Restoration of Autophagic Flux Improves Endothelial Function in Diabetes Through Lowering Mitochondrial ROS-Mediated eNOS Monomerization Free

Cardiovascular complications are the leading cause of morbidity and mortality in patients with diabetes (1,2). Endothelial dysfunction is featured by increases in vascular tone, oxidative stress, and adhesion of monocytes to endothelial cells (ECs), as well as decreases in barrier function, EC proliferation and migration, and angiogenic properties (2). Endothelial dysfunction is also critically involved in diabetic vasculopathy, retinopathy, and nephropathy (2). Impaired vascular tone regulation is an initial event of diabetes-associated endothelial dysfunction. Hyperglycemia, advanced glycation end products (AGEs), and oxidized low-density lipoprotein (ox-LDL) are among the major inducers of endothelial dysfunction in diabetes (2). Endothelial nitric oxide synthase (eNOS)-derived nitric oxide (NO) is the key local determinant of vascular tone, particularly in larger vessels. eNOS dimerization is necessary for maintaining normal eNOS catalytic activity, which requires the presence of a cofactor, tetrahydrobiopterin (BH4) (3). Under a hyperglycemic condition, overproduction of reactive oxygen species (ROS) lowers BH4 availability, eNOS thus produces peroxynitrite (ONOO⁻) instead of NO, further increasing oxidative state in ECs—a situation termed eNOS uncoupling (4). BH4 deficiency destabilizes eNOS by turning eNOS dimers into monomers, the latter lack eNOS activity and are more susceptible to ubiquitination and proteosome degradation (5). eNOS uncoupling or decreased eNOS dimer to monomer ratio is widely observed in vessels of patients or rodents with diabetes and also in ECs exposed to high glucose levels (6–9). Results of previous studies suggest that dysfunctional mitochondria are the principal source of ROS overproduction responsible for eNOS uncoupling (10). However, what causes endothelial mitochondrial dysfunction in diabetes is not fully understood.

Autophagy is a multistep intracellular process for eliminating damaged proteins and organelles (11). Upon autophagy induction, cytosolic LC3-I is conjugated to phosphatidylethanolamine to form LC3-II, a marker for the double-membrane autophagosome (12). The autophagosome fuses with lysosome to form an autophagolysosome, where the enclosed contents are degraded by cathepsins. Sequestosome 1 (SQSTM1/p62) is degraded along with the targeted organelles and proteins, and thus widely serves as a marker for monitoring autophagic flux, an indicator of the degradative activity of the autophagic system (12). Accumulated LC3-bound puncta increase the conversion of LC3-I to LC3-II and thus reduce p62 content, reflecting degradation within the autophagolysosome and autophagy activation. Reduced LC-3-bound puncta, decreased LC3-II, and increased p62 content collectively indicate an impaired autophagy at the initiation stage, whereas a preserved number of puncta and protein levels of LC3-II and p62 in the presence of lysosome inhibitor suggest an inability of cells to clear autophagosomes, which is indicative of impairment of autophagic flux (13). Existing evidence reveals that autophagy and autophagic flux play crucial roles in modulating mitochondrial turnover. For example, knockout of autophagy-related protein 5 (ATG5) or ATG7 increases the generation of ROS and accumulation of enlarged and defective mitochondria (14,15). Inhibition of autophagosome fusion with lysosome by chloroquine (CQ) increases mitochondrial ROS (mtROS) production (16). In a recent study, researchers found that autophagic flux was impaired in peripheral vein ECs of patients with diabetes. Autophagy or autophagic flux modulates the eNOS activity and NO production (17). However, the role of autophagic flux in regulating eNOS dimerization and monomerization in diabetes is unknown.

Transcription factor EB (TFEB) is a master regulator of autophagy and lysosomal biogenesis. TFEB overexpression increases the number of autophagosomes and generation of new lysosomes, promotes the fusion of lysosomes, and thus facilitates autophagic flux (18). In the aging heart, impaired TFEB nuclear translocation inhibits autophagic flux, leading to increased oxidative stress in cardiomyocytes (19). Meanwhile, TFEB drives the autophagy-lysosome system in macrophages to protect against atherogenesis (20,21). TFEB overexpression reduces accumulation of dysfunctional mitochondria and cellular debris via restoration of autophagosome formation (22). By contrast, TFEB knockout mice exhibit reduced expression of mitochondrial bioenergetics-related genes in skeletal muscle, leading to defective ATP production and exercise intolerance (23). Mammalian target of rapamycin kinase complex 1 (mTORC1), the primary negative regulator of autophagy, is the key upstream kinase that directly phosphorylates and inactivates TFEB (24). In the presence of nutrients, the hyperactivated mTORC1 phosphorylates TFEB at Ser142 and Ser211 to inhibit its nuclear translocation (25). Conversely, calorie restriction or pharmacologic inhibition of mTORC1 activates TFEB by promoting its nuclear translocation, thereby enhancing autophagic activity (24). We recently demonstrated that TFEB protein is downregulated and inactivated in diabetic mouse aortas (26). Overexpressing TFEB attenuates EC inflammation in diabetes by restraining the canonical NF-κB pathway (26). However, whether mTOR-TFEB-autophagy signaling plays a role in regulating eNOS dimerization and monomerization, which, in turn, affect endothelial function in diabetes, is yet to be established.

In this study, we investigated whether and how impaired autophagic flux reduces eNOS activity and endothelial function through suppression of eNOS dimerization in ECs. We also determined whether restoration of autophagic flux through TFEB induction, mTOR inhibition, and calorie restriction can rescue the impaired endothelial function in diabetic mice by restoring eNOS dimerization.

Male and female db/m⁺ and db/db mice (ages 8–10 weeks) were provided by the Chinese University of Hong Kong Laboratory Animal Service Center. Most of the experiments were carried out using male mice unless otherwise mentioned. The mice were housed in a temperature-controlled holding room (22–23°C) with a 12-h light/dark cycle and were fed standard chow diet and water. The experimental protocol was approved by the Chinese University of Hong Kong Laboratory Animal Experimentation Ethical Committee and is consistent with the Guide for the Care and Use of Laboratory Animals. The mice were treated with rapamycin (2 mg/kg every 2 days; Sigma-Aldrich, St. Louis, MO) or vehicle via intraperitoneal injection for 12 days. This dosage of rapamycin was reported to be safe for animal treatment (27).

The protocol of calorie restriction was described previously (28). Briefly, db/db mice were randomly assigned into two groups: an ad libitum feeding group and a calorie restriction group. For the calorie restriction group, the daily food consumption was recorded in the first 3 days. On days 4–7, each mouse was given 50%, 10%, 10%, and 10% of the weight of the average amount of daily food. Then these mice were subjected to ad libitum feeding for 7 days. A period from day 4 to day 14 was considered 1 cycle of calorie restriction. Before each cycle, body weight and random blood glucose levels were measured. After 3 cycles, mice were sacrificed and arterial tissues were collected for functional and molecular assays.

Briefly, the right ventricle was perfused with PBS to empty blood from the lung. The lung tissues were cut into small pieces and digested for 40 min in 10 mL of Hanks’ balanced salt solution containing type I collagenase (450 units/mL) and hyaluronidase (60 units/mL), which was stopped by addition of an equal volume of FACS buffer (PBS with 2 mmol/L EDTA plus 2% FBS). Cells were filtered using a 70-μm strainer to remove the undigested tissue and centrifuged at 300g for 10 min at 4°C. The cell pellet was resuspended in 200 μL of FACS buffer and then added with 1:20 CD31 microbeads. After 15-min incubation on ice, 2 mL of FACS buffer was added to each sample, and each sample was centrifuged at 300g for 5 min at 4°C to remove the supernatant. The cell pellet was again resuspended in 1 mL of FACS buffer. An LS column on the magnet was set up and prewashed with 3 mL of FACS buffer, and then a 1-mL cell pellet suspension was added into the column, followed by washing twice in another 4 mL of FACS buffer. The column was taken off the magnet and placed on the top of a 15-mL tube. The beads in the column were washed with 4 mL of FACS, using a syringe plug. The collected solution in the tube was finally centrifuged at 300g for 5 min and the cell pellet was labeled as CD31⁺ lung microvascular ECs.

Human umbilical vein ECs (HUVECs) were purchased from Lonza (catalog no. CC-3317; Walkersville, MD) and cultured in EC growth medium (DMEM/F12 supplemented with 20% FBS, unless mentioned otherwise), EC growth supplement (50 μg/mL), and 100 units/mL penicillin plus 100 μg/mL streptomycin) in a 37°C incubator. Cells between passages 4 and 8 were used.

Aortic rings (∼2 mm long) were dissected out in ice-cold PBS and incubated in low-glucose (5.5 mmol/L) DMEM (Gibco, Gaithersburg, MD) supplemented with 10% FBS (Gibco), 100 μg/mL streptomycin, and 100 units/mL penicillin. CQ (an autophagy inhibitor; 10 µmol/L), bafilomycin A1 (BafA1; an autophagy inhibitor; 10 nmol/L), 3-methyladenine (3-MA) (an autophagy inhibitor; 10 mmol/L), rapamycin (300 nmol/L), Mito TEMPO (100 µmol/L), Mito Q (10 µmol/L), or AGEs (100 µg/mL), all from Sigma-Aldrich, were added to the culture medium that bathed aortic rings in an incubator at 37°C. After the incubation, rings were transferred to a chamber containing fresh Krebs solution for functional studies on myograph or Western blotting.

After mice were sacrificed, thoracic aortas were dissected out and placed in oxygenated ice-cold Krebs solution containing (mmol/L): 119 NaCl, 4.7 KCl, 2.5 CaCl₂, 1 MgCl₂, 25 NaHCO₃, 1.2 KH₂PO₄, and 11 d-glucose. The thoracic aortas were cut into 2 mm–long segments, and changes in isometric tone of the aortic rings were recorded on a wire myograph (Danish Myo Technology, Aarhus, Denmark). The rings were stretched to an optimal baseline tension of 3 mN. After equilibration for 60 min at baseline tension, rings were contracted with 60 mmol/L KCl solution (NaCl in Krebs solution was substituted with an equimolar amount of KCl) and rinsed in Krebs solution. Endothelium-dependent relaxations (EDRs) induced by acetylcholine (Ach; 0.003–10 μmol/L; catalog A6625, Sigma) were measured in phenylephrine-contracted rings. Endothelium-independent relaxations induced by the NO donor, sodium nitroprusside (0.001–10 μmol/L; catalog no. 71778, Sigma) were recorded in aortas in the presence of l-N^G-nitro-l-arginine methyl ester (100 μmol/L; catalog 72760, Fluka). The relaxation was presented as a percentage reduction of phenylephrine-induced contraction.

Protein samples prepared from mouse aorta homogenates or cells were electrophoresed on the 10% or 12.5% SDS–polyacrylamide gels and then transferred to an Immobilon-P polyvinylidene difluoride membrane (Millipore Corp., Bedford, MA). The membranes were blocked by 3% BSA in 0.05% Tween-20 PBS, and incubated overnight at 4°C with primary antibodies including LC3 (Abcam, ab51520), p62 (Cell Signaling Technology, Danvers, MA; catalog no. 5114), eNOS (BD, catalog no. 610297), p-eNOS (Ser1177; BD, catalog no. 612392), mTOR (Cell Signaling Technology, catalog no. 2972), p-mTOR (Ser2448; Cell Signaling Technology, catalog no. 2971), TFEB (Cell Signaling Technology, catalog no. 4240), TFEB (Proteintech, catalog no. 13372-1-AP), p-TFEB (Ser142; Merck Millipore, catalog no. ABE1971-I), β-actin (Cell Signaling Technology, catalog no. 4967), and GAPDH (Cell Signaling Technology, catalog no. 5174). Then, the membranes were washed three times in TBST and incubated with corresponding secondary antibodies conjugated with horseradish peroxidase (DakoCytomation, Carpinteria, CA) at room temperature for 2 h. The membranes were developed using an enhanced chemiluminescence detection system (ECL reagents, Amersham Pharmacia Biotech, Buckinghamshire, U.K.) and subsequently exposed on X-ray films.

The protocols of low-temperature Western blotting were described previously (29). Briefly, protein samples were prepared in radioimmunoprecipitation assay buffer (in mmol/L: NaCl 150, Tris-HCl 50, NP-40 1%, SDS 0.1%; pH 7.2) supplemented with protease and phosphatases inhibitors. After 3 cycles of freezing and thawing in liquid nitrogen, samples were centrifuged at 15,000g for 10 min. Protein concentrations were measured using the bicinchoninic acid method. Then, 4× sample loading buffer (0.5 mol/L Tris-HCl, 8% SDS, 40% glycerol, 28% [v/v] 2-mercaptoethanol, and 0.0025% bromphenol blue; pH 6.8) was added. Samples were loaded on 6% or 12% polyacrylamide gels and subjected to electrophoresis at constant current of 20 mA/gel. Gels and buffers were cooled to 4°C prior to electrophoresis. Gels were also run in a refrigerated cold box to maintain the temperature of the gel below 15°C. The transfer of proteins to polyvinylidene difluoride membrane and the antibody incubation process we followed were the same as the described normal Western blotting method.

HUVECs or segments of thoracic aortas were fixed in 4% paraformaldehyde and permeabilized by 0.01% Triton X-100 in PBS for 15 min. The cells or segments were washed three times in PBS, blocked by 5% normal donkey serum in PBS for 1 h at room temperature, and subsequently incubated with primary antibodies for 48 h at 4°C. The cells or segments were again washed three times in PBS and incubated with AlexaFluor secondary antibodies for 2 h at room temperature in the dark. Nuclei were counterstained with DAPI. The aortic segments were cut open longitudinally and placed upside down between two coverslips. Images were captured by Fluoview FV1000 laser scanning confocal system (Olympus, Tokyo, Japan).

The intracellular NO level was determined using an NO-sensitive dye, 4-amino-5-methylamino-2′,7′-difluorofluorescein diacetate (DAF-FM DA; Molecular Probes, catalog D23844). HUVECs were incubated in 2 µmol/L DAF-FM DA in the dark at room temperature for 10 min and then washed twice for 5 min each time. NO production was stimulated by 1 µmol/L calcium ionophore A23187 (Sigma-Aldrich) and evaluated by measuring fluorescence intensity excited at 495 nm and emitted at 515 nm on the Fluoview FV1000 laser scanning confocal system (Olympus). Changes in NO production were calculated as relative fluorescence intensity F1/F0 (F0 is the average of fluorescence signals before addition of A23187, and F1 is the fluorescence signal at defined time intervals after A23187 administration).

The amount of intracellular mtROS production in the vascular wall and en face endothelium was measured by MitoSOX staining (Invitrogen, Waltham, MA; catalog M36008). For en face MitoSOX staining, aortic rings were incubated with MitoSOX (1 µmol/L) in extracellular medium (in mmol/L: 121 NaCl, 5 NaHCO₃, 10 Na-HEPES, 4.7 KCl, 1.2 KH₂PO₄, 1.2 MgSO₄, 2 CaCl₂, 10 glucose; pH 7.4) for 20 min at 37°C. The rings were washed twice in extracellular medium and cut open longitudinally. The endothelium was placed upside down between two coverslips, and the fluorescence images were taken on the Olympus FV1000 confocal microscope (MitoSOX: excitation: 510 nm, emission: 580 nm; autofluorescence of elastin excitation: 488 nm, emission: 520–535 nm). For determination of intracellular mtROS in vitro, HUVECs were incubated with MitoSOX (1 µmol/L) for 20 min at 37°C, protected from light. Fluorescence images were captured on a confocal system and analyzed by FLUOVIEW software (version 3.05a; Olympus).

To measure BH4 concentration, 1 × 10⁶ HUVECs were washed three times in PBS and resuspended in 200 µL of PBS. Cells were ultrasonicated three times, each for 3 s, and centrifuged at 1,000g for 15 min at 4°C. The supernatant was collected and assayed using a BH4 ELISA kit (MyBiosource, catalog MBS756420) according to the manufacturer’s protocol. Briefly, 100-μL samples (cell lysates or serum) or standards were added to the well, followed by addition of 50 µL of the conjugate. The plate was incubated at 37°C for 1 h. Each well was washed five times in 200 µL of washing buffer and then added to 50 µL each of substrates A and B. After 15-min incubation, 50 µL of stopping buffer was added and the absorbance at 450 nm was measured using the microplate reader (Bio-Rad, Hercules, CA). The concentration of BH4 was calculated using the standard curve method and corrected by the protein concentration of each sample.

Adenovirus (Ad) encoding full-length mouse Tfeb coding sequence (NM_011549.3) was prepared with Gateway Cloning System (Hanbio Biotechnology, Shanghai, China) according to the manufacturer’s protocol. For the in vivo study, Ad-TFEB or Ad-GFP with titer at 10¹⁰ pfu/mouse was delivered to db/m⁺ or db/db mice via tail-vein injection, and experiments were carried out 7 days after injection. To assess autophagic flux, HUVECs were transduced with mCherry-GFP-LC3 double-labeled Ad (Hanbio Biotechnology, Shanghai, China) for 24 h. After changing the medium, the cells were subjected to different treatments, and autophagosomes and autolysosomes were subsequently measured using the Olympus FV1000 confocal microscope.

HUVECs were transduced with Ad-TFEB and Ad-GFP for 24 h. RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany). A total of 2 µg of RNA was dried and sent to Macrogen (Seoul, South Korea) for paired-end RNA sequencing analysis. The reference genomes and the annotation file were from the ENSEMBL database. Bowtie2, version 2.2.3 (http://bowtie-bio.sourceforge.net/bowtie2/), was used to build the genome index, and clean data were then aligned with the reference genome using HISAT2, version 2.1.0 (http://www.ccb.jhu.edu/software/hisat/). The raw data of each sample comprised >6 Gb, with clean data comprising >5.5 Gb. The Q score 30 rates of clean data were >85%. Reads was counted by HTSeq, version 0.6.0 (https://htseq.readthedocs.io/en/release_0.11.1/overview.html), and fragments per kilobase million mapped reads were then calculated to estimate the expression level of genes in each sample.

For enrichment analysis, genes with adjusted P value <0.01 were included and analyzed in the DAVID Functional Annotation Tool (https://david.ncifcrf.gov/summary.jsp). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis result was sorted by P value, and the top 18 pathways were selected for construction of the bubble plot using an online tool (https://www.bioinformatics.com.cn/plot_basic_gopathway_enrichment_bubbleplot_081).

For the oral glucose tolerance test (OGTT), mice were first fasted for 8 h and then administered glucose (1.2 g/kg body weight) via oral gavage and the levels of blood glucose were measured with a glucometer (Ascenia Elite XL; Bayer Corp, Mishawaka, IN) at 0, 15, 30, 60, 90, and 120 min. For the insulin tolerance test (ITT), mice fasted for 2 h were injected intraperitoneally with insulin (0.75 units/kg body weight; catalog I6634, Sigma), and blood glucose levels were determined at 0, 15, 30, 60, 90, and 120 min after injection. The area under the curve was calculated.

All data generated or analyzed in this study are included in the article or in the online Supplementary Material.

Results are reported as mean ± SD of a given number of separate experiments. The relative protein expression was normalized to the level of GAPDH or β-actin and quantified using Quantity One software (Bio-Rad). The statistical analysis and calculations were performed using Prism, version 8 (GraphPad Software, San Diego, CA). Normality and equal variance tests were performed for all variables before additional analysis. For data that passed the equal variance test, an unpaired two-tailed Student t test was used to analyze differences between two groups; one- or two-way ANOVA followed by the Tukey multiple comparison test was used for comparing multiple groups. If a normal distribution was not achieved, the Mann-Whitney test was used for the two independent groups and Kruskal-Wallis test followed by Dunn multiple comparison test was used for comparing three or more groups. P < 0.05 was considered statistically different.

To investigate autophagic flux in aortas of diabetic mice, we determined the expression of autophagy markers LC3-II and p62. Compared with that of nondiabetic db/m⁺ mice, the expression of LC3-II and p62 was increased in aortas of male diabetic db/db mice (Fig. 1A and B). To observe whether the increased p62 protein expression was within ECs, we performed en face immunofluorescence staining, which showed that the p62 signal was elevated in aortic ECs of male db/db mice (Supplementary Fig. 1A and B). Furthermore, we isolated lung ECs from female db/m⁺ and db/db mice. Western blotting results showed that LC3-II and p62 levels were increased (Supplementary Fig. 1C and D), indicating that female db/db mice also have altered autophagic flux in ECs (either enhanced autophagosome formation or impaired autophagic flux).

To further assess the autophagic activity, we performed the turnover assay of LC3 and p62 and found that treatment with the autophagic flux inhibitor CQ (10 µmol/L, 6 h) increased the content of both LC3-II and p62 in aortas of male db/m⁺ mice but not in aortas of db/db mice (Fig. 1C and D), suggesting an impaired autophagic flux in diabetic mouse aortas. These results indicate the dysregulated autophagic flux in ECs of diabetic mouse arteries.

AGEs and ox-LDL are the chemical risk factors that are positively associated with the severity of diabetes and participate in mediating diabetic endothelial dysfunction (30,31). To evaluate their effects on autophagic flux, HUVECs were treated for 24 h with AGEs (100 µg/mL) or ox-LDL (100 μg/mL), and the relative abundance of autophagosomes and autolysosomes were determined using mCherry-GFP tandem-tagged LC3 Ad (13). Both AGEs and ox-LDL increased the abundance of autophagosomes (yellow dots in Fig. 1E and F). Consistently, AGEs and ox-LDL increased the levels of LC3-II and p62 (Fig. 1G). These results indicate that autophagosome clearance is defective in ox-LDL- or AGEs-treated HUVECs.

Concomitant with impaired autophagic flux, eNOS monomerization was increased in male db/db mouse aortas and in ox-LDL- or AGEs-exposed HUVECs (Fig. 1A and G–I). To investigate whether inhibition of autophagic flux has direct impact on eNOS dimerization, we treated HUVECs with CQ (10 µmol/L) or BafA1 (10 nmol/L) for 3 h or 12 h. Both CQ and BafA1 decreased the protein level of eNOS dimer or the ratio of dimers to monomers (Fig. 2A–C). Consistently, ex vivo incubation of db/m⁺ mouse aortas individually with 3-MA, CQ, or BafA1 attenuated ACh-induced EDRs (Fig. 2D), whereas endothelium-independent relaxations in response to the NO donor sodium nitroprusside were comparable among groups (Supplementary Fig. 2A).

Previous studies demonstrated that ROS-induced BH4 deficiency is a key determinant of eNOS monomerization (32). H₂O₂ is sufficient to reduce dimer and increases the monomer level of eNOS, which can be reversed by pretreatment with Mito TEMPO (100 µmol/L) or BH4 (100 µmol/L) (Supplementary Fig. 2B), suggesting that ROS is both sufficient and required for conversion of eNOS dimer to its monomers. Furthermore, the serum level of BH4 in both male and female db/db mice was lower than db/m⁺ mice (Fig. 1J), whereas the mtROS level was higher in db/db mouse aortic endothelium (Supplementary Fig. 2C and B), suggesting that reduced BH4 level might be the consequence or acts as a contributor to the increased ROS in vascular endothelium of diabetic mice. Ex vivo supplementation with Mito TEMPO (100 µmol/L) or MitoQ (10 nmol/L) rescued the impaired EDR in male db/db mouse aortas (Supplementary Fig. 2E). Ox-LDL and AGEs also enhanced the production of mtROS in HUVECs (Supplementary Fig. 2F), and AGEs-induced impairment of EDR was reversed by Mito TEMPO (Supplementary Fig. 2G). These results further demonstrate the crucial role of mtROS in mediating diabetic endothelial dysfunction (33).

Treatment with an autophagic flux inhibitor, CQ, increased mtROS formation and decreased NO production, as measured by MitoSOX and DAF-DA staining, respectively (Fig. 2E–G). The inhibitory effect of CQ on NO production was reversed by Mito TEMPO (Fig. 2E–G). More importantly, BafA1-induced eNOS dimer dissociation can be reversed by Mito TEMPO (100 µmol/L), BH4 (100 µmol/L), and N-acetyl-l-cysteine (2.5 mmol/L), suggesting that autophagy inhibition–induced eNOS dimer dissociation is most likely mediated by ROS (Fig. 2H and I). Taken together, these results demonstrate that inhibition of autophagic flux leads to increased production of mtROS, which reduces eNOS activity by disassociating eNOS dimers into monomers to mediate endothelial dysfunction in diabetic mice.

TFEB, a member of the basic helix-loop-helix leucine-zipper family of transcription factors, is the master regulator of lysosomal biogenesis and autophagic flux (34). To explore whether induction of autophagic flux by TFEB facilitates eNOS dimerization, we overexpressed TFEB in HUVECs for 24 h using Ad. KEGG enrichment analysis showed that TFEB overexpression was closely associated with AGE-RAGE signaling in diabetic complications, shear stress and atherosclerosis, PI3K-Akt-mTOR signaling, and autophagy, as well as lysosomal biogenesis (Fig. 3A). Heat map data revealed that the expression of autophagy and lysosomal biogenesis markers (BCL2, SQSTM1, MAP1LC3B, and CTS(s)) was increased by TFEB (Fig. 3B).

To explore the effect of TFEB-driven autophagic flux on the production of ROS, we used MitoSOX and dihydroethidium staining to determine ROS generation in mitochondria and cytosol and showed that TFEB overexpression suppressed MitoSOX and dihydroethidium fluorescence intensity, confirming the antioxidative effect of TFEB in HUVECs (Fig. 3C, Supplementary Fig. 3A–C). TFEB overexpression further increased BafA1-induced elevation in LC3-II and p62 levels, suggesting TFEB-enhanced autophagic flux. More importantly, TFEB attenuated BafA1-induced dissociation of eNOS dimers and CQ-induced reduction of the eNOS dimer to monomer ratio (Fig. 3D and E). These results indicate that TFEB-mediated autophagic flux can attenuate eNOS dimer dissociation at least partially by lowering the cellular ROS level.

In our previous study, we showed that TFEB expression and nuclear localization were reduced in db/db mouse aortic endothelium (Supplementary Fig. 5A and B) (26). In the present study, we found that AGEs and ox-LDL induced a marked reduction of TFEB nuclear localization (Supplementary Fig. 5C–E), which may partially explain the cytosolic retention of TFEB in db/db mouse arterial ECs. To determine whether TFEB can restore the impaired autophagic flux in db/db mice, we overexpressed TFEB in male db/db mice for 1 week via tail-vein injection of Ad-TFEB. TFEB was successfully overexpressed in db/db mouse aortas, as evidenced by Western blotting and immunofluorescence staining (Fig. 4A and B, Supplementary Fig. 5F). TFEB overexpression decreased p62 level in aortas of db/db mice, indicating the preservation of autophagic turnover activity (Fig. 4A and C, Supplementary Fig. 6A and B). In addition, TFEB overexpression inhibited mtROS production in aortic ECs of male and female db/db mice (Fig. 4D and E, Supplementary Fig. 6C and D). The impaired EDRs in male and female db/db mice were also reversed by Ad-TFEB administration (Fig. 4F and Supplementary Fig. 6E). Collectively, these results demonstrate that restoration of autophagic flux by overexpressing TFEB is effective in lowering ROS and rescuing endothelial function in diabetic mice of both sexes. TFEB may be a target for pharmaceutical intervention to preserve endothelial autophagic flux and function in diabetes.

TFEB activity is tightly modulated by nutrient availability through posttranslational modification. Under nutrient-rich conditions, TFEB is phosphorylated by mTORC1 at serine 142 and serine 211 residues and is primarily localized in the cytosol for inactivation. Under starvation conditions, suppressed mTORC1 activity induces TFEB dephosphorylation, resulting in TFEB nuclear translocation and enhanced autophagic flux (25). The protein level of p-mTOR (S2448) was increased in male db/db mouse aortas compared with db/m⁺ mouse aortas (Supplementary Fig. 7A and B), suggesting that the mTOR pathway was hyperactivated. Rapamycin and torin1 are widely used mTORC inhibitors that mimic the effects of calorie restriction, which extends life span in yeast, worms, flies, and mice (35).

Next, we examined whether mTOR inhibition could promote TFEB activity and rescue endothelial function in diabetes. Indeed, rapamycin treatment reversed AGEs-induced increase of p-mTOR (S2448) and p-TFEB (S142) (Supplementary Fig. 7C and D). Rapamycin prevented AGEs- or ox-LDL-induced mtROS production in HUVECs (Supplementary Fig. 7E and F) and restored the impaired EDR in AGEs-treated male db/m⁺ mouse aortas, whereas the autophagy inhibitor CQ counteracted the effect of rapamycin (Fig. 5A). By contrast, endothelium-independent relaxations were comparable among groups (Supplementary Fig. 7G). Moreover, inhibition of mTOR by rapamycin (300 nmol/L, 12 h) or torin1 (200 nmol/L, 12 h) elevated the level of BH4, enhanced autophagic flux, increased eNOS dimerization, and the ratio of dimers to monomers in HUVECs, further suggesting that the mTOR-TFEB-autophagic flux cascade plays critical roles in regulating eNOS dimerization (Fig. 5B–D, Supplementary Fig. 7H). Consistent with both in vitro or ex vivo results, 12 days of administration of rapamycin (2 mg/kg every 2 days) to male db/db mice lowered the aortic p62 accumulation (Fig. 5E and F), suppressed endothelial mtROS production (Fig. 5G and H) and restored the impaired EDR (Fig. 5I) without affecting SNP-induced relaxations (Supplementary Fig. 8A) in db/db mouse aortas. These beneficial effects of mTOR inhibition were similar to those induced by TFEB overexpression. OGTT (Supplementary Fig. 8B and C) and ITT (Supplementary Fig. 8D and E) data showed that rapamycin did not affect glucose or insulin tolerance in diabetic mice, suggesting that the beneficial effect of rapamycin was independent of blood glucose level. Taken together, these results reveal that activation of TFEB through inhibiting mTOR improves eNOS dimerization and endothelial function in diabetic mice by enhancing autophagic flux and inhibiting mtROS production.

Calorie restriction is an increasingly popular means of losing weight to reduce cardiovascular risk (36). A previous study showed that 3 cycles of short-term fasting-mimicking diet, improved insulin secretion and glucose homeostasis in type 1 and type 2 diabetes (28). The induction of autophagy and autophagic flux is one of the most crucial mechanisms of calorie restriction to protect against diabetic vascular complications (37). Meanwhile, fasting-induced autophagy is mainly mediated by TFEB activation due to mTOR suppression. However, whether calorie restriction–induced TFEB activation and autophagic flux can ameliorate diabetic vascular dysfunction is still unclear. To address this question, male db/db mice were given 3 cycles of calorie restriction; each contained a 4-day fasting-mimicking diet followed by a 7-day ad libitum feeding (Fig. 6A). Body weight and blood glucose levels were reduced by the end of the second cycle (Supplementary Fig. 9A and B). After 3 cycles, the impairment of EDR in db/db mouse aortas was significantly attenuated by calorie restriction (Fig. 6B). Meanwhile, db/db mouse aortas showed reduced p62 and increased LC3-II and TFEB, indicating an enhancement in autophagic turnover capacity (Fig. 6C and D).

In addition, mitochondrial dynamic- and quality control-related proteins, such as sirtuin 1 and mitofusion 2 were increased, whereas mitochondrial fission-related protein dynamin-1-like protein was reduced by calorie restriction (Fig. 6E and F). These results indicate that dietary intervention by calorie restriction is another effective approach to rescue endothelial function in diabetes partially through inducing a TFEB-autophagic flux cascade that improves mitochondrial quality in ECs.

In this study, we demonstrated that autophagic flux is impaired in arterial ECs of diabetic mice and in HUVECs treated with AGEs or ox-LDL. We show that 1) impaired autophagic flux increases mtROS formation and decreases eNOS dimerization, leading to a reduced NO production and endothelial dysfunction in diabetes; 2) TFEB overexpression restores autophagic flux, suppresses mtROS generation, and rescues endothelial function in diabetic mice; and 3) activation of TFEB by either mTOR inhibition or intermittent fasting enhances autophagic flux, decreases ROS level, and improves endothelial function in diabetic mice. Taken together, the results show that restoration of autophagic flux by targeting TFEB is an effective therapeutic strategy against diabetes-associated vascular dysfunction (Fig. 6G).

Adequate production and function of eNOS-derived NO maintains endothelial function in healthy blood vessels. Normal catalytic activity of eNOS requires the formation of its homodimer and availability of substrate and cofactors (e.g., l-arginine, BH4, and NADPH) (3). Previous studies showed that in prediabetes or in patients and rodents with diabetes, total eNOS protein level was unchanged or increased, whereas NO production and EDR were impaired, indicating an uncoupling of eNOS (8,38–40). The major cause of eNOS uncoupling is the BH4 insufficiency, a condition that facilitates eNOS monomerization and its proteasome degradation (6,8,41). As an electron donor, BH4 is readily oxidized to BH2 by increased ROS, which primarily originate from dysfunctional mitochondria (42,43). Reducing ROS production by improving mitochondria quality or scavenging ROS by enhancing antioxidant capacity in ECs are effective to preserve endothelial function in diabetes (43,44).

Mitochondria are the major source of ROS in ECs. High levels of glucose, AGEs, ox-LDL, and free fatty acids inhibit mitochondrial biogenesis and impair mitochondria renewal, thereby increasing the content of dysfunctional mitochondria. Intact autophagy or autophagic flux plays crucial roles in mitochondria quality control (44). Impaired endothelial autophagic flux is reported in arterial aging and atherosclerosis (14,19). In the present study, db/db mouse aortas contained higher levels of LC3-II and p62, suggesting either the activation of autophagy or inhibition of autophagosome degradation. An additional p62 turnover assay showed that the autophagy inhibitor CQ increased LC3 and p62 levels in aortas of db/m⁺ mice but not in those of db/db mice, indicating that autophagic flux is impaired in diabetic mouse aortas. Incubation of HUVECs with AGEs and ox-LDL, which are well-established diabetes risk factors, also reduced the autophagosome clearance capacity and increased ROS generation. Concomitantly, db/db mouse aortas and AGEs- and ox-LDL-treated HUVECs had more eNOS monomer and reduced dimer to monomer ratio, indicating that impaired autophagic flux is likely associated with eNOS monomerization and suppressed eNOS activity in ECs under diabetic conditions. To test this likelihood, we treated HUVECs and db/m⁺ mouse aortas with the autophagy inhibitors CQ or BafA1 and measured the eNOS dimer level and endothelial function. Both CQ and BafA1 decreased eNOS dimerization, increased mtROS generation, reduced NO production in HUVECs, and inhibited EDR in mouse aortas. More importantly, inhibition of mtROS by Mito TEMPO was sufficient to reverse BafA1-induced eNOS dimer dissociation and to rescue endothelial function in diabetic mouse aortas or in AGEs-treated nondiabetic mouse aortas, suggesting that ROS is indispensable in autophagic flux deficiency–induced eNOS inactivation and endothelial dysfunction. It is of interest to note that serum BH4 levels were lower in db/db mice of both sexes, compared with db/m⁺ mice, suggesting that BH4 insufficiency may participate in diabetic endothelial eNOS monomerization in both sexes.

Collectively, these results clearly indicate that under diabetic conditions, the defective autophagic flux leads to overproduction of mitochondrial ROS; the latter reduces eNOS dimerization and activity, evolving in endothelial dysfunction. Therefore, enhancing autophagic flux in vascular endothelium may represent as a new approach to suppress ROS and associated eNOS dimer dissociation, thereby improving vascular function in diabetes.

Recent studies have revealed a pivotal role of TFEB in cardiovascular health through maintaining autophagic flux (21,45–47). In our previous study, we demonstrated that TFEB inactivation partially mediated the increased inflammation in db/db diabetic mouse arteries (26). However, it was unknown whether defective the TFEB-autophagic flux-eNOS cascade participated in endothelial dysfunction in diabetes. The present study shows a reduced TFEB nuclear localization in aortic ECs of db/db mice and in HUVECs exposed to AGEs or ox-LDL. Both in vivo and in vitro TFEB overexpression effectively increase autophagic flux, lower mtROS, reverse CQ- or BafA1-induced eNOS dimer dissociation, and attenuate endothelial dysfunction in db/db mice. These rescue experiments clearly demonstrated that defective autophagic flux closely associated with TFEB downregulation is likely responsible for diabetic endothelial dysfunction of male and female db/db mice. However, the overexpression TFEB using Ad is not restricted to the ECs (Supplementary Fig. 5F). Therefore, we cannot exclude that the beneficial effects of TFEB overexpression in vivo were contributed, at least in part, by other cell types (e.g., smooth muscle cells and fibroblasts), which is a limitation of our study. It is of interest to note that the inhibitory effect of TFEB on the expression of vascular cell adhesion molecule-1 and monocyte chemoattractant protein-1 is still preserved in the presence of CQ or BafA1 (Supplementary Fig. 4A), suggesting that the anti-inflammatory effect of TFEB appears to be independent of its role in autophagic flux. TFEB activity is tightly regulated by the nutrition sensor mTOR (24). In line with previous studies, we also observed that mTOR is hyperactivated, whereas TFEB nuclear localization is reduced in the aortas of db/db mice and in AGEs-treated HUVECs. Inhibition of mTOR by rapamycin or torin1 suppresses mtROS, enhances autophagy and autophagic flux, increases eNOS dimerization, and attenuates db/db aortic dysfunction. Taken together, these observations indicate that induction of autophagic flux by TFEB overexpression or through mTOR inhibition-associated TFEB activation is effective in preserving eNOS dimerization and endothelial function in diabetes. Although the effect of rapamycin in inducing autophagy is largely mediated through TFEB, TFEB-independent mechanisms exist (48). Therefore, in this study, we could not exclude that the effect of rapamycin on eNOS dimerization may be partially TFEB independent.

Diet intervention such as calorie restriction or intermittent fasting is a well-accepted, safe approach to delay aging, increase longevity, and prevent metabolic syndrome–associated cardiovascular diseases (36). Inhibition of autophagy attenuates the antiaging effect of calorie restriction, indicating a significant role of autophagy in mediating the health benefit of calorie restriction (37). mTOR inhibition and TFEB activation are crucial mediators of calorie restriction–induced autophagy. In this study, we showed that calorie restriction inhibited the gain of body weight, lowered blood glucose level, and improved endothelial function in diabetic mice. Calorie restriction resulted in increased autophagic flux, as supported by elevated protein levels of TFEB and LC3-II accompanied by reduced p62 level in mouse aortas. In addition, the increased expression of mitochondrial quality control–associated markers suggests that calorie restriction leads to improved mitochondrial function in diabetic mouse aortas.

The lack of experimental data from endothelium-specific transgenic mice is a limitation of this study. The important role of endothelial TFEB in reducing inflammation and atherosclerosis (49), promoting postischemic angiogenesis (46), and improving glucose tolerance (50) has been clearly demonstrated in recent studies. However, we did not use either endothelium-specific knockdown or overexpression of TFEB to confirm the role of endothelial TFEB in increased eNOS dimerization in this study. Therefore, we cannot rule out a possible minor involvement of TFEB expressed in other types of vascular cells to the improved EDRs in diabetic mice, although the expression of both eNOS and TFEB in mouse aortas are primarily confined to the luminal endothelial layer of mouse aortas, as indicated by immunofluorescence staining and Western blot results in the present study and our previous study (26).

In this study, we demonstrated a causal relationship between impaired autophagic flux and endothelial dysfunction in diabetic mice. This may be the first study to show a critical role of impaired autophagic flux–induced mtROS accumulation and subsequent eNOS dimer dissociation in diabetic endothelial dysfunction. Induction of TFEB-driven autophagic flux by inhibiting mTOR or through diet intervention is effective for treating diabetes-associated endothelial dysfunction.

Funding. This study is supported by the Hong Kong Research Grants Council (grants C4024-16W, SRFS2021-4S04, 14124216, AoEM-707/18, R4012-18) and the Natural Science Foundation of China (grants 91939302, 82000056).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.Y., W.L., and Z.C.L. contributed to the conceptualization and design of the study. Data curation was performed by Z.L., Z.C.L., H.L., C.Q., L.L., K.L., L.J.Y., L.J., W.L., G.L., S.W., Q.D., and L.C.W. Resources were provided by H.Y., L.L., K.H., M.V., and T.X.Y. Research funding was provided by H.Y. and Z.C.L. H.Y. also contributed to study supervision and revision of the original draft. Z.L. and Z.C.L. contributed to the formal analysis, investigation, methodology, and writing original draft. H.Y. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.