β3 Adrenergic Stimulation Restores Nitric Oxide/Redox Balance and Enhances Endothelial Function in Hyperglycemia

S961 was expressed in Escherichia coli as a fusion protein with an affinity tag and had the peptide sequence GSLDESFYDWFERQLGGGSGGSSLEEEWAQIQCEVWGRGCPSY with a disulfide bridge connecting the 2 cysteine residues (a gift from Novo Nordisk, Denmark).¹⁸ We infused S961 dissolved in normal saline for 7 days at a rate of 12 μg/kg per hour via osmotic mini‐pumps (Alzet, Palo Alto, CA) implanted subcutaneously in male New Zealand White rabbits, weighing 2.2 to 2.6 kg while they were under general anesthesia. The selective β₃ AR agonist CL (Sigma‐Aldrich, St. Louis, MO) was dissolved in normal saline and infused via mini‐pumps at a rate of 40 μg/kg per hour during the last 3 days of hyperglycemia. Since infusion of the vehicle alone (normal saline) in 5 rabbits did not change oxidative modification of the target proteins or endothelium‐dependent vasorelaxation (either identical or almost‐identical means within nearly equal SEM), we did not use sham infusions in controls. Durations of infusions of S961 and CL were limited by the amount of S961 we had available and the cost of CL needed in a relatively large animal such as rabbit. Blood samples for biochemical and metabolic analysis were obtained from the central ear artery of the anesthetized rabbits before and after each treatment. Blood glucose was measured in a drop of blood from the marginal ear vein using a glucometer and strips (Optium Xceed, Abbott Diabetes Care Ltd, Australia). Heart rate and blood pressure were measured via a catheter in the ear artery after anesthesia by subcutaneous injection of ketamine hydrochloride (50 mg/kg) and xylazine hydrochloride (50 mg/kg) prior to euthanasia by intravenous bolus injection of ketamine. The rabbit thoracic aorta was harvested and cleared of adhering tissue in Krebs buffer. The total number of rabbits used for this study was 50 (diabetogenesis with alloxan in vivo, n=9, control rabbits for in vitro assessment of diabetogenic agents n=6, sham infusions with normal saline n=5, control rabbits for vasomotor studies and molecular experiments n=10, rabbits with hyperglycemia induced by S961 infusion including dose titration experiments n=20). Pooled, multiple‐donor human umbilical vein endothelial cells (Lonza, Basel, Switzerland) were used for in vitro experiments. Surplus segments of human internal mammary or radial artery were obtained from patients undergoing coronary artery bypass graft operation at our institution (n=4 samples from diabetic patients and n=4 from nondiabetic patients). Myocardial tissue was obtained from a patient undergoing cardiac biopsy at another institution. The study protocols were in accordance with institutional guidelines and were approved by the appropriate research ethics committees at our institution. Informed written consent was obtained from patients.

To detect glutathionylation of eNOS and β₁ Na⁺‐K⁺ pump subunit in co‐immunoprecipitation experiments, an antibody against glutathionylated protein (anti‐glutathione antibody) was used to detect glutathionylation.¹⁰ Aorta was homogenized in ice‐cold lysis buffer containing 150 mmol/L NaCl, 50 mmol/L Tris‐HCl (pH 8.0), 1% Triton X‐100, 2 mmol/L EDTA, and protease inhibitor (Complete EGTA‐free, Roche Diagnostics), followed by centrifugation at 16 000g for 20 minutes. The supernatant (0.5–1 mg protein) was incubated with the appropriate antibody at a ratio of 1 μg of β₁ Na⁺‐K⁺ subunit antibody:1 mg protein and 2.5 μg anti‐eNOS antibody:1 mg protein at 4°C for 1 hour and then with protein A/G‐Plus agarose beads. The proteins bound to the collected beads were eluted in Laemmli buffer, subjected to SDS‐PAGE, and probed with anti‐glutathione antibody. This protocol was also used to detect co‐immunoprecipitation of eNOS/ glutaredoxin‐1 (Grx1), β₁ subunit/Grx1, and p47^phox/eNOS. To detect glutathionylation of β₁ Na⁺‐K⁺ subunit in cardiac myocytes, freshly isolated myocytes were loaded with biotinylated glutathione ester (500 μmol/L) for 1 hour at room temperature. The biotin‐tag was used to precipitate glutathionylated proteins as previously described.¹⁰ Western blot chemiluminescence was read by a LAS‐4000 image reader and quantified by densitometry using Multi Gauge 3.1 software (Fujifilm Life Science, Tokyo, Japan). Exposure times were adjusted to ensure that the variation in signal intensity was in the linear dynamic range.

To synthesize the colloid spin trap, sodium diethyldithiocarbamate (sodium DETC, 4.5 mg) and FeSO₄ 7H₂O (2.8 mg) were dissolved separately in 10 mL of deoxygenated Krebs bubbled with a nitrogen purge and mixed rapidly, resulting in a final concentration of Fe(DETC)₂ of 0.5 mmol/L. The spin trap was used immediately. After harvesting the aorta from rabbits and separating the loose adhering tissue, the aorta was opened longitudinally and cut into 10×10 mm segments. The segments were placed in a 6‐well plate containing 500 μL Krebs, with 500 μL of Fe(DETC)₂ added (final concentration of 0.25 mmol/L), and incubated at 37°C for 30 minutes. The tissue samples were placed in electron paramagnetic resonance tubes, snap‐frozen, and stored at −80°C until analysis. Spectra were recorded at 77 K with a Bruker EMX X‐band spectrometer using the following parameters: microwave power 2.5 mW; microwave frequency 9.4 GHz; modulation amplitude 0.2 mT; modulation frequency 100 kHz; conversion time 81.9 ms, time constant 163.8 ms, scan time 83.9 s with 12 scans averaged. The amplitude of NO signal was determined as the perpendicular height between the top of the first low‐field signal and the valley of the third high‐field signal.²⁰

The stock solution of DHE was prepared by dissolving DHE in deoxygenated dimethyl sulfoxide under argon in the dark with the concentration checked by UV‐VIS. Immediately after harvesting from rabbits, five 2‐mm segments of aorta were incubated in Krebs buffer containing the metal chelator diethylenetriaminepentaacetic acid (DTPA, 100 μmol/L) to minimize artificial oxidation, and DHE (50 μmol/L) in 1.5 mL Eppendorf tubes (37°C, 30 minutes, in the dark). The tissue was washed of the DHE with Krebs/DTPA 3 times and homogenized in 300 μL of cold methanol by centrifugation at 12 000g for 10 minutes at 4°C in the dark. Fifty microliters of homogenate was taken to determine protein concentration. Equal amounts of the homogenate were then added to equal amount of 0.2 mol/L of HClO₄ in methanol, vortexed, and placed on ice for 1 hour in the dark to allow precipitation of protein. After centrifugation at 20 000g at 4°C for 30 minutes, the supernatant was stored at −80°C until analysis. High‐performance liquid chromatography equipped with a fluorescence and a CoulArray electrochemical detector was used to separate the O₂^•−‐dependent 2‐hydroxy‐ethidium (2‐OH‐E⁺) product from the nonspecific product ethidium (E⁺) following DHE oxidation in rabbit aorta.²¹ Samples (50 μL) were separated by high‐performance liquid chromatography (Shimadzu) after injection onto a Synergi Polar RP C18 column (250×4.6 mm, 4 μmol/L, 80 Å, Phenomenex) equilibrated at 30°C with 60% mobile phase A (10% CH₃CN in potassium phosphate buffer, 50 mmol/L, pH 2.6) and 40% mobile phase B (60% CH₃CN in potassium phosphate buffer, 50 mmol/L, pH 2.6). Separation of products was performed by a linear increase to 100% mobile phase B over 30 minutes at a flow rate of 0.5 mL min⁻¹. Products were quantified by fluorescence (DHE and 2‐OH‐E⁺: λ_EX 358 nm, λ_EM 440 nm; E⁺ λ_EX 490 nm, λ_EM 565 nm) and electrochemical oxidation using a CoulArray detector (Environmental Sciences Associates; electrodes at 0, 200, 280, 365, 400, 450, 500, and 600 mV). The quantified 2‐OH‐E⁺ levels were normalized to protein concentration.

Aortic rings measuring 5 mm in length were mounted on 2 stainless steel hooks immersed in 25 mL organ bath chambers filled with Krebs buffer containing (in mmol/L) NaCl 118, KCl 4.6, NaHCO₃ 27.2, MgSO₄ 1.2, CaCl₂ 2.5, KH₂PO₄ 1.2, and glucose 11.1 (pH 7.4). The isometric contractions were measured in organ chambers at 37°C by using a force transducer (Radnoti) and recording with Powerlab software (AD Instruments, United States). Vascular rings were allowed to equilibrate for 0.5 hours and tension was gradually increased to 2 g. Vessels were then contracted with the addition of 80 mmol/L KCl to determine integrity of the rings. After maximum contraction was achieved, the aortic rings were washed with Krebs buffer and re‐equilibrated. Rings were then precontracted with phenylephrine (100 nmol/L). Once a plateau was achieved, increasing concentrations of acetylcholine were added. The endothelium‐dependent relaxation was expressed as the percentage of relaxation to phenylephrine‐induced contraction. In separate experiments, Krebs buffer was replaced with Krebs buffer that was nominally K⁺‐free. After 20 minutes in K⁺‐free Krebs buffer, rings were precontracted with phenylephrine (100 nmol/L). Once a plateau was achieved, increasing concentrations of K⁺ were administered as previously described.^{22, 23} As the Na⁺‐K⁺ pump is inhibited in the absence of extracellular K⁺ and reactivated by its reintroduction at low concentrations, K⁺‐induced relaxation has been characterized as an index of Na⁺‐K⁺ pump activity in arteries, which is dependent on external K⁺ and intracellular Na⁺ and Mg²⁺ and inhibited by the pump inhibitor ouabain.²²

Sodium diethyldithiocarbamate, FeSO₄ 7H₂O, diethylenetriaminepentaacetic acid, and glutathione ethyl ester were purchased from Sigma‐Aldrich. N‐Hydroxysulfosuccinimidobiotin was obtained from Merck, dithiothreitol from Promega, streptavidin–Sepharose from GE Healthcare Bio‐Sciences, and Protein A/G‐Plus agarose from Santa Cruz Biotechnology. Osmotic mini‐pumps were purchased from Alzet and CL from Sigma‐Aldrich. Dihydroethidium and N‐(Biotinoyl)‐N′‐(iodoacetyl) ethylenediamine were obtained from Invitrogen. Monoclonal antibodies were purchased from the following vendors: β₁ subunit of Na⁺‐K⁺ ATPase from Upstate Biotechnology; anti‐glutathione from Virogen, eNOS antibody from Sigma‐Aldrich, and p47^phox, phosphorylated eNOS116, phosphorylated eNOS1177, neuronal NOS (nNOS), and α tubulin from Santa Cruz Biotechnology. All chemicals used in Krebs solutions were analytical grade and were obtained from BDH.

Results are presented as mean±SEM and median with 25% to 75% interquartile range where necessary. Kolgomorov–Smirnov test was performed to check for normality. For comparison between 2 groups either a non‐paired Student t test or a Mann–Whitney test were used where appropriate. Paired t test was used for analysis of changes in blood biochemistry and body weight before and after the induction of hyperglycemia or treatment with CL (Table). For multiple comparisons, 1‐way ANOVA was used with Tukey's post‐hoc analysis. For vasorelaxation data, concentrations of acetylcholine and vasorelaxation responses in rabbits from different experimental groups were included in 2‐way repeated‐measures ANOVA with Tukey or Sidak's post‐hoc tests used for multiple comparisons. P value <0.05 was considered statistically significant.

Since we aimed to examine the effects of hyperglycemia on oxidative modification of proteins, we first determined whether alloxan and STZ, agents commonly used to establish type 1 diabetes, would have direct redox effects independently of diabetes per se. We tested this by examining glutathionylation of the highly abundant, ubiquitous membrane protein Na⁺‐K⁺ pump, which mediates inhibition of the pump function in cardiac myocytes²⁴ and vascular smooth muscle cells.²³ Alloxan, at high concentrations in vitro (1 mmol/L, 37°C, 30 minutes), reflecting the pancreatotoxic diabetogenic doses used in vivo, increased glutathionylation of the β₁ Na⁺‐K⁺ pump subunit (Figure 1A). Moreover, glutathionylation of the β₁ pump subunit was also increased in rabbits that received a typically diabetogenic dose of alloxan (100 mg/kg) in vivo but did not develop diabetes due to insufficient necrosis of pancreatic β cells (Figure 1B). Higher glutathionylation levels were detected 7 days after alloxan injection and were similar to levels in rabbits that did develop diabetes (Figure 1B), thus showing that alloxan can alter glutathionylation independent of diabetogenesis in vivo. In contrast, exposure of isolated rabbit cardiac myocytes to STZ in vitro (37°C, 30 minutes) decreased glutathionylation of β₁ Na⁺‐K⁺ pump subunit in a concentration‐dependent manner (Figure 1C). In contrast, S961, the peptide antagonist of the insulin receptor, had no effects on β₁ pump subunit glutathionylation in vitro (Figure 1D).

The details of how we determined the appropriate dose of S961 to establish stable, persistent hyperglycemia is described by us elsewhere.²⁵ Briefly, after checking for transient hyperglycemia induced by a bolus injection of S961 in rabbits, we examined the effects of continuous subcutaneous infusion at different rates, and found that mean blood glucose values increased with an increase in the rate of S961 infusion.²⁵ On the basis of our dose titration experiments, we selected an infusion rate of 12 μg/kg per hour for the study. Continuous S961 infusion at this rate induced a relatively rapid rise in blood glucose levels in rabbits that remained stably high during the 7 days of infusion (Figure 2A). Since infusion of vehicle (normal saline) alone in 5 rabbits had no effect oxidative modification of the target proteins (data not shown), we did not use sham infusions in controls. Mean blood glucose levels in rabbits infused for 7 days with this rate compared to levels in controls are shown in Figure 2B. The S961‐induced increase in blood glucose levels was associated with an increase in plasma insulin levels (Table) consistent with a counter‐regulatory response to the competitive insulin receptor blockade shown in in vitro studies.¹⁸ C‐peptide levels were not within the detection limits of the assay (Table). The rabbits gained weight normally during the study (Table). S961 infusion had no effect on serum Na⁺ or K⁺, renal function, and cholesterol or triglyceride levels (Table). To examine the impact of S961 model on vascular physiology, we measured endothelium‐dependent relaxation in aortic rings precontracted with phenylephrine in organ chambers. S961 hyperglycemia reduced acetylecholine‐induced relaxation (Figure 2C). In contrast, sodium nitroprusside–induced relaxation was unchanged (Figure 2D).

We co‐administered the β₃ AR agonist CL by continuous infusion in hyperglycemic rabbits for the last 3 days of the study. Treatment with CL had no effect on plasma glucose or insulin levels (Table). CL also had no effect on heart rate but it reduced blood pressure (Table). S961‐induced hyperglycemia impaired endothelial function as measured by acetylcholine‐induced relaxation in aortic rings of hyperglycemic rabbits, which was restored by in vivo CL treatment (Figure 3A). There was a marked increase in O₂^•− levels in aorta of hyperglycemic rabbits as measured by 2‐OH‐E⁺ levels, the specific product of DHE oxidation by O₂^•− (Figure 3B). Treatment with CL reduced O₂^•− levels in aorta of hyperglycemic rabbits (Figure 3B). While NO levels, measured by electron paramagnetic resonance analysis of NO‐Fe(DETC)₂ complexes, were decreased in aorta of hyperglycemic rabbits, they were restored to almost normal levels by treatment with CL (Figure 3C).

Since hyperglycemia was associated with increased O₂^•− and decreased NO levels, we next examined whether hyperglycemia had induced eNOS uncoupling. Indeed, consistent with the pattern of changes in the free radicals, we found an increase in eNOS glutathionylation in aorta of hyperglycemic rabbits (Figure 3D). In vivo infusion of CL abolished the increase in hyperglycemia‐induced eNOS glutathionylation (Figure 3D). Taken together, these data show that CL enhances NO bioavailability, decreases oxidative stress, recouples eNOS, and improves endothelial function in S961‐induced hyperglycemia.

Since the Na⁺‐K⁺ pump plays a major role in regulation of many essential cellular processes in the vasculature including regulation of vascular tone, we next examined the effects of CL on the function of the pump, previously reported by our group to be regulated by redox‐dependent mechanisms in the vasculature.²³ Hyperglycemia decreased K⁺‐induced relaxation, which is dependent on Na⁺‐K⁺ pump reactivation upon a switch from nominally K⁺‐free to K⁺‐containing solutions in the tissue bath²² (Figure 4A). CL enhanced K⁺‐induced relaxation in aortic rings of hyperglycemic rabbits (Figure 4A). Consistent with this functional effect, CL abolished hyperglycemia‐induced increase in β₁ subunit glutathionylation (Figure 4B).

Because O₂^•− levels were significantly higher in vessels of hyperglycemic rabbits, and since NADPH oxidase is a major source of vascular O₂^•− in diabetes, we examined whether activation of the oxidase contributed to the redox stress in hyperglycemia. Aortic rings from control and hyperglycemic rabbits were incubated with the membrane‐permeable peptide gp91ds‐tat ex vivo. The peptide had no effect on acetylcholine‐induced relaxation in control rings (Figure 5A), but it improved relaxation in rings from hyperglycemic rabbits (Figure 5B), suggesting that increased constitutive NADPH oxidase activity contributed to the hyperglycemia‐induced endothelial dysfunction. Accordingly, gp91ds‐tat had no effect on eNOS glutathionylation in control rabbits (Figure 5C), but it decreased eNOS glutathionylation in aorta of hyperglycemic rabbits (Figure 5D).

Activation of NADPH oxidase depends on translocation of its cytosolic p47^phox subunit to the membrane, a translocation that is blocked by the cell‐permeable peptide gp91ds‐tat. We examined co‐immunoprecipitation of p47^phox subunit with eNOS as an index of p47^phox translocation to the membrane. p47^phox/eNOS co‐immunoprecipitation was increased in hyperglycemic rabbits, with CL treatment significantly reducing it (Figure 5E), thus pointing to a probable effect of CL in decreasing hyperglycemia‐induced activation of the oxidase.

Glutathionylation is a reversible reaction, with de‐glutathionylation catalyzed by Grx1. Since activation of Grx1 is believed to occur via translocation to its targets,²⁶ we examined association of Grx1 with eNOS and β₁ Na⁺‐K⁺ pump subunit by co‐immunoprecipitation as a surrogate for its activation. While co‐immunoprecipitation of both eNOS/Grx1 and β₁ pump subunit/Grx1 was markedly decreased in hyperglycemia, CL treatment significantly increased these associations without changing Grx1 abundance (Figure 5F), thus suggesting that CL might have enhanced de‐glutathionylation by Grx1.

To assess the relevance of our findings in the S961 model of hyperglycemia to human diabetes, we examined the changes in eNOS glutathionylation in vessels harvested from diabetic patients and their matched controls (both groups had hypertension, dyslipidemia, and ischemic heart disease with no differences in age [70±3 versus 69±4 years]). Similar to the experimental model, there was an increase in eNOS‐GSS in vessels of diabetic patients (Figure 6A). CL stimulated NO release in human endothelial cells (Figure 6B) and it decreased the levels of eNOS‐GSS in the vessels of diabetic patients (Figure 6C). In contrast to signaling in cardiac myocytes, β₃ AR stimulation in human vessels was not mediated by differential phosphorylation of eNOS^{27, 28} (Figure 7A) nor through nNOS²⁸ as we did not detect a signal for nNOS expression in human endothelial cells or vessels of diabetic and nondiabetic patients, regardless of exposure to CL (Figure 7B). This might be due to very low constitutive expression of nNOS in endothelial cells.²⁹ Taken together, these human data suggest the presence and support the potential significance of the redox pathways that we have examined in detail in the S961 model.