M sodium citrate buffer (pH 6.0) at 95 °C for 15 min. After cooling to room temperature, sections were permeabilized with 0.3% Triton X-100 for 15 min and blocked with 10% normal goat serum in PBS at 37 °C for 1 h. Sections were then incubated overnight at 4 °C with the following primary antibodies: anti-Occludin (1:200), anti-VE-cadherin (1:200), anti-von Willebrand factor (vWf, 1:100), anti-Matk (1:200), anti-phospho-Src (Tyr419, 1:200), and anti-phospho-Caveolin-1 (Tyr14, 1:200). Fluorescence images were captured using a Leica TCS SP8 DIVE laser scanning confocal microscope (Leica Microsystems, Mannheim, Germany) equipped with × 40 and × 63 oil immersion objectives.

Proteomic analysis

To investigate the mechanisms underlying Ani-mediated protection against rtPA‑induced blood‑brain barrier disruption, a label‑free quantitative proteomic analysis was performed on cortical tissues from the right hemisphere of mice in three groups: Sham, Thrombosis+rtPA, and Thrombosis+rtPA+Ani, with three 8‑week‑old male C57BL/6 J mice per group. At 28.5 h after modeling, cortical tissue was rapidly collected, lysed, and digested with trypsin. Peptides were analyzed on a Thermo Orbitrap Exploris 480 mass spectrometer coupled to a Vanquish Neo nano‑LC and FAIMS Pro interface (−45 V CV) in data‑independent acquisition (DIA) mode. Separation used a 75 µm × 25 cm C18 column with a 120‑min 5–30% acetonitrile gradient at 300 nL/min. MS1 scans (m/z 400–1000) were acquired at 60,000 resolution; DIA scans used 40 variable windows with MS2 resolution of 15,000 and 28% normalized HCD energy. Raw data were processed with DIA‑NN (v1.8.1) against the UniProt mouse proteome database (ID UP000000589), allowing one missed cleavage, fixed carbamidomethylation, variable oxidation/acetylation, and requiring ≥2 unique peptides per protein at 1% FDR. Proteins with fold‑change >1.5 or <0.67 and p < 0.05 were considered differentially expressed and further analyzed via GO enrichment using the OmicsBean platform (http://www.omicsbean.cn).

HuProt™ human proteome microarray

The HuProt20K Human Proteome Microarray—with a coverage around 20,000 full-length human proteins—was used for the high-throughput screening of Ani-binding targets³⁸. Recombinant proteins expressed in eukaryotic systems with GST affinity tags were immobilized on this array with a technical replicates per protein. Ani was conjugated to D-biotin for detection, while unmodified D-biotin served as a negative control.

Molecular docking

The three-dimensional structure of Ani was retrieved from PubChem (CID: 11616712) and structurally optimized with Openbabel (v3.1.1)³⁹, Using the ReverseDock tool, we determined the binding affinities of Ani with Matk, Dido1, Glrx2, Psmc3, and Cuta proteins⁴⁰.The Matk structure was retrieved from the AlphaFold Protein Structure Database (AlphaFold ID: AF-P42679-F1-v4) and prepared with Dock Prep within ChimeraX (v1.7.1)⁴¹. Subsequently, the structure of Matk and Ani were processed by MGLTools (v1.5.7) and binding modes was predicted with molecular docking by AutoDock Vina (v1.2.5)⁴². The result was visualized by PyMOL (v3.1.0).

Molecular dynamics simulation

Molecular dynamics (MD) simulations were performed using GROMACS 2024.5 with the AMBER ff14SB force field for the protein. The partial atomic charges of Ani were obtained using the RESP2 method by Multwfn. Then the General Amber Force Field (GAFF) was used for the generation of forcefield parameter by Sobtop (1.0(dev5), http://sobereva.com/soft/Sobtop). Long-range electrostatic interactions were calculated using the particle mesh Ewald (PME) method with a 1.0 nm cutoff and van der Waals interactions were truncated at 1.0 nm with a long-range dispersion correction applied to energy and pressure. The system was first energy-minimized using the conjugate gradient algorithm until the maximum force dropped below 100 kJ·mol⁻¹·nm⁻¹, followed by a 100 ps position-restrained equilibration under NPT conditions at 298.15 K and 1 bar using the V-rescale thermostat and the Berendsen barostat. Production MD was then performed for 100 ns under NPT conditions at 298.15 K and 1 bar with the V-rescale thermostat and the Parrinello-Rahman barostat, using a 2 fs integration time step, with coordinates saved every 2 ps for subsequent analysis. The binding free energy was estimated for the 95-100 ns trajectory using the gmx_MMPBSAtool implementing the MM/GBSA method⁴³.

Pull-down assay

Cell suspensions in PBS containing protease inhibitor cocktail was lysed through three freeze-thaw cycles in liquid nitrogen. Lysates were centrifuged at 20,000 × g for 20 min at 4 °C to collect the supernatant, which was divided into three experimental groups while retaining an Input sample as control. Subsequently, groups were treated with: (1) PBS (vehicle control), (2) 50 μM Ani-biotin, or (3) 50 μM Ani-biotin plus 100 μM Ani (competition control), followed by 12 h incubation at 4 °C. Next, each group was incubated with streptavidin (SAV) magnetic beads (SAV MagBeads, Yeasen Biotechnology, Shanghai, China) for additional 12 h at 4 °C. The SAV MagBeads were adsorbed by a magnetic grate, washed with PBS and then denatured in 1 × SDS loading buffer (100 °C, 10 min). 20 μL aliquots from each group were used for western blot analysis.

SPR assay

The binding affinity between Matk and Ani was assessed using surface plasmon resonance (SPR) assays with the Biacore 8 K system (Cytiva, Marlborough, MA, USA). Purified Matk protein was immobilized on CM5 sensor chips through amine coupling. Binding interactions were characterized using a multi-cycle kinetic assay with serial dilutions of analytes. Data analysis was performed using Biacore Evaluation Software (Version 2.0).

CETSA

The cellular thermal shift assay (CETSA) was conducted following established protocols. Briefly, HCMECs were incubated with Ani or vehicle control for 2 h. Cells were then washed with ice-cold PBS, detached using trypsin, and resuspended in PBS containing protease inhibitors. Equal volumes of cell suspensions were exposed to a temperature gradient using a PCR thermal cycler, with each temperature maintained for 3 min followed by a 3-min incubation on ice. The samples underwent three freeze-thaw cycles, after which they were centrifuged, and the supernatants were analyzed by western blotting. Protein band intensities were quantified using ImageJ software.

HCMECs permeability assay and transendothelial electrical resistance (TEER) measurement

HCMECs were cultured in Transwell inserts until reaching confluency. Following monolayer formation, cells underwent OGD/R. At 2 h after reoxygenation, Ani was administered. After 24 h of reoxygenation, endothelial permeability was assessed through two complementary approaches: FITC-dextran leakage assay, where the tracer was added to the upper chamber for 2 h incubation followed by fluorescence measurement in the lower chamber using a microplate reader; and TEER measurement using an EVOM2 voltohmmeter (World Precision Instruments, Sarasota, FL, USA) to evaluate barrier integrity.

Src activity assay in HCMECs and the mouse brain tissues

Src kinase activity in HCMECs and mouse brain tissues was measured using the Src Kinase Activity Quantitative Detection Kit (GMS 50060.1 for HCMECs; GMS 50060.2 for tissue; Genmed Scientifics, MA, USA) according to the manufacture’s instruction.

Transmission electron microscopy (TEM)

Samples were fixed in 2.5% glutaraldehyde at 4 °C, followed by secondary fixation in 1% osmium tetroxide. After dehydration through a graded ethanol series, samples were embedded in EPON resin. Ultrathin sections (70 nm) were prepared using a Leica UC7 ultramicrotome (Leica Microsystems, Wetzlar, Hesse, Germany) and double-stained with uranyl acetate and lead citrate. High-resolution images were acquired using a Tecnai G2 Spirit TEM (FEI, USA) operating at 120 kV.

Knock-down of Matk in HCMECs

The Matk siRNA (siMatk: 5′-GUUCACCAGCAAGUCGGAUGU-3′) and control siRNA (siNC: 5′-UUCUCCGAACGUGUCACGUTT-3′) were synthesized by Generay Biotech (Shanghai, China). Cell transfection was performed using Lipofectamine RNAiMAX transfection reagent (Invitrogen) according to the manufacturer’s instructions.

Matk Overexpression in HCMECs

The coding sequence (CDS) of human Matk (NM_139355) was amplified by PCR and subcloned into the pcDNA3.1-his vector (pcDNA3.1-matk-his, MatkOE for short) (Tsingke Biotech, Beijing, China). The empty pcDNA3.1-his vector was used as control (NC for short). Transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions.

Cellular uptake of Ani and OCT2-dependence in endothelial cells

To investigate whether Ani enters endothelial cells and whether the entry is OCT2‑mediated, HCMECs were incubated with Ani-biotin for 24 h at 37 °C, washed three times with PBS, fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X‑100, and stained with streptavidin‑FITC followed by DAPI counterstaining. Fluorescence images were acquired using a Cytaion 7 Cell Imaging system (Agilent). For pharmacological inhibition, cells were incubated with the OCT2 inhibitor cimetidine for 30 min before Ani‑biotin treatment; for genetic knockdown, cells were transfected with 50 nM OCT2 siRNA (siOCT2: 5′-GUGAUCAUGAGGUCCUUCCAU-3′) for 48 h before Ani‑biotin treatment.

AAV-BI30-tie1 vector-mediated matk knockdown and overexpression in mice

The serotype 2/BI30 of adeno-associated virus (AAV2/BI30) carrying the miR30 knockdown sequence (5′- UGUUGCUGAAGGCAUGGAAUATT -3′) of mouse Matk under Tie1 promoter (AAV2/BI30-Tie1-mir30-shMatk, AAV-shMatk for short), the AAV2/BI30 carrying the coding sequence of mouse Matk (NM_010768) under Tie1 promoter (AAV2/ BI30-Tie1-Matk-3xFlag, AAV-Matk for short), as well as the control AAV (AAV2/BI30-Tie1-null, AAV-NC for short) were constructed and purchased from Hanbio Biotechnology (Shanghai, China).

Statistical analysis

Data were analyzed using GraphPad Prism 9.0 and expressed as mean ± SD. Normality was assessed by Shapiro–Wilk test. Data were analyzed using a two-tailed unpaired t-test. No adjustments were made for multiple comparisons. p < 0.05 was considered statistically significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.G. and A.Q.L. performed the experiments, analyzed the data, and wrote the article. F.K.C., C.Y., Y.J.Z., Y.Z. and H.B.L participated in the animal and cell experiments. D.Z.W. and Z.W. contributed to the SPR and MD Simulation experiment. X.C. contributed to electron microscopy. J.L. designed and supervised the research, provided key research directions, wrote the article and provided the funding of this study. All authors read and agreed with the final article. All data were generated in-house, and no paper mill was used. All authors agree to be accountable for all aspects of work ensuring integrity and accuracy.

Peer review

Peer review information

Nature Communications thanks Christine Cheung, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

J.L. discloses support for the research of this work from Chengdu First Pharmaceutical [grant numbers 2023016 and 2024129].

Data availability

All data supporting the findings of this study are included in the article and following public data repositories. Source data are provided with this paper, and have also been deposited in the figshare database under accession ID 31828060⁴⁴, and the raw proteomics mass spectrometry data are publicly available in the ProteomeXchange Consortium via the iProX partner repository under accession code PXD068816. The mass spectrometry data for determining the plasma concentration of Ani have been deposited in the figshare repository under accession ID 32084754⁴⁵. The HuProt Human Proteome Microarray Chip data are available in figshare under accession ID 32086650⁴⁶. All data can be publicly accessed via the corresponding hyperlinks. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73995-0.