T-bet Controls Autoaggressive CD8 Lymphocyte Responses in Type 1 Diabetes

Modulating the cytokine balance of autoreactive T lymphocytes can be of great value for preventing development of autoimmune disorders. In type 1 diabetes (T1D), autoaggression is supported by IFN-γ, IL-12, and cytotoxic activity, whereas counter-regulation is achieved by Th2 cytokines such as IL-4, IL-5, and possibly IL-10. Therefore, we wished to explore the role of the recently discovered (1) transcription factor T-bet in T1D because it was known that T-bet controls lineage commitment and IFN-γ production by CD4 T cells (1, 2). The significant Th2 shift in T-bet–deficient mice can lead to an asthmalike syndrome (3). This type of skewing of the cytokine equilibrium occurs on a systemic level and also extends to the mucosal compartment (4, 5). Thus, T-bet appears to be a master controller of the Th1 cytokine pathway and might constitute a mechanism for “cytokine memory” in T helper cells, as much as GATA-3 controls Th2 lineage commitment (6). Interestingly, although Th1 cell function is markedly impaired in the absence of T-bet, IFN-γ production by CD8 T cells after polyclonal stimulation is unaffected (2).

A suitable diabetes model to test the role of T-bet on CD8 responses in vivo during T1D is the rat insulin promoter (RIP)–lymphocytic choriomeningitis virus (LCMV) mouse developed by Oldstone and Zinkernagel (7, 8). These transgenic mice express viral antigens derived from LCMV under control of the RIP in pancreatic β cells (9). Viral infection with LCMV induces rapid CD8-dependent diabetes in RIP-LCMV-glycoprotein (GP) transgenic lines. Importantly, the infection is first cleared in these mice before the autoimmune process fully develops. Therefore, islet autoimmunity is governed by different immunological rules compared with clearance of the systemic LCMV infection, and treatments that abrogate the diabetogenic process need not necessarily impair viral clearance (10, 11). Lymphocytes infiltrating the islets are of the experienced (effector/memory) phenotype, which is characterized functionally by concomitant production of TNF-α and IFN-γ, but not IFN-γ alone (12, 13). Islet destruction depends on production of inflammatory cytokines such as IL-12 and IFN-γ in addition to cytolytic effector function by autoaggressive (anti-LCMV) CD8 lymphocytes (10, 14).

Based on these arguments, the RIP-LCMV model appeared to be an ideal tool for evaluating the role of T-bet in CD8 responses in autoimmunity. The suitability of RIP-LCMV mice as a valid diabetes model for our analysis is further supported by the fact that more recently a strong role for CD8 cells as the final effectors in β cell destruction has been found (14–17). In addition, antigenic spreading occurs in our RIP-LCMV-GP lines as evidenced by autoantibodies to insB and GAD preceding the onset of T1D (18). Finally, diabetes in RIP-GP mice occurs in a CD4-independent fashion (9, 19), and the effect of T-bet on CD8 responses can be segregated from its known effect on CD4 T cells. In the studies reported here, we found a dramatic reduction of disease. This is associated with a decrease of islet infiltration and significant reduction of LCMV-specific (“anti-self”) memory CD8 responses. Blocking the activity of T-bet might be of unique therapeutic value in preventing T cell–mediated autoimmune disorders.

Mice deficient in T-bet (T-bet^−/−) were generated on a C57BL/6 × 129 background as described previously (2). T-bet^−/− mice were backcrossed three times to C57BL/6 before being crossed to RIP-GP (7, 9) to obtain T-bet^+/− × RIP-GP mice. These mice were intercrossed to obtain T-bet^+/+, T-bet^+/−, and T-bet^−/− × RIP-GP. In some experiments, nontransgenic T-bet^−/− or ^+/− and ^+/+ littermate control mice were used. Mice were housed under specific pathogen-free conditions at La Jolla Institute for Allergy and Immunology.

LCMV Armstrong clone 53b (20) plaque-purified three times on Vero cells and stocks prepared by a single passage on BHK-21 cells were used throughout experiments. 6–10-wk-old mice were infected with a single dose of 10⁵ PFU i.p.

Blood glucose was monitored with OneTouch Ultra at weekly intervals. Diabetes was defined as blood glucose values (BGVs) over 300 mg/dl (21).

Peptides used for viral studies were the dominant D^b-restricted LCMV epitopes GP_33–41 and NP_396–404 (The Scripps Research Institute Peptide Core Facility; reference 22) as well as I-A^b–restricted epitope GP_61–80 (Chiron Corp.; reference 23).

Tissues were immersed in Tissue-Tek OCT (Bayer) and quick frozen on dry ice. Using cryomicrotome and sialin-coated Superfrost Plus slides (Fisher Scientific), 6–10-μm tissue sections were cut. Sections were fixed with 90% ethanol at −20°C, and, after washing in PBS, an avidin/biotin-blocking step was included (Vector Laboratories). Primary and biotinylated secondary antibodies (Vector Laboratories) were reacted with the sections for 60 min each, and color reaction was obtained by sequential incubation with avidin–peroxidase conjugate (Vector Laboratories) and diaminobenzidine–hydrogen peroxide. Primary antibodies used were rat anti–mouse CD8a (Ly2) and rat anti–mouse CD4 (L3T4; BD Biosciences).

For intracellular stains, single cell suspensions were restimulated for 5 h with 1 μg/ml MHC class I–restricted viral peptides, 2 μg/ml MHC class II–restricted viral peptides, or 2 μg/ml anti-CD3 and anti-CD28 in the presence of brefeldin A (Sigma-Aldrich). Cells were stained for surface expression of CD4 and CD8, fixed, permeabilized, and stained for intracellular IFN-γ, TNF-α, or IL-2. CCR7 staining was performed by incubation with ELC-Ig followed by goat anti–human Fc-FITC (24). D^bGP₃₃ and D^bNP₃₉₆ MHC class I tetramers were produced as described previously (25). Samples were acquired using a FACSCalibur™ (Becton Dickinson).

Membrane-bottom 96-well plates were coated overnight with IL-4 or IFN-γ capture antibodies. Splenocytes were plated at various dilutions and cultured at 37°C, 5% CO₂ with GP₆₁ MHC II–restricted viral peptide. After 24 h (IFN-γ) or 48 h (IL-4), the plates were washed, the cells were removed, and an IFN-γ or IL-4 detection antibody was added at 4°C overnight. A streptavidin/angiotensin-converting enzyme color reaction was used to identify cell cytokine secretion patterns, and the spots were counted with the aid of a dissection microscope.

For LCMV-specific responses, memory splenocytes were cultured for 6 d on LCMV-infected irradiated peritoneal exudate cells. After culture, viable cells were collected, and cells were incubated for 5 h with peptide in the presence of brefeldin A before staining for intracellular IFN-γ. The fold increase of specific CD8⁺ cells was calculated by dividing the absolute number of IFN-γ⁺ CD8⁺ cells after in vitro culture by the absolute number of IFN-γ⁺ CD8⁺ cells at the start of the culture.

Quantitation of cytokines in cell culture supernatants after in vitro stimulation was performed in a sandwich ELISA. For IL-4, IL-10, and IFN-γ assays, 96-well plates were precoated overnight at 4°C with purified capture antibodies (BD Biosciences). After washing and blocking, serial dilution of supernatants and recombinant cytokine standards (BD Biosciences) were incubated overnight at 4°C. Plates were further incubated with matched biotinylated detection antibodies and a streptavidin–peroxidase conjugate. For color reaction, H₂O₂-activated 2,2'-Azino-di-[3-ethylbenzthiazolin-6-sulfonic acid] (ABTS; Sigma-Aldrich) solution was added. Plates were read at 405 nm. For TGFβ quantitation, the TGFβ1 Emax ImmunoAssay kit (Promega) was used.

CD8⁺ T cells were purified from GP₃₃ TCR-tg (26) or T-bet^{+/+, +/−}, or ^−/− donors by negative selection (Dynal). 10⁴ GP₃₃ TCR-Tg cells were transferred i.v. into T-bet^{+/+, +/−}, or ^−/− recipients.

Bone marrow was isolated from tibias and femurs of donor T-bet^−/−, T-bet^+/−, or C57Bl/6 Thy1.1 congenic mice. Bone marrow was depleted of CD4 and CD8 T cells, and a 1:1 mixture containing a total of 10⁷ cells was injected i.v. into lethally irradiated (1,100 rad) T-bet^+/− or T-bet^−/− recipients. Mice were allowed to reconstitute over a period of 6 wk, which was confirmed by FACS^® analysis of peripheral blood lymphocytes before infection with LCMV.

To examine a role for T-bet in autoimmune diabetes (T1D), T-bet^−/− mice were crossed to RIP-GP transgenic mice. T-bet^+/+, T-bet^+/− or T-bet^−/− × RIP-GP littermates were infected with 10⁵ PFU LCMV Armstrong to induce diabetes. BGVs were measured weekly, and by 2 wk after infection, 80–100% of T-bet^+/+ or T-bet^+/− × RIP-GP mice were diabetic (Fig. 1 A), with average BGVs of ∼500 mg/dl (Fig. 1 B). In these experiments, a small percentage of the diabetic mice reverted to normal BGVs at >3 wk after infection. This has been described previously in mice of a mixed 129 × B6 background (27). However, by 35 d after infection, 75–80% of T-bet^+/+ and T-bet^+/− × RIP-GP mice remained diabetic. In contrast, most of the T-bet^−/− × RIP-GP had normal BGVs on day 14 (Fig. 1 B), averaging ∼200 mg/dl. In addition, diabetes incidence in T-bet^−/− × RIP-GP remained dramatically reduced with only 10% diabetic (Fig. 1 A) 35 d after infection. Thus, elimination of T-bet results in a highly significant reduction (P < 0.0001 by Student's t test) of BGVs and autoimmune diabetes.

Pancreata from T-bet × RIP-GP mice were examined by immunohistochemistry for the presence of CD4 and CD8 T cells. Islets from diabetic T-bet^+/− × RIP-GP mice were heavily infiltrated with both CD4 and CD8 T cells. In contrast, islets from T-bet^−/− × RIP-GP had less severe infiltration, with CD4 and CD8 T cell present mostly around the periphery of the islets (Fig. 1 C).

The CD8 immune response to LCMV is characterized by an initial primary expansion of LCMV-specific, IFN-γ single positive T cells, peaking ∼7–8 d after infection. This is followed by a contraction phase leaving resting LCMV memory CTLs that can turn into IFN-γ/TNF-α double positive effector/memory CD8 T cells upon antigenic restimulation. Subsequently, memory CD8 T cells are maintained at a similar frequency for life. The frequency of LCMV-specific CD8 lymphocytes was evaluated over time. Primary CD8 T cell responses examined on day 8 after LCMV infection were not impaired in T-bet–deficient mice as evidenced by unimpaired IFN-γ production (Fig. 2 A) and cytotoxicity toward LCMV-infected targets (not depicted). Accordingly, viral clearance was not affected (unpublished data). This was not unexpected because this particular primary antiviral response occurs independently of many immunological parameters, including B7, IFN-γ, CD4 help, CD28, and CD40L (28–35). In addition, previous studies have demonstrated that the primary immune response on day 8 does not correlate with disease, whereas effector/memory cells do (36, 37).

In contrast, and consistent with the reduction of autoimmune diabetes, antigen-specific effector/memory CD8 T cells were dramatically reduced by day 14 after LCMV infection, with 2% IFN-γ positive cells in T-bet^−/− mice after culture with GP₃₃, compared with 10% in T-bet^+/+ mice (Fig. 2 A). The reduction seen at this time point is likely to be highly relevant in terms of diabetes induction because it corresponds exactly to the peak of clinical disease in control mice. This lower number of LCMV-specific CD8 T cells after day 14 is maintained in T-bet^−/− mice over time (Fig. 2 A). In agreement with the reduced number of IFN-γ positive effector/memory CD8 lymphocytes after 5 h stimulation with peptide is the reduced amount of IFN-γ in tissue culture supernatants assessed by ELISA after a 3-d stimulation period (Fig. 2 A). Despite the significantly reduced IFN-γ production, there was no compensatory increase in the production of Tc2 cytokines, such as IL-4, IL-10, and TGF-β. LCMV-specific production of IL-4, IL-10, or TGF-β by CD8 T cells could not be detected by ELISA or ELISPOT in T-bet^−/− or T-bet^+/− mice (see Fig. 6 D and not depicted). Thus, antigen-specific CD8 effector/memory responses in T-bet–deficient mice are markedly impaired.

Phenotypically, effector/memory CD8 cells are predominantly present in islets of diabetic RIP-GP mice and secrete TNF-α in addition to IFN-γ (12). To assess the activity of autoaggressive (LCMV specific) CD8 lymphocytes in proximity to the target organ, antigen-specific production of IFN-γ and TNF-α by CD8 T cells was examined in T-bet × RIP-GP mice 1 mo after infection. Cells were pooled from the spleen or pancreatic draining LNs of T-bet^+/− and T-bet^−/− × RIP GP mice and were stimulated in vitro with the MHC class I LCMV peptides GP₃₃ or NP₃₉₆, or with anti-CD3 and anti-CD28. Intracellular cytokine staining was performed to determine antigen-specific IFN-γ and TNF-α production. Compared with diabetic T-bet^+/− mice, T-bet^−/− mice exhibited an ∼50% decrease in double cytokine–producing (TNF-α and IFN-γ) effector/memory CD8 T cells specific for GP₃₃ and NP₃₉₆ in the spleen (Fig. 2 B) and pancreatic lymph node (Fig. 2 C) at 1 mo after LCMV infection. IFN-γ and TNF-α production by CD8 T cells was also reduced in T-bet^−/− mice after polyclonal stimulation with anti-CD3 and -CD28 1 mo after LCMV infection (Fig. 2, B and C). Thus, effector functions of LCMV-specific autoaggressive lymphocytes were significantly reduced in the pancreatic draining lymph node of T-bet–deficient RIP-GP mice.

The aforementioned results demonstrated a reduced frequency of LCMV-specific CD8 T cells in T-bet^−/− mice. To determine if this was due to an overall numeric reduction in antigen-specific T cells, the number of LCMV-specific T cells per spleen was determined after peptide restimulation and IFN-γ staining. There were fewer IFN-γ⁺ LCMV-specific CD8 T cells in spleens of T-bet^−/− mice compared with T-bet^+/+ or ^+/− controls 1 mo after infection (Fig. 3 A). In addition, splenocytes were stained with H-2D^b tetramers loaded with GP₃₃ or NP₃₉₆ peptides to ensure that the reduced frequency of IFN-γ⁺ cells as measured by intracellular cytokine staining reflected a reduction in the actual number of LCMV-specific CD8 memory T cells. As shown in Fig. 3 B, there was also a reduction of GP₃₃ or NP₃₉₆-specific CD8 T cells in T-bet^−/− mice compared with controls after tetramer staining. Thus, deficiency of T-bet results in reduced numbers of CD8 memory lymphocytes in vivo. Splenocytes were also stained for the expression of CCR7 to enumerate the frequency of effector/memory versus central memory T cells. On day 14 after LCMV infection, there was a significantly lower percentage of LCMV-specific effector/memory CD8 T cells (CD8⁺, NP₃₉₆⁺, CCR7neg) in T-bet^−/− mice (Fig. 3 C). CCR7 staining was also performed on day 28. T-bet^−/− mice still had a lower percentage of effector/memory T cells, but the difference from controls was not statistically significant, and much less dramatic than observed on day 14 (Fig. 3 C).

The ability of T-bet^−/− memory T cells to expand in response to reexposure to LCMV in vitro was examined. As shown in Fig. 4 A, CFSE-labeled memory splenocytes from T-bet^−/− mice do proliferate, although the overall percentage of GP₃₃-specific cells is still reduced compared with controls after 6 d of culture. Because the LCMV-infected T-bet^−/− mice typically have approximately threefold fewer LCMV-specific T cells ex vivo, the fold increase in LCMV-specific T cells was determined by dividing the absolute number of GP₃₃-specific T cells present after culture by the number present ex vivo. Using this calculation, the proliferation of T-bet^−/− GP₃₃-specific CD8 T cells after 6 d in culture is as good or better than T-bet^+/+ T cells (Fig. 4 B). Furthermore, the mean fluorescent intensity of IFN-γ production was equivalent comparing memory CTL from T-bet^+/+ and ^−/− mice (Fig. 4 C). It is interesting that, despite a numeric reduction in CD8 memory T cells in vivo, these cells were still competent to expand in vitro in the presence of antigen (virus or peptide). Thus, although the numbers of effector/memory cells are reduced in T-bet–deficient mice, their ability to respond to antigen with proliferation, cytolytic activity (unpublished data), and IFN-γ production is completely intact when assessed on a per cell basis. Therefore, we conclude that generation of memory CTL in T-bet^−/− mice is defective, but that the remaining effector/memory cells found in T-bet^−/− mice are functionally competent.

It has been shown previously that one of the functions of T-bet is to repress the IL-2 promoter (1). Accordingly, we examined IL-2 production by CD8 T cells after infection with LCMV. On day 8 after infection, LCMV-specific CD8 T cells from T-bet^−/− and T-bet^+/− mice produced more IL-2 compared with T-bet^+/+. As shown in Fig. 5 A, T-bet^−/− mice have a higher frequency of IFN-γ, IL-2⁺ GP₃₃-specific CD8 T cells compared with T-bet^+/+ mice (∼6 vs. 1%, respectively). T-bet^+/− mice showed an intermediate phenotype with ∼3% of CD8 T cells producing IL-2 in response to GP₃₃. IL-2 production was also examined at the memory stage. Although the absolute number of LCMV-specific CD8 T cells is reduced in T-bet^−/− mice, the ratio of IL-2⁺ to IL-2⁻ LCMV-specific CD8 T cells is still much higher in T-bet^−/− compared with T-bet^+/+ (Fig. 5 A). The higher levels of autologous IL-2 production by CD8 T cells in T-bet^−/− mice was unable to rescue the memory defect in vivo, despite the fact that IL-2 is a key helper factor for programming CD8 memory (38).

Interestingly, a recent paper demonstrated that in vivo administration of IL-2 during the expansion phase (days 0–8) resulted in fewer antigen-specific T cells and more apoptosis (39). Thus, it was possible that the enhanced IL-2 production by T-bet^−/− CD8 T cells during the acute phase of infection might actually be partially responsible for the defect in effector/memory T cells. To examine this, annexin V staining was performed on LCMV tetramer positive cells at days 9, 10, and 12 after infection with LCMV. Compared with T-bet^+/− or ^+/+ controls, T-bet^−/− mice did not have significantly enhanced levels of annexin V⁺ antigen-specific cells (Fig. 5 B). To ensure that possible differences in apoptosis were not masked by the rapid clearance of apoptotic cells in vivo, these experiments were repeated in vitro. Splenocytes were harvested on day 8 after LCMV infection and were cultured for 1, 2, or 3 d in medium before staining for annexin V on tetramer positive cells. These studies also revealed that T-bet^−/− mice did not have elevated levels of apoptosis compared with controls (Fig. 5 C). Thus, increased contraction of the primary immune response by apoptosis is not the cause for reduced levels of effector/memory CD8 CTL in T-bet^−/− mice.

Autoimmune diabetes occurs independently of CD4 help in RIP-GP mice (9, 19). Thus, the resistance to diabetes of T-bet^−/− GP⁺ mice cannot directly be explained by the profound CD4 helper Th1 defect that has been demonstrated in previous studies (2). However, it was still possible that the reduced number of CD4 Th1 lymphocytes in T-bet^−/− mice affected generation of CD8 memory. Indeed, memory CD8 responses after LCMV infection depend on CD4 helper activity for maintenance and effector programming (35, 38). We sought to resolve this issue.

As expected, LCMV-specific, IFN-γ–secreting CD4 Th1 cells were decreased in T-bet^−/− mice. At both the memory stage (Fig. 6 A), and at the acute phase of infection on day 8 (Fig. 6 C), IFN-γ production by CD4 T cells from T-bet^−/− mice was greatly reduced. LCMV-specific CD4 T cells from heterozygous T-bet^+/− mice were also substantially reduced in number (Fig. 6, A and C), consistent with the gene-dosage effect observed previously (1, 2). Because both T-bet heterozygous (+/−) and T-bet^−/− mice exhibit a similar defect in LCMV-specific CD4 T cells (Fig. 6), but the former maintains completely normal LCMV-specific memory CD8 T cell responses (Fig. 2), it is unlikely that a T-bet–dependent CD4 defect alone can account for the defective α-LCMV CD8 memory response observed. Indeed, attempts to correct the CD8 memory generative defect in T-bet^−/− mice by adoptive transfer of 10–30 × 10⁶ naive CD4 T-bet^+/+ lymphocytes before LCMV infection failed to correct the memory defect (Fig. 7 A).

It was possible that in the absence of T-bet, CD4 T cells would default to the Th2 pathway in context of the viral (LCMV) infection as described previously in unprimed mice (3). Therefore, ELISPOT assays for IL-4 or ELISAs for IL-10 and TGF-β were performed using splenocytes stimulated with GP₆₁. Although the CD4 T cells from T-bet^−/− and T-bet^+/− mice produced reduced levels of IFN-γ compared with controls, there was no significant increase in antigen-specific IL-10, IL-4, or TGF-β production (Fig. 6, B–D). Thus, it is very unlikely that Th2 deviation would affect CD8 memory responses in T-bet^−/− via bystander suppression, as described previously in other experimental settings for antigen-induced regulatory cells (11).

To further examine the notion that T-bet^−/− CD8 T cells have an intrinsic defect, a series of adoptive transfer experiments was performed. GP₃₃-TCR-tg C57Bl/6 (WT) T cells were transferred into T-bet^{+/+, +/−}, or ^−/− recipients before infection with LCMV. Although the adoptive transfer of TCR transgenic T cells reduced the overall frequency of the endogenous CD8 T cell response, T-bet^−/− mice still had fewer endogenous NP₃₉₆-specific T cells compared with T-bet^+/+ on day 14 (Fig. 7 B). In contrast, the frequency of GP₃₃-specific T cells was normal in T-bet^−/− mice (Fig. 7 B). Thus, WT GP₃₃-TCR-tg CD8 T cells were not impaired in an otherwise T-bet–deficient environment. In addition, mixed bone marrow chimeras were created to further demonstrate an intrinsic defect in T-bet^−/− CD8 T cells. Bone marrow was isolated from donor T-bet^−/−, T-bet^+/−, or C57Bl/6 (WT) Thy1.1 congenic mice. A mixture of T-bet^−/− and WT Thy1.1 bone marrow was injected into lethally irradiated recipients. As a control, T-bet^+/− and WT Thy1.1 mixed bone marrow chimeras were also generated. Mice were rested for 6 wk, and the percentage of Thy1.2⁺ (T-bet^+/− or T-bet^−/−) CD8 T cells was determined in peripheral blood. Both T-bet^+/− and T-bet^−/− chimeras expressed Thy1.2 on 50–60% of peripheral blood CD8 T cells, indicating that the chimerism had been achieved (Fig. 7 C). Mice were subsequently infected with LCMV, and the frequency of Thy1.1 (WT) and Thy1.2 (T-bet^+/− or ^−/−) LCMV-specific CD8 T cells was determined on day 14. The generation of LCMV-specific T-bet^−/− CD8 T cells was dramatically impaired in WT/KO mixed bone marrow chimeras. Thy1.2⁺ (T-bet^−/−) CD8 T cells accounted for ∼60% of total CD8 T cells before infection, whereas they represented only ∼16% of GP₃₃-specific T cells on day 14 after infection (Fig. 7 C). In contrast, control WT/T-bet^+/− chimeras had a similar fraction of Thy1.2⁺ LCMV-specific CD8 T cells compared with preinfection. These results demonstrate that the generation of T-bet^−/− memory CD8 T cells is impaired in the very same immunological environment that allows the successful generation of WT CD8 memory T cells.

Here, we have demonstrated a critical role for the transcription factor T-bet in the autoaggressive behavior of CD8 lymphocytes. It is striking that the absence of T-bet so profoundly affects the development of insulitis and diabetes in a solely CD8-mediated model, as is the case in the RIP-GP transgenic mouse (9). Our data appear to conflict with earlier studies in T-bet^−/− mice that failed to demonstrate an IFN-γ production defect in polyclonally activated primary CD8 cells (2). However, we have recently obtained evidence for an essential role for T-bet in antigen-driven as opposed to polyclonal, antigen-nonspecific generation and effector function of CD8 cells. In these studies, a profound defect in CD8 function in the absence of T-bet was revealed in an antigen-specific, TCR-transgenic setting (40). The findings reported here extend these observations, and demonstrate an intrinsic defect in the generation of effector/memory CD8 T cells after viral infection in vivo.

Our data provide novel mechanistic insight as to how T-bet functions in CD8 cells and demonstrate an intrinsic defect in generation of effector/memory CTLs in the absence of this factor. Primary antiviral CD8 responses develop equally in T-bet–competent or –deficient mice. This is conceivable if one considers that primary immune responses to LCMV may follow different rules compared with effector/memory and indeed, have been shown to be independent of many immunological parameters including B7, CD4 T cells, CD28, and CD40. However, in the absence of T-bet, lower numbers of effector/memory CD8 lymphocytes resulting in overall reduced IFN-γ and TNF-α production are found after LCMV infection. Their lack is also evident in the pancreatic draining lymph node and target organ that harbors autoaggressive effector CD8 lymphocytes in RIP-LCMV mice.

The difference between T-bet^−/− and T-bet^+/+ mice is most dramatic on day 14, which corresponds to the peak of clinical disease in the RIP-GP model. Studies performed on day 14 revealed the largest difference in terms of the number of IFN-γ–producing LCMV-specific T cells, as well as a reduced percentage of CCR7neg effector/memory cells. It is likely that effector/memory cells, as opposed to central memory, are the most important type of cell for mediating immune pathology during T1D, due to their ability to home to peripheral tissues (41).

What is the cause for defective antiviral CD8 memory in T-bet–deficient mice? Although antigen-specific studies performed with transgenic T cells demonstrated a Tc2 shift in terms of IL-4 and IL-10 production (40), we did not observe any significant Tc2 (CD8) or Th2 (CD4) deviation in the LCMV viral system. This discrepancy might be due to the intrinsic Th1 skewing properties of virus infection as compared with peptide stimulation alone. The absence of Tc2/Th2 cytokines in the LCMV system makes the decrease in autoaggressive LCMV-specific CD8 response unlikely to be due to counter-regulation, as has been described previously in the RIP-LCMV system (11).

Because our studies demonstrate that reduced CD4 helped and/or increased IL-2 were likely not the reasons for decreased CD8 memory, we assessed a role for another cytokine important in CTL generation, IL-12. However, systemic administration of IL-12 to T-bet knockout mice did not resuscitate CD8 memory (unpublished data). Along these lines, levels of IL-15, another cytokine critical in the generation and maintenance of CD8 memory cells, was normal in the bone marrow of T-bet^−/− mice (unpublished data).

Lastly, some consideration should be given to the role of IFN-γ in our system. IFN-γ is an important factor in β cell destruction and viral clearance (14, 37, 42). Thus, the reduction of T1D in T-bet^−/− mice agrees well with reduced levels of CD8 lymphocytes resulting in an overall decrease in IFN-γ and lytic activity. Furthermore, IFN-γ–deficient mice exhibit intact primary CTL responses to LCMV, whereas memory CTLs were reduced (37). If one considers that the memory response is programmed in the LCMV system during the primary antiviral response phase (38), possible reduction of IFN-γ levels at the time of the primary antiviral response might have resulted in an altered program in memory CTL. Indeed, major amounts of IFN-γ by CD8 CTLs are generated early (days 6–9) after LCMV infection. IFN-γ can also be produced by other cell types, such as dendritic cells and macrophages. A recent paper demonstrated that IFN-γ production by myeloid dendritic cells after culture with IL-12/IL-18 is reduced in the absence of T-bet (43); this resulted in a modest reduction in IFN-γ production, but no proliferative defect by T cells after in vivo priming with antigen-pulsed DCs (43). However, we have not detected any difference in the ability of LCMV-infected T-bet^+/+ or T-bet^−/− APCs to stimulate LCMV-specific CD8 memory or naive T cells in vitro (unpublished data) or in vivo (Fig. 7 B). In addition, because IFN-γ levels made by primary LCMV CTLs in T-bet^−/− mice match those of T-bet^+/− or ^+/+ littermates (Fig. 2), we argue that lack of IFN-γ at the acute phase of infection is not responsible for the observed defect in memory CTLs. All of these data make it likely that the defect is intrinsic to the CD8 cell rather than attributable to the cytokine microenvironment. Most convincingly, our studies using bone marrow chimeras demonstrate that the generation of T-bet^−/− memory CD8 T cells is impaired in the very same immunological environment that allows the successful generation of WT CD8 memory T cells.

In conclusion, our observations indicate that small molecule inhibitors of crucial transcription factors such as T-bet might offer a unique therapeutic opportunity for preventing autoimmunity. T1D likely depends on both autoaggressive CD4 as well as CD8 lymphocytes. Indeed, a potent role for CD8 effector cells has been found more recently in many T1D models (15–17). Therefore, the intrinsic CD8 defect specific for effector/memory cells but not primary CD8 lymphocytes reported here in the absence of T-bet strengthens the rationale for targeting T1D. It will need to be determined whether control of chronic infections, or long-term immunity after vaccination is sufficient in the absence of T-bet, although our results here demonstrate that lack of T-bet does not render mice susceptible to an acute viral infection.

We thank Dr. U. Christen and A. Rhode for the generation of MHC-I tetramers and D. Frye for assistance with the manuscript.

M.G. von Herrath was supported by National Institutes of Health grants AI44451, DK51091, and U-19 AI51973 and a Juvenile Diabetes Research Foundation (JDRF) grant. A.E. Juedes is the recipient of a JDRF postdoctoral fellowship. L. Glimcher was supported by a grant from the JDRF Center for Islet Transplantation at Harvard Medical School. This is manuscript no. 538 from the Department of Developmental Immunology at the La Jolla Institute for Allergy and Immunology.