Introduction

Skeletal muscle plays a pivotal role in insulin-induced glucose uptake, which is impaired in insulin-resistant states such as obesity and type 2 diabetes [1, 2]. We previously demonstrated, using endothelium-specific Irs2-knockout (ETIrs2KO) mice, that the hyperinsulinaemia in high-fat (HF) diet-fed mice decreased the expression levels of Irs2 in the endothelial cells, and this insulin receptor substrate 2 (IRS2) downregulation caused impaired insulin-induced capillary recruitment and insulin delivery, resulting in a reduction of skeletal muscle glucose uptake [3]. Moreover, beraprost sodium, a stable prostaglandin analogue, activated endothelial nitric oxide synthase (eNOS) independently of IRS2, and reversed the reduction in capillary blood volume/insulin-induced skeletal muscle glucose uptake in both ETIrs2KO and HF diet-fed mice [3]. Thus, agents inducing activation of eNOS in the endothelial cells may be promising candidates for amelioration of skeletal muscle insulin resistance.

Glucagon-like peptide 1 (GLP-1) receptor (GLP-1R) agonists improve fasting and postprandial glucose levels and pancreatic beta cell function, and are widely used in the treatment of type 2 diabetes mellitus [4–8]. In addition to these actions, GLP-1R agonists have been reported to attenuate insulin resistance in humans [9–11]. Recently, Chai et al demonstrated that a bolus injection of a GLP-1R agonist improved both the muscle microvascular blood volume and skeletal muscle glucose uptake in HF diet-fed rats [12]. Moreover, the GLP-1R agonist exendin-4 has been shown to enhance the phosphorylation of eNOS via the protein kinase A (PKA) pathway in endothelial cell lines [13]. These data suggest that GLP-1 may improve skeletal muscle glucose uptake through activating eNOS in the endothelial cells. As physiological GLP-1 is easily degraded by dipeptidyl peptidase-4 (DPP-4) [14, 15], DPP-4 inhibitors that induce GLP-1 activation have been used in the treatment of type 2 diabetes [16–21]. In addition to exerting actions mediated by GLP-1, DPP-4 inhibitors have been shown to inhibit cleavage of several non-incretin substrates, such as cytokines, chemokines and neuropeptides [22–24]; thus, DPP-4 inhibitors may have additional actions independent of those mediated by GLP-1. We therefore investigated whether the DPP-4 inhibitor anagliptin improved insulin-induced skeletal muscle glucose uptake through eNOS activation using ETIrs2KO mice.

Methods

Animals

ETIrs2KO mice were generated as described previously [3], and housed under a 12 h light/dark cycle. The mice were fed regular chow, CE-2 (CLEA Japan, Tokyo, Japan), or CE-2 containing 0.3% (wt/wt) anagliptin, for 8 weeks. Anagliptin provided by Sanwa Kagaku Kenkyusho (Nagoya, Japan) was used as the DPP-4 inhibitor [25]. Male ETIrs2KO mice, 8 weeks old, were assigned to receive anagliptin or not, and Irs2 flox/flox mice were used as the control. The animal care and experimental procedures used in this study were accepted by the Animal Care Committee of the University of Tokyo.

Measurement of the plasma variables

Plasma anagliptin levels were measured with a validated liquid chromatography/tandem MS (LC-MS/MS) assay, as described previously [26]. Plasma DPP-4 activity was determined by a fluorometric assay using Gly-Pro-MCA (Peptide Institute, Osaka, Japan), slightly modified from a previously described protocol [27]. Further information is provided in the electronic supplementary material (ESM) Methods.

Insulin tolerance test

Mice were intraperitoneally administered with human insulin (Humulin R; Eli Lilly Japan, Kobe, Japan) at 0.75 U/kg body weight after 1 h of fasting. Blood glucose levels were measured at 0, 20, 40, 60, 80, 100 and 120 min after insulin administration using an automatic glucometer (Sanwa Kagaku Kenkyusho).

Hyperinsulinaemic–euglycaemic clamp

Clamp studies were performed as described previously [28], with some modifications. Briefly, 2 days before the study, an infusion catheter was inserted into the right jugular vein under general anaesthesia. Studies were conducted under conscious and unstressed conditions after 3 h of denying the animals access to food. Insulin was continuously administered (5.0 mU kg−1 min−1), and the blood glucose concentration, monitored every 5 min, was maintained at approximately 6.67 mmol/l by the infusion of glucose (5 g of glucose/10 ml enriched to about 20% [wt/vol.] with [6,6-2H2]glucose [Sigma-Aldrich, Tokyo, Japan]) for 60 min. Blood samples were obtained via tail-tip bleeds at 30, 45 and 60 min to determine the rate of glucose disappearance (Rd). Rd was calculated according to non-steady-state equations, and endogenous glucose production (EGP) was calculated as the difference between the Rd and the exogenous glucose infusion rate (GIR).

Capillary blood volume

The capillary blood volume was determined using contrast-enhanced ultrasound, by slight modification of a protocol described previously [29]. In brief, the hindlimb muscles were imaged in the short axis using an ultrasound system equipped with a 40 MHz transducer (RMV 704) (Vevo 770; VisualSonics, Toronto, ON, Canada). Infusion of Sonazoid (Daiichi Sankyo, Tokyo, Japan) was administered to the animals, which were divided into two groups for the measurements at 0 and 60 min after the hyperinsulinaemic–euglycaemic clamp. A high-power ultrasound with a frequency of 1 MHz was applied to the lower leg muscles, and images were obtained for 30 s to assess the enhancement. The ultrasound intensity in decibels within the region of interest was converted to acoustic intensity after background subtraction of the 0.5 s ultrasound images, using the formula of microvascular volume, fill rate constant and capillary blood volume was y = A(1 − e-βt), where y is capillary blood volume, A is plateau video intensity, and t is the pulsing interval.

Interstitial concentrations of insulin in the skeletal muscle

Muscle microdialysis was conducted in the hindlimb muscles using a 4 mm microdialysis tube (CMA-20; CMA Microdialysis, Stockholm, Sweden), at a rate 0.3 μl/min, as described previously [30]. We carried out calibration using the no-net-flux technique. In brief, four known concentrations of insulin (0 pmol/l, 8.75 pmol/l, 17.5 pmol/l and 26.25 pmol/l) above and below the expected concentration in the skeletal muscle were selected. The insulin solutions were added to the perfusate and the net changes in the concentrations of the analytes in the dialysate were archived (insulinout − insulinin = net change). According to regression analysis, a linear relationship was obtained between the concentrations in the perfusates and the dialysates. The intercept with the x-axis shows the insulin concentrations in the perfusate at equilibrium with the surrounding medium, and the slope of the line exhibits the dialysis recovery by the no-net-flux technique. The insulin concentrations in the interstitial fluid were determined from the dialysis recovery by the no-net-flux technique and in vivo dialysate insulin concentration, as described previously [31].

Insulin signalling in the skeletal muscle in vivo and ex vivo

To investigate insulin signalling in the skeletal muscle, insulin (105 pmol/l) was injected via the inferior vena cava. The soleus and gastrocnemius muscles were dissected 60 min after the insulin infusion and immediately frozen in liquid nitrogen. Insulin signalling in the isolated skeletal muscle was carried out as described previously [32], with slight modifications. In brief, the soleus and gastrocnemius muscles were dissected out, cut into strips of 10–15 mg, and incubated for 10 min at 37°C in 1% BSA/DMEM with or without 100 nmol/l insulin. The muscles were rapidly washed twice with PBS buffer to stop the reaction. For the western blot analysis, the muscle pieces were immediately frozen in liquid nitrogen. Information about western blot analysis and antibodies is provided in ESM Methods.

Administration of exendin-(9-39)

Exendin-(9-39) (Ex-9-39) (Sigma-Aldrich) was administered as described previously [33], with a slight modification of the method. In brief, Ex-9-39 or vehicle (0.9% [wt/vol.] NaCl, 1% [wt/vol.] BSA) was administered subcutaneously at a rate of 50 pmol kg−1 min−1 using mini-osmotic pumps (model 1002; DURECT Corporation, Cupertino, CA, USA), which were replaced every 2 weeks. Ex-9-39 was administered for 8 weeks.

Cell culture

Human coronary arterial endothelial cells (HCAECs) and human cardiac microvascular endothelial cells (HCMVECs) were purchased from Lonza Japan (Tokyo, Japan) and maintained at 37°C under 5% CO2 in endothelial growth medium-2 (Lonza Japan) supplemented with hydrocortisone, human epidermal growth factor (hEGF), FBS, vascular endothelial growth factor (VEGF), human fibroblast growth factor (hFGF)-B, R3 insulin-like growth factor (IGF)-1, ascorbic acid and gentamicin/amphotericin B. The HCAECs and HCMVECs were grown to confluence, then incubated overnight under serum starvation. H89 or vehicle was added 30 min prior to the GLP-1 stimulation. Insulin and 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) were incubated for 30 min. All experiments were performed with cells at passage 5–10.

NOS inhibition

Mice were infused with either insulin or insulin with a superimposed infusion of N ω-nitro-l-arginine methyl ester (l-NAME) (50 mg/kg bolus followed by continuous infusion at 10 mg kg−1 h−1) during the hyperinsulinaemic–euglycaemic clamp study. We measured the capillary blood volume at 60 min after the infusion.

RNA preparation and RT-PCR

Total RNA was prepared from the tissues and cell lines using the Trizol reagent (Thermo Fisher Scientific, Yokohama, Japan) and RNeasy Mini Kit (Qiagen Sciences, Germantown, MD, USA), respectively, according to the manufacturer’s instructions. RNA, 1 μg, was used for generating cDNA using random hexamers with MultiScribe reverse transcription reagents (Thermo Fisher Scientific). In the tissue samples, TaqMan quantitative PCR (50°C for 2 min, 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 60°C for 1 min) was then performed with ABI Prism 7900 PCR (Thermo Fisher Scientific) to amplify Tnfα (also known as Tnf), Mcp1 (also known as Ccl2), resistin, Rbp4, Vegf (also known as Vegfa), Gapdh and β-actin cDNA. The expression levels of each of the transcripts were normalised to the constitutive expression levels of Gapdh and β-actin mRNA. In the cell lines, RT-PCR was performed in a DNA thermal cycler for 40 cycles (1 min at 94°C, 30 s at 56°C, and 1 min at 72°C). Information about TaqMan probes and Glp1r specific primers is provided in ESM Methods.

Statistical analysis

Results are expressed as means ± SEM. The statistical significance of all possible pairwise differences in the means were determined by one-way ANOVA followed by post hoc Tukey–Kramer test. The statistical significance level was set at p < 0.05.

Results

Treatment with anagliptin ameliorated impaired insulin-induced skeletal muscle glucose uptake in ETIrs2KO mice

When anagliptin was administrated to ETIrs2KO mice, plasma anagliptin levels were increased by about 1200 ng/ml (ESM Fig. 1), similar to those reported for healthy volunteers receiving anagliptin [34]. Plasma DPP-4 activity was inhibited by approximately 90% in the anagliptin-treated ETIrs2KO mice (Fig. 1a). However, plasma levels of active GLP-1 were markedly increased in the anagliptin-treated ETIrs2KO mice (Fig. 1b). There were no significant differences in the body weight gain or food intake among the three groups (Fig. 1c, d). These data suggest that anagliptin treatment increased the active GLP-1 levels in ETIrs2KO mice without inducing any changes in body weight or food intake.

Fig. 1
figure 1

Treatment with anagliptin restored the impaired insulin-induced skeletal muscle glucose uptake in the ETIrs2KO mice. Plasma DPP-4 activity (a) and active GLP-1 levels (b) in the control, ETIrs2KO and anagliptin-treated ETIrs2KO mice (n = 21–24). Values are the means ± SEM. ***p < 0.001, ETIrs2KO mice vs anagliptin-treated ETIrs2KO mice. (c) Body weight in the control (white circles), ETIrs2KO (white triangles) and anagliptin-treated ETIrs2KO (white squares) mice (n = 15). (d) Food intake in the control, ETIrs2KO and anagliptin-treated ETIrs2KO mice (n = 7). (e) Blood glucose levels during the insulin tolerance test in the control (white circles), ETIrs2KO (white triangles) and anagliptin-treated ETIrs2KO (white squares) mice (n = 13–15). Values are the means ± SEM. *p < 0.05, control mice vs ETIrs2KO mice; † p < 0.05, ETIrs2KO mice vs anagliptin-treated ETIrs2KO mice. GIR (f), Rd (g) and EGP (h) during the hyperinsulinaemic–euglycaemic clamp study in the control, ETIrs2KO and anagliptin-treated ETIrs2KO mice (n = 7–9). Values are the means ± SEM. *p < 0.05 vs ETIrs2KO mice

Full size image

To investigate the effect of anagliptin on the insulin resistance observed in the ETIrs2KO mice, an insulin tolerance test was performed. Although the hypoglycaemic response to insulin was lower in the ETIrs2KO mice than in the control mice, as reported previously by us [3], it did not differ between the control and anagliptin-treated ETIrs2KO mice (Fig. 1e). These results suggest that anagliptin treatment ameliorated the insulin resistance in the ETIrs2KO mice. Consistent with these data, the decrease in GIR in the ETIrs2KO mice was completely reversed by anagliptin treatment during the hyperinsulinaemic–euglycaemic clamp study (Fig. 1f). Moreover, anagliptin treatment also increased Rd in the ETIrs2KO mice to a level similar to that observed in the control mice (Fig. 1g). However, there were no significant differences in EGP among the three groups (Fig. 1h). These data suggest that anagliptin treatment ameliorated the impaired insulin-induced skeletal muscle glucose uptake in the ETIrs2KO mice. No significant differences were observed in the rectal temperature, plasma lipid or adipocytokine profiles, or mRNA expression levels of inflammatory cytokines and adipocytokines among the three groups (ESM Fig. 2a–m).

Anagliptin ameliorated the decreased insulin-induced capillary blood volume and interstitial insulin concentrations in ETIrs2KO mice

To delineate the mechanism underlying the improvement in insulin-induced skeletal muscle glucose uptake in the anagliptin-treated ETIrs2KO mice, the capillary blood volume was measured in the three groups. Treatment with anagliptin restored the decreased insulin-induced capillary blood volume in the ETIrs2KO mice to a level similar to that observed in the control mice (Fig. 2a, ESM Video 1a–c). Moreover, the reduced insulin-induced interstitial concentration of insulin observed in the ETIrs2KO mice was completely restored in the anagliptin-treated ETIrs2KO mice, even though no significant differences were observed in the plasma insulin levels among the three groups during the hyperinsulinaemic–euglycaemic clamp study (Fig. 2b, c). These data suggest that anagliptin treatment ameliorated the decreased insulin-induced capillary blood volume and interstitial insulin concentrations in the ETIrs2KO mice. Consistent with the results for the interstitial insulin concentrations, the insulin-induced phosphorylation levels of insulin receptor β (IRβ) and Akt in the skeletal muscle at 60 min after insulin infusion into the inferior vena cava were also significantly restored in the anagliptin-treated ETIrs2KO mice (Fig. 2d, e). However, in the isolated skeletal muscle, the phosphorylation levels of IRβ and Akt did not differ between the control and ETIrs2KO mice, and no significant differences in the phosphorylation levels between the anagliptin-treated and non-anagliptin-treated ETIrs2KO mice were observed (Fig. 2f, g). These data suggest that the improved insulin delivery induced by anagliptin treatment ameliorated impaired insulin signalling in the skeletal muscle. Taken together, anagliptin treatment restored the decreased insulin-induced capillary blood volume and interstitial insulin concentrations, resulting in amelioration of the impaired insulin signalling in the skeletal muscle of the ETIrs2KO mice.

Fig. 2
figure 2

Treatment with anagliptin restored the insulin-induced increase in capillary blood volume and interstitial concentration of insulin in the ETIrs2KO mice. Capillary blood volume (a), and plasma (b) and interstitial (c) insulin levels during the hyperinsulinaemic–euglycaemic clamp study in the control (white bars), ETIrs2KO (black bars) and anagliptin-treated ETIrs2KO (grey bars) mice (n = 9–12). Values are the means ± SEM. *p < 0.05 vs ETIrs2KO mice; **p < 0.01 and ***p < 0.001, 0 min vs 60 min. Phosphorylation levels of IRβ (d) and Akt (e) in the skeletal muscle of control (white bars), ETIrs2KO (black bars) and anagliptin-treated ETIrs2KO (grey bars) mice at 60 min after insulin infusion (n = 4). *p < 0.05 vs ETIrs2KO mice. Insulin-stimulated phosphorylation levels of IRβ (f) and Akt (g) in the isolated skeletal muscle of control (white bars), ETIrs2KO (black bars) and anagliptin-treated ETIrs2KO (grey bars) mice (n = 4). Values are the means ± SEM. IB, immunoblot; pY, phosphorylated tyrosine

Full size image

Capillary blood volume in the control, ETIrs2KO and anagliptin treated ETIrs2KO mice at 60 min after insulin infusion. Representative video clips of capillary recruitment in the control (a), ETIrs2KO (b) and anagliptin-treated ETIrs2KO mice (c) at 60 min after insulin infusion. Injected Sonazoid (microbubbles) was destroyed by a high-power ultrasound with a frequency of 1 MHz in the lower leg muscles, and images were collected for 30 s thereafter to assess the bubble replenishment. Green labels indicate Sonazoid. (MP4 4.54 mb)

(MP4 5.80 mb)

(MP4 4.54 mb)

Treatment with anagliptin increased insulin-induced skeletal muscle glucose uptake in the ETIrs2KO mice via a GLP-1-dependent pathway

To address the extent to which GLP-1 activation by anagliptin is involved in the improvement of the insulin-induced skeletal muscle glucose uptake in ETIrs2KO mice, Ex-9-39, an antagonist of GLP-1R, was administered to the anagliptin-treated ETIrs2KO mice. The improvement in GIR and Rd in the anagliptin-treated ETIrs2KO mice was significantly abrogated by treatment with Ex-9-39 in the hyperinsulinaemic–euglycaemic clamp study (Fig. 3a, b). However, no significant difference in EGP was found among the four groups (Fig. 3c). These data suggest that anagliptin treatment improved insulin-induced skeletal muscle glucose uptake in the ETIrs2KO mice via a GLP-1-dependent pathway.

Fig. 3
figure 3

GLP-1 activation by anagliptin treatment mainly restored the insulin-induced skeletal muscle glucose uptake in ETIrs2KO mice. GIR (a), Rd (b) and EGP (c) in the anagliptin-treated ETIrs2KO mice following Ex-9-39 treatment in the hyperinsulinaemic–euglycaemic clamp study (n = 8–11). Values are the means ± SEM. *p < 0.05

Full size image

Treatment with l-NAME completely abrogated the insulin-induced increase in capillary blood volume in anagliptin-treated ETIrs2KO mice

To determine whether the improvement in capillary blood volume in the anagliptin-treated ETIrs2KO mice is mediated by an NO-dependent mechanism, we administered l-NAME to the anagliptin-treated ETIrs2KO mice. Although there were no significant differences among the six groups in the capillary blood volume before insulin infusion, treatment with l-NAME completely reversed the insulin-induced increase in capillary blood volume in the control mice. This result suggests that the insulin-induced increase in capillary blood volume in the control mice was mediated by an NO-dependent mechanism (Fig. 4). Similarly, the insulin-induced increase in capillary blood volume in the anagliptin-treated ETIrs2KO mice was completely abrogated by treatment with l-NAME. These data suggest that anagliptin treatment increased the insulin-induced capillary blood volume via an NO-dependent pathway in ETIrs2KO mice.

Fig. 4
figure 4

Treatment with l-NAME completely abrogated the insulin-induced increase in capillary blood volume in the anagliptin-treated ETIrs2KO mice. Capillary blood volume in the anagliptin-treated ETIrs2KO mice following l-NAME treatment (n = 8–12). Values are the means ± SEM. *p < 0.05, **p < 0.01 and ***p < 0.001

Full size image

GLP-1 elevated the phosphorylation levels of eNOS via the PKA pathway in endothelial cells

The expression levels of Glp1r mRNA were confirmed in HCAECs (Fig. 5a). To investigate whether GLP-1 activated eNOS in the HCAECs, the protein and phosphorylation levels of eNOS were measured after GLP-1 stimulation in these cells. Although the levels of eNOS protein did not differ significantly before and after GLP-1 stimulation, the eNOS phosphorylation levels were significantly increased 30 and 60 min after GLP-1 stimulation (Fig. 5b). However, the protein and phosphorylation levels of eNOS did not differ significantly before or after anagliptin treatment (Fig. 5c). These data suggest that GLP-1, and not directly anagliptin, activates eNOS in the endothelial cells. Moreover, in the presence of the PKA inhibitor H89, the GLP-1-stimulated eNOS phosphorylation was completely abolished (Fig. 5d). However, GLP-1 stimulation had no significant effect on AMPK phosphorylation levels (Fig. 5e). Just as in the case of the HCAECs, levels of Glp1r were confirmed in the HCMVECs (ESM Fig. 3a). While there was no change in the phosphorylation level of eNOS associated with anagliptin administration, the phosphorylation level of eNOS was significantly increased after GLP-1 stimulation (ESM Fig. 3b). Moreover, in the presence of the PKA inhibitor H89, the GLP-1-stimulated eNOS phosphorylation was almost completely abolished (ESM Fig. 3b). These results suggest that GLP-1 activates eNOS via the PKA pathway in endothelial cells.

Fig. 5
figure 5

GLP-1 stimulated phosphorylation of eNOS via the PKA pathway in HCAECs. (a) Glp1r expression in HCAECs using RT-PCR (n = 3). (b, c) GLP-1- and anagliptin-stimulated phosphorylation levels of eNOS in HCAECs (n = 4–6). Values are the means ± SEM. **p < 0.01 and ***p < 0.001 vs untreated condition. (d) GLP-1-stimulated phosphorylation levels of eNOS following treatment with H89, an inhibitor of PKA, in HCAECs (n = 4–6). (e) GLP-1-stimulated phosphorylation levels of AMPK in HCAECs (n = 4–6). Values are the means ± SEM. **p < 0.01 and ***p < 0.001

Full size image

Discussion

In this study, we demonstrated that anagliptin treatment enhanced insulin-induced capillary recruitment and interstitial insulin concentrations, resulting in improvement of skeletal muscle glucose uptake in ETIrs2KO mice. These effects were abrogated by treatment with a GLP-1 antagonist. Moreover, pre-treatment with l-NAME also significantly attenuated the insulin-induced capillary recruitment in anagliptin-treated ETIrs2KO mice. These data suggest that anagliptin treatment enhances the insulin-induced capillary recruitment and interstitial insulin concentrations, resulting in improved skeletal muscle glucose uptake in vivo via direct actions on the endothelial cells through GLP-1- and NO-dependent pathways.

As no effective antibodies to GLP-1R are currently available [35], the protein levels of GLP-1R could not be assessed in endothelial cells. However, the expression of Glp1r mRNA in HCAECs has been confirmed in a previous study [36]. GLP-1 has been shown to increase the plasma NO levels and improve endothelial-dependent vascular relaxation [37]. These data suggest that functional GLP-1R is expressed in endothelial cells. In the current study, although anagliptin treatment did not change the phosphorylation level of eNOS, eNOS phosphorylation was increased by GLP-1 via the PKA pathway in HCAECs. Consistent with our results, Matsubara et al reported that while DPP-4 inhibitors did not change the phosphorylation levels of eNOS, treatment with GLP-1 or GLP-1 plus DPP-4 inhibitor significantly increased the eNOS phosphorylation level in HCAECs [36]. These data suggest that anagliptin stimulates eNOS activity by a GLP-1-dependent mechanism, potentially via PKA–eNOS phosphorylation.

Nakagami et al previously demonstrated using a spontaneously hypertensive corpulent congenic (SHR)/NDmcr-cp rat model that DPP-4 inhibitors improved endothelial function [38]. This rat model, a spontaneously hypertensive strain with a nonsense mutation in the leptin receptor, shows obesity, severe hypertension, insulin resistance, hyperglycaemia, increased levels of inflammatory cytokines and hyperlipidaemia. In the aforementioned study, DPP-4 inhibitors ameliorated metabolic abnormalities such as glucose intolerance and inflammatory cytokines, as well as endothelial function. Thus, it was difficult to determine whether the DPP-4 inhibitor improved endothelial function via a direct effect or via indirect effects such as improvement in the risk factor profile. In contrast, ETIrs2KO mice do not show any obesity or changes in risk factor profile, including levels of triacylglycerol, FFA, leptin, inflammatory cytokines and adipocytokines, on a normal diet, despite also exhibiting endothelial dysfunction (ESM Fig. 2a–m). Moreover, the body weight, risk factor profile and levels of inflammatory cytokines and adipocytokines remained unaltered by anagliptin treatment in the ETIrs2KO mice (ESM Fig. 2a–m). Thus, ETIrs2KO mice are thought to be appropriate as a model for investigating the effects of DPP-4 inhibitors.

We demonstrated that eNOS phosphorylation level was significantly increased from 30 min after GLP-1 stimulation in vitro (Fig. 5b). This activation of eNOS was completely abolished in the presence of the PKA inhibitor, H89. Dong et al reported that muscle microvascular recruitment induced by bolus infusion of GLP-1 was inhibited by the administration of H89 [39]. These findings suggest that anagliptin probably improves skeletal muscle capillary recruitment via the PKA pathway in mice. However, to investigate whether anagliptin indeed regulates skeletal muscle capillary recruitment via the PKA pathway in vivo, further study is needed using endothelial cell-specific Pka-knockout mice (Pka is also known as Prkaca).

Although it has been demonstrated that bolus injection of a GLP-1R agonist improved muscle microvascular blood volume and skeletal muscle glucose uptake in HF diet-fed rats [12], the causal relationship between increased muscle microvascular blood volume and amelioration of skeletal glucose uptake is yet to be clarified. This study is the first to demonstrate this causal relationship using ETIrs2KO mice; we showed that chronic treatment with anagliptin enhanced insulin-induced capillary recruitment and interstitial insulin concentrations, thereby improving the skeletal muscle glucose uptake in vivo, by directly acting on endothelial cells via NO- and GLP-1-dependent mechanisms. The findings of our study may be of clinical relevance, as GLP-1R agonists and DPP4 inhibitors have been reported to attenuate insulin resistance in humans [9–11, 40]. Taken together, these data suggest that anagliptin might be a promising agent to improve skeletal muscle insulin resistance in obese patients with type 2 diabetes.