The ANGPTL3-integrin α5 axis drives retinal vascular leakage in diabetic retinopathy

Abstract

Background

Breakdown of the blood-retinal barrier (BRB) is a key driver of diabetic retinopathy (DR). While Angiopoietin-like protein 3 (ANGPTL3) is a clinical predictor of DR, its causal role and mechanism in retinal vascular pathology are unknown.

Methods

This study combined genetic (Angptl3 knockout and overexpression) and pharmacological (intravitreal evinacumab) approaches in streptozotocin-induced diabetic mice to define the in vivo function of ANGPTL3. In vivo assessments included fluorescein angiography and optical coherence tomography for retinal vascular integrity, alongside electron microscopy for ultrastructural analysis of endothelial junctions. Mechanistic investigations involved co-immunoprecipitation to probe the ANGPTL3-integrin α5 interaction, and in vitro permeability assays using human endothelial cells. The therapeutic efficacy of a neutralizing antibody, evinacumab, was subsequently evaluated via intravitreal administration in diabetic mice.

Results

ANGPTL3 overexpression exacerbated STZ-induced retinal vascular leakage, increased retinal thickness, and disrupted endothelial junctions. Plasma ANGPTL3 levels were significantly lower in knockout mice and elevated in overexpression mice, with no observed effects on body weight or systemic glucose tolerance. Co-immunoprecipitation assays confirmed a direct interaction between ANGPTL3 and integrin α5. In HUVECs, integrin α5 knockdown attenuated ANGPTL3-induced downregulation of ZO-1 and VE-cadherin expression, thereby preserving endothelial barrier function. In vivo, intravitreal evinacumab reduced vascular leakage and restored junctional protein levels in diabetic mice, without altering body weight or systemic glucose metabolism.

Conclusions

This study identifies a novel mechanism whereby ANGPTL3 disrupts endothelial junctions via integrin α5-dependent pathways, thereby contributing to DR progression. The ANGPTL3-integrin α5 axis represents a promising therapeutic target, with pharmacological inhibition offering a potential strategy to ameliorate diabetic vascular leakage.

Clinical trial registration number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-07710-4.

Background

Diabetic retinopathy (DR) remains a leading cause of blindness in working-age adults, primarily driven by the breakdown of the blood-retinal barrier (BRB) and subsequent vascular hyperpermeability [1]. Globally, more than 100 individuals (approximately 22.3% of diabetic adults) are affected by some form of DR, with nearly 6.2% progressing to vision-threatening stages. This number is projected to rise to 160 million by 2045 [2]. Notably, adolescents with type 2 diabetes exhibit a 13.7% prevalence of DR within 8 years of diagnosis, highlighting the critical need for early therapeutic intervention [3].

The progression of DR is driven by vascular hyperpermeability, a hallmark of initial stages characterized by tight junction disruption and pericyte loss [4–6]. These pathologic changes contribute to retinal oedema, hemorrhage, and ultimately, neovascularization [7]. Despite advances in anti-VEGF therapies, over 40% of patients exhibit suboptimal responses [8], underscoring the need for identifying alternative therapeutic targets that regulate endothelial barrier integrity.

Angiopoietin-like protein 3 (ANGPTL3) is a hepatocyte-secreted glycoprotein and has emerged as a dual regulator of lipid metabolism and vascular function [9, 10]. Although the ANGPTL3 inhibitor evinacumab has been approved for the treatment of dyslipidemia [11], its role in microvascular diseases remains unexplored. Beyond involvement in lipid modulation, ANGPTL3 is increasingly recognized for its diverse vascular functions, including angiogenesis promotion through integrin αvβ3 and association with endothelial dysfunction in obesity [12, 13]. Notably, our previous clinical study has identified serum ANGPTL3 as an independent predictor of DR progression across all stages [14]. However, whether this clinical correlation reflects a causal role for ANGPTL3 in driving BRB disruption remained a critical unanswered question.

To unravel the mechanism of ANGPTL3-mediated vascular pathology, we turned our attention to the integrin family of receptors, which are known to mediate ANGPTL3’s vascular effect [12]. While ANGPTL3 interacts with integrin αvβ3 to regulate angiogenesis, the regulation of barrier permeability often involves distinct signaling pathways. Notably, integrin α5β1, a primary fibronectin receptor on endothelial cells, is a critical stabilizer of endothelial junctions, directly influencing the localization and expression of key barrier proteins like ZO-1 and VE-cadherin [15]. Its dysfunction contributes to barrier impairment during pathological angiogenesis [16], suggesting a potential role in mediating ANGPTL3-induced vascular leakage.

Building upon previous findings, we hypothesized that ANGPTL3 directly exacerbates DR by compromising endothelial barrier function. Employing genetic mouse models, pharmacological approaches, and in vitro mechanistic investigations, we demonstrate that ANGPTL3 directly interacts with retinal endothelial cells, disrupting tight junction integrity. Crucially, we also show that local inhibition of ANGPTL3 rescues vascular leakage in diabetic models. Our findings thus unveil a novel ANGPTL3-dependent mechanism underlying BRB breakdown in DR and highlight ANGPTL3 as a promising, locally amenable therapeutic target for this devastating complication.

Methods

Proteomic analysis of aqueous humour in DR

Proteomic data from human aqueous humour samples, comprising 10 healthy controls and 15 patients with DR, were obtained from a publicly available dataset [17]. Differential expression analysis was performed using the limma package in R (version 4.4.1). Proteins with an absolute log2-fold change greater than 1 and an adjusted P-value below 0.05 were considered statistically significant. Results were visualized using the ggplot2 package.

Protein–protein interaction and functional enrichment analysis

To explore potential interacting partners of the target protein ANGPTL3, the GeneMANIA database (https://genemania.org/) was used with Homo sapiens selected as the organism. The platform identified interacting proteins and performed functional enrichment analysis to associate relevant biological processes.

Animals and the induction of a diabetic model

Eight-week-old male wild-type C57BL/6 mice and global Angptl3 knockout mice (Angptl3^−/−, strain: C57BL/6) were obtained from Viewsolid Biotechnology Co., Ltd (Beijing, China) and housed under specific pathogen-free (SPF) conditions. The mice were maintained on a 12-hour light/dark cycle at 22–24℃ with 50%-60% humidity. Diabetes was induced using intraperitoneal streptozotocin injection (STZ; Sigma-Aldrich, Cat# S0130, St. Louis, MO, USA), 80 mg/kg body weight, for 5 consecutive days. Blood glucose levels were monitored weekly using a Roche Accu-Chek glucometer (Roche, Mannheim, Germany). Mice were classified as diabetic if non-fasting blood glucose levels exceeded ≥ 16.7 mmol/L in two consecutive measurements. Only diabetic mice were included in the subsequent experiments. Body weight and blood glucose levels were recorded biweekly until 21 weeks post-STZ injection (Supplementary Figure S1). The C57BL/6 mice were randomized into the control and diabetic groups (n = 10 per group), while the Angptl3^−/− mice were assigned to the corresponding groups (n = 9 per group). All animal procedures were approved by the Institutional Animal Care and Use Committee of Beijing Capital Medical University (Approval Number AEEI-2021-158).

ANGPTL3 overexpression mediated by adeno-associated virus

To induce systemic ANGPTL3 overexpression, an adeno-associated virus serotype PHP.Eb construc encoding mouse Angptl3 (AAV-Angptl3; full name: PHP.Eb-EGFP-Angptl3) was generated by OBiO Technology (Shanghai, China). The mice in the AAV-group received tail vein injections of AAV-Angptl3 at 2 × 10¹¹ genome copies/mL in 200 µL. Control mice were injected with an equal volume of vehicle.

Evinacumab treatment

Evinacumab (Evi; MCE, Cat# HY-P99194, USA) was administered to diabetic mice. At 21 weeks post-STZ injection (a model of long-term diabetes [18, 19]), diabetic mice were randomly allocated to the STZ group (intravitreal phosphate-buffered saline (PBS) injection) and the Evi group (intravitreal evinacumab 10 mg/mL, 1 µL per eye, weekly for 8 weeks). All intravitreal injections were performed using a 32-gauge Hamilton syringe (Hamilton, Reno, NV, USA) under stereomicroscope guidance (Olympus, SZX10, Tokyo, Japan). Fundus fluorescein angiography (FFA) was performed 8 weeks after the first injection [20].

Intraperitoneal glucose tolerance test (IPGTT)

Following a 14-hour overnight fast (with free access to water), mice received an intraperitoneal injection of 10% glucose solution (2 g/kg body weight). Blood glucose levels were measured in tail vein blood samples at baseline (0 min) and 15, 30, 60, 90, and 120 min post-injection. Glucose tolerance was assessed by calculating the area under the curve (AUC) for blood glucose levels over time.

Optical coherence tomography (OCT)

The mice were anesthetized using intraperitoneal tribromoethanol (250 mg/kg; 2123 A, Nanjing AIBI Biotech, China). Pupil dilation was achieved using 0.5% tropicamide (Sinqi Pharmaceutical, Shenyang, China). A medical-grade ultrasonic coupling gel (Cofoe, China) was applied to the ocular surface to maintain corneal hydration. Retinal cross-sectional images were acquired using a Micron IV OCT system (Phoenix Research Labs, USA) with a dedicated murine objective lens. The system was configured with an A-scan resolution of 1024 points per scan, with 20 linear scans averaged per eye to minimize motion artifacts. Quantitative analysis of retinal thickness, measured from the inner limiting membrane to the retinal pigment epithelium, was performed using the integrated InSight 4.0 software.

FFA

Mice received intraperitoneal 2% sodium fluorescein (34 mg/kg; Lishede Pharmaceutical, China). After 1 min, retinal fluorescein leakage was imaged using an OPTO-RIS retinal imaging system (Optoprobe, UK). Quantitative analysis was performed using ImageJ software (v1.52a; NIH, USA), with fluorescein leakage expressed as the ratio of extravascular to intravascular fluorescence intensity. Three predefined regions of interest (ROIs) were analyzed per eye, with measurements averaged for analysis.

Periodic acid-schiff (PAS) staining

Retinal PAS staining was performed to quantify acellular capillaries (AC-Cap), endothelial cells (EC), and pericyte cells (PC), as previously described [21]. The eyes were fixed in 4% paraformaldehyde for 24 h, and the retinas were carefully dissected under a microscope. Flat-mounted retinas were fixed in 4% paraformaldehyde for 1.5 h, followed by digestion with 3% trypsin for 2 h. Non-vascular tissues were gently removed using a 3 mL transfer pipette with PBS, and the remaining vascular networks were mounted and spread onto glass slides under microscope guidance. After air-drying overnight, the samples were stained using a PAS stain kit (Solarbio, Cat# G1281) according to the manufacturer’s instructions. Retinal capillary degeneration was assessed under light microscopy (Olympus BX43, Tokyo, Japan). The numbers of AC-Cap, ECs, and PCs were counted using ImageJ software (v1.52a, NIH, USA).

Transmission electron microscope (TEM)

Retinal tissue samples (1 mm³ sections) were fixed in 2.5% glutaraldehyde for 2 h and post-fixed in 1% osmium tetroxide for 1 h. The samples were dehydrated through a graded ethanol series and embedded in Epon 812 resin. Ultrathin sections (70 nm) were prepared, stained with uranyl acetate and lead citrate, and examined using a Hitachi HT7800 TEM (Japan) at 80 kV. Tight junction integrity was assessed.

Immunofluorescence staining

Eyeballs were fixed in FAS eyeball fixative solution (ServiceBio, Cat# G1109) for 24 h, followed by secondary fixation in paraformaldehyde. The samples were embedded in paraffin and sectioned at a thickness of 8 μm. Human umbilical vein endothelial cells (HUVECs) were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.1% Triton X-100. Tissue sections were blocked with 5% bovine serum albumin (BSA) and incubated overnight at 4 °C with rabbit anti-ZO-1 (1:200; Abcam, Cat# ab276131, RRID: AB_3083081), mouse anti-VE-cadherin (1:200; Cell Signaling Technology, Cat# 2500, RRID: AB_10839118), or mouse anti-eNOS (1:500; ServiceBio, Cat# GB12086, RRID: AB_3678903). Following thorough washing, the sections were incubated for 1 h at room temperature with species-appropriate Alexa Fluor 488- or 594-conjugated secondary antibodies (1:500; Molecular Probes, Cat# A-11004, RRID: AB_2534072 and Thermo Fisher Scientific, Cat# A-11008, RRID: AB_143165). Nuclei were counterstained with DAPI (1:500; Southern Biotech, Cat# 0100-20). Fluorescence images were acquired using a Zeiss LSM 700 confocal microscope (Germany).

Measurement of lipid profile

Cholesterol (TC), triacylglycerol (TG), high-density lipoprotein (HDL), and low-density lipoprotein (LDL) levels were quantified using a fully automatic biochemical analyzer (Chemray 800, Rayto, Shenzhen, China) by Servicebio Technology CO., Ltd. (Wuhan, China)

Measurement of ANGPLT3

Plasma ANGPTL3 levels were quantified using a commercial ELISA kit (IBL, 27410, Immuno-Biological Laboratories Co., Ltd., Japan) following the manufacturer’s protocol. Absorbance measurements were performed using an Enspire multilabel plate reader (Perkin Elmer, Waltham, MA, USA).

Cell culture

HUVECs (BNCC342438) were obtained from BeNa Culture Collection (Hebei, China) and maintained in endothelial cell growth medium (ECM; ScienCell, Cat# 1001) at 37 °C in a humidified 5% CO₂ atmosphere. The cells were seeded in 6-well plates and cultured to confluence. The cells were serum-starved overnight in ECM containing 1% fetal bovine serum (FBS). Confluent cell monolayers were treated with high glucose (30.5 mmol/L) and human recombinant ANGPTL3 simultaneously.

Real-time cell analysis (RTCA) proliferation assay

Cell proliferation dynamics following ANGPTL3 treatment were monitored using the xCELLigence RTCA system (ACEA Biosciences, Inc., San Diego, CA, USA). Briefly, 50 µL of ECM medium was added to each well of a 16-well E-plate, which was then placed on the RTCA station inside a 37 °C, 5% CO₂ incubator for 30 min to allow temperature equilibration. Background impedance was automatically recorded, and wells with a cell index (CI) < 0.063 were selected for further analysis. HUVECs were resuspended in ECM at 5 × 10⁴ cells/mL, and 50 µL of the cell suspension (about 2.5 × 10³ cells/well) was gently added to the selected wells. The cells were incubated at room temperature for 30 min before being transferred back to the RTCA station. CI values were recorded every 5 min until a stable baseline (CI = 1.0 ± 0.2) was reached, typically 24 h post-seeding; this time point was designated as the normalization reference (T₀). At T₀, the cells were treated with recombinant human ANGPTL3 (ProSpec, Cat# CYT-986-B) at 0, 0.25, 0.50, and 0.75 µg/mL. Short-term proliferative responses were recorded every 5 min for 2 h, while long-term effects were monitored at 30-minute intervals thereafter.

In vitro permeability assay

HUVECs were seeded at 2 × 10⁵ cells/cm² on gelatin-coated 12-well Transwell inserts (0.4 μm polyester membrane; Corning, Cat# 3378) and cultured in ECM medium for 72 h to form confluent monolayers. The cells were equilibrated in ECM containing 1% FBS (Gibco, Cat# 26140079) for 12 h before treatment with recombinant ANGPTL3 (0.25, 0.50, 0.75 µg/mL) or recombinant vascular endothelial growth factor (VEGFA; 50 ng/mL, R&D Systems, Cat# 293-VE/CF) as a positive control. For the permeability assay, 500 µL of ECM was added to the lower chambers, and 100 µL of treatment medium was added to the upper chambers. Paracellular permeability was assessed using fluorescein isothiocyanate (FITC)-labelled dextran (Sigma-Aldrich, Cat# FD70S-1G). FITC-dextran was added to the upper chambers at a final concentration of 1 mg/mL (100 µL per well) and incubated for 1 h. Subsequently, 50 µL of culture medium was collected from both upper and lower chambers. Fluorescence intensity was measured using an Enspire multimode plate reader (Perkin Elmer, Waltham, MA, USA) at an excitation and emission wavelength of 485 nm and 510 nm, respectively. The apparent permeability was calculated as the ratio of fluorescence intensity in the lower chamber to that in the upper chamber.

Co-immunoprecipitation (CO-IP) assays

Protein lysates were extracted from HUVECs using tissue lysis buffer (Applygen, Cat# C1051) supplemented with a protease inhibitor cocktail (Solarbio, Cat# P0100). The lysates were precleared by incubation with 100 µL of protein A/G agarose beads in 1 mL lysis buffer. The precleared lysates were incubated overnight at 4 °C with gentle rotation in the presence of anti-ANGPTL3 (Thermo Fisher Scientific, Cat# MA5-35681, RRID: AB_2849581), anti-integrin α5 (Abcam, Cat# ab179475, RRID: AB_2716738), or anti-IgG. Following immunoprecipitation, the antibody-protein complexes were collected by centrifugation and washed three times with ice-cold lysis buffer. The immunoprecipitates were resuspended in 2× SDS loading buffer and denatured at 95 °C for 10 min. The samples were separated by SDS-PAGE, and immunoprecipitated proteins were detected by immunoblotting using appropriate secondary antibodies.

siRNA transfection

HUVECs at 60% confluence were transfected with 50 nmol/L of integrin α5-specific siRNA (Santa Cruz Biotechnology, Cat# sc-29372) or a scrambled control siRNA using Lipofectamine RNAiMAX (Invitrogen, Cat# 11668-019) in accordance with the manufacturer’s protocol. Six hours post-transfection, the culture medium was replaced. Experiments were conducted 48 h later, and knockdown efficiency was confirmed by western blot.

Western blot analysis

Retinal tissues or HUVEC lysates were homogenized in RIPA buffer supplemented with protease inhibitors. Total protein (20 µg per sample) was separated by SDS-PAGE on 7.5%-12% polyacrylamide gels and transferred onto nitrocellulose membranes. The membranes were blocked with 5% BSA for 1 h at room temperature and incubated overnight at 4 °C with anti-ZO-1 (1:1000, Abcam, Cat# ab276131, RRID: AB_3083081), anti-VE-cadherin (1:1000, Cell Signalling Technology, Cat# 2500, RRID: AB_10839118), or anit-β-actin (1:5000; ZSGB-Bio, Cat# TA-09, RRID: AB_2636897). After washing, the membranes were incubated with HRP-conjugated secondary antibodies (1:3000; ZSGB-Bio, Cat# ZB-2305, RRID: AB_2747415 and ZSGB-Bio, Cat# ZB-2301, RRID: AB_2747412) for 1 h at room temperature. Protein bands were visualized using enhanced chemiluminescence (ECL) (Thermo Fisher, Cat# WBKLS0500), and band intensities were quantified using Image Lab software (v6.0; Bio-Rad).

Statistical analysis

Data were presented as means ± standard error of the mean (SEM). Comparisons between two groups were analyzed using the unpaired Student’s t-test. For multiple group comparisons, one-way ANOVA with Tukey’s post-hoc test was employed. Statistical significance was defined as P < 0.05. All analyses were conducted using GraphPad Prism 9.0 (GraphPad Software, San Diego, USA).

Results

ANGPTL3 is elevated in the ocular microenvironment of DR: evidence from human aqueous humor and diabetic mouse retinas

In our analysis of proteomics data from human aqueous humor samples of healthy individuals and patients with DR [17], ANGPTL3 was identified as a significantly upregulated gene in the DR group compared to the DM group (Fig. 1A). This finding is consistent with previous clinical observations reporting a positive correlation between serum ANGPTL3 levels and DR progression [17]. To validate this in our preclinical model, we assessed ANGPTL3 expression in the retinas of STZ-induced diabetic mice. Consistent with the human data, both western blotting (Fig. 1B, C) and immunofluorescence staining (Fig. 1D, E) demonstrated a marked increase in ANGPTL3 protein levels, particularly within the retinal vasculature of diabetic mice compared to healthy controls. These findings establish that ANGPTL3 levels are elevated in the diabetic retinal microenvironment, positioning it as a potential contributor to DR pathology.

Fig. 1.

ANGPTL3 is elevated in the diabetic ocular microenvironment. (A) Bioinformatics analysis showing an upregulation of ANGPTL3 and ANGPTL4 in aqueous humor. (B, C) Western blot analysis of ANGPTL3 and ANGPTL4 in retinal tissue from the DR group and controls. (D, E) Immunofluorescence staining for ANGPTL3 in retinal sections from control and STZ-diabetic mice (nuclei counterstained with DAPI)

Genetic targeting of ANGPTL3 rescues diabetic retinal vascular leakage, while its overexpression exacerbates pathology

To establish a causal role for ANGPTL3 in DR, we utilized global Angptl3 knockout (Angptl3⁻/⁻) mice. Successful gene deletion was confirmed (Supplementary Figure S2A, B). Importantly, Angptl3 deletion had no effect on body weight or glucose tolerance in either non-diabetic or diabetic mice (Fig. 2A-C), suggesting that its retinal effects are independent of systemic glucose metabolism. While Angptl3 knockout altered systemic lipid profiles as expected (Supplementary Figure S3), its primary protective effects were observed directly within the retina.

Fig. 2.

Genetic targeting of ANGPTL3 rescues diabetic retinal vascular leakage. (A) Body weight of the mice. (B and C) IPGTT and corresponding AUC. (D and E) Representative FFA and quantification of extravascular/intravascular fluorescence intensity ratio. (F and G) Representative OCT images and quantification of retinal layer thickness. (H) PAS staining of retinal vasculature (scale bar = 50 μm). (I) Quantification of AC-Cap, EC, and PC based on PAS-stained images in (H)

Functionally, fundus fluorescein angiography (FFA) revealed that while diabetic wild-type (WT) mice developed significant vascular leakage, Angptl3⁻/⁻ diabetic mice were largely protected, with leakage rates reduced by 25.4% (P < 0.01) and comparable to non-diabetic controls (Fig. 2D and E). Structurally, optical coherence tomography (OCT) showed that Angptl3 deletion prevented diabetes-induced thinning of vascularized retinal layers, including the inner nuclear layer (INL) and the ganglion cell-inner plexiform layer (GCL + IPL) complex, without affecting non-vascular layers (Fig. 2F and G; Supplementary Figure S4). At the microvascular level, analysis of retinal trypsin digests demonstrated that Angptl3 knockout attenuated key hallmarks of diabetic microangiopathy, significantly reducing pericyte loss and the formation of acellular capillaries (Fig. 2H and I). Together, these results demonstrate that loss of ANGPTL3 preserves retinal vascular integrity and function during diabetes.

Systemic overexpression of ANGPTL3 exacerbates retinal vascular leakage

To complement the knockout studies, we induced systemic ANGPTL3 overexpression in WT mice using an AAV vector, which resulted in a 1.84-fold increase in plasma ANGPTL3 levels (Fig. 3A). Similar to the knockout model, ANGPTL3 overexpression did not alter body weight or glucose tolerance (Fig. 3B and C). However, in stark contrast to the protective phenotype of Angptl3 deletion, ANGPTL3 overexpression significantly exacerbated STZ-induced retinal vascular leakage by 26.38% compared to diabetic controls (P < 0.05) (Fig. 3D and E). This gain-of-function evidence reinforces the conclusion that elevated ANGPTL3 directly promotes retinal vascular pathology.

Fig. 3.

ANGPTL3 overexpression exacerbates STZ-induced retinal vascular leakage in diabetic mice. (A) Plasma ANGPTL3 levels following AAV-mediated ANGPTL3 overexpression. (B and C) IPGTT and AUC following ANGPTL3 overexpression. (D and E) Representative FFA images and quantification of vascular leakage in ANGPTL3-overexpressing mice. ^*P < 0.05, ^**P < 0.01

ANGPTL3 knockout preserves retinal endothelial barrier function by upregulating tight junction proteins

Given the profound effects of ANGPTL3 on vascular leakage, we investigated its impact on endothelial tight junctions, the structural basis of the BRB. Transmission electron microscopy (TEM) of retinal capillaries revealed prominent gaps and disrupted junctional complexes between endothelial cells in diabetic WT mice. In contrast, diabetic Angptl3⁻/⁻ mice maintained near-normal, intact tight junction ultrastructure (Fig. 4A).

Fig. 4.

ANGPTL3 knockout preserves retinal endothelial barrier function by upregulating tight junction proteins. (A) Representative TEM images showing tight junction structures in retinal capillary endothelial cells of control, diabetic, and Angptl3 knockout mice. (B) Immunofluorescence staining of retinal sections for tight junction proteins ZO-1 and VE-cadherin. Scale bars = 10 μm. (C and D) Quantification of ZO-1 and VE-cadherin fluorescence intensity from (B). (E) Proliferation assay of HUVECs treated with recombinant human ANGPTL3. (F) Western blot analysis of ZO-1 and VE-cadherin protein levels in HUVECs following ANGPTL3 treatment. (G-H) Quantification of protein expression from (F). ^*P < 0.05, ^**P < 0.01

This structural integrity was mirrored at the molecular level. Immunofluorescence analysis for the key tight junction proteins ZO-1 and VE-cadherin showed continuous, linear staining along the vasculature in control and Angptl3⁻/⁻ diabetic mice, but fragmented and diminished staining in diabetic WT mice (Fig. 4B). Quantification confirmed that Angptl3 deletion significantly rescued the diabetes-induced reduction of both ZO-1 and VE-cadherin expression (Fig. 4C and D). To confirm that ANGPTL3 directly acts on endothelial cells, we treated cultured HUVECs with recombinant ANGPTL3. This led to a dose-dependent decrease in ZO-1 and VE-cadherin protein levels, without affecting cell proliferation (Fig. 4E-H). These findings indicate that ANGPTL3 compromises the endothelial barrier by directly promoting the loss of essential tight junction proteins.

ANGPTL3 compromises endothelial barrier function through integrin α5-dependent signaling

To identify the receptor mediating ANGPTL3’s effects, we performed a protein-protein interaction network analysis, which highlighted integrin α5 (ITGA5) as a potential binding partner (Fig. 5A). We confirmed a direct physical interaction between endogenous ANGPTL3 and integrin α5 in HUVECs via co-immunoprecipitation (Fig. 5C; protein docking simulation shown in Fig. 5B).

Fig. 5.

ANGPTL3 compromises the endothelial barrier via integrin α5-dependent signaling. (A) Protein interaction network analysis. (B) Protein docking simulation (C) Co-IP assay between integrin α5 and ANGPTL3 in HUVECs. (D) Quantification of fluorescence intensity in endothelial permeability. (E) Representative immunofluorescence images of HUVECs stained for ZO-1 and VE-cadherin. Scale bars = 20 μm. (F and G) Quantification of ZO-1 and VE-cadherin fluorescence intensity is shown in (E). *P < 0.05, **P < 0.01

To determine if this interaction was functionally essential, we used siRNA to silence ITGA5 expression in HUVECs (Supplementary Figure S5). Remarkably, knockdown of integrin α5 completely abolished the ANGPTL3-induced increase in endothelial permeability (Fig. 5D). Furthermore, silencing integrin α5 prevented the ANGPTL3-driven downregulation of ZO-1 and VE-cadherin, restoring their expression to control levels (Fig. 5E-G). These results establish that ANGPTL3 compromises endothelial barrier function through a mechanism critically dependent on its interaction with integrin α5.

Therapeutic targeting of ANGPTL3 with evinacumab rescues retinal vascular leakage without altering systemic metabolism

Finally, to evaluate the therapeutic potential of targeting ANGPTL3, we administered the clinically approved neutralizing antibody evinacumab via intravitreal injection into mice with established long-term diabetes (Fig. 6A). This local treatment strategy had no significant effect on body weight or systemic blood glucose levels (Fig. 6B and C). Despite the lack of systemic metabolic changes, evinacumab treatment resulted in a dramatic and significant reduction in retinal vascular leakage, decreasing the leakage rate by nearly 20% (P < 0.01) and restoring barrier function to near-healthy levels (Fig. 6D and E). Furthermore, evinacumab could improve the thickness of the outer nuclear and photosensitive layers, but with limited effect on the thickness of the inner vascular layer, which may reflect that its local vascular barrier protection precedes or mainly targets the outer structures (Supplementary Figure S6). This finding demonstrates that local inhibition of ANGPTL3 is sufficient to ameliorate diabetic retinal vascular pathology, validating the ANGPTL3-integrin α5 axis as a viable therapeutic target.

Fig. 6.

Therapeutic targeting of ANGPTL3 with evinacumab rescues retinal vascular leakage without altering systemic metabolism. (A) Flowchart of the study. (B) Fasting blood glucose level. (C) Body weight of the mice. (D-E) Representative FFA images and quantification of vascular leakage. ^*P < 0.05, ^**P < 0.01

Discussion

DR is fundamentally a disease of vascular leakage, but > 40% of patients show suboptimal response to anti-VEGF therapies. Our study addresses this critical gap by identifying ANGPTL3 as a direct and causal driver of BRB disruption in DR. The study demonstrated that ANGPTL3 exacerbates retinal vascular leakage. Most significantly, we uncover a novel pathogenic mechanism: ANGPTL3 physically interacts with integrin α5 on endothelial cells to dismantle tight junctions, a pathway that is therapeutically tractable via local ANGPTL3 blockade.

Although ANGPTL3 is well known for its role in hepatic lipid metabolism, its pathological function in extrahepatic tissues, particularly the retina, remains poorly understood. Using both systemic knockout and ANGPTL3 overexpression models, the results strongly suggest a direct correlation between ANGPTL3 levels and BRB integrity. ANGPTL3 deficiency preserved retinal structure, reduced AC-Cap density, and attenuated pericyte loss, whereas its overexpression exacerbated vascular leakage.

Consistent with ANGPTL3’s role as an inhibitor of lipoprotein lipase (LPL) and endothelial lipase (EL), our Angptl3⁻/⁻ mice exhibited reduced systemic levels of TG and TC [22]. However, we observed an intriguing discrepancy: while pharmacological inhibition of ANGPTL3 with evinacumab in human clinical trials consistently lowers LDL, our global knockout mice showed an elevation in LDL. This difference may arise from several factors. First, it could reflect a complex, long-term compensatory response to constitutive gene deletion in mice versus acute pharmacological blockade in humans. Second, species-specific differences in lipid metabolism regulation cannot be excluded. Crucially, this paradox, combined with our finding that local intravitreal evinacumab robustly reversed leakage without altering systemic lipids, strongly reinforces our central hypothesis: ANGPTL3’s pathogenic role in the diabetic retina is a primary, localized effect on the vasculature, not merely a downstream consequence of its influence on systemic lipid profiles.

The present study identified integrin α5 as a novel binding partner, a finding with dual significance. First, it distinguishes retinal ANGPTL3 signaling from its reported mechanism of nephropathy, involving integrin αVβ3 [23], suggesting tissue-specific receptor usage. Second, this discovery reveals the functional plasticity of ANGPTL3 across vascular beds. While ANGPTL3-αVβ3 interaction promotes angiogenesis, the ANGPTL3-α5β1 axis in the retina specifically compromises endothelial barrier integrity. This receptor-context specificity likely underlies the divergent outcomes of ANGPTL3 signaling, which appear to be dependent on local integrin expression patterns, microenvironmental cues, and disease context. Further investigation into the selective integrin partnership of ANGPTL3 across tissues may yield valuable insights into its multifaceted roles in vascular biology.

The findings demonstrate that ANGPTL3 interacts with integrin α5 to compromise retinal endothelial barrier integrity by downregulating the junction proteins ZO-1 and VE-cadherin. The integrin α5β1 complex is well established to regulate endothelial permeability through activation of downstream signaling cascades, including FAK, Src family kinases, and the RhoA/ROCK pathway, which collectively orchestrate cytoskeletal remodeling and junctional complex stability [24, 25]. While the direct exploration of these pathways was beyond the scope of this study, it can be proposed that ANGPTL3-integrin α5 engagement likely triggers analogous signaling events that ultimately lead to junctional disassembly. Elucidating these downstream molecular mechanisms will be essential for a comprehensive understanding of ANGPTL3-mediated blood-retinal barrier disruption in DR.

In recent years, significant progress has been made in the development of drugs for ANGPTL3 due to the in-depth understanding of ANGPTL3’s role in vascular endothelial damage, inflammatory response, and metabolic regulation [10]. Evinacumab is a fully human monoclonal antibody targeting ANGPTL3 and has been approved for the treatment of homozygous familial hypercholesterolemia, showing superior lipid-lowering and vasoprotective effects in several studies [26–28]. Although there are currently no studies directly evaluating the potential of evinacumab in DR, its ability to effectively inhibit circulating ANGPTL3 levels makes it a therapeutic candidate of interest. Given the evidence that ANGPTL3 is involved in retinal vascular injury, neoangiogenesis, and BRB disruption, further efficacy and safety studies of evinacumab or other ANGPTL3-targeted agents in DR models may provide a useful complement to existing treatments and provide new strategies for patients who do not benefit from conventional treatments.

The translational implications of this newly identified axis were validated by the pharmacological intervention. The use of intravitreal evinacumab yielded highly significant results. Local ocular delivery dramatically reduced retinal vascular leakage in a long-term diabetes model, validating our target in a therapeutically relevant setting [29]. This approach holds immense promise, particularly for the significant cohort of patients with DR refractory to anti-VEGF therapy. It offers a complementary mechanism of action focused on restoring the structural integrity of the BRB. Furthermore, our findings may provide a mechanistic explanation for the observed benefits of fenofibrate in DR. Fenofibrate is known to reduce ANGPTL3 expression, and its protective effects may, at least in part, be mediated through the inhibition of the ANGPTL3-integrin α5 pathway we have described [30]. The strategy of local ANGPTL3 blockade, however, presents a distinct advantage by directly targeting the ocular pathology while circumventing the potential systemic side effects associated with PPARα agonists like fenofibrate.

The novelty of the present study is the study of the ANGPTL3-Integrin α5 axis in DR from in vivo association to causation. While earlier work observed correlations in patient samples and animals and performed therapeutic modulations of ANGPTL3 in DR models [9–14, 22, 29–31], the present study provides direct causal evidence that ANGPTL3 drives retinal vascular leakage in diabetic patients using the genetic loss-of-function (global Angptl3⁻/⁻) and gain-of-function (systemic AAV-mediated ANGPTL3 overexpression) methods. These genetic manipulations suggest that altering ANGPTL3 levels is sufficient to alter BRB integrity in diabetic animals, creating causality rather than just associations. Second, a novel tissue-specific receptor and mechanism were identified: ANGPTL3-integrin α5-tight junction disassembly. Previous mechanistic work in other tissues involved integrin αVβ3 or angiogenesis-related pathways [12, 16, 23, 25]. For probably the first time, the present study (i) identified integrin α5 as a direct binding chaperone (co-IP docking) of ANGPTL3 in retinal endothelial cells, and (ii) showed that ANGPTL3-integrin α5 signaling leads to loss of key tight junction proteins (ZO-1 and VE-cadherin) and increases endothelial permeability. It reveals a VEGF-independent pathway for retinal-specific receptor background and BRB disruption. A third novelty is the therapeutic translation through local blockade. Although fenofibrate and ANGPTL3-targeted approaches have been systematically explored, the present study showed that local intravitreal delivery of the clinically approved ANGPTL3-neutralizing antibody evinacumab restores barrier integrity in mice with long-term diabetes without systemic metabolic changes. It suggests that local ocular ANGPTL3 activity is sufficient to drive pathology and that local ANGPTL3 inhibition is a viable, VEGF-independent therapeutic strategy.

In this study, neither systemic deletion nor systemic overexpression of Angptl3 significantly altered glucose tolerance or blood glucose levels, and local vitreous injection of evinacumab did not affect systemic metabolic parameters. It suggests that the pathogenic effects of ANGPTL3 in retinal blood vessels are mainly local and direct, rather than indirectly mediated by systemic metabolic changes.

While our findings are robust, we acknowledge certain limitations that open avenues for future research. This study demonstrates ANGPTL3’s pro-leakage effect in DR, which appears to contrast with its reported anti-angiogenic effects in some tumor models [31]. This highlights a critical need to understand how the diabetic microenvironment might modulate ANGPTL3 function, perhaps through post-translational modifications like glycation, shifting its signaling output from protective to pathogenic. Our study relied on rodent models; therefore, two critical next steps are warranted. First, validating the elevation and potential modification of ANGPTL3 in vitreous humor samples from human patients across different stages of DR is essential. Second, assessing the long-term efficacy and safety of intravitreal ANGPTL3 inhibition in higher-order animal models, such as non-human primates, will be crucial before clinical translation. Furthermore, a single dose of evinacumab was used, preventing the investigation of a dose-response effect, which should be examined in future studies.

Besides, we only evaluated the effect of ANGPTL3 blockade separately. There are currently no experimental data on the synergistic, superposition, or antagonist effects of combined interventions. In future work, experiments on the ANGPTL3 block and anti-VEGF combination therapy should be designed to explore the possible interaction between the two in the protection of the BRB, given that VEGF inhibitors have been shown to improve DR [32].

Conclusions

In conclusion, this study deciphers a novel and fundamental pathogenic axis in DR, where ANGPTL3 acts via integrin α5 to dismantle endothelial barrier integrity (Fig. 7). By demonstrating that local, targeted inhibition of ANGPTL3 effectively ameliorates vascular leakage independent of systemic metabolic control, we provide compelling preclinical evidence for a new therapeutic strategy. This work positions the ANGPTL3-integrin α5 pathway as a highly promising target to combat diabetic vascular disease, offering hope for a more effective and personalized approach to preserving vision in patients with diabetes.

Fig. 7.

Schematic of the proposed mechanism for ANGPTL3-mediated BRB disruption in DR. Under diabetic conditions, increased ANGPTL3 interacts with endothelial integrin α5, leading to the downregulation of tight junction proteins (ZO-1, VE-cadherin). This compromises BRB integrity and causes vascular leakage. Both genetic knockout of Angptl3 and local inhibition with evinacumab rescue this phenotype by preserving tight junction structure and function

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

DR

Diabetic retinopathy

ANGPTL3

Angiopoietin-like protein 3

STZ

Streptozocin

FFA

Fundus fluorescein angiography

OCT

Optical coherence tomography

PAS

Periodic acid-Schiff

HUVECs

Human umbilical endothelial cells

ECM

Endothelial cell growth medium

PBS

Phosphate-buffered saline

RTCA

Real time cellular analysis

CI

Cell Index

VE-cad

vE-cadherin

ZO-1

Zonula occludens-1

IPGTT

Intraperitoneal glucose tolerance test

AUC

Area under the curve

AC-Cap

Acellular capillaries

EC

Endothelial cells

PC

Pericyte cells

Author contributions

Jing Ke: Study conceptualization, experimental design, and original manuscript drafting. Yongsong Xu and Yingjun Zhu: Conducted in vitro and in vivo experiments, performed statistical analysis, and verified data reproducibility. Yongsong Xu: Methodology documentation and manuscript revision. Xiaotong Feng: Established rodent models and assisted in histological interpretation. Haodi Cao: Performed proteomic data processing and pathway enrichment analysis. Dong Zhao and Longyan Yang: Led study conceptualization, provided project supervision and data interpretation, and critically revised the manuscript with intellectual input. All authors participated in the discussion of results, approved the final manuscript, and accept accountability for all aspects of the work.

Funding

This work was supported by the National Natural Science Foundation of China (grant No. 81800723), the Natural Science Foundation of Beijing Municipality (grant No. 7222099), and the Beijing Nova Program (grant No. 20220484039).

Data availability

The datasets generated and/or analyzed during this study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The animal experiment was executed with the agreement of the Institutional Animal Care and Use Committee of Beijing Capital Medical University (Approval Number AEEI-2021-158), and all methods are reported in accordance with ARRIVE guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.