Spontaneous Autoimmunity Sufficiently Potent to Induce Diabetes Mellitus Is Insufficient to Protect against Insulinoma¹

Most tumor Ags represent proteins derived from normal “self” proteins that are inappropriately expressed relative to the tissue of origin (1, 2, 3). Truly tumor-specific Ags are rare. Therefore, with some exceptions, tumors are not typically immunologically distinct. Recognizing that tumors are derived from self, inhibiting tolerance currently represents the major strategic focus of cancer immunotherapists (4). Adoptive immunotherapy and vaccination strategies each effectively bypass tolerogenic mechanisms with occasional benefits in clinical trials. Failures have largely been attributed to the emergence of Ag-loss variants (1, 2) and down-regulation of MHC class I (3, 5). To some degree, these problems can be avoided by targeting numerous tumor epitopes. If the autoreactive nature of the immune system is to be exploited against cancer, understanding the factors that limit effective tumor immunity despite a breakdown in tolerance would be useful. We considered that highly effective inhibition of tolerance would be sufficient to eradicate a tumor. To investigate this hypothesis, we created a double-transgenic murine model in which CD8⁺ T cells that recognize an Ag expressed by spontaneously arising insulinoma and by normal β-islet cells are pathologically primed. Mice were bred on a nonobese diabetic (NOD)⁴ background, associated with a predisposition to autoimmunity due to multiple defects in various tolerogenic mechanisms (6). The rat insulin promoter (RIP)-driven large T Ag (TAg) transgene (RIPTAg) confers a propensity to develop spontaneous insulinomas. The RIPTAg phenotype has been well characterized, and all insulinomas appear to progress through similar histological stages (7, 8, 9, 10). In individuals with the 8.3-TCR transgene, the vast majority of CD8⁺ T cells recognize residues 206–214 of the autoantigen IGRP (islet-specific glucose 6 phosphatase catalytic subunit-related protein), which is present on normal β-cells and on insulinoma cells. As a result of enhanced recruitment of CD8⁺ T cells to islets, mice with the 8.3-TCR transgene on the NOD background develop diabetes much earlier than NOD mice, although the cumulative incidence of diabetes at 32 wk is similar (11, 12). This model would enable the study of the interactions between pathologically primed diabetogenic CD8⁺ T cells and (normal or transformed) β-islet cells.

We demonstrate that a genetic predisposition to mount a vigorous autoimmune response does indeed provide protection against an autologous tumor bearing the targeted autoantigen. Nevertheless, this protection is incomplete, as the tumor eventually develops. The tumor microenvironment was thought to impede effective tumor immunity. These findings illustrate that even the best tumor vaccination strategies would not likely be sufficient to eradicate tumors. Rather, approaches targeting the microenvironment may be necessary to enhance successful tumor immunity.

8.3-NOD mice express the TCRαβ rearrangements of the H2K^d-restricted, β-islet cell-specific CD8⁺ T cell clone NY8.3. These mice have an accelerated onset of diabetes mellitus (DM) compared with NOD mice due to increased recruitment of diabetogenic CD8⁺ T cells to pancreatic islets (11). The NOD/Lt-TgN(RIP-TAg)1Lt-Prkdc-scid mouse (common name NOD-scid-RIPTAg) was provided by Dr. D. Serreze (The Jackson Laboratory, Bar Harbor, ME). NOD-RIPTAg mice, which have the normal complement of immune effectors, were generated by crossing NOD-scid-RIPTAG mice to NOD mice, eliminating the scid mutation. 8.3-NOD-RIPTAg mice were produced by crossing 8.3-NOD and NOD-RIPTAg mice. Such double-transgenic mice have tolerance to TAg (data not shown). The 8.3-TCR and RIPTAg transgenes were screened by PCR. Heterozygote scid mice were identified using PCR. Homozygotes were identified using the Ouchterlony assay. In some experiments, hypoglycemic mice were fed sucrose-containing water to ensure health in the later stages of insulinoma development. All mice were housed under specific pathogen-free conditions at the University of Calgary (Calgary, Alberta, Canada), and experimental procedures were approved by the University of Calgary Animal Care Committee in agreement with the Canadian Guidelines for Animal Research.

Blood glucose levels were measured 2–3 times weekly using a FastTake glucose monitor (Johnson & Johnson). Animals were considered diabetic after two readings ≥14 mM (hyperglycemia). The presence of insulinoma was marked by two readings ≤4 mM (hypoglycemia).

The following were purchased from BD Biosciences: PE-Vβ8.1,8.2 TCR (MR5-2), Cy-chrome-CD8α (53-6.7), allophycocyanin-CD8α (53-6.7), biotin-CD25 (7D4), biotin-CD44 (IM7), biotin-CD69 (H1.F23), streptavidin-Cy-chrome, FITC-IFN-γ (XMG1.2), FITC-CD107α (1D4B), PE-CD11c (HL3) and PE-H2Kd (SF1-1.1). A FITC-conjugated rat anti-mouse IgG1 Ab was purchased from Cedarlane Laboratories. Anti-CD8 was derived from Lyt2.2 hybridoma and anti-CD4 was derived from GK1.5 hybridoma. Anti-F4/80 was derived from hybridoma supernatant. Anti-MAC-1 was purified from the M1/70 hybridoma and biotinylated.

CD8⁺ T cells were isolated using Abs against CD4, MAC-1, F4/80, and CD45R (derived from the B220 hybridoma) for negative selection with Dynal beads. Cells were analyzed by flow cytometry using a FACScan (BD Biosciences) after conventional staining. Analysis software used included CellQuest (BD Biosciences) and FlowJo (Tree Star).

To assess cytokine production by 8.3-TCR CD8⁺ T cells, splenocytes or peripancreatic lymphocytes from 8.3-NOD and 8.3-NOD-RIPTAg mice were incubated in 24-well plates (2 × 10⁶ cells/ml/well) with 100 nM NOD-related peptide (NRP-A7; KYNKANAFL) in RPMI 1640 containing 5% FCS, glutamine (30 μg/ml), penicillin (100 U/ml), streptomycin (100 μg/ml), and 2-ME (5 × 10⁻⁵ M) (complete medium). Stimulation with 100 nM OVA peptide (SIINFEKL) served as a negative control. GolgiStop (BD Biosciences) was added for the last 5 h of culture. After 72 h, cells were stained for surface expression of CD8 and the 8.3-TCR, and intracellular IFN-γ was stained using the Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer’s protocol.

In vitro proliferation was measured using [³H]thymidine incorporation. Splenocytes or peripancreatic lymphocytes (1 × 10⁵) were incubated in flat-bottom, 96-well plates with the indicated concentrations of NRP-A7, anti-CD3 (145-2C11 supernatant), or OVA for 48 h. Isolated 8.3-TCR CD8⁺ T cells or nontransgenic CD8⁺ T cells (1 × 10⁵) were incubated with mitomycin C (Sigma-Aldrich)-treated tumor cells (1 × 10⁴) from 8.3-NOD-RIPTAg or NOD-RIPTAg mice for 72 h in flat-bottom, 96-well plates. In conditions that used dendritic cells, bone marrow was harvested from NOD mice and cultured for 7 days with 10% IMDM (Invitrogen) and 20 ng/ml GM-CSF (from the F10.9 hybridoma provided by C. Brown, University of Pittsburgh, Pittsburgh, PA). [³H]Thymidine incorporation assay was performed as previously described (1). At 16–18 h, cells were harvested.

In vivo proliferation of 8.3-TCR CD8⁺ T cells was monitored by evaluating the reduction of CFSE (Invitrogen)-staining intensity. Purified 8.3-TCR CD8⁺ T cells were labeled with 5 μM CFSE. Naive 8.3-TCR CD8⁺ T cells (10 × 10⁶) in 0.5 ml of HBSS were injected via the tail vein. Six days following adoptive transfer, recipients were sacrificed and donor cells were identified using flow cytometry. Results were expressed as percentage of division calculated using the formula M2/(M2 + M1) × 100, where M2 represents a gate on cells that had diluted CFSE and M1 represents a gate on cells that had not undergone division.

In vivo proliferation was also measured by evaluating BrdU uptake. 8.3-NOD-RIPTAg mice received i.p. injections of 9.2 μg/ml BrdU in 200 μl of PBS. Mice were sacrificed 24–31 h after injection, and lymphocytes were isolated from spleen, lymph node (LN), thymus, and tumor. Following surface staining, an intracellular staining protocol (13) was used and cells were stained with either FITC-anti-BrdU or FITC-conjugated mouse IgG1 isotype control (BD Biosciences). BrdU uptake by T lymphocytes was evaluated by flow cytometry.

RIPTAg tumors >2 mm in diameter were morselized in 3 ml of type V collagenase (470U/ml; Sigma-Aldrich). Erythrocytes were lysed using an ammonium chloride lysis solution. The flasks were rinsed with PBS to remove nonadherent cells. Adherent cells were stained with 0.01 μg/ml 7-aminoactinomycin D and stained for MAC-1, CD11c, and H2K^d.

CTL were developed by isolating splenocytes from 8.3-NOD mice and stimulating the cells with NRP-A7 (1 μg/ml) in complete medium with 10% FCS for 3 days. CTLs (20 × 10⁶) were injected i.v. into 8.3-NOD-RIPTAg and NOD-RIPTAg mice. Blood glucose levels were followed daily, and booster injections of 20 × 10⁶ CTLs were given weekly.

Freshly removed pancreata were cut longitudinally. If tumors were present they were also bivalved. The degree of inflammation was evaluated on formalin-fixed paraffin-embedded samples after staining with H&E (Sigma-Aldrich). Islets were scored for insulitis over five serial sections, as previously described (12).

IGRP, insulin, and cell differentiation markers were stained on 6-μm sections from frozen blocks. Sections were fixed using equal volumes of 100% methanol and acetone. Avidin and biotin were blocked (Vector Laboratories). Serial sections were stained using the following dilutions: IGRP (1/800 dilution of polyclonal rabbit anti-mouse IGRP provided by J. Hutton (University of Colorado, Denver, CO) (14); insulin (1/25 dilution guinea pig anti-swine polyclonal Ab; DakoCytomation); CD8 (neat Lyt2.2 supernatant); CD4 (neat GK1.5 supernatant); and MAC-1 (1/5 dilution of M1/70 supernatant). Primary Abs were detected using 1/200 dilutions of biotinylated rabbit anti-rat IgG for CD4, CD8, and MAC-1 and biotinylated goat anti-rabbit IgG for insulin, followed by incubation with HRP conjugated to streptavidin (Zymed Laboratories). Endogenous peroxidase activity was blocked using 3% H₂O₂/Tris-buffered saline. Color was developed using diaminobenzidine tetrachloride, and counterstaining was done with hematoxylin.

Lesion size was evaluated using Openlab software (Improvision) by measuring the perimeter of lesions that had been stained for insulin. (All IGRP-positive lesions were confirmed in serial sections to be insulin positive.) Lesion size was evaluated using a ×4 objective. Those lesions that were <10% of the field of view were defined as “normal” islets, whereas those that were >10% of the field of view were “intermediate” lesions. Those that exceeded the field of view were classified as “large” tumors. Infiltration by CD8⁺ T cells, CD4⁺ T cells, or macrophages was evaluated using a scoring system similar to that used to score insulitis. The observer was blinded to information on mouse age and genotype.

All values are expressed as mean ± SEM. Deviations from Gaussian distribution were tested using the Kolmogorov-Smirnov test. The differences between two means were tested by the two-tailed Student t test for two independent samples. The Mann-Whitney U test was used for nonparametric data. Categorical data were evaluated using the Fisher exact test. The time to hypoglycemia was evaluated using Kaplan-Meier analysis. The log-rank test was used to test for differences in incidence of hypoglycemia between groups. It was decided a priori that p ≤ 0.05 would be considered statistically significant.

In the absence of the 8.3-TCR transgene, the RIPTAg transgene had little influence on the inflammatory response, as there was no difference between the degrees of inflammatory infiltrate in islets from NOD mice and age-matched NOD-RIPTAg mice. 8.3-NOD-RIPTAg mice developed insulitis that was generally more severe than that of age-matched NOD-RIPTAg mice (Fig. 1,A). Infiltration of normal islets by CD8⁺ T cells at 13–14 wk and at >14 wk was significantly more severe in 8.3-NOD-RIPTAg islets compared with NOD-RIPTAg islets (Fig. 1 B). These trends were also observed for CD4⁺ lymphocytes at these age groups (data not shown). This was consistent with the more rapid onset of CD8⁺-mediated autoimmune DM seen with other models expressing the 8.3-TCR transgene (11).

To determine whether pathologically primed autoreactive CD8⁺ T cells protect against a tumor expressing the targeted autoantigen, tumor incidence was compared in 8.3-NOD-RIPTAg and NOD-RIPTAg cohorts. NOD-scid-RIPTAg mice served as a baseline control. Blood glucose was measured beginning at 3 wk of age and continued until the time of hypoglycemia. Of 52 8.3-NOD-RIPTAg mice, 12 (23%) became diabetic and died before the end of the observation period; 11 (21%) became diabetic before becoming hypoglycemic, and 29 (56%) were euglycemic before developing hypoglycemia. 8.3-NOD-RIPTAg mice that did not die before developing insulinoma had a delayed onset of hypoglycemia compared with NOD-RIPTAg and NOD-scid-RIPTAG cohorts (Fig. 1 C). The median time to hypoglycemia was 125 days for 8.3-NOD-RIPTAg mice compared with 90 and 94 days for NOD-RIPTAg and NOD-scid-RIPTAG mice, respectively (p < 0.0001). Despite the delay in tumor emergence in 8.3-NOD-RIPTAg mice, ultimately 100% of individuals became hypoglycemic. Therefore, 8.3-NOD-RIPTAg mice were incompletely protected from a tumor.

To investigate whether diabetogenic 8.3-TCR CD8⁺ T cells could inhibit tumors in the absence of T cells displaying endogenous TCRs, we introduced the 8.3-TCR and RIPTAg transgenes into RAG2^−/− mice, which cannot rearrange endogenous TCR or Ig genes. 8.3-NOD-RIPTAg.RAG2^−/− (n = 9) were not protected against tumors; median time to hypoglycemia was 91 days, which was not significantly different from NOD-RIPTAg.RAG2^−/− mice (n = 13; median time to hypoglycemia was 89 days) or other mice not bearing the 8.3-TCR transgene (data not shown). This is consistent with the previous observation that RAG2^−/− 8.3-NOD mice developed DM much less frequently and significantly later than RAG2⁺ 8.3-NOD mice (12). In that instance, the diabetes deceleration in RAG2^−/− 8.3-NOD mice could be reversed by adoptive transfer of CD4⁺ T cells, suggesting the need for CD4⁺ help for the diabetogenicity of 8.3-TCR CD8⁺ T cells.

We considered that overt DM represented a surrogate marker for a more exuberant autoimmune response. Therefore, we postulated that the emergence of overt DM before insulinoma provides enhanced protection against insulinoma. However, there was no significant difference in the incidence of hypoglycemia in diabetic mice compared with nondiabetic mice (Fig. 1 D). Therefore, the development of overt DM did not confer additional protection against insulinoma.

It was possible that the high blood glucose levels associated with DM masked the development of small insulinomas. To evaluate this possibility, mice were sacrificed at 100 days and examined for the appearance of gross tumor regardless of blood glucose level. 8.3-NOD-RIPTAg mice had a reduced prevalence of insulinoma at 100 days of age compared with NOD-RIPTAg mice (69 vs 95%, respectively; p = 0.002). This confirmed the initial observation that the 8.3-TCR transgene was protective. Moreover, there was no significant reduction in the prevalence of gross tumor in 8.3-NOD-RIPTAg mice that had developed overt DM before insulinoma (data not shown). Altogether, these data show that insulinoma develops despite an immune response that is sufficiently potent to induce the destruction of normal β-islet cells.

It is possible that peripheral tolerogenic mechanisms, possibly related to the tumor itself, impeded a sustained immune response against a tumor. In other experimental models, tumor-specific T cells become tolerant to tumor Ags due to clonal anergy (15, 16). Prolonged T cell anergy can lead to activation-induced cell death, resulting in peripheral deletion of tumor-specific T cell populations (17, 18, 19). Thus, 8.3-TCR CD8⁺ T cell anergy and clonal deletion were evaluated as potential mechanisms for incomplete protection against insulinoma in 8.3-NOD-RIPTAg mice.

The absolute numbers of T cells present in secondary lymphoid tissue were determined in 8.3-NOD-RIPTAg mice as they aged. There was no significant difference in the absolute number of 8.3-TCR CD8⁺ T cells derived from spleens or peripancreatic LNs of 8.3-NOD-RIPTAg mice compared with age-matched 8.3-NOD mice (Fig. 2 A). To evaluate the possibility that thymic output masked deletion, we monitored in vivo persistence of 8.3-TCR CD8⁺ T cells following adoptive transfer. We injected 10 × 10⁶ CFSE-labeled 8.3-TCR CD8⁺ T cells into mice aged 13–14 wk. In NOD-RIPTAg mice 6 days after adoptive transfer, an average 8.74 × 10⁵ 8.3-TCR CD8⁺ T cells were recovered from spleen, peripancreatic LN, and peripheral LN. In NOD mice, 8.23 × 10⁵ 8.3-TCR CD8⁺ T cells were recovered, which was not significantly different from the results seen in NOD-RIPTAg mice. Therefore, there is no evidence of CD8⁺ T cell clonal deletion secondary to tumor.

The functional capacity of the 8.3-TCR CD8⁺ T cell population was then examined. The proliferative response against increasing doses of the 8.3-TCR peptide ligand NRP-A7 was assessed by measuring [³H]thymidine incorporation. 8.3-TCR CD8⁺ T cells derived from peripancreatic LNs from tumor-bearing 8.3-NOD-RIPTAg mice responded to NRP-A7 with similar kinetics and magnitude as those derived from tumor-free 8.3-NOD-RIPTAg mice and prediabetic 8.3-NOD controls (Fig. 2 B). Similar results were observed for splenic 8.3-TCR CD8⁺ T cells (data not shown). Thus, a tumor does not impair the proliferative capacity of 8.3-TCR CD8⁺ T cells systemically (i.e., in spleen) or regionally (i.e., in peripancreatic LNs). Moreover, because proliferative responses to Ag are seen in whole LN and spleen (data not shown), it is unlikely that the responsiveness of 8.3-TCR CD8⁺ T cells is due to the removal of a suppressive cell population during CD8⁺ T cell isolation.

Rather than inhibiting the 8.3-TCR CD8⁺ T cell population as a whole, tumor progression may have inhibited individual T cells. Upon stimulation with NRP-A7, there was no difference in the proportion of IFN-γ secreting 8.3-TCR CD8⁺ T cells derived from spleens or peripancreatic LNs of tumor-bearing 8.3-NOD-RIPTAg mice compared with those of their 8.3-NOD counterparts (Fig. 2 C). Thus, the proportion of CD8⁺ 8.3-TCR T cells that produce IFN-γ in response to peptide stimulation is not influenced by the presence of tumors.

Overall, these experiments demonstrate that the function of peripheral 8.3-TCR CD8⁺ T cells was not impaired as insulinoma developed. The in vitro response by CD8⁺ T cells derived from 8.3-NOD-RIPTAg mice with established tumor was comparable to that of T cells derived from tumor-free 8.3-NOD-RIPTAg mice, and 8.3-NOD controls.

Tumors may have persisted secondary to immunological ignorance. Ignorance in the context of tumor immunity may occur because CD8⁺ T cells are exposed to insufficient levels of tumor Ag (20, 21, 22). Although recognition of untransformed β-cell Ag resulted in CD8⁺ T cell activation (as indicated by the emergence of DM), the elaboration of insulinoma Ags may have been insufficient for continued cross-presentation and activation of CD8⁺ T cells in secondary lymphoid tissue.

To determine whether insulinoma escaped the CD8⁺ T cell response due to ignorance, naive 8.3-TCR CD8⁺ T cells were labeled with CFSE and adoptively transferred into NOD-RIPTAg recipients. Donor cells were identified in spleen, peripancreatic LNs, and peripheral LNs (i.e., axillary and inguinal LNs). No division was observed in cells that had trafficked to the spleen (data not shown) and peripheral LNs (Fig. 3, A and B). Interestingly, CD8⁺ T cell proliferation in the peripancreatic LNs was significantly increased in NOD-RIPTAg recipients across all ages compared with NOD controls (Fig. 3,B). Nontransgenic CD8⁺ T cells were adoptively transferred in a similar fashion. These cells did not proliferate in peripancreatic LNs (Fig. 3 A), demonstrating that the proliferation seen in adoptively transferred 8.3-TCR CD8⁺ T cells was Ag dependent.

Ignorance is therefore not responsible for tumor progression in this model. Moreover, naive 8.3-TCR CD8⁺ T cells are cross-primed and are capable of proliferating in peripancreatic LNs in response to Ags derived from normal as well as transformed β-islet cells. We speculate that the enhanced proliferation seen in NOD-RIPTAg mice is secondary to increased Ag load, although this does not completely exclude other variables such as the elaboration of tumor-derived factors that support T cell proliferation.

Dysfunctional Ag presentation is well documented as a mechanism of tumor escape, arising secondary to a reduction in MHC class I expression or to the generation of Ag loss variants (1, 23, 24). Therefore, RIPTAg insulinoma cells were evaluated for their suitability as targets for CTL lysis. As shown in Fig. 4,A, 8.3-NOD-RIPTAg tumor cells retain H2K^d expression. There was no significant difference in mean fluorescence intensity in 8.3-NOD-RIPTAg tumors (n = 5) compared with NOD-RIPTAg tumors (n = 6) (data not shown). Both islets and intermediate size tumors from 8.3-NOD-RIPTAg mice retain IGRP expression (Fig. 4 B). In fact, 100% of lesions, including normal islets and intermediate size lesions, were IGRP positive at 13–14 wk of age.

To confirm that 8.3-TCR CD8⁺ T cells recognize tumors derived from established 8.3-NOD-RIPTAg mice, tumors were cocultured with CD8⁺ T cells isolated from 8.3-NOD spleen or NOD spleen and with dendritic cells grown from NOD bone marrow. [³H]Thymidine uptake was greater in tumors cocultured with 8.3-CD8⁺ T cells compared with nontransgenic CD8⁺ T cells (p = 0.014; Fig. 4 C). This is consistent with our observation that adoptively transferred naive 8.3-TCR CD8⁺ T cells divide in response to tumor in an Ag-dependent manner, even in older 8.3-NOD-RIPTAg mice.

Finally, the ability of 8.3-TCR CD8⁺ T cells to eliminate tumor cells from 8.3-NOD-RIPTAg mice with established tumors was evaluated. Activated 8.3-TCR CTLs were adoptively transferred into hypoglycemic 8.3-NOD-RIPTAg mice. Following CTL administration, blood glucose rapidly increased before returning to hypoglycemic levels a few days later (Fig. 4 D). Upon receiving a booster injection of CTL, blood glucose levels in 8.3-NOD-RIPTAg mice again increased to normoglycemic levels before decreasing. This demonstrated that RIPTAg tumors, even in 8.3-NOD-RIPTAg mice, remained targets for freshly activated CD8⁺ T cells, although the inhibitory activity of these CTLs was temporary.

In hypoglycemic 8.3-NOD-RIPTAg and NOD-RIPTAg mice 16–31 wk of age, we analyzed the inflammatory infiltrate separately in small, intermediate, and large lesions. It was considered that this would illustrate differences in the inflammatory microenvironments in normal islets vs tumors. There was no significant difference in the severity of inflammation in islets of NOD-RIPTAg vs 8.3-NOD-RIPTAg mice (Fig. 5 A). However, a higher proportion of intermediate size lesions in 8.3-NOD-RIPTAg mice had scores reflecting severe inflammation (i.e., grade 3 or 4) than in NOD-RIPTAg mice (p < 0.0001). Thus, the 8.3-TCR transgene was associated with increased leukocyte recruitment to intermediate size lesions. Interestingly, this result was not observed in larger tumors. Large lesions in both 8.3-NOD-RIPTAg and NOD-RIPTAg mice typically had a paucity of infiltrating leukocytes, particularly in their central components. Therefore, the inflammatory infiltrate is most pronounced in intermediate lesions, yet large lesions contain inflammatory cells only in their periphery.

Infiltration of islets and tumors by CD8⁺ and CD4⁺ lymphocytes was also analyzed in mice with established tumors (16–31 wk of age). 8.3-NOD-RIPTAg mice had a higher proportion of intermediate size lesions with severe CD8⁺ T cell infiltration (i.e., grade 3 or 4) (p = 0.0013), whereas NOD-RIPTAg mice had a higher proportion of intermediate size lesions with mild infiltration (i.e., grades 0–2) (Fig. 5,B). CD4⁺ T cell infiltration was also markedly increased in 8.3-NOD-RIPTAg mice (data not shown). Therefore, the 8.3-TCR transgene conferred a vigorous T cell-mediated response against intermediate size adenomas. Remarkably, in 8.3-NOD-RIPTAg mice that had more advanced insulinomas (i.e., 16–31 wk of age), normal islets and intermediate size lesions remained densely infiltrated by CD8⁺ T cells, yet CD8⁺ T cell infiltration in large tumors from both 8.3-NOD-RIPTAg (Fig. 5,C) and NOD-RIPTAg (data not shown) mice was minimal. In fact, 100% of large tumors in RIPTAg mice with and without the 8.3-TCR transgene were <25% infiltrated. The same was noted with CD4⁺ T cell infiltrate (Fig. 5 C). Importantly, in these larger lesions CD8⁺ and CD4⁺ T cells were located primarily around the perimeter of the tumor, with a paucity of both T cell subsets in their middle.

Although the above data suggest that tumors are recognizable by 8.3-TCR CD8⁺ T cells, it is possible that the T cells recruited to the lesion no longer recognized tumors and therefore ceased to proliferate at the tumor site. To test for this, BrdU uptake in tumor-infiltrating T cells was studied in animals that had been persistently hypoglycemic while ingesting sucrose water (aged 20–24 wk). 8.3-TCR CD8⁺ and CD4⁺ T cells in these tumors readily incorporated BrdU but nontransgenic CD8⁺ T cells did not (Fig. 5 D), suggesting that the CD8⁺ T cells contained within tumors proliferated in an Ag-dependent manner.

Together, these data demonstrate a change in microenvironment as tumors progress. That is, the islets and intermediate lesions in mice with established tumors still harbored intense inflammation, including CD8⁺ T cells. This was distinct from the more restricted inflammatory response seen in large tumors within the same pancreata. As tumors enlarged, it appeared that the permissive microenvironment converted to an immunosuppressive one. Although the proportion of tumor-infiltrating 8.3-TCR CD8⁺ T cells that incorporated BrdU within large tumors was similar to that seen in the thymus, the majority of cells did not take up BrdU. It is these latter cells that may represent dysfunctional cells.

We considered that the CD8⁺ T cell response could be self-limiting by recruiting other immune effectors that contribute to the immunosuppressive tumor microenvironment. To evaluate other cellular constituents that might have influenced intratumoral trafficking, we sought to better characterize the cells contained in larger tumors. There was no difference between the numbers of B cells or dendritic cells present in tumors from 8.3-NOD-RIPTAg and NOD-RIPTAg mice (data not shown). The increase in the number of CD4⁺ T cells in 8.3-NOD-RIPTAg tumors did not reach statistical significance (p = 0.055) (Fig. 6,A). Macrophages and CD4⁺ T suppressor cells were present in all established insulinomas. There was no significant difference in the proportions of CD4⁺CD25⁺Foxp3⁺ T cells between 8.3-NOD-RIPTAg and NOD-RIPTAg lesions (Fig. 6,B). Interestingly, there was a significant increase in the number of CD11b⁺ cells identified in 8.3-NOD-RIPTAg insulinomas compared with those derived from NOD-RIPTAg mice (Fig. 6 C).

The CD11b⁺ infiltrate was further characterized. The majority (86.9 ± 0.67%) consisted of F4/80⁺ macrophages. Of those, 85.6 ± 2.9% expressed macrophage scavenger receptor 1, which is typically found on alternatively activated (M2) macrophages (25) (supplemental Fig. 1).⁵ Myeloid-derived suppressor cells also have immunosuppressive activity and are recognized as CD11b⁺Gr-1⁺ (26, 27). Only 1.6 ± 1.1% of CD11b⁺ cells coexpressed Gr-1 within the tumor infiltrate (data not shown).

Macrophage infiltration was evaluated as a function of lesion size and correlated with CD8⁺ T cell infiltration. In 8.3-NOD-RIPTAg mice, macrophage infiltrate increased steadily as lesion size increased (Fig. 6 D). This was accompanied by augmented CD8⁺ T cell infiltration in small and intermediate size lesions. However, in large tumors macrophage infiltration continued to rise, and suddenly CD8⁺ T cells were constrained to the tumor periphery.

Together, these data demonstrate that the enhanced CD8⁺ T cell response in 8.3-NOD-RIPTAg mice was accompanied by the recruitment of other immune effector populations, including macrophages and CD4⁺ T cells. However, the increase in CD4⁺ T cells within the tumor compartment of 8.3-NOD-RIPTAg mice was not significantly larger than that seen in NOD-RIPTAg animals. We postulate that the inflammatory response secondary to the influx of CD8⁺ T cells altered the tumor microenvironment to an immunosuppressive one and ultimately may have altered the capability of CD8⁺ T cells to effectively lyse tumors.

Most immunotherapeutic strategies for generating immunity against tumors rely on the inhibition of tolerogenic mechanisms. Such strategies have dominated cancer immunotherapeutics because most tumor Ags reflect the antigenic repertoire of “self” and are therefore not recognized by the immune system as foreign. We have created a spontaneous tumor model that fully tests the capability of this strategy to inhibit tumor development. The model is important for the following reasons: 1) the tumor develops spontaneously; 2) the Ag recognized by CD8⁺ T cells is a naturally occurring endogenous Ag that is expressed in normal β-islet cells as well as in transformed β-islet cells; 3) diabetogenic CD8⁺ T cells are primed spontaneously due to the breakdown of tolerance mediated by multiple mechanisms (12); and 4) the relative magnitudes of autoimmunity and tumor immunity can be monitored. Even in the presence of overt autoimmunity, despite the initial inhibition of tumor development, tumors ultimately emerged. This effectively demonstrates that immunotherapeutic strategies that rely solely on bypassing tolerance are insufficient for cancer therapeutics.

In a TCR transgenic model in which a single Ag is targeted, a number of factors could have contributed to the ultimate emergence of tumors. Such factors included clonal deletion, clonal anergy, poor delivery of tumor-specific immune effectors to tumor, Ag loss, or MHC down-regulation. None of these factors were found to contribute to tumor progression. Rather, CD8⁺ T cells efficiently homed to early tumors, as well as to normal islets. Tumors appeared capable of priming CD8⁺ T cells in draining peripancreatic LNs even at later stages of development. Moreover, intratumoral CD8⁺ T cells retained at least some of their capability to respond to larger tumors, as evidenced by their intratumoral proliferation. However, unlike in normal islets and in smaller tumors, CD8⁺ T cells were excluded from the centers of large tumors, where the absence of any inflammatory response mediated by other immune effectors was striking.

Indeed, in our model we argue that the immunosuppressive tumor microenvironment limited CD8⁺ T cell-mediated tumor lysis, particularly near the center of the tumor. That immunosuppressive microenvironment appeared not to be present in normal islets, as autoimmune DM frequently preceded tumor development. We speculate that these microenvironmental factors were present early in tumor development, as one might surmise that even “early tumors” should not have developed in the presence of a sufficiently inhibitory antitumor immune response. We further speculate that this immunosuppressive microenvironment is a product of the other cell populations that comprise the tumor mass.

The role of regulatory T cells in tumor immunity has recently been extensively investigated, and it is well established that CD4⁺CD25⁺ T cells can influence T cell function in the tumor microenvironment (28, 29). For example, CD4⁺CD25⁺ T cells were found to inhibit intratumoral T cell function in a TAg transgenic model of prostate dysplasia (29). We were unable to identify any alterations in the numbers of CD4⁺CD25⁺FoxP3⁺ cells in insulinomas derived from 8.3-TCR transgenic mice as compared with individuals without this transgene. However, this does not exclude the possibility that CD4⁺CD25⁺ suppressor T cells contribute to the dysfunction of intratumoral CD8⁺ T cells.

Although tumor-associated macrophages have the potential to induce a protective innate antitumor immune response, they have also been shown to inhibit T cell function directly at the tumor site (30). Their polarization to an inhibitory phenotype may be a product of the tumor microenvironment, including cytokines produced by other inflammatory cell populations (31). We identified an impressive number of macrophages in both 8.3-NOD-RIPTAg and NOD-RIPTAg insulinomas. The fact that they were particularly numerous in individuals with the 8.3-TCR transgene suggests that the homing of macrophages into a tumor is a function of diabetogenic CD8⁺ T cells, although this must be investigated further. It is also possible that the prominent macrophage population is simply secondary to the diabetogenic inflammatory response, as insulitis in NOD mice is similarly accompanied by macrophage infiltration (32). The potentially immunosuppressive role of intratumoral macrophages must be investigated further in this model.

Bone marrow-derived myofibroblasts may also influence the nature of the intratumoral inflammatory response. Myofibroblasts resemble muscle cells and fibroblasts and have been identified within RIPTAg insulinomas (33). They serve not only as structural components in the tumor microenvironment but also as key regulators of the intratumoral immune response (34). The distribution of T cells and macrophages has been shown to be associated with large sheaths of myofibroblasts at the tumor margin of both rat colorectal tumors and RIPTAg insulinomas (33, 35). Evidence suggests that tumor-associated myofibroblasts not only prevent T cell infiltration in established tumors in vivo, but also have the ability to shield tumor cells from T cell-mediated lysis (33). The paucity of T cells at the center of large tumors in our model may therefore be secondary to the presence of insulinoma-associated myofibroblasts, although this was not specifically evaluated.

In summary, we have demonstrated that even an immune response sufficiently robust to induce autoimmunity is not sufficiently robust to retard tumor progression. Although functional CD8⁺ T cells were delivered to tumors, initially inhibiting tumor progression, the tumor microenvironment differed from that of normal islets. These microenvironmental perturbations resulted in the exclusion of inflammatory cells from the central portions of larger tumors, ultimately allowing unfettered tumor growth. Our observations demonstrate that immunotherapeutic strategies that rely solely on breaking down tolerance are insufficient for cancer therapeutics. Rather, success will depend on the identification and inhibition of factors that contribute to an immunosuppressive intratumoral microenvironment.

We acknowledge Dr. Chris Mody for helpful advice during this work. We also thank Sandy Eyton-Jones for technical assistance.

The authors have no financial conflict of interest.