The role of endothelial nitric oxide in the anti-restenotic effects of liraglutide in a mouse model of restenosis

The chemical agents were purchased as follows: liraglutide from Novo Nordisk Japan (Chiyoda, Tokyo, Japan); exendin-(9–39) (Ex-9) from AnaSpec (Fremont, CA, USA); l-NAME, SQ22536 (SQ), cAMP-dependent protein kinase (PKA) inhibitor fragment 14–22 myristoylated trifluoroacetate salt (PKI[14–22]), ESI-09, STO-609, and N-omega-nitro-l-arginine methyl ester (l-NAME) from Sigma-Aldrich Japan (Shinagawa, Tokyo, Japan); 10-(4′-[[N-diethylamino]]butyl)-2-chlorophenoxazine (Akt inhibitor X) was purchased from Santa Cruz (Dallas, Texas, USA) and dorsomorphin dihydrochloride from Wako (Osaka, Osaka, Japan).

The study design was approved by the animal care committee of Showa University School of Medicine, and animal experiments were conducted in strict adherence to the Guide for the Care and Use of Laboratory Animals (8th edition, 2011; Office of Laboratory Animal Welfare, National Institutes of Health, MD, USA). All invasive procedures were performed under general anesthesia using isoflurane. Seven-week-old male C57BL6J (wild-type) and db/db mice were purchased from Sankyo Labo Service (Edogawa, Tokyo, Japan) and housed in clean cages (3 mice per cage, maximum) in the Division of Animal Experimentation of Showa University School of Medicine. After a 2-week acclimatization period, mice were randomly divided into vehicle or liraglutide treatment groups. We conducted five animal experiments. The experimental design and treatment groups are shown in Fig. 1. The dose of liraglutide was 17 nmol/kg/day, unless otherwise specified [3, 7, 8]. Agents were delivered using osmotic pumps (Alzet minipump 1002; Cupertino, CA, USA) subcutaneously implanted on Day 1, and the pumps were replaced on Day 15 to avoid liraglutide degradation. l-NAME (20 mg/kg/day), a NOS inhibitor, was administered via drinking water [26]. On Day 3, mice underwent left femoral artery wire injury, as previously reported by Sata and co-workers, with a few modifications [21]. Briefly, the left femoral artery was carefully isolated from the surrounding connective tissues through a small skin incision made in the left inguinal region. A straight spring wire (0.38-mm diameter, No. C-SF-15-15; Cook Japan, Nakano, Tokyo, Japan) was retrogradely inserted into the femoral artery via a small cut on the muscle branch. The wire was inserted to a depth of 7.5 mm and withdrawn after 1 min. The muscle branch was ligated proximal to the cut to prevent bleeding, and the skin incision was sutured. Xylocaine (1%) was added dropwise to the incision area to maintain tissue hydration keep tissues wet and relieve pain. Following the procedure, mice recovered in a separate cage for 6 h under careful observation, after which they were returned to their original cages. The mice did not show any signs of severe complications such as claudication or necrosis of the injured extremity, surgical site infection, or massive weight loss (20% of the baseline). Mice were killed by an overdose of inhaled isoflurane, and the injured arteries (5.5 mm long, from the site of the ligation) were collected on Day 29 for morphometric analyses (experiments 1–4) or on Day 11 for gene expression analyses (experiment 5).

Blood samples were collected after 6–8 h fasting. Plasma glucose levels were measured using a dextrometer (Stat strip XP2, Nipro, Osaka, Osaka, Japan). Hemoglobin A1c (HbA1c) and plasma lipid levels were determined using a latex-enhanced immunoassay and an enzymatic colorimetric assay, respectively (Cobas, Roche Diagnostics Japan, Minato, Tokyo, Japan). Plasma active GLP-1 levels were measured by ELISA (High Sensitivity GLP-1 Active ELISA Kit, EMD Millipore, Billerica, MA, USA).

Blood pressure and pulse rates were measured at the end of the experiments, using the tail-cuff method (Model MK-2000ST; Muromachi Kikai, Chuo, Tokyo, Japan). The average value obtained from 3 to 5 consecutive measurements was used as a single data point.

Injured arteries were collected for morphometric analysis following perfusion fixation with PBS and 4% paraformaldehyde. Three serial cross-sections taken from the proximal end of the paraffin-embedded femoral arteries at 0.5-mm intervals were stained with Elastica van Gieson (EVG), and the average of three serial cross-sections was used as a single data point. The sections were digitized using an inverted microscope (Model IX71, Olympus, Shinjyuku, Tokyo, Japan) and analyzed using Image J software (National Institutes of Health, Bethesda, MD, USA) by an investigator blinded to the treatment. The areas within the internal elastic lamina and the external laminal perimeter were defined as the neointima and the artery perimeter. Arteries completely occluded by the thrombus were excluded from analysis (Additional file 1: Figure S1a–c). Sections showing the following features were also excluded from analysis (Additional file 1: Figure S1d–f): broken wall structure, a branch of another artery, and missing elastic lamina. The percentage of excluded arteries and sections in each treatment group was approximately 10–20 and 15–25%, respectively. Fisher’s exact test revealed no difference between the groups.

Proliferating and endothelial cells were identified by immunostaining using an anti-Ki67 antibody (Thermo Fisher Scientific Japan, Yokohama, Kanagawa, Japan, RM-9106-S1, RRID: AB_149792, raised in rabbit, 1: 250) and an anti-CD31 antibody (Abcam Japan, Chuo, Tokyo, Japan, ab28364, RRID: AB_726362, raised in rabbit, 1: 200), respectively. Nuclei were counterstained with hematoxylin. The averages of three serial cross-sections were used as single data points.

In experiment 5, uninjured and injured femoral arteries were harvested 7 days after injury (Day 11). Total RNA extracted from the arteries was used for synthesizing complementary DNA as previously described [32]. Gene expression was assessed by real-time RT-PCR using the TaqMan gene expression assay and sequence detection system (ABI PRISM 7900; Life Technologies Japan, Minato, Tokyo, Japan). The following pre-designed TaqMan probe sets were used: Glp-1r, Mm00445292_m1; interleukin (Il)-1, Mm00434228_m1; Il-6, Mm00446190_m1; tumor necrosis factor (Tnf)-α, Mm00443258_m1; monocyte chemotactic protein (Mcp)-1, Mm00441242_m1; transforming growth factor (Tgf)-β, Mm01178820_m1. The 18S ribosomal RNA probe (18sRNA, Mm03928990_g1) was used as an internal control.

Human umbilical vein endothelial cells (HUVECs) were obtained from Lonza Japan (Chuo, Tokyo, Japan) and cultured in EBM-2 medium supplemented with FBS and growth factors (Lonza Japan). Collagen I-coated flasks or plates were used for all cell culture experiments. The cells were seeded onto 24-well plates at the following density per well: 2.5 × 10⁴ for NO measurements, 1.0 × 10⁴ for cell proliferation assays, and 4 × 10⁴ for protein extraction and siRNA transfection. Cells were starved in M199 containing 0.3% FBS for 1 h for NO measurements, and in 0.1% FBS containing EBM-2 (without growth factors) for 6 h for protein extraction. For cell culture experiments involving treatment with high levels of glucose, cells were exposed to 25 mmol/L glucose for 48 h before seeding. Cells between passages 4–8 were used for the experiments.

Plasma NO levels and NO production by HUVECs was determined by measuring the stable metabolites of NO (NO₂ and NO₃) in the culture medium (Nitrate/Nitrite Fluorometric Assay Kit; Dojin, Kamimashiki, Kumamoto, Japan) [33]. Protein in plasma samples was removed by ultrafiltration (Pierce concentrator, PES, 10 K MWCO; Thermo Fisher Scientific Japan) before measurement. HUVECs were stimulated with the indicated concentrations of agents (saline or liraglutide 0.1–100 nmol/L) for 2 h, following serum starvation in M199 with 0.3% FBS for 1 h. Inhibitors were added 30 min prior to stimulation at the following concentrations: GLP-1R antagonist Ex-9, 100 nmol/L; cAMP inhibitor SQ, 100 μmol/L; PKA inhibitor PKI (14–22), 1 μmol/L; exchange factor directly activated by cAMP (EPAC) inhibitor ESI-09, 6.4 μmol/L; calcium-calmodulin-dependent protein kinase kinase (CaMKK) inhibitor STO-609, 10 μmol/L; AMP-activated protein kinase (AMPK) inhibitor dorsomorphine, 10 μmol/L; Akt inhibitor X, 5 μmol/L; NOS inhibitor l-NAME, 1 mmol/L [3, 27, 34‐39]. The NO₂ in the plasma samples and medium was converted to NO₃ by NO₂ reductase after stimulation, and the medium was incubated with fluorescent 2,3-diaminonaphthalene for 15 min. The fluorescence intensities were measured using an Infinite M200 PRO instrument (TECAN Japan, Kawasaki, Kanagawa, Japan).

Proteins were extracted from HUVECs or mouse aortas using CelLytic M, containing cocktails of protease inhibitors and phosphatase inhibitors (all from Sigma-Aldrich Japan). Western blot was performed as previously described [40]. Briefly, 7 μg of protein per lane was electrophoresed in polyacrylamide gels and transferred to polyvinylidene fluoride membranes. The membranes were blocked with PVDF Blocking Reagent (TOYOBO, Osaka, Osaka, Japan) for 1 h, and then incubated overnight with antibodies at the following dilutions: phosphorylated LKB1 at Ser 428 (p-LKB1; Cell Signaling Technology Japan, Chiyoda, Tokyo, Japan, #3482, RRID: AB_2198321, raised in rabbit, 1: 2500), LKB1 (Cell Signaling Technology Japan, #3050, RRID: AB_823559, raised in rabbit, 1: 5000), p-AMPKα at Thr 178 (p-AMPK; Cell Signaling Technology Japan, #2535, RRID: AB_331250, raised in rabbit, 1: 2500), AMPKα (Cell Signaling Technology Japan, #2532, RRID: AB_330331, raised in rabbit, 1: 5000), p-eNOS at Ser 1177 (p-eNOS; Cell Signaling Technology Japan, #9570, RRID: AB_2298588, raised in rabbit, 1: 2500), eNOS (Cell Signaling Technology Japan, #32,027, RRID: not available, raised in rabbit, 1: 5000), and β-actin (Sigma-Aldrich Japan, A2066, RRID: AB_476693, raised in rabbit, 1: 8000). Goat anti-rabbit IgG conjugated with horseradish peroxidase (GE Healthcare Japan, Minato, Tokyo, Japan, NA934, RRID: AB_772206, raised in donkey) was used at a 1: 40,000 dilution. The Can Get Signal® Immunoreaction Enhancer Solution (TOYOBO) was used for antibody dilutions. The bands on the immunoblot were detected using the Amersham ECL Prime kit (GE Healthcare Japan) and quantified using Image J software.

Predesigned siRNAs for the negative control and LKB1 were purchased from Applied Biosystems (Silencer® Select Negative Control No. 2 siRNA; LKB1 siRNA, S74498; Foster City, CA, USA). The siRNAs were incubated for 30 min with the TransIT-TKO transfection reagent (Mirus Bio, Madison, WI, USA) in serum-free EBM-2, and the mixture was added to the medium at a final siRNA concentration of 25 nmol/L. HUVECs were used for experiments 72 h post-transfection.

First, we investigated the dose–effect relationship of liraglutide against restenosis after arterial injury (animal experiment 1). Wild-type C57BL6 mice were treated with vehicle or increasing doses of liraglutide (5.7, 17, or 107 nmol/kg/day). The physiological and biochemical parameters measured are shown in Table 1. No differences were detected between the groups, except for elevated levels of plasma active GLP-1 in groups treated with liraglutide. When evaluating morphometric changes, liraglutide treatment at 17 and 107 nmol/kg/day significantly suppressed neointimal hyperplasia without inducing medial thinning or arterial dilation. These changes resulted in reductions in the intima to media (I/M) ratio. In contrast, treatment with a 5.7 nmol/kg/day dose of liraglutide did not suppress neointimal hyperplasia (Fig. 2a–e). The Jonckheere-Terpstra trend test revealed a significant trend between the decreases in neointimal area and the increases in liraglutide doses (p < 0.001).

Next, we focused on endothelial NO as a potential mediator of the anti-restenotic effects of liraglutide (animal experiment 2). Vehicle or liraglutide (17 nmol/kg/day) were administered to mice in the presence or absence of the l-NAME NOS inhibitor. In a subset of animals, we observed NOS inactivation by l-NAME treatment in vivo. Plasma NO levels were significantly lower in mice treated with l-NAME than in those treated with vehicle (Additional file 1: Figure S2a). Consistently, l-NAME treatment significantly suppressed phosphorylation of eNOS in the aorta compared to vehicle treatment (Additional file 1: Figure S2b, c). Table 2 shows the physiological and biological parameters of each treatment group. Mice treated with l-NAME exhibited higher systolic blood pressure levels than those not administered the inhibitor, as previously reported [41]. Co-treatment with l-NAME completely abolished the suppression of neointimal hyperplasia by liraglutide, while the medial area and the arterial perimeter were not affected (Fig. 3a–e). Furthermore, liraglutide treatment decreased the percentages of intimal and medial proliferating cells, as assessed by cells that stained positive for the Ki-67 marker; however, these effects were not observed in mice co-treated with l-NAME (Fig. 3f–h). The number of proliferating cells in the neointima and media was correlated with neointimal hyperplasia and medial thinning, respectively (Table 3). In contrast, the density of neointimal or medial cells, calculated as the number of total cells divided by the area, was not affected by treatment with liraglutide or l-NAME (Fig. 3i, j).

To determine the optimal treatment regimen, we compared the anti-restenotic effects of early and delayed treatment (Days 1–14 and Days 15–29 days, respectively) with those achieved by full treatment (Day 1–29) (animal experiment 3). Figure 4a shows the experimental design. The formation of the neointima and endothelial regeneration, at different time points after arterial injury, are presented in Additional file 1: Figure S3. On Day 14, at the beginning of the delayed treatment, the neointima was not clearly noticeable; however, endothelial regeneration, as assessed by the CD31-positive area, was almost complete. Physiological and biochemical parameters were comparable between the groups (Table 4). Although treatment was terminated at Day 14, the early treatment group exhibited suppression of neointimal hyperplasia on Day 29, which was similar to that observed in the full treatment group (Fig. 4b–f). In contrast, this suppression was reduced in the delayed treatment group. Furthermore, early treatment was effective in reducing the proliferation of vascular cells, as assessed on Day 29 (Fig. 4g–i). Similar to neointimal hyperplasia, delayed treatment failed to suppress the proliferation of vascular cells.

To investigate the mechanism underlying the observed effects, we examined the molecules involved in liraglutide-stimulated NO production in vitro. Liraglutide treatment dose-dependently stimulated NO production in HUVECs, which was blocked by the Ex-9 GLP-1R antagonist (Fig. 5a). This effect was also completely abolished by the cAMP, PKA, AMPK, or NOS inhibitors, but not by the EPAC, CaMKK, or Akt inhibitors (Fig. 5b). Taken together, our data suggest that endothelial NO production induced by liraglutide is dependent on the GLP-1R/cAMP/PKA/AMPK/eNOS pathway in vitro.

Next, we investigated the role of LKB-1 as a molecule that potentially links PKA to AMPK. Liraglutide treatment significantly increased the phosphorylation of LKB1 in HUVECs, consistent with the phosphorylation of AMPK and eNOS (Fig. 6a–d). Cells transfected with siRNA against LKB1 exhibited a ~ 95% decrease in LKB1 protein levels compared with cells transfected with scrambled siRNA (Fig. 6e, f). LKB1 knockdown abrogated the effects of liraglutide on NO production (Fig. 6g). Moreover, LKB1 knockdown inhibited liraglutide-induced phosphorylation of AMPK and eNOS (Fig. 6h–j).

We examined whether the anti-restenotic effects of liraglutide were preserved under severe hyperglycemia in db/db mice, a model of obesity-induced diabetes (animal experiment 4). At the time of treatment initiation (at the age of 9 weeks), all db/db mice exhibited severe hyperglycemia (random plasma glucose levels over 300 mg/dL), and a significantly decreased Glp-1r expression in the aorta, compared to that in nondiabetic wild-type mice (Fig. 7a). First, we determined the dose of liraglutide to be administered. The body weights and the fasting plasma glucose levels of db/db mice were significantly reduced following liraglutide treatment with 107 nmol/kg/day compared with those of mice administered vehicle treatment, while treatment with 17 nmol/kg/day liraglutide did not affect body weight, and caused a slight decrease in fasting plasma glucose levels (Additional file 1: Figure S4a, b). To avoid the potential influence of systemic effects, we chose a 17 nmol/kg/day dose of liraglutide for this experiment. The physiological and biochemical parameters are presented in Table 5. Fasting plasma glucose and HbA1c levels tended to be lower in liraglutide-treated mice (p = 0.20 and 0.66, respectively). Other parameters were similar between the groups, except for the pulse rate and plasma active GLP-1 levels. Treatment with liraglutide significantly suppressed neointimal hyperplasia compared with that achieved by treatment with the vehicle, without inducing medial thinning or arterial dilation (Fig. 7b–f). In addition, liraglutide treatment reduced the percentage of neointimal and medial proliferating cells (Fig. 7g–i). Further, the reduction in the number of neointimal proliferating cells was correlated with the neointimal area (Table 6). Liraglutide treatment also decreased intimal cell density, which was not changed in normoglycemic mice (Fig. 7j, k).

We examined the effects of hyperglycemia on liraglutide-stimulated endothelial NO production in vitro. HUVECs were exposed to a high glucose concentration (25 mmol/L) 96 h before the experiments. Liraglutide treatment significantly increased NO production in HUVECs cultured in high concentrations of glucose, as well as in those cultured under normal glucose conditions (Fig. 8a). Similarly, liraglutide treatment significantly increased the phosphorylation of LKB1, AMPK, and eNOS in HUVECs cultured in high-glucose media (Fig. 8b–e).

Finally, we investigated the anti-inflammatory effects of liraglutide in vivo (animal experiment 5). Femoral arteries were collected from a different set of animals 7 days after injury (Day 11). In normoglycemic wild-type mice, liraglutide did not affect the injury-induced vascular expression of pro-inflammatory cytokines, chemokines, or growth factors (Fig. 9a–e). In hyperglycemic db/db mice, inflammatory responses to the injury were substantially enhanced compared with those in normoglycemic mice. Further, liraglutide treatment significantly suppressed the hyperglycemia-enhanced vascular expression of Il-1, Tnf-α, and Tgf-β, and tended to reduce Il-6 and Mcp-1 levels (Fig. 9a–e).