We developed a tg mouse model where a viral transgene, the LCMV full-length GP, was expressed only in β cells of the pancreatic islets of Langerhans 
. While insertion and expression of the viral gene, per se
did not lead to diabetes (incidence <1%), the induction of an anti-LCMV CD8+
-specific CTL response that reacted with the viral expressed transgene effectively and routinely produced T1D (incidence >94%). Previous studies indicated that CD4 T cells failed to play a role in this model, since their depletion did not alter either the kinetics or incidence of the resultant diabetic disease 
. This result stands in contrast to a model in which the viral transgene was expressed both in the β cells and in the thymus. In this model, anti-LCMV CD8 T cells of high affinity are selected out in the thymus with antiviral CD8 T cells of lower affinity going to the periphery. To cause T1D, virus-induced specific CD4 T cells in addition to virus-specific CD8 T cells were required 
The full-length viral GP in H-2b mice has only four distinct CD8 CTL epitopes that recognize two immunodominant LCMV epitopes, GP33 and GP276, as well as two weaker subdominant CD8 CTL epitopes, GP92 and GP118. Deletion of both immunodominant GP epitopes from LCMV (GPV) failed to cause T1D despite the generation of virus-specific CD8 CTL to the two weak subdominant GP epitopes. The percentages of GP-specific CD8 CTL generated in the spleen following infection with GPV were as follows: GP33 (1.2%), GP276 (0.1%), GP92 (<1%), and GP118 (2.7%) for a total of 5% as determined by tetramer analysis. This number of virus-specific CD8 T cells was insufficient to cause diabetes. Islets in RIP-LCMV GP tg mice infected with GPV routinely failed to show an accumulation of T cells; there was no β cell damage, and blood glucose levels were normal. Although virus-specific CD4 T cells were generated, they were inefficient in providing help to virus-specific CD8 T cells to damage the islets or incite T1D. In contrast, when the total percentage of LCMV GP-specific CTL in the spleen reached 20%, diabetes was observed in >94% of RIP-LCMV GP tg mice, and if the percentage was 9%, T1D occurred in 50% of RIP-LCMV GP tg mice. These findings indicate that there is a quantifiable number of LCMV GP CD8 T cells required to cause diabetes and that a therapeutic approach directed to lowering or limiting; but not necessarily eliminating virus (self) specific CD8 T cells is likely to be successful in treating T1D.
Following LCMV infection, 12.5% of CD8 T cells in the spleen are specific to the dominant GP33 epitope which leads to migration of these virus-specific CD8 T cells into the islets. Unexpectedly, only a small number of these virus-specific CD8 CTL, on the order of 1–2% of the total virus-specific CD8+
T cell population were found in the islets during the peak of disease. Earlier studies in the LCMV model indicated that, at any one snapshot in time, a single CTL could engage up to three target cells in vivo
, and under optimal in vitro culture conditions, one virus-specific CD8 CTL could interact and lyse up to 10–12 virus-infected target cells 
. Thus, it is likely that in vivo CTLs amplify their efficiency in killing β cells by engaging and moving to multiple targets. Nevertheless, since only a low number of effector cells are required to cause disease, a therapeutic focus on reducing the numbers of such cells may be sufficient to curtail β cell damage and prevent T1D. Indeed, past studies using specifically designed peptides that competed for MHC binding to immunodominant LCMV CD8 T cells were able to diminish or abort diabetes 
. Additionally, the number of effector CD8 T cells, as documented in Listeria infection, appears tightly controlled by TGF-β-mediated apoptosis 
. This suggests that TGF-β and likely other factors may also play a role in our in vivo analysis of the generation of effector CD8 T cells.
Analysis of interactions at the immunologic synapse between virus-specific CTL and β cell targets following in vivo infection showed a proboscis or tongue-like extension from the CTL cytoplasm to the target cell ( and ). This CTL extension appears anchored by LFA molecules that connect the target cell to the killer cell and provide a bridge for the passage of cytotoxic granules onto and into the β cell. Indeed, in vitro studies clearly showed that cytoplasmic granules containing serine esterases travel down that proboscis-like bridge to the target cell. This leads to a puncturing of the target cell membrane followed by actin/myosin dissociation, blebbing, and cellular destruction 
. The binding of virus-specific CD8 T cells to β cells and the demonstration of perforin-like molecules released by CTL onto β cells strongly suggest a lytic role for CD8 T cells in destroying β cells to cause T1D. In addition, inflammatory cytokines like interferon-γ generated by virus-specific CD8 T cells are important contributors to T1D 
. Interestingly, in LCMV-induced acute leptomeningitis, LCMV-specific CD8 T cells caused damage indirectly by first identifying the virus-infected target followed by a release of cytokines and chemokines which call in myeloid cells that also participate in cellular injury 
. Whether a similar event occurs or plays a role in virus-induced diabetes is unknown.
While our data clearly demonstrate that an overlap between a viral and self-protein can give rise to T1D following infection, others have suggested that bystander T cell activation might also serve as a disease trigger 
. Bystander activation occurs when an infection causes T cells to acquire effector functions in a non-specific manner 
. Infection of BDC2.5 T cell receptor transgenic mice with coxsackie virus CB4 strain (a pancreatrophic virus) results in a 90% incidence of T1D 
. Because BDC2.5+
T cells recognize an islet antigen, but do not cross-react with coxsackie virus, it was concluded that T1D was induced by bystander activation of self-reactive T cells. It is important to note, however, that bystander T cell activation is a relatively inefficient process estimated to occur in only 1 in 200 virus-specific CD8+
T cells infected with a non-overlapping pathogen 
. Thus, the high frequency of islet-specific T cells in the repertoire of BDC2.5 mice likely favors the induction of T1D following coxsackie virus infection 
. Importantly, when non-transgenic diabetes-prone NOD mice were infected with coxsackie virus, the pathogen could accelerate but not initiate T1D 
. These data suggest that viruses may accelerate autoimmune diseases through bystander activation once a critical threshold of self-reactive T cells is reached. A shared epitope between a viral and self-protein (molecular mimic) is presently the most efficient mechanism to generate the number of self-reactive T cells required for an autoimmune disease to develop, and once initiated, a pathogenic disease process does not need to rely on bystander T cells 
, although they may contribute to tissue pathology 
. With the low levels of virus-specific T cells in the islets it is unlikely that bystander T cells play a role in T1D in our model.
Host genes, autoimmune responses, cytokines/chemokines and virus have all been implicated in the initiation and progression of T1D. In our view, these various components and activities can be integrated into one theme associated with two distinct events. First, a viral or other environmental exposure occurs early in life and results in restricted but low level expression of the viral or microbial gene(s) in β cells of the islets of Langerhans. This is an event that by itself need not cause disease since most hosts are hyporesponsive or tolerant to the viral product. Later, a second event occurs. Infection with the same virus (or microbe) or one bearing cross-reacting antigenic determinants or conformation shapes lead to an immune response that activates sufficient CD8 T cells to migrate to and infiltrate pancreatic islets, injure β cells, and cause T1D. Manipulations that lower the numbers or functions of activated CTL or impair their migration to the islets should be beneficial as therapy for T1D.