Islet Expression of M3 Uncovers a Key Role for Chemokines in the Development and Recruitment of Diabetogenic Cells in NOD Mice

NOD and NOD.SCID mice were obtained from The Jackson Laboratories (Bar Harbor, ME). Rat insulin promoter (RIP)-M3 mice, described previously (17), were backcrossed from a mixed B6D2 (C57Bl/6 × DBA/2) background onto the NOD background for 11 generations and onto B6 background for >11 generations. BDC2.5 NOD mice were provided by Dr. Ralph Steinman (The Rockefeller University, New York). In all experiments, transgenic mice were compared with their corresponding littermates. All mice were housed under specific pathogen-free conditions in individually ventilated cages at the Mount Sinai School of Medicine Animal Facility. All experiments were performed following institutional guidelines.

The blood glucose was monitored weekly using a one-touch blood Ascensia Elite XL glucometer (Bayer, Elkhart, IN). Animals were considered diabetic when their blood glucose levels were >250 mg/dl in two consecutive daily measurements.

For immunohistochemical staining, slides were incubated for 1 h at room temperature with purified primary antibodies followed by incubation with the appropriate labeled secondary antibodies for 30 min. Primary antibodies used were anti-CD45, CD3, CD31, CD11c, B220, CD11b from BD Biosciences Pharmigen (San Diego, CA); F4/80 from Serotec; anti-CCL1, anti-CCL21, anti-CCL2, and rat anti-insulin from R&D Systems (Minneapolis, MN); rabbit anti-CXCL9 (provided by J. Farber [National Institutes of Health, Rockville, MD]); anti-CCL3, -CCL4, -CCL11, -CCL22, and -CXCL10 (provided by S. Kunkel); and guinea pig polyclonal anti-insulin from DAKOCytomation (Carpinteria, CA). Secondary antibodies used were Alexa Fluor 488–and Alexa Fluor 594–goat anti-rat IgG from Molecular Probes (Eugene, OR) and Cy3-goat anti-Armenian hamster, fluorescein isothiocyanate-donkey anti–guinea pig IgG, and Cy5-goat anti-rat from Jackson ImmunoResearch (West Grove, PA).

To determine the degree of islet infiltration in NOD and NOD-M3 female mice and in NOD.SCID mice in transfer experiments, groups of mice were analyzed at 10 or 30 weeks of age or 9 weeks after transfer, respectively, assessing at least 80–100 islets per animal. Insulitis was scored as follows: grade 0, no lesions; grade 1, peri-insular leukocytic aggregates, usually periductal infiltrates; grade 2, <25% islet destruction; and grade 3, >25% islet destruction. An insulitis score for each mouse was obtained by dividing the total score for each pancreas by the number of islets examined. Data are presented as mean insulitis score ± SD for the indicated experimental groups.

Islets of Langerhans were isolated as previously described (18). Briefly, the common bile duct was clamped distal to the pancreatic duct junction at its hepatic insertion. The proximal common bile duct was then cannulated using a 27-gauge needle, and the pancreas was infused by retrograde injection of 2 ml ice-cold collagenase-V solution (1 mg/ml; Sigma, St. Louis, MO) in Hanks’ balanced salt solution (HBSS). Pancreatic tissue was recovered and subjected to a 15-min digestion at 37°C. Ice-cold HBSS was added, and the suspension was vortexed at full speed for 10 s. Islets were hand-picked under a dissecting microscope.

Isolated islets or lymph nodes were incubated in 5 mg/ml collagenase D (Roche Applied Science, Indianapolis, IN) at 37°C for 45 min. Samples were strained through a 70-μm diameter nylon mesh to obtain a single-cell suspension, centrifuged, and washed in PBS. All cell suspensions were resuspended in FACS staining buffer (PBS containing 2% fetal calf serum and 0.01% sodium azide). Islet suspension cells were incubated for 20 min at 4°C with 5 μg/ml Fc block (BD PharMingen, San Diego, CA) and then stained with directly conjugated primary monoclonal antibodies (BD PharMingen). Samples were analyzed in a FACSCanto instrument (Becton Dickinson, San Jose, CA). Data were analyzed using the FlowJo software (Tree Star).

Splenocytes (20 × 10⁶ cells) obtained from newly diabetic NOD female mice were injected intravenously into 5- to 6-week-old NOD.SCID and NOD-M3.SCID mice. In other experiments, splenocytes (20 × 10⁶ cells) obtained from diabetic NOD and nondiabetic NOD-M3 female mice were injected intravenously into 5- to 6-week-old NOD.SCID mice. BDC2.5 (6–11 weeks of age) splenocytes (5 × 10⁷) were incubated with 10 mmol/l carboxyfluoroscein succinimidyl ester (CFSE) for 10 min at 37°C. The reaction was quenched with 10 ml cold PBS and followed by washing twice in PBS. The 2.5 × 10⁷ CFSE-labeled total splenocytes in 100 μl sterile PBS were injected intravenously per recipient (4- to 6-week-old NOD and NRM3 mice). At day 4 after transfer, peripancreatic and cervical lymph nodes were removed, and a single cell suspension was done per each lymph node and each recipient mouse. For monitoring the extent of cell death in the adoptively transferred cells, lymph node cells from the recipients were also stained for PI, CD3, CD4, and Vβ4 antibodies (BD Biosciences Pharmigen).

One-way ANOVA and unpaired t-test were used to determine statistical significance. Differences were considered significant when P < 0.05.

In NOD mice, infiltration of mononuclear cells into pancreatic islets begins as early as 4–6 weeks of age, accumulating in the periphery (peri-insulitis) of a few pancreatic islets (19). The autoimmune destruction of the islets is extensive by 10 weeks of age, 8–12 weeks before the onset of overt diabetes (20,21). To analyze the expression of chemokines in islets of NOD female mice, we performed immunostaining using antibodies against CXCL9, CXCL10, CCL1, CCL3, CCL4, CCL22, and CCL24; insulin; CD31; and CD45. By 4–6 weeks of age (n = 5), the islets of Langerhans had a small number of CD45⁺ cells that localized to the periphery of the islets (Fig. 1A and B). All chemokines studied were expressed in islets with peri-insulitis. CCL1, CCL3, CCL4, CCL22, CXCL9, and CXCL10 were primarily expressed by infiltrating cells (Fig. 1C, D, E, G, I, and J, respectively). CXCL10 was also expressed by β-cells (Fig. 1J), in agreement with a previous study (22). CCL24 was expressed by infiltrating cells (Fig. 1H) and by CD31⁺ endothelial cells in the islets (Supplemental Fig. 1, which is detailed in the online appendix [available at http://dx.doi.org/10.2337/db07-1309]). As described previously (12), CCL21 was expressed in paraductal areas, and its expression was increased in areas rich in inflammatory cells (Fig. 1F). Analysis of pancreata of 10-week-old mice (n = 5) showed increased in the expression of all chemokines studied compared with the earlier time points (data not shown). At this point, large infiltrates occupying the whole islet were observed. Some of these aggregates had T- and B-cells (Fig. 1K) and peripheral lymph node addressin (PNAd)–positive high endothelial vessels that coexpressed CXCL9 (Fig. 1L). Taken together, these results show that numerous chemokines are expressed in NOD islets and suggest that the expression of several chemokines may be necessary to drive the recruitment of inflammatory cells required for the development of diabetes.

To examine whether chemokine blockade would affect diabetes development in NOD mice, we backcrossed mice expressing M3, a viral chemokine binding protein, in β-cells (17) onto the NOD background. N11 female NOD-M3 and their NOD littermates were monitored weekly for the development of hyperglycemia. By 45 weeks of age, >80% of the NOD female mice (n = 61) were diabetic (Fig. 2A). Remarkably, expression of M3 in islets of NOD mice (NOD-M3 littermates, n = 66) completely abrogated the development of diabetes (Fig. 2A, P < 0.0001). Furthermore, NOD-M3 mice did not develop disease over the next 15 weeks (n = 16, Fig. 2A).

To investigate the effect of M3 expression by β-cells on insulitis, we examined pancreata from NOD and NOD-M3 female mice at 10 weeks of age (n = 5/group). Semiquantitative analysis of islet infiltrates showed that, as expected, most (95%) islets from NOD nondiabetic mice had peri-insulitis and in some cases developed a destructive inflammatory infiltrate (Fig. 2B and D), with marked loss of β-cell mass. In contrast, islets from NOD-M3 mice were virtually devoid of infiltrating cells and showed a normal complement of insulin-producing cells (Fig. 2C and D, P = 0.004). None of the NOD-M3 females that were more than 60 weeks of age (n = 13) showed insulitis. Only 12 ± 10% of islets had peri-insulitis, and in some cases, inflammatory cells were present in the exocrine tissue (Supplemental Fig. 2).

Lymphocytes are often observed in the salivary and lacrimal glands of NOD mice (23). We analyzed the presence of infiltrates in salivary and lacrimal glands from NOD and NOD-M3 mice at 10 weeks of age and onward. CD45⁺ infiltrates were found in salivary glands of both NOD (20 of 20) and NOD-M3 (22 of 22) mice (Fig. 2E and F, respectively). Small infiltrates were also found in lacrimal glands of NOD (3 of 5) and NOD-M3 (2 of 5) mice. Overall, these findings indicate that β-cell expression of M3 prevented islet mononuclear infiltration and diabetes development but did not prevent cellular infiltration into other organs.

Because the absence of mononuclear cells in the pancreas of NOD-M3 mice could be due to a blockade of tissue accumulation of diabetogenic cells, we tested whether M3 blocked entry of diabetogenic cells into the islets. To generate mice deficient in mature T- and B-cells that also expressed M3 in islets, we crossed NOD-M3 mice to NOD.SCID mice (referred as NOD-M3.SCID mice). When splenocytes from newly diabetic NOD females were transferred into young NOD.SCID mice, all NOD.SCID mice (9 of 9) developed diabetes within 7 weeks of transfer (Fig. 3A). Although autoreactive T-cells within NOD splenocytes could also induce diabetes in all NOD-M3.SCID recipient mice (11 of 11) (Fig. 3A), the development of diabetes was significantly delayed (P = 0.0086). To test whether the delay in diabetes development was caused by a reduced influx of the cells into the islets, we examined pancreata of the recipients 2 weeks after transfer. NOD.SCID recipients had insulitis and peri-insulitis in 75% islets (Fig. 3B and D). In contrast, NOD-M3.SCID mice (n = 5) had mild infiltrates of leukocytes in ∼25% of the islets and a normal complement of insulin-producing cells (Fig. 3C and D). However, at later time points, infiltrates in islets from NOD-M3.SCID mice were comparable with those observed in NOD.SCID recipients (data not shown). These results indicate that M3 expression delayed the insulitis and islet destruction induced by diabetogenic splenocytes derived from NOD mice.

Because M3 expression in the islets delayed but did not completely prevent diabetes after transfer of diabetogenic cells, we considered the possibility that M3 expression also inhibited development of diabetogenic cells in NOD-M3 mice. To determine whether functional diabetogenic splenocytes could be detected in NOD-M3 mice, we harvested spleen cells from NOD and NOD-M3 mice and transferred them into young NOD.SCID female mice. After transfer of NOD splenocytes, all NOD.SCID recipients (n = 7) became diabetic in 8 weeks. No diabetes was observed in the recipients (n = 19) up to 16 weeks after transfer of splenocytes from NOD-M3 mice (Fig. 3E). Histological analyses of the pancreata from the recipients showed that 100% the islets of the animals transferred with splenocytes from NOD mice had massive cellular infiltration and few remaining insulin-producing cells (Fig. 3F and H). Animals that received NOD-M3 splenocytes had diffuse, low-grade infiltrates of leukocytes in the islets and a normal complement of insulin-producing cells (Fig. 3G and H). Interestingly, most of the NOD.SCID mice that received NOD-M3 splenocytes had infiltration in the islets but only in 60% of them. Furthermore, these infiltrates were less severe than those observed in recipients of NOD cells (Fig. 3H). Lymphocytic infiltrates consisting mainly of CD3⁺ T-cells and B220⁺ B-cells were present in the salivary glands of NOD.SCID mice transplanted with both splenocytes from NOD and NOD-M3 mice (Supplemental Fig. 3A and B), suggesting that the presence of M3 blocked development of diabetogenic cells but did not affect autoimmune disease to salivary glands. Taken together, these results indicate that islet-specific expression of M3 delays insulitis induced by adoptively transferred NOD splenocytes and impairs the development of diabetogenic cells.

The initiating events in diabetes are not fully understood, but β-cell stress and death during early islet restructuring are thought to provide autoantigens, which induce the formation of diabetogenic cells in peripancreatic lymph nodes (PPLNs) (24,25). Therefore, we asked whether expansion of islet-specific CD4⁺ T-cells was affected in NOD-M3 mice. To this end, we took advantage of BDC2.5 NOD T-cell receptor (TCR)-transgenic mice, which carry the rearranged TCR-α (Vα1) and -β (Vβ4) chain genes from a diabetogenic, β-cell–specific, CD4⁺ T-cell clone isolated from a diabetic NOD mouse (26). Adoptively transferred CFSE-labeled naive BDC2.5 splenocytes were analyzed for their proliferation in PPLNs and irrelevant cervical lymph nodes on the basis of CFSE dilution on day 4 after transfer into NOD and NOD-M3 recipients. BDC2.5 cells showed significantly lower levels of proliferation in PPLNs of NOD-M3 recipients (13 ± 6% n = 12; Fig. 4A, bottom right panel), whereas NOD recipients displayed a significantly higher (P < 0.0001) percentage of proliferating BDC2.5 T-cells (35 ± 11% n = 8; Fig. 4A, bottom left panel). BDC2.5 cells did not proliferate in cervical lymph nodes from NOD and NOD-M3 recipients (Fig. 4A, top panel).

Dendritic cells are believed to be essential in the initial activation of islet-specific T-cells in PPLNs (27). Because islet-specific BDC2.5 cells CD4⁺ T-cells proliferated less when transferred into NOD-M3 mice, it could be argued that the observed differences could be due to alterations in the number of dendritic cells in PPLNs. To test this hypothesis, we compared the composition of dendritic cell subsets in PPLNs from 30- to 40-day-old NOD (n = 7) and NOD-M3 (n = 8) mice. To investigate the different dendritic cell subsets, single cell suspensions were stained with anti-CD11c and CD11b antibodies and analyzed by flow cytometry. We observed that the relative number of CD11c⁺CD11b⁺ cells in PPLNs from NOD-M3 mice was reduced compared with that in NOD mice (0.48 ± 0.08% in NOD mice vs. 0.3 ± 0.02% in NOD-M3 mice, P = 0.02; Fig. 4B). This difference was more evident when we analyzed the absolute number of CD11c⁺CD11b⁺ dendritic cells (4.5 ± 1.5 × 10³ cells in NOD mice vs. 2.1 ± 0.6 × 10³ cells in NOD-M3 mice, P = 0.0011; Fig. 4C). The relative and absolute numbers of the other two populations (CD11c⁺CD11b⁻ and CD11c⁻CD11b⁺ cells) were not affected (Fig. 4B and C). There were no differences among these cell subsets in cervical lymph nodes (Supplemental Fig. 4A and B). Because dendritic cells capture antigen in tissues and then migrate into draining lymph nodes to present their antigens to T-cells, we asked whether the presence of M3 in the islets would alter the composition of dendritic cells present in the islets. We isolated islets from NOD and NOD-M3 mice and stained single cell suspensions with different surface markers. As shown in Fig. 4D–F, we observed a significant reduction in the relative number of CD11c⁺CD11b⁺ cells in islets of NOD-M3 mice (22.9 ± 3.4% in NOD mice vs. 11.1 ± 1.5% in NOD-M3 mice; P = 0.0054 Fig. 4E) and in the absolute number of this dendritic cell subset (Fig. 4F). NOD-M3 249 ± 47 CD11c⁺CD11b⁺ cells/100 islets, whereas the number of CD11c⁺CD11b⁺ cells from islets of NOD mice was increased fourfold (965 ± 85 cells/100 islets) (P < 0.0001; Fig. 4F). Interestingly, there was no difference in the number of CD11c⁺CD11b⁺ cells in the blood of NOD-M3 mice when we compared with blood of NOD mice (Supplemental Fig. 4C). To investigate whether this difference was driven solely by the expression of M3, we measured the number of CD11c⁺CD11b⁺ cells in islets of mice that are not prone to autoimmunity. To this end, we isolated islets from B6 and 6RIP-M3 mice (RIP-M3 mice (17) crossed over 10 times into the B6 background). We found that the expression of M3 in a nonautoimmune background does not alter the relative or the absolute numbers (Fig. 5A and B) of any dendritic cell subset analyzed. Taken together, these results indicate that M3 expression by β-cells in NOD mice decreases inflammation in the islets, impairs the migration of a subset of dendritic cells into the islets and PPLNs, and reduces the proliferation of diabetogenic cells in the draining lymph nodes.

In this study, we examined the role of the chemokine system in an experimental model of diabetes. We show that chemokine expression precedes development of diabetes and that islet-specific blockade of chemokine function reduces the local inflammation that results in a defective development and pancreatic accumulation of diabetogenic cells. These results thus constitute formal demonstration that chemokines are critical determinants of autoimmune diabetes.

A pattern of multiple chemokine expression is found in the islets of NOD mice before disease. Our results confirm and expand results obtained by other groups (12,22,28). Prediabetic NOD islets expressed inflammatory (CCL1, CCL3, CCL4, CCL22, CCL24, CXCL9, and CXCL10) as well as homeostatic chemokines (CCL21). The expression of all chemokines was closely associated, but not limited, to the inflammatory cells. For instance, CCL24 was expressed in vascular endothelial cells and CXCL9, previously reported in lining capillaries in the islets (22), was also expressed by PNAd⁺ vessels within the islets. This finding is of interest because CXCL9 expression has been detected in high endothelial vessels of inflamed lymph nodes (29) but not in pancreatic islets. The combined expression of CXCL9 and CCL21 by endothelial cells may be important for recruitment of CXCR3⁺ effector T-cells and for recruitment of dendritic cells and memory T-cells expressing CCR7. Increased expression of CCL21 may also contribute to the development of segregated T- and B-cell aggregates reported within prediabetic NOD islets and to the recruitment of CCR7⁺ monocytes and dendritic cells (12,30).

To date most attempts to interfere with the chemokine pathways in diabetes have relied on the use of neutralizing antibodies to chemokines or chemokine receptors. Neutralization of CXCL10, partially prevents autoimmune disease in the (RIP)-LCMV mouse model (31) and in NOD mice after cyclophosphamide administration (22). CXCL10 DNA vaccination of NOD mice also partially prevents diabetes (32). Neutralization of CCL22 with an antibody causes a significant reduction of CCR4⁺ T-cells within the pancreatic infiltrates and partially inhibits the development of diabetes in adoptive transfer model in NOD-SCID (14). Furthermore, treatment of NOD mice with a neutralizing anti-CCR5 antibody reduces incidence of diabetes (33). In all of these studies, the individual treatments reduce inflammatory infiltrates, but do not completely block diabetes development, suggesting additional requirements.

To test the hypothesis that diabetes is caused by the combined activities of several chemokines, we used M3, a chemokine binding protein encoded by the mouse herpesvirus 68 (MHV68) that binds several murine and human chemokines with high affinity (34,35). M3 interacts with the N-loop of chemokines and mimics elements of G protein-coupled receptor recognition to block interaction with cognate receptors by a competitive inhibition mechanism (36,37). In vivo, M3 reduces mononuclear cellular responses after MHV68-induced meningitis in mice (38). Systemic expression of M3 reduces intimal hyperplasia subsequent to femoral artery injury (39) and aortic allograft transplantation (40). Finally and perhaps more important in the context of these studies, expression of M3 in transgenic islets inhibits recruitment of leukocytes induced by islet-specific expression of CCL21 (17), CXCL13, and CCL2 (41) and prevents insulitis and diabetes induced by multiple low doses of streptozotocin (18). Here, we show that the scavenging properties of M3 are essential for the reduced inflammation observed in the NOD-M3 transgenic mice. Many of the chemokines upregulated in the NOD model, such as CCL2, CCL3, CCL4 CCL5, and CCL21, bind to M3 with high affinity (34,35).

We suggest that M3 blocks pancreatic autoimmunity by disrupting both the influx and generation of diabetogenic cells. Our results indicate that local M3 expression blocks influx of diabetogenic cells. Adoptive transfer of NOD splenocytes into NOD.SCID and NOD-M3.SCID recipients causes diabetes. However, the course of the disease is delayed in the NOD-M3.SCID mice, suggesting a partial blockade in the recruitment of diabetogenic cells. The failure of M3 to prevent development of diabetes in this context could be due to either the amount or kind of chemokines produced relative to those present in NODM3 islets during development. Experiments addressing these hypotheses are currently underway. Finally, the failure of splenocytes from NOD-M3 mice to transfer disease suggests that either β-cell–specific T-cells are not formed in the NOD-M3 animals or that such cells may be very reduced in number (or poorly activated) and thus incapable of inducing disease.

The activation of naive T-cells requires not only TCR stimulation but also the simultaneous delivery of a costimulatory signal by a specialized antigen-presenting cells in draining lymph nodes. Activation of transgenic, islet-specific CD4⁺ T-cells (BDC2.5 cells) and diabetogenic CD8⁺ T-cells occurs in the pancreatic lymph nodes at 3–4 weeks of age (24,27). Interference with priming and diabetes occurs if PPLNs are ablated (42,43) or if important cellular constituents are ablated (44). In this context, dendritic cells are believed to be essential in the initial activation of islet-specific T-cells in PPLNs (27). Both CD11c⁺CD11b⁺ and CD11c⁺CD11b⁻ (or CD11c⁺CD8⁺) dendritic cells have been shown to be responsible for presentation of islet antigens in the draining PPLNs (45,46). Recently, Harbers et al. (47) have shown that T-cell proliferation is abolished in dendritic cell–depleted RIP-mOVA/CD11c-DTR mice, and that T-cell proliferation can be reverted when mice are reconstituted with CD11c⁺ splenic cells, confirming the importance of dendritic cells in islet antigen presentation. Here, we show that NOD-M3 mice have a reduced number of CD11c⁺CD11b⁺ cells in both islets and PPLNs at 4 weeks of age. Because dendritic cell migration is highly dependent on chemokines, it is likely that blockade of chemokine function promoted by M3 directly interfered with dendritic cell migration from the blood into islets and draining lymph nodes. Finally, the effects of M3 in our system appear to be local. We have previously shown that M3 is highly expressed by the islets. M3 is detected in the supernatants of transgenic islets cultured in vitro (17,18), and low levels of M3 are present in circulation (41). Thus, the pancreatic levels of expression of M3 are sufficient to block diabetes but do not affect disease development in the salivary and lacrimal glands.

In summary, expression of the chemokine decoy receptor M3 in β-cells prevented development of insulitis and autoimmune diabetes. To date most antagonists targeting single chemokine receptors have not been successful in treating autoimmune diseases in the clinic (48). Our results suggest that chemokine receptor antagonists that can block multiple receptors may be a viable strategy to ameliorate autoimmune diabetes. This concept can be further tested in experimental models using broad-spectrum chemokine antagonists in both prophylactic and therapeutic contexts.

J.S.B. has received funding from the Juvenile Diabetes Research Foundation and National Institutes of Health Grant AI41428. S.A.L. has received funding from the Irene Diamond Fund and National Institutes of Health Grant DK-067381.

We thank Julie Blander, Miriam Merard, Jay Unkeless, Ralph Steinman, and Joshua Farber for reagents and critical comments to the manuscript.