Anti-CD3/Anti-CXCL10 Antibody Combination Therapy Induces a Persistent Remission of Type 1 Diabetes in Two Mouse Models

Type 1 diabetes (T1D) is a serious autoimmune-mediated disease characterized by the progressive destruction of insulin-producing β-cells in the islets of Langerhans in the pancreas. Several attempts have been made to block the autoimmune destruction of these β-cells. One of the most promising therapies targets T cells using anti-CD3 antibodies such as hOKT3γ1, teplizumab, and otelixizumab (also known as ChAglyCD3), which have been evaluated in several clinical trials of patients with new and recently diagnosed T1D, including the DEFEND-1 study (otelixizumab) and the Protégé study (teplizumab) (1–4). Similar to preclinical animal models (5,6), treatment with anti-CD3 antibodies was effective in decelerating the pathogenesis of T1D in clinical trials (1,7,8). Administration of anti-CD3 antibodies results in the inactivation of conventional T cells and the expansion of previously constrained regulatory T-cell (Treg) populations (9,10). Unfortunately, in the majority of treated patients, prevention of the decline in β-cell function lasted no longer than 1–2 years (1,2,8,11,12). In addition, many patients did not respond to the treatment, for unknown reasons (13).

Several combination therapies (CTs) to achieve long-term protection in the majority of patients with T1D have been assessed in preclinical models. In addition to anti-CD3 antibodies, several immunomodulatory agents have been used, including administration of nasal proinsulin (14), Lactococcus lactis–secreting interleukin (IL)-10/proinsulin (15), cyclosporine A and vitamin D3 analog (TX527) (16), IL-1 receptor antagonist (17), anti-CD20 antibody (18), fingolimod (FTY720) (19), the selective sphingosine 1 phosphate 1 modulator ponesimod (20), dipeptidyl peptidase-4 inhibitor MK626 (21), and HSP60 peptide p277 (22). Many of these CTs were superior to monotherapies. However, with the exception of CTs with anti-CD3 antibody/fingolimod tested in the LEW.1AR1-iddm rat model (19) and anti-CD3 antibody/ponesimod investigated in the NOD mouse model (20), none included blocking cell migration into the islets. To avoid side effects associated with traditional immunosuppressive drugs, such as cytostatic drugs or glucocorticoids, the anti-CD3 antibody therapy aimed at the short-term deletion/inactivation of T cells. However, the T-cell repertoire regenerates; therefore, one major problem is the reentry of regenerated autoaggressive lymphocytes into the islets of Langerhans.

Here we present data from a CT with a non-Fc-binding anti-CD3ε F(ab′)2 fragment (clone 145–2C11) (aCD3) (6) and a neutralizing anti-CXCL10 antibody (clone 1F11) (aCXCL10) (23). The chemokine CXCL10 (IP-10, a 10-kDa interferon [IFN]-γ–inducible protein) has been demonstrated to play a key role in the pathogenesis of T1D in animal models (24–27) and is elevated in islets of patients with T1D (28,29). As an inducible model for T1D, we used the RIP-LCMV-GP model, in which transgenic mice express the glycoprotein (GP) of the lymphocytic choriomeningitis virus (LCMV) under the rat insulin promoter (RIP) in the β-cells of the islets of Langerhans (30). In addition, we used the NOD mouse as a model for spontaneous T1D. We could demonstrate in both diabetic RIP-LCMV-GP and NOD mice that partial and temporal T-cell inactivation, followed by a blockade of CXCL10-mediated T-cell migration, resulted in the persistent remission of T1D. CT was superior to monotherapy in both models. A detailed analysis of the frequency and activity of islet antigen–specific T cells in the spleen and pancreas of reverted RIP-LCMV-GP mice revealed that T cells indeed recovered after aCD3 treatment but were prevented from islet reentry by CXCL10 neutralization. This observation was also supported by data obtained in the NOD model following islet-antigen peptide mimotope NRP-V7–specific T cells. Our data suggest that CT with aCD3 and aCXCL10 might constitute a novel therapy for patients with T1D.

H-2^b RIP-LCMV-GP transgenic mice were generated and screened by PCR as previously described (30,31). CXCL10 mice were generated as previously described (32) and have been backcrossed to C57BL/6 mice for more than 10 years. NOD mice were from The Jackson Laboratory and were bred in the local breeding facility of the Georg-Speyer Haus, Frankfurt, Germany. Approximately 60% of female NOD mice developed T1D within 14 to 26 weeks of age (data not shown). LCMV Armstrong clone 53b was produced as described elsewhere (31). Animal experiments were approved by the local Ethics Animal Review Board, Darmstadt, Germany (V54–19c20/15-F143/56). Blood glucose (BG) was monitored at weekly intervals using a dynaValeo glucometer from dynamiCARE. Animals with BG concentrations >300 mg/dL were considered to be diabetic (33).

Armenian hamster anti-mouse CXCL10 monoclonal antibody (clone 1F11) (23) was purified from hybridoma cell supernatant on a HiTrap Protein G HP column. Monoclonal Armenian hamster anti-CD3 antibody (145–2C11 F(ab′)2 fragment, pepsin digested) was obtained from Bio X Cell (Be0001–1FAB 4294/0212; West Lebanon, NH). Armenian hamster IgG (MBL/Biozol, Eching, Germany) was used as an isotype-matched control monoclonal antibody for the anti-mouse CXCL10 antibody.

Tissues were immersed in Tissue-Tek OCT and quick-frozen on dry ice; 7-μm tissue sections were cut, then fixed in ethanol or ethanol/acetone (1:1) at −20°C. Primary antibodies used include rat anti-mouse CD8a and rat anti-mouse CD4 (BD Biosciences) and a polyclonal guinea pig anti-swine insulin antibody (DakoCytomation). Images of pancreas sections were acquired with an Axioscope 2 microscope (Zeiss).

Tissues were immersed in Tissue-Tek OCT and quick-frozen on dry ice; 7-μm tissue sections were cut and fixed in ethanol at −20°C. Rat anti-mouse FoxP3 (eBioscience) and goat anti-rat Alexa Fluor 594 (Invitrogen) antibodies were used for detection of FoxP3+ cells; directly conjugated rat anti-mouse CD8a-FITC and rat anti-mouse CD4-FITC (Southern Biotech) antibodies also were used. Images were acquired with a confocal microscope (Zeiss LSM 510 META).

The pancreas was extracted and 3 mL of collagenase P solution (1.2 U/mL in RPMI 1640) were injected. After 30 min of incubation at 37°C, the collagenase solution was removed and cold RPMI 1640 containing 20 µg/mL DNase I was added. The pancreas was shaken for 1 min and pressed through a 70-µm cell strainer. The suspension was washed with RPMI 1640 containing 20 µg/mL DNase I and resuspended again in RPMI containing 20 µg/mL DNase I. Then, 40% Ficoll in PBS was overlayed with the cell suspension and a gradient was performed. The pellet was washed with RPMI 1640 and resuspended in RPMI 1640 containing 10% FCS.

Single-cell suspensions of spleen and pancreatic draining lymph nodes (PDLNs) were stimulated overnight with 2 μg/mL LCMV peptides GP33 (CD8) and GP61 (CD4), or with 2 μg/mL of the NOD islet-antigen peptide mimotope NRP-V7 (34) in the presence of Brefeldin A. Cells were stained for surface expression of CD8 and CD4 and fixed, permeabilized as previously described (33), and stained for intracellular IFN-γ and FoxP3. V450-conjugated rat anti-CD4 antibody, allophycocyanin-Cy7-conjugated rat anti-CD8 antibody, and allophycocyanin-conjugated rat anti–IFN-γ antibody all were obtained from BD Biosciences. The phycoerythrin-conjugated rat anti-FoxP3 antibody was obtained from eBioscience. Samples were acquired with a FACSCanto II flow cytometer (BD Biosciences).

In vivo cytotoxicity assay was performed as previously described (35). Briefly, splenocytes from C57BL/6 mice were divided into two groups. One group was pulsed overnight with 2 µg/mL LCMV-GP33 peptide. Peptide-pulsed cells were labeled at a final concentration of 0.5 μmol/L (carboxyfluorescein succinimidyl ester [CFSE]^lo) and unpulsed control cells at 5 μmol/L (CFSE^hi). Equal amounts (1.5 × 10⁷) of CFSE^hi and CFSE^lo cells were mixed and injected intravenously into recipient mice. Specific in vivo cytotoxicity was determined by collecting blood and assessing the amounts of differentially CFSE-labeled target cell populations by flow cytometry (FACSCanto II; BD Biosciences). The data obtained at different times after target cell injection were normalized against the ratio between CFSE^lo and CFSE^hi cells detected 10 min after transfer.

Diabetes incidence curves were analyzed using the Mantel-Cox log-rank test. T-cell frequencies and counts were analyzed using the unpaired, two-tailed t test (Prism software version 5.02; GraphPad).

To find the optimal window for aCXCL10 administration, we first performed a single round of aCD3 therapy in diabetic RIP-LCMV-GP mice. At days 7 to 9 (when T-cell response peaked) and days 10 to 12 (the start of diabetes with BG concentrations >300 mg/dL) after LCMV infection, RIP-LCMV-GP mice received three daily intravenous injections of either 3 or 30 µg aCD3. One day after the final aCD3 injection, the frequency of CD4 T cells was significantly decreased in the blood, spleen, pancreatic lymph nodes, and pancreas of mice receiving aCD3 at days 10–12 after infection (Fig. 1A and Supplementary Fig. 1). Whereas administration of 30 µg aCD3 almost completely depleted CD4 T cells, injection of 3 µg resulted in a reduction of 40–60% (Fig. 1A and Supplementary Fig. 1). By contrast, the reduction of CD8 T cells was much less pronounced (Fig. 1A and Supplementary Fig. 1). These data confirm recent observations that differential expression of CD3 on T-cell subsets influences aCD3-induced T-cell depletion (36). Importantly, the aCD3 treatment affected the T cells in the pancreas and reduced the local frequency of both CD8 and CD4 T cells (Fig. 1A and Supplementary Fig. 1). Our data also confirm that the frequency of FoxP3+ Tregs is increased after aCD3 treatment (Fig. 1A). Importantly, we also found that, after aCD3 treatment, the total number of islet antigen (LCMV-GP)–specific T cells in the pancreas was significantly decreased and their frequency was reduced in the spleen (Fig. 1A). Examination of sections of pancreas tissue obtained 1 day after the last aCD3 injection revealed that administration of both 3 and 30 µg of aCD3 reduced insulitis and preserved the production of insulin (Supplementary Fig. 2). Administration of aCD3 at days 7–9 after LCMV infection had a less pronounced effect on the T cells’ frequency, possibly because of the relatively high number of T cells present at that time (Supplementary Fig. 1A).

To avoid a general immune suppression, we used an aCD3 dose of 3 µg for the following CT: Diabetic mice received three intravenous injections of 3 µg aCD3 on days 10–12 after infection with LCMV, followed by two initial intraperitoneal injections of 100 µg aCXCL10 on days 13 and 14 after infection, then six additional intraperitoneal injections of 100 µg aCXCL10 from day 17 to 28 after infection (three injections per week). The last aCXCL10 injection was given on day 28 after infection (Fig. 1B). CT resulted in a remission of T1D in more than 60% of diabetic mice (Fig. 1C). This remission was significantly greater than the reduction of T1D after single therapy with either aCD3 (38%) or aCXCL10 (36%) (Fig. 1C). At days 12 (last aCD3 injection) and 35 (1 week after the end of CT), the majority (70%) of mice treated with CT showed a reduced BG concentration, and numerous diabetic mice remitted to normoglycemia (BG <200 mg/dL) (Fig. 1D). By contrast, BG concentrations of most mice receiving no or only aCXCL10 monotherapy increased from days 12 to 35 (Fig. 1D). Mice receiving aCD3 monotherapy displayed a mixed outcome; a large fraction of mice (47%) showed increased BG concentrations, but a significant fraction (47%) also presented with decreased BG concentrations (Fig. 1D).

The remission of T1D and the blockade of further destruction of β-cell mass was reflected by the degree of insulitis at day 31 after LCMV infection. In untreated mice, islet infiltration and β-cell destruction progress over time, leaving almost no functional islets behind (Fig. 2A and B). By contrast, mice receiving CT showed a reduced degree of insulitis and retained more functional β-cells (Fig. 2A and B). Similarly, a single administration of aCD3 resulted in the delay of islet destruction and decreased insulitis (Fig. 2A and B). Assessment of islet integrity in mice that did not develop T1D until the end of the observation period (day 182 after infection) revealed that islets of mice treated with CT were largely devoid of cellular infiltrations, whereas islets of mice that received aCD3 alone displayed substantial insulitis, although they did not display hyperglycemia (Fig. 2C). It has to be noted here that only nondiabetic mice in the aCD3 and CT groups were analyzed at day 182 after infection. Isotype control and aCXCL10-treated mice developed severe T1D and had to be killed much earlier because of severe T1D. These data indicate that administration of aCXCL10 in addition to aCD3 results in a persistent reduction of cellular recruitment to the islets, thereby blocking the destructive process and reversing T1D.

The observed T1D remission was associated with a reduced presence of islet antigen–specific T cells in the pancreas. T-cell frequencies were analyzed at days 20 and 31 after infection (Fig. 3). At the end of CT (day 31 after infection), the frequency of islet antigen–specific CD8 T cells was decreased by approximately 73% in the spleen of CT-treated mice compared with the spleen of isotype control mice (Fig. 3A and C). aCD3 monotherapy reduced the frequency by 56%, and aCXCL10 monotherapy had no significant effect (Fig. 3A and C). By contrast, total CD8 and CD4 T cells recovered completely in all treatment groups. Importantly, in the pancreas of mice receiving CT, the total number of islet antigen–specific T cells was also diminished by 78% compared with isotype control animals (Fig. 3A and C). Monotherapy with aCD3 or aCXCL10 reduced the frequency by only 23% or had no effect, respectively (Fig. 3A and C). Furthermore, the number of pancreas-infiltrating total CD8 and CD4 T cells was lower after CT, indicating that decreased overall inflammation of the pancreas also results in the reduced recruitment of bystander T cells (Fig. 3A and C). The observed effect was already detectable at day 20 after infection (i.e., after four of eight aCXCL10 injections). At that time, CD4 T cells had not yet completely recovered from the aCD3 treatment, whereas CD8 T-cell levels were already similar to those in isotype control mice (Fig. 3B). Islet antigen–specific CD8 T cells were already significantly diminished in the spleen of CT- and aCD3-treated mice (Fig. 3B). Importantly, in the pancreas of CT-treated mice, the total number of islet antigen–specific CD8 T cells is significantly reduced by 81% (Fig. 3B). At day 20 after infection, aCD3 or aCXCL10 monotherapy resulted in a reduction of islet antigen–specific CD8 T cells by 66% or 49%, respectively. Thus, when comparing the total number of LCMV-GP33–specific CD8 T cells in the pancreas at days 20 and 31, it becomes apparent that a reentry of cells is prevented only in CT-treated mice, not in mice receiving aCD3 or aCXCL10 monotherapy (Fig. 3B and C). LCMV-specific CD4 T cells play only a minor role in the RIP-LCMV-GP model, and the number of LCMV-GP61–specific CD4 T cells was below 1% in these experiments (Supplementary Fig. 3). Nevertheless, we could detect a reduction in the frequency of LCMV-GP61–specific CD4 T cells after treatment with aCD3 and CT (Supplementary Fig. 3).

In contrast to the absolute number of total CD4 T cells, the number of FoxP3+ CD4 T cells was not significantly reduced in the pancreas after treatment with either antibody or CT (Fig. 3B and C). Hence, calculating the ratio between islet antigen–specific CD8 T cells and FoxP3 CD4 T cells reveals a clear shift toward a regulatory milieu. In particular, at day 31 after infection, the pancreas of CT-treated mice shows a significant three- to fourfold elevation of the FoxP3+ CD4 versus islet antigen–specific CD8 T cells compared with mice receiving monotherapy or isotype control treatment (Fig. 3D). A tendency toward a more regulatory milieu could also be detected in the spleen of CT-treated mice (Fig. 3D).

After CT and aCD3 monotherapy, the remaining islet antigen–specific CD8 T cells found in the pancreas at day 20 after infection show little activity. They produced less IFN-γ (Fig. 3E and F), and only a small fraction of IFN-γ–producing T cells also generated tumor necrosis factor (TNF)-α upon stimulation with the LCMV epitope GP33 (Fig. 3E and F). By contrast, in the pancreas of untreated or aCXCL10-treated mice, a large fraction (>30%) of islet-specific CD8 T cells also produced TNF-α (Fig. 3E and F). In the spleen of isotype control mice, almost 80% of islet antigen–specific CD8 T cells generated both IFN-γ and TNF-α. There was only a slight decrease of such T cells in CT-treated mice and no significant difference in aCD3- or aCXCL10-treated mice (Fig. 3F). Such a low frequency of TNF-α–producing T cells in mice receiving CT or aCD3 monotherapy indicates that most of the islet antigen–specific CD8 T cells were indeed newly regenerated, since, in contrast to experienced T cells, newly activated CD8 T cells predominantly express IFN-γ and only small amounts of TNF-α (33,37–39).

The overall cytotoxicity to islet antigen peptide–loaded target cells was reduced in CT-treated mice. At day 31 after infection, we performed an in vivo cytotoxicity assay using differentially CFSE-labeled splenocytes as target cells. LCMV-GP33–loaded CFSE^lo splenocytes were mixed with an equal amount of unloaded CFSE^hi splenocytes and were transferred into LCMV-infected mice treated with CT, aCD3, aCXCL10, or isotype control. The ratio of CFSE^lo to CFSE^hi splenocytes was determined at several time points after the adoptive transfer of the target cells. In isotype control and aCXCL10-treated mice, a considerable fraction of epitope-loaded target cells was killed within 6 h, and most of the target cells were gone after 24 h (Fig. 4A and B). In CT- and aCD3-treated mice, the killing was less efficient and delayed. Almost no target cells were lost after 6 h, and a large proportion survived for more than 24 h (Fig. 4A and B). Evaluation of the half-life of the epitope-loaded target cells revealed a twofold longer mean survival time in CT-treated mice, indicating the reduced overall cytotoxic potency of the islet-specific immune response (Fig. 4B and C).

To further demonstrate that a lack of the key chemokine CXCL10 improves the efficacy of anti-CD3 therapy, we treated CXCL10-deficient RIP-LCMV-GP mice with aCD3. For this experiment, regular RIP-LCMV-GP mice (30) were crossed with CXCL10-deficient mice (32); homozygous CXCL10^−/− × RIP-LCMV-GP and CXCL10^+/+ × RIP-LCMV-GP littermates were infected with LCMV and were treated at days 10–12 with three doses of 3 µg aCD3 or were left untreated. Whereas aCD3 treatment of RIP-LCMV-GP mice resulted in protection similar to that gained before (∼30% remission; compare with Fig. 1), all diabetic CXCL10^−/− × RIP-LCMV-GP mice reverted to a nondiabetic state within 8 weeks after aCD3 administration and remained free of diabetes until the end of the observation period at week 28 after infection (Fig. 5). It has to be noted here that the frequency of T1D was also reduced in untreated CXCL10^−/− × RIP-LCMV-GP mice and that many of these mice displayed only a mild form of T1D, allowing remission over time after LCMV infection (Fig. 5). These data indicate that the impact of a total absence of CXCL10 on the course of T1D in aCD3-treated mice is even stronger than a neutralization with aCXCL10.

The incidence data from the inducible RIP-LCMV mouse model were confirmed in the spontaneous NOD mouse model. Diabetic female NOD mice were treated with aCD3, isotype control, or CT within 1 week after becoming diabetic (BG >300 mg/dL). Dose-finding studies with NOD mice revealed that, in contrast to the RIP-LCMV-GP mice, a dose of 30 µg aCD3 was required to induce remission from T1D in a fraction of the mice. However, even at a dose of 30 µg, large clusters of infiltrating T cells remained around the islets of Langerhans (Supplementary Fig. 4), which stands in contrast to the situation in aCD3-treated RIP-LCMV mice (Supplementary Fig. 2). Similar to data from earlier studies (14,15,21), aCD3 monotherapy resulted in T1D remission in 30% of female NOD mice (3 of 10) (Fig. 6A). CT of diabetic NOD mice improved the outcome and resulted in T1D remission in 55% of mice (6 of 11) (Fig. 6A). Immunohistochemistry revealed that pancreata of isotype control–treated NOD mice did not contain any functional islets, and most lymphocytes had already left the pancreas, leaving behind residual islet scars (Fig. 6B). By contrast, both aCD3 and CT-treated NOD mice displayed remaining functional islets producing insulin (Fig. 6B). However, large cellular infiltrates were still present in pancreas of NOD mice receiving either treatment regimen (Fig. 6B). Scoring of islet infiltration revealed that, in contrast to untreated NOD mice, which displayed massive infiltrations in all islets, aCD3 and CT treatment resulted in reduced overall insulitis (Fig. 6C).

At day 21 after the first aCD3 dose, lymphocytes isolated from spleen and PDLNs were stimulated with the NOD islet antigen peptide mimotope NRP-V7 (34) to determine the frequency of islet antigen–specific T cells by intracellular cytokine assay for IFN-γ. Because of the lower frequency of islet antigen–specific T cells, the smaller number of NOD mice used, and the poor synchronicity of the pathogenesis between individual NOD mice, the data obtained were not as evident as in the RIP-LCMV model. However, we could detect a tendency toward a reduced frequency of islet antigen–specific CD8 T cells in the PDLNs of aCD3- and CT-treated diabetic NOD mice compared with isotype control mice (Fig. 7A and B). In addition, we found a tendency toward a higher frequency of FoxP3+ T cells after treatment with aCD3 or CT in the spleen and the PDLNs (Fig. 7A and B). Similar to the experiments with the RIP-LCMV model, we assessed the immune balance locally in the pancreas and determined the ratio of aggressive, islet antigen–specific CD8 T cells and FoxP3+ T cells in the NOD mouse model. As detected at day 21 after the first dose of aCD3, CT caused a significant shift (>17-fold increase) in the ratio of FoxP3+ versus islet autoantigen–specific T cells toward a more regulatory and less aggressive milieu in the PDLNs (Fig. 7B). In contrast to the RIP-LCMV-GP model, the change of the insulitis phenotype was dominated by the relative increase of FoxP3+ T cells rather than the relative decrease in islet antigen–specific aggressive CD8 T cells.

Interestingly, the majority of these FoxP3+ T cells was of the CD8, rather than the CD4, phenotype (Fig. 7A and B). Therefore, we performed double fluorescence immunohistochemistry of pancreas sections obtained from CT-, aCD3-, and isotype control–treated diabetic NOD mice at day 21 after the first aCD3 dose. Using confocal microscopy, we found an increased number of FoxP3+ T cells in the infiltrated islets of Langerhans of aCD3- and CT-treated mice (Fig. 7C). Differential counting of the cells revealed that the absolute number of FoxP3+ CD4 T cells was about 10- to 20-fold higher than FoxP3+ CD8 T cells in all treatment groups (Fig. 7D). The highest frequency of FoxP3+ CD4 T cells was found in the aCD3-treated group (Fig. 7D). There was no significant difference in the frequency of FoxP3+ CD8 T cells. However, the highest frequency was found in the pancreas of CT-treated mice (Fig. 7D).

In contrast to many preclinical studies of rodents, anti-CD3 monotherapy caused only a temporary halt of T1D progress rather than permanently curing the disease (1,2,4,8,11,12). One reason for this lack of persistence is the regeneration of the inactivated T-cell repertoire. Hence, to achieve long-term protection in the majority of patients with T1D, several CTs pairing anti-CD3 therapy with a secondary treatment have been assessed in preclinical models (14–19,21,22). However, none of the investigated CTs directly targeted the reentry of autoaggressive lymphocytes, which regenerated after inactivation by anti-CD3, into the islets of Langerhans. Here we report that a CT of aCD3 and aCXCL10 persistently blocks T1D pathogenesis by preventing the reentry of autoaggressive T cells into the islets. We used two different mouse models for T1D, and we treated diabetic mice with three doses of aCD3 followed by eight injections of neutralizing aCXCL10. In the inducible RIP-LCMV-GP and the spontaneous NOD mouse model, CT induced T1D remission in 60% and 55% of diabetic mice, respectively. CT was superior to monotherapies with either aCD3 or aCXCL10, and the observed T1D remission was reflected in restored insulin production and reduced insulitis. Mechanistically, we found that the sequential inactivation of T cells followed by a blockade of T-cell migration resulted in a change in the composition of T cells in the pancreas. In the RIP-LCMV-GP model the total number of islet antigen–specific CD8 T cells was significantly reduced in the pancreas, whereas in NOD mice we found only a tendency toward a reduction of specific T cells. It was recently demonstrated by in situ MHC-peptide tetramer staining of pancreas tissue from cadaveric donors with T1D that islet autoantigen–specific T cells are indeed present in the islets of patients with T1D up to 8 years after clinical diagnosis (40). These data suggest that a long-lasting reduction of islet autoantigen–specific T cells in the vicinity of the islets might be important in a persistent abrogation of disease. In contrast to RIP-LCMV-GP mice, we found a marked increase in FoxP3+ T cells in NOD mice receiving CT. Direct analysis of infiltrated islets by double immunofluorescence staining and confocal microscopy revealed that the majority of FoxP3+ T cells were CD4+, as previously reported (9,10). However, we also detected some FoxP3+ CD8 T cells in the pancreas and, to a larger extent, in the PDLNs. Such CD8+FoxP3+ T cells have also been detected in patients with T1D after aCD3 therapy (41). Importantly, in both the RIP-LCMV and the NOD mouse models, the ratio between FoxP3+ T cells and islet antigen–specific CD8 T cells was significantly higher in CT-treated mice, indicating that a shift in the immune balance toward regulation in the proximity of the islets of Langerhans might be responsible for the persistent blockade of the autodestructive process. In our hands the aCD3 monotherapy was also effective. However the CT was far more effective (60% vs. 38% remission), particularly in the RIP-LCMV model. An aCD3 monotherapy was evaluated in the RIP-LCMV model before (42) and resulted in a higher rate of remission (75–100%, depending on the time of aCD3 administration). However, in contrast to our studies, which used a rather low aCD3 dose (three injections of 3 µg), five daily doses of 100 µg were administered in the previous study (42).

CXCL10 has been found to be elevated in islets of patients with T1D (28,29). In the RIP-LCMV model, CXCL10 was induced very soon after LCMV infection (25). Neutralization of CXCL10 resulted in reduced T1D frequency and decreased recruitment of CXCR3-positive T cells to the pancreas (25). Reduced incidence and delayed onset of T1D have also been reported for CXCR3-deficient RIP-LCMV mice (24). However, a recent follow-up study using aCXCL10-treated or CXCR3-deficient RIP-LCMV mice suggested a certain redundancy of the CXCL10/CXCR3 axis (27). Here we found distinct T1D remission after aCXCL10 monotherapy. In contrast to our earlier studies of T1D prevention (25,27), mice were treated with aCXCL10 for 19 days after diabetes onset. A similar discrepancy exists for CXCL10-deficient RIP-LCMV-GP mice, which, in a previous study, developed T1D between day 9 and 14 after LCMV infection, just like regular RIP-LCMV-GP mice (27). In our hands the majority of CXCL10-deficient RIP-LCMV-GP mice also developed T1D within 2 weeks after infection. Many of these diabetic mice reverted in the following weeks, however, and when additionally treated with aCD3, all of the diabetic CXCL10-deficient RIP-LCMV-GP mice display full remission. Thus, one can speculate that CXCL10 plays a critical role in retaining aggressive T cells in the islets, whereas at earlier times other inflammatory factors might compensate for the relative lack of CXCL10.

Still, aCXCL10 monotherapy seems unlikely to be effective in patients with T1D. The reason for a likely failure is that, even if the concept of pathogens being involved in T1D etiology (43,44) holds true, at the time of diagnosis the responsible pathogenic infection would lie in the past, and it therefore would be impossible to interfere with the initial chemokine burst. In the CT setting, the administration of aCXCL10 follows a precise schedule, in which diabetic mice are treated first with low-dose aCD3 to inactivate a significant portion of T cells, including islet antigen–specific CD8 T cells. Thereafter, aCXCL10 administration prevents the de novo migration of reactivated/regenerated T cells to the islets. Such a regimen seems to be realistic for therapeutic application in patients with T1D to prevent the reported relapse within 2 years of aCD3 treatment (1,2,8,11,12). The effect of CT in the RIP-LCMV-GP and the NOD mouse models was long-lasting. Even 20 weeks after remission, no relapse was detected. Insulin production was maintained; in particular, islets of CT-treated RIP-LCMV mice remained largely without insulitis. In NOD mice we detected a state of peri-insulitis similar to that in young mice that have not yet developed T1D. Such a steady state might be preserved by the presence of a high frequency of FoxP3+ T cells maintaining a long-term regulatory milieu. Interestingly, no further administration of aCXCL10 is needed to maintain T1D remission after the initial eight injections following aCD3 therapy. Note that at the end of CT, no infectious virus particles were detected in the spleen and pancreas of RIP-LCMV-GP mice, as determined by virus plaque assay (45) (data not shown). Thus, reactivation of the LCMV-GP–specific islet-destructing CD8 T cells is likely to occur via presentation of β-cell–derived transgenic LCMV-GP rather than through remaining LCMV.

Among different aCD3 CTs, it seems to be important to target two distinct mechanisms to achieve a significant improvement over the corresponding monotherapy. Whereas some CTs aim at generating islet antigen–specific Tregs (14,15), others use two antibodies in parallel, such as aCD3 combined with an anti-CD20 antibody (aCD20) (18). Similar to our study, the aCD3/aCD20 CT of NOD mice demonstrated that intravenous injection of two antibodies is successful in reversing T1D. In contrast to the persistent effect of our aCD3/aCXCL10 CT, however, the parallel inactivation of T and B cells with aCD3/aCD20 caused T1D remission within 1 month of treatment, but most of the mice showed a relapse by month 3 after treatment (18). In any case, the therapeutic window relative to T1D onset might be critical for success. We treated RIP-LCMV and NOD mice early after onset. At this stage, most of the β-cells are functionally inactivated by inflammatory stress rather than completely eradicated. Thus, stress relief through the inactivation of T cells and prevention of de novo insulitis restored β-cell function. Later, when most of the β-cells are physically desztroyed, such CT might be ineffective, since β-cells’ natural regeneration might be insufficient.

In conclusion, we showed that CT with aCD3 and aCXCL10 results in persistent T1D remission in diabetic RIP-LCMV-GP and NOD mice. Both models do not entirely reflect the pathogenesis of human T1D. On the one hand, in contrast to human T1D, the RIP-LCMV-GP model is independent of CD4 T-cell help but has the advantage of being inducible by a well-defined triggering event (i.e., virus infection), which results in highly synchronized pathogenic events. This allows for the precise timing of a given therapeutic intervention in relation to pathogenic status. The NOD mouse, on the other hand, is a spontaneous model in which a genetic predisposition results in T1D-like disease independent of environmental triggering events. Just as in human T1D, however, the destruction of β-cells is dependent on both CD4 and CD8 T cells. Although our data from those two models are not completely identical, it is important to acknowledge that neutralization of the critical inflammatory chemokine CXCL10 directly after the transient inactivation of T cells with low-dose aCD3 results in the diminished presence of autoaggressive T cells in the pancreas. The ongoing functional inactivation and destruction of β-cells is thereby halted, and this truce is maintained by a shift in the local immune balance toward regulation. Since the administration of both low-dose aCD3 as well as aCXCL10 is only transient, such a CT might be highly attractive for an application in patients with T1D.

Funding. This study was supported by the Else Kröner-Fresenius Foundation (EKFS), Research Training Group Translational Research Innovation—Pharma (TRIP), the German Research Foundation (DFG), and the Goethe University Hospital Frankfurt.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.L. designed the study, performed experiments, interpreted data, and drafted the manuscript. P.M. and M.B. performed experiments. J.M.P. critically revised the manuscript. A.D.L. designed the study and critically revised the manuscript. E.H. performed experiments and critically revised the manuscript. U.C. conceived of and designed the study, interpreted data, and drafted the manuscript. U.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.