In this study, we examined the contribution of NK cells to spontaneous diabetes in NOD and NOD.NK1.1 mice. NK cells from NOD mice do not express a marker that can be targeted by currently available antibodies for depletion and therefore this study required the use of the NK1.1-congenic NOD strain to allow for in vivo targeted depletion of NK cells. Using NKp46 as a marker for NK cells in NOD mice, we found that NK cells were found in comparable numbers in the diseased pancreata of NOD mice and NOD.NK1.1 mice.
The presence of significant numbers of NK cells in the pancreas is specific to the NOD mouse and disease state. Few (or in some experiments no) NK cells were found in the pancreas of B6.g7 mice as detected by flow cytometry; however, these levels were consistently lower than in Rag2−/−
NOD mice, which had much fewer numbers of NK cells in the pancreas compared to diseased NOD mice. These results differ from another study in which they found significant numbers of NK cells in the pancreas of B6.g7 mice 
. The method of cell isolation from the pancreas might be responsible for this difference as we perfused our mice to remove any circulating NK cells from the blood prior to tissue harvest; it is unknown if the previous study followed a similar protocol. We also did not detect NK cells in the pancreas of Rag2−/−
NOD mice and B6.g7 by immunohistochemistry. Collectively, data from experiments evaluating Rag2−/−
and wild-type NOD mice suggest that the NOD pancreas harbors a unique environment for NK cells, but the infiltration of significant numbers of NK cells requires the presence of adaptive immunity.
Pancreatic NK cells in NOD and NOD.NK1.1 mice have an activated surface phenotype, but results from our ex vivo
stimulation assays suggest that pancreatic NK cells are somewhat hyporesponsive as measured by their capacity to degranulate or produce IFNγ, consistent with prior finding 
. Yet, because the pancreatic NK cells were similar in their responsiveness to liver NK cells in NOD mice, this hyporesponsiveness is not specific to pancreatic NK cells, but might be due to the methods used to isolate NK cells from solid organs or the unique microenvironment of tissues. Consistent with this interpretation, we have also noted a lower responsiveness of liver NK cells compared to splenic NK cells in C57BL/6 mice (our unpublished observations). In contrast, the in vivo
response to poly I:C by liver and pancreatic NK cells was equal to that of splenic NK cells, suggesting that indeed these pancreatic NK cells are capable of responding quickly, similar to NK cells from the spleen.
Dendritic cell maturation is important to T cell activation and autoimmune disease progression 
. DCs from Rag2−/−
NOD mice displayed the same maturational phenotype as DCs from the NK cell-deficient NOG strain mice. Thus, at steady-state NK cells did not promote DC maturation in the absence of an adaptive immune system in NOD mice. Moreover, the activation status of DCs in NOD mice was not affected by the depletion of NK cells. Therefore, NK cells are not required for pancreatic DC maturation or activation in NOD mice. Thus, when using DCs as sentinels of disease status 
, the absence of NK cells doesn't appear to inhibit the progression of inflammation in the islets.
The early deletion of NK cells in NOD.NK1.1 mice by depleting antibody treatment was used to determine the role of NK cells in disease initiation or their contribution as effectors. The adoptive transfer of only CD4+ T cells from diabetic NOD mice into Rag-deficient NOD hosts bypasses the need to prime and activate pathogenic CD4+ T cells and excludes participation of CD8+ T cells in induction of disease. Additionally, the time period to disease onset in this adoptively transferred CD4+ T cell model is similar to that of spontaneous disease in NOD mice. By depleting NK cells in the Rag-deficient NOD hosts, we used this sensitive in vivo assay to demonstrate that NK cells have no role in the effector phase of disease in this model. Thus, NK cells do not mediate islet destruction. Likewise, when NK cells were depleted in the model of spontaneous disease in intact NOD mice, no differences in BrdU incorporation by CD4+ T cells or CD8+ T cells were demonstrated. These data also support the finding that depletion of NK cells had no significant impact on the onset of spontaneous disease in NOD mice.
Targeting of either NKG2D or NKp46 in NOD mice prevents the onset of diabetes, but our results here suggest that these outcomes are not due to the targeting of NK cells. Indeed, NK cells can kill islets in vitro
, and islets express ligands for the activating receptors NKG2D and NKp46 
. NKG2D is expressed by activated T cells and was suggested as the target for anti-NKG2D therapy 
. The study that targeted the NKp46 pathway by treating mice with a NKp46-Fc fusion protein did not directly implicate NK cells by addressing the efficacy of this therapy in the absence of NK cells in NOD mice, an important consideration given that NKp46 is expressed on some γδ T cells, some non-NK innate lymphoid cells, and a subset of αβ TCR-bearing T cells identified in our study. It is also possible that a receptor other than NKp46 recognizes this NKp46 ligand and that NKp46-Fc therapy prevents that interaction. With respect to this study, the NKp46-Fc therapy does not target the same cells as anti-NK1.1 therapy. Anti-NK1.1 depletes NK cells and NKT cells. NKT cells have been shown to be protective in NOD mice if their activity is induced 
. NK1.1 depletion does not target NKp46+
cells in the intestinal lumen that produce IL-22, which is important for gut homeostasis 
. As noted above, IFNγ production by activated NK cells is protective from diabetes. The loss of NKp46 does appear to affect homeostasis of NK cells in the gut and their ability to produce IFNγ in response to IL-12 and IL-18. Gut homeostasis is directly linked to diabetes in NOD mice as MyD88-deficient mice only develop diabetes in germ-free conditions 
. Of note, the important NK cell activation factor IL-18 is not required for diabetes as MyD88 signaling is necessary to process functional IL-18.
A possible explanation for low NK cell function in NOD mice is that these mice have a poorly expressed Il15
. The rescue of NK cell activity in NOD mice by injecting IL-15 and IL-15R complexes dramatically accelerated disease onset in BDC2.5 NOD mice, suggesting that fully functional, activated NK cells may contribute to diabetes. Indeed, in our studies presented here the deletion of NK cells had modest, but not statistically significant, effects on the rate of spontaneous disease in NOD mice. However, significant delays in disease in NOD mice were demonstrated with transient depletion of either CD4+
T cells 
. Thus, the recruitment and activation of NK cells in the pancreas of diseased mice appears to be a consequence of the disease pathogenesis and not a required driver of disease. In some instances, these activated pancreatic NK cells might contribute enough to the inflammatory milieu to promote disease progression, which might explain why primed CD4+
T cells take a little longer to promote disease in the absence of NK cells and why there is a slight delay in the spontaneous development of diabetes in NOD mice. Our results do not support a direct role of NK cells in islet destruction or a significant role for NK cells in the priming of pathogenic T cells, but future studies might explore the potential role of NK cells in the recruitment of T cells into the pancreas (e.g. by production of chemokines) since the absence of NK cells during adoptive transfer demonstrated a clear delay in onset. Finally, our finding, combined with other recent studies, suggest that NK cells can play a role in diabetes in mice on the NOD background when manipulated, but that spontaneous disease ultimately does not require NK cells.