Immunotherapy can be used to induce immunological tolerance by a number of different protocols. During the last decade the ability to use tolerogenic dendritic cells (DCs) to prevent autoimmunity has received much attention. Many studies have attempted to use immature or semi-mature DCs to induce tolerance in the non-obese diabetic (NOD) mouse model of human type 1 diabetes. However, most studies to date have used protocols in which generation of DCs involved a culture step in fetal bovine serum (FBS)-supplemented medium which may affect tolerance induction in a non-specific fashion. Indeed, several studies have shown that DCs cultured in the presence of FBS will induce a powerful T helper type 2 (Th2) immune response towards FBS-related antigens which can suppress an ongoing immune response. Hence, this may interfere with diabetes development in the NOD mouse by induction of immune deviation rather than by antigen-specific tolerance. In order to test whether antigen-specific tolerance induction by DC therapy is feasible in the NOD mouse, we therefore generated immature DCs using autologous serum [normal mouse serum (NMS)-supplemented cultures] instead of FBS, and we show that these DCs can protect NOD mice from diabetes, if pulsed with insulin-peptide antigens before adoptive transfer. Our data therefore support that DC therapy is able to prevent diabetes in the NOD mouse in an antigen-specific manner.
Antigen-specific tolerance induction is the attempt to induce immunological unresponsiveness or active tolerance to one antigen but not another as opposed to more general immunosuppression induced by systemic agents, e.g. blockade of co-stimulation by cytotoxic T lymphocyte antigen–immunoglobulin (CTLA–Ig) or cytokines by anti-tumour necrosis factor (TNF)-α agents. Tolerance induction to one specific antigen can be induced very efficiently by promoting presentation of an antigen in the context of an immature (or semi-mature) dendritic cell (DC), by pulsing immature DCs with an antigen ex vivo, by generating large numbers of immature DCs in vitro or by targeting antigens directly to immature DCs in vivo, e.g. by coupling to anti-CD205 (DEC-205) antibodies [1,2].
In the most widely used model of human type 1 diabetes – the non-obese diabetic (NOD) mouse – a number of studies have shown that therapy with various forms of DCs can protect mice from diabetes development [3–7]. In most cases, however, DCs were generated from bone marrow culture or isolated from tissue and cultured briefly in the presence of fetal bovine serum (FBS). Furthermore, protection from diabetes development was not antigen-specific in the sense that it was not necessary to pulse the protective DCs with any islet-associated antigen. Hence, both unpulsed DCs and DCs pulsed with xenobiotic antigens (e.g. human γ-globulin) could protect from disease [4,7]. Importantly, one exception is a study in which only DCs from pancreatic lymph nodes (panLN) – but not DCs from other anatomical sites – could protect from disease, presumably because panLN-derived DCs would present β cell-derived antigens after adoptive transfer .
While it is formally possible that unpulsed FBS-derived DCs could take up islet-associated antigens in vivo and present these in a tolerogenic fashion, we and others have generated data suggesting an alternative interpretation: DCs cultured in the presence of FBS induce a T helper type 2 (Th2)-dominated immune response towards FBS-derived antigens which could interfere with diabetes development in the NOD mouse, e.g. by suppressing a concomitant Th1 response. Indeed, in these studies DCs presenting FBS-derived antigens are capable of suppressing an otherwise robust anti-lymphocytic choriomeningitis virus (LCMV) virus response [8,9]. Thus, it is likely that DC therapy with FBS-exposed DCs may induce a dominant FBS-specific immune response rather than presenting islet-associated antigens in a tolerogenic and autoantigen-specific manner. This questions the translational value of DC therapy for T1D as DC therapy in the clinic aims ultimately at induction of antigen-specific tolerance.
We have demonstrated recently that by generating DCs in the presence of normal mouse serum (NMS) rather than FBS, immature DCs capable of inducing antigen-specific immunological unresponsiveness can be generated from bone marrow cultures in vitro. The mechanism was shown, at least partially, to involve abortive activation and induction of apoptosis in CD8+ T cells. Furthermore, treatment with these immature DCs was capable of preventing virus-induced diabetes in the rat insulin promoter (RIP)–LCMV model. Here, we present data showing that NMS-derived, interleukin (IL)-10-treated DCs can protect NOD mice from diabetes development, but only if pulsed with antigenic peptides from insulin B chain (InsB15–23 and InsB9–23), whereas unpulsed DCs had no effect on diabetes development.
Materials and methods
Female NOD mice were purchased from Taconic (Ry, Denmark). Diabetes incidence in our colony is approximately 80%. All animal experiments have been conducted according to Danish legislation and have been approved by the Danish Animal Inspectorate and the local ethical review board. Blood glucose was measured by tail vein blood sampling and analysed using the OneTouchUltra system (Lifescan, Milpitas, CA, USA), and mice were considered diabetic upon two consecutive measurements of blood glucose levels >16·6 mmol/l (300 mg/dl). All samples were taken from non-fasting mice.
Generation of DCs and treatment of mice
DCs were generated largely as described [8,10,11]. In brief, bone marrow cells were harvested from the femur and tibiae from female prediabetic NOD mice (5–7 weeks of age) and washed in ice-cold Hanks's balanced salt solution (HBSS) following lysis of red blood cells. T cells were depleted using baby rabbit complement (Harlan Sera Laboratories, Leicestershire, UK) and antibodies from hybridomas against mouse CD4 (RL172·4), CD8 (31M) and Thy1 (AT83). Remaining cells were washed extensively and plated in RPMI-1640 (Invitrogen, Carlsbad, CA, USA) with either 10% FBS (Invitrogen) or 1·5% normal mouse serum harvested from female, prediabetic NOD mice (Taconic M&B) plus 100 U/ml penicillin, 100 µg/ml streptomycin, 10 mM HEPES and 50 µM 2-mercaptoethanol at 4–5 × 106 cells/well in a six-well plate (Nunc, Roskilde, Denmark). Fresh medium was added every other day. Bone marrow DCs developed in the presence of recombinant murine granulocyte macrophage colony-stimulating factor (rmGM-CSF) (Pharmingen, BD, Brondby, Denmark; 20 ng/ml days 0–4 and 10 ng/ml days 5–8) plus 20 ng/ml rmIL-10 (Pharmingen) when indicated (days 0–8). The yield of DCs was similar between the groups and was in the range of 4·5–8·0 × 106 DCs per 107 precursor cells, which is in line with previous reports . The amount of dead cells at the time of harvest was also similar between the different types of DCs and was in the range of 4–12% of the CD11c+ cells by 7-amino-actinomycin D (7-AAD) staining (data not shown). Cells were pulsed with 10 µg/ml InsB9–23 (SHLVEALYLVCGERG) and InsB15–23 peptide (LYLVCGERG) on day 7 as indicated and were harvested on day 8 and washed extensively in PBS. Peptides were purchased from Schaeffer-Nielsen (Copenhagen, Denmark).
NOD mice were treated by weekly injections with pulsed or unpulsed NMS–IL-10 DCs from 5 to 12 weeks with age with 106 DCs intraperitoneally (i.p.) in 100 µl phosphate-buffered saline (PBS).
For flow cytometric analysis, DCs were harvested on day 8 after set-up unless indicated otherwise. Cells were washed in PBS and stained with PBS/1%FBS in the presence of anti-CD16/CD32 to block Fc receptors. The following antibodies were used: CD11c-biotin plus streptavidin-allophycocyanin (APC), CD11b-phycoerythrin (PE), CD40-PE, CD86-fluorescein isothiocyanate (FITC), H2Kd-FITC, OX6-PE (which cross-reacts with I-Ag7 in the NOD), CD54-FITC, CD3e-PE, CD8-FITC, CD4-peridinin chlorophyll (PerCP), CD19-PE and Gr1-FITC and appropriate isotype controls, all from BD Pharmingen. Furthermore, anti-ICOSL-PE, CD45·2-FITC, F4/80-PE and 33D1-PE from eBioscience (San Diego, CA, USA) was used. Cells were stained for 1 h at 4°C and then washed in PBS/1%FBS before being analysed on a LSRII flow cytometer (Becton-Dickinson, Brondby, Denmark). Compensation was determined by using compensation beads bound to the appropriate fluorochromes as described by the manufacturer (Becton-Dickinson). Samples were analysed using Becton Dickinson FACS Diva software, version 6·0.
For analysis of regulatory CD4+CD25+forkhead box P3 (FoxP3+) cells, blood samples were stained with antibodies against CD4, CD25 and FoxP3, using the regulatory T cell detection kit according to the manufacturer's recommendations (eBioscience). In the case of intra-pancreatic cells, pancreata were forced through a cell-strainer, and leucocytes were purified using Lympholyte-M (Cedarlane, Burlington, ON, Canada). Then, cells were stained for CD4, CD25 and FoxP3 as described above. FoxP3+ cells were determined by comparing FoxP3 stained cells with isotype-control stained cells.
Measurement of insulin antibodies
Insulin antibodies were measured with a 96-well filtration plate micro insulin autoantibody assay, largely as described previously , on serum samples of treated and untreated mice at 9, 13 and 20 weeks of age.
Induction of diabetes by adoptive transfer
Spleen cells were procured from protected mice treated with InsB-peptide pulsed NMS–IL-10 DC at termination of study (‘protected’, 40 weeks of age), from prediabetic, young NOD mice (‘prediabetic’) or from new-onset diabetic NOD mice (‘diabetogenic’). Single cells were generated by mashing spleens through a cell strainer in PBS, followed by lysis of red blood cells; 10 × 106 cells of each of the cell-pools in PBS were transferred i.p. to 8-week-old NOD/severe combined immunodeficiency (SCID) mice as indicated. Diabetes development was monitored as described above.
Immunohistochemistry for insulin and glucagon was performed largely as described previously . All sections were counterstained with haematoxylin.
For detection of insulin on frozen sections, tissue was embedded in TissueTek, snap-frozen on dry ice and cut into 6-µm sections, fixed in cold (−20°C) 96% ethanol for 15 min, washed in PBS and blocked using FBS and normal goat serum. Slides were then blocked using an avidin/biotin kit (Dako, Glostrup, Denmark) and stained for insulin (guinea-pig anti-insulin; #A0564, Dako). As secondary antibody, we used biotinylated goat anti-guinea pig antibody (Vector #BA7000, Burlingame, CA, USA) and colour reaction was obtained by sequential incubation with avidin–peroxidase conjugate (Vectastain Elite, Vector, Burlingame, CA, USA) and diaminobenzidine–hydrogen peroxide. Fluorescent detection of CD4 or CD8 and insulin was performed as described previously . In order to classify the degree of islet infiltration, islets (≥10 from three mice in each group) were scored as follows: 0, intact islets; 1, insulin-positive islets with peri-insulitis; 2, insulin-positive islets with invasive insulitis; 3, islets without detectable insulin-staining but with invasive insulitis; and 4, end stage-islets.
For detection of FoxP3+ T cells, frozen sections were prepared as described above, but were fixed in cold (−20°C) acetone with 1% hydrogen peroxide for 15 min. Sections were washed in PBS and blocked with 5% donkey and goat serum in PBS/0·25% bovine serum albumin (BSA). Slides were then blocked with avidin/biotin as described above, and anti-FoxP3 antibody (FJK-16s) or isotype-matched control antibody was applied for 2 h at room temperature at 5 µg/ml. Then, secondary antibody (biotinylated goat anti-rat was applied, and colour reaction was obtained by sequential incubation with avidin–peroxidase conjugate (Vectastain Elite) and diaminobenzidine–hydrogen peroxide. Next, primary antibody against the T cell receptor (anti-TCR-FITC, H57-597; BD Pharmingen) or isotype control was applied for 2 h at room temperature. Finally, samples were washed, nuclei were counterstained using Hoechst (Invitrogen) and mounted. Images were analysed using Adobe Photoshop (San Jose, CA, USA). Control slides stained with isotype control for either FoxP3 or TCR antibodies were always blank.
Generation of immature DCs in the NOD mouse
Our aim was to generate FBS-free DCs capable of preventing development of diabetes in the NOD mouse model. We therefore wanted to generate immature DCs without the use of FBS. As demonstrated earlier, FBS-free DCs can be generated from bone marrow precursors by culturing in granulocyte–macrophage colony-stimulating factor (GM-CSF) for 6–8 days if FBS is replaced by NMS in the culture medium. However, because defects in DC development in the NOD mouse have been reported earlier [14,15], we first tested how NOD DCs developed from bone marrow cultures in vitro. We compared DCs generated in the presence of FBS or NMS and analysed the presence of DCs by flow cytometry. As demonstrated in Fig. 1a, NMS-supplemented cultures supported development of CD11c+ DCs after 6–8 days, which was comparable to FBS-supplemented cultures where CD11c+ DCs started to develop a little earlier (e.g. after 4–6 days). No significant contamination by Gr-1+ granulocytes, CD19+ B cells or CD3+ T cells could be detected. The small increase in Mac-3+ cells at day 8 in both culture systems was a weak staining of a subset of CD11c+ cells (data not shown). To show that CD11c+ cells expressed other relevant DC markers, we stained day 8 DCs from both FBS- and NMS-supplemented cultures for CD40, CD86 and major histocompatibility complex (MHC)-II (Fig. 1b). As shown, CD11c+ cells from FBS-supplemented cultures expressed low levels of CD40, CD86 and MHC-II, which could be up-regulated after incubation with lipopolysaccharide (LPS) for 24 h (Fig. 1b). In contrast, CD11c+ cells from NMS-supplemented cultures had a more immature phenotype with lower expression of CD40 and CD86, and very few cells being MHC-IIhigh. As described earlier, NMS-derived DCs were somewhat maturation-resistant as the cells only up-regulated CD40, CD86 and MHC-II to a limited extent .
Because our aim was to generate immature DCs capable of preventing diabetes in the NOD mouse we next generated NMS-derived DCs in the presence of IL-10, as this has been shown by us and others to prevent maturation of DCs [16,17]. NMS-derived IL-10 DCs were CD11chighCD11b+F4/80+H2Kd+I-Ag7lowICOSLdimCD40-CD86-CD54+ (Fig. 1c), which is consistent with immature DCs. Expression of MHC-II on NOD DCs from NMS-supplemented IL-10-treated cultures was lower than on NOD DCs generated without IL-10 in both FBS- or NMS-supplemented cultures, showing that IL-10 ensured a more immature state of the cells (data not shown). No expression of CD3, CD4, CD8, CD19, 33D1 or Gr-1 could be detected (data not shown).
Prevention of diabetes development in the NOD mouse by treatment with immature DCs
As discussed above, DC therapy has been tested with success in the NOD mouse before; however, often with little antigen-specificity. We wanted to use DCs pulsed with a relevant autoantigen to prevent diabetes development in the NOD mouse. Recent evidence has pointed to the insulin epitopes B9–23 and B15–23 to be pivotal in diabetes development in the NOD mouse . We therefore treated prediabetic NOD mice with weekly injections of NMS–IL-10 DCs pulsed with the insulin peptides B9–23 and B15–23 from weeks 5 to 12. As a control, NOD mice were injected either with PBS or unpulsed NMS–IL-10 DCs. As shown in Fig. 2, only NMS–IL-10 DCs pulsed with insulin peptides protected mice from diabetes, whereas diabetes incidence was high in mice injected with PBS or unpulsed NMS–IL-10 DCs. Thus, DCs generated in the presence of IL-10 in NMS-supplemented cultures could prevent diabetes developing in an antigen-specific manner.
Preservation of β cell-containing islets in DC-treated NOD mice
Histological analysis of pancreatic sections of protected NOD mice treated with InsB-peptide pulsed NMS–IL-10 DCs showed the presence of insulin+ islets surrounded by peri-insulitic infiltrates (Fig. 3e–h). This was observed in seven of seven mice. In contrast, seven of seven diabetic PBS- or NMS–IL-10-treated mice had no or extremely few insulin+ islets left, and inflammation was rarely observed (Fig. 3a–d). Thus, protection from diabetes by treatment with InsB-peptide-pulsed NMS–IL-10 DCs correlated with preservation of insulin-containing β cells in the islets of Langerhans.
Presence of FoxP3+ T cells in peri-insulitic infiltrates in DC-treated NOD mice
To analyse further the content of the peri-insulitic infiltrates in protected mice, we stained sections for CD4 and CD8 or T cell receptor and FoxP3. Both CD4+ and CD8+ cells were present in the inflammatory infiltrates surrounding the islets in protected mice, but CD4+ T cells were most abundant (Fig. 4a,b). When these infiltrates were double-stained for T cell receptor and the regulatory T cell-specific transcription factor FoxP3, it was evident that a large number of T cells expressed FoxP3 in all protected mice analysed (four of four); see Fig. 4c. Thus, protection of NOD mice from diabetes development by treatment with InsB-peptide pulsed NMS–IL-10 DCs correlates with peri-insulitic infiltrates, consisting largely of FoxP3+ T cells.
Adoptive transfer and analysis of autoantibodies in DC-treated NOD mice
We wanted to test whether diabetes protection in NMS–IL-10 DC-treated mice could be transferred, which would indicate the presence of circulating regulatory T cells. We therefore isolated spleen cells from protected mice and transferred the cells into NOD/SCID mice together with diabetogenic spleen cells from regular recent-onset diabetic NOD mice. However, as shown in Fig. 5a, cells from protected NOD mice treated with NMS–IL-10 DCs pulsed with InsB peptides could not prevent diabetes development; if anything, diabetes development was faster in mice transferred with both diabetogenic cells and cells from protected mice. Furthermore, mice transferred only with cells from protected mice also developed diabetes quickly. This result is in accordance with measurements of circulating regulatory T cells as the percentage of CD4+CD25+FoxP3+ regulatory T cells was similar in NMS–IL-10 DC/InsB-treated mice and PBS-treated mice when measured 2 weeks after the last DC injection (age 13–14 weeks, Fig. 5b). Thus, we conclude that protection from diabetes induced by treatment with InsB-peptide pulsed NMS–IL-10 DCs does not lead to increased numbers of circulatory regulatory T cells which can protect from adoptive transfer-induced diabetes.
Levels of insulin autoantibodies in DC-treated NOD mice
It was possible that protection of NOD mice induced by InsB-peptide pulsed NMS–IL-10 DCs could also affect development of humoral immunity. In the NOD mouse, insulin antibodies are induced during diabetes development and are often predictive for disease, but some protocols which protect from diabetes development also induce increased amounts of insulin antibodies [12,19]. Interestingly, as shown in Fig. 6, treatment with InsB-peptide pulsed NMS–IL-10 DCs did not protect against induction of insulin autoantibodies; if anything, treatment with NMS–IL-10 DCs increases the levels of and number of animals with detectable anti-insulin antibodies, especially at 13 and 20 weeks of age. This is in accordance with data showing that prevention of diabetes with InsB9–23 in incomplete Freund's adjuvant (IFA) induces insulin autoantibodies, as described earlier .
Our study shows that it is possible to generate immature DCs from bone marrow cultures in NOD mice and that these immature DCs can protect from diabetes development in an antigen-dependent fashion. We chose to use the insulin B-chain-derived peptides B9–23 and B15–23 as antigens because these have been shown recently to be essential in the development of type 1 diabetes in the NOD mouse, i.e. changing this particular epitope resulted in protection from insulitis and diabetes whereas another autoimmune manifestation – sialitis – was unaffected . Furthermore, others have demonstrated T cell reactivity to these two epitopes in prediabetic NOD mice [20,21].
Replacement of FBS with NMS in DC cultures in vitro enabled us to exclude any contributions from FBS-derived antigens to the protective effect of NMS–IL-10 DCs. Indeed, previous studies have shown that NMS–IL-10 DCs generated by this protocol do not induce any FBS-specific immune response, whereas FBS-derived DCs do . Previous studies in the NOD mouse have shown that FBS-derived DCs could protect against diabetes development in the NOD mouse in an autoantigen-independent manner [3–7], but subsequent analyses have also shown a significant induction of Th2 cytokines, including IL-4, IL-5 and IL-10 in at least one of the studies. Thus our data, combined with a recent study showing that protection of NOD-mice by treatment with FBS-derived DCs correlated with a strong Th2-type humoral and cellular immune response to FBS antigens , strongly support the interpretation that treatment of NOD mice with FBS-exposed DCs results in a Th2-type anti-FBS response which affect disease development; however, the translational implications of these findings are not entirely clear.
The requirement for IL-10 in the generation of NMS–IL-10 DCs has not been addressed directly in this study. However, earlier results from our group and others indicate that IL-10 functions to keep the DCs in an immature state . Specifically for NMS DCs, we have shown that IL-10 treatment conveys a more immature phenotype which results in lower levels of maturation markers (MHC-II, CD40, CD86) and lower stimulatory capacity towards naive T cells . Also, we have shown previously that NMS–IL-10 DCs can induce CD8+ T cell unresponsiveness by inducing abortive activation of CD8+ T cells followed by apoptosis and deletion, and it is therefore likely that NMS–IL-10 DCs pulsed with InsB15–23 may also mediate deletion of CD8+ T cells specific for this epitope . However, due to the extremely low frequency of these cells in the blood we did not attempt to detect these cells . Indeed, it has been shown previously that deletion of both the dominant CD4 and CD8 epitope is sufficient to prevent diabetes development in the NOD mouse , which fits well with our observation that protection against diabetes by DC vaccination was dependent upon antigen-pulsing with InsB-peptides. We were not able to detect any increase in the levels of circulating CD4+CD25+FoxP3+ regulatory T cells after DC therapy, and protection from diabetes could not be transferred to NOD/SCID mice. This would indicate that DC therapy did not induce infectious tolerance – neither by increasing the number of circulating CD4+CD25+ regulatory T cells nor by inducing another spleen-residing cell type capable of transferring protection against diabetes development (e.g. IL-10-producing Tr1 cells or CD8 suppressor cells). Alternatively, it is possible that insulin-specific regulatory T cells induced by DC therapy were present in circulation in too-low numbers to mediate protection in adoptive transfer experiments . In favour of the latter hypothesis is the accumulation of FoxP3+ T cells around the islets of protected NOD mice. When analysed by flow cytometry, the fraction of CD4+CD25+FoxP3+ cells was 6·4 ± 2·4% of all CD4+ cells in the pancreas (data not shown); however, due to the low number of regulatory T cells which can be retrieved from the islets, we have not yet been able to demonstrate this formally by adoptive transfer experiments.
Although the contribution of FBS-derived antigens may be problematic for DC therapy to prevent diabetes in the NOD mouse, it has been demonstrated that for other animal models of human autoimmune diseases, FBS-derived antigens do not contribute greatly. One example is in experimental autoimmune encephalomyelitis (EAE), which is a mouse model with some resemblance to human multiple sclerosis. Here, antigen-dependent tolerance-induction is possible, even by using FBS-derived DCs. Thus, although FBS-derived DCs also induced Th2 cytokines in this system, protection from EAE was obtained only if the DCs were pulsed with a relevant autoantigen before injection [24,25]. This is in contrast to the situation in the NOD mouse and to two examples of virus infections – respiratory syncytial virus  or lymphocytic choriomeningitis virus – where FBS DCs clearly affected the immune response. The reasons for this discrepancy are unknown, but may reflect that the development of autoimmune diabetes or anti-viral immunity is more sensitive to a concomitant Th2-dominated response towards FBS. It is also possible that the sensitivity towards priming by DCs to FBS-derived antigens is dependent on the magnitude of the FBS-specific response and/or the timing relative to other responses.
Taken together, we would argue that it is important to consider the right controls when DC therapy is tested in preclinical animal models. If it is not possible to avoid the use of FBS-exposed DCs altogether, e.g. by using NMS-derived DCs or freshly isolated splenic DCs, a group of animals treated with unpulsed DCs should at least be included in the treatment groups so that any effects of FBS immunization can be tested experimentally.
The data presented in this study support DC therapy as a means to induce antigen-specific tolerance for treatment of T1D and also, perhaps, other autoimmune diseases. However, as human T1D patients are identified most often after disease onset, it will be important to establish and test protocols in which recent-onset diabetic NOD mice are treated with tolerogenic DCs to test whether such a treatment protocol is able to restore normoglycaemia. Protocols are under way to test the feasibility of this. If so, it is possible that DC therapy could contribute to the treatment of T1D, possibly in combination with other treatment therapies, e.g. monoclonal antibodies to anti-CD3 where the combination of antigen-specific therapy together with general immune modulation has proved efficient .
The authors wish to thank Rose Andresen, Trine Larsen, Stine Bisgaard and Mie Berndorff for excellent technical assistance.