Reversal of Diabetes in NOD Mice by Clinical-Grade Proinsulin and IL-10–Secreting Lactococcus lactis in Combination With Low-Dose Anti-CD3 Depends on the Induction of Foxp3-Positive T Cells

Clinical translation of antigen-based therapies has been disappointing so far, and, specifically in autoimmune type 1 diabetes (T1D), the administration of oral insulin or GAD has not been efficacious in preventing or halting the disease in high-risk individuals or in new-onset patients to date (1–3). Issues like the route and timing of vaccination and dosing strategy, but also the reliability of delivery of full protein or peptide due to gastric digestion may lie at the basis of these failures (4). The introduction of a delivery vehicle like the gram-positive food-grade lactic acid bacterium Lactococcus lactis (L. lactis), which is able to deliver intact antigen and immunomodulating cytokines directly into the gut, in proximity to the gut-associated lymphoid tissue, is an appealing tool (5,6). Another reason why antigen-based therapies did not succeed in stopping ongoing autoimmune processes may be that by the time the disease manifests, the strong pathogenic immune reactions overpower the regulatory mechanisms induced by the therapy. In light of this consideration, we and others (7,8) advocate that combinations of robust antigen-based interventions and systemic immune modulators may ultimately be needed to successfully reinstate long-term tolerance in ongoing autoimmunity without compromising immune function. Previously, we reported that a combination therapy consisting of a 5-day course of anti-CD3 antibodies at disease onset along with 6 weeks of oral administration of live, genetically modified L. lactis producing human proinsulin (PINS) and interleukin-10 (IL-10) safely restored durable normoglycemia in ∼60% of nonobese diabetic (NOD) mice and elicited forkhead box p3–positive (Foxp3⁺) T cells with a regulatory phenotype (9). The route to bring this successful antigen-based therapy to new-onset T1D patients will depend both on the generation of a clinical-grade self-containing L. lactis strain (10), but also on a profound understanding of the processes underlying this disease-modifying approach, and, consequently, on the implementation of certified biomarkers of both immune and therapeutic success.

Here, we demonstrated similar therapeutic efficacy in autoimmune diabetes remission using a clinical-grade self-containing L. lactis vaccine compared with the plasmid-driven L. lactis strain previously reported (9). In addition, we identified both functional β-cell reserve (initial blood glucose concentrations <350 mg/dL) and pretherapy insulin autoantibody (IAA) positivity as predictors of therapeutic efficacy, and proved that Foxp3⁺ T cells are a prerequisite for the induction and maintenance of active tolerance induced by the L. lactis–based therapy.

NOD mice, originally obtained from Dr. C.Y. Wu (Department of Endocrinology, Peking Union Medical College Hospital, Beijing, People’s Republic of China), were housed and inbred in the animal facility of Katholieke Universiteit Leuven (Leuven, Belgium) since 1989. NOD.Foxp3.DTR (diphtheria toxin receptor) mice and NOD.Foxp3.hCD2 mice were bred from stocks provided by Dr. C. Benoist (Harvard Medical School, Boston, MA) and Dr. S. Hori (RIKEN Research Center for Allergy and Immunology, Yokohama, Japan), respectively. The housing of all mice occurred under semibarrier conditions, and animals were fed sterile food and water ad libitum. Mice were screened for the onset of diabetes by evaluating glucose levels in urine (Diastix Reagent Strips; Bayer, Leverkusen, Germany) and venous blood (AccuCheck; Roche Diagnostics, Vilvoorde, Belgium). Mice were diagnosed as diabetic when they had glucosuria and two consecutive blood glucose measurements exceeding 200 mg/dL. NOD-scid and NOD-scid γc^−/− mice were bred from stocks purchased from The Jackson Laboratory (Bar Harbor, ME). Animals were maintained in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and all experimental procedures were approved and performed in accordance with the Ethics Committees of the Katholieke Universiteit Leuven under project no. 185-2009.

Details on the construction, culture, and in vitro quantification of the lactococcal vectors used in the current study are available in the Supplementary Research Design and Methods (9). For intragastric inoculations, stock suspensions were diluted 1,000-fold in growth media and incubated for 16 h at 30°C, reaching a saturation density of 2 × 10⁹ colony-forming units/mL. Bacteria were harvested by centrifugation and concentrated 10-fold in BM9 medium. Treatment doses consisted of 100 μL of this bacterial suspension.

Upon diabetes determination, NOD or NOD transgenic mice were treated for 5 consecutive days intravenously (days 0–4; 2.5 µg/mouse) with hamster anti-mouse CD3 antibodies (clone 145–2C11; BioXCell, West Lebanon, NH). This therapy was given in combination with oral administration of either plasmid-driven or clinical-grade L. lactis strains (2 × 10⁹ colony-forming units) 5 times/week over 6 weeks. Control mice were left untreated. Individual blood glucose concentrations at the start of treatment were recorded. Mice were tested three times weekly for their weight and blood glucose status. Remission was defined as the absence of glucosuria and a return to normal blood glucose concentrations. Experimental animals were sacrificed immediately or long after stopping therapy (6 or 14 weeks after treatment initiation). Peripheral blood, lymph organs, and pancreas were harvested, and single cells were assessed for phenotyping as described in the Supplementary Research Design and Methods. Detailed methodology and references on in vitro suppression assays are described in the Supplementary Research Design and Methods. Mice were removed from the study prior to the 14-week end point when blood glucose concentrations exceeded 600 mg/dL in two consecutive measurements.

One or two weeks prior to sacrifice, intraperitoneal glucose tolerance tests (IPGTTs) were performed. Mice were fasted for 16 h, injected with glucose (2 g/kg i.p.), and blood glucose concentrations were measured at 0, 15, 30, 60, 90, and 120 min.

Heparinized plasma was collected from NOD mice with new-onset diabetes before treatment randomization and at therapy discontinuation, and IAAs were analyzed at the University of Florida Department of Pathology, Immunology and Laboratory Medicine, College of Medicine (Gainesville, FL), as previously described (11).

Mice tolerized by L. lactis–based therapy were injected intraperitoneally after therapy withdrawal with blocking antibodies against cytotoxic T-lymphocyte antigen 4 (CTLA4) (clone UC10–4F10; Bioceros) and transforming growth factor-β (TGF-β) (clone 1D11.16.8; BioXCell) in the following dose regimen: 250 µg at days 1 and 3, and then 100 µg at days 6, 8, 10, 13, and 18 for CTLA4; 200 µg three times per week over 3 weeks for TGF-β. Blood glucose concentrations were measured daily up to 25 days after the first injection.

To assess the diabetogenic potential of effector T cells (Teffs), all T cells from the spleens (1 × 10⁷ cells) of control mice with new-onset diabetes, responders and nonresponders to L. lactis–based therapy, were transferred intravenously into the tail veins of 6- to 8-week-old immune-deficient NOD-scid mice. Recipient mice were monitored twice weekly for the development of diabetes up to 100 days after cell transfer.

NOD.Foxp3.DTR mice (expressing the human DTR under the control of Foxp3 transcriptional control elements) allow for the depletion of Foxp3⁺ T cells upon diphtheria toxin (DT) administration (12). For regulatory T-cell (Treg) depletion, NOD.Foxp3.DTR mice (unmanipulated or tolerized after stopping the L. lactis–based therapy) were injected intravenously with 40 µg/kg body wt DT (Sigma-Aldrich) on days 1, 2, 4, and 7, and examined on day 8. After DT injections, weight, urine, and blood glucose status of mice were monitored. Foxp3⁺ T cells were monitored in peripheral blood and pancreas by flow cytometry and histology, respectively, as previously described (13).

P60 (a 15-mer synthetic peptide that can bind to and block FOXP3, 50 μg/dose i.p. daily, up to 14 doses) was administered at start of the L. lactis–based therapy, as previously described (14).

Six-micrometer sections from formalin-fixed, paraffin-embedded pancreas tissues were cut and collected 100 µm apart, then stained with hematoxylin-eosin. Islets were observed under light microscopy at 20× or 40×, enumerated, and graded by an independent investigator in blinded fashion. At least 25 islets per pancreatic sample were scored for islet infiltration as follows: 0, no infiltration; 1, peri-insulitis; 2, islets with lymphocyte infiltration in <50% of the area; and 3, islets with lymphocyte infiltration in >50% of the area or completely destroyed.

Pancreas tissues were snap frozen in 2-methyl-butane 99% (ACROS Organics, Geel, Belgium), and cut into 12-μm tissue sections. Foxp3⁺ T-cell detection was performed as previously described (9).

All data were analyzed using GraphPad Prism 6 (GraphPad, La Jolla, CA). Survival curves were computed with the Kaplan-Meier test and were compared with the log-rank test. Groups were analyzed by ANOVA (nonparametric Kruskal-Wallis test) with the Dunn multiple comparison test or with the Mann-Whitney U test, as appropriate. Error bars represent the SEM. Unless otherwise indicated, differences are not significant. Symbols for P values are defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Using a clinical-grade self-containing L. lactis strain secreting human PINS along with IL-10 in combination with subtherapeutic doses of anti-CD3 antibodies, 66% of mice (23 of 35 mice) reverted to normoglycemia for at least 14 weeks after disease onset, which was significantly superior to the 43% of mice treated by anti-CD3 alone (Fig. 1A). This therapeutic efficacy obtained with the clinical-grade L. lactis strain was comparable to the combination therapy with plasmid-driven L. lactis strain (72% of mice [18 of 25 mice]; not significant). As expected, animals left untreated (n = 20) or treated with the empty vector bacterial strain L. lactis-pT1NX (n = 9) remained hyperglycemic and were sacrificed when 20% of their starting body weight was lost. Monotherapy with either the clinical-grade or plasmid-driven L. lactis strain secreting PINS and IL-10 was significantly less effective than the combination with anti-CD3 (0% [n = 8] and 17% [n = 8], respectively) (Fig. 1A).

During follow up, controls with new-onset diabetes and mice protected or not by L. lactis–based therapy were subjected to IPGTT and sacrificed 6 weeks after treatment initiation, at which time their pancreas tissues were assessed by histology. Only in the successfully treated animals, residual β-cell function (i.e., assessed as the area under the glucose tolerance curve) was preserved, and smaller proportions of islets had severe insulitis (Fig. 1B). Of interest, at the end of the combination therapy no difference in the severity of insulitis was observed between responders and nonresponders (Fig. 1C).

No influence of the age or sex of mice was observed on the therapeutic success of the L. lactis–based therapy (data not shown). However, as previously shown (9), glycemic concentrations at the beginning of therapy predicted success, with 82% of mice starting with a glycemia <350 mg/dL cured (n = 22), compared with 38% of mice with a starting glycemia >350 mg/dL (n = 13) (Fig. 2A). In addition, positivity for IAAs at study entry seemed to correlate with therapeutic success (Fig. 2B). Interestingly, mice with blood glucose concentrations <350 mg/dL and IAA positivity at therapy start had a clearly superior diabetes remission rate (89%; n = 8) than mice with blood glucose levels >350 mg/dL and being IAA negative (33%; n = 5; P = 0.07) (Fig. 2C). Moreover, the L. lactis–based therapy significantly decreased IAA levels, particularly in mice responsive to the therapy (Fig. 2D).

The mechanisms underlying disease remission induced by the L. lactis–based treatment were investigated by dissociating between the therapeutic immune effects in mice responsive or not to the intervention. We found that the percentages of CD4⁺Foxp3⁺ (both CD25⁺ and CD25⁻) T cells observed in the peripheral blood (Fig. 3A), the pancreas-draining lymph nodes (Fig. 3B), and the pancreas (Fig. 3C) were significantly higher in mice treated with the L. lactis–based therapy in comparison with untreated controls. Interestingly, in the pancreas-draining lymph nodes and pancreas, but not in peripheral blood, the increased frequency of CD4⁺Foxp3⁺ T cells was less pronounced in responders than in nonresponders. Using multicolor flow cytometry, we identified that most CD4⁺Foxp3⁺ Tregs were positive for CTLA4 and that the expression of this inhibitory marker was significantly higher in pancreas-draining lymph nodes (for both responders and nonresponders) and pancreata (only for responders) of treated mice compared with untreated controls (Supplementary Fig. 1B and C). Of interest, no differences in the percentages of CD4⁺Foxp3⁺CTLA4⁺ T cells were observed in the peripheral blood of treated mice compared with untreated controls (Fig. 3D).

The percentages of naive (CD44⁻CD62L⁺CCR7⁺), effector memory (CD44⁺CD62L⁻CCR7⁻) and central memory (CD44⁺CD62L⁺CCR7⁺) CD4⁺ T cells were not altered in any recipient group with respect to therapeutic success or failure (data not shown). The transfer of splenocytes from responders and nonresponders of L. lactis–based treatment caused diabetes in NOD-scid recipients with similar disease kinetics as seen with the transfer of splenocytes isolated from untreated controls with new-onset diabetes, suggesting that circulating diabetogenic cells were not depleted from treated mice (Supplementary Fig. 2).

Using NOD.Foxp3.hCD2 mice treated by L. lactis–based therapy, we could isolate CD4⁺CD25⁺Foxp3⁺ T cells for functional in vitro studies, in which they suppressed proliferation, CD69 activation, and interferon-γ production of pathogenic CD4⁺CD25⁻ Teffs. These Tregs produced IL-10 (and TGF-β) when they were cocultured and stimulated with anti-CD3 antibody in the presence of splenic antigen-presenting cells (APCs) isolated from NOD-scid γc^−/− mice (Fig. 4 and data not shown). No difference in the regulatory capacity of CD4⁺CD25⁻Foxp3⁺ T cells was seen between therapy responders and nonresponders.

The addition of anti-CTLA4 Ig (clone UC10–4F10) or a TGF-β neutralizing antibody (clone 1D11.16.8) significantly reduced the suppression by the CD4⁺CD25⁺Foxp3⁺ T cells (Fig. 5A), suggesting that CD4⁺CD25⁺Foxp3⁺ Tregs of cured mice inhibit Teff proliferation via a CTLA4-dependent and TGF-β–dependent fashion in vitro. Adding anti–IL-10 (clone JES5–2A5) did not alter the direct suppressive effect of the Tregs. On the other hand, these regulatory mechanisms were not demonstrated with the CD4⁺CD25⁻Foxp3⁺ T-cell fraction from therapy responders and nonresponders (Fig. 5B). Treating stably cured mice in vivo with a combination of anti-CTLA4 Ig (clone UC10–4F10) and anti–TGF-β (clone 1D11.16.8) led to diabetes recurrence in two of five mice (Fig. 5C).

Finally, we investigated whether the therapeutic success of L. lactis–based therapy was dependent on the presence and functionality of Foxp3⁺ T cells. For this, NOD mice with new-onset diabetes were simultaneously treated with the L. lactis–based therapy and the FOXP3-inhibitory peptide P60 for a period of 14 days (Fig. 6A). Interestingly, none of the mice (n = 6) developed normoglycemia, while mice treated with the L. lactis–based therapy and vehicle (n = 11) already had a 60% diabetes remission rate, indicating that Tregs are crucial for the induction of therapy-induced tolerance (Fig. 6B).

Next, NOD.Foxp3.DTR mice with new-onset (spontaneous) diabetes were treated with the L. lactis–based therapy, and, after stable diabetes reversal was observed, Foxp3⁺ T cells were eliminated using DT as described in the scheme depicted in Fig. 7A. First, we established in unmanipulated NOD.Foxp3.DTR mice that the selected DT regimen eliminated >90% of CD4⁺Foxp3⁺ T cells, with the remaining Tregs expressing low or no CD25, in the peripheral blood within 3 days after the first DT injection (Supplementary Fig. 3A–C). A progressive repopulation of these cells started from day 5 after the first DT injection, as has been reported for several Foxp3.DTR strains (15–17). This DT regimen also dramatically decreased the number of Foxp3⁺ T cells residing in the pancreas, consequently leading to the development of autoimmune diabetes (Supplementary Figs. 2D and 3B). Next, comparable to wild-type NOD mice, the L. lactis–based treatment induced autoimmune diabetes remission in 57% of NOD.Foxp3.DTR mice (four of seven mice) (Fig. 7B). Transient Foxp3⁺ T-cell depletion resulted in a complete reversal to the diabetic state in all mice (n = 4) that were initially cured by the therapy, as evidenced by the reappearance of glucosuria along with severe hyperglycemia starting from day 2 after the first DT injection (Fig. 7B). This breach of immune tolerance to insulin-producing β-cells was also accompanied by the induction of severe insulitis (Fig. 7C) and the ablation of the islet-resident Foxp3⁺ Treg pool (Fig. 7D). Collectively, these data demonstrated that the therapeutic effect from the L. lactis–based intervention depended on the presence and functionality of Foxp3⁺ Tregs.

Oral tolerance as a means of intervention to arrest disease has been extensively explored in various animal models of autoimmune disease, including T1D (18). We previously described the reversal of new-onset autoimmune diabetes in mice by the oral administration of plasmid-driven L. lactis strains secreting diabetes-relevant antigens (i.e., whole PINS or GAD65 peptide) and IL-10 in combination with systemic low-dose anti-CD3 (9,11). In both antigen-based therapies, induction of CD4⁺CD25⁺Foxp3⁺ T cells accompanied the therapeutic success.

In the current study, we designed an oral clinical-grade self-containing L. lactis strain secreting chromosomal-integrated human PINS and IL-10. When combined with a short course of subtherapeutic doses of anti-CD3, the intervention was safe and highly effective in inducing long-term normoglycemia in mice with new-onset diabetes. Initial blood glucose concentrations (<350 mg/dL) in addition to IAA positivity at disease onset were predictors of therapeutic outcome, whereas the preservation of residual β-cell function and decline in IAA positivity were markers of therapeutic success. It is encouraging that studies with anti-CD3 monotherapy in patients with new-onset T1D already revealed that subjects who enrolled within 6 weeks of diagnosis and with higher levels of C-peptide at entry responded better to therapy (19–21). These observations suggest that some degree of residual β-cell mass will be necessary for therapeutic success when intervening at the moment of diabetes diagnosis, namely when dysglycemia is present. Likewise, a post hoc analysis of participants with new-onset T1D of the otelixizumab trial found a good correlation between pre-existing IAA levels and clinical outcome (22). IAA positivity at study entry was also found to distinguish responders from nonresponders among recipients of oral insulin (1).

Our previous studies suggested that the mechanism of L. lactis–based therapy involved the induction of CD4⁺CD25⁺Foxp3⁺ T cells (9). By dissociating between the immune effects of the L. lactis–based intervention in mice responsive or not to the therapy, we were able to further characterize the nature and role of the immune processes accompanying the treatment. The L. lactis–based therapy induced suppressive IL-10–secreting CD4⁺Foxp3⁺ (both CD25⁺ and CD25⁻) T cells in the pancreas-draining lymph nodes and pancreata of responders, and even more so of nonresponders, suggesting enhanced recruitment of Tregs to the inflamed target tissues. In the periphery, the frequency of CD4⁺Foxp3⁺ T cells was also increased in treated mice compared with untreated controls, pointing toward a possible value for this cell population as an immune marker. Interestingly, the frequency of CTLA4⁺ T cells among various Treg subsets was significantly higher in the pancreata of combination therapy-treated responder mice compared with mice with new-onset diabetes and in contrast to combination therapy–treated nonresponder mice. CTLA4 by Tregs has a nonredundant role to limit lymphopenia-induced uncontrolled proliferation of autoreactive Teffs in vivo (23). Alternatively, CTLA4 expression by Tregs may prolong the contact time between Tregs and dendritic cells via lymphocyte function–associated antigen 1 activation (24), increasing the efficiency of Treg suppression in a (transient) lymphopenic environment, similar to the one induced by low-dose anti-CD3. Because various immune- or tissue-specific mediators are important for Treg-suppressive functions at these inflammatory sites, more in-depth studies looking at the expression of chemokines, adhesion molecules, and extracellular matrix components of site-specific Tregs can provide more insights into follow-up studies. Of note, no difference was seen in the degree of insulitis between responder and nonresponder mice, suggesting also alterations in other lymphocyte subsets besides Tregs.

It is still not fully understood how Tregs control immune effector responses in autoimmune diseases and inflammation. Several studies (25) demonstrated that peripheral Tregs can use different regulatory mechanisms according to their environmental milieu and stimulatory conditions. In our case, CTLA4 and TGF-β were important for the regulatory activity of therapy-expanded CD4⁺CD25⁺Foxp3⁺ T cells in vitro, and partially in vivo, whereas IL-10 was not. There is no consensus on the role of CTLA4 for Treg function, and several effects have been reported: the induction of cell-intrinsic negative signals to activated Teffs, the modulation of the development of APCs, and the stable function of Foxp3⁺ Tregs (26,27). With regard to the involvement of TGF-β in therapy-mediated suppression, this cytokine can regulate several immunological processes, such as inflammation, lineage commitment, antibody generation, as well as tolerance induction (28). Moreover, it can preserve Foxp3 expression and support the differentiation of other T cells into Treg-like cells (29). In fact, TGF-β can promote the development of IL-10–secreting Tregs, because treatment of mice with anti–TGF-β prevented the conversion of CD4⁺Foxp3⁻ cells into CD4⁺Foxp3⁺IL-10⁺ cells in intestine-associated lymphoid tissues (30). Although IL-10 seemed to be a good marker for the identification of our L. lactis–based therapy–induced Tregs, the role of IL-10 in their regulator function remains controversial. Others also demonstrated (31) that anti–IL-10 antibodies did not abrogate established tolerance in vivo. Here, it has been discussed whether IL-10 modulated the maturation phenotype of the APCs inducing anergy in both antigen-specific CD4⁺ and CD8⁺ T cells, and preferentially converting truly naive CD4⁺ T cells into suppressor cells expressing Foxp3, rather than through direct activity on T cells (32). Based on our observations, it is intriguing to speculate that Treg production of IL-10 is a major mechanism by which Tregs regulate inflammation at environmental interfaces, whereas TGF-β- and CTLA4-dependent regulation of the function of APCs may be a regulatory mechanism that predominates in secondary lymphoid tissues, where it controls the initial activation and expansion of naive autoreactive T-cells.

As in mice, human Tregs are defined by having a suppressive phenotype endowed by high and sustained expression of the transcription factor Foxp3 (33), and the loss of function/mutation in the Foxp3 gene leads to severe fatal autoimmune disorders (15,34). In the current study, we discovered that the specific inhibition of Treg functionality by the P60 peptide (14) at the start of L. lactis–based therapy completely impaired the induction of therapy-induced tolerance. Moreover, transient depletion of Foxp3⁺ Tregs from therapy-tolerized NOD.Foxp3.DTR mice was sufficient to induce complete disease relapse in all animals, demonstrating that the presence of Foxp3⁺ T cells was crucial to maintain therapeutic tolerance and control pathogenic Teffs, which were still present in mice responsive to L. lactis–based therapy. A recent study (35) suggested that antigen-specific Foxp3⁺ Tregs can also mediate tolerance both by diminishing recruitment of antigen-carrying inflammatory APCs to lymph nodes and by impairing their function.

In conclusion, our data demonstrated that combining a clinical-grade self-containing L. lactis–secreting human PINS and IL-10 with low-dose anti-CD3 increased the frequency of diabetes reversal compared with anti-CD3 monotherapy. Both therapy responders and nonresponders had increased frequencies of CD4⁺Foxp3⁺ T cells, suggesting that immune effects induced by the L. lactis–based therapy occurred in each individual recipient, but that therapeutic success (defined as a return to stable normoglycemia) depended on other parameters, such as functional β-cell mass still present at disease onset. This idea was further strengthened by the observation that therapeutic success was correlated with starting glycemia. Next to initial blood glucose concentrations at study entry, IAA levels also predicted the outcome of this L. lactis–based therapy using PINS as antigen. Finally, we showed that Foxp3⁺ Tregs were essential to induce and maintain active tolerance and control diabetogenic immune responses in tolerized mice. These findings provide all the ingredients for testing this intervention in humans: a clinical-grade self-containing L. lactis–secreting islet antigens, biomarkers for predicting therapeutic success, and the demonstration that the induction of Foxp3⁺ T cells is the basis of the L. lactis–based therapy–induced cure.

Acknowledgments. The authors thank Sofie Robert, Jos Laureys, and Elien De Smidt (Laboratory of Clinical and Experimental Endocrinology, Katholieke Universiteit Leuven, Leuven, Belgium) for excellent technical support.

Funding. This work was supported by the European Community’s Health Seventh Framework Programme (FP7/2009-2014 under grant agreement 241447 with acronym NAIMIT), JDRF (grant 17-2011-524), the Fund for Scientific Research Flanders (FWO-Vlaanderen grant G.0554.13N), the KU Leuven (GOA 2014/010), the European Foundation for the Study of Diabetes (Sanofi Innovative Approaches Programme 2014), and by gifts from Hippo & Friends Type 1 Diabetes Fonds and Carpe Diem Fonds voor Diabetesonderzoek. D.P.C. is a PhD fellow of the FWO-Vlaanderen (11Y6716N). H.K. is a postdoctoral fellow and C.M. is a clinical researcher of the FWO-Vlaanderen. F.D. received support from the Italian Ministry of Research (grant 2010JS3PMZ_008), the Italian Ministry of Health, and Fondazione Roma.

Duality of Interest. L.S. and P.R. have financial interests in Intrexon Actobiotics NS, including employment and stock options. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.T. contributed to the design of the study, the conducting of the experiments, the interpretation of the data, and the drafting of the manuscript. D.P.C. contributed to the design of the study, the conducting of the experiments, the interpretation of the data, and the critical revision of the manuscript. H.K., N.C., J.J.L., L.S., and P.R. contributed to the critical revision of the manuscript. G.S., F.M., and J.P.M.C.M.C. contributed to the conducting of the experiments. C.W. contributed to the conducting of the experiments and the critical revision of the manuscript. F.D. contributed to the interpretation of the data and the critical revision of manuscript. C.G. and C.M. contributed to the idea for and design of the study, the interpretation of the data, the drafting of the manuscript. C.G. and C.M. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.