Our data confirm and extend the hypothesis that certain viral agents have the capacity to trigger autoimmune diabetes in genetically susceptible rodent hosts that develop little (LEW.1WR1) or no (BBDR) spontaneous disease when maintained in VAF housing. Triggering of type 1 diabetes by virus alone has previously been documented only for KRV in BBDR and LEW.1WR1 rats (22
). Pretreatment with poly I:C, an activator of innate immunity, increases the penetrance of KRV-induced diabetes in both strains, and it induces diabetes in a third strain, PVG.RT1u, which does not become diabetic after inoculation with KRV alone (22
). Class II MHC–identical WF and PVG.R8 rats do not become diabetic in response to either KRV alone or KRV after poly I:C (22
We now show that diabetes reliably occurs in LEW.1WR1 rats exposed to KRV (a single-stranded DNA virus) and RCMV (a double-stranded DNA virus) and to a lesser extent after exposure to VV (a double-stranded DNA virus) and CoxB4 (a single-stranded RNA virus). The data document that viral diabetogenicity is a complex phenotype with epidemiological characteristics that may help explain why it has been difficult to prove a role for infection in human diabetes pathogenesis and identify the relevant mechanisms.
First, for any given combination of diabetogenic virus and susceptible inbred host genotype, disease penetrance is incomplete; not all infected animals become diabetic and the fraction that does varies from experiment to experiment. For example, RCMV triggered diabetes in a fraction of LEW.1WR1-inoculated rats in every experiment performed, but despite the use of identical virus preparations and inbred rats in VAF housing, that fraction varied from ~20% in the dose response study to as high as ~60% in the maternal immunization study. In the case of CoxB4 virus, diabetes after inoculation was dependent on pretreatment with poly I:C, and even then the frequency of disease was low (~17%). At the other extreme, H-1, a parvovirus with high homology to KRV (33
) does not trigger diabetes in BBDR or LEW.1WR1 rats with or without poly I:C.
Second, the inflammatory substrate of diabetes, insulitis, also exhibited variable penetrance. LEW.1WR1 rats that became diabetic after infection with either KRV or RCMV alone, or CoxB4 after poly I:C, exhibited intense insulitis and selective loss of islet β-cells. In contrast, identically treated animals that did not become diabetic were largely free of insulitis and exhibited no detectable β-cell loss. There was essentially no intermediate phenotype in the islets. It appears that the diabetogenic immune response is difficult to initiate with these three viruses, but once started it goes to completion.
Third, the data document the importance of genetic background. The frequency of spontaneous diabetes in VAF environments is ~2.5% in LEW.1WR1 rats and 0% in BBDR or PVG.RT1u rats (22
). Noninfectious immunological perturbation of innate or adaptive immunity dramatically increases the penetrance and tempo of the disease in all three strains (22
) but not in MHC class II–identical LOU or WF rats (35
). LEW.1WR1 rats become diabetic after exposure to either KRV or RCMV, or after poly I:C to VV and CoxB4. In contrast, BBDR rats develop KRV-induced diabetes but have minimal susceptibility to RCMV-induced disease. PVG.RT1u rats become diabetic in response to KRV but only if pretreated with poly I:C (22
). In their aggregate, the data indicate that susceptibility to virus-induced diabetes is dependent on specific combinations of host and virus. Background genes other than the class II MHC that may modulate the expression of type 1 diabetes after infection are under investigation. Generation of a (WF × BBDR)F2 cohort has identified a genetic locus, Iddm20
, on chromosome 17 that is specifically associated with susceptibility to KRV-induced diabetes (36
). A different locus on chromosome 20, Iddm37
, has been identified in a (LEW.1WR1 × BBDR)F2 cohort treated with KRV (37
). Fourth, susceptibility to virus-triggered diabetes in the rat declines with age. This was true both for older VAF animals and for older rats deliberately exposed 3 or 6 weeks earlier to a different virus. This age dependence suggests that, even in a VAF environment, developmental changes in immune response are determinants of viral diabetogenicity. Interestingly, genetic studies of spontaneous autoimmune diabetes in lymphopenic rats have identified loci that determine age at onset (38
). Speculatively, these loci may harbor genes that modify the immune response over time.
Fifth, and perhaps most importantly, the data show that viruses require a favorable immunological environment—a “fertile field” (40
)—to be diabetogenic. Poly I:C enhances disease induction by KRV in three rat strains (33
) and, in LEW.1WR1 rats, by CoxB4 and VV. We also show, however, that poly I:C reduces the penetrance of RCMV-induced diabetes in the LEW.1WR1 rat. Similarly, antecedent RCMV infection enhances the diabetogenicity of KRV, but inverting the order of infection has little or no effect.
While not specifically investigated in these studies, the underlying mechanisms are likely to involve antiviral responses evoked by the production of type 1 interferons (IFNs) and other proinflammatory cytokines (34
). IFNs can promote the action of β cytotoxic effector cells and, when expressed in β-cells in insulitic islets, can elicit β-cell death (41
Acting through different signaling pathways, both RNA and DNA viruses induce these cytokine responses. At least three classes of innate pattern recognition receptors are involved: the TLRs, the RLHs (like MDA5), and nucleotide-binding oligomerization domain–like receptors (NLRs) (34
). Viruses of different families differ greatly in their ability to induce type 1 IFN responses via each of these receptors, as well as in their ability to evade these responses (34
). New human genome-wide association data suggest that genetic control of these responses may be critical for antiviral responses that protect against type 1 diabetes (42
). Nejentsev et al. (42
) have reported four variants in IFIH1 that lowered type 1 diabetes risk. These variants were predicted to alter the expression and structure of IFIH1, a cytoplasmic helicase that mediates induction of the type 1 interferon response to viral RNA and is a target of poly I:C.
We speculate that genetically determined “tuning” of innate immune responses to virus determines their diabetogenicity (43
). The proinflammatory capability of poly I:C clearly enhances viral diabetogenicity for agents like KRV, but, in contrast, perhaps by impairing viral replication it reduces the diabetogenicity of RCMV. Our sequential infection data support this possibility. The data are consistent with the suggestion that innate immune system can act as a double-edged sword, having both a beneficial role in host defense while leading, in genetically susceptible individuals, to upregulation of proinflammatory autoimmune responses, islet destruction, and diabetes (4
Studies to quantify the type 1 IFN responses in infected rats and to modulate diabetogenicity by manipulating these responses are underway in our laboratories. In their aggregate, the data suggest that the diabetogenicity of a given virus is dependent on both the innate immune environment in which infection occurs and on background genes that modify the immune response.
Our observation that RCMV after KRV or poly I:C reduced or had no effect on diabetes penetrance in the LEW.1WR1 also suggests that, properly timed, an innate immune response might engender the kind of protective response implicit in the “hygiene hypothesis” (18
). Apropos of the hygiene hypothesis, it should be pointed out that spontaneous diabetes does occur in the absence of viral infection, and disease penetrance tends to increase in an increasingly clean environment. This has been reported in the BBDP (25
), KDP (45
), and LEW.1AR1-iddm (46
) strains. BBDP and KDP rats have mutations known to affect T-cell number or function (25
). LEW.1AR1-iddm rats have a normal immunophenotype and the underlying genetic mutation is not yet known (47
Our data do provide one insight into mechanism. They demonstrate the absence of RCMV early antigen in the islets of diabetic LEW.1WR1 rats and suggest that β-cell infection and cytotoxicity are not the mechanism of diabetes induction. This observation is consistent with similar findings in BBDR rats infected with KRV (48
Because of its relevance to human diabetes, our CoxB4 data deserve comment. Enteroviruses including CoxB4 are among the leading candidate viral triggers of human type 1 diabetes (15
). CoxB4 did not by itself induce the disease in the rat strains we studied, but brief pretreatment with poly I:C was associated with subsequent diabetes in 17% of CoxB4-inoculated LEW.1WR1 rats. Our pretreatment protocol using three small doses of poly I:C does induce low-grade insulitis in the LEW.1WR1 rat, and this finding parallels the observation that CoxB3 accelerates diabetes in NOD mice but only in the context of preexisting insulitis (20
). As is true of other viral infections in the rat, diabetes was associated with 3 to 4+ insulitis and selective destruction of insulin-producing, but not glucagon-producing, cells, whereas islet histology in animals that did not progress to diabetes was normal. In addition, there was no evidence of the exocrine pancreatitis that can occur after Coxsackie virus infection in the mouse (21
). The LEW.1WR1 rat may thus provide a platform for additional investigation of the diabetogenicity of enteroviral infection.
Finally, we have shown that maternal exposure to a diabetogenic virus prior to pregnancy enables dams to produce progeny that are resistant to the diabetogenicity of viral inoculation. Our data suggest (but do not prove) that the passive transfer of maternal antibodies can prevent viral triggering of diabetes. These data suggest that it is the immune response to virus that is present in a susceptible genetic context that determines the diabetogenicity (or protectiveness) of infection.
In conclusion, exposure to viruses, including KRV, RCMV, VV, and CoxB4, can affect the penetrance of autoimmune diabetes in genetically susceptible animals. In keeping with previous analyses of both rodents and humans (49
), the connection between infections and autoimmunity is multifaceted and complex. We show that low-frequency “viral footprints” (49
) may be hard to detect. Such low and variable rates of penetrance by viruses of different families acting in the context of outbred genetic backgrounds may account for the lack of firm evidence that viruses trigger human type 1 diabetes. The effect of infection in target tissues may also be dependent on preinitiation of autoimmunity, as in the NOD mouse (20
), and multiple infections might act in concert to precipitate clinical autoimmunity. As posited by the “fertile field hypothesis” (40
), viral infection alone might not be able to induce disease in the absence of other inflammatory factors. Filippi and von Herrath (18
) have written that, “based on current evidence, it…appears impossible to assess the capacity of viruses to modulate [human] type 1 diabetes without knowledge of the state of advancement of autoimmunity and infection history of affected individuals. This is no easy task …” We suggest that the task may be made more tractable by exploring rat models of virus-induced pathogenesis in which the genetic background can be dissected and the permissive cytokine milieu identified and manipulated. They provide insight into the difficulties that have impeded analysis of viral triggering and prevention of diabetes in humans and may assist in the analysis of human datasets such as those being generated by the TEDDY study. Finally, the suggestion that immunization can prevent at least some cases of autoimmune diabetes reanimates a longstanding prevention strategy (50