Effects of Autoimmunity and Immune Therapy on β-Cell Turnover in Type 1 Diabetes

Female NOD/LtJ mice and female NOD/scid mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and maintained under pathogen-free conditions at our facility. NOD mice were screened for glycosuria three times a week beginning at 12 weeks of age and were considered diabetic when two random glucose levels were >200 mg/dl. Glucose levels were measured in whole blood from the tail vein with the Glucometer Elite XL (Bayer, Elkhart, IN). Animal use was approved by the Columbia University Animal Use Committee.

NOD mice were treated with either whole or F(ab′)₂ fragments of the mAb 145-2C11 (anti-mouse CD3) at 20 and 50 μg/day i.p., respectively, for 5 days within 0–2 days of onset of hyperglycemia (21,22). Reversal of diabetes was considered to have occurred when random glucose levels were <200 mg/dl.

A total of 10 million splenocytes were purified from NOD mice with diabetes and administered, via the tail vein, to 8-week-old NOD/scid mice. Some mice also received 2 million CD4⁺CD25⁺ or CD4⁺CD25⁻ cells within 12 days after the transfer of diabetogenic splenocytes. These cells were purified by magnetic bead–activated cell sorting of splenocytes from mice that had been treated with whole or F(ab′)₂ fragments of the anti-CD3 mAb and showed reversal of diabetes (22).

A subset of mice underwent a glucose tolerance test after a 14- to 16-h fast. Dextrose at 1 mg/g body wt i.p. was injected, and glucose levels were measured in whole blood from the tail vein taken before and 15, 30, and 60 min after the injection. Random glucose levels are reported in Fig. 1. These values were measured with a glucose meter in the morning in mice that had access to food.

Pancreata were resected, weighed, fixed, and embedded in paraffin. Noncontiguous 5-μm pancreatic sections were stained using antibodies to insulin, 4′,6-diamidine-2-phenylindole dihydrochloride (DAPI), and either Ki67, active caspase-3, or 5-bromo-2-deoxyuridine (BrdU). The bound antibodies were detected by immunofluorescent secondary antibodies. The anti-Ki67, anti-cleaved caspase-3, and anti-BrdU antibodies were purchased from Novocastra (New Castle upon Tyne, U.K.), Trevigen (Gaithersburg, MD), and Accurate Chemical and Scientific (Westbury, NY), respectively.

Insulin-positive and total pancreatic area was calculated from 8–10 noncontiguous 5-μm pancreatic sections from each mouse at ×10 magnification using a Nikon light microscope, Spot Insight digital imaging camera, and Image-Pro Plus software. β-Cell area was determined by calculating the total insulin-positive cells/islet area (in millimeters squared) with the assumption that all insulin-positive area represented β-cells. To determine the percentage of replicating and apoptotic β-cells, ×40 images were taken of pancreatic sections stained with antibody to insulin, DAPI, and either Ki67 or active caspase-3, respectively. The percentage of replication was defined as the number of Ki67⁺ insulin⁺ cells/total insulin⁺ cells × 100. The percentage of apoptotic cells was expressed as the number of active caspase-3⁺ insulin⁺ cells/number of insulin⁺ cells × 100.

In certain experiments, the total number of newly formed β-cells was determined by administering BrdU (Fisher Scientific International) in the drinking water of the mice for 3 weeks. Both diabetic NOD mice treated with anti-CD3 mAb and NOD/scid mice were used for these studies. BrdU was added to the drinking water (1 mg/ml), protected from light, and the drinking water was changed weekly. After 3 weeks, the animals were killed and pancreatic sections stained for BrdU, DAPI, and insulin using previously described methods (23).

To improve the sensitivity of detection of β-cells, additional immunohistochemical staining was performed on pancreatic sections from mice with new-onset diabetes. Two secondary antibodies were used to detect guinea pig–derived anti-insulin antibodies—donkey-derived anti–guinea pig antisera congugated to Cy2 (Jackson Immunoresearch Laboratories, West Grove, PA). β-Cells were also identified by staining with anti-GLUT2 antibody (Research Diagnostics Division of Fitzgerald Industries International, Concord, MA) conjugated to donkey-derived anti-goat antisera conjugated to Cy3 (Jackson) (24).

Sections of 5 μm were stained for hematoxylin and eosin and graded for insulitis. The inflammatory lesions in the islets were graded (multiplier) as no lymphocytes present (1), scattered single cells in the islet (2), periinsulitis (3), more extensive inflammatory cell infiltrates but preservation of the islet architecture (4), or extensive inflammatory infiltrate and loss of the architecture (5). A composite insulitis score was calculated by multiplying the number of islets in each category by the multiplier and dividing by the total number of islets.

Groups of 4–10 mice were studied. The data are expressed as means ± SE. Data from treatment groups of mice were compared with a two-sided Student’s t test. The frequency of the response to treatment between groups was compared by a χ² test. P < 0.05 was considered statistically significant.

To understand the changes in β-cells during diabetes and treatment with immune modulatory agents, we first characterized β-cell area and the rates of β-cell replication during progression of spontaneous diabetes in NOD mice and compared these measurements with those of NOD/scid mice, which develop neither diabetes nor insulitis. The mice that were studied were between 6 and 16 weeks of age, which spans the period of time from the first appearance of insulitis until the development of diabetes (Fig. 1). Between 6 and 16 weeks of age, there was a progressive increase in insulitis in the NOD mice. The random, fasting, and stimulated glucose levels were indistinguishable from NOD/scid mice until 12–14 weeks, when the random glucose levels were modestly but significantly higher than the level in age-matched NOD/scid mice (P < 0.05). β-Cell area was reduced in NOD compared with NOD/scid mice as early as 6 weeks of age but was statistically different only at the time of diagnosis of diabetes (P < 0.001, Fig. 1). At diagnosis of diabetes, the β-cell area, identified by insulin staining, was 8.7% of the area found in age-matched NOD/scid mice.

Beginning at 6 weeks of age, the first time point studied, the percentage of replicating β-cells was increased in the NOD mice compared with the NOD/scid controls (P < 0.05) and was significantly higher than in NOD/scid until diagnosis of diabetes (Figs. 1 and 2). There was a significant reduction in the rates of β-cell replication in NOD/scid mice with age (P < 0.02 at 10 and 13 weeks of age and P < 0.05 at 16 weeks of age vs. 6 weeks of age), but the replication rate increased with age in NOD mice. The initial increase in β-cell replication preceded detectable differences in random, fasting, and stimulated glucose levels. The percentage of replicating β-cells more than doubled at 13 weeks compared with 9–11 weeks. The β-cell replication in diabetic NOD mice was >10-fold greater than in aged-matched NOD/scid mice at that time (Figs. 1 and 2).

The stimuli for the enhanced β-cell replication rates might involve a modest elevation in glucose levels, a feedback from the declining β-cell mass, and/or a developmental feature of NOD mice. The β-cell replication rates were marginally (inversely) correlated with β-cell area (r = −0.54, P = 0.05), but replication rates were not associated with glucose levels (r = 0.35, P = 0.22).

The changes we observed in NOD mice before and after the development of diabetes could have been due to a developmental phenomenon or a unique feature of the NOD strain and not related directly to islet inflammation. Therefore, to study directly the effects of the inflammatory cells on β-cell replication, we transferred diabetogenic spleen cells into NOD/scid recipients and studied the changes in β-cell turnover in the recipient mice. The recipient NOD/scid mice were studied before and 2, 3, 4, and 5 weeks after transfer of diabetogenic splenocytes, compared with age-matched NOD/scid mice that did not receive a transfer. The rates of diabetes in the NOD/scid recipients of splenocytes from diabetic NOD mice were 0% at 2, 13% at 3, 25% at 4, and 67% at 5 weeks after transfer. The random glucose levels were significantly different from controls beginning at 4.5 weeks posttransfer (196 ± 44 vs. 84 ± 4 mg/dl, P = 0.003, Fig. 3). The loss of β-cell area coincided with the increase in glucose levels, and the area was significantly reduced from 3 weeks onward. β-Cell replication, however, increased significantly before there was a significant increase in the glucose level, beginning 2 weeks after transfer, and reached a peak level at 3.5 weeks after transfer (3.14 ± 1.07% in recipients of diabetogenic splenocytes vs. 0.22 ± 0.05% in controls, P = 0.007, Fig. 3).

Studies in diabetic NOD mice and clinical studies have shown maintenance or even improvement in C-peptide responses with anti-CD3 mAb treatment, but β-cell mass cannot be directly measured in humans (16–18). We therefore tested whether immune modulatory therapy will result in recovery of β-cell area in NOD mice treated with anti-CD3 mAb or Tregs. Newly diabetic NOD mice received a 5-day course of anti-CD3 mAb, and mice with a persistent glucose level of <200 mg/dl were considered to be responders. Some mice were studied at 3 weeks after mAb treatment, and others were followed for a longer duration (6–19 weeks). Overall, diabetes reversal was 48% (14 of 29) after mAb treatment. From an average glucose level of 404 ± 58 mg/dl at diagnosis (n = 5), the glucose levels fell to 166 ± 24 mg/dl (n = 7) 3 weeks after anti-CD3 mAb (P = 0.01). β-Cell area increased more than fourfold from 0.086 ± 0.04 (n = 5) to 0.367 ± 0.065 (n = 10) (P = 0.01) (Fig. 4).

During recovery following anti-CD3 mAb, there are opposing processes that may affect the β-cell area, including death of β-cells, heightened rates of β-cell replication that exist at the time of diagnosis, and general improvements in glucose levels. Furthermore, the assessment of β-cell replication rates may be affected by the specificity of the autoimmune response; if the autoimmune process selectively affects matured β-cells, the replication rate may appear to be elevated because the replicating cells are not destroyed while the total number of cells is decreasing. We therefore determined whether arresting the autoimmunity does lead to increased production of new β-cells. We added BrdU to the drinking water of the diabetic mice that were treated with anti-CD3 mAb for the entire 3-week period following anti-CD3 mAb and compared the proportion of β-cells that were BrdU⁺ in the treated NOD mice to NOD/scid mice that received BrdU for the same period of time. An average of 10.9 ± 1.38% (n = 6) of the insulin⁺ cells were BrdU⁺ in the mice that had been treated with anti-CD3 mAb vs. 4.22 ± 0.49% (n = 5) in control NOD/scid mice (P = 0.0003) (Fig. 5), indicating that the heightened rates of replication that were identifed in the mice with diabetes did result in an increased production of new β-cells.

However, ∼90% of the β-cells identified after anti-CD3 mAb did not stain with BrdU. Our finding of BrdU⁻ insulin⁺ cells could be explained by transdifferentiation of existing cells into insulin⁺ β-cells or a failure to identify β-cells at the diagnosis because of the absence of insulin staining. We believed that the former explanation was unlikely, particularly considering the time course of the recovery. The latter could have occurred if, for example, β-cells had been degranulated as a result of hyperglycemia but not destroyed. We therefore restudied islets in newly diabetic NOD mice using two secondary antibodies against insulin and by staining for GLUT2 to identify insulin⁻ β-cells. The second secondary antibody used contained a Cy5 conjugate that allowed us to reduce background staining and enhance the insulin-specific signal. In islets from four NOD mice newly diagnosed with diabetes, we were able to identify very weak insulin staining with the use of the use of the two anti-insulin secondary antibodies that could not be identified by conventional staining. Moreover, we found GLUT2⁺ cells within the islets that had not been previously identified as β-cells when stained for insulin only (Fig. 6). These findings indicate that there are β-cells present at the diagnosis of diabetes that are not identified with conventional approaches and that these cells account for the majority of the recovered β-cell mass.

To confirm the importance of islet inflammation in stimulating β-cell replication, we studied β-cell dynamics in these models after the insulitis had been dampened by immune therapy. We examined β-cell replication in diabetic NOD mice treated with anti-CD3 mAb with a resultant correction of clinical diabetes and in NOD/scid mice following adoptive transfer of diabetogenic lymphocytes and CD4⁺CD25⁺ Tregs and prevention of the development of clinical diabetes. After anti-CD3 mAb treatment, the percentage of Ki67⁺ β-cells was decreased from 3.09 ± 0.8 to 1.57 ± 0.28% (P < 0.05, Fig. 4). The percentage of apoptotic, caspase-3⁺ β-cells was also decreased from 60.5 ± 10 to 26.6 ± 4.6% (P < 0.01).

In adoptive transfer studies, we harvested CD4⁺CD25⁺ T-cells from NOD mice that had been successfully treated with anti-CD3 mAb and mixed these cells with splenocytes from diabetic NOD mice that were transferred to NOD/scid recipients. The mixture of the Tregs reduced the rates of diabetes from 11 of 14 to 1 of 6 (χ² = 6.71, P < 0.01). Similar to our findings with anti-CD3 mAb treatment, although β-cell area was increased from 0.018 ± 0.005 (n = 10) to 1.98 ± 0.273 (n = 5) (P = 0.003) when CD4⁺CD25⁺ T-cells were added to the adoptively transferred splenocytes, β-cell proliferation was significantly reduced from 2.95 ± 0.27 (n = 5) to 1.98 ± 0.26 (n = 5) (P = 0.035) (Table 1).

NOD mice that had been treated with anti-CD3 mAb and showed correction of hyperglycemia were followed for longer periods of time (n = 7, range of follow-up 6–18.5 weeks) and compared with mice that were studied 3 weeks after anti-CD3 mAb. The random glucose levels were similar 3 weeks (166 ± 24 mg/dl, n = 7) and 6 weeks (146 ± 16 mg/dl, n = 7) after mAb treatment. However, there was a reduction in the β-cell area over time (0.367 ± 0.065 vs. 0.166 ± 0.038, P = 0.03) (Fig. 4), which was due to an increase in the percentage of apoptotic β-cells (26.6 ± 4.64 vs. 70.3 ± 10.2%, P < 0.01). Unlike the situation in NOD mice developing diabetes in which there was a compensatory increase in β-cell proliferation with a declining β-cell mass, β-cell proliferation did not increase in the mice followed for longer durations but were instead slightly reduced compared with mice studied 3 weeks after anti-CD3 mAb (1.57 ± 0.28 vs. 1.11 ± 0.15%, P = 0.25).

We have shown that there is increased β-cell replication during the development of type 1 diabetes in the NOD mouse that precedes both a significant elevation of glucose levels and a reduction in β-cell mass. Unlike NOD/scid and other mice in which β-cell replication decreased with age, the rates of β-cell replication increased more than twofold as the islet inflammatory lesions progressed (23). The increased β-cell proliferation is due to the inflammatory infiltrate itself because a significant induction of β-cell replication also occurred when diabetogenic splenocytes were transferred into NOD/scid recipients. Immune therapies, namely anti-CD3 mAb or Tregs, that render the inflammatory cells nonpathogenic also reduce β-cell replication.

Hyperglycemia is the best characterized stimulant of β-cell replication, and modest degrees of elevated glucose are thought to be responsible for the recovery of islet mass following partial pancreatectomy and in settings of increased insulin resistance (5–7,25). Significant metabolic stress, as suggested by the numbers of degranulated β-cells could also be a potent stimulator of β-cell replication; thus, although circulating blood glucose concentrations were only modestly elevated, the metabolic control could be achieved by exhaustion of insulin secretion. However, in our studies, β-cell replication was not correlated with glucose levels in the pre-diabetic mice. We cannot exclude that modest changes in glucose levels that are not reflected by significant changes in the random glucose levels can serve as a stimulus for replication in the pre-diabetic mice. Indeed, our findings of degranulated β-cells in NOD mice at the time of diagnosis would support a role of metabolic stress in stimulating β-cell replication.

The decline in β-cell replication after immune therapy with either anti-CD3 mAb or Tregs indicates that the pathogenic immune response is responsible for the proliferation rates. An increased metabolic load on each β-cell, created by the immunologic destruction of a percentage of β-cells, could create a stimulus for β-cell replication. However, glucose per se does not appear to account for the stimulus because the glucose levels were slightly elevated after anti-CD3 mAb compared with the pre-diabetic mice in which the β-cell replication rates were higher. Furthermore, the failure of β-cell area to recover fully after anti-CD3 mAb therapy would likewise suggest that total β-cell area alone is not a regulator of β-cell replication.

Other studies have suggested that the inflammatory cells and/or factors they produce may be responsible for β-cell replication. Investigations in transgenic models have shown loss of immunologic tolerance and induction of β-cell replication when interferon-γ was expressed as a transgene in the β-cells (26,27). Likewise, more recent studies have indicated that transforming growth factor-β may have a direct effect on β-cell replication (28). Maedler et al. (8) found that upregulation of cellular FLICE (caspase-8)-inhibitory protein by culture of islets with transforming growth factor-β protected β-cells from glucose-induced apoptosis, restored β-cell replication, and improved β-cell function. These effects were mediated through Fas receptor activation. Expression of antiapoptotic genes such as heme oxygenase-1 may contribute to β-cell hypertrophy following partial pancreatectomy and can be induced by the cytokine interleukin-10 (29,30).

β-Cell area, measured by conventional approaches, increased approximately fourfold after treatment with anti-CD3 mAb. The β-cell area remained below the area found in nondiabetic age-matched NOD/scid mice but was significantly greater than the area at the time of diagnosis of diabetes in NOD mice. Our studies with BrdU labeling showed that the increased rate of β-cell replication that was found in the mice at the time of onset of diabetes does, in fact, result in an increased number of new β-cells. Approximately 10% of the β-cells were produced during the 3-week period following anti-CD3 mAb therapy, which is about threefold higher than in NOD/scid mice in which BrdU was given for the same period of time. However, it is important to note that ∼90% of the β-cell area that is identified after mAb therapy does not represent new β-cells, and our additional studies with GLUT2 staining and more sensitive detection of insulin confirmed that the majority of the β-cells detected after anti-CD3 mAb were present at the time of diagnosis of hyperglycemia but not identified by conventional staining for insulin. This suggests that the recovery of the β-cell area after mAb treatment represents, at least in part, regranulation of β-cells with insulin. These findings are consistent with an important metabolic contribution to the β-cell dysfunction that occurs at the time of onset of disease and may help to explain the “honeymoon” that is seen soon after onset in which insulin production improves after institution of metabolic control. They may account for the apparent effect of intensive glucose control on improvement in β-cell function that was seen in the Diabetes Control and Complications Trial (31). Importantly, these findings also suggest that the number of β-cells present at diagnosis may be greater than can be appreciated by metabolic studies done at that time because the cells may be dysfunctional and/or degranulated. Nonetheless, they appear to have the capacity to recover.

After the initial recovery, however, there was loss of β-cell area. The random glucose level was not a good reflection of the changes in β-cell area, since the significant decline in area was not mirrored by changes in the glucose levels. Most likely, the residual β-cell area was able to meet the metabolic requirements. The decline in area was due to increased rates of β-cell apoptosis. In contrast to the decline in β-cell area in pre-diabetes or after adoptive transfer of diabetes, there was not an increased rate of β-cell replication after treatment with the anti-CD3 mAb. Before diabetes, the infiltrating pathogenic T-cells stimulated β-cell proliferation, whereas after immune treatment with anti-CD3 mAb, the affected immune response is no longer able to stimulate β-cell replication. The same conclusion was reached from adoptive transfer studies in which cotransfer of regulatory CD4⁺CD25⁺ T-cells improved β-cell area but decreased β-cell replication.

We did not follow mice for longer than 18 weeks to determine whether the loss would continue or remain at a lower but stable β-cell area. The absence of the compensatory β-cell replicative response suggests that the loss of β-cell area after week 3 was due to an apoptotic process that was initiated at or soon after diagnosis and had evolved over time without a compensatory increase in replication. We cannot be certain of the relevance of our findings of β-cell dynamics for the changes that have been seen in human disease, but it is of interest that the same pattern of slow decline in measured β-cell capacity has been seen in human trials of anti-CD3 mAb and in long-term follow-up of islet allografts in patients who are maintained on immune suppressive medications (16,18,32). These findings suggest that approaches that can directly inhibit apoptotic pathways and/or stimulate β-cell replication would be a useful adjunct to immune modulation in these setting.

This study was supported by grants DK068678, DK57846, AI-98-010, P30 DK063608, DK064101 (to J.A.K.), and DK063061 (to A.B.B.), a Pediatric Endocrine Fellowship and Clinical Scholar award from the Lawson Wilkins Pediatric Endocrine Society, the March of Dimes, Friends United for Juvenile Diabetes Research, the New Jersey Foundation for Diabetes Research, and the Juvenile Diabetes Research Foundation.

The authors thank Youping Liu and Michelle Nowroozi for their technical assistance.