The effect of rotavirus infection on diabetes development in NOD mice is modified by the age of infection: inoculation of infant NOD mice with rotavirus has been shown to delay or even prevent the development of clinical diabetes [11], whereas inoculation of older NOD mice with established insulinitis has been shown to accelerate progression to diabetes [12]. This accelerating effect appears to be strain-specific, occurring with inoculations of rhesus rotavirus (RRV), as well as a reassortant human-RRV strain, but not with porcine rotavirus CRW-8 [13, 14]. Increased antibody response has also been associated with diabetes acceleration [12, 13].

Several mechanisms have been proposed by which rotavirus infection may precipitate T1D: direct pancreatic infection (leading to beta cell damage), molecular mimicry (in which rotavirus proteins bearing similar sequences to autoantigens activate autoreactive T cells), and bystander activation (in which preexisting autoreactive T cells are independently activated during an immune response against rotavirus) [10].

Although RRV can directly infect NOD mouse islet cells in vitro [15], pancreatic infection by RRV has only been seen in infant NOD mice, and has not been demonstrated to extend to the islet cells in vivo [11, 12]. RRV infection of weanling non diabetes-prone mice has been associated with a temporary decrease in pancreas size as well as transient hyperglycemia, and is apparently dependent on expression of toll-like receptor 3 (TLR3) [16]; however, this effect has not been studied in NOD mice, and no immediate hyperglycemia was observed in rotavirus-infected infant NOD mice [11].

Several other studies have attempted to investigate the NOD mouse immune response to rotavirus in more detail, to provide better understanding of the possibility of molecular mimicry, bystander activation, or other mechanisms. Although sequence similarity has been observed between RRV and islet-specific glucose 6-phosphatase catalytic subunit-related protein, an autoantigen associated with diabetes development in NOD mice, no evidence of molecular mimicry was found upon investigation [17].

It seems likely that the age-based differences in the diabetogenic effect of RRV infection on NOD mice are related to differences in the immune response. In infant mice, RRV results in intestinal as well as extra-intestinal infection, with viral replication demonstrated in the mesenteric lymph nodes (MLN) and thymus [18]. Thymus infection has been associated with changes in T cell development in NOD mice; specifically, infant NOD mice inoculated with RRV displayed an increased ratio of regulatory T cells, which promote self-tolerance, to T effector cells, which act to defend the body against perceived threats [18]. When these same mice were subjected to a strong stimulus as adults, they displayed reduced T cell response, suggesting persistent alterations in immune function [18].

The immune response to RRV inoculation of adult NOD mice has also been investigated [14, 17, 19]. In adult NOD mice, RRV infection induces minimal intestinal inflammation, but RRV can be detected in the pancreatic lymph nodes (PLN) and MLN [14, 17]. Increased expression of pro-inflammatory cytokines has been noted in the PLN, MLN, and spleen of adult NOD mice following inoculation with RRV [14, 17]. In contrast, infectious CRW-8, a porcine rotavirus, does not appear to spread to the PLN or MLN, and significant increases in pro-inflammatory cytokines were not seen in these regions following inoculation of weanling NOD mice [17]. In adult NOD mice, Major Histocompatibility Complex molecule (MHC) expression has been associated with RRV infection [14]. Further, increased B cell MHC I expression in the PLN and MLN was seen in the first week following infection with RRV, but not with CRW-8 (which does not accelerate diabetes development), suggesting that B cell MHC I expression may be related to diabetes acceleration in this mouse model [17]. Ex vivo, infection of mouse splenocytes by RRV, inactivated RRV, or CRW-8 induced B cell activation in a dose-dependent manner, with higher levels of activation in splenocytes from NOD mice as compared to non-diabetes-prone mice; however, this activation was dependent on the activity of rotavirus VP7, toll-like receptor (TLR) 7, and type 1 interferon (IFN), and the presence of CD11c+ dendritic cells (DC) [20]. Further investigation of DC activation in rotavirus-infected NOD mice showed that plasmacytoid DCs (pDCs), which produce type 1 IFN after activation, were activated by RRV but not CRW-8, and that RRV infection transiently increased the ratio of pDCs to conventional DCs (cDCs) [19]. Non-diabetes-prone mice infected with RRV appeared more resistant to these effects, and showed lower infection of the PLN and MLN [19]. RRV also induced type 1 IFN signaling in the MLN and PLN of infected NOD mice, but not non-diabetes-prone mice [19]. Even in the absence of type 1 IFN signaling, B cell activation in the PLN and MLN still occurred, albeit delayed [19]. However, type 1 IFN signaling was necessary for RRV to accelerate diabetes onset in NOD mice [19]. The role of type 1 IFN signaling in rotavirus--induced diabetogenesis was further assessed in relation to melanoma differentiation-associated factor 5 (MDA5), which is encoded by the IFIH1 gene (associated with T1D risk) [21]. Experiments showed that MDA5 was important to limiting murine infection with a human rotavirus isolate, especially in the pancreas, and that this activity occurred both dependently and independently of IFN [21].

Taken together, this body of evidence suggests that rotavirus-induced diabetes acceleration in NOD mice is strongly related to the immune response to the infection, specifically the activation of pDCs and B cells in the lymph nodes, as well as B cell MHC I expression and the activity of IFN and other pro-inflammatory cytokines. Further, preexisting diabetic pathology (insulinitis, and the existence of autoreactive cells) appears necessary for acceleration of disease progression.

The possibility of molecular mimicry between rotavirus and T1D-associated autoantigens guided specific research questions within large cohorts designed to elucidate T1D etiology. The two cohorts in which this potential association has been studied in most detail are the BabyDiab cohort in Australia, and the Type 1 Diabetes Prediction and Prevention project (DIPP) in Finland [22, 23]. In both cohorts, infants at high risk of T1D are enrolled from birth and followed longitudinally, with regular serum samples taken and assayed for T1D-associated autoantibodies (antibodies to IA-2, antibodies to GAD, and insulin antibodies [IIA]).

Researchers attempted to compare the timing of autoantibody seroconversion to the timing of rotavirus seroconversion in such cohorts, but findings were inconsistent (Table 1). Honeyman et al. studied the relationship between the appearance of autoantibodies and that of rotavirus antibodies among 24 children from the BabyDiab cohort who had developed clinical T1D or in whom autoantibodies had been detected at least twice during follow-up. Rotavirus infection, defined as an increase in rotavirus antibodies of ≥40% between consecutive time points (equivalent to an increase of ≥2 interassay coefficients of variation [CV] in this study), was found to be significantly associated with the appearance of autoantibodies in the 24 children during that same time period [22]. However, when a similar analysis was conducted by Blomqvist et al. of 29 cases from the DIPP cohort, no significant association was found in the appearance of rotavirus and autoantibody seroconversion [23]. A matched case-control analysis also demonstrated no significant differences in the occurrence of rotavirus infections during the time periods when cases developed autoantibodies [23]. Blomqvist’s research group later updated this analysis with a larger population from DIPP, comprising 43 children with incident T1D, 36 children with at least 2 T1D-associated autoantibodies, and 104 children without any autoantibodies (controls) [24]. In this study, they again found no significant association between the presence of rotavirus antibodies and the presence of T1D-associated autoantibodies [24]. Cellular responsiveness to rotavirus was also not associated with T1D-associated antibodies. Production of interleukin-4 (IL-4) in response to stimulation with purified rotavirus was stronger in children with autoantibodies as compared to controls, but no significant difference was seen between children with clinical T1D and controls [24].

These discrepant findings may be attributable to population differences (both in host genetic background as well as circulating virus genotype) or to methodological differences. For instance, Honeyman et al. used human rotavirus in their EIA, and defined rotavirus infection by an increase in IgA or IgG, whereas the first Blomqvist et al. study used an indirect EIA method based on the Nebraska Calf Diarrhea Virus, and defined rotavirus infection by a 2-fold absorbance increase in IgG [22, 23]. Associations in Honeyman et al. appeared primarily driven by IgA, which was not measured by Blomqvist et al. It is also possible that evidence of rotavirus infection in the Blomqvist study population could have been masked by high levels of maternally transferred IgG early in life, which might have contributed to the null results seen. However, in the subsequent analysis of the DIPP cohort, which again showed a null association, both IgA and IgG were measured [24], suggesting additional factors at play. Definitions of separate rotavirus infection were also not consistent across populations: whereas Honeyman et al. interpreted consecutive increases in rotavirus antibody titers as additional infections, Blomqvist et al. considered such situations as part of the same initial infection. Thus, rotavirus infections may have been more sensitively detected in Honeyman’s study, but more specifically detected in Blomqvist et al.’s research.

Outside the BabyDiab and DIPP cohorts, several other studies have assessed the potential relationship between T1D and rotavirus infection. Ataei-Pirkooh et al. conducted a case-control study of children ≤14 years of age attending the Children’s Medical Center in Tehran, with cases defined by incident T1D diagnosis and controls recruited from children who presented to the hospital but did not have any diabetes-related history [25]. Although anti-rotavirus antibodies were significantly higher in cases as compared to controls, and rotavirus antibodies were significantly correlated with islet autoantibodies, the analysis was not adjusted for age, a potentially important confounder [25]. In a sex-, age-, geographical-, and time-matched case-control study, Bian et al. screened serum from 42 U.S.-based cases with new-onset T1D and 42 healthy controls (verified to be autoantibody-negative) [26]. Multiplex viral protein arrays were used to test for IgA and IgG reactivity to viral proteins from 23 viruses, including rotavirus and 6 others that had been previously associated with T1D; reactivity to the auto-antigen IA-2 was also tested [26]. Only responses to IA-2 and Epstein-Barr Virus (EBV) were significantly higher in cases as compared to controls [26]. Shulman et al. tested maternal sera and cord blood samples from healthy deliveries in Israel during the winter viral season [27]. Antibodies to rotavirus and the T1D autoantigen GAD65 were found to co-occur in some maternal and cord blood samples, while in other samples, only cord blood was antibody positive, suggesting the possibility of an independent fetal antibody response. However, the cross-sectional study design precludes any conclusions about whether these antibodies had any effect on clinical development of T1D [27].

Makela et al. later analyzed a larger subset of children from the DIPP study to assess correlations among antibodies to enteric viruses (including rotavirus) and antibodies to bovine and human insulin [28]. Of 238 children available for analysis, 211 had information on rotavirus antibodies, 61 developed at least one autoantibody, and 46 developed at least 2 autoantibodies [28]. Rotavirus infections (as well as enterovirus and adenovirus infections) before the age of 6 months were found to be significantly associated with human and bovine insulin-binding antibodies, whereas infections occurring later in infancy were not found to be significantly associated with insulin-binding antibodies; this association appeared to be stronger in children receiving cow-milk-based formula before the age of 3 months [28]. Insulin-binding antibodies were significantly associated with development of autoantibodies as well as development of clinical T1D [28]. These interactions were further studied in another, larger case-control study drawn from the ongoing DIPP study cohort [29]. Rotavirus antibodies, respiratory syncytial virus antibodies, bovine insulin-binding antibodies, and autoantibodies were analyzed among 107 autoantibody-positive cases and 446 controls. As previously documented by this research group, rotavirus infection before the age of 6 months was associated with development of T1D-associated autoantibodies with an indication that this association was strongest in those infants exposed to cow’s milk formula before the age of 3 months [29].

Zhao et al. selected a subset of the DIPP cohort for a case-control study of the intestinal virome [30]. Eleven children who had developed autoimmunity were matched to 11 controls, and sequential stool samples were analyzed for each group [30]. No significant difference was found in rotavirus prevalence or abundance when comparing stool samples from cases (before seroconversion) and controls [30].