Unconjugated Bilirubin Mediates Heme Oxygenase-1–Induced Vascular Benefits in Diabetic Mice

Cardiovascular disease is the leading cause of death and disability for patients with diabetes (1). Hyperglycemia and insulin resistance reduce nitric oxide (NO) bioavailability and diminish the antiatherogenic capacity of the endothelium, resulting in platelet aggregation, increased vascular contractility, and accelerated atherosclerosis. Thus, strategies that preserve endothelial function may help alleviate diabetic vasculopathy (2).

Heme oxygenase (HO) catalyzes heme to form carbon monoxide (CO), Fe²⁺, and biliverdin; the latter is converted into unconjugated bilirubin by biliverdin reductase (BVR) (3). Two distinct isoforms of HO have been identified: heme oxygenase 1 (HO-1) and heme oxygenase 2 (HO-2). Although HO-2 is constitutively expressed, the expression of HO-1 is normally low, but inducible. HO-1 can be upregulated by stimuli, such as lipopolysaccharide and H₂O₂, in various cells or organs (4,5). Limited studies show an altered HO-1 expression under metabolic conditions. HO-1 expression and activity is increased in human umbilical vein endothelial cells (HUVECs) when exposed to 10 mmol/L glucose for 48 h compared with 5.5 mmol/L glucose but remains unchanged when exposed to 20 mmol/L glucose for 48 h (6). HO-1 mRNA is downregulated in skeletal muscle of diabetic patients but is increased in the liver and visceral fat of obese patients (7,8), suggesting that HO-1 expression can be differently regulated in major metabolic organs under different disease states.

Growing evidence suggests HO-1 is a potential target for intervention of diabetes. Induction of HO-1 in obese mice increases plasma adiponectin, improves insulin sensitivity, and decreases visceral and abdominal adiposity and plasma proinflammatory cytokines (9). HO-1 overexpression reduces lymphocytic infiltration in Langerhans islets and retards the progression of type 1 diabetes (10). HO-1 also protects against high glucose (HG)–induced impairment of endothelial-dependent relaxations (EDRs) in rat aortas (11) and HG-induced endothelial cell apoptosis (6,12). These studies indicate that HO-1 induction may serve as a potential therapeutic strategy to treat diabetic vascular complications. However, the precise intracellular mechanisms mediating the vasoprotective benefits of HO-1 remain largely unexplored.

Bilirubin is converted from biliverdin by BVR. Clinical studies show an inverse association between the serum bilirubin concentration with the incidences of myocardial infarction, peripheral artery disease, and stroke (13–16), and the serum bilirubin level is lower in diabetic patients (17). Of note, Gilbert syndrome patients, who have higher levels of serum unconjugated bilirubin, are less prone to major adverse cardiovascular events (18), and individuals with higher serum bilirubin were better protected from developing metabolic disorders in a 4-year retrospective longitudinal study (19). Bilirubin is a potent antioxidant, and efforts have been directed to determine whether bilirubin can be used to treat diseases associated with oxidative stress. Indeed, atazanavir, a drug that raises unconjugated bilirubin levels, enhances plasma antioxidant capacity and improves EDRs in diabetic patients (20).

This study examined the hypothesis that unconjugated bilirubin mediates the vascular benefits of HO-1 induction to restore impaired endothelial function in diabetic db/db mice and investigated the possible mechanisms involved. The present new findings support the clinical observation that bilirubin is vasoprotective in diabetes.

Animal care and the experimental protocol were approved by the Chinese University of Hong Kong (CUHK) Animal Research Ethics Committee in compliance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health Publication No. 85-23, revised 1996).

Twelve-week-old male diabetic db/db mice on a C57BL/KsJ background and nondiabetic littermates db/m⁺ mice were supplied by the CUHK Animal Service Center, kept in a temperature-controlled room (∼23°C) with a 12-h light/dark cycle, and fed a standard diet and water ad libitum. The db/db mice were treated with vehicle, hemin (HO-1 inducer, 25 mg/kg body weight [BW], three times weekly intraperitoneally [i.p.]), hemin+SnMP (HO-1 inhibitor; 20 mg/kg BW, three times weekly i.p.), hemin+scramble virus (10⁹ plaque-forming units [pfu]), hemin+HO-1 short hairpin (sh)RNA virus (10⁹ pfu), or bilirubin (5 mg/kg BW, three times weekly i.p.). The db/m⁺ mice were treated with vehicle or hemin (25 mg/kg BW, three times weekly i.p.).

Diet-induced obese (DIO) mice were generated by feeding 6-week-old C57BL/6J mice a high-fat rodent diet with 45% kcal% fat (D12451; Research Diets Inc., New Brunswick, NJ) for 10 weeks. Mice were then treated with hemin or vehicle for 2 weeks (25 mg/kg BW, three times weekly i.p.).

SnMP was purchased from Frontier Scientific (Logan, UT), and all other reagents were from Sigma-Aldrich (St. Louis, MO), unless otherwise stated.

After mice were killed, thoracic aortas were removed, placed in ice-cold Krebs solution (in mmol/L: 119 NaCl, 4.7 KCl, 2.5 CaCl₂, 1 MgCl₂, 25 NaHCO₃, 1.2 KH₂PO₄, and 11 d-glucose), and cut into ring segments. Changes in isometric tension in aortic rings were recorded by Multi Myograph System (Danish Myo Technology A/S, Aarhus N, Denmark) (21). The baseline tension was set at 3 mN, and all rings were equilibrated for 60 min before the start of the experiments.

Aortic rings were contracted by 1 μmol/L phenylephrine (Phe) to induce a sustained tension before acetylcholine (ACh) (10⁻⁸ to 10⁻⁵ mol/L) was added cumulatively to trigger EDRs. EDRs induced by the NO donor sodium nitroprusside (SNP) (10⁻⁸ to 10⁻⁵ mol/L) were compared in arteries from different groups.

Mouse aortas were cultured for 24 h at 37°C in DMEM (Gibco, Gaithersburg, MD) supplemented with 10% FBS (Gibco) and 100 IU/mL penicillin plus 100 μg/mL streptomycin, as previously described (22). The db/db mouse aortas were treated with 5 μmol/L hemin or 1 μmol/L unconjugated bilirubin, in control and in the presence of 30 μmol/L SnMP or 5 μmol/L Akt inhibitor V.

For ex vivo viral transduction, mouse aortas were cultured with 10⁹ pfu BVR shRNA or GFP shRNA adenovirus for 24 h and then treated with hemin or unconjugated bilirubin for another 24 h.

The use of human specimens was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee. Human renal arteries were obtained after informed consent from patients without cardiometabolic complications and from diabetic patients undergoing nephrectomy due to neoplasm at ages between 50 and 80 years. Diabetes in patients was defined as a fasting plasma glucose level ≥7.0 mmol/L.

Mice were loaded with glucose (1.2 g/kg BW) for the oral glucose tolerance test after an 8-h fast. For the insulin tolerance test, mice were injected with insulin at 0.75 units/kg BW after a 2-h fast. Blood glucose was measured at 0, 15, 30, 60, and 120 min with an Ascensia ELITE glucometer (Bayer HealthCare, Mishawaka, IN).

Plasma insulin levels were assayed by enzyme immunoassay (Mercodia AB, Uppsala, Sweden). Plasma levels of total cholesterol, triglyceride, HDL, and non-HDL were determined using enzymatic methods (Stanbio Laboratory, Boerne, TX).

A total of 400 μL of 60% (v/v) acetonitrile in 0.01 mol/L phosphate buffer (pH 8.0) was added to 100 μL serum or culture medium, vortexed for 30 seconds, and centrifuged for 5 min at 1,000 rpm (23). Supernatant (100 μL) was applied on the Agilent 6460 Triple Quadrupole Mass Spectrometer (Agilent Technologies, Santa Clara, CA) with ultra-high-performance liquid chromatographic (UPLC) tandem mass spectroscopy (MS). The chromatographic separation was performed on an ACQUITY UPLC BEH C18 column (2.1 mm × 100 mm, ID, 1.7-μm particle size; Waters, Milford, MA) at 4°C, with the mobile phase consisting of A) 0.1% formic acid in water and B) 0.1% formic acid in acetonitrile by a gradient elution method (55–100% of B from 0 to 12.5 min, 100% of B from 12.5 to 13 min, and 55% of B from 13.1 to 16 min). The injection volume was 3 μL, and the flow rate was 0.4 mL/min. The elution was directed to the MS without splitting. The temperature of the autosampler was set at 4°C throughout the analysis. The MS, which was equipped with an ion source of electrospray ionization (Agilent Jet Stream) in negative ion mode, was used for bilirubin detection. The source parameters were set as gas temperature, 300°C; gas flow, 8 L/min; nebulizer, 30 ψ; sheath gas temperature, 350°C; sheath gas flow, 10 L/min; capillary voltage, 3,500 V; and nozzle voltage, 1,000 V. The MS recordings were done in multiple reactions monitoring mode. Nitrogen was used as the collision gas, and the monitor ion and collision energy were m/z 583.2 → 285.2 and 20 eV, respectively.

Mouse aortas were treated with hemin (5 μmol/L) or bilirubin (1 μmol/L) for 24 h, followed by the addition of 10 μmol /L ACh (10 min) to stimulate NO generation with the presence of nitrate reductase to reduce nitrate to nitrite. Aortas were homogenized, and the nitrite level in the supernatants was measured by a Griess reagent kit (Molecular Probes, Eugene, OR) and normalized by the protein content.

shRNA sequence targeting mouse HO-1 was obtained from Sigma-Aldrich (TRCN0000234077). The 5′-GATCCACAGTGGCAGTGGGAATTTATCTCGAGATAAATTCCCACTGCCACTGTTTTTTA-3′ and 5′-AGCTTAAAAAACAGTGGCAGTGGGAATTTATCTCGAGATAAATTCCCACTGCCACTGTG-3′ were annealed and cloned into the pAAV-ZsGreen -shRNA (YRGene) shuttle vector to construct pAAV-ZsGreen-HO-1 shRNA plasmid, and then it was cotransfected into HEK-293T with RGDLRVS-AAV9 cap plasmid (a gift of Dr. O.J. Müller, University Hospital Heidelberg, Heidelberg, Germany) (24) and pHelper (Stratagene, La Jolla, CA). Adeno-associated virus (AAV) viral particles were harvested as previously reported (25). Mice were injected with 10⁹ pfu HO-1 shRNA or scramble (empty vector) virus via tail vein. The knockdown efficiencies were confirmed by Western blotting.

The U6 promoter and 1.9-kb stuffer sequence were excised from pLKO.1 (Addgene, Cambridge, MA) with NotI/XhoI and cloned into pShuttle. shRNA targeting mouse BLVRA (Sigma-Aldrich, TRCN0000042128) was generated similarly to the pLKO.1 system (26). Briefly, the forward (CCGGGCCAAATGTAGGAGTCAATAACTCGAGTTATTGACTCCTACATTTGGCTTTTTG) and reverse (TCGAGAAAAAGCCAAATGTAGGAGTCAATAACTCGAGTTATTGACTCCTACATTTGGC) oligos were annealed and ligated to pShuttle-U6 predigested with AgeI and XhoI. The pAd-U6-shBLVRA plasmid was produced as reported previously (27) and linearized by PacI and transfected to Hek-293 cells to produce the viral particle.

The db/db mice were injected with HO-1 overexpressing adenovirus (10⁹ pfu, a gift from Dr. Jun Yu, Prince of Wales Hospital, Hong Kong) through the tail vein and kept for 4 days before being killed.

HUVECs were purchased from Lonza (San Diego, CA) and cultured in Endothelial Cell Growth Medium (EGM, Lonza) supplemented with 10% FBS plus 1% penicillin/streptomycin. HUVECs were transfected with a dominant-negative Akt construct (DN-Akt) by electroporation using the Nucleofector II machine (Amaxa/Lonza, Walkersville, MD) according to the manufacturer’s instruction.

NO production in HUVECs was measured by Fluoview FV1000 laser scanning confocal system (Olympus, Tokyo, Japan) using 4-amino-5-methylamino-2′,7′-difluorofluorescein diacetate (DAF-DA, Invitrogen, Carlsbad, CA) as the indicator. The amount of NO produced in response to A23187 (100 nmol/L) was evaluated by measuring fluorescence intensity at excitation 488 nm and emission 515 nm. Changes in [NO]_i were displayed as a ratio of fluorescence intensity after (F₁) and before (F₀) the addition of A23187 (F₁-to-F₀). For reactive oxygen species (ROS) measurement, HUVECs were incubated for 30 min in 5 μmol/L CM-2′,7′-dichlorodihydrofluorescein diacetate (Invitrogen) and measured using Fluoview FV1000 confocal system at excitation 488 nm and emission 520 nm. To detect the NO-mediated antioxidant capacity of bilirubin, NO scavenger–oxidized hemoglobin (20 μmol/L) was used together with bilirubin in HG-treated HUVECs. HUVECs were incubated with 5 μmol/L dihydroethidium (Invitrogen) for 30 min, and fluorescence intensity was measured using the Fluoview FV1000 confocal system at excitation 515 nm and emission 585 nm (Olympus).

Protein lysates from mouse aortas or HUVECs were separated by electrophoresis and transferred onto an Immobilon-P polyvinylidene difluoride membrane (Millipore Corp., Bedford, MA). The blots were blocked with 1% BSA in 0.05% Tween-20 PBS for 1 h and incubated with primary antibodies overnight at 4°C, including polyclonal anti-eNOS (1:500; Abcam, Cambridge, U.K.), anti–phospho (p)-eNOS Ser¹¹⁷⁷ (1:1,000; Abcam), polyclonal anti–HO-1 (1:1,000; Assay Designs, Ann Arbor, MI), monoclonal anti–p-Akt Thr³⁰⁸ (1:1,000; Cell Signaling, Danvers, MA), monoclonal anti-Akt1 (1:1,000; Cell Signaling), and polyclonal anti-insulin receptor substrate 1 (1:1,000; Cell Signaling). Blots were washed and incubated with horseradish peroxidase (HRP)–conjugated secondary antibody (DakoCytomation, Carpinteria, CA). Protein expression was normalized by monoclonal anti-GAPDH (1:10,000; Ambion, Austin, TX). Protein expression was determined by a Flurochem densitometer (Alpha Innotech Corp., San Leandro, CA).

Human renal arteries were frozen in optimal cutting temperature compound (Sakura Finetek, Torrance, CA), cut into 10-μm sections on a microtome (Leica Microsystems, Wetzlar, Germany), and fixed with 4% paraformaldehyde. Sections were washed in PBS, blocked in 5% normal donkey serum (Jackson ImmunoResearch, West Grove, PA), and incubated with anti-BVR antibody (1:200; Enzo Life Sciences, Farmingdale, NY) or PBS overnight at 4°C. Biotin-conjugated goat anti-rabbit secondary antibodies (1:500; Jackson ImmunoResearch), streptavidin-HRP conjugate (1:500; Zymed Laboratories, South San Francisco, CA), and DAB substrate (Vector Laboratories, Burlingame, CA) were used to visualize positive staining. Photomicrographs were taken under a Nikon TI-S microscope (Tokyo, Japan).

Results are means ± SEM of n experiments. EDR was expressed as percentage reduction in Phe-induced contraction. Concentration-response curves were constructed using GraphPad Prism 4.0 software (GraphPad Software, Inc., San Diego, CA). Statistical significance was determined by two-tailed Student t test or one-way ANOVA, followed by Bonferroni post hoc tests when more than two treatments were compared. P < 0.05 indicates statistical significance.

ACh-induced EDRs were impaired in aortas of diabetic db/db mice compared with those from db/m⁺ mice. Hemin treatment for 2 weeks restored EDRs in db/db mouse aortas, which were reversed by cotreatment with the HO-1 inhibitor SnMP (Fig. 1A). By contrast, SNP-induced endothelium-independent relaxations were similar in all groups (Supplementary Fig. 1). Hemin treatment induced an approximately threefold increase of HO-1 expression in aortas from db/db and db/m⁺ mice, which was unaffected by SnMP (Fig. 1B). Ex vivo treatment with hemin (5 μmol/L for 24 h) or tail vein injection of HO-1 overexpressing adenovirus (4 days) also improved EDRs in db/db mouse aortas (Fig. 1C and D) and elevated HO-1 expression (Supplementary Fig. 2A and B). The effect of hemin was again reversed by cotreatment with 30 μmol/L SnMP (Fig. 1C). To confirm that HO-1 mediates vascular benefits of hemin in db/db mice, HO-1 shRNA AAV was constructed and injected into db/db mice via tail vein, and these mice were treated with hemin for 2 weeks. Compared with scramble virus, HO-1 shRNA AAV reversed the effect of hemin on EDRs and HO-1 expression (Fig. 1E and F). In addition, we also used DIO mice to confirm the findings in db/db mice. Hemin administration for 2 weeks augmented EDRs and HO-1 expression in DIO mouse aortas (Supplementary Fig. 3A and B), suggesting HO-1 is most likely to mediate hemin-induced improvement of EDRs in diabetic mice.

The phosphorylation levels of Akt (Thr³⁰⁸) and eNOS (Ser¹¹⁷⁷) in aortas were lower in db/db mice than in db/m⁺ mice. In vivo hemin treatment restored the diminished phosphorylation of Akt and eNOS, and such effects were reversed by cotreatment with SnMP (Fig. 2A and B). Chronic hemin treatment also increased phosphorylation of Akt and eNOS in DIO mouse aortas (Supplementary Fig. 3C and D). HO-1 shRNA AAV transduction inhibited hemin-stimulated phosphorylation of Akt and eNOS (Fig. 2C–E). The hemin-improved EDRs were reversed by the phosphatidylinositide 3-kinase (PI3K) inhibitor wortmannin (100 nmol/L) or Akt inhibitor V (5 μmol/L) (Fig. 2F). However, neither inhibitor affected ACh-induced relaxations of db/m⁺ mouse aortas (data not shown).

The serum concentration of unconjugated bilirubin was 77.5 ± 7.8 nmol/L in db/db mice and 350.0 ± 3.9 nmol/L in db/m⁺ mice. Hemin treatment increased bilirubin to 201.4 ± 13.6 nmol/L, which was inhibited by SnMP (Fig. 3A). Ex vivo bilirubin treatment (1 μmol/L for 24 h) improved EDRs in db/db mouse aortas, and this effect was inhibited by coincubation with Akt inhibitor V (5 μmol/L) but not by SnMP (30 μmol/L) (Fig. 3B). To investigate whether bilirubin mediated the beneficial effect of HO-1, BVR shRNA adenovirus was constructed and it reversed the vascular benefit of hemin (Fig. 3C) but not that of bilirubin (Fig. 3D). Successful inhibition of BVR expression by shRNA adenovirus was shown by Western blotting (Fig. 3E) and immunohistochemical staining (Fig. 3F). Next, we treated db/db mice with bilirubin for 2 weeks. Chronic bilirubin administration improved EDRs in db/db mouse aortas (Fig. 4A), accompanied by increased phosphorylation of Akt (Thr³⁰⁸) and eNOS (Ser¹¹⁷⁷) (Fig. 4B–D). These results further indicate that bilirubin is most likely to mediate the effect of HO-1 induction to improve endothelial function.

HG for 36 h reduced Ca²⁺ ionophore A23187 (100 nmol/L)–stimulated NO elevation detected by DAF-DA in HUVECs compared with normal glucose (NG) (Fig. 5A and B). Hemin or bilirubin restored the HG-impaired NO production (Fig. 5A, B, and C). Ex vivo treatment with hemin and bilirubin also raised the nitrite level (which reflects tissue NO level) in db/db mouse aortas (Supplementary Fig. 4). Akt inhibitor V (5 μmol/L) abolished the effect of hemin and bilirubin, whereas SnMP only inhibited the effect of hemin on NO production (Fig. 5B and C). Inhibition of Akt activity by Ad-DN-Akt, the plasmid expressing DN Akt, abolished the effect of both hemin and bilirubin (Fig. 5D). In addition, treatment with hemin or bilirubin increased phosphorylation of Akt and eNOS in HG-treated HUVECs. Again, the effect of hemin was abolished by SnMP, whereas the effect of hemin and bilirubin was reversed by Akt inhibitor V (Fig. 5E–H).

GFP shRNA adenovirus served as the control virus. Hemin increased the bilirubin level in the medium culturing HUVECs, which was diminished by BVR shRNA adenovirus (Supplementary Fig. 5A). The stimulatory effect of hemin on Akt and eNOS phosphorylation in HG-treated HUVECs was reversed by BVR shRNA adenovirus (Fig. 6A–C). Likewise, hemin-stimulated NO production in HUVECs was inhibited by BVR shRNA adenovirus (Supplementary Fig. 5B and C). By contrast, the effect of bilirubin remained unchanged by BVR shRNA virus (Fig. 6D–F and Supplementary Fig. 5B and C).

The HbA_1c concentrations in diabetic patients are presented in Supplementary Table 1. Compared with renal arteries from patients without cardiometabolic complications (control), EDRs in renal arteries from diabetic patients were impaired, which were augmented after 24-h exposure to bilirubin (1 μmol/L) (Fig. 7A and B). BVR expression was detected in the endothelium in human renal arteries by immunohistochemistry (Fig. 7C).

Although HO-1 is known to benefit vascular function, whether such beneficial effects are directly or indirectly induced by HO-1 remains poorly understood. HO-1 degrades heme to form biliverdin, CO, and Fe²⁺; the former is further converted into unconjugated bilirubin. Bilirubin is increasingly recognized to be vasoprotective in cardiometabolic diseases (18,20), and the total plasma bilirubin concentration is lower in diabetic patients (17). The current study also shows a reduced serum level of unconjugated bilirubin in db/db mice, which is increased after 2 weeks of hemin treatment. Importantly, ex vivo treatment with unconjugated bilirubin at 1 μmol/L (a concentration comparable to that detected in the serum of nondiabetic mice) and intraperitoneal administration of bilirubin to db/db mice for 2 weeks resulted in vascular benefits similar to those of hemin to restore EDRs. More definitely, silencing BVR using shRNA adenovirus reverses ex vivo hemin-induced improvement of EDRs in db/db mouse aortas and hemin-stimulated NO production in HUVECs. By contrast, the beneficial effects of bilirubin were unaffected by BVR shRNA. Taken together, bilirubin is the most likely mediator of the vasoprotective action of HO-1 induction. Likewise, bilirubin is also reported to mediate the antiatherogenic action of HO-1 through inhibiting vascular inflammation (28).

The current study provides the first piece of evidence showing that ex vivo treatment with bilirubin improves EDRs in renal arteries from diabetic patients, further suggesting that bilirubin is vasoprotective in humans. It should be noted that the renal vascular response to bilirubin in the diabetic patients might have been influenced by the presence of renal carcinoma and by the different duration of diabetes and HbA_1c levels (Supplementary Table 1). Also, it will be important in the future to investigate whether bilirubin produces similar vascular benefit in other arteries, such as coronary arteries, carotid arteries, and femoral arteries, which are more clinically relevant.

PI3K/Akt signaling is important to preserve endothelial function in diabetic mice by increasing eNOS phosphorylation at Ser¹¹⁷⁷ (22) and endothelial progenitor cell regenerative capacity (29,30). The current study reveals a critical role of Akt signaling in mediating the vascular benefits of hemin and bilirubin, because inhibition of Akt activity abolishes the ability of hemin and bilirubin to improve EDRs in db/db mice and to restore the impaired NO production in HG-treated HUVECs.

The biological action of bilirubin has long been considered due to its antioxidant property (31,32). Bilirubin scavenges peroxyl radicals and reduces lipid peroxidation in vitro (33), and also downregulates NAD(P)H oxidase activity and protects against diabetic nephropathy in rodents in vivo (34). Bilirubin also mediated the effect of HO-1 to reduce ROS and nitrogen species in lipopolysaccharide-treated HUVECs (35). Reduced ROS contribute to the improved EDRs; therefore, the antioxidant activity of bilirubin may partially account for its vascular benefits. The current study shows that bilirubin also increases NO production. NO can interact with superoxide anions to lower ROS levels. Our results show that bilirubin inhibited HG-stimulated ROS generation in HUVECs, which was attenuated by Akt inhibitor V and NO scavenger oxidized–hemoglobin (Supplementary Fig. 6), suggesting that increased NO is likely to mediate a significant part of the effect of bilirubin to reduce ROS in endothelial cells. Thus, the increased NO bioavailability may directly contribute to bilirubin-induced improvement in EDRs and also indirectly by reducing ROS.

CO, one of the metabolic products of heme degradation, is reported to affect vascular reactivity, causing either endothelium-independent relaxations (36) or contractions (37). Most studies show that CO inhibits NO-mediated vasodilatation or NO production in endothelial cells (38,39). Unlike hemin or bilirubin, ex vivo treatment with the CO-releasing molecule tricarbonyl-dichloro-ruthenium dimmer does not improve EDRs in db/db mouse aortas (Supplementary Fig. 7). Taken together with results from experiments using BVR shRNA virus and chronic bilirubin treatment, the present results suggest a minimal involvement of CO in hemin-induced endothelial protection in diabetic mice.

It is worthwhile to note that HO-1 induction in vivo moderately reduced fasting glucose and insulin levels in db/db mice and also improved insulin sensitivity (Supplementary Fig. 8). Improved insulin sensitivity is more likely a systematic effect, because HO-1 induction increased expression of insulin receptor substrate 1 and Akt phosphorylation in liver, adipose tissue, and skeletal muscle, which play an important role in regulating insulin sensitivity (Supplementary Fig. 9). Previous studies also showed that HO-1 induction improves insulin sensitivity and reduces adiposity in diabetic mice and rats, which is associated with increased adiponectin release (9,40). Thus, the metabolic benefit may also contribute, albeit to a lesser degree, to the restoration of endothelial function in hemin-treated diabetic mice through HO-1 upregulation in adipose tissues (40,41).

In summary, the current study provides novel experimental evidence that bilirubin mediates HO-1 induction–induced vascular benefits through activation of the PI3K/Akt/eNOS signaling cascade in diabetic mice. The present findings enhance the prospects of using HO-1 inducers or bilirubin to ameliorate diabetic vasculopathy.

Funding. This study was supported by the Hong Kong Research Grants Council (CUHK2/CRF/12G, T12-402/13-N, T12-705/11), the National Basic Research Program of China (2012CB517805), and the Natural Science Foundation of China (91339117).

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

Author Contributions. J.L. designed and conducted the experiments, analyzed the data, and prepared the manuscript. L.W., L.L., Y.Z., and S.L.W. conducted the experiments and analyzed the data. X.Y.T., W.T.W., and Y.H. designed the experiments and prepared the manuscript. Q.-B.H. and H.-M.H. performed bioassays. N.W., Z.-Y.C., J.Y., and X.Y. conducted biochemical assays, provided plasmids, and assisted with discussion. C.-F.N. provided the human specimen. 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.