Research ArticleAUTOIMMUNITY

Suppression of diabetes by accumulation of non–islet-specific CD8+ effector T cells in pancreatic islets

See allHide authors and affiliations

Science Immunology  23 Mar 2018:
Vol. 3, Issue 21, eaam6533
DOI: 10.1126/sciimmunol.aam6533

Infiltration inhibition

Type 1 diabetes (T1D) is associated with the infiltration of islet-specific autoreactive cytotoxic CD8+ T cells (CTLs) in pancreatic islets, which leads to islet destruction and loss of insulin production. Most of the CTLs in islets are non–islet-specific, and their contribution to T1D is not well understood. Christoffersson et al. observed that the accumulation of these “bystander” CTLs is associated with decreased activation and proliferation of islet-specific CTLs. The abundance of non–islet-specific CTLs in islets reduced the accessibility of islet-specific CTLs to autoantigens, which led to a state of unresponsiveness. A similar form of nonspecific suppression by CTLs was observed in a viral meningitis model. Together, these results reveal an immune-regulatory role for nonautoreactive CTLs.


The inflammatory lesion at the pancreatic islet in type 1 diabetes (T1D) contains a heterogeneous infiltrate of T cells. In human and mouse studies, a large majority (98 to 99%) of the cytotoxic CD8+ T cells (CTLs) within islets are not specific to any islet antigen and are thought to passively add to tissue damage. We show by intravital confocal microscopy the opposite, immune-regulatory function of this cohort of CTLs. Diabetes did not develop in mice with islets showing high levels of infiltration of non–islet-specific CTLs not recognizing local antigens. Accumulation of such CTLs resulted in lower activation and proliferation of islet-specific CTLs, leading them to enter a state of unresponsiveness due to limited access to antigens at the inflammatory lesion. This nonspecific suppression by nonautoreactive CTLs was recapitulated in a model of viral meningitis, may explain viral interference in autoimmunity, and provides insight into the regulation of organ-specific autoimmune responses.


A hallmark of type 1 diabetes (T1D) is insulitis, the accumulation of an inflammatory infiltrate in the islets of Langerhans, leading to the subsequent destruction of the insulin-producing β cells. The infiltrate is populated by a wide range of immune cell subsets, including the main cytotoxic effector, the CD8+ T cell. The retention of these effectors is thought to be governed by the presence of local antigen (Ag) (1), whereas their entry into solid tissues is mediated by the activation and expression of chemokines (2). Current models of the insulitic lesion in the pancreas postulate the accumulation of autoreactive cytotoxic CD8+ T cells (CTLs). However, results from both human pathology samples (3) and mouse models (4) have shown that the level of islet Ag specificity can be as low as 1 to 2% among infiltrating CD8+ T cells, and although an autoreactive CTL clone is present in the body, it needs to reach a threshold quota of the total CTL repertoire to induce T1D (4). The importance of the non–islet-specific majority of CD8+ T cells present at the islets during autoimmune diabetes and their contribution to the course of the disease are unknown.

These nonspecific CTLs are present at inflammatory lesions, areas devoid of their cognate Ag, and they have been shown to be able to add to tissue damage through the nonspecific release of cytokines without local T cell receptor (TCR) engagement (5, 6). However, it is still unclear to what extent they affect the course of an inflammatory or autoimmune disorder. The dual nature of immune cell responses adds to the complexity of what is occurring within the inflammatory lesion. A balance between aggressive immune responses and counteracting suppressive responses appears to be critical to avoiding excessive and unresolved inflammation. Endogenous immune suppression by the immune system has been observed for decades (7), and CD8 expression on suppressor cells was associated with this response. Failure to identify a molecular basis for the suppressive effect of these cells brought studies to a halt. The field was revived by the discovery of CD4+/CD25+ suppressors, and since then the most studied type of suppressive immune response is that of the bona fide regulatory T cell (Treg) (8). However, with recent insights into the regulatory mechanisms used by CD8+ T cells, the interest in the immune-suppressive effects of these cells is reawakening (9, 10). These, in combination with the understanding that the immune system is extremely plastic and situation-dependent, have widened the preconceptions of immune-regulatory cells to not only include the classical CD4+ Tregs.

In this study, we have investigated the trafficking and influence of these “bystander” CTLs not recognizing local tissue Ags on the progression to T1D in different Ag-driven mouse models of this disease and one model of viral meningitis. We found that activated CTLs not recognizing local tissue Ags can elicit powerful suppressive effects when present in inflammatory lesions by limiting the local access to target cells and Ag-presenting cells for the target-specific effectors and, thereby, acted protectively against disease.


Ag governs CTL behavior at islets, and inflammation is required for their accumulation in larger numbers

T cells can gain access to sites of inflammation even though their cognate Ag is not expressed at that location, which is evident at the pancreatic islets in T1D (3, 4). We decided to investigate the requirements for the accumulation of non–islet-specific CD8+ T cells at pancreatic islets in well-defined Ag-driven models of autoimmune diabetes. Using mice expressing the lymphocytic choriomeningitis virus (LCMV) glycoprotein (GP) under the control of the rat insulin promoter (RIP), we adoptively transferred TCR-transgenic (TCR-tg) GP-specific P14 CD8+ T cells (fig. S1A). Transferred naïve P14 cells did not accumulate in the pancreas of RIP-GP mice but required stimulation through LCMV infection of the host or by GP33–41 (KAVYNFATC) peptide immunization to enter the islets and destroy the β cells (fig. S2). In the face of ongoing insulitis, we also investigated the trafficking of non–islet-specific OT-I CD8+ T cells to the islets. These cells are specific for chicken egg ovalbumin (OVA), do not recognize Ag in the RIP-GP mouse, and did not accumulate in the insulitic lesions in their naïve state. However, when they were activated by either recombinant LCMV-OVA virus (fig. S1A) or OVA257–264 (SIINFEKL) immunization (fig. S1B), they infiltrated and accumulated at the inflamed islets to a similar extent to and in the same pattern as their activated islet-specific P14 counterparts (Fig. 1A and movie S1). Simultaneous islet inflammation and the expression of an auto-Ag were requirements for the accumulation of both activated P14 and activated OT-I CD8+ T cells because, in wild-type (wt) C57Bl/6J mice, where neither the GP nor the OVA Ag is expressed on the islet β cells, no accumulation of transferred and activated T cells was observed (Fig. 1B and movie S2). The expansion of the transferred T cells was comparable between the wt and RIP-GP strains, as measured by their numbers in the spleen (Fig. 1, C and D). In groups where both transferred P14 and OT-I CD8+ T cells were activated and GP Ag was expressed on β cells, both populations of CTLs were found to specifically accumulate at the islets and not dwell in the exocrine parenchyma to any larger extent (Fig. 1, E to G). This highly focal accumulation of non–islet-specific CTLs observed at the islets was unexpected. This was also found to occur to the same extent in the RIP-mOVA mouse model where the antigenic situation is reversed (OVA is expressed on the β cells; fig. S1C): Non–islet-specific P14 CTLs accumulated at the islets in the presence of islet-specific OT-I CTLs. These data indicate that, irrespective of the model (RIP-GP or RIP-mOVA), the transferred non–islet-specific CTLs were found to have the same tissue distribution patterns as their islet-specific counterparts despite the complete lack of cognate Ag in the pancreas.

Fig. 1 Ag governs CTL behavior at islets but is not required for their accumulation in large numbers.

(A) Islet (dashed line) in the pancreas of a diabetic RIP-GP mouse where islet-specific P14 CTLs (red) and non–islet-specific OT-I CTLs (green) have gathered. (B) An islet (dashed line) in the pancreas of a wt C57Bl/6 mouse immunized in the same way as in (A) displaying a few CTLs in the pancreatic parenchyma. (C and D) The levels of T cell proliferation were similar between the two strains of mice, as judged by T cell content in spleens. (E to G) The densities of accumulated CTLs of both specificities (P14 and OT-I) were higher in the islets than in the surrounding pancreatic parenchyma, as quantified in (E) (n = 5 mice) and visualized in (F) and (G), where the yellow line across a part of the pancreas in (F) is represented in the histograms over fluorescent signals in (G). (H) Islets in mouse pancreata were imaged in anesthetized mice using a vacuum imaging window where the pancreas was immobilized, enabling long-term imaging. (I and J) A still frame (z-projection) (I) from an intravital recording of an inflamed islet in a RIP-GP mouse (red P14 CD8+ T cells and green OT-I CD8+ T cells) and the resulting T cell tracks (J). (B) and (C) correspond to movie S3. (K and L) OT-I CTLs consistently displayed higher migration speeds than the islet-specific P14 CTLs (K) and were kept, on average, further away from the islet center (L). Data in (J) to (L) are from one of six representative experiments. Data are means ± SEM. *P < 0.05, two-tailed unpaired Mann-Whitney U tests. Scale bars, 20 μm (A to D), 30 μm (F), and 50 μm (I and J).

We further analyzed the behavior of the two cell populations at the islets using intravital confocal microscopy of the pancreas (Fig. 1H) (11). Imaging was performed 7 to 9 days after adoptive transfer of TCR transgenic CD8+ T cells, when islet inflammation is at its peak in these models (Fig. 1I). Tracking the migration of the transferred CTLs revealed that islet-specific CTLs were moving substantially slower than their Ag-ignorant counterparts (Fig. 1, J and K). We also found that the non–islet-specific cells were, on average, more distant to the center of the islet, which was mostly occupied by the islet-specific CTLs (Fig. 1L). This migration pattern points to the notion that although Ag was dispensable for the accumulation of non–islet-specific CTLs at inflammatory foci, it governed the behavior of the CTLs in the microenvironment of the islet.

Abundance of activated non–islet-specific CTLs at islets is immune-suppressive

We found the OT-I transgenic CD8+ T cells to differ in their activation kinetics to their cognate Ag compared with the CD8+ T cells of the P14 transgene [also observed by others (12)]. The OT-I TCR-tg cells expanded several logs faster than the P14 TCR-tg in similar induction protocols (LCMV-OVA infection or SIINFEKL + KAVYNFATC immunization). This required optimization of immunization and adoptive transfer protocols to have quantifiable amounts of cells at the islets for imaging purposes. During the process of finding an appropriate balance between the numbers of the two TCR transgenes, we found that the transfer of equally high amounts (1 × 106 cells; ratio, 1:1) of P14 and OT-I CD8+ T cells (a scenario hereon denoted “HI”) to the RIP-GP model did not result in β cell death and hyperglycemia (Fig. 2, A to E) despite undergoing an intensive immunization protocol (fig. S1B). In these animals, we observed massive insulitis at days 7 to 9 after adoptive transfer (Fig. 2A) and the local accumulation of both P14 and OT-I CTLs, with particularly high amounts of the latter. However, when imaging the same group of animals 1 week later, both types of CTLs had retracted from the pancreas (Fig. 2B). Islets were mostly intact, and six of eight mice were normoglycemic (Fig. 2F). We then lowered the amount of transferred OT-I CD8+ T cells in logarithmic decrements and found that when 1 × 104 (1:100) or 1 × 103 (1:1000) cells were cotransferred with 1 × 106 P14 CD8+ T cells, diabetes incidence again became normal for this model (95 to 100%; Fig. 2E) (13). In contrast to the observations made in the HI scenario, mice that had received only 1 × 103 (1:1000) non–islet-specific OT-I CD8+ T cells (a scenario hereon denoted “LO”) also presented with insulitis on days 7 to 9, along with signs of β cell destruction (Fig. 2C). One week later, these animals were overtly diabetic (seven of eight mice were hyperglycemic; Fig. 2F), and very few β cells remained intact in these pancreata (Fig. 2D). In addition, in this scenario, the CTLs had largely retracted from the pancreas. Similar observations were made when animals, instead of being immunized by peptides, were infected with the recombinant LCMV-OVA virus to activate both TCR transgenes (fig. S1A).

Fig. 2 Abundance of non–islet-specific CTLs at islets is immune-suppressive.

(A and C) Adoptive transfer of equal amounts (1 × 106) of P14 and OT-I CD8+ T cells resulted in an abundance of non–islet-specific OT-I CTLs at islets on day 9 in RIP-GP mice (A) compared with when the ratio was changed to 1000:1 (P14:OT-I) (C). (B and D) Follow-up studies of the mice on day 16 showed that the infiltrate was gone, and in the HI (1:1 ratio) group, the islets were found to be mostly intact (B), whereas in the LO (1000:1 ratio) group, very few β cells were found (D). Scale bars, 30 μm. (E) A majority of the HI group (six of eight) displayed normoglycemia, whereas only one of eight was normoglycemic in the LO group (*P = 0.0058, 1:1 versus 1:1000, log-rank test). Different ratios of cell transfers suggested a dose-dependent mechanism because 50% of the mice in the group receiving a 10:1 ratio remained normoglycemic. (F to H) These scenarios were repeated in three different transgenic models that express model Ags on β cells: RIP-GP (six of eight were normoglycemic, representative of four independent experiments) (F), RIP-mOVA (five of six were normoglycemic, representative of two independent experiments) (G), and RIP-OVAlow (three of four were normoglycemic, representative of two independent experiments) (H).

To rule out the fact that the observations made in the RIP-GP mouse were not model-specific, we reversed the antigenic situation in the RIP-mOVA mouse (fig. S1C). We adjusted the numbers of transferred CTLs to accommodate the differences in antigenic reactivity between the P14 and OT-I transgenes. We found that 1 × 104 transferred OT-I CD8+ effectors in the RIP-mOVA model resulted in diabetes induction kinetics largely equivalent to that of 1 × 106 transferred P14 CD8+ T effectors in the RIP-GP model. Transferring 2 × 106 P14 CD8+ T cells (HI) alongside the OT-I CD8+ T cells in the RIP-mOVA model mimicked the protection from diabetes seen in the RIP-GP model (five of six mice were normoglycemic; Fig. 2G). A similar pattern (three of four mice were normoglycemic) was also seen in the RIP-OVAlow model (fig. S1D), where Ag is expressed at very low levels on the β cells and Ag is not cross-presented in pancreatic lymph nodes (Fig. 2H).

To rule out the possibility of adoptive transfer bias (12) or the idea that an abnormal accumulation of also islet-specific CTLs at the islets could have suppressive effects, we made single transfers of 2 × 106 and 1 × 107 P14 CD8+ T cells in the RIP-GP model. These mice displayed normal amounts of P14 CTLs in insulitic lesions and had similar diabetes induction kinetics to the normal scenario when 1 × 106 P14 CD8+ T cells were transferred (fig. S3). We were also interested to see whether an Ag-experienced but not highly activated non–islet-specific T cell population could home to the pancreas and induce protection. We therefore induced a memory phenotype of the OT-I CD8+ T cells in vitro by culturing the cells in the presence of SIINFEKL and interleukin-7 (IL-7). In contrast to their CTL counterparts, these cells did not accumulate in the islets or confer any protection against autoimmunity after transfer to the RIP-GP model, underlining the observation that a high level of activation of CTLs is required for the migration to the islets and the nonspecific interference with an Ag-driven immune response.

Protection from autoimmunity is not due to the expansion of Treg subsets

We next looked into the distribution of immune cells in the pancreas of the RIP-GP mouse during the HI and LO scenarios of adoptive transfer and induction of autoimmunity. By flow cytometry, the CD4+ T cell populations were similar in numbers between the groups, whereas, as expected, the CD8+ T cells were three times more abundant in the mice that had received high amounts of OT-I CD8+ T cells (Fig. 3, A to D). However, the P14 CD8+ T cells were significantly less frequent in the mice that had received high (HI) amounts of non–islet-specific OT-I CD8+ T cells (1 × 106 cells), despite receiving the same amount of islet-specific driver T cells (1 × 106 cells) as the LO group (Fig. 3E). Although this dimidiation of the amount of P14 CD8+ T cells in the pancreas might have contributed to a slower progression of the autoimmune process, the amount of islet-specific cells present at the islets in these mice should be more than enough to cause β cell demise and hyperglycemia. Looking into the distribution of the TCR transgenes and the endogenous CD8+ T cell populations in the pancreas, the high abundance of OT-I CD8+ T cells in the HI scenario left little space for other populations (~13%) compared with the transferred ones, whereas in the LO scenario, the space was taken up to ~87% by endogenous CD8+ T cells (Fig. 3, F to I).

Fig. 3 Protection from autoimmunity is not due to the expansion of Treg subsets.

(A to D) Flow plots showing similar representation of CD4+ T cells at islets in the two scenarios (A and B) but different amounts of CD8+ T cells (C and D). (E) Despite having lower amounts of total CD8+ T cells, LO mice had more islet-specific P14 T cells in the pancreas. (F and G) The distributions of Ag specificity of the CD8+ T cells present in the pancreas are visualized in (F) and (G), where each dot represents 1of 100. (H and I) Representative fixed, frozen tissue sections from HI and LO pancreata displaying the TCR-tg cells [red, P14 (islet-specific); green, OT-I (non–islet-specific)] and endogenous CD8+ T cells [blue, anti-CD8 monoclonal antibody (mAb) (polyclonal)] in islets (dashed lines). Scale bar, 20 μm. (J to L) CD4+/CD25+/FoxP3+ Treg fractions were not different between the two scenarios. (M and N) The abundant population of non–islet-specific T cells in the HI scenario had not transformed into a CD8+ Treg type as judged by the expression of FoxP3 (M) and CD122 (N). Groups are representative of at least three independent experiments. Data are means ± SEM. *P < 0.05, two-tailed unpaired Mann-Whitney U tests. MFI, mean fluorescence intensity.

An explanation to the protection from autoimmunity observed in the HI scenario could be the permissive expansion of a Treg population. We did not find any significant differences in the CD4+/CD25+/FoxP3+ subset of regulatory cells in the pancreas (Fig. 3, J to L) or pancreatic draining lymph nodes between the groups to explain the immune suppression. Markers for regulatory-like CD8+ T cells, such as FoxP3 (Fig. 3M) or CD122 (Fig. 3N) (14), did not show any differences in numbers or levels of expression between the HI and LO scenarios. Thus, differences in Treg subsets did not seem to be the factor limiting the expansion and actions of the islet-specific CD8+ T cells.

Accumulation of CTLs not recognizing local tissue Ag attenuates effector functions of islet-specific CTLs

We extended the profiling of the CD8+ T cells present in the pancreas to assess markers for their activation status, effector functions, and exhaustion (Fig. 4A). The non–islet-specific OT-I CD8+ T cells did not display any significant differences between the two scenarios. A few markers stood out for the islet-specific CD8+ T cells: P14 T cells in the HI scenario expressed significantly less CD69 (Fig. 4, B and C), consistent with their inability to execute their effector functions and kill Ag-expressing β cells (15). PD-1, an inhibitory receptor that is highly expressed on CD8+ T cells after exhaustion (16), was up-regulated on islet-specific cells in the HI scenario (Fig. 4, D and E). However, other hallmark inhibitory receptors for exhaustion [lymphocyte-activation gene 3 (LAG-3) and T cell immunoglobulin and mucin-domain containing-3 (TIM-3)] did not differ (Fig. 4A), although the decreased killer cell lectin-like receptor subfamily G member 1 (KLRG1) expression on islet-specific CD8+ T cells also points to an unresponsive phenotype (16).

Fig. 4 High numbers of non–islet-specific CTLs attenuate effector functions of islet-specific CTLs.

(A) Flow cytometric analysis of a range of surface markers on islet-specific P14 CD8+ T cells visualized as a heat map [based on arbitrary intensity units (MFI)]. Significant differences were found in CD69, KLRG1, and PD-1. (B and C) The expression of the activation marker CD69 on Ag-specific P14 CTLs was higher in the group receiving low amounts of non–islet-specific CD8+ T cells. (D and E) The exhaustion marker PD-1 was found to be highly expressed in the HI scenario (D), quantified in (E). Results are representative of two to three independent experiments. Data are means ± SEM. *P < 0.05, two-tailed unpaired Mann-Whitney U tests. (F to J) Islet-specific P14 CTLs isolated from pancreatic draining lymph nodes were stimulated in vitro with their cognate Ag GP33–41 and revealed less effector cytokine production (IFN-γ) [(F) and (G)] and increased IL-10 production [(H) and (I)] in mice receiving HI amounts of non–islet-specific TCR-tg CD8+ T cells; this is quantified in (J). *P < 0.05, two-tailed unpaired Mann-Whitney U tests. FSC, forward scatter.

To further determine the potential functional differences in the CTLs, we isolated T cells from the pancreatic draining lymph nodes of both HI and LO RIP-GP mice and stimulated them in vitro with KAVYNFATC (cognate Ag for P14) or SIINFEKL (cognate Ag for OT-I). Subsequent intracellular cytokine staining revealed that the P14 CD8+ T cells from the HI scenario produced lower amounts of interferon-γ (IFN-γ) (Fig. 5, C and D) and higher amounts of IL-10 (Fig. 5, E and F) than their counterparts in the LO scenario (Fig. 5G). We found no differences between the OT-I CD8+ T cells from the two scenarios. The apparent down-regulation of effector functions and the increased production of IL-10 in the islet-specific cells are symptomatic of exhausted T cells (16).

Fig. 5 Low accessibility to Ag for CTLs leads to low proliferation of effectors.

Intravital observations in the pancreas of RIP-GPxCX3CR1+/GFP mice from the HI and LO scenarios revealed less interactions between islet-specific P14 CTLs and GFP+ APCs in the HI scenario compared with the LO scenario. (A and B) Still frames from representative recordings for HI (A) (movie S4) and for LO (B) (movie S5). (C) Interactions are quantified as the arrest coefficient (fraction of time spent in a nonmoving state). Data are representative of groups of six mice per treatment. (D and E) APC occupancy by islet-specific CTLs and non–islet-specific CTLs was further analyzed in paraformaldehyde-fixed, frozen pancreas sections from RIP-GPxCX3CR1+/GFP mice receiving HI (D) or LO (E) amounts of non–islet-specific CD8+ T cells (n = 4 mice per group). Bars next to images display fractional occupancy of APCs by the indicated cell type. (F to H) The difference in APC occupancy could not be explained by changes in islet APC subsets, as shown in (F) for macrophages, in (G) for migratory DCs, and in (H) for plasmacytoid DCs (n = 3 to 4 mice per group). (I and J) The low level of APC interactions in the HI scenario likely led to a lower level of proliferation of effectors, as assessed by transfer of CellTrace Violet–stained Ly5.1+ P14 CD8+ T cells. n = 5 mice per group. Results are representative of two independent experiments. Scale bars, 10 μm. Data are means ± SEM. *P < 0.05, two-tailed unpaired Mann-Whitney U tests.

We further hypothesized that the high amount of CD8+ T cells that had gathered at the inflamed islets in the HI scenario could passively deplete the organ of inflammation-supportive and proliferative cytokines. To this end, we assessed the pancreatic content of 10 different cytokines to see whether increased CTL content could shift the environment in an organ to a less permissive and more immune-suppressive state. None of the assessed cytokines was differentially expressed between the HI and LO scenarios when assessing the whole pancreas (fig. S4) speaking against a depletion mechanism.

Low accessibility to Ag and professional Ag-presenting cells for effector CTLs leads to reduced proliferation of effectors

We sought to find the underlying reason for the P14 CD8+ T cells to display a nonresponsive phenotype (Fig. 4). Exhaustion occurs in situations such as chronic viral infections, malignancies, and autoimmunity where CTLs are exposed to high amounts of Ag for extended periods of time, while they are not capable of completing their task (i.e., target cell killing). In the current study, the P14 CD8+ T cells in RIP-GP animals were initially exposed to high amounts of Ag (1 + 1 mg of KAVYNFATC) and adjuvant (50 + 50 μg of CpG) during the immunization protocol (fig. S1B). What was apparent in the HI scenario was that the islet-specific T cells did not fully execute their effector functions and left most β cells unharmed. We performed intravital imaging in RIP-GPxCX3CR1+/GFP mice undergoing the HI and LO scenarios. Professional antigen-presenting cells (APCs) in the islets expressed green fluorescent protein (GFP) in this model and thereby enabled interrogation of the interactions between CTLs and APCs. In the in vivo recordings, it was evident that interactions between islet-specific CD8+ T cells and APCs were few in the HI scenario compared with the LO scenario. Although islet-specific P14 CD8+ T cells were fewer at the islets in the HI scenario (Fig. 3E), they were fractionally less engaged to APC than the islet-specific CD8+ T cells in the LO scenario (28 ± 17% versus 74 ± 11% occupancy, respectively; Fig. 5, A to C, and movies S4 and S5). By performing anti-CD8 immunofluorescence stainings of fixed, frozen pancreata, we could assess the APC occupancy by other CD8+ T cells. In the HI scenario, non–islet-specific T cells were found to occupy a majority of the APCs in the islets (61%; Fig. 5D) compared with the LO scenario where only 7% of the APCs were occupied by non–islet-specific CD8+ T cells (Fig. 5E). In both scenarios, a majority of the islet APCs were found to interact with CD8+ T cells, but a slightly larger proportion of the APCs in the LO scenario were unoccupied (HI versus LO, 8% versus 18%, respectively).

The phenotype and accumulation of APCs in islets can have a major impact on disease progression (17, 18). We therefore isolated islet APCs from pancreata from the two scenarios and performed flow cytometry, looking for differences in three major macrophage/dendritic cell (DC) subsets—resident macrophages (Fig. 5F), migratory DCs (Fig. 5G), and plasmacytoid DCs (Fig. 5H)—without finding any differences in these compartments.

The importance of cross-presentation of Ag to CD8+ T cells at the islets is unclear, but the T cell proliferative effect of Ag presentation at nonlymphatic peripheral sites of inflammation is becoming increasingly appreciated (1922). We therefore investigated whether the apparent blocking of CTL-APC interactions that occurred in the HI scenario affected the proliferation of islet-specific CD8+ T cells. By transferring 1 × 106 CellTrace Violet–loaded CD45.1+ P14 CD8+ T cells on day 6, we were able to track the proliferation of islet-specific cells during the onset of disease by assessing the dilution of dye on days 8 and 9 in pancreatic draining lymph nodes (Fig. 5, I and J). Conforming to the findings on the amount of P14 CD8+ T cells and their activation in the HI scenario (Fig. 3E), these cells also seemed to be restricted in their proliferation. Thus, the high amount of non–islet-specific CD8+ T cells at the insulitic lesion did not seem to be detrimental to the β cells but rather suppressed the autoreactive response partly by interfering with local Ag presentation, a conclusion that was further supported by the suppression observed in the RIP-OVAlow model (Fig. 2H), where islet Ag is not cross-presented in lymph nodes.

Non–viral Ag–specific CTLs dampen inflammation in a mouse model of viral meningitis

Intracerebral challenge of mice with LCMV results in infection of the meninges, leptomeninges, ependyma, and choroid plexus cells. Immunocompetent mice normally succumb to the infection about 7 days after inoculation due to excessive inflammation leading to edema and brain herniation (23). To explore the wider relevance of non–Ag-specific immune suppression, we infected mice intracerebrally with LCMV after receiving either 1 × 103 or 1 × 106 activated OT-I CD8+ T cells and followed their disease progression. Mice with induced memory to LCMV through previous intraperitoneal infection with the virus did not show any signs of disease, whereas mice receiving no cell transfers had a median survival of 7.25 days after intracerebral infection, as expected (Fig. 6A). The group of mice receiving 1 × 103 OT-I CD8+ T cells (corresponding to the LO groups in the T1D models) mimicked the disease kinetics of the group receiving no transfer and had a median survival of 7.5 days. The group of mice receiving a high amount of OT-I CD8+ T cells (1 × 106 cells corresponding to the HI groups in the T1D models) survived, on average, 3 days longer, with a median survival of 10.5 days, with no extended period of being moribund (Fig. 6A).

Fig. 6 Non–viral Ag–specific CTLs dampen inflammation in a mouse model of viral meningitis.

(A) Mice were inoculated intracerebrally with LCMV. One group (black line) had been infected with LCMV intraperitoneally 1 month before intracerebral infection and mounted a memory response, and mice were thereby protected. Mice receiving intracerebral LCMV without previous induced memory succumbed to the infection 7.25 days (median) after inoculation (red line). In cell transfer experiments similar to the HI and LO scenarios in the T1D models, 1 × 103 or 1 × 106 non–LCMV-specific OT-I CD8+ T cells were transferred and activated in vivo through peptide immunization. High amounts of OT-I led to a significant delay (P = 0.0006, 1 × 103 OT-I versus 1 × 106 OT-I, log-rank test) in time to death (n = 6 mice per group; data are representative of two independent experiments). Brains from the mice were examined 7 days after intracerebral infection. (B and C) No morphological differences or major differences in immune infiltrates could be detected between the groups receiving 1 × 106 (B) or 1 × 103 (C) OT-I CD8+ T cells (n = 4 mice per group). Scale bars, 1 mm. (D and E) Micrographs of periventricular areas in (D) and (E) represent the squares in (B) and (C) and have been stained using anti-CD8 mAb (gray), Hoechst for nuclei (blue), and OT-I cells expressing GFP (green). Scale bars, 20 μm. Bars next to the micrographs represent the clonal distribution of CD8+ T cells. (F) Overall numbers of CD8+ T cell clones per field of view in periventricular regions of brains from mice receiving transfers of 1 × 106 (HI) or 1 × 103 (LO) OT-I CD8+ T cells. *P < 0.05, two-way ANOVA.

In a separate experiment, we sacrificed animals from the two groups receiving different amounts of OT-I CD8+ T cells on day 7 after infection to evaluate the cerebral inflammation. In hematoxylin and eosin stainings of the brains, we could not observe any major dissimilarity in morphology or overall cellular infiltration between the groups (Fig. 6, B and C). When staining for CD8 in fixed, frozen brain sections, we found accumulation of lymphocytes in periventricular regions to the same extent in both groups (Fig. 6, D and E). However, when looking for GFP+ OT-I T cells, we found differences first in the accumulation of OT-I cells, which were more or less absent in the groups receiving 1 × 103 OT-I, but constituted more than 50% of the CD8+ lymphocytic infiltrate in the group receiving 1 × 106 OT-I (Fig. 6F).


In this study, we describe a mechanism of immune regulation by CD8+ T cells by which effectors not recognizing local tissue Ags were recruited to a site of inflammation, and their abundance at these sites limited tissue damage elicited by CD8+ T cells specific for islet or viral Ags. Non–islet-specific CTLs were found to efficiently, and with anatomic specificity, localize to the inflamed islets of Langerhans in models of T1D. When present in high amounts, these cells limited the autoaggressive response from the islet-specific effectors through nonspecific hindrance at the target tissue, resulting in reduced effector T cell proliferation, activation, and down-regulation of effector functions and, ultimately, reduced tissue destruction. We believe that the current study exposes an intrinsic local immune-regulatory mechanism that limits excessive immune responses. As judged by findings in autoimmune and other inflammatory lesions, the CD8+ T cell repertoires at these sites are heterogeneous with only a small fraction of them bearing TCR specificity to auto-Ags or viral epitopes.

T1D is characterized by immune infiltration of the pancreatic islets, resulting in progressive β cell decay and eventually onset of hyperglycemia (24). CD8+ T cells constitute the principal immune cell subset within insulitic lesions and have the potential to directly recognize and kill β cells (25, 26). Regardless of the molecular events that trigger T cell activation or convey cytotoxic functionality, it has been questioned whether TCR specificity against islet-related epitopes is a requirement for islet recruitment. Savinov and colleagues (1) showed that preexistent inflammation in the pancreas is dispensable for the migration of activated insulin-specific CD8+ T cells across the endothelium. Moreover, endothelial cells were found to actively cross-present islet Ags in the pancreas and thus mediate local T cell adhesion in an Ag-specific manner. In mouse models of allotransplantation, it was shown that, without preexisting inflammation, only islet-specific activated CTLs or memory CD8+ T cells were able to gain access to tissue through presentation of Ag by endothelial cells or intravascular APC dendrites (27). However, when inflammation was established by the islet-specific cells, non–islet-specific CD8+ T cells gained access through the leukocyte recruitment cascade involving signaling through chemokine receptors and integrins (27). In the current study, we show that islet auto-Ags are not required for non–islet-specific T cell accumulation but that local inflammation and T cell activation are critical to this response.

In the RIP-GP mouse model of autoimmune diabetes, only 1 to 2% of the CD8+ T cells in the islet infiltrate were specific to the GP Ag driving the disease (4). This abundance of activated “bystander” CD8+ T cells not recognizing islet Ags has led to hypotheses regarding their detrimental effect on β cells through the nonspecific release of cytokines without TCR engagement (5, 6). The current study does not argue against the possibility of bystander CTLs promoting β cell damage. Instead, we found here that the collective behavior of large groups of activated CTLs not receiving local TCR engagement does not have a disease-accelerating effect. The non–islet-specific CTLs did not differ largely in their expression of surface markers or effector functions dependent on whether they had been transferred to mice in high or low amounts. Neither did they express any evidence of converting into a specialized Treg type. Islet-specific CTLs were the only lymphocyte subset that displayed major differences after being cotransferred with high or low amounts of non–islet-specific CTLs. At the time point when hyperglycemia normally occurs in these models, the amounts of islet-specific CTLs in the pancreas were lower in mice receiving high amounts of non–islet-specific CD8+ T cells. They also displayed signs of less activation (low expression of CD69), less effector function (low expression of IFN-γ and high expression of IL-10), and nonresponsiveness (high expression of PD-1 and low expression of KLRG1). All these features thus resulted in the sparing of β cells and sustained normoglycemia in these animals.

We examined the cytokine environment in the pancreas, hypothesizing that the large influx of “uninvited” non–islet-specific cells would deplete the area of cytokines and metabolic factors, resulting in the impaired proliferation and function of effectors. This was not the case at the organ level because none of the 10 cytokines assessed was differentially expressed between the two scenarios.

Local interactions within the islets are crucial for the development of effector CD8+ T cells in T1D (28). Interactions between CTLs and APCs in islets during the onset of T1D have also been shown to increase effector functions and the production of IFN-γ (29). We used our imaging model to interrogate whether interference in these contacts could explain the decreased effector function when high amounts of non–islet-specific CD8+ T cells were present. Normally in the RIP-GP and RIP-mOVA mouse models, CTL-APC contacts are common in the insulitic lesion (29). In mice receiving high numbers of non–islet-specific CD8+ T cells, these contacts were fewer in vivo, and islet-associated APCs were occupied by a vast majority of transferred non–islet-specific TCR-tg cells or endogenous CD8+ T cells and not by islet-specific effectors. A “scanning” behavior of CTLs interacting with APCs has been described before, and this behavior was unrelated to Ag specificity or Ag presented by either part of the pair (30).

These findings may have relevance for several aspects of human disease. In human T1D, insulitis does not seem to occur as aggressively as in current mouse models. Along with other autoimmune disorders, T1D has a seemingly slow onset with the appearance of islet autoantibodies (and insulitis?) several years before clinical manifestation of the disease. In pancreata from autoantibody-positive organ donors without T1D, evidence for insulitis was found in only 10% of islets or less (31). This has led to hypotheses regarding an apparent relapsing-remitting pattern of the disease (32). A low-grade autoimmune response could therefore be interrupted by mechanisms similar to the one presented here by the expansion of large sets of antiviral CTLs to interfere at the insulitic lesion. Certain viral infections have been shown to induce pancreatic inflammation and be protective in mouse models of T1D (33, 34). Nonspecific inflammation through treatment with polyinosinic/polycytidylic acid [poly(I:C)] protects diabetes-prone biobreeding rats and nonobese diabetic mice from disease (35, 36). A recent study has also identified poly(I:C) to condition CD4+ T cells to express FoxP3 and induce a nonspecific bystander suppressor population in tumor and diabetes models (37). In a recent study, the depletion or functional impairment (CCR7-knockout, unable to reach lymph nodes) of Tregs in a mouse model of colitis led to the local induction of IL-10–producing CD8+ T cells that partly controlled inflammation (38).

A recent study found that T cell exhaustion plays an important role in the relapsing-remitting pattern of autoimmune pathologies where the “exhausted appearance” of the nonresponsive CD8+ T cells (PD-1hi, IFN-γlo, and IL-10hi) was found to be associated with poor clearance of chronic viral infections but conversely predicted better prognosis in multiple autoimmune diseases (39). Diversifying the CD8+ T cell repertoire at an autoinflammatory site and thereby limiting the access to costimulation for the target-specific effectors may be an intrinsic way for the immune system to self-regulate. This was underlined in the current study with the intracerebral LCMV infection model, where an otherwise aggressive course of disease was slowed by the entrance into the central nervous system by high numbers of a non–LCMV-specific CTL clone. The cause of death in this model is cerebral edema caused by the influx of myelomonocytic cells, which in turn is guided to the site by the secretion of chemokines by the CTLs at the site (40). These data are in line with what was seen in the T1D models in this study, in which islet-specific CTLs shifted their cytokine expression profile in the presence of high numbers of non–islet-specific CTLs. This may be part of the explanation as to why autoimmune attacks are seemingly interrupted and may also be a possible avenue for immune intervention by which nonspecific activation of the immune system could possibly dampen autoaggressive responses.

The antigenic situation in human autoimmunity is complex, with several auto-Ags directing the demise of the target tissue. In this study, we have used autoimmune models in which a single known Ag was driving the disease. Although we have used models with different driver Ags, these situations are still simplified. However, it is evident in human disease and experimental models that T cells lacking specificity to the target tissue will infiltrate inflamed regions (3, 4, 41). Combined with our data from a viral infection model, it seems plausible that similar events could also occur in human disease and partly explain viral interference and relapsing-remitting patterns in autoimmunity (34, 39).

In conclusion, this study presents a previously unknown mechanism of immune regulation in autoimmunity. The accumulation of effector CD8+ T cells not recognizing local tissue Ags at the pancreatic islets during the onset of T1D halted an aggressive immune response, and the animals went into remission with very little evidence of tissue damage. This may reflect inherent mechanisms in the immune system for controlling excessive inflammation and tissue damage and a possible avenue to explore for immune interventions in disease.



C57Bl/6J mice were purchased from the Jackson Laboratory. All other strains were bred in-house under specific pathogen–free conditions at the La Jolla Institute for Allergy and Immunology and kept on C57Bl/6J background. Transgenic RIP-LCMV-GP (Berlin strain) (42), RIP-mOVA (43), and RIP-OVAlow (44) mice have been described previously. TCR transgenic strains P14 (45) and OT-I (46) mice were bred on DsRed-, GFP-, or Ly5.1 (B6.SJL-Ptprca Pepcb/BoyJ)–expressing backgrounds. B6.129P-Cx3cr1tm1Litt/J mice (CX3CR1GFP/GFP) were bred with RIP-LCMV-GP mice. Female and male mice were used at 8 to 16 weeks of age. All animal experiments were approved by the Institutional Animal Care and Use Committee at the La Jolla Institute for Allergy and Immunology.

Adoptive cell transfers

Single-cell suspensions were prepared from the spleen and lymph nodes. Red blood cell lysis was performed, and CD8+ T cells were isolated using negative selection (Miltenyi Biotec). Isolated cells were transferred in single-cell suspensions into mice through retro-orbital injections.

Induction of autoimmune diabetes

Diabetes was induced in transgenic animals either by infection with 1 × 104 plaque-forming units of recombinant LCMV expressing OVA (LCMV-OVA) or by peptide immunization protocols: Twenty-four hours after the adoptive cell transfer, 1 mg of GP33–41 (KAVYNFATC, GenScript) and 100 μg of OVA257–264 (SIINFEKL, InvivoGen) in 50 μg of CpG adjuvant (Integrated DNA Technologies) were injected subcutaneously at the same site in the scapular area. The treatment was repeated 48 hours later. On the sixth day after the adoptive cell transfer, 500 μg of poly(I:C) (InvivoGen) was injected intraperitoneally (fig. S1). Blood glucose levels were measured using test strips (OneTouch Ultra, LifeScan). Mice with blood glucose levels above 250 mg/dl in two consecutive measurements were considered diabetic.

Intravital imaging

A previously described protocol for confocal and multiphoton microscopy of the mouse pancreas (47) was adapted to also incorporate the use of a custom-made immobilization device using vacuum [Fig. 1A and described in (11)]. Briefly, mice were anesthetized with an initial dose of ketamine hydrochloride (90 mg/kg) and xylazine hydrochloride (15 mg/kg), which was iterated as needed. A longitudinal incision was made in the left flank, and the vacuum device was lowered in place to immobilize the pancreas for imaging using a Leica HC Fluotar L 25×/0.95-W objective on a Leica SP5 confocal microscope. The pancreatic islets were in most experiments visualized using the reflected light from the 633-nm laser line.

Flow cytometry

All flow cytometry experiments were run on BD LSR II or Fortessa (BD Biosciences) instruments. Dead cells were discriminated using an Aqua Live/Dead staining kit (Life Technologies). All flow cytometry data were initially gated on lymphocytes>singlets in two steps>live cells>. Data were analyzed using FlowJo software (Tree Star).


Data are means ± SEM. Comparisons between the different groups were performed with two-tailed unpaired nonparametric Mann-Whitney U tests (two groups), two-way analysis of variance (ANOVA), or log-rank tests (specified in the respective figure legends). A P value of less than 0.05 was considered significant.



Fig. S1. Ag-driven mouse models of T1D.

Fig. S2. CTL activation and expansion is a requirement for pancreatic recruitment.

Fig. S3. Immune suppression by non–Ag-specific CD8+ T cells was not due to a “transfer bias.”

Fig. S4. No signs of differences in the cytokine environment in the pancreas.

Movie S1. Activated CTLs infiltrate islets irrespective of the presence of their cognate Ag.

Movie S2. No islet infiltration in the absence of cognate Ag and islet inflammation.

Movie S3. Differential behavior of islet-specific CTLs and non–islet-specific CTLs at islets.

Movie S4. High amounts of CTLs not recognizing local Ag reduces the number of interactions with APCs.

Movie S5. High amounts of CTLs not recognizing local Ag reduces the number of interactions with APCs.

Reference (48)


Acknowledgments: We are grateful to G. Aguila (La Jolla Institute for Allergy and Immunology) for expert mouse colony management and P. Colby (La Jolla Institute for Allergy and Immunology) for administrative assistance. Funding: G. Christoffersson was supported by a postdoctoral fellowship from the Swedish Research Council. This study was funded by the NIH. Author contributions: G. Christoffersson, G. Chodaczek, K.C., and M.G.v.H. conceptualized the study. G. Christoffersson, G. Chodaczek, S.S.R., and K.C. provided the methodology. G. Christoffersson, G. Chodaczek, S.S.R., and K.C. conducted the investigation. G. Christoffersson performed statistical analysis. G. Christoffersson and M.G.v.H. wrote the original draft of the manuscript. G. Christoffersson, G. Chodaczek, K.C., and M.G.v.H. wrote, reviewed, and edited the manuscript. M.G.v.H. supervised the study. Competing interests: M.G.v.H. and K.C. are employees of Novo Nordisk. All other authors declare that they have no competing interests.

Stay Connected to Science Immunology

Navigate This Article