Dendritic Cells Induce Autoimmune Diabetes and Maintain Disease via De Novo Formation of Local Lymphoid Tissue

C57BL/6 mice were obtained from the Institut für Labortierkunde (University of Zürich). Mice expressing LCMV-GP under the control of the rat insulin promoter (RIP-GP) and perforin-deficient mice crossed with RIP-GP mice (PKO × RIP-GP) have been described previously (18, 22). Mice transgenic for a Vα2/Vβ8 TCR specific for H2-D^b and the major LCMV-GP epitope, GP33-41 (GP33), were used as donors of transgenic T cells (24). Transgenic mice expressing the LCMV GP33 epitope in all tissues were generated using a construct containing the H-2K^b regulatory elements and the coding sequence for amino acids 1–60 of the LCMV glycoprotein (H8-mice [25]). All animals were kept under specific pathogen–free conditions. Experiments were carried out with age- (8–16 wk) and sex-matched animals.

LCMV-WE was originally obtained from Dr. F. Lehmann-Grube (University of Hamburg, Germany) and was propagated on L929 cells. EL-4 (H-2^b), a thymoma cell line, was used as target cell. LCMV GP33-41 (KAVYNFATM; GP33) was produced by Neosystem Laboratoire (Strasbourg, France).

Generation of DC from C57BL/6 and H8 bone marrow cultures has been described previously (26). In brief, bone marrow was flushed from femurs and tibias and subsequently depleted of red cells with ammonium chloride. Bone marrow cells were depleted of lymphocytes, B cells, and I-A^b positive cells using a cocktail of mAbs (YTS191.1, YTS169.4.2, RA3-3A1/6.1, B21-2) and goat anti–rat IgG–coated Dynabeads (Dynal A.S., Oslo, Norway). Cells were plated at 0.5 × 10⁶/ml in RPMI 1640 supplemented with 5% FCS, penicillin/streptomycin, 10 ng/ml recombinant murine (rm)GM-CSF (supplied by Novartis Pharma Schweiz AG, Bern, Switzerland) and IL-4–containing supernatant from cell line X63-IL4 (provided by Dr. M. Kopf, Basel Institute of Immunology, Basel, Switzerland) to a final concentration of 100 ng/ml. At days 2 and 4 of culture, 50% of the supernatant was removed, and fresh medium and fresh cytokines were added. At day 6, nonadherent cells were collected and further purified over metrizamide (14.5% in RPMI 1640/5% FCS; Sigma Chemical Co., St. Louis, MO) to remove cell debris and high density cells. The resulting DC populations showed a purity of 80–90% as determined by cytofluorometry using mAb against MHC class I and II antigens, costimulatory molecules CD80 and CD86, and the DC marker NLDC-145 (not shown). Cells were washed three times with BSS (balanced salt solution), resuspended at 2 × 10⁶/ml in BSS, and serial 10-fold dilutions were made. DC were intravenously injected in a volume of 0.5 ml.

For detection of primary ex vivo cytotoxicity, effector cell suspensions were prepared from spleens of immunized mice at the indicated time point after priming. EL-4 cells were labeled with GP33 (10⁻⁶ M) and 250 μCi ⁵¹Cr for 1.5 h at 37°C. 10⁴ target cells per well were incubated in 96-well round-bottomed plates with serial threefold dilutions of spleen effector cells, starting with an E/T ratio of 90:1, for 15 h. EL-4 cells without peptide served as controls. The supernatant of the cytotoxicity cultures was counted in a gamma counter (Cobra II; Canberra Packard, Berkshire, UK). Percentage of specific lysis was calculated as (experimental release − spontaneous release)/ (total release − spontaneous release) × 100. Spontaneous release was always <29%.

Quantification of GP33-specific precursor CTL (CTLp) per spleen was performed by limiting dilution analysis as described (27). Responder spleen cells were titrated and cultured with 10⁴ LCMV-infected, irradiated peritoneal macrophages and 10⁵ irradiated feeder spleen cells in IMDM/10% FCS and 10% Con A supernatant using 16 wells per dilution step. After 6 d of culture, cytotoxicity was tested on GP33-labeled or unlabeled EL-4 cells in a standard ⁵¹Cr-release assay and CTLp frequencies were calculated.

To detect expansion of transgenic TCR- expressing T cells (24), peripheral blood cells, spleen cells, and lymph node cells from hepatic lymph nodes were stained for CD8, Vα2, and Vβ8.1 using FITC-conjugated rat anti–mouse CD8, PE-conjugated rat anti–mouse Vα2a, and biotinylated rat anti–mouse Vβ8.1 (all from PharMingen). Erythrocytes were lysed with FACS^® lysis solution (Becton Dickinson, San Jose, CA), and the cell suspensions were analyzed on a FACScan^® flow cytometer (Becton Dickinson) using logarithmic scales. Viable lymphocytes were gated using a combination of forward light scatter and 90° side scatter.

Freshly removed organs were immersed in HBSS and snap-frozen in liquid nitrogen. Tissue sections of 5-μm thickness were cut in a cryostat and fixed in acetone for 10 min. Sections were incubated with anti–mouse mAb against CD4⁺ cells (YTS191.1 [28]), CD8⁺ cells (YTS169.4.2 [28]), B cells (RA3-3A1/6.1; American Type Culture Collection, Rockville, MD), follicular DC (4C11 [29]); CD11c⁺ DC (30), peripheral node addressin (PNAd) (MECA-79; PharMingen), and polyclonal guinea pig antibodies against insulin (Dakopatts A/S, Glostrup, Denmark). Alkaline phosphatase–labeled, species-specific goat antibodies (Tago Inc., Burlingame, CA) were used as secondary reagents. The substrate for the red color reaction was AS-BI phosphate/New Fuchsin. Sections were counterstained with hemalum. 15–20 islets from 2–3 histological sections of each mouse were evaluated.

The glucose concentration in blood obtained from a tail vein was measured using Haemo-Glucotest strips (Boehringer Mannheim, Mannheim, Germany). Mice were considered diabetic with values >14 mM at two consecutive measurements.

To study whether DC can initiate and promote IDDM in the RIP-GP mouse, different doses of DC from H8-mice (H8-DC) were administered intravenously using various protocols, and blood glucose levels were monitored. With a single injection of a high dose of 10⁶ H8-DC, mice remained normoglycemic or displayed only transient hyperglycemia with intermediate blood glucose levels of 15–20 mM and eventually returned to normoglycemia within a few days (Fig. 1,A). In some mice (2/8), blood glucose quickly rose to values >44 mM (Fig. 1,A), followed by death after 4–6 d. After a single injection of an intermediate dose of 10⁵ H8-DC, diabetes did not develop (Fig. 1,B). In contrast, repetitive injection of intermediate doses of H8-DC, i.e., three doses of 10⁵ DC in 6-d intervals (Fig. 1,C) or four doses of 10⁴ DC in 2-d intervals (Fig. 1 D), caused development of diabetes. 50% of the mice repetitively immunized developed diabetes between days 10 and 14, whereas 40% of the mice developed hyperglycemia later, by days 18–21. One out of eight mice repetitively immunized with 10⁴ DC showed a transient hyperglycemia, and one out of eight mice from this group remained normoglycemic. 4 out of 16 mice repetitively immunized with H8-DC rapidly developed diabetes and died before day 21. This was surprising because the same mice infected with LCMV developed diabetes (with blood glucose values >44 mM), but survived usually for 40 d or more (not shown). Thus, under the conditions used, a severe and rapidly lethal IDDM was induced by DC, in a dose- and timing-dependent fashion.

To determine how dose and timing of antigen delivery by DC influences CTL responses, CTL activity and expansion at various time points after single high dose (10⁶), single low dose (10⁴), and repetitive intermediate doses (10⁵) of DC was monitored. Injection of a single high dose of DC induced CTL activity by day 4, as determined by direct ex vivo cytotoxicity in an overnight ⁵¹Cr-release assay; maximum CTL activity was reached by day 8 and waned between days 12 and 16. With low doses of DC, CTL activity was lower and the period during which CTL activity could be detected was shorter (Table 1). Importantly, repetitive stimulation with DC on days 0 and 8 prolonged CTL activity, leading to a weak but detectable CTL activity also on day 16 (Table 1).

To follow the expansion of GP33-specific CTLp in the spleen during the activation by DC, CTLp frequencies were determined by limiting dilution assay. The maximal expansion of GP33-specific CTLp was detected on day 4, followed by a continuous decrease until day 16 (Table 1). A 100-fold increase of DC used for immunization resulted only in a modest elevation of CTLp in the spleen, e.g., 1:15,000 and 1:5,000 at day 4 for 10⁴ and 10⁶ H8-DC, respectively (Table 1). Repetitive inoculation of 10⁵ DC did not further enhance CTLp frequency (Table 1).

To test whether the limited expansion of CTL after high dose or repetitive DC immunization was due to a lack of GP33-specific precursor cells or whether CTL leave the spleen immediately after activation by DC, we followed the expansion of H8-DC–activated GP33-specific TCR transgenic CTL (24). After adoptive transfer of 5 × 10⁵ spleen cells from TCR transgenic mice into C57BL/6 mice, 2–3% of the CD8⁺ T cells expressed the transgenic Vα2/Vβ8 TCR. Single DC immunization led to a maximal expansion of these cells by day 4 to ∼8% in spleen and blood and 5% in the hepatic lymph node (Fig. 2). Similarly, repetitive DC administration in 2-d intervals induced maximal expansion on day 4, with 22% in spleen and blood and 14% in hepatic lymph nodes (Fig. 2). In both single and repetitive immunization protocols, TCR transgenic T cell expansion was transient and TCR transgenic T cells persisted at slightly elevated levels between 3.5 and 5% (Fig. 2). After intravenous injection, H8-DC have been shown to migrate mainly to the spleen and, to a smaller proportion, to hepatic lymph nodes (26). Thus, it is reasonable to assume that GP33-specific T cells were first activated and expanded by H8-DC in spleen and hepatic lymph nodes and subsequently entered the blood to emigrate into peripheral nonlymphoid tissues.

Using immunohistochemical analysis, we evaluated the cellular composition and organization of pancreatic infiltrates in the course of DC-induced autoimmune diabetes. To this end, RIP-GP mice were immunized intravenously at days 0, 6, and 12 with 10⁵ H8-DC, and pancreata and submandibular salivary glands were analyzed on days 7, 14, and 21 for the presence of CD8⁺ and CD4⁺ T cells, B220⁺ B cells, F4/80⁺ macrophages, CD11c⁺ DC, and insulin-producing islet cells. Mice analyzed by immunohistochemistry on day 7 were normoglycemic, whereas mice analyzed on day 14 had been hyperglycemic for 1–2 d (blood glucose 17–25 mM), and mice analyzed on day 21 had been hyperglycemic for at least 7 d (blood glucose >44 mM). At the early time point, CD8⁺ T cells infiltrating the pancreatic islets clearly outnumbered CD4⁺ T cells, and B cells were barely detectable (Fig. 3, A1–5). Interestingly, APCs characteristic of lymphoid organs could be detected in these early infiltrates. Particularly DC staining positive with N418, an antibody that stains the gp150/ 90 α_x integrin CD11c (30), were found in the infiltrates (Fig. 3,A5). F4/80⁺ macrophages were less abundant in these early infiltrates (not shown). The infiltrates were always associated with insulin-positive islets or periductal in their close proximity (Fig. 3, A1–5). At day 14, more cells had infiltrated the insulin-producing islets, particularly CD4⁺ T cells, B cells, and N418⁺ dendritic cells (Fig. 3, B1–5). Correlating with the increased infiltration by immune cells, staining for insulin decreased (Fig. 3,B1), indicating the progressive destruction of islet cells. By day 21, infiltrating cells had nearly completely replaced islet cells (Fig. 3, C1–5). Control stainings of submandibular salivary glands in DC-immunized RIP-GP mice (Fig. 3,D) and liver and pancreas sections of DC-immunized C57BL/6 mice (not shown) showed only a few scattered CD8⁺ and CD4⁺ T cells. A striking finding was the formation of lymphoid structures in close proximity to remaining insulin-producing islets (Fig. 4). These newly formed lymphoid tissues were found mainly periductally in the pancreatic parenchyma (Fig. 4,A, arrow), whereas pancreatic lymph nodes were clearly located outside of and loosely attached to the pancreas (Fig. 4,A, arrowhead). CD4⁺ T cell infiltrates were found in insulin-positive islets as well as in the newly formed lymphoid structures (Fig. 4,B). In contrast, B cell infiltrates were detected mainly in the periductal lymphoid structures (Fig. 4,C), where they showed distinct compartmentalization particularly with CD8⁺ T cells (Fig. 4, E and F), comparable to that of peripheral lymphoid organs. Cells staining with 4C11, an antibody specific for follicular DC, were restricted to the newly formed lymphoid tissues and colocalized with B cells (Fig. 4, G and H). Their presence in the DC-induced pancreatic infiltrates is particularly interesting, since follicular DC are involved in B cell Ig-switch, affinity maturation, and memory. PNAd, an adhesion molecule found on endothelial cells in peripheral lymph nodes and involved in recirculation of lymphocytes into peripheral lymphoid tissues, was expressed in blood vessels in the periductal organized lymphoid structures (Fig. 4 I). Pancreatic endothelial cells outside these lymphoid structures did not express PNAd.

It has been shown previously that MHC-restricted destruction of β cells by perforin-dependent CTL is an important initiating event for autoimmune diabetes in RIP-GP mice (22). To evaluate whether contact-dependent T cell cytotoxicity is involved in DC-induced autoimmune diabetes, we studied perforin-deficient RIP-GP mice (PKO × RIP-GP; reference 22). 45% of the PKO × GP mice that were repetitively immunized with H8-DC developed diabetes between days 12 and 18 (Table 2). Immunohistological analysis revealed that all PKO × RIP-GP mice developed insulitis with >50% of the islets affected. Furthermore, PKO × RIP-GP mice rapidly progressed to diabetes and, like perforin-competent RIP-GP mice, showed lymphoid neogenesis in periductal areas in proximity to remaining insulin-producing islet cells (summarized in Table 2). In contrast to this and confirming earlier results (22), LCMV infection of RIP-GP mice with a disrupted perforin gene did not lead to the development of autoimmune diabetes (Table 2). These mice showed a pronounced peri- and intrainsular infiltrate; however, some islet cells remained intact, and lymphoid neogenesis did not develop (Table 2). Taken together, these results show that DC immunization with the appropriate self-antigen leads to destructive autoimmune diabetes accompanied by local formation of organized lymphoid tissues.

A major question in the pathogenesis of autoimmune disorders is to define the triggering events leading to activation of previously indifferent but potentially autoreactive T cells. Expression of the LCMV-GP in pancreatic β cells using the rat insulin promoter (RIP-GP mice) allowed us to study the role of virus infections as possible triggers for breaking tolerance against peripheral self-antigens leading to destruction of pancreatic islets (18, 19). Development of virus-induced autoimmune diabetes in RIP-GP mice is largely dependent on the number of activated self-reactive CTL; e.g., only double-transgenic mice expressing the LCMV-GP in the pancreas and a GP33-specific TCR on many of their CTL (RIP-GP × TCR), but not single transgenic mice (RIP-GP), become acutely diabetic after infection with the less immunogenic LCMV-GP recombinant vaccinia virus, whereas both develop diabetes upon LCMV infection (20). Therefore, it is surprising that immunization of RIP-GP mice with DC constitutively expressing the endogenous peptide led to development of a slow form of clinically manifest and rapidly lethal diabetes, despite the fact that they induced 10–100-fold lower CTL activity and CTLp frequencies compared with LCMV infection (26). Although this lower response corresponded to about that induced by LCMV-GP recombinant vaccinia virus, the data here suggest that the CTL stimulation by repetitive DC immunization was an important prerequisite for the eventual progression to IDDM. Thus, with the number of autoreactive cytotoxic T cells (18, 20), the duration of antigenic stimulation by professional APCs, i.e., the integral of CTL activity over time, determines the disease outcome in this model of autoimmune diabetes. Although perforin usually plays a key role also in the induction of acute diabetes in the RIP-GP model (22), repetitive immunization with GP33-presenting DC overcame the need for perforin in causing diabetes in 45% of the mice.

Cell types normally found in peripheral lymphoid tissues were present in DC-induced islet infiltrates and in periductal lymphoid structures, including T cells, B cells, bone marrow–derived APCs, i.e., DC, and macrophages with a distinct compartmentalization into T and B cell areas. Follicular DC were found exclusively in these newly formed lymphoid structures, suggesting their functional involvement in local B cell responses. Only mice developing early infiltrations and de novo formation of lymphoid follicles developed diabetes, suggesting that autoimmunity is linked to lymphoid neogenesis in the target organ. Similarly, de novo formation of lymphoid tissue can be observed in mice expressing lymphotoxin-α under the control of the RIP (31); however, despite massive pancreatic infiltrates, these mice do not develop diabetes (32). Therefore, it seems likely that the appropriate and repeated activation of CTL, e.g., via DC, together with development of lymphoid structures in the target organs, plays a major role in the development of destructive autoimmunity.

Human IDDM is often associated with chronic infiltration of mononuclear cells in the diseased tissue, leading to tissue destruction (33, 34) and occasional development of lymphoid structures (35). Additional evidence for the possible involvement of DC in the induction of IDDM comes from the nonobese diabetic (NOD) mouse model. In prediabetic NOD mice, DC form an organized network in and around pancreatic islets, facilitating efficient stimulation of infiltrating lymphocytes (36). Taken together, the presented and the discussed findings suggest that DC are not only crucially involved in the initiation of an anti-self immune response but are also essential for the maintenance of autoimmune lesions in peripheral organs. Several antibody-dependent organ-specific autoimmune diseases (including Hashimoto's thyroiditis and IDDM) and the data shown here support the notion that peripheral self-antigen, usually ignored by the immune system, can trigger an immunopathological, autoimmune response if the antigen enters draining organized lymphoid tissues in sufficient quantity and for a sufficient time. This autoimmune response is transient if antigen ceases to reach organized lymphoid tissues. It is maintained if sufficient antigen reaches draining lymphoid organs, or, alternatively, if lymphoid tissue is formed within the target organ itself.

These results may impinge on our rationales to use repetitive immunization with DC against tumors. Due to their high immunostimulatory capacity and their efficient migration into organized lymphoid tissues, autologous DC are well suited for immunization against tumors (15, 16). In mice, antitumor immunity can be induced using DC exogenously pulsed with tumor-derived peptides (37), preincubated with crude tumor cell preparations (38, 39), transfected with tumor cell-derived mRNA (39, 40), infected with recombinant viruses (41, 42), or by fusion of DC with tumor cells (43). By presenting tumor antigens together with other self-antigens by DC or by using tumor antigens that are not exclusively expressed on tumor cells, there is a potential of inducing undesirable anti-self responses, as exemplified in this study. Therefore, it may be important to use antigens solely expressed on tumor cells for DC-based anti-tumor therapy to avoid possible autoimmune complications.

We thank Maries van den Broek, Peter Aichele, and Paul Klenerman for helpful discussions and critical reading of the manuscript, Lenka Vlk and Anne Henzelin for expert technical assistance, and Norbert Wey for excellent microphotography.

This work was supported by the Swiss National Science Foundation, the Deutsche Forschungsgemeinschaft (to B. Ludewig), and the Kanton Zürich.

CTLp

CTL precursor(s)

DC

dendritic cell(s)

GP

glycoprotein

IDDM

insulin-dependent diabetes mellitus

PNAd

peripheral node addressin

RIP

rat insulin promoter

LCMV

lymphocytic choriomeningitis virus