By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Despite encouraging results in the NOD mouse, type 1 diabetes prevention trials using subcutaneous insulin have been unsuccessful. To explain these discrepancies, 3-week-old NOD mice were treated for 7 weeks with subcutaneous insulin at two different doses: a high dose (0.5 U/mouse) used in previous mouse studies; and a low dose (0.005 U/mouse) equivalent to that used in human trials. Effects on insulitis and diabetes were monitored along with immune and metabolic modifications. Low-dose insulin did not have any effect on disease incidence. High-dose treatment delayed but did not prevent diabetes, with reduced insulitis reappearing once insulin discontinued. This effect was not associated with significant immune changes in islet infiltrates, either in terms of cell composition or frequency and IFN-γ secretion of islet-reactive CD8+ T cells recognizing the immunodominant epitopes insulin B15-23 and islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP)206-214. Delayed diabetes and insulitis were assoc-iated with lower blood glucose and endogenous C-peptide levels, which rapidly returned to normal upon treatment discontinuation. In conclusion, high- but not low-dose prophylactic insulin treatment delays diabetes onset and is associated with metabolic changes suggestive of β-cell “rest” which do not persist beyond treatment. These findings have important implications for designing insulin-based prevention trials.
Type 1 diabetes (T1D) is an autoimmune disease in which failures in central and peripheral tolerance mechanisms lead to T-cell-mediated destruction of insulin-producing β cells. Several β-cell Ags have been identified, among which insulin and its precursor (pre)proinsulin ((pre)PI) are major targets of islet-reactive T cells, both in humans [] and in NOD mice []. This led to the hypothesis that subcutaneous insulin administration could prevent disease, the rationale of which is twofold. First, insulin administration could induce a state of “β-cell rest” that would make β cells less vulnerable to metabolic stress, apoptosis [], and possibly to immune-mediated destruction []. Second, repeated subcutaneous insulin injections could act as a vaccination protocol potentially restoring immune tolerance []. The relative role of these two putative mechanisms remains unsettled.
Studies in the NOD mouse showed that subcutaneously administered insulin, started prior to disease development, is capable of preventing insulitis and diabetes []. This observation prompted the Diabetes Prevention Trial-Type 1 (DPT-1)  and the smaller European Prediabetes Prevention–Subcutaneous Insulin Trial (EPSCIT) European trial in relatives of T1D patients at risk of developing disease . However, both trials were unsuccessful, as no difference in T1D incidence was observed between insulin- and placebo-treated individuals [[7, 8]].
It is still not clear why this translation from animal studies to human clinical trials failed, but two major differences between mouse and human interventions emerge as possible explanations. First, insulin doses used in animal studies were extremely high, that is 0.5 U per animal, corresponding to ∼25 U/kg, a dose which is ∼200-fold higher to that later used in human trials (0.125 U/kg) [[7, 8]]. The other major difference is the duration of treatment. Mice were treated for more than 150 consecutive days (until 26 weeks of age) []. Since the majority of female NOD mice develop diabetes before this age, it cannot be excluded that a delay in disease onset overshadowed bona fide protection. Although mice were left untreated for one week before testing [], hyperglycemia could have been attenuated by prior insulin treatment.
All of these studies were performed at a time when the specific epitope sequences of insulin and other β-cell Ags targeted by autoreactive T cells were not yet identified. It was therefore not possible to track modifications induced in these pathogenic T-cell populations. In more recent years, insulin B15-23 and islet-specific glucose-6-phophatase catalytic subunit-related protein (IGRP)206-214 have emerged as major epitopes targeted by autoreactive CD8+ T cells []. Diabetes is induced when T-cell clones recognizing these sequences are transferred into healthy mice [[10, 11]] or when transgenic T cells are primed by these epitopes in vivo [[12, 13]]. Tracking of circulating IGRP-specific cells is able to predict subsequent diabetes development in NOD mice []. Thus, islet-reactive T cells offer relevant biomarkers capable of reflecting the underlying immune modifications induced by insulin administration []. Moreover, biomarkers of putative β-cell rest, which may be further induced, would also clarify therapeutic mechanisms, but have not been explored in previous studies [].
Using these biomarkers, we investigated the clinical, immunological, and metabolic effects induced by short-term insulin therapy in NOD mice. The purpose was threefold. First, to assess whether a short-term treatment—similar to regimens proposed for human trials—could achieve any long-term effect. Second, to address whether the different insulin doses employed in animal and human studies could explain the lack of efficacy in clinical trials. Third, to study whether insulin treatment induces modifications in T-cell or metabolic markers suggestive of tolerance restoration or β-cell rest respectively. Our results identify some important elements to consider in the design of prophylactic insulin trials.
High-dose but not low-dose insulin delays but does not prevent diabetes onset
To address whether differences in regimens could explain the lack of efficacy in human preventative trials, NOD mice were monitored for diabetes incidence following treatment at two different doses of insulin glargine: 0.5 U/mouse, corresponding to the maximum tolerable dose used in early animal studies [] (referred to as high dose); and 0.005 U/mouse, equivalent to the dose used in human trials [] (referred to as low dose). Moreover, treatment was started at the earliest possible age (3 weeks) and stopped at week 10, at a time when all mice are not yet diabetic, to avoid the confounding factor of treating ongoing disease.
Only high-dose insulin-treated mice showed a delay in diabetes incidence compared with the control-treated group (Fig. 1A). This difference reached statistical significance between week 21 and 23 (p = 0.046). However, high-dose insulin-treated mice subsequently developed diabetes at the same rate of control-treated mice. Hence, diabetes protection was no longer significant at the end of follow-up (week 30; 38 and 48% diabetic mice in the insulin- and control-treated groups, respectively). On the contrary, mice treated with low-dose insulin showed a trend toward faster diabetes development compared with the control group, although this difference did not reach statistical significance at any time point (Fig. 1B). Thus, only high-dose insulin exerts a diabetes-protective effect, but this effect is limited in time, ultimately resulting only in delayed onset.
Insulitis is reduced during high-dose insulin treatment, but reappears when it is ceased
Given the lack of efficacy of low-dose insulin treatment, we focused subsequent immunological analyses on the high-dose regimen. To investigate whether high-dose insulin administration affects islet infiltration, we compared islet histology in insulin- and control-treated animals, scoring insulitis according to the number of normal, peripherally, and invasively infiltrated islets (Fig. 2A). Reduced mononuclear cell infiltrates were observed in islets of insulin-treated animals at the end of treatment (Fig. 2B). However, this difference was not sustained over time, since it was no longer present already 5 weeks after treatment discontinuation (week 15; Fig. 2C). Thus, high-dose insulin treatment is effective at initially reducing islet infiltration, but it has no long-term effect, in line with the observed delay in diabetes onset.
High-dose insulin treatment does not yield any durable immune modification
Next, we analyzed whether, despite their similar magnitude after treatment discontinuation, the composition of islet infiltrates was different between treated and control mice. The percentages of CD4+ and CD8+ T cells as well as of dendritic cells (DCs) were, however, similar in both groups (Fig. 3A–C). Similarly, we did not notice any difference in the number of regulatory (CD4+FoxP3+) T cells present (Fig. 3D).
We further analyzed the islet Ag specificity and IFN-γ secretion of infiltrating CD8+ T cells specific for the two immunodominant epitopes IGRP206-214 and PIB15-23, using surface tetramer (TMr) and intracellular IFN-γ staining upon in vitro peptide recall. Representative stainings are shown in Fig. 4A and B. We did not observe any difference in the fraction of these two CD8+ T-cell specificities, either in terms of total percent (number of TMr+CD8+ T cells; Fig. 4C and D) or of their capacity to secrete IFN-γ in response to their cognate epitope (Fig. 4E and F). Although we observed a trend toward lower IFN-γ production in PIB15-23-reactive CD8+ T cells in insulin-treated animals (Fig. 4F), this difference was not statistically significant and could simply reflect the delay in islet destruction observed during the treatment course. Similarly, spleen and pancreatic lymph nodes showed no difference in percentages of different cell populations (T cells, B cells, DCs), in their phenotype (T-cell activation, DC maturation) or IFN-γ T-cell response to PIB15-23 and IGRP206-214. Even when these analyses were performed at the end of treatment (week 10), no difference was observed in IFN-γ production upon stimulation with PIB15-23 and IGRP206-214 peptides (data not shown).
C-peptide secretion is only acutely suppressed during high-dose insulin treatment
Since we did not find any enduring immunological effect following insulin treatment, we investigated whether there was any long-term metabolic effect of β-cell rest induced. To obtain such effect, insulin doses would need to be sufficient to lower blood glucose levels, thus reducing the stimulus for endogenous insulin secretion. Indeed, in line with its delaying effect, only high- but not low-dose insulin was capable of acutely lower glycemia 1 h after a single injection (Fig. 5A). This effect of insulin glargine was similar to that of isophane insulin used in early animal studies []. Of further note, both insulin glargine and isophane resulted in blood glucose levels in the hypoglycemic range (0.56 ± 0.06 and 0.61 ± 0.08 g/L, respectively). Moreover, in line with its longer half-life, the glucose-lowering effect was more sustained for insulin glargine than for isophane insulin (Fig. 5B). Indeed, glargine- but not isophane-treated mice displayed marginally lower glycemic levels even 10 h after administration.
We then analyzed the effect of high- and low-dose insulin treatment on endogenous insulin secretion by measuring serum C-peptide concentrations after an i.p. glucose challenge. Also in this case, a lowering effect on glycemia and C-peptide secretion was observed when NOD mice were glucose challenged 2 h after the last insulin administration (Fig. 6A and B), but only when insulin was given at high doses. Indeed, the i.p. load did not cause any significant increase in glycemia, with C-peptide levels of 833 ± 169 and 1195 ± 327 pmol/L in control and low-dose insulin-treated animals, respectively, compared with 92 ± 4 pmol/L in high-dose insulin-treated ones (p < 0.0001). However, also this effect was rapidly lost, as this difference was no longer observed when the same experiment was repeated 20 h after the last insulin injection (Fig. 6C and D). In this case, the glucose challenge induced a similar increase in blood glucose levels in all groups, paralleled by similar C-peptide concentrations. Low-dose insulin treatment did not significantly decrease glucose nor C-peptide concentrations, even when the i.p. load was given 2 h after the last insulin injection (Fig. 6A and B). Thus, a condition of β-cell rest characterized by suppressed endogenous insulin secretion is induced only during high-dose treatment. However, this effect is not maintained over time, as C-peptide responses rapidly return to control levels in the absence of further insulin administration.
T1D prevention trials employing subcutaneous insulin administration have so far been uniformly unsuccessful [[7, 8, 16-21]]. Given this failure in translating mouse studies to human trials, we took the opposite approach and tried to translate the strategy proposed for human trials into mouse, taking into account two key differences. First, the insulin dose used that was much higher in murine protocols. Second, the duration of treatment which, conceivably, was shorter for human (median 3.7 years) [] than for mice (until day 180, that is 26 weeks, corresponding to 50–70% of the NOD lifespan) []. We employed insulin glargine that, in light of its pharmacokinetics, has a steadier bioavailability, a theoretical advantage both in terms of metabolic action and of Ag delivery. Indeed, low-level Ag persistence without significant pulsatility may be more effective at inducing tolerance [] and the amino acid substitution of insulin glargine (G→N at position A21) does not alter major T-cell epitopes.
We show that a short-term (7 weeks) treatment during the prediabetic phase with subcutaneous insulin administered at low doses (0.005 U/mouse), equivalent to that used in human trials, does not have any effect on disease protection. On the other hand, high-dose insulin treatment (0.5 U/mouse) delayed disease but was not sufficient to prevent it in the long term. This transient effect on diabetes incidence was not associated with modifications in IGRP206-214-reactive CD8+ T cells, and was only associated with marginal changes in PIB15-23-reactive CD8+ T cells. Notwithstanding the importance of CD4+ T cells in diabetes pathogenesis, we focused on CD8+ T cells since there is no known CD4+ T-cell specificity infiltrating the islets that is as immunodominant as the CD8+ subsets recognizing IGRP206-214 and PIB15-23. Moreover, only frequencies of β-cell-specific CD8+ T cells have been shown to be predictive of future diabetes development in NOD mice [].
Conversely, metabolic modifications were more substantial, with lower blood glucose and endogenous C peptide levels reflecting a state of β-cell rest. However, in line with the transient therapeutic effect, these metabolic modifications were limited to the treatment period and were not maintained once insulin was withdrawn. In the absence of major immune modifications following treatment, our data suggest that a mechanism of transient β-cell rest could be predominant in the observed delay in diabetes onset. Transient β-cell rest may also explain the reduced islet infiltration observed at the end of treatment, which was also lost few weeks later. In other words, β cells remain protected as long as they are kept at rest. Once the inhibitory effect of exogenous insulin on endogenous secretion is released, the disease resumes its course.
Our hypothesis of a predominant β-cell resting effect is at variance with the conclusions of Karounos et al. []. These authors could prevent diabetes in NOD mice by treating with a B25Asp insulin analogue, which binds very poorly to the insulin receptor [], but preserves the sequences targeted by autoimmune T cells. Given that the B25Asp analogue is metabolically inactive, its efficacy was considered a proof of immune protective mechanisms being at play. This discrepancy may be explained by the fact that NOD mice were treated starting at week 12, that is during the insulitic phase rather than in the earlier period targeted in our study and in previous ones []. It is therefore possible that insulin may acts through immune mechanisms when administered later in the disease process, once a considerable proportion of islet-reactive T cells is already present in pancreatic infiltrates. Even more relevant is the fact that mice were continuously treated daily throughout their lifetime, with insulin administration only stopped for 48 h every 2–4 weeks to monitor blood glucose [].
A similar approach of continuous insulin administration was followed by Atkinson et al. [], who started treatment at week 4 and continued until week 26. Mice were subsequently followed up for only 1 week after insulin withdrawal. It is thus possible that a longer follow up could have revealed an effect of diabetes delay rather than protection. The same caveat applies to mouse adoptive transfer models, which documented a similar protective effect, but with no off-treatment follow-up period [[25, 26]]. Indeed, in the study by Thivolet et al. [] protection was associated with significant hypoglycaemia, similar to the values we observed, and was lost upon early treatment discontinuation.
One major question remains unanswered: by which mechanisms does β-cell rest exert its protective effect? Two major hypotheses can be proposed. First, β-cell rest could inhibit homing and expansion of autoreactive T cells in the islets by lowering secretion of chemokines and presentation of immunogenic epitopes, respectively. The marginal decrease in islet-reactive CD8+ T cells observed suggests that this mechanism remains accessory. Second, β-cell rest could reduce local inflammation mediated by innate immune cells and β-cell-intrinsic apoptosis, independent of adaptive T-cell responses. The observation of reduced insulitic infiltrates during treatment suggests that this second mechanism may be predominant.
Our data have important implications in the design of insulin-based prevention trials. First, if our findings were translated into clinical settings, at-risk subjects would need to be treated at doses far above the safe window to avoid hypoglycemia. It should however be noted that all mouse studies to date have been performed using therapeutic preparations of human insulin, given the lack of suitable mouse formulations. Thus, we cannot exclude that lower doses would be sufficient to obtain similar effects on NOD mice if murine insulin was to be used. Indeed, much lower doses would be sufficient, on a weight basis, to lower glycemia in humans to the levels here observed. This caveat does not, however, solve the problem of safe therapeutic windows that could be used in human. Indeed, the glycemic levels associated with delayed diabetes onset were in the range of life-threatening hypoglycemia. Although mice are quite resistant to insulin-induced hypoglycemia, this risk would be unacceptable in clinical trials. Second, treatment would need to be continued indefinitely to maximize the odds of achieving a protective effect in the long term. These findings are reminiscent of what observed in a recent follow-up study of insulin autoantibody (IAA)+ subjects enrolled in the oral insulin arm of the DPT-1 study []. Despite initial T1D protection, the rate of disease development increased to a rate similar to the placebo group once therapy stopped.
The inherent difficulties in translating these findings into human invite alternative approaches. Conventional insulin immunization using suitable adjuvants is a first alternative, which would favor immune modulation over β-cell rest. Indeed, several mouse studies suggest efficacy even when vaccinating with the insulin B chain [] or the B9-23 peptide [[28, 29]] alone. The former approach is the subject of ongoing trials []. Second, insulin administration by oral or intranasal routes should be preferred to the subcutaneous one, as they have shown promising results in preclinical studies [[31, 32]] and, to a limited extent, also in clinical trials [[15, 19]]. Third, it could be questioned whether inducing tolerance to insulin would be sufficient to halt β-cell autoimmunity. Apart from the early appearance of IAA in children, there is no conclusive evidence for an initiating role of insulin in human T1D as is the case in NOD mice []. Even if we assume insulin to have an initiating role in the human autoimmune cascade, prevention trials are performed at a time when the first signs of β-cell autoimmunity (i.e. autoantibodies) are already present. It is thus likely that, at this stage, Ag spreading has already occurred, thus necessitating to target additional islet Ag specificities (e.g. glutamic acid decarboxylase) to achieve complete tolerance.
In conclusion, prophylactic insulin treatment in NOD mice may be effective only when long-term, high-dose regimens are employed, thus hampering translation into clinical trials. More targeted strategies of tolerogenic vaccination via the subcutaneous or mucosal routes using insulin and other islet Ags may offer more favorable risk-benefit profiles.
Materials and methods
Mice and insulin treatment
NOD mice, bred and housed in specific pathogen-free conditions, were injected s.c. with 0.5 U or 0.005 U per day of insulin glargine (Lantus; Sanofi-Aventis) diluted in the same vehicle used for insulin glargine preparations (water pH 4.0 containing 3 mg/L of Zn2+ and 1.7 g/L glycerol). The control group was injected with diluent only. Treatment was started at 3–4 weeks of age until week 10 and diabetes incidence followed weekly until week 30. Mice were considered diabetic after a positive urine glucose test (Glucotest strips, Roche) and confirmed by elevated glycemia (>2 g/L) measured with Accu-chek Performa reagents (Roche). This study was approved by the local ethics committee for animal experimentation.
Tissue preparation for insulitis score
Freshly isolated pancreata were fixed in 4% paraformaldehyde (PFA) for 24 h. PFA-fixed pancreata was then dehydrated for 24 h at increasing ethanol concentrations and embedded in paraffin. Sections of paraffin-embedded pancreata (4 μm) were stained with hematoxylin and eosin, ∼70 islets per mouse from different sections were analyzed, and degree of mononuclear cell infiltration was scored as: none, no infiltration; peri-insulitis, infiltration surrounding islets; and insulitis, infiltration covering more than 50% of the islet surface area.
Preparation of pancreatic islet infiltrates
Mice were sacrificed and pancreatic islets isolated after perfusion with 0.75 mg/mL Collagenase P (Roche), as described elsewhere []. If used for ex vivo analyses, islets were incubated with cell dissociation solution (Sigma) for 9 min and filtered to obtain a single-cell suspension. To measure Ag-specific CD8+ T-cell responses, islets were isolated and cultured for 7 days in the presence of 50 U/mL recombinant human IL-2 (R&D) [[33, 34]]. Infiltrating cells were then collected and analyzed.
Islet-infiltrating cells were collected and stained with PE-labeled Kd TMrs synthesized by the NIH Tetramer Core Facility and loaded with NRP-V7 (a mimotope of the IGRP206-214 epitope; KYNKANVFL), PIB15-23 (LYLVCGERL), and control TUM peptide (KYQAVTTTL). IFN-γ responses to IGRP206-214 and PIB15-23 epitopes were measured using intracellular IFN-γ staining after incubation for 5 h in the presence of 10 μM peptide and 10 μg/mL Brefeldin A. CD8-Alexa Fluor 700 (clone 53-6.7; eBioscience), CD4-PerCP (clone RM4-5; BD), CD11c-Alexa Fluor 700 (clone N418; eBioscience), IFN-γ-PE (clone XMG1.2; BD), CD45-FITC (clone 30-F11; BD), and FoxP3-PE (clone FJK-16s; eBioscience) were used. For intracellular staining, Cytofix/Cytoperm (BD) and a FoxP3 staining kit (eBioscience) were used for IFN-γ and FoxP3, respectively.
Serum C-peptide measurements
Mice were treated with insulin or diluent for 8 days prior to measurement. At day 8, they were challenged with an i.p. glucose load (1 mg/g) either 2 or 20 h after the last insulin injection and murine C-peptide levels measured after 30 min by ELISA (Alpco).
Diabetes incidence was plotted according to the Kaplan–Meier method. Incidences between groups were compared with the log rank test. For other experiments, means ± SEM are represented and compared using the nonparametric Mann–Whitney U-test or paired t-test for paired observations as appropriate. p values <0.05 were considered statistically significant. All data were analyzed using Prism software (version 5; GraphPad).
This research was supported by the Juvenile Diabetes Research Foundation (JDRF grant 1-2008-106), the European Foundation for the Study of Diabetes (EFSD/JDRF/Novo Nordisk European Programme in Type 1 Diabetes Research 2007), and the Ile-de-France CODDIM (grant Soutien aux Jeunes Equipes) (to R.M.). V.B. was supported by Ile-de-France CODDIM scholarship. R.M. is an INSERM Avenir Investigator and an Interface Assistance Publique—Hopitaux de Paris INSERM awardee. We wish to thank the National Institute of Health Tetramer Core Facility for production of Kd tetramers; and Prof. E. Larger for critical review of the manuscript.
Conflict of interest
The authors declare no financial or commercial conflict of interest.
Diabetes Prevention Trial-Type 1
islet-specific glucose-6-phosphatase catalytic subunit-related protein