Induction of KLF2 by Exercise Activates eNOS to Improve Vasodilatation in Diabetic Mice Free

Vascular complications of diabetes predispose the development of atherosclerosis that leads to disability and death (1). Endothelial dysfunction, an initial pathological event during atherogenesis, is closely associated with diabetes (2). The current recommended interventions to improve endothelial function in diabetes include physical exercise and drug treatment with statins, metformin, and angiotensin type 1 receptor antagonists (3–6).

Krüppel-like factor 2 (KLF2) is a transcription factor induced by laminar flow and statins in endothelial cells (ECs) (7,8). By contrast, KLF2 expression is inhibited by proinflammatory cytokines and atheroprone disturbed flow (7,9). A large number of vasoprotective genes induced by laminar flow are reported to be KLF2-dependent (10,11), thus explaining an antiatherosclerotic property of KLF2 via its anti-inflammatory and antioxidant effects in ECs (12,13). Of note, risk factors that inhibit KLF2 expression, such as proinflammatory cytokines and disturbed flow, also impair endothelial function, suggesting that KLF2 downregulation is likely to be a mechanism underlying endothelial dysfunction in diabetes. However, the precise role of KLF2 in the development of diabetic endothelial dysfunction is largely unclear, and whether augmenting KLF2 expression/activity improves endothelium-dependent relaxation (EDR) in diabetic mice remains underexplored.

Endothelial nitric oxide (NO) synthase (eNOS) mediates NO production, endothelial function, and vascular tone regulation. In diabetic rodent vasculature, increased oxidative stress and inflammation cause eNOS dysfunction manifested by reduced phosphorylation of eNOS at serine 1177 (p-eNOS S1177) and uncoupled eNOS (14,15). Instead of producing NO, the uncoupled eNOS releases superoxide anion (16). The fully functional eNOS requires its coupling with cofactors such as tetrahydrobiopterin (BH4) (16). Sufficient cellular level of BH4, mainly synthesized de novo through the rate-limiting guanosine triphosphate cyclohydrolase I (GCH-1), promotes eNOS dimerization for efficient NO production (17). The association with Hsp90 facilitates agonist- and fluid shear stress-induced eNOS activation (18), and Akt-mediated p-eNOS S1177 greatly enhances eNOS activity and NO generation (19,20). The transcription of the eNOS gene can be increased by KLF2, and fluid shear stress-mediated upregulation of eNOS expression is KLF2-dependent (11,21). Nevertheless, whether KLF2 is protective in preserving eNOS function under diabetic conditions remains basically unknown.

The current study shows that KLF2 overexpression improved endothelial function through increasing NO bioavailability in diabetic db/db mice. Mechanistically, the KLF2-mediated transcriptional program creates a signaling network to restore eNOS activity, which is partially mediated by phosphoinositide 3-kinase (PI3K)–Akt signaling and Hsp90. The present results suggest that upregulation of the KLF2 level is effective to rescue eNOS activity and thus preserve endothelial function in diabetic mice.

The animal use and experimental protocols were approved by the Laboratory Animal Experimentation Ethical Committee (Ref no. 16-029-MIS), Chinese University of Hong Kong. Male and female C57BL/6J, db/m⁺ and db/db mice (10–16 weeks old) were provided by Chinese University of Hong Kong Laboratory Animal Service Centre, housed in a temperature-controlled room (22–24°C) with a 12-h light/dark cycle, fed with standard research diet, and had free access to water.

Male and female mice were both trained to run on a motorized treadmill. For C57BL/6J mice, the running speed was initially set at 8 m/min for 30 min and then gradually increased to the target speed at 12 m/min for 30 min, for 7 days. For db/db mice, the initial running speed was set at 5 m/min for 30 min and then increased to 8 m/min for 30 min, 6 days per week, for 4 weeks (5). The sedentary mice were placed on a nonmoving treadmill for the same duration as the exercising mice. For simvastatin treatment, 12-week-old male db/db mice were orally administered with the drug (40 mg/kg/day) for 4 weeks before sacrifice for collection of blood vessels for examination.

The culture of human umbilical vein ECs (HUVECs) from Lonza (CC-2519), human aortic ECs (HAECs) from Thermo Fisher Scientific (C0065C), and HEK293A/T cells from ATCC were performed using previously reported method (22).

Full-length human KLF2 coding sequence was cloned into a pAdTrack-cytomegalovirus vector and packaged to recombinant adenovirus using an AdEasy system (23). Briefly, the recombinant construct was transfected to HEK293A cells for virus package. The condensation of adenovirus (Ad) was achieved by polyethylene glycol 6000 precipitation before use in cell culture. Purification of Ad for animal studies was achieved by using the commercial kit Adeno-X Maxi Purification Kit (Clontech, Mountain View, CA). Ad-KLF2, Ad-Cdh5-Klf2, or Ad-green fluorescent protein (GFP) (10¹⁰ plaque forming units/mouse) were administered to mice via tail vein injection for 7 days. The EC-specific Klf2 overexpression virus Ad-Cdh5-Klf2 (Mouse Klf2 coding sequence) and human KLF2 knockdown virus Ad-shKLF2 were generated using the same method.

The mice were sacrificed, and the vessels were rapidly removed and placed in oxygenated ice-cold Krebs-Henseleit solution. Aortas and main mesenteric arteries were carefully dissected out and cut into ∼2-mm-long segments. Two segments from each mouse were amounted to a Multi Wire Myograph System (Danish Myo Technology, East Jutland, Denmark) to measure isometric tension. Each arterial ring was first precontracted by 60 mmol/L KCl to test viability. Phenylephrine (Phe) at 1 μmol/L was used to contract the rings, followed by cumulative addition of acetylcholine (ACh) to cause endothelium-dependent relaxations (24). Endothelium-independent relaxations to sodium nitroprusside (SNP), an NO donor, were determined in rings pretreated with a NOS inhibitor l-N^G-nitro-l-arginine methyl ester (l-NAME; 100 μmol/L).

The segments of second-order mouse mesenteric arteries were dissected and cannulated between two glass cannulas. The intraluminal pressure and vessel diameter were monitored by a light-inverted microscopy with a video camera and the MyoVIEW software (Danish Myo Technology), as previously described (24). Briefly, Phe at 3 μmol/L was added to induce constriction after the vessel’s diameter stabilized, and flow-mediated dilatation (FMD) was triggered by pressure change that equals 15 dynes/cm² shear stress. Passive dilatation was obtained at the end of the experiment by changing the bathing solution to Ca²⁺-free Krebs solution with 2 mmol/L EGTA. FMD was calculated as the percentage of diameter changes: (flow-induced dilatation-Phe tone)/(passive dilatation-Phe tone).

The mice were euthanized, and aortas were dissected and placed in ice-cold Krebs solution. Each aorta was cut into several ring segments before being cultured in DMEM supplemented with 40% serum collected from db/m⁺ and db/db mice (10). Ad-GFP and Ad-Cdh5-Klf2 were added to the culture plate to infect the aortic rings for 16 h before mounting to a wire myograph for isometric force measurement.

Laminar shear stress (12 dynes/cm²) was generated using a previously reported method (25). Briefly, 5 × 10⁵ HAECs were seeded onto fibronectin (50 μg/mL)-coated glass slides (75 mm × 38 mm; Corning) and cultured in Medium 200 supplemented with 10% FBS. During the laminar shear stress experiment, the culture medium was replaced with Medium 200 supplemented with 2% FBS, and the cells were exposed to shear force for 24 h before harvest.

The quantitative PCR analysis was performed on the ViiA7 Real-Time PCR system (Applied Biosystems) (22). Briefly, samples were homogenized and lysed in TRIzol reagent, and 1 μg total RNA was reversely transcribed using RevertAid Reverse Transcriptase (Thermo Scientific, no. EP0442) to synthesize cDNA. Gene expression was analyzed by the comparative C_T method and expressed as the value relative to GAPDH. Please refer to Supplementary Table 1 for the DNA sequences of primers.

For immunoprecipitation, cells were homogenized in NP-40 lysis buffer containing protease inhibitors and phosSTOP phosphatase inhibitors (Roche Life Sciences, 04906845001). The lysate was precleared by Sepharose beads (GE Healthcare, 17088601) before incubation overnight with anti-Hsp90 or anti-eNOS antibodies. After incubation with Sepharose beads at 4°C for 2 h the next day, the immune complex was precipitated, washed, and boiled in 1 × denaturing Laemmli sample buffer for 10 min before electrophoresis. For Western blotting, cells or tissues were homogenized in radioimmunoprecipitation assay lysis buffer. The proteins were separated by SDS-PAGE and probed with primary antibodies at 4°C overnight. The polyvinylidene fluoride membranes were washed and subsequently incubated with corresponding secondary antibodies conjugated with horseradish peroxidase (DakoCytomation California, Inc.) before being developed with enhanced chemiluminescence detection on X-ray films. The antibody information is provided in the Supplementary Table 2.

Low-temperature SDS-PAGE was used for detection of eNOS dimer and monomer. Briefly, protein samples were incubated with loading dye without β-mercaptoethanol at 37°C for 5 min before loading to 7.5% SDS-PAGE gel preequilibrated in 4°C electrophoresis buffer. The blots were probed with anti-eNOS and anti-GAPDH antibodies, followed by horseradish peroxidase-conjugated secondary antibody.

En face staining of reactive oxygen species (ROS) generation using dihydroethidium (DHE, Ex515/Em585 nm) was performed in thoracic aortas from db/db mice according to the previously described method (26). Briefly, freshly isolated aortic rings were incubated with 5 μmol/L DHE for 15 min at 37°C in extracellular medium. The aortic rings were cut open longitudinally, and the endothelium was placed upside down between two coverslips before being captured on a confocal microscope (FV1000, Olympus, Tokyo, Japan).

Frozen sections of mouse aortas were used for the immunohistochemistry for KLF2. Antigen retrieval was performed in 0.01 mol/L citrate buffer at pH 6.0. After blocking by 5% BSA for 2 h at room temperature, the tissue sections were incubated with the primary antibody anti-KLF2 (1:500; Millipore, 09-820) at 4°C overnight. The signals were developed by using UltraSensitive SP kit (Fuzhou MAXIM Biological Technology Development Co., Ltd., Fujan, China) according to the manufacturer’s instructions. The nuclei were counterstained in hematoxylin.

NO content was presented as its metabolite, and the nitrite level was measured by a Griess assay kit (Life Technologies) according to manufacturer’s instructions. Briefly, HUVECs were pretreated with Ad-GFP and Ad-KLF2 for 24 h, and culture medium was replaced with serum-free endothelial growth medium containing A23187 (1 μmol/L) for 16 h. Absorbance was measured in a spectrophotometric microplate reader at wavelength of 548 nm, and the data were analyzed and converted to nitrite concentrations by comparing calibration readings. For measurement of nitrite level in aortas, the db/db mouse aortas were treated with ACh (10 μmol/L) for 10 min, followed by incubation with nitrate reductase to reduce nitrate to nitrite. Aortas were thereafter homogenized, and the supernatants were collected for quantification of total nitrite level by using the kit. The protein content of tissue homogenates was determined by the Bradford method and used to normalize the nitrite values.

Results represent means ± SEM of n separate experiments. Gene and protein expression were normalized to the expression level of GAPDH and then expressed as relative to the control. Concentration-response curves were constructed using GraphPad Prism 6 software (GraphPad Software, San Diego, CA). Unpaired t test was used to compare differences between two groups, and one-way ANOVA was used to analyze dose-response curves, followed by the Dunnett posttest. P < 0.05 was considered as statistically different.

All analyzed data in this study are included in this article (and its online supplementary files). Resources available upon request.

We isolated aortas from db/m⁺ and db/db mice and determined levels of KLF2 along with phosphorylated and total eNOS. The results showed that the protein level of eNOS was not changed, but the levels of KLF2 protein and p-eNOS S1176 were reduced in aortas from db/db mice (Fig. 1A). We then mechanically denuded the endothelium of aortas of db/m⁺ and db/db mice, as evidenced by a marked reduction of the Nos3 gene, which encodes eNOS (Supplementary Fig. 1A). The results showed that the level of Klf2 mRNA was significantly lower in endothelium-intact aortas from db/db mice compared with those from db/m⁺ mice (Fig. 1B). However, the Klf2 mRNA levels in endothelium-denuded aortas from both strains of mice were comparable (Fig. 1B), suggesting that downregulation of Klf2 mRNA in db/db mouse aortas is likely caused by reduced endothelial expression of KLF2. Diabetes increases circulating proinflammatory mediators such as interleukin 1-β (IL-1β) and advanced glycation end products (AGEs) (27). To test whether diabetic risk factors inhibit KLF2 expression, we treated HAECs with oxidized LDL (Ox-LDL), lysophosphatidylcholine (LPC), IL-1β, and AGEs and showed that KLF2 mRNA expression was suppressed by these risk factors (Fig. 1C). KLF2 expression in ECs is responsive to shear stress, but whether its expression can be upregulated by running exercise remains unclear. We next subjected C57BL/6J mice to treadmill exercise and found, as compared with sedentary mice, 1-week running exercise induced mRNA expression of Klf2 and Nos3 (a known KLF2 downstream target) but not Cnn1 (a smooth muscle marker) in mouse aortic endothelium (Fig. 1D and E and Supplementary Fig. 1B), suggesting that endothelial KLF2 expression and KLF2-mediated transcription is exercise-responsive. Additionally, db/db mice receiving 4-week exercise or simvastatin treatment showed increased expression of Klf2 mRNA in their aortas (Fig. 1F and G), and exercise also increased p-eNOS S1176 and KLF2 protein levels in db/db mouse aortas (Fig. 1H and K). These results demonstrate that reduced endothelial KLF2 expression in diabetes can be reversed by exercise and simvastatin, indicating that the upstream pathway controlling KLF2 expression remains responsive to vasoprotective signals under diabetic conditions.

Given that KLF2 expression is reduced in diabetic mouse aortic endothelium, we used Ad-mediated overexpression of KLF2 (Ad-KLF2) to determine whether KLF2 upregulation is sufficient to improve the impaired EDR in diabetic mice. Ad-KLF2 administration to db/db mice for 7 days does not affect body weight, insulin sensitivity, and lipid profile (Supplementary Fig. 2). Compared with the Ad-GFP–treated group, Ad-KLF2 increased KLF2 mRNA expression (Supplementary Fig. 3A) and protein level of p-eNOS S1176, eNOS, and KLF2 in db/db mouse aortas (Fig. 2A and Supplementary Fig. 3C–E), indicating the successful overexpression of KLF2. Ad-KLF2 also improved ACh-induced EDR in db/db mouse aortas (Fig. 2B), while SNP-induced endothelium-independent relaxations were similar in aortas from both groups (Fig. 2C).

Endothelium-derived relaxing factors (EDRFs) include NO, prostacyclin I₂, and endothelium-derived hyperpolarizing factor (EDHF) in smaller arteries (28–30). To dissect the role of these EDRFs in contributing to ACh-induced relaxation in Ad-KLF2–treated main mesenteric arteries, we used l-NAME, indomethacin, and 20 mmol/L KCl as inhibitors for NOS, cyclooxygenase, and EDHF, respectively. Ad-KLF2 treatment produced an improvement of EDR in the main mesenteric arteries (Fig. 2D), which was not affected by indomethacin or 20 mmol/L KCl but was reversed by l-NAME treatment (Fig. 2E and F). These results show that NO is the primary form of EDRFs in response to KLF2 overexpression to contribute to the improved EDR in mesenteric arteries of db/db mice.

To exclude a possible endothelium-independent effect of KLF2 overexpression using Ad-KLF2, we generated an endothelium-specific overexpression vector Ad-Cdh5-Klf2 (31) and delivered it to db/db mice for 7 days (Fig. 2G and Supplementary Fig. 3B). Similar to the results observed in the Ad-KLF2 experiment, Ad-Cdh5-Klf2 also improved EDR in aortas from db/db mice (Fig. 2H). SNP-induced relaxation was similar in both Ad-GFP– and Ad-Cdh5-Klf2–treated groups (Fig. 2I). To show a direct effect of endothelial KLF2 overexpression on improvement of EDR, ex vivo Ad-Cdh5-Klf2 transduction was performed in aortic rings isolated from db/db mice, and the results showed a marked improvement of EDR in Ad-Cdh5-Klf2–transduced mouse aortas, again without affecting SNP-induced relaxations (Fig. 4H and I). Taken together, these results demonstrate that augmenting endothelial KLF2 level is effective to improve endothelial function, most likely through increasing NO bioavailability in diabetic mouse arteries.

FMD is one of the fundamental mechanisms controlling vascular tone, with NO as an important mediator of flow-induced vasodilatation (31). The impaired FMD in db/db mice was significantly improved by Ad-KLF2 treatment (Fig. 3A–C). Flow-induced p-eNOS S1177 is critical for flow-induced NO production. To determine whether KLF2 is required for flow-induced p-eNOS S1177, we silenced KLF2 using shRNA in HAECs (Fig. 4G and Supplementary Fig. 4A). The Western blotting result showed that KLF2 is required for both the basal and laminar flow-induced p-eNOS S1177 and eNOS in HAECs (Fig. 3D–F).

The ROS level indicated by dihydroethidium staining in the endothelium of db/db mouse aortas was reduced following Ad-KLF2 treatment (Fig. 4A and B). Exposure of HUVECs to high glucose (HG, 30 mmol/L) for 48 h increased cellular ROS level indicated by chloromethyl–2′,7′-dichlorodihydrofluorescein diacetate fluorescence intensity, and 24 h Ad-KLF2 treatment reduced ROS levels in both normal glucose (5.5 mmol/L) and HG conditions (Fig. 4C). Additionally, KLF2 decreased levels of tyrosine nitrated proteins in db/db mouse aortas (Fig. 4G), suggesting attenuated oxidative stress by KLF2 overexpression. To test whether KLF2 promotes eNOS coupling, Western blotting with low-temperature SDS-PAGE under nonreduced condition was performed. The results showed that KLF2 overexpression for 24 h increased levels of both eNOS dimer and monomer in cultured HUVECs (Supplementary Fig. 4B). To determine whether KLF2 promotes eNOS dimerization in vivo, protein samples from db/m⁺ mouse aortas treated with Ad-GFP and db/db mouse aortas treated with Ad-GFP or Ad-Cdh5-Klf2 were detected. Compared with db/m⁺ mouse aortas, the eNOS dimer level was reduced in db/db mouse aortas, which was restored by Ad-Cdh5-Klf2 (Fig. 4D). To examine whether KLF2 increases NO release, A23187 as the calcium ionophore was used as an eNOS activator to induce NO biosynthesis in HUVECs. Ad-KLF2 treatment increased NO production reflected by elevated extracellular nitrite level (Fig. 4E). Consistently, measurement of the nitrite level in KLF2-overexpressed mouse aortas showed that KLF2 increases the nitrite level in these tissues (Fig. 4F).

To reveal the mechanism by which KLF2 enhances eNOS activity and NO production, we analyzed the RNA-sequencing data showing the changes of mRNA expression in HUVECs overexpressed with KLF2 (22). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of KLF2-upregulated genes showed that KLF2 is critical for the regulation of cyclic guanosine monophosphate–protein kinase (PKG)/cAMP-dependent protein kinase (PKA)/insulin pathways (Fig. 5A). Analysis of KLF2-suppresed genes revealed that KLF2 inhibits p53/transforming growth factor-β/FoxO signaling pathways (Fig. 5B), which are activated in diabetic conditions (32–34). Further analysis of KLF2-upregulated genes showed that KLF2 increased the expression of genes (PLCB2, PLCB3, PLCB4, PLCD3, PLCG2) encoding phospholipase C (PLC) isotypes (Fig. 5C, D, and G), whose activation is critical for intracellular Ca²⁺ signaling and diacylglycerol formation (35). KLF2 also induced the expression of INSR (insulin receptor), TEK (angiopoietin-1 receptor), and FLT1 (vascular endothelial growth factor receptor 1) (Fig. 5C, E, and G), whose upregulation can lead to activation of the PI3K-Akt cascade. In addition, KLF2 induces upregulation of genes in the integrin family and genes coding for signaling transducers that lead to activation of PKA (Fig. 5C and F). At the translational and posttranslational level, Ad-KLF2 increased the phosphorylated levels of Akt at serine 473/threonine 308 and eNOS at serine 1177, together with the increased total contents of Akt1 and eNOS (Fig. 5G), thus confirming a stimulatory regulation of Akt-eNOS activity by KLF2 in human ECs.

To determine whether PI3K-Akt activation participates in KLF2-induced p-eNOS S1177, HUVECs pretransduced by Ad-KLF2 were treated for 12 h individually with PI3K inhibitors wortmannin (1 μmol/L) and LY294002 (25 μmol/L) as well as a selective Akt inhibitor triciribine (10 μmol/L). The results showed that KLF2-induced an increase in p-eNOS S1177 but not in the total eNOS level, which was reduced by these inhibitors (Fig. 6A and B and Supplementary Fig. 4C), suggesting the involvement of PI3K-Akt activation in KLF2-induced p-eNOS S1177. The abolishment of p-eNOS at serine 615, a site mainly regulated by activated Akt, by triciribine treatment further supports that Akt is critically involved in KLF2-induced eNOS activation in ECs (Fig. 6B).

The activation of eNOS by extracellular stimuli, such as fluid shear stress, requires the binding of Hsp90 to facilitate eNOS dimer formation (18). To examine whether Hsp90 plays a role in KLF2-induced eNOS activation, we performed Hsp90 immunoprecipitation in KLF2-overexpressed lysate from HUVECs and found that KLF2 promoted the association of eNOS with Hsp90, which can be disrupted by Hsp90 chaperoning function inhibitor 17-N-allylamino-17-demethoxygeldanamycin (17-AAG) (Fig. 6C). Moreover, eNOS immunoprecipitation in KLF2-overexpressed HUVEC lysate also showed that increased association of Hsp90 with eNOS was inhibited by 17-AAG (Fig. 6D). Further analysis using Western blotting showed that KLF2-induced Akt-eNOS activation, GCH-1 expression, and eNOS dimerization were all attenuated by 17-AAG (Fig. 6E and Supplementary Fig. 6), thus supporting a required role of Hsp90 in KLF2-mediated activation of Akt-eNOS signaling.

The current study provides several lines of novel and important findings. First, the endothelial KLF2 level is inducible by running exercise and simvastatin treatment, indicating that targeting KLF2 against diabetic endothelial dysfunction is achievable. Second, KLF2 overexpression improves the impaired EDR and FMD in arteries from db/db mice, thereby revealing a vasoprotective role of KLF2 in diabetic conditions. Third, the correction of eNOS dysfunction, a key abnormality in diabetes, reflected by KLF2-induced increase in both phosphorylation and dimerization of eNOS, is likely the major mechanism responsible for vascular benefits of KLF2 activation. Moreover, KLF2 attenuates oxidative stress in aortic ECs in db/db mice and in HUVECs exposed to HG, suggesting an additional pleiotropic effect of KLF2 in protecting endothelial function in diabetes. Finally, the mechanistic study uncovered that KLF2 regulates the transcription of the genes coding for upstream regulators of eNOS activity, thereby revealing a previously unrecognized role of KLF2 in the regulation of PI3K-Akt and phospholipase C-inositol trisphosphate signaling. Taken together, the protective role of KLF2 in endothelial function in db/db mice indicates the effectiveness of targeting KLF2 for restoration of endothelial function in diabetes.

We first observed KLF2 downregulation in diabetic aortic endothelium and in HAECs in response to diabetes risk factors, showing the likely involvement of KLF2 downregulation in diabetic endothelial dysfunction. KLF2 is a flow-responsive transcription factor that is crucially involved in the maintenance of flow-dependent EC phenotype (11). One key upstream regulator of KLF2 under laminar flow is extracellular signal–regulated kinase 5 (ERK5), whose suppression has been identified as a critical mediator of endothelial dysfunction in diabetes and atherogenesis (36). The involvement of ERK5 in diabetic endothelial dysfunction suggests that KLF2 as a key downstream transcriptional effector of ERK5 might also play important roles in pathogenesis of diabetic endothelial dysfunction.

To investigate whether reduced KLF2 can be corrected by vasoprotective strategies, db/db mice were subjected to running exercise on treadmill and simvastatin treatment. The present results show that endothelial Klf2 mRNA level is increased by exercise and simvastatin and that exercise also upregulates aortic KLF2 protein expression in db/db mice, indicating that enhancement of the KLF2 level is a feasible therapeutic option to rescue endothelial function in diabetes. An unclear question in the current study is whether exercise and statin treatment activate KLF2 through a common mechanism. Previous study found that human cardiac ECs treated with statin show increased activity of mitogen/extracellular signal-regulated kinase-5 (MEK5)-ERK5 signaling (37), which is also activated by exercise in skeletal muscle (38). It is possible that the downstream signaling of these two distinct approaches converges on the MEK5-ERK5 signaling cascade to promote KLF2 expression and activity.

A noticeable abnormality in diabetic vasculatures is the impaired EDR and FMD in arteries. The central enzyme controlling EDR and FMD is eNOS, which is inhibited in diabetic situations, although levels of other EDRFs, such as prostacyclin I₂, were also decreased in diabetic vasculature (39). The present results show that KLF2 overexpression improved both EDR and FMD in the arteries of db/db mice. The eNOS activity is regulated via multiple mechanisms, including phosphorylation (20), S-glutathionylation (40), subcellular localization (18), and coupling (41). The mechanism underpinning KLF2-increased eNOS activity remains elusive. The current study shows that the improved EDR was inhibited by l-NAME, suggesting that KLF2 preserves EDR through enhancing NO bioavailability in diabetes.

Increased oxidative stress in the vascular wall was reported to quench NO and damage vascular function (16). The expression of KLF2 is required for flow-mediated resistance to oxidative stress, and KLF2 promotes nuclear translocation of Nrf2 (11,42). These findings indicate that KLF2 might be able to reduce overproduction of ROS in diabetic endothelium. In diabetes, increased vascular wall oxidative stress uncouples eNOS and induces vascular dysfunction (43). An important finding on eNOS regulation in the current study is that KLF2 increases eNOS dimerization in vitro and in vivo. We found that KLF2 overexpression increased the expression of GCH-1 (Supplementary Fig. 5A–C), the rate-limiting enzyme for de novo synthesis of BH4, which is critical for eNOS dimerization (44,45). This result suggests that KLF2 is a previously unrecognized nuclear effector in the regulation of GCH-1 expression. The Western blotting results showed a difference in KLF2-promoted eNOS dimerization in vitro and in vivo. In cultured human ECs, KLF2 also increased eNOS monomer while KLF2 decreased eNOS monomer in db/db mouse aortas. It is likely that in the presence of other promoting factors of eNOS coupling, such as shear stress, KLF2 in vivo is a determinant mediator of eNOS dimer formation. However, in a cultured condition, no flow is present to facilitate eNOS dimerization. Hence, it is interesting to study whether KLF2 is required for eNOS dimer formation under flow.

To investigate the molecular basis by which KLF2 improves eNOS function, we analyzed the RNA-sequencing data in HUVECs overexpressed with KLF2. The results showed that KLF2-upregulated genes are enriched in the cGMP-PKG–signaling pathway, cAMP-signaling pathway, and insulin-signaling pathway, which are critical upstream regulators of eNOS activity. Further analysis revealed that genes encoding PLC isotypes (PLC-β2–4, PLC-γ2, PLC-δ3) are upregulated by KLF2. Increased expression of PLC facilitates the generation of intracellular inositol 1,4,5-trisphosphate (IP₃) and diacylglycerol (DAG), resulting in Ca²⁺ release and PKC activation (46). IP₃ triggers intracellular Ca²⁺ mobilization from endoplasmic reticulum via interaction with the IP₃ receptor, which plays an essential role in Ca²⁺-dependent eNOS activation in ECs and agonist-induced EDR (47,48). The ability of KLF2 to strongly induce expression of PLC isotypes suggests that KLF2 enhancement is likely to provide an augmented response of vascular endothelium to agonists to improve endothelial function. The other arm of the PLC signaling leads to the activation of PKC by the membrane-retained DAG. Increased PKC activity promotes p-eNOS at threonine 495 (p-eNOS T495) to reduce overall eNOS enzymatic activity (49). We detected p-eNOS T495 level in KLF2-overexpressed HUVECs and found that although KLF2 increased the p-eNOS T495 level, such increase was proportional to the elevated total eNOS level (Supplementary Fig. 5D), suggesting that KLF2 does not increase PKC activity to cause eNOS inhibition, which was confirmed by Western blotting results showing that KLF2 does not affect phosphorylation of PKC-β (Supplementary Fig. 5D) in ECs.

The present results show that KLF2 potently activates Akt to increase p-eNOS S1177, a required step toward eNOS activation. Although the detailed mechanism by which KLF2 activates Akt is not fully resolved in the current study, our results suggest that the KLF2-induced upregulation of growth factor receptors, such as INSR, FLT1, and TEK, is most likely responsible for KLF2-induced Akt activation in ECs. We confirmed that KLF2 increases INSR expression at mRNA and protein levels (Fig. 5E and G), indicating that through regulating insulin receptor expression, KLF2 level is likely critical to endothelial insulin signal transduction. The interaction between Hsp90 and eNOS increases overall enzymatic activity of eNOS. Hsp90 functions as a scaffold for the optimal regulation of eNOS by protein kinases such as Akt (50). Inhibition of Hsp90 reversed KLF2-induced phosphorylation of Akt and eNOS, placing Hsp90 activation at the upstream of the Akt-eNOS axis in response to KLF2 activation. We attempted to understand how KLF2 promotes the Hsp90 activity and showed that KLF2 upregulated genes coding for Hsp90α and Hsp90β without affecting the mRNA level of heat shock factor 1 (HSF1) (Supplementary Fig. 7). The transcription of Hsp90 is mainly controlled by HSF1, and statins were found to activate HSF1 (51,52). The simultaneous activation of HSF1 and KLF2 by statins suggests that posttranslational mechanisms, such as protein-protein interaction between KLF2 and HSF1, are likely to explain the increased Hsp90 activity by KLF2.

In summary, the current study provides several lines of novel and important insights into the regulation of eNOS activity and endothelial function by KLF2 in diabetic mice (Fig. 6F). It is conceivable that pharmacologically targeting KLF2 has potential beneficial effects on preservation of endothelial function in diabetes.

Acknowledgments. The authors thank Dr. Yujie Pu (City University of Hong Kong) for help in obtaining research materials for this project. The authors are also grateful to Dr. Jinghui Dong (Hebei Medical University) for critical comments on the manuscript.

Funding. This study was supported by Research Grants Council of Hong Kong (SRFS2021-4S04, AoE/M/707-18, 14112919, 14109618) and National Natural Science Foundation of China (32241016, 91939302).

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

Author Contributions. J.-Y.L. conceived the project, performed most of the animal and cellular experiments and data analysis, and drafted and revised the manuscript. C.K.C. and L.G. performed mechanistic studies, assisted in the cellular and animal experiments, and made revisions of the manuscript. L.H., L.Z., and L.W. contributed to the study design and performed flow experiments. Y.Z., C.W.L., A.X., and A.F.C. provided crucial reagents and made critical comments on the manuscript. Y.H. supervised the study, provided materials, and contributed to the experimental design and the manuscript preparation and revision. Y.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.