Diabetogenic Potential of Human Pathogens Uncovered in Experimentally Permissive β-Cells

SOCS-1-Tg (3) and non-Tg NOD mice and SOCS-1-Tg and non-Tg NOD.scid mice were bred and maintained in a specific pathogen-free environment at the Scripps Research Institute. Breeding pairs of IFNαβγR^−/− mice (129S6/SvEv) were kindly provided by Dr. S. Virgin (Washington University School of Medicine, St. Louis, MO), and wild-type 129S6/SvEv mice were purchased from Taconic Laboratories. Female and male mice were infected at the age of 8–9 weeks. All experiments were conducted in accordance with institutional guidelines for animal care and use.

MCMV (Smith strain American Type Culture Collection VR-1399) was isolated and the viral titer was determined as described (38). Propagation and determination of titers of CVB3 (Nancy strain) and LCMV has been described previously (30,39). Mice were infected with one intaperitoneal injection of LCMV (10⁵ plaque-forming units [PFU]), MCMV (10³–10⁵ PFU), or CVB3 (50–100 PFU). Age-matched animals, mock infected with Hanks’ balance salt solution only, served as controls.

Venous blood glucose (BG) concentrations were measured under nonfasting conditions using Glucometer Elite strips (Bayer, Pittsburgh, PA). Mice were considered diabetic after two consecutive measurements of BG >13.8 mmol/l (250 mg/dl). Onset of diabetes was dated from the first of the sequential measurements.

Harvested organs were fixed in formalin and embedded in paraffin. Tissue sections (5 μm thick) were stained with hematoxylin and eosin or with a primary antibody against insulin, glucagon (Dako, Carpinteria, CA), VP-1 (Dako, Carpinteria, CA), or the intranuclear MCMV protein immediate early 1 (IE1) 89-kDa phosphoprotein pp89. The clone (Croma 101) generating the latter antibody was originally generated by S. Jonjic (University of Rijeka, Rijeka, Croatia) and kindly provided by Dr. S. Virgin (Washington University School of Medicine, St. Louis, MO). Both VP-1 and the croma-101 antibodies were biotinylated in house. Slides were counterstained in Mayer’s hematoxylin. Alternatively, organs were fixed in OCT compound embedding medium (Miles, Diagnostic Division, Elkhart, IN) before sectioning and fixation in 95% ethyl alcohol.

Pancreatic sections (2–3 levels per organ) were stained with hematoxylin and eosin and ranked for insulitis using the following classes: A, normal islet morphology with no peri-insulitis or insulitis; B, peri-insulitis; C, insulitis; and D, islet remnant (40). Because of the difficulty in separating periinsulitic infiltrates from mononuclear cells infiltrating the exocrine part of pancreata of infected mice, A and B were combined when ranking the CVB3-infected mice (3).

Accumulated incidence of diabetes was determined by χ² test.

SOCS-1-expressing β-cells fail to mount an antiviral response after a challenge with type I or II IFNs (3). To determine whether IFN-induced antiviral pathways regulate β-cell survival during a systemic LCMV infection, 8- to 9-week-old SOCS-1-Tg and non-Tg nonobese diabetic (NOD) mice were infected with LCMV (10⁵ PFU). At this age neither SOCS-1-Tg nor non-Tg mice had developed spontaneous diabetes, which occurs only after the mice have reached an age ≥16 weeks (M.F. and N.S., unpublished observations; and rev. in 41). The mice (SOCS-1-Tg mice, n = 6; non-Tg mice, n = 10) were followed up to 21 days postinfection (p.i.). During this time period, none of the mice, including uninfected SOCS-1-Tg and non-Tg mice (data not shown), developed elevated BG levels (Fig. 1A and B).

Lymphocytes surround and infiltrate the islets before the onset of spontaneous diabetes in NOD mice. This so-called insulitis is already well established at an age of 6–8 weeks (41). A histological evaluation of pancreata from another set of infected mice killed on day 7 (SOCS-1-Tg, n = 4; non-Tg, n = 3) or 14 p.i. (SOCS-1-Tg, n = 8; non-Tg, n = 4) revealed many intact islets with a normal distribution of endocrine cells (data not shown). Some islets showed peri-insulitis or insulitis, but there was no difference in the degree of peri-insulitis or insulitis between infected and uninfected SOCS-1-Tg and non-Tg mice (data not shown).

To determine whether β-cell permissiveness to LCMV is regulated by IFNs, the presence of LCMV in organs of infected mice was examined by immunohistochemistry. Each of seven SOCS-1-Tg and six non-Tg mice was infected with LCMV and then killed on day 7 p.i. As expected (42), LCMV-positive cells were detected in the splenic red pulp of both non-Tg and SOCS-1-Tg mice (Fig. 2B and data not shown). Few LCMV-positive cells, confined to the endothelium lining of larger blood vessels and to the ductal epithelium, were found in pancreata of infected mice (Fig. 2D and data not shown). Cells harboring LCMV were not seen in islets of SOCS-1-Tg or non-Tg mice (Fig. 2D and data not shown) or in organs of uninfected mice (Fig. 2C and data not shown).

We next determined whether β-cells depend on IFN signals to survive a CMV infection by infecting SOCS-1-Tg (n = 17) and non-Tg (n = 11) littermates with MCMV (10⁴ PFU). A decrease in BG levels was observed on day 3 p.i. (Fig. 1C and D). However, this decrease was transient and, with the exception of 2 SOCS-1-Tg mice (2 of 17), the majority of the mice reverted to normoglycemia by day 5 p.i. The mice that did not return to normal glucose homeostasis developed persisting severe hyperglycemia. By 21 days p.i., diabetes had not occurred in any of the other infected mice or in uninfected controls (Fig. 1C and D and data not shown). Additional experiments showed that an increased dose of MCMV (5 × 10⁴ or 1 × 10⁵ PFU) failed to trigger diabetes in both non-Tg and SOCS-1-Tg mice (data not shown).

Systemic MCMV infection induces dose-dependent production of IFN-α (26,43), and there was a possibility that the high dose of MCMV used in the experiments until now resulted in amounts of IFN-α that could overcome SOCS-1’s ability to block IFN signaling. However, diabetes did not occur in any mice challenged with a lower dose of virus (1 × 10³ PFU; SOCS-1-Tg, n = 18; non-Tg, n = 10) (data not shown).

To investigate whether MCMV was able to cause pathologies in pancreata of infected mice, another set of mice (SOCS-1-Tg, n = 7; non-Tg, n = 6) infected with MCMV (10⁴ PFU) was killed on day 7 p.i. Spleens and livers of infected SOCS-1-Tg and non-Tg mice revealed classic MCMV-related pathologies (Fig. 3 and data not shown). No comparative differences were seen in the organs of SOCS-1-Tg and non-Tg mice.

The pancreatic exocrine tissue of infected mice had diffuse infiltrations of mononuclear cells. Interestingly, insulitis appeared to have progressed in pancreata of nondiabetic SOCS-1-Tg mice (Table 1). Indeed, compared with uninfected mice, there was a decreased number of intact islets free of peri-insulitis and insulitis. This increased recruitment of cells to the islets compared with uninfected control mice was not observed in infected non-Tg mice (Table 1). Despite the presence of peri-insulitis and insulitis, numerous intact islets, with a normal distribution of endocrine cells, were observed in both SOCS-1-Tg and non-Tg mice (data not shown). One exception was the pancreas from one of the SOCS-1-Tg mice, which had succumbed to diabetes early after infection (see above; the other diabetic mouse succumbed to the infection and could therefore not be studied). Its pancreas was nearly devoid of islets. Mononuclear infiltrates were observed in the few remaining islet-like structures, and insulin-positive cells were absent (data not shown).

MCMV was detected by immunohistochemistry in the hepatic tissue of infected SOCS-1-Tg and non-Tg mice killed on day 4 (SOCS-1-Tg, n = 4; non-Tg, n = 4) (Fig. 3F and data not shown) or day 7 p.i. (SOCS-1-Tg, n = 8; non-Tg, n = 6) (data not shown). As previously demonstrated (44), few MCMV-positive hepatocytes were detected on days 11 (SOCS-1-Tg, n = 2; non-Tg, n = 2) and 14 p.i. (SOCS-1-Tg, n = 2; non-Tg, n = 2) (data not shown). Little, if any, MCMV was detected in the pancreata of infected SOCS-1-Tg and non-Tg mice killed on days 4, 7, or 14 p.i. (Fig. 3H and data not shown).

We next asked whether intact β-cell IFN responses are important for β-cell survival during systemic CVB3 infection. The majority of non-Tg mice challenged with CVB3 (10 of 14) survived the initial observation period of 21 days, and none of the mice became hyperglycemic (Fig. 1E). Infected SOCS-1-Tg mice showed an initial drop in BG levels, similar to that of non-Tg mice (compare Figs. 1E and F). However, 5–7 days after infection, the mice developed severe and persistent hyperglycemia (nonfasting BG levels >20 mmol/l). By day 9 p.i., all SOCS-1-Tg mice (16 of 16, P < 0.01 vs. non-Tg mice) were overtly diabetic. Age-matched SOCS-1-Tg and non-Tg mice that remained uninfected did not develop diabetes during the study period (data not shown).

A histological analysis of organs from infected mice revealed a nearly complete absence of exocrine tissue in pancreata from CVB3-infected SOCS-1-Tg and non-Tg mice killed 7–21 days p.i. (Fig. 4). Despite the massive destruction of the surrounding exocrine tissue, intact islets with a normal distribution of insulin-staining (Fig. 4A), glucagon-staining (Fig. 4B), and somatostatin-staining cells (data not shown) were present in pancreata of non-Tg mice. The degree of insulitis and islet destruction was similar to that of uninfected non-Tg mice (Table 1). The islets of SOCS-1-Tg mice were strikingly different; some islets exhibited a collapsed appearance, with the presence of only glucagon-staining (Fig. 4D) and a few somatostatin-staining cells (data not shown). Other islets not only lacked insulin-staining cells but were also decorated by infiltrating mononuclear cells (Fig. 4C, Table 1, and data not shown).

The NOD mouse harbors a pool of autoreactive T-cells that upon activation attack and destroy β-cells (41). To test whether autoreactive T-cells or any other component(s) of the adaptive immune system were causing β-cell destruction after CVB3 infection, non-Tg and SOCS-1-Tg mice crossed to a NOD.scid background, lacking functional T- and B-cells (45), were used for the next set of infections. By day 21 p.i., none of the non-Tg.scid mice (n = 8) had developed diabetes. Uninfected mice also remained free of diabetes (SOCS-1-Tg.scid, n = 8; non-Tg.scid, n = 7) (data not shown). Unlike the non-Tg mice, all CVB3-infected SOCS-1-Tg.scid mice (n = 5) developed hyperglycemia on days 5–6 p.i. (data not shown). These mice were killed on day 7 p.i. for histological analysis. Their pancreata were compared with those of two non-Tg mice killed on day 7 p.i. The examination revealed that both SOCS-1-Tg and non-Tg mice had suffered a major loss of pancreatic acinar cells (data not shown). The islets of non-Tg mice were mainly intact. Only a few islets showed a minor degree of disruption by infiltrating immune cells, and the distribution of insulin- and glucagon-positive cells was similar to that of uninfected mice (data not shown). In contrast, the islets of SOCS-1-Tg mice were either completely infiltrated by immune cells, with only a few, if any, remaining insulin-positive cells, or they consisted only of endocrine cells other than β-cells (data not shown).

Next, we wished to determine whether the unresponsiveness to IFN resulted in permissiveness of SOCS-1-expressing islet cells to CBV3 infection. CVB3 was present in the exocrine tissue of both SOCS-1-Tg and non-Tg mice on days 3 and 4 p.i. (Fig. 4F, G, and H and data not shown). On these time points, i.e., shortly before diabetes occurred in CVB3-infected SOCS-1-Tg mice, no virus could be detected in the islets of non-Tg mice (day 3, n = 5; day 4, n = 8) (Fig. 4F and data not shown). In stark contrast, CVB3 antigen was observed in islet cells within the pancreata of all infected SOCS-1-Tg mice on both of these days (day 3, n = 4; day 4; n = 4) (Fig. 4G and H).

SOCS-1 is primarily an inhibitor of IFN signaling, but it can block the signaling of other cytokines as well (rev. in 20,46). To determine the specific role for intact IFN responses in regulating permissiveness to CVB3, mice lacking intact IFN receptors (IFNαβγR^−/−) and their wild-type counterparts were infected with CVB3. Because IFNαβγR^−/− mice do not survive for more than 2–3 days after challenge with CVB3 (M.F. and N.S., unpublished observations), the mice were killed on days 2 or 3 p.i. As evaluated by immunohistochemistry, CVB3 was found in the exocrine tissue of both IFNαβγR^−/− (n = 6) and wild-type (n = 4) mice but was absent from the islets of infected wild-type mice (data not shown). In contrast, by day 2 p.i., cells within the islets of IFNαβγR^−/− mice were already infected by CVB3 (data not shown).

In the present study, three viruses in which tissue tropism and/or pathogenicity is regulated by IFNs (22,30,35–37) were used to demonstrate that β-cells are likely to depend on IFN-induced antiviral defense for their survival during systemic infection by CVB3 but not LCMV. Interestingly, the studies have also uncovered a role for MCMV during disease progression.

LCMV infection can occur in humans (22), but to the best of our knowledge LCMV infections have not been associated with the development of type 1 diabetes. In adult immunocompetent mice, LCMV is cleared within 14 days (22,47). The virus can cause a persistent infection if given to newborn mice or mice with immune deficiencies. During persistent infections, LCMV antigens have been found in endocrine cells such as the islet cells (48,49). However, LCMV has only on rare occasions been detected in β-cells cells during acute infection of adult mice (22,47). The present study showed that adult mice harboring β-cells lacking IFN responses remained free of both diabetes and obvious β-cell damage after an acute LCMV infection. Preliminary studies suggest that LCMV also fails to precipitate diabetes in older (11 weeks old) SOCS-1-Tg mice (M.F. and N.S., unpublished observations). Little if any LCMV could be detected in the islets of infected SOCS-1-Tg and non-Tg mice, suggesting that LCMV tropism for β-cells is not primarily controlled by IFNs. LCMV uses α-dystroglycan (α-DG) for cell entry (50,51), although alternative receptors have been proposed (51). Expression of α-DG has been shown in the pancreas (52), but it remains to be established to what extent this receptor is expressed by β-cells and whether the level of α-DG expression is a primary factor determining LCMV’s tropism for the β-cell.

Several findings have led to the suggestion that CMV infections may contribute to the onset of type 1 diabetes in humans (53–55). HCMV can infect human fetal islet cells in vitro (56), and characteristic CMV inclusion bodies have been found in islet, acinar, and ductal cells of children who died of disseminated CMV infection (12). Although these observations have suggested that a direct infection of β-cells may occur after systemic CMV infection, the factor(s) determining β-cell permissiveness to CMV infection have remained unexplored. Furthermore, it has been unclear whether a direct infection of β-cells could contribute to the proposed diabetogenic potential of this virus.

In the present study, diabetes was observed at a low frequency (∼12% of mice given 10⁴ PFU) in infected SOCS-1-Tg mice, revealing a cryptic diabetogenic potential by MCMV. However, MCMV was only occasionally found in pancreata, including the islets, of infected mice. Thus, IFNs may not be major regulators of β-cell permissiveness to MCMV infection. It is also possible that the overall low tropism of MCMV for the pancreas results in only low levels of virus reaching the islets. Accordingly, the rapid destruction of β-cells and onset of diabetes in a few SOCS-1-Tg mice after MCMV infection may have occurred without a direct infection of islet cells. Indeed, the nature and expression level of the cell surface receptor used by MCMV to infect islet cells remains to be established.

A mild pancreatitis was observed both in non-Tg and SOCS-1-Tg mice soon after MCMV infection. Notably, the percentages of islets showing either peri-insulitis or insulitis were somewhat higher in MCMV-infected SOCS-1-Tg mice than in infected non-Tg and uninfected mice (see Table 1). SOCS-1-Tg mice that succumbed to diabetes after MCMV infection suffered a selective loss of islet cells staining positive for insulin, and mononuclear cells had infiltrated some of the remaining islet-like structures. It is therefore possible that those infiltrating cells may have played a role in β-cell destruction. Accordingly, autoreactive T-cells known to contribute to β-cell destruction in the NOD mouse (41) could have been involved. However, other components of the immune system could also have participated. In a previous study (3), it was demonstrated that CVB4-instigated β-cell destruction in the SOCS-1-Tg mouse was in part dependent on the actions of natural killer (NK) cells. NK cells are critical in controlling early viral replication after infection with MCMV and possibly also HCMV (57). Moreover, NK cells may contribute to autoimmunity (58). Thus, it cannot be excluded that MCMV could have driven the activation and expansion of such cells, which could have precipitated diabetes in SOCS-1-Tg mice. An elucidation of the mechanism(s) behind β-cell destruction may provide better understanding for how a CMV infection might contribute to type 1 diabetes. Clearly, our results suggest that individuals with impaired β-cell antiviral defense may have an increased risk to progress to type 1 diabetes after infection with CMV.

CVB3 infections have been associated with type 1 diabetes in humans (28,29), and IFN-α is important for human islet cell survival after CVB3 infection in vitro (21). Here, we demonstrated that upon infection with CVB3, the SOCS-1-Tg mice rapidly succumbed to diabetes. At the time of diabetes onset, few intact β-cells remained, and β-cell destruction occurred without the actions of T- and/or B-cells. Moreover, there was a remarkable difference between SOCS-1-Tg and non-Tg mice in that CVB3 had infected the islets of SOCS-1-Tg mice but not those of non-Tg mice. CVB3 also infected islets of mice lacking functional IFN receptors. Taken together, these observations strongly suggest that CVB3 tropism for the β-cell is regulated by IFNs produced early during the infection. Moreover, the finding that CVB3 infection of islet cells preceded the disappearance of β-cells suggests that the virus may have caused a direct cytolysis of infected β-cells. Although the precise mechanism(s) by which CVB3 causes β-cell destruction remains to be elucidated, these findings propose that CVB3 can cause β-cell destruction and diabetes in a host lacking proper β-cell antiviral defense. Moreover, the findings also suggest that the ability of CVB3 (or any other virus with a tropism for the pancreas) to cause β-cell damage could be inversely related to its capacity to trigger the production of IFN in the infected host. A low IFN response could favor the virus’ ability to infect β-cells, whereas a robust IFN response would allow the β-cell to mount an efficient and lifesaving antiviral response.

Enteroviral infections may trigger or precipitate type 1 diabetes (5). The SOCS-1-Tg mouse model has now highlighted a crucial role for IFN-induced β-cell antiviral defense in preventing β-cell destruction and diabetes during not only CVB4 (3) but also CVB3 infection. Interestingly, a recent study demonstrated variations in susceptibility to enterovirus infection among human islet cell preparations (19). Therefore, differences in β-cell antiviral defense may naturally exist. It is likely that these differences reside downstream of the host’s ability to produce IFNs because recent findings have indicated that individuals suffering from type 1 diabetes may in fact have increased serum levels of IFN-α (59) and enhanced IFN-γ responses of T-cells to CVB4 antigens (60) as compared with healthy individuals. It is possible that the efficiency by which the β-cell mounts an antiviral defense in response to IFN determines its susceptibility to infection and subsequent damage. Therefore, it will be of interest to characterize the IFN-induced antiviral defense mechanism in β-cells and then to determine whether individual variations in this could explain why certain islet preparations were less vulnerable to CVB3 infection than others. All together, the findings presented here suggest that target cell antiviral responses critically regulate an individual’s risk for developing viral-induced type 1 diabetes.

This work was supported by National Institutes of Health Grant ROI-AI42231 and an advanced postdoctoral fellowship from the Juvenile Diabetes Research Foundation (to M.F.).

We thank Drs. N. Gascoigne and K. Holmberg of the Scripps Research Institute for sharing their stock of the MCMV virus and Dr. M. Buchmeier of the Scripps Research Institute for the LCMV antibody. We also thank Dr. Horwitz and other members of the Sarvetnick lab for fruitful discussions.

This is manuscript number 15541-IMM from the Scripps Research Institute.