NKG2D Signaling Within the Pancreatic Islets Reduces NOD Diabetes and Increases Protective Central Memory CD8⁺ T-Cell Numbers

Type 1 diabetes is a T cell–mediated autoimmune disorder resulting in the destruction of insulin-producing pancreatic β-cells. β-Cell loss results in an inability of the body to produce insulin. Despite significant progress in type 1 diabetes research, the immune defects that lead to the development of this disease remain poorly understood. The immune receptor NKG2D, expressed by natural killer (NK) cells and subsets of T cells (1–5), is implicated in type 1 diabetes development. However, its role in disease progression remains unclear, with conflicting reports describing pathogenic (6,7), nonpathogenic (8), and protective (9) effects of NKG2D signaling.

NKG2D recognizes multiple NKG2D ligands in both humans and mice. These are endogenous ligands, which are all distantly related to MHC class I in sequence, and are generally believed to be functionally redundant (10,11). NKG2D ligands are considered stress ligands, with their expression induced by cellular stressors, including viral infection or DNA damage (10,11). Engagement of these ligands by NKG2D on NK cells is sufficient to activate NK cell killing of NKG2D ligand-bearing target cells (1). On T cells, NKG2D engagement generally costimulates T-cell receptor–driven activation, enhancing T-cell responses (1,7,9,12–14). In addition, NKG2D signaling has been shown to be important in the development of mouse memory CD8⁺ T cells (15–17).

Early studies with the NOD mouse model led to the hypothesis that NKG2D signaling, induced by NKG2D ligands expressed on β-islet cells, enhances diabetes development (6,7). However, conflicting results were later reported that brought into question the importance of NKG2D to spontaneous autoimmune diabetes in this preclinical model (8,9). Our previous data implicated differential effects of NKG2D signaling induced by microbiota or islet antigen as a source of this controversy (9). Therefore, new experimental approaches are required to determine the various roles of NKG2D in diabetes.

Here, we show that mRNAs encoding both NKG2D and NKG2D ligands are expressed in human islets, demonstrating that NKG2D signaling in the pancreas is likely relevant to type 1 diabetes pathogenesis. To directly test the effect of NKG2D signaling in the pancreas on autoimmune diabetes, we generated NOD mice with constitutive expression of the mouse NKG2D ligand RAE1ε on β-islet cells (i.e., NOD.RIP-RAE1ε mice). We found that despite earlier infiltration of immune cells into the pancreas, these mice developed significantly less diabetes compared with littermate NOD mice. This diabetes reduction corresponded with a reduced ratio of CD8⁺ effector T (T_eff) and effector memory T (T_em) cells (T_{eff + em}) to CD8⁺ central memory T cells (T_cm). Correlating with these in vivo data, we found that stimulation of NKG2D on NOD and human CD8⁺ T cells during in vitro activation resulted in a reduced CD8⁺ T_{eff + em}:T_cm ratio. Finally, we found that CD8⁺ T_cm cells actively suppress NOD diabetes development in vivo. Taken together, these results indicate that NKG2D ligand expression in pancreatic islets protects against autoimmune diabetes development by enhancing the generation or survival of a protective CD8⁺ T_cm population.

NOD/ShiLtJ (NOD), NOD.CB17-Prkdc^scid/J (NOD.Scid), and C57BL/6 (B6) mice were purchased from The Jackson Laboratory (stock numbers 001976, 001303, and 000664, respectively). B6.Cg-Tg(Ins2-cre)25Mgn/J (B6.RIP-cre) mice were purchased from The Jackson Laboratory (stock number 003573). B6.PCCALL-RAE1ε and ubiquitous B6.RAE1ε mice were generated previously (7,18) and provided by Dr. Andrey Shaw (Washington University School of Medicine, St. Louis, MO). The PCCALL-RAE1ε (PCCALL) allele and rat insulin promoter (RIP)-cre transgene were transferred to the NOD genetic background using speed congenic methods (19) by the Washington University School of Medicine Mouse Genetics Core. Both NOD.PCCALL and NOD.RIP-cre strains maintain >98% of the NOD genome, with no alterations in known Idd loci. Experimental NOD.PCCALL-RAE1ε, NOD.RIP-cre, and NOD.RIP-RAE1ε littermates of both sexes were generated by interbreeding NOD.PCCALL and NOD.RIP-cre mice. NOD.B6-Klrk1^tm1.1Bpol (NOD.Klrk1^−/−) mice were previously generated in our laboratory (9). All mice were housed under specific pathogen–free conditions at either the Washington University School of Medicine or the University of Kansas Medical Center animal facility in accordance with institutional guidelines and with approval from the institutional animal care and use committee.

Frozen tissue from cadaveric donors was provided by the Network for Pancreatic Organ Donors (nPOD) (20) with approval from the University of Florida Health Center Review Board. The characteristics of the donors used are shown in Supplementary Table 1. Tissue slides were fixed, and laser capture of islets was conducted as previously described (21). All islets in two to five sections of tissue from each donor (a minimum of 30 islets each) were captured and pooled, and RNA was extracted using the Arcturus PicoPure RNA Isolation Kit (Applied Biosystems, Grand Island, NY). An Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA) was used to assess quality and quantity of RNA. Samples with sufficient RNA quantity and quality were then subjected to gene expression analysis using Affymetrix expression arrays (GeneChip Human Gene 2.0 ST), and scaled normalized gene expression values were produced as previously described (22).

Insulitis was scored using standard methods (23). The pancreata were fixed in formalin, and 5 μmol/L sections were generated from the whole tissue and stained with hematoxylin-eosin by the Washington University School of Medicine Histology Core or the Kansas Intellectual and Developmental Disabilities Research Center Histology Core. Thirty islets per pancreas in nonoverlapping sections were scored.

Diabetes was determined by blood glucose measurements taken weekly through tail vein nick. A mouse was defined as diabetic on the date when the first of two consecutive blood glucose measurements ≥250 mg/dL was obtained, as previously described (24).

Spleens were dissociated by pressing through a 40-μm cell strainer in isolation buffer containing PBS, 2% FCS, and 2 mmol/L EDTA. Pancreata were chopped into small sections and digested with 1 mg/mL collagenase IV (Life Technologies) in Iscove’s modified Dulbecco’s medium for 10 min at 37° with shaking, and a single-cell suspension was generated by pressing through a 40-μm cell strainer. The cells were then washed at least five times, followed by antibody staining at 4°C for at least 15 min. After staining, cells were washed and analyzed. Cultured cells were washed once in isolation buffer and stained before similar staining and analysis.

Male and female mice were used, with all transfers containing male cells going only into male NOD.Scid recipients. Spleens and lymph nodes from nondiabetic 6- to 12-week-old NOD mice were harvested and dissociated by passing through a 40-μm cell strainer into isolation buffer containing PBS, 2% FCS, and 2 mmol/L EDTA. CD8⁺ T cells were then enriched by negative selection using magnetic bead separation (cat. no. 558471; BD Biosciences), according to the manufacturer’s instructions. CD8⁺ T cells were then stained with anti-mCD3-PE-Cy7, anti-mCD8-APC, anti-mCD44-BV650, and anti-mCD62L-BV510 antibodies (BD Biosciences). CD3⁺CD8⁺CD44⁺CD62L⁻ (T_{eff + em}) and CD3⁺CD8⁺CD44⁺CD62L⁺ (T_cm) populations were separated by FACS by the University of Kansas Medical Center Flow Core using a FACSAria IIIu (BD Biosciences). Separately, splenocytes from nondiabetic 6- to 12-week-old mice were depleted of CD8⁺ T cells by positive selection using anti-mCD8-PE (BD Biosciences) and anti-PE magnetic bead separation (cat. no. 557899; BD Biosciences); 2.5 × 10⁶ T_{eff + em}, 2.5 × 10⁶ T_cm, or a mixture of 2.5 × 10⁶ T_{eff + em} and 2.5 × 10⁶ T_cm were each combined with 1.5 × 10⁷ CD8⁺ T cell–depleted NOD splenocytes and adoptively transferred into 2- to 6-month-old NOD.Scid mice by retro-orbital injection.

All cells were grown in Iscove’s modified Dulbecco’s medium supplemented with 0.1 units/mL penicillin, 0.1 units/mL streptomycin, 0.29 μg/mL l-glutamine, 5.5 μmol/L 2-mercaptoethanol, and 10% FCS. Cells were grown at 37°C at 5% CO₂.

CD8⁺ T cells were isolated from the spleens and lymph nodes of 6- to 8-week-old NOD mice by negative selection using magnetic bead separation (cat. no. 558471; BD Biosciences), according to the manufacturer’s instructions. Separately, splenocytes from ubiquitous RAE1ε mice or control C57BL/6 mice were dissociated by passing through a 40-μm cell strainer and irradiated with 15 Gy using a ¹³⁷Cs irradiator. Then, 10⁶ CD8⁺ T cells were plated with either irradiated C57BL/6 RAE-1ε or C57BL/6 control splenocytes (3 × 10⁶) along with 1 μg/mL anti-mCD3 (clone 2C11) in a 24-well plate.

Total CD8⁺ T cells were isolated from human peripheral blood mononuclear cells by negative selection using a human CD8 T-cell enrichment set (BD Biosciences), according to manufacturer’s instructions. Twenty-four–well plates were incubated with 3 μg/mL anti-hCD3 (clone: OKT3) and anti-hCD28 (clone: 9.3), along with 20 μg/mL anti-hNKG2D (clone: 149810), or isotype control in PBS overnight at 4°. The plates were washed with PBS, and then the CD8⁺ T cells were added to the plate (10⁶ cells/well). After 3 days, cells were split 1:2 and maintained in the presence of 20 μg/mL plate-bound anti-hNKG2D. After 5 days of culture, the cells were harvested for analysis.

Data were collected and analyzed with GraphPad Prism (GraphPad Software). The log-rank (Mantel-Cox) test was used for analysis of diabetes incidence or human NKG2D ligand expression incidence. Differences between groups were compared using a two-tailed Mann-Whitney U test unless otherwise indicated. Linear regression analysis was performed to determine relationships in gene expression in human islets. P < 0.05 was considered to be significant and is marked in the figures. Results are shown as mean ± SEM unless otherwise indicated.

The primary data generated and analyzed during the current study are either included in the published article or are available from the corresponding author upon reasonable request. All resources generated during the current study, including all novel mouse strains, are available from the corresponding author upon reasonable request.

To determine whether studying NKG2D-ligand interaction within the pancreas is relevant to type 1 diabetes, we measured the expression of the mRNAs encoding NKG2D and NKG2D ligands within human pancreatic islets. Pancreatic islets were harvested by laser capture from tissue sections donated to nPOD by people with type 1 diabetes, people with islet antigen-specific autoantibodies but without diabetes (likely prediabetes), or age- and sex-matched control subjects without diabetes. Increased transcripts encoding both NKG2D and NKG2D ligands were detected in pancreatic islets harvested from donors with type 1 diabetes, but not individuals positive for islet-specific antibody, compared with donors without diabetes (Fig. 1). Specifically, the average mean intensity value for the mRNA encoding NKG2D was increased by 20% in the islets from patients with type 1 diabetes compared with those from donors without diabetes (P < 0.05, Mann-Whitney U test) (Fig. 1A). Additionally, we found a positive relationship between NKG2D expression and islet T-cell infiltration (Supplementary Fig. 1A). In the case of NKG2D ligands, 40% of donors with type 1 diabetes expressed levels of mRNA encoding of at least one ligand at a level determined by Grubb’s test to be an outlier from the values of the control subjects compared with 5.3% of control subjects without diabetes (P < 0.009) (Fig. 1B–J). We also found a positive relationship between NKG2D ligand expression and islet T-cell infiltration or NKG2D expression (Supplementary Fig. 1B and C). Together, these data indicate that NKG2D-ligand interactions occur within T cell–infiltrated islets in the pancreas of patients with type 1 diabetes.

We wished to directly test the effect of NKG2D ligand expression in the pancreas with an animal model. However, because both NKG2D and NKG2D ligands are expressed on immune cells (25–28), we could not eliminate these proteins specifically within the pancreas. We therefore tested this question by increasing NKG2D signaling within the pancreas, which also allowed us to directly test the hypothesis that NKG2D ligand expression on β-islet cells enhances diabetes (6,7). We moved a cre-inducible RAE1ε transgene we generated previously (PCCALL-RAE1ε) (7) and a RIP-cre transgene onto the NOD background. We then interbred these mice to generate NOD mice with constitutive expression of RAE1ε in the pancreatic islets (NOD.RIP-RAE1ε) (Fig. 2A–C).

Engagement of NKG2D by ligand results in not only signaling but also internalization of NKG2D from the cell surface (29). Therefore, to confirm that NKG2D signaling is enhanced in the pancreas of NOD.RIP-RAE1ε mice as it is in B6.RIP-RAE1ε (7), we compared NKG2D expression on CD8⁺ T cells in the pancreas of NOD.RIP-RAE1ε and single transgenic control NOD mice, NOD.RIP-cre. NKG2D expression was lower on CD8⁺ T cells in the pancreas of NOD.RIP-RAE1ε mice compared with NOD.RIP-cre (Fig. 2D). By contrast, NKG2D expression was similar within the spleen between the two strains (Fig. 2E). These data demonstrate that NKG2D signaling is enhanced specifically in the pancreas of NOD.RIP-RAE1ε mice.

To determine whether overexpression of RAE1ε in β-cells altered NOD islet-specific immunity, we compared diabetes development between NOD.RIP-RAE1ε (RAE1ε+ in islets) and control NOD.PCCALL (RAE1ε− in islets) mice. We observed a significant decrease in diabetes in the NOD.RIP-RAE1ε compared with the NOD.PCCALL littermates (Fig. 3B and C). Because of sexual dimorphism in the NOD model (30–32), a difference in diabetes incidence between male and female mice was expected (Fig. 3B and C). However, in both sexes, RAE1ε expression in the β-islet cells significantly reduced diabetes. NKG2D is the only known receptor to interact with RAE1ε. Therefore, we crossed the RIP-RAE1ε NOD mice to NOD mice genetically deficient in the gene encoding NKG2D (Klrk1) (9,33). Diabetes developed similarly in NOD.RIP-RAE1ε.Klrk1^−/− and NOD.PCCALL.Klrk1^−/− mice (Fig. 3D). This demonstrates that NKG2D was required for the reduced diabetes in NOD.RIP-RAE1ε mice.

To determine the mechanism by which diabetes is reduced in NOD.RIP-RAE1ε mice, we first compared insulitis development between NOD.RIP-RAE1ε and control littermates (NOD.PCCALL). Consistent with our earlier studies of the B6.RAE1ε mice (7), we observed increased insulitis in the NOD.RIP-RAE1ε mice compared with NOD.PCCALL mice at 9 weeks of age (Fig. 3E). However, at 12 weeks of age, insulitis was similar in NOD.PCCALL and NOD.RIP-RAE1ε mice (Fig. 3F). Therefore, we concluded that the decrease in diabetes development in the NOD.RIP-RAE1ε mice was not the result of reduced immune infiltration into islets.

We next compared the immune infiltrate in the pancreas of NOD.RIP-RAE1ε and control littermates (NOD.PCCALL and NOD.RIP-cre) at 12 weeks of age, which is a time at which these strains have similar insulitis and have not yet developed diabetes. We focused on T cells and NK cells because these are the cells on which we detect NKG2D in NOD pancreata and spleens (9). We found similar numbers of T cells, including CD4⁺ T and CD8⁺ T, as well as NK cells, in the pancreas (Supplementary Fig. 2A–D) and spleen (Supplementary Fig. 2E–H) of NOD.RIP-RAE1ε and control littermates (NOD.PCCALL and NOD.RIP-cre). However, NOD.RIP-RAE1ε mice had a lower percentage of a CD8⁺ T_eff or T_em phenotype (CD44^hiCD62L^lo) and a greater percentage of the CD8⁺ T_cm phenotype (CD44^hiCD62^hi) in the pancreas (Fig. 4A–C and Supplemental Fig. 5B), resulting in a decreased CD8+ T_{eff + em}:T_cm ratio compared with control littermates (NOD.PCCALL and NOD.RIP-cre) (Fig. 4D). The CD8⁺ T_{eff + em}:T_cm ratio was even lower in the spleen of NOD.RIP-RAE1ε mice (Fig. 4H–K and Supplemental Fig. 5B), and this difference was consistently observed in both male and female mice (Supplementary Fig. 3). Demonstrating that NKG2D signaling was required for this effect on CD8⁺ T-cell populations, no difference was observed between NOD.RIP-RAE1ε-Klrk1^−/− versus control littermate NOD.Klrk1^−/− (PCCALL/Klrk1^−/− and RIP-cre/Klrk1^−/−) (Fig. 4E, F, L, and M). In contrast to these CD8⁺ T-cell population alterations, we did not observe a difference in the CD4⁺ T_{eff + em}:T_cm ratio when comparing NOD.RIP-RAE1ε and control littermates (NOD. PCCALL and NOD.RIP-cre) (Fig. 4G and N and Supplemental Fig. 4).

We hypothesized that the greater number of CD8⁺ T_cm in the NOD.RIP-RAE1ε was directly responsible for the decreased diabetes development in these mice. To test this, we purified CD8⁺ T_{eff + em} (CD44⁺CD62L⁻) and T_cm (CD44⁺CD62L⁺) populations from nondiabetic NOD mice. We then adoptively transferred T_{eff + em}, T_cm, or T_{eff + em} and T_cm, along with CD8⁺ T cell–depleted NOD splenocytes, into NOD.Scid recipient mice (Fig. 5A). We measured blood glucose weekly, and time to diabetes was determined within each experiment relative to the first mouse to become diabetic. The mice that received the CD8⁺ T_{eff +} T_em cells developed diabetes significantly earlier compared with the ones that received T_cm or T_cm + T_{eff + em} (Fig. 5B). These results demonstrate that NOD CD8⁺ T_cm cells not only are less diabetogenic compared CD8⁺ T_{eff + Tem} but also actively suppress diabetes development.

NKG2D has been shown to be important in CD8⁺ T-cell memory formation in other mouse strains, especially CD8⁺ T_cm (15–17). We therefore hypothesized that NKG2D signaling directly in NOD CD8⁺ T cells was responsible for the decreased CD8⁺ T_{eff + em}:T_cm ratio in the NOD.RIP-RAE1ε mice. To test this, we purified total CD8⁺ T cells from wild-type NOD mice and activated them in vitro with splenocytes from B6.RAE1ε, which have ubiquitous transgenic expression of RAE1ε (18), or control B6 splenocytes. Similar to our observation with CD8⁺ T-cell populations in vivo in the NOD.RIP-RAE1ε mice (Fig. 4), NKG2D stimulation during in vitro activation with the B6.RAE1ε splenocytes resulted in a significantly reduced ratio of T_{eff + em}:T_cm NOD CD8⁺ T cells (Fig. 6).

We wanted to determine whether NKG2D stimulation of human CD8⁺ T cells similarly decreases the ratio of T_{eff + em}:T_cm CD8⁺ T cells. To do this, we purified total CD8⁺ T cells from human peripheral blood and activated these cells in vitro for 5 days in the presence of activating anti-NKG2D or control antibody. Similar to the mouse CD8⁺ T cells, we found that NKG2D stimulation during in vitro activation significantly reduced the ratio of human CD8⁺ T_{eff + em}:T_cm (Fig. 7).

NKG2D and NKG2D ligand expression has been reported in the pancreas of NOD mice (6,9). However, the expression of NKG2D and NKG2D ligands in the human pancreas had not been previously assessed. We demonstrate here that mRNAs encoding NKG2D and NKG2D ligands are detectable in human pancreatic islets, with increased expression in patients with type 1 diabetes compared with age- and sex-matched control subjects without diabetes, and positively correlate with T-cell infiltration into islets. These findings establish that NKG2D signaling likely occurs in the pancreas in inflamed islets during type 1 diabetes, supporting continued study of NKG2D signaling in the pancreas in diabetes pathogenesis.

NKG2D has been implicated in the development of type 1 diabetes. How it affects disease development and whether these effects are positive or negative have been a point of uncertainty. We previously showed a protective effect of NKG2D signaling on CD8⁺ T cells in NOD diabetes but found evidence that NKG2D signaling in the gut may have detrimental effects on disease development (9). These findings suggest that NKG2D signaling plays different roles in autoimmune diabetes progression in different anatomical locations. In this study, we endeavored to more closely assess the role played by NKG2D signaling within the pancreas in autoimmune diabetes.

Because both NKG2D and NKG2D ligands are expressed on immune cells (25), we could not eliminate these proteins specifically within the pancreas. Therefore, to test the effect of NKG2D signaling in the pancreas, we instead increased NKG2D signaling specifically within the β-islet cells of the pancreas. In our NOD.RIP-RAE1ε model, we selectively increased NKG2D signaling in the islets by transgenically expressing RAE1ε using the RIP. This expression resulted in earlier insulitis in the NOD.RIP-RAE1ε mice, which is consistent with our previous findings with B6.RIP-RAE1ε (7). However, despite this earlier increase, insulitis was similar in NOD.RIP-RAE1ε and control littermates by 12 weeks, and NOD.RIP-RAE1ε mice had reduced diabetes development. This diabetes reduction correlated with a decreased ratio of CD8⁺ T_{eff + em} cells to CD8⁺ T_cm cells_. Given the central role that CD8⁺ T_eff cells play in autoimmune diabetes pathology (34,35), it is not surprising that reduction in these cells correlated with a reduction in diabetes development. In contrast, the increase in CD8⁺ T_cm in NOD.RIP-RAE1ε mice was somewhat surprising. This increase is unlikely simply a reflection of the lower percentage of NOD.RIP-RAE1ε mice on their way to developing diabetes. If this was the case, we would have observed a similar increase in CD8⁺ T_cm in male NOD mice, which have a lower diabetes incidence, compared with female NOD mice, but we did not. Therefore, our data support the hypothesis that this increase in CD8⁺ T_cm does not just correspond to the reduced diabetes in both female and male NOD.RIP-RAE1ε but also is the cause for the RAE1ε-mediated reduction. This hypothesis is further supported by the results of the NOD.Scid adoptive cell transfer studies that demonstrated suppression of diabetes development by CD8⁺ T_cm.

The cell adoptive transfer studies revealed that CD8⁺ T_cm not only blocked diabetes onset but also delayed disease transfer by CD8⁺ T_{eff + em}. CD8⁺ T_{eff + em} cells correlate with islet pathology (36), and CD8⁺ T_eff cells play a central role in autoimmune diabetes pathology (34,35). In contrast, CD8⁺ T_cm cells have reduced cytotoxicity and effector functions (37), correlating with reduced diabetes transfer by these cells. However, the delay in diabetes we observed with cotransfer of CD8⁺ T_cm cells with CD8⁺ T_{eff + em} cells demonstrates a regulatory effect of the CD8⁺ T_cm population. Such a regulatory role for NOD CD8⁺ T_cm was suggested by previous studies performed by Santamaria and colleagues (38–40) using transgenic T-cell systems; however, this had not been investigated in parental NOD mice. T_cm cells can also be precursors to T_eff cells (41). Therefore, the protective cells are likely a subset of the CD8⁺ T_cm population. CD8⁺ T_cm cells are a heterogeneous population, which has been shown in other mouse strains to contain a subpopulation of immune regulatory CD8⁺ T cells (42). Future studies will need to be performed to determine the identity of the protective NOD CD8⁺ T_cm population, but our results demonstrate that this population is expanded by NKG2D signaling.

Our results indicate that the reduced T_{eff + em}:T_cm ratio in NOD.RIP-RAE1ε mice is due to enhanced NKG2D signaling directly in CD8⁺ T cells. We found that NKG2D stimulation during in vitro activation of wild-type NOD CD8⁺ T cells drove a similar decrease in the T_{eff + em}:T_cm ratio as observed in NOD.RIP-RAE1ε mice. This finding that NKG2D stimulation drives a shift toward CD8⁺ T_cm in the NOD mouse is consistent with a role for NKG2D in promoting memory cell development or survival, particularly T_cm, demonstrated by others with C57BL/6 mice (15–17,43). Importantly, demonstrating the translatability of the findings of these mouse studies to the human, we showed that NKG2D signaling during the activation of human CD8⁺ T cells also favors T_cm generation over T_em/T_eff. Our findings presented here were from studies performed with total NOD and human CD8⁺ T cells. In similar studies we performed beginning with naive CD8⁺ T cells, we did not observe this increase in T_cm. This is consistent with the findings of Wensveen et al. (15), who found that NKG2D signaling in C57BL/6 CD8⁺ T cells increased CD8⁺ T_cm production by promoting the survival of precursor memory cells rather than by affecting cell proliferation or differentiation. However, additional studies need to be done to confirm that the change in CD8⁺ T_{eff + em}:T_cm ratio in vivo is entirely dependent on NKG2D signaling directly in NOD CD8⁺ T cells, as well as when during the process of memory cell differentiation NKG2D signaling is important. Further, although our data demonstrate that constitutive high expression of an NKG2D ligand in pancreatic islet cells reduces NOD diabetes and the CD8⁺ T_{eff + em}:T_cm ratio, it still needs to be determined whether endogenous NKG2D ligand expressed in the pancreas (9) has similar effects.

Initial reports demonstrated that RAE1 family members were expressed by β-islet cells in NOD mice (6), that RAE1ε expression by β-islet cells enhanced the recruitment of CD8⁺ T cells to pancreatic islets (7), and that NKG2D inhibition reduced diabetes in both the NOD and the RIP-lymphocytic choriomeningitis virus mouse models of autoimmune diabetes alone or together with regulatory CD4⁺ T cells, respectively (6,44). This led to the hypothesis that inappropriate NKG2D ligand expression by β-islet cells contributes to autoimmune diabetes development or progression. However, we and others have been unable to detect RAE1 expression in the pancreas of NOD mice and do not detect NKG2D ligand expression by β-islet cells in our NOD colony (9,45,46). The reason for these differing reports of RAE1 expression in islets of NOD mice from different mouse colonies is unclear but may be the result of variation that is due to genetic drift or differences in microbiota composition. Regardless, the lack of RAE1 expression in the islets of mice in our NOD colony afforded us a model with which we could directly test the hypothesis that NKG2D ligand expression on β-islet cells promotes NOD diabetes. In contrast to the original hypothesis, our data demonstrate that RAE1ε expression by islet cells is protective against NOD diabetes and acts through interaction with NKG2D.

Although constitutive exposure to ligand can lead to NKG2D downregulation and dysfunction (47,48), our data are consistent with enhanced, not reduced, NKG2D signaling in pancreatic islets, resulting in reduced diabetes in the NOD.RIP-RAE1ε mice. While we observed reduced NKG2D expression on CD8⁺ T cells of NOD.RIP-RAE1ε mice in the pancreas, NKG2D expression by CD8⁺ T cells in the spleen was unaffected. This is indicative of increased local NKG2D signaling because NKG2D is internalized upon ligand engagement and signaling. Additionally, the reduced diabetes in NOD.RIP-RAE1ε mice correlates with the increased rate of diabetes in NKG2D-deficient NOD mice that we observed with treatment with gut microbiota-depleting antibiotics in previous studies (9). Finally, our in vitro data demonstrate that NKG2D signaling in CD8⁺ T cells enhances T_cm generation or survival, which correlates with the increase in CD8⁺ T_cm that is observed in NOD.RIP-RAE1ε mice. NKG2D is not expressed on mouse CD8⁺ T cells until 3–4 days following initial activation (9,14). This is a time when these cells start to migrate to the site of antigen (49). Taken together, these data support the conclusion that NKG2D signaling in islet-specific CD8⁺ T cells present in the islets during activation enhances T_cm generation or survival.

In conclusion, we found an increased presence of NKG2D and NKG2D ligand mRNA within the pancreatic islets of patients with type 1 diabetes and developed a novel mouse model, NOD.RIP-RAE1ε, to test the effects of NKG2D signaling specifically within the pancreas. We showed that increased NKG2D signaling within the pancreas of NOD mice had a protective effect, delaying diabetes in NOD.RIP-RAE1ε mice. This correlated with a decrease in the CD8⁺ T_{eff + em}:T_cm ratio in NOD.RIP-RAE1ε mice. We demonstrated that NKG2D signaling on CD8⁺ T cells drives a similar shift in both human and NOD CD8⁺ T-cell populations in vitro. Finally, we found that CD8⁺ T_cm cells transferred significantly less NOD diabetes than CD8⁺ T_{eff + em} cells and had a protective effect, delaying diabetes when cotransferred with CD8⁺ T_{eff + em} cells. We therefore propose a protective role for NKG2D signaling within the pancreas in autoimmune diabetes by increasing a protective CD8⁺ T_cm population relative to the more pathogenic CD8⁺ T_{eff + em} population. These findings reiterate the importance of NKG2D in autoimmune diabetes and further stress the importance of anatomical location when determining the role that NKG2D signaling plays in disease development. This is particularly critical when designing therapeutic strategies that target NKG2D or NKG2D-expressing cells, such as T cells.

Acknowledgments. Organ procurement organizations partnering with nPOD to provide research resources are listed at https://www.jdrfnpod.org//for-partners/npod-partners.

Funding. This research was performed with the support of nPOD (RRID:SCR_014641), a collaborative type 1 diabetes research project sponsored by JDRF (nPOD: 5-SRA-2018-557-Q-R), and The Leona M. & Harry B. Helmsley Charitable Trust (grant number 2018PG-T1D053). This work made use of the University of Kansas Medical Center Flow Cytometry Core Laboratory (supported in part by National Institute of General Medical Sciences [NIGMS] Centers of Biomedical Research Excellence grant P30-GM-103326 and National Cancer Institute Cancer Center Support grant P30-CA-168524) and the Kansas Intellectual Developmental Disabilities Research Center Histology Core (supported by National Institute of Child Health and Human Development grant U54-HD-090216). Funding for this work was provided by The University of Kansas Biomedical Research Training Program (to A.P.T. and K.L.K.), JDRF (JDRF 17-2012-595 to I.C.G. and C.E.M.), American Association of Immunologists (to N.S. and M.A.M.), the American Diabetes Association (1-12-JF-41 to M.A.M.), the Washington University School of Medicine Diabetes Research Center (National Institute of Diabetes and Digestive and Kidney Diseases [NIDDK] grant P30-DK-020579 to M.A.M.), the Molecular Regulation of Cell Development and Differentiation Center of Biomedical Research Excellence (NIGMS grant P20 GM104936/P30-GM-122731 to M.A.M.), and NIDDK grants UC4-DK-104155 (to I.C.G. and C.E.M.) and UC4-DK-104167 (to C.E.M.).

The content and views expressed are the responsibility of the authors and do not necessarily reflect the official view of nPOD.

Duality of Interest. M.A.M. is a consultant for Johnson & Johnson Global Services. No other conflicts of interest relevant to this article were reported.

Author Contributions. A.P.T., K.L.K., and M.A.M. designed and planned the experiments. A.P.T., K.L.K., N.S., I.C.G., and M.A.M. performed the experiments. A.P.T., K.L.K., N.S., C.E.M., and M.A.M. analyzed and discussed the data. A.P.T. and M.A.M. wrote the manuscript. All authors reviewed the manuscript. M.A.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this article were presented in abstract form at the 80th Scientific Sessions of the American Diabetes Association, 12–16 June 2020.