Systemic Expression of Heme Oxygenase-1 Ameliorates Type 1 Diabetes in NOD Mice

We used the AAV helper-free system from Stratagene (La Jolla, CA) for vector construction. Mouse HO-1 cDNA was obtained by RT-PCR of RNA isolated from heme-treated murine J774 cells and subcloned into the pCMV-MCS plasmid. Recombinant AAV vector was then prepared according to the manufacturer's instructions. The large-scale production and purification of AAV virus were conducted using heparin affinity chromatography as described by Clark et al. (27). The recombinant AAV particles were quantified by real-time PCR using a PE-Applied Biosystems Prism 7700 sequence detector (Foster City, CA) and fluorescent probe sets. The primers and probe used for quantification of AAV-HO-1 were as follows: HO-1 5′ primer, 5′-AGGTGATGCTGACAGAGGAACAC-3′; HO-1 3′ primer, 5′-TAGCAGGCCTCTGACGAAGTG-3′; HO-1 probe, 5′-6-carboxy-fluorescein-AAGACCAGAGTCCCTCACAGATGGCG-6-carboxy-tetramethyl-rhodamine-3′.

Female NOD and NOD.scid mice were purchased from The Jackson Laboratories (Bar Harbor, ME) and kept in specific pathogen-free conditions. Female NOD mice at 9 weeks of age were subjected to retro-orbital plexus injection of saline or the indicated amounts of AAV-HO-1 in 50 μl saline. Some NOD mice at 8–9 weeks of age received two intraperitoneal injections of cyclophosphamide (200 mg/kg body wt) with a 1-week interval to facilitate the onset of diabetes. To test the effect of CO, animals were exposed to 250 ppm CO gas as described previously (28) for 2 h twice per week. Animals were monitored twice per week for hyperglycemia as defined by two consecutive nonfasting blood glucose levels >240 mg/dl. Mice were killed at 24 weeks of age. All animal procedures were approved by the institutional animal care and use committee of Academia Sinica.

Splenocytes from all mice killed at 24 weeks of age were isolated and cultured at 5 × 10⁵ cells/well in 200 μl RPMI 1640 containing 10% fetal bovine serum in a 96-well, flat-bottom plate. After incubation with or without concanavalin A (ConA; 5 μg/ml per well) for 48 h, media were collected for the measurement of cytokine production. For flow cytometric experiments, splenocytes (1 ×10⁶ cells/ml) were preincubated with purified rat anti-mouse CD16/CD32 (eBioscience) to block Fc receptor binding, followed by staining with fluorescein isothiocyanate (FITC)-conjugated anti-CD4 antibody (BD Biosciences), phycoethrin-conjugated anti-CD25 antibody (BD Biosciences), APC-conjugated anti-CD11c antibody (BD Biosciences), or FITC-conjugated anti–I-A^g7 MHC II antibody (BD Biosciences) as indicated. To examine intracellular Foxp3 expression, splenocytes were immunostained with an APC–anti-mouse Foxp3 staining set (eBiosciences) according to the manufacturer's instruction. Flow cytometry data were acquired and analyzed using a FACS Calibur (Becton Dickinson) and Cell Quest software (version 3.3; Becton Dickinson).

Freshly isolated splenocytes from 24-week-old saline-treated or AAV-HO-1–treated (2.5 × 10¹⁰ genome particles) NOD mice were injected at 1 × 10⁷ cells/mouse into 5-week-old female NOD.scid mice via the retro-orbital plexus. Hyperglycemia was then assessed twice per week.

Cytokines interleukin (IL)-2, IL-4, IL-10, and γ-interferon (IFN-γ) were measured by ELISA kits (R&D Systems).

The pancreas was removed, fixed in Bouin's solution (Sigma) for 16 h, embedded in paraffin, and sectioned at 5 μm. Sections were hematoxylin/eosin stained and evaluated on a blinded basis by two individuals using scoring criteria described by others (29): stage 0, normal islet; stage 1, peri-insulitis; stage 2, insulitis in <50% of the islet; and stage 3, insulitis in >50% of the islet. At least 20 islets from each pancreas were examined.

Sections were deparaffinized, rehydrated, and pretreated with target retrieval solution (Dako) at 95°C for 30 min. Endogenous peroxidase was blocked in a solution of 3% H₂O₂ for 10 min at room temperature. After incubation with 5% bovine serum albumin in PBS at 37°C for 30 min, sections were incubated with rabbit polyclonal anti–HO-1 (1:100; Stressgen) at 37°C for 1 h, followed by three washes with PBS containing 0.1% Tween-20 (PBST). Sections were then incubated with horseradish peroxidase–conjugated secondary antibody (1:200) for another 1 h at 37°C. After three washes, antigen was visualized after incubation with diaminobenzidine/H₂O₂ followed by counterstain with hematoxylin reagent. Negative control was performed by using rabbit normal IgG as the primary antibody.

Tissues were homogenized in lysis buffer containing 25 mmol/l Tris-HCl, pH 7.9, 1% Triton X-100, 10 mmol/l EDTA, 1 mmol/l sodium vanadate, 1 mmol/l phenylmethylsulfonyl fluoride (PMSF), and 1 μg/ml leupeptin. After centrifugation at 14,000 rpm for 30 min, the supernatant was removed, and protein concentration was determined by Bio-Rad protein assay. An equal amount of protein (50 μg) from each sample was separated by SDS-PAGE and transblotted onto nitrocellulose membranes. After blocking for 12 h with 5% nonfat milk in PBST, membranes were incubated with rabbit anti–HO-1 antibody (Stressgen), rabbit anti–HO-2 antibody (Santa Cruz), or rabbit anti–glyceraldehyde-3-phosphate dehydrogenase antibody (Santa Cruz) for 1 h at 37°C. After three washes with PBST, membranes were incubated with horseradish peroxidase–conjugated secondary antibody for 1 h. Antigen was then detected using the ECL system.

Tissues were homogenized in ice-cold 0.1 mol/l potassium phosphate buffer, pH 7.4, containing 1 mmol/l EDTA, and 0.5 mmol/l PMSF and centrifuged at 13,000 rpm at 4°C for 10 min. Supernatant proteins (1 mg) were then incubated in dark with 200 μl of reaction mixture containing mouse liver cytosol (1 mg protein), 50 μmol/l hemin, 1 mmol/l NADPH, 2 mmol/l glucose-6-phosphate, and 0.2 unit glucose-6-phosphate dehydrogenase in 0.1 mol/l potassium phosphate buffer, pH 7.4, at 37°C for 1 h. Bilirubin was then extracted with 1 ml chloroform and measured by the absorbance difference between 464 and 530 nm with an extinction coefficient of 40 mmol/l per centimeter.

Data were presented as mean ± SD and analyzed using the Mann-Whitney-Wilcoxon test. Hyperglycemia frequencies and survival rates were analyzed by the Mantel-Haenszel χ² test. P < 0.05 was considered statistically significant.

The expression levels of HO-1 and HO-2 and the net HO activity in NOD mice were first examined. Western blot analysis revealed no significant difference in the expression levels of HO-1 or HO-2 in lung, pancreas, and liver between young (10 weeks) and old (24 weeks) mice (Fig. 1A and B). However, the HO activities measured in indicated tissues appeared to be lower in old NOD mice compared with young counterparts (Fig. 1C). To evaluate the potential effect of HO-1 on type 1 diabetes development, we constructed a recombinant AAV vector carrying mouse HO-1 gene for the in vivo gene transfer in NOD mice. AAV-mediated transgene expression was first confirmed in HT1080 cells (data not shown). To assess the effect of HO-1 on type 1 diabetes, 9-week-old female NOD mice received saline or various doses of AAV-HO-1 via retro-orbital plexus injection. The incidence of hyperglycemia in each group was examined for 15 weeks. As shown in Fig. 2A, 10% of the saline-treated control mice developed hyperglycemia in the following 1–3 weeks. In contrast, the onset of diabetes was substantially delayed in mice receiving AAV-HO-1. The salutary effect of AAV-HO-1 was dose dependent, and 85% of mice receiving the highest dose of AAV-HO-1 (2.5 × 10¹⁰ virus particles/mouse) remained normoglycemic at the time of killing (P < 0.001 vs. control group). Histological assessment of pancreatic sections from the animals revealed that the extent of insulitis was inversely correlated with the dose of AAV-HO-1 received (Fig. 2C). In contrast to the control group in which only ∼15% of the examined islets were devoid of lymphocyte infiltration, >50% of the examined islets in the group treated with the highest dose of AAV-HO-1 were normal (P < 0.01 vs. control group). To confirm that this protective effect resulted from sustained transgene expression, Western blot analysis was performed to examine HO-1 expression in various tissues 15 weeks after gene delivery. As shown in Fig. 3A and B, the levels of HO-1 protein in liver and spleen, which express high basal levels of HO-1, were not significantly enhanced by AAV-HO-1 administration. In contrast, HO-1 expression in lung, heart, and pancreas was significantly higher in mice receiving AAV-HO-1 (2.5 × 10¹⁰ virus particles/mouse) compared with control counterparts. Likewise, HO activities determined in these tissues were also higher in AAV-HO-1–treated mice (Fig. 3C). When immunostaining experiment was performed with pancreatic sections, it was noted that HO-1 was primarily expressed in β-cells of islets and was much prominent in mice receiving AAV-HO-1 (Fig. 3D). Together, these results support an inverse association between HO-1 expression and the onset of autoimmune diabetes in NOD mice.

To examine whether HO-1 expression affects the balance of Th1/Th2 responses, we measured Th1 and Th2 cytokines in serum samples and in media from ConA-activated splenocytes isolated from surviving animals 15 weeks after saline or AAV-HO-1 (2.5 ×10¹⁰ particles) administration. As shown in Fig. 4A, significant decreases in serum levels of IL-2 and IFN-γ were observed in mice receiving AAV-HO-1. However, the levels of IL-4 and IL-10 were not significantly different between the saline- and AAV-HO-1–treated groups. Likewise, the production of IL-2 and IFN-γ, but not of IL-4 and IL-10, by activated splenocytes was substantially lower in AAV-HO-1–treated mice (Fig. 4B). To further confirm that HO-1 overexpression reduced the diabetogenicity of lymphocytes, an adoptive transfer experiment was performed in female NOD.scid mice. Splenocytes isolated from 24-week-old NOD mice with or without AAV-HO-1 treatment for 15 weeks were transferred into 5-week-old NOD.scid mice. Results showed that 9 of 10 mice receiving splenocytes from control donors became diabetic by 4 weeks after transfer, whereas the onset of diabetes was significantly delayed in recipients receiving splenocytes from AAV-HO-1–treated donors (P < 0.05) (Fig. 4C).

To determine whether systemic HO-1 overexpression affects the level of CD4⁺CD25⁺ Treg cells, splenocytes isolated from 24-week-old mice pretreated without or with AAV-HO-1 (2.5 × 10¹⁰ particles) for 15 weeks were subjected to flow cytometry analysis. The percentage of CD4⁺CD25⁺ Treg cells in total splenocytes of the AAV-HO-1–treated group (2.60 ± 0.12%, n = 3) was not significantly different from that of the saline-treated group (2.73 ± 0.28%, n = 5) (Fig. 5A). When the frequencies of Foxp3 expression were examined, again there was no significant difference in the percentage of CD4⁺Foxp3⁺ cells between saline and AAV-HO-1–treated groups (3.74 ± 0.35 vs. 3.57 ± 0.11%). We next examined the population of CD11c⁺ dendritic cells, which are crucial for the activation of autoreactive T-cells. The percentage of CD11c⁺ splenocytes was similar between the control and AAV-HO-1–treated groups (0.70 ± 0.14 vs. 0.74 ± 0.26%). However, when the expression level of MHC II, an index of dendritic cell maturation, in this cell population was analyzed, it was found that the percentage of CD11c⁺MHC II⁺ cells was reduced by 35% in the AAV-HO-1–treated group compared with the saline-treated control (Fig. 5B and C).

To determine whether CO, a by-product of heme degradation, has a role in HO-1–mediated type 1 diabetes protection, we exposed 9-week-old NOD mice to a low dose of CO gas (250 ppm for 2 h) twice per week. As shown in Fig. 6A, CO treatment for a 15-week period did not significantly affect the expression levels of HO-1 in tissues examined. However, the incidence of diabetes in the CO-treated group was significantly reduced compared with that of the control group (P < 0.001) (Fig. 6B). Consistently, insulitis was less severe in CO-treated animals (Fig. 6C). In contrast to control mice, in which only ∼12% of the examined islets were normal, ∼40% of examined islets were normal in CO-treated mice (P < 0.01 vs. control group). To examine whether CO treatment also affects the Th1/Th2 response, cytokine expression in ConA-activated splenocytes isolated from control and CO-treated mice was assessed. As shown in Fig. 7, the levels of IL-2 and IFN-γ, but not of IL-4 and IL-10, were significantly lower in the CO-treated group compared with the control group. To further explore whether CO exerts protective effect on mice developing overt diabetes, NOD mice at 8–9 weeks of age received two intraperitoneal injections of cyclophosphamide (200 mg/kg body wt) to accelerate the onset of diabetes. Fifty percent of mice became diabetic within 3 weeks after the first injection of cyclophosphamide (data not shown). Mice with diabetes onset were either left untreated or treated with CO gas as described above beginning at day 3 after the disease onset (day 0) twice per week. It was found that CO treatment did not affect the hyperglycemia of diabetic mice (data not shown). However, CO significantly prolonged the survival of these animals (Fig. 8). In contrast to the untreated group in which 50% of the animals died at the first week after diabetes onset, all CO-treated animals were still alive at 6 weeks after the disease onset (P < 0.05 vs. control group).

Decreased HO-1 expression has been shown in human patients with diabetes (30,31). Studies on experimental diabetes also revealed lower levels of HO-1 expression and HO activity in streptozotocin-induced diabetic rats compared with nondiabetic controls (32,33). In the present study, we did not observe significant changes in HO-1 and HO-2 protein expression along with aging process in NOD mice. However, a significant reduction in net HO activity was noticed in old NOD mice. This finding appears to be similar to a recent report showing that hyperglycermia did not affect HO-1 and HO-2 expression but caused a decrease in HO-1 activity (33). It is conceivable that development of hyperglycermia in old NOD mice may have an impact on HO activity. Earlier studies by others have demonstrated that HO-1 overexpression can attenuate vascular endothelial damage and prevent cardiac ischemia/reperfusion injury in experimental diabetes (32,34–36). Furthermore, overexpression of HO-1 in islets provided protection from apoptosis and increased allograft survival (37–40). These findings support the multiple protective roles of HO-1 in diabetes. However, whether HO-1 has an immunoregulatory function in the course of type 1 diabetes development remains to be established. To address this issue, in the present study, we performed in vivo HO-1 gene transfer in NOD mice using AAV vector that has been shown to mediate sustained and stable transgene expression in animals. Our results demonstrated that a single intravenous injection of AAV-HO-1 in pre-diabetic NOD mice significantly attenuated the progression of destructive insulitis and the onset of diabetes. The protection conferred by AAV-HO-1 appears to be correlated with the systemic induction of HO-1 in several tissues, including the pancreas. Immunostaining of pancreatic sections revealed that HO-1 was predominantly expressed in β-cells. Because we did not specifically investigate whether pancreatic HO-1 overexpression can prevent the autoimmune destruction of β-cells in situ, this possibility cannot be completely ruled out in the present experimental setting. Nevertheless, we provide evidence showing that systemic HO-1 gene transduction resulted in suppression of the Th1 cell response. AAV-HO-1 administration significantly reduced the levels of the Th1-type cytokines IL-2 and IFN-γ in circulation and in activated splenocytes, without affecting the levels of the Th2-type cytokines IL-4 and IL-10. An influence of HO-1 on the Th1/Th2 cytokine profile has also been reported in a study of liver allografts after HO-1 gene transfer (41). A recent study on HO-1–deficient mice revealed a profound Th1 response after T-cell stimulation (23). The preferential suppression of Th1 effector cell function by HO-1 suggests that HO-1 has a beneficial effect in various diseases involving the Th1-mediated immune response. Because destructive insulitis is closely associated with Th1 cell activation and Th1-type cytokine production, it is apparent that HO-1 suppresses type 1 diabetes in NOD mice by influencing the diabetogenicity of T-cells. Adoptive transfer of splenocytes from AAV-HO-1–treated mice demonstrated late onset of diabetes in NOD.scid recipients, confirming the immunosuppressive effect of HO-1 on autoreactive T-cells.

Over the past few years, considerable evidence has supported the vital role of CD4⁺CD25⁺ Treg cells in maintaining peripheral tolerance (11–13). A decline in the CD4⁺CD25⁺ Treg cell population was noted in NOD mice (14,15). Transfer of polyclonal CD4⁺CD25⁺ Treg cells prevented diabetes in NOD mice (42). Furthermore, systemic overexpresssion of IL-10 in NOD mice increased the CD4⁺CD25⁺ Treg population and ameliorated diabetes (43). Recently, HO-1 has been implicated in the suppressive capacity of CD4⁺CD25⁺ Treg cells (44). To examine whether HO-1–mediated type 1 diabetes protection in NOD mice involves regulation of CD4⁺CD25⁺ Treg cells, we analyzed the percentage of CD4⁺CD25⁺ Treg cells in splenocytes. Our results showed that the AAV-HO-1–treated group did not show a higher proportion of CD4⁺CD25⁺ Treg cells in splenocytes, indicating that HO-1–induced suppression of diabetogenic T-cells is not achieved by elevating the Treg population. Because the present study did not assess the suppressive activity of CD4⁺CD25⁺ Treg cells, it is unclear whether HO-1 affected CD4⁺CD25⁺ Treg function in our experimental setting. Further study is required to clarify this issue. Along with Treg cells, dendritic cells play an integral role in controlling autoimmunity and self tolerance. Recently, it has been demonstrated that immature dendritic cells induce peripheral tolerance by either increasing Treg cells or inducing T-cell anergy (45). Dendritic cells in NOD mice have abnormally high immunostimulatory and Th1-inducing abilities (8–10). Downregulating the phenotypic maturation of dendritic cells suppresses the immune response and induces peripheral tolerance in NOD mice (46). A recent study showed that HO-1 expression inhibits dendritic cell maturation and proinflammatory and allogeneic immune responses while preserving IL-10 production (26). It is of great interest to determine whether dendritic cell maturation is affected by HO-1 in NOD mice. Flow cytometry analysis revealed that the total number of CD11c⁺ splenocytes was not altered by AAV-HO-1 administration in NOD mice. However, the percentage of CD11c⁺ cells expressing I-A^g7 MHC II was reduced by 35% in AAV-HO-1–treated mice compared with the control group. This finding supports the idea that HO-1 overexpression in NOD mice suppressed the functional maturation of dendritic cells, which in turn limited Th1 induction and disease progression.

HO-1 is an endoplasmic reticulum–anchored protein, and its protective functions are primarily mediated through its reaction by-products (16,17). Recently, studies have shown that both CO and bilirubin can prevent endothelial cell sloughing and prolong the survival of allogeneic islets in type 1 diabetes, likely through different mechanisms (33,47,48). In the present study, we specifically assessed the potential role of CO in HO-1–mediated type 1 diabetes protection in NOD mice. Animals were subjected to a 2-h exposure of low-dose CO twice per week for a period of 15 weeks. We admit that this treatment may not exactly resemble the same situation in AAV-HO-1–treated mice in which CO was constantly produced by HO-1 in situ. Nevertheless, it was interesting to note that the pre-diabetic mice subjected to the periodic CO treatment had a significantly lower incidence of diabetes compared with the untreated control mice. Similar to what we observed in the AAV-HO-1–treated mice, Th1 cytokines produced by activated splenocytes were markedly reduced in the CO-treated group. Previous studies on various disease settings in animals have revealed multiple cellular mechanisms underlying the protections conferred by CO inhalation. For example, CO has been shown to activate heme-containing soluble guanylyl cyclase to increase cellular cGMP level or modulate p38 kinase-mediated signaling transduction (17). Recently, there are reports showing that CO treatment can protect animals against liver failure and chronic colitis through induction of HO-1 expression (49,50). We did not observe changes in the level of HO-1 protein expression in NOD mice after CO treatment, suggesting that the beneficial effect of CO observed in present experimental setting is unlikely mediated through upregulating HO-1 expression. Nevertheless, to completely rule out the implication of HO-1, additional experiments with the condition of HO-1 activity inhibition or HO-1 deficiency in NOD mice will be needed. The detailed mechanism responsible for the protective effect of exogenous CO treatment on type 1 diabetes remains to be determined. Further experiment demonstrated that the periodic CO treatment also prolonged the survival of NOD mice with chemical-induced overt diabetes. Together, these results suggest a possible role for CO in the protective function of HO-1 in NOD mice.

In conclusion, the present study provides convincing evidence to support an immunosuppressive function of HO-1 during the course of type 1 diabetes. The beneficial effect of HO-1 is most likely through downregulating dendritic cell maturation, thereby limiting T-cell activation and Th1 response. These findings suggest a new therapeutic approach for treating type 1 diabetes by manipulating the expression level of HO-1 with pharmacological agents or gene-based therapy.

L.-Y.C. has received the Academia Sinica Investigatorship. This work has received support from the Institute of Biomedical Sciences, Academia Sinica.