central nervous system
inducible nitric oxide synthase
multiple low doses
IL-23, a proximal regulator of IL-17, may be a major driving force in the induction of autoimmune inflammation. We have used a model of subdiabetogenic treatment with multiple low doses of streptozotocin (MLD-STZ; 4 × 40 mg/kg body weight) in male C57BL/6 mice to study the effect of IL-23 on immune-mediated β cell damage and the development of diabetes, as evaluated by blood glucose, quantitative histology, immunohistochemistry and expression of relevant cytokines in the islets. Ten daily injections of 400 ng IL-23, starting on the first day of MLD-STZ administration led to significant and sustained hyperglycemia along with weight loss compared with controls (no IL-23), and a significant increase in the number of infiltrating cells, a lower insulin content, enhanced apoptosis, expression of IFN-γ and IL-17 (not seen in the controls) and a significant increase in the expression of TNF-α and IL-18 in the pancreatic islets. IL-23 treatment started 5 days prior to MLD-STZ administration had no effect on diabetogenesis or cytokines expression in the pancreatic islets. We provide the first evidence in an animal model that IL-23 is involved in the development of type-1 diabetes, by inducing IL-17 and possibly IFN-γ production in the target tissue.
Streptozotocin (STZ), a D-glucopyranose derivative of N-methyl–N-nitrosourea endowed with potent alkylating properties 1, acts also as a producer of reactive oxygen species 2. While at high doses the diabetogenic potential of STZ is explained by its capacity to selectively promote insulin-producing β cell death by apoptosis or necrosis, at low doses, STZ generates H2O23 and induces expression of glutamic acid decarboxylase autoantigens 4. Glutamic acid decarboxylase is a strong trigger of β cell-specific autoimmunity, both in humans and experimental models of diabetes 5–7. It has also been demonstrated that multiple low dose (MLD)-STZ-induced diabetes in susceptible strains of mice requires a Th-1 cell-dependent inflammatory reaction 6, 7. We 8–10 and others 11–13 have used this model to study the role of proinflammatory cytokines in the development of type-1 diabetes. The model offers an opportunity to study the effects of cytokines on synchronized group of diabetic animals.
The IL-12 family of cytokines, which includes IL-12, IL-23 and IL-27 (reviewed in 14), are required for Th-1 cell differentiation. This family of cytokines is produced by antigen-presenting cells such as macrophages, dendritic cells (reviewed in 15) and microglia in the central nervous system (CNS) 16. While both macrophages and a small number of dendritic cells are present in intact adult pancreatic islets 17, blood-borne macrophages are the first cells to infiltrate the islets after MLD-STZ diabetes induction 18, 19. Dendritic cells are also known to either enhance 20 or prevent 21 autoimmunity leading to diabetes, depending on their activation status.
IL-12 induces IFN-γ production, a process which appears to down-regulate rather than stimulate T cell-dependent autoimmunity, at least in the CNS 22. In contrast, IL-23 and IL-27 play pivotal roles in the establishment and maintenance of organ-specific autoimmunity as demonstrated in experimental allergic encephalomyelitis (EAE) 23 and collagen-induced arthritis 24. Increased expression of IL-23 has also been demonstrated in skin lesions of patients with psoriasis vulgaris 25.
Whereas the pro-diabetic effect of IL-12 produced by macrophages and dendritic cells in situ, appears to be controversial, the role of the other two members of the IL-12 family, IL-23 and IL-27, has not been studied. In humans, polymorphism of the IL-12B gene, which encodes the p40 subunits of IL-23, is associated with late onset of type-1 diabetes 26. Here we provide the first evidence that in vivo administration of IL-23 induces the onset of diabetes in mice administered subdiabetogenic doses of MLD-STZ. This effect of IL-23 was associated with expression of IFN- γ and IL-17 in the pancreatic islets and an enhancement of TNF-α and IL-18 expression early after diabetes induction. These effects were accompanied also by enhanced mononuclear cellular infiltration and β cell loss in the islets.
The MLD-STZ model
To study the role of IL-23 in diabetes induction, we used a subdiabetogenic regimen of MLD-STZ treatment. It has been established that five daily doses of 40 mg/kg/day are required for delayed onset, sustained and progressive hyperglycemia and insulitis in male C57BL/6 mice 7, 8. As shown in Fig. 1, four injections of STZ did not induce significant hyperglycemia and only mild single-cell insulitis (Fig. 2a vs. b). We proceeded to determine whether IL-23 would enhance the diabetogenic process by repeated administration of IL-23 with control animals receiving the carrier solution only. In accordance with a previous report 3 and our preliminary data, female C57BL/6 mice were resistant to the MLD-STZ regimen (data not shown), and therefore the modulatory effects IL-23 were studied in male mice only.
IL-23 promotes development of MLD-STZ diabetes in male C57BL/6 mice
In the first series of experiments we administered ten daily i.p. injections of 400 ng IL-23 starting on the first day of a 4-day regimen of MLD-STZ. The control animals likewise received 4 × 40 mg/kg STZ but only the carrier solution (without IL-23), i.e., for 10 days. IL-23 significantly enhanced diabetogenesis. At the end of the experiment, four out of seven control animals were mildly diabetic, while all animals treated with IL-23 had significant hyperglycemia (Fig. 1a); venous glucose levels were significantly higher in IL-23-treated mice than controls after day 16 of MLD-STZ treatment. This effect was associated with weight loss in IL-23-treated animals compared to controls (Fig. 1b). Similar results were obtained in the two additional experiments (data not shown).
IL-23 greatly enhances influx of mononuclear cells
Diabetogenesis after MLD-STZ treatment is mediated by an early influx of macrophages into the pancreatic islets, 18, 19, efficient presentation of released diabetogenic antigens and recruitment of T cells. By day 23 after diabetes induction, IL-23 stimulated this process, leading to peri-insulitis and insulitis (Fig. 2b vs. c) in comparison with single-cell infiltrates of islets in control animals treated with STZ only. Quantitative analysis of islets of similar size (370.45 ± 22.6 µm) confirmed significant enhancement in the influx of inflammatory cells in IL-23-treated animals vs. controls (Table 1).
|Treatment||STZ only (controls)||STZ + IL-23|
|Number of infiltrating mononuclear cells||24 ± 4.0||80 ± 5*|
|Percent Insulin positivity||52.7 ± 1.4||27.2 ± 3.1**|
|Number of apoptotic cells||25 ± 6||50 ± 12**|
IL-23 induces β cell loss
As shown in Table 1 and illustrated in Fig. 2e vs. f, insulin content was significantly lower in IL-23-treated animals than controls treated with MLD-STZ only. The number of apoptotic cells was twofold greater in MLD-STZ + IL-23-treated mice compared with controls (p<0.05) (Fig. 2 g vs. h; Table 1).
Effect of IL-23 on the expression of proinflammatory cytokines in islets and spleen
TNF, IFN-γ and IL-18 are all reported to be involved in the pathogenesis of type-1 diabetes. More recently, IL-17 has been considered to be an important effector cytokine in Th1 cell-mediated immunopathology in the CNS 27. Therefore, we analyzed by RT-PCR, expression of these cytokines in isolated islet cells and mononuclear spleen cells.
As shown in Fig. 3, IL-23 had profound effects on the expression of these cytokines in the target tissue and to a much lesser extent in the spleen. Diabetogenesis stimulated by IL-23 was associated with enhanced expression of TNF-α in the islets. Additionally, both IFN-γ and IL-17 were expressed in the islets of animals treated with IL-23 but not in MLD-STZ controls. Interestingly, IL-18 was expressed to a lesser extent in control islets, suggesting that its involvement in diabetes induction is upstream of IL-23 and IFN-γ.
IL-23 induces inducible nitric oxide synthase expression in islet-infiltrating cells
We have shown previously that nitric oxide (NO) is the major pathogenic factor in MLD-STZ-induced diabetes 28, 29. This finding has been extended to other experimental models such as NOD mice 30, 31 and BB rats 32. It appears that NO production is important in IFN-γ-mediated but not in IFN-γ-independent β cell death 33. Results presented in Fig. 4 indicate that IL-23 induces the expression of inducible NO synthase (iNOS) in islet isolates (Fig. 4a line 7), but not in the spleen (Fig. 4a line 5). Subdiabetogenic MLD-STZ treatment alone did not induce iNOS expression in either cell population (Fig. 4b lines 5 and 7).
Effects of varying the duration and the timing of IL-23 treatment
Since IL-23 may be involved in the early differentiation of T cells 15, 16, we attempted to induce MLD-STZ diabetes in an environment where the cytokine was already present. Five to ten daily administrations of IL-23 were initiated 5 days prior to starting MLD-STZ injections. There was no diabetogenic effect of this treatment (data not shown), indicating that IL-23 was required only in the later phase, after the diabetogenic antigen-driven pathological process had already been initiated in the target tissue. Similarly, expression of effector cytokines IFN-γ and IL-17 was not induced in the islets by such early administration of IL-23, while stimulation of IL-18 and TNF-α expression was seen in the spleen and to a much lesser extent in the islets (Fig. 5).
Previous study has suggested that IL-23 may play an important role in the establishment and maintenance of autoimmunity in the CNS 23. It has also been demonstrated that the IL-23 receptor complex is not detectable in resting naïve T cells, but is rapidly induced after polyclonal activation 34. Indeed, it has been argued that, whereas IL-23 does not promote the development of IFN-γ−producing Th1 cells, it is an essential factor for the development of a pathogenic CD4+ T cell population, which is characterized by the production of IL-17, IL-6 and TNF 35.
Here we provide evidence that prolonged administration of IL-23 after subdiabetogenic treatment with MLD-STZ induced hyperglycemia and weight loss that was not seen in control MLD-STZ-treated animals. These findings were accompanied by immunopathological parameters of type-1 diabetes, including enhanced infiltration of mononuclear cells and apoptotic loss of insulin-producing β cells.
Whereas five daily injections of 40 mg/kg are known to induce delayed and sustained hyperglycemia and weight loss, a 4-day regimen is insufficient to induce diabetes in up to 40 days of observation. However, immunohistological analysis of the pancreata of these mice revealed weak to moderate mononuclear cell infiltration and β cell loss. This finding is compatible with the notion that protracted, effective stimulation of islet antigen-specific cells is required for the development of “clinical” disease. Our data indicate that IL-23 is able to facilitate or enhance this process of diabetogenesis. Interestingly, when the same regimen of IL-23 treatment was instituted earlier, prior and during MLD-STZ treatment, this diabetogenic effect is not seen. This observation tends to confirm a recent report that IL-23 suppressed, rather than enhanced relapsing clinical EAE in female SJL mice when given during the induction of the disease, whereas CD4+ and CD8+ T cell infiltration was not significantly affected 35. Therefore, it could be assumed that IL-23 is required in the final effector phase of autoimmune destruction of the target cells, as previously suggested 27, 36. In fact, Langrish et al. 27 demonstrated that IL-23-dependent IL-17-producing cells are highly pathogenic and essential for the establishment of organ-specific inflammation in the CNS. Our data are compatible with the notion that this new IL-23-driven subset of Th1 cells called Th1-IL-1736 are of particular importance in the induction of organ-specific autoimmunity. IL-17 expression correlated with the intensity of mononuclear cell infiltrate, and β cell loss. IL-17 can enhance secretion of TNF-α by activated macrophages 37 as indicated by our results (Fig. 1). Recent evidence suggests that IFN-γ is not required or may be protective in collagen-induced arthritis 24 as well as in EAE 23. Its role in type-1 diabetes is not completely understood. Our data (Fig. 3) indicate that the diabetogenic effect of IL-23 is accompanied by a dramatic increase in the expression of IFN-γ in the islets by day 16 after diabetes induction (Fig. 3). In mouse models of type-1 diabetes, CD4+ cells, which produce large amounts of IFN-γ and little IL-4, induce disease and tissue destruction upon adoptive transfer (reviewed in 38). Gysemans et al. 39 suggested that IFN-γ signaling transduction at the β cell level is not essential for immune β cell destruction in vivo. However, the same group has recently reported that exposure of β cells to double-stranded RNA in combination with IFN-γ significantly increased apoptosis of β cells 40, and that disruption of the IFN-γ signaling pathway at the level of signal transducer and activator of transcription-1 prevented immune destruction of β cells 41.
In the model of Bettelli and Kuchroo 36, when the T-bet-expressing precursor is activated to induce Th1 differentiation, these cells express both IL-12 receptors and IL-23 receptors on their surfaces. It can be assumed that additional autoantigen(s) released by the TNF-α-mediated β cell damage stimulates the IL-12-dependent pathway and leads to the induction of IFN-γ-producing Th1 cells. Trembleau et al. 33 have shown that NO is produced in high quantity by pancreas-infiltrating cells in NOD mice, through a mechanism involving IL-12-induced IFN-γ accompanied by enhanced iNOS expression. Conversely, in IL-12-treated IFN-γ-deficient mice, apoptosis of β cells appears to be mediated by the up-regulation of FAS ligand on Th1 cells. Since iNOS is expressed in islet cells of IL-23-treated animals, but not in controls (Fig. 4), we suggest that IFN-γ-mediated pathway in addition to IL-17 plays a role in β cell death, and that IFN-γ is pathogenic rather than protective factor in our experimental model.
It has been suggested that the proinflammatory action of IL-17 depends considerably on its ability to trigger the expression of iNOS (reviewed in 42). However, in mice and rat macrophages, IL-17 was completely unable to elicit NO production or stimulate that triggered by suboptimal doses of IFN-γ 42. Additionally, even though IL-17 synergizes with TNF to induce cartilage destruction in vitro43, and enhances macrophage-driven cartilage damage in streptococcal cell-induced arthritis 44, iNOS activation and NO production are not involved in these autoimmune processes. These findings support the notion that the overexpression of iNOS is IFN-γ dependent.
Therefore, as recently reported in host defense against Klebsiella45 both IL-12-mediated IFN-γ production and IL-23-mediated IL-17 production may contribute to the inflammatory damage seen in our model of experimental diabetes.
Materials and methods
Male C57BL/6 mice, 8–10 weeks of age, were used in the experiments. They were kept under strict pathogen-free conditions and were given rodent diet and free access to drinking water.
Mice received MLD-STZ treatment as described previously 3, 4. Briefly, STZ was dissolved in citrate buffer, pH 4.5, and injected i.p. at a dose of 40 mg/kg/day. This was carried out for 4 or 5 consecutive days. Diabetes was defined as a blood glucose level 10 mmol/l in non-fasting animals. Glucose concentrations in venous blood were measured by the glucose oxidase method using a glucometer. Body weights were recorded weekly. Animals were killed by cervical dislocation and their pancreas excised. Ethical approval was obtained from the FMHS Animal Research Ethics Committee.
Injections of recombinant IL-23 (0.5 mL, 400 ng in 0.1% BSA/saline) were administered i.p. in two regimens: (a) ten daily injections starting on the first day of MLD-STZ treatment; (b) ten daily injections starting 5 days prior to the first day of MLD-STZ. Control animals received matching i.p. injections of carrier without IL-23.
Histology and immunohistochemistry
Pancreata from animals were fixed in 4% paraformaldehyde or 10% buffered formaldehyde. The latter specimens were used for studies of apoptosis. All specimens were subsequently embedded in paraffin wax.
Insulin-containing β cells
Sections, 5–7 µm thick, of 4% paraformaldehyde-fixed pancreata were stained by a direct immunofluorescence technique. Briefly, sections were dehydrated, and after three 5-min washes in 0.1 M PBS, the slides were incubated with prediluted guinea pig anti-swine insulin (Dako Cytomation, Copenhagen) overnight at 4°C. The sections were then incubated with the link antibody comprising fluorescein isothiocyanate (FITC) bound anti-guinea IgG (Jackson Immuno Research Laboratories Inc., USA) diluted 1:100 in 0.3% Triton X in 0.1 M PBS for 1 h. The specimens were washed three times in 0.1 M PBS for 5 min and incubated with propidium iodide (PI, 1 mg/mL) for 30 min at 37°C. After a quick wash in 0.1 M PBS, the sections were coverslipped with glycerol as mountant and examined by fluorescence microscopy. As controls, sections were treated with the universal negative control (rabbit, code N1699, DakoCytomation, Copenhagen, Denmark).
The percentage of insulin immunopositivity was established by calculating the number of FITC-immunoreactive cells per total number of PI-stained nuclei in non-consecutive sections. PI-stained infiltrating mononuclear cells were identified by staining the same sections with hematoxylin and eosin, and excluded from the quantification.
Quantitative histology of infiltrating cells
After examination of insulin-immunostained sections, the coverslips were gently removed, the slides rehydrated and stained with hematoxylin and eosin. The number of infiltrating cells per islet was quantified in non-consecutive sections of hematoxylin and eosin-stained pancreata by light microscopy with a ×40 oil immersion objective. To remove discrepancies due to variations in islet size, the mean perimetric size of all islets in five non-consecutive sections from the pancreata was computed with an Axiocam digital camera attached to a Zeiss Axiophot. The mean value of 370.45 ± 22.6 µm was then used as the “standard size islet” to calculate the number of cells per islet. Any islet larger or smaller than the mean ± SEM was not used for quantification. Values obtained in groups of animals were expressed as mean ± SEM.
Cleaved caspase-3 was detected using the SignalStain cleaved caspase-3 (Asp175) immunohistochemistry detection kit (Cell Signaling Technology Inc., MA, USA). Procedures recommended by the manufacturers were used, with a few modifications. Briefly, wax sections were dehydrated and, after blocking endogenous peroxidase activity, were transferred into 0.1 M citrate buffer and boiled in a 750-W microwave to retrieve antigen. The sections were incubated with the primary antibody comprising prediluted polyclonal synthetic peptide corresponding to Asp175 in human caspase-3. The sections were washed and incubated with biotinylated secondary antibody for 1 h, tertiary antibody for another hour, and immunoreactivity detected using the NovaRed substrate. Both secondary and tertiary antibodies were supplied in the kit. Controls consisted of incubating slides with negative control solution provided in the kit under the same conditions as the experimental sections.
Isolation of islet cells
Islets were isolated according to the method of Lacy and Kostianovsky 46 modified by Lernmark 47. Briefly, each pancreas was carefully removed and placed immediately in 10 mL washing solution comprising 0.1 M PBS, 5% fetal calf serum and 10 mM Hepes. The pancreas was cut into smaller pieces, squashed between two frosted slides, and placed in a 15-mL tube containing washing buffer and centrifuged at 1100 rpm for 8 min at 4°C. The supernatant was discarded and the pellet resuspended in 12 mL washing solution and centrifuged again as above. After discarding the supernatant, the pellet was suspended in 1 mL 2 mg/mL collagenase P (Roche Diagnostic GmbH, Germany) and 3 mL washing solution, and shaken vigorously in a 37°C water bath for 10–15 min. The homogenate was diluted with 10 mL washing solution and centrifuged twice as above, with the supernatant being discarded between the steps.
The pellet was resuspended in 3.0 mL Ca2+ and Mg2+-free Hanks’ balanced salt solution containing 3 mM EGTA, 20 mg/mL BSA and 2.2 mg/mL glucose (GIBCO, USA) and 2.5 mg/mL bovine-derived pancreatic trypsin (Sigma Co, USA). The suspension was then incubated in CO2 for 20–25 min at 37°C, with intermittent shaking, and then transferred into a 15-mL centrifuge tube containing 7.0 mL RPMI 1640 culture medium and centrifuged for 10 min at 1100 rpm and 4°C. After washing three times in 0.1 M PBS containing 10 mM Hepes, the cells were suspended in 5.0 mL culture medium then employed in the studies.
Isolation of splenic cells
Pieces of splenic tissue were minced between two frosted microscope slides to dislodge splenic cells. The dislodged cells were collected after centrifugation in RPMI 1640 (GIBCO, USA) containing 5% fetal calf serum. ACK lysing buffer (0.15 M NH4Cl, 1.0 mM KHCO3, 0.1 mM Na2EDTA in distilled H20, pH 7.4) was used to lyse red blood cells.
RNA isolation and RT-PCR of cytokine mRNA
Five million cells from pooled islet or splenic tissues from three mice were lysed with SV RNA lysis buffer (Promega, USA) and stored at –70°C until all samples were collected. mRNA was isolated using the SV total RNA isolation system (Promega, USA) according to the manufacturer's instructions. cDNA synthesis and subsequent RT-PCR was performed strictly according to the manufacturer's instructions. The sequences of the specific oligonucleotide primer pairs 5′ and 3′, are as follows: for IFN-γ: TTACTGCCACGGCACAGTCA and CGAATCAGCAGCGACTCCTT; TNF: TCGAGTGACAAGCCTGTAGC and TGAGACAGAGGCAACCTGAC; IL-17: ACCTCACACGAGGCACAAGT and CTTCATTGCGGTGGATC; IL-18: GCTGCCATGTCAGAAGACTC and GGTCACAGCCAGTCCTCTTA; and iNOS: ACAGCCTCAGAGTCCTTCAT and TTGTCACCACCAGCAGTAGT. The sequences of the specific oligonucleotide primer pairs 5′ and 3′ for the positive control (RNA from the Kanamycin resistance gene supplied by the manufacturer) are: GCCATTCTCACCGGATTCAGTCGTC and AGCCGCCGTCCCGTCAGATCAG.
Two-tail Student's t-test (glycemia) and ANOVA (quantitative histology) were applied; p<0.05 was considered to be statistically significant.
We thank Mrs. G. K. Dawood for secretarial help, Ms. Katija Parekh for help with PCR and Professors F. Y. Liew and M. G. Nicholls for review of the manuscript. This work was supported by an interdisciplinary grant (01–10-8-12/04) to M.L.L. from the UAE University Research Sector and in part by FMHS grants.