Specific Inhibition of Autoimmune T Cell Transmigration Contributes to β Cell Functionality and Insulin Synthesis in Non-obese Diabetic (NOD) Mice*

Human diabetes mellitus (IDDM; type I diabetes) is a T cell-mediated disease that is closely modeled in non-obese diabetic (NOD) mice. The pathogenesis of IDDM involves the transmigration of autoimmune T cells into the pancreatic islets and the subsequent destruction of insulin-producing β cells. Therapeutic interventions leading to β cell regeneration and the reversal of established IDDM are exceedingly limited. We report here that specific inhibition of T cell intra-islet transmigration by using a small molecule proteinase inhibitor restores β cell functionality, increases insulin-producing β cell mass, and alleviates the severity of IDDM in acutely diabetic NOD mice. As a result, acutely diabetic NOD mice do not require insulin injections for survival for a significant time period, thus providing a promising clue to effect IDDM reversal in humans. The extensive morphometric analyses and the measurements of both the C-peptide blood levels and the proinsulin mRNA levels in the islets support our conclusions. Diabetes transfer experiments suggest that the inhibitor specifically represses the T cell transmigration and homing processes as opposed to causing immunosuppression. Overall, our data provide a rationale for the pharmacological control of the T cell transmigration step in human IDDM.

Human diabetes mellitus (IDDM; type I diabetes) is a T cellmediated disease that is closely modeled in non-obese diabetic (NOD) mice. The pathogenesis of IDDM involves the transmigration of autoimmune T cells into the pancreatic islets and the subsequent destruction of insulin-producing ␤ cells. Therapeutic interventions leading to ␤ cell regeneration and the reversal of established IDDM are exceedingly limited. We report here that specific inhibition of T cell intra-islet transmigration by using a small molecule proteinase inhibitor restores ␤ cell functionality, increases insulin-producing ␤ cell mass, and alleviates the severity of IDDM in acutely diabetic NOD mice. As a result, acutely diabetic NOD mice do not require insulin injections for survival for a significant time period, thus providing a promising clue to effect IDDM reversal in humans. The extensive morphometric analyses and the measurements of both the C-peptide blood levels and the proinsulin mRNA levels in the islets support our conclusions. Diabetes transfer experiments suggest that the inhibitor specifically represses the T cell transmigration and homing processes as opposed to causing immunosuppression. Overall, our data provide a rationale for the pharmacological control of the T cell transmigration step in human IDDM.
The pathogenesis of IDDM 2 involves the activation of autoimmune T killer cells. Activated autoimmune T cells transmigrate from the bloodstream through the pancreatic endothelium and into the islets of Langerhans, where they destroy insulin-producing ␤ cells. We suggested and then proved in our current work that the inhibition of T cell transmigration and homing would effectively control islet destruction and stimulate the functional recovery of the insulin-producing ␤ cells and the regeneration of the pancreatic islets.
Mice of the NOD inbred strain develop a spontaneous disease closely resembling human IDDM and have been widely and successfully used as a model of IDDM (1)(2)(3). CD8 ϩ T lymphocytes are involved in diabetogenesis in NOD mice; mice lacking CD8 ϩ T cells do not develop diabetes (4,5). The cell surface CD44 levels are elevated in activated T cells (6). Via its interactions with endothelial hyaluronan, CD44 functions as a prominent adhesion receptor in autoimmune T cell adhesion on the endothelium and the subsequent transmigration events (7)(8)(9)(10)(11). Membrane type-1 matrix metalloproteinase (MT1-MMP) (12,13) is the most important cell surface-associated proteinase that contributes to the shedding CD44 in the adherent autoimmune CD8 ϩ T cells (11,14). As demonstrated in our earlier cell-based tests (8,9,14), MT1-MMP cleavage releases the extracellular domain of CD44 from T cell surfaces and inactivates the CD44 cell receptor function. Similarly, the cleavage of CD44 by MT1-MMP plays a significant role in the regulation of tumor cell migration (15,16). By means of this regulatory proteolysis, which our earlier data suggest (8,9,14), MT1-MMP appears to control the severity of the diabetic disease and mediates the transition of T cell adhesion on endothelium to transendothelial migration that results in T cell homing into the pancreatic islets. The inhibition of MT1-MMP proteolysis of CD44 leads to an extended instead of a temporal adhesion of cytotoxic autoimmune T cells on the vascular endothelium. In a similar fashion, tissue inhibitor of metalloproteinases-2 (TIMP-2; a potent inhibitor of MT1-MMP) but not tissue inhibitor of metalloproteinases-1 (TIMP-1; a poor inhibitor of MT1-MMP) decreased T-cell transmigration and preserved insulin production in a type 1 diabetes organ culture model (17). The long term immobilization on the endothelium impedes the transmigration and homing efficiency of the diabetogenic CD8 ϩ T cells. These combined events delayed the onset of the transferred diabetes in NOD mice (8). Consistent with our data, a recent publication suggests that MT1-MMP plays an important role in regulating monocyte transendothelial migration (18).
Therapies of IDDM will require the repair of immunological tolerance breakdown, restoration of insulin-producing ␤ cell mass, or both (19). Because in IDDM the de novo developing ␤ cell populations, which are believed to originate from stem cell precursors (20 -25), are continually destroyed in the islets by the transmigrating autoimmune T cells, we hypothesized that diminishing the rate of T cell transmigration and homing would lead to both the restoration of the ␤ cell mass (25) and a clinically relevant increase in insulin production in acutely diabetic NOD mice. Here, using acutely diabetic NOD mice and performing subsequent extensive morphometric analyses and the measurement of both the C-peptide blood levels and the pro-insulin mRNA in the islets, we experimentally confirm that this hypothesis is correct.

MATERIALS AND METHODS
Animals-Mice of NOD/LtJ (NOD), C57Bl/6J, and NOD.CB17-Prkdc scid /J (NOD-scid) strains were obtained from the Jackson Laboratory (Bar Harbor, ME) and housed in the pathogen-free animal facility of the Burnham Institute for Medical Research. All animal treatment protocols have been reviewed and approved by the Burnham Institute Animal Care Committee. NOD female mice developed diabetes in ϳ5 months following birth. The onset of spontaneous diabetes was identified by assessing urine glucose levels with Diastix strips (Bayer, Tarrytown, NY) and verified by blood glucose measurement using an Ascensia Elite 1 one-touch blood glucose monitor (Bayer). Mice with blood glucose levels Ͼ450 mg/dl for three consecutive measurements were considered diabetic.
Adoptive Transfer-Adoptive transfer of diabetes was performed by intravenous injection of freshly isolated splenocytes from diabetic NOD animals into irradiated (725 Roentgens, 24 h in advance) syngeneic NOD recipients (1.5 ϫ 10 7 cells/ animal) (26). Both males and females 6 -10 weeks of age were used. Because neither sex nor age affected the incidence and the time of diabetes onset, the results were pooled. The recipients of adoptive transfer were monitored daily by blood glucose measurements for 30 days for diabetes development. Animals with blood glucose levels Ͼ450 mg/dl for three consecutive measurements were considered diabetic.
Multidose STZ Treatment-To induce diabetes, streptozotocin (STZ) (40 mg/kg in 0.1 M citrate buffer, pH 4.5) was injected intraperitoneally daily for 5 consecutive days into 8-week-old C57BL/6 female mice (27). Blood glucose levels were monitored every second day. Animals were considered diabetic when their blood glucose levels were Ͼ450 mg/dl in three consecutive measurements. Mice in the control group received an equal volume of sodium citrate buffer alone.
AG3340 Treatment-Porcine insulin (Sigma; Ն27 USP units/mg; 15-20 units/kg; one injection every 2-3 days) was injected subcutaneously into female NOD mice that had already developed acute spontaneous diabetes and also into C57BL/6 animals with STZ-induced diabetes. Insulin injections lasted for 20 days to ensure total destruction of any residual degranulated ␤ cells (28). The spontaneously diabetic NOD-and STZ-treated C57BL/6 control animals (four and five mice/group, respectively) then continued to receive insulin alone. Experimental groups of spontaneously diabetic NODand STZ-treated C57BL/6 (six mice/group each) received insulin subcutaneously jointly with AG3340 intraperitoneally (5 mg/kg; one injection every 2-3 days) for 30 days, and then insulin injections were stopped in all groups, whereas experimental groups continued to receive AG3340 intraperitoneally (5 mg/kg; one injection in every 2-3 days) for the following 30 days. Every other day, animals were subjected to measurements of blood glucose. Animals with blood glucose levels Ͼ450 mg/dl for three consecutive measurements were considered diabetic and were then sacrificed.
Immunohistochemistry, Morphometry, and Pancreatic Regenerative Capability Analyses-Pancreata were weighed, fixed by immersion in 10% formaldehyde, and paraffin-embedded. Paraffin-embedded sections were stained with guinea pig anti-insulin polyclonal serum (Linco Research, St. Charles, MO) and rabbit polyclonal antibody to glucagon (DacoCytomation, Carpinteria, CA) followed by species-specific secondary horseradish peroxidase-conjugated antibodies. The Vector VIP kit (Vector Laboratories, Burlingame, CA) was used to visualize horseradish peroxidase-stained tissue. Sections were counterstained with hematoxylin, mounted, and analyzed.
Morphometry and a pancreatic regenerative capability analyses were performed using serial 5-m-thick longitudinal paraffin sections of pancreata separated by a 50-m interval and immunostained for insulin. The microscopic images of pancreatic sections were taken using the Olympus BX51 microscope (Olympus America, Center Valley, PA) connected through a video camera to a computer. Images were analyzed using the Metamorph (Molecular Devices, Sunnyvale, CA) and Image Pro Plus software (Media Cybernetics, Bethesda, MD). The ␤ cell representation was measured using our assumption that all insulin-positive areas were represented by ␤ cells. Morphometric analysis was performed using a grid system. At least 300 fields/mouse were examined. Relative ␤ cell, non-␤ cell, and total exocrine tissue areas were calculated. The ␤ cell mass was then calculated by multiplying the relative ␤ cell representation by the corrected pancreatic weight.
For the analysis of the pancreatic regenerative capability, we used 5-m-thick longitudinal paraffin sections of pancreata separated by a 50-m interval. The sections were immunostained for insulin. The size of the individual functional islets, defined as the insulin-positive area of the sections, was measured using an Olympus BX51 microscope with a ϫ200 magnification connected through a video camera to a computer. The data were analyzed by Metamorph software and expressed in pixels (29). According to our observations, 85-90% of the functional islets in adult NOD-scid mice exceeded 80 m (2000 pixels) in size. Infiltration-free, insulin-positive, and Յ2000 pixel in size islets were considered to be either regenerating or newly formed. These miniature islets were counted in the sections of the entire pancreas and recorded. 4 -6 mice/group were analyzed by an observer under double-blinding conditions. The data were expressed as mean value ϩ S.E.
Measurement of Serum Insulin and C-peptide Levels-Mice were fasted for 16 h. Glucose (2.0 g/kg body weight, 20% solution in 0.9% NaCl) or an equal volume of 0.9% NaCl was administered intraperitoneally. The blood was withdrawn from the orbital sinus 30 min after injection. The serum insulin and the C-peptide levels were measured in triplicate using the insulin enzyme-linked immunosorbent assay kit (Crystal Chemistry, Downers Grove, IL) and the C-peptide radioimmune assay kit (Linco), respectively. The data were expressed as mean Ϯ S.E.
Q-RT-PCR-Total RNA was extracted from fresh pancreatic tissue using the RNeasy Maxi kit (Qiagen, Valencia, CA). Reverse transcription was performed using 3 g of RNA. Quan-titative real-time PCR (Q-RT-PCR) was conducted in duplicate using 25 ng of cDNA and 0.4 M primers in a 30-l final volume of 1xSYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA). Forward and reverse primers (5Ј-GCTCTCTACCT-GGTGTGTGG-3Ј and 5Ј-CTGGTAGAGGGAGCAGATG-3Ј, respectively) were specifically designed to amplify both murine Ins1 and Ins2 genes. PCR conditions were as follows: 95°C for 15 min, and 40 cycles of 95°C for 15 s, 58°C for 1 min, and 72°C for 30 s. The melt curve was analyzed, and the absence of primer dimers was verified. Relative expression levels were calculated using the following equation: 2 Ct ␤-actin Ϫ Ct insulin , where Ct is the cycle threshold. ␤-Actin RNA was used as an intrinsic control.
Statistical Analysis-The Fisher's least significant difference test and one-way analysis of variance were used to determine statistical significance. A difference was considered statistically significant if the p value was Ͻ0.05.

Inhibition of T Cell MT1-MMP in Spontaneously Diabetic
NOD Mice-To generate proof-of-principle data that the inhibition of T cell MT1-MMP and the subsequent reduction of autoimmune T cell transmigration and homing would lead to regeneration of insulin-producing ␤ cells, NOD female mice were first allowed to develop spontaneous IDDM. To compare the effects of an AG3340 treatment on spontaneous and STZinduced diabetes (27), we also used multiple injections of STZ (40 mg/kg/day; intraperitoneally for 5 days) to cause diabetes in 8-week-old C57BL/six mice. Diseased mice then received insulin alone (15 units/kg subcutaneously) for 20 days to ensure the complete destruction of residual degranulated ␤ cells (28). For the next 30 days, the control mice continued to receive insulin, whereas the experimental groups received insulin jointly with AG3340 (5 mg/kg).
The K i values of AG3340 against MMP-2, MMP-3, and MMP-13 are higher, and hence, AG3340 is less efficient (ϳ100, 300, and 200 pM, respectively). Other individual MMPs are significantly less sensitive to AG3340 inhibition (e.g. the K i values for MMP-1 and MMP-7 are 10 and 55 nM, respectively) (30,32,33). Our previous extensive studies involving AG3340, a recently designed thiirane MMP-2/MMP-9 inhibitor SB-3CT ((4-phenoxyphenylsulfonyl) butane-1,2-dithiol) (34,35), and epigallocatechin-3-gallate (EGCG), a major natural catechin of green tea and a nonspecific inhibitor of MMPs (36 -38), demonstrated that in a transfer diabetes model in NOD mice, only AG3340, the potent antagonist of MT1-MMP, delivered clinically relevant effects (14). Although both EGCG and SB-3CT are poor inhibitors of MT1-MMP, they are capable of targeting MMPs distinct from MT1-MMP. Because of the wide-ranging specificity of the MMP inhibitors, a simultaneous assessment of AG3340, SB-3CT, and EGCG permitted us to conclude that only T cell MT1-MMP plays a significant role in IDDM, whereas the combined effect of all other MMPs, including MMP-2 and MMP-9, both of which are efficiently inhibited by SB-3CT, is far less important. Overall, our earlier comprehensive studies led us to conclude that because of the primary role of MT1-MMP in shedding of T cell CD44 and the subsequent effects of CD44 shedding in T cell transendothelial migration, only MT1-MMP antagonists such as AG3340, as opposed to other broad range inhibitors of MMPs, are efficient in delaying IDDM transfer to NOD mice (8,9,14).
Insulin injections were then suspended in both the control and the experimental group mice (Fig. 1, A and B). Those mice that received insulin alone became hyperglycemic in a matter of days following the cessation of insulin injections and were then sacrificed in accordance with National Institutes of Health guidelines. Simultaneously, the spontaneously diabetic NOD mice that received insulin jointly with the inhibitor continued to be normoglycemic/mildly hyperglycemic (Ն450 mg/dl) for several weeks as long as the inhibitor injections were continued but without the injections of external insulin (Fig. 1A). These results support and extend our preliminary observations, which we reported earlier (9).
On the other hand, mice with STZ-induced diabetes did not improve their hyperglycemia following AG3340 treatment. It is well established that STZ uses Glut-2 as the receptor for entering into the target cell that eventually kills (39 -41). STZ does not provoke an autoimmune process; rather, it directly poisons the Glut-2 receptor-positive ␤ cells. Because T cell activity is not part of the picture, it is understandable why AG3340 does not bring any benefit to STZ-treated animals (Fig. 1B).
Glucose Tolerance, C-peptide, and Insulin mRNA-A routine glucose tolerance test further confirmed that spontaneously diabetic NOD mice that received AG3340 and healthy NODscid animals were capable of clearing injected glucose (intraperitoneally, 2 g/kg of body weight) from the blood at a similar rate (Fig. 1C). NOD-scid mice share an NOD genetic background but lack T and B cells and thus are broadly used as diabetes-and insulitis-free controls for NOD mice. To confirm that insulin was produced by the body, we determined the C-peptide levels in the blood of healthy NOD-scid mice and diseased NOD mice that received the inhibitor. There was an ϳ50% reduction in the C-peptide level in both fasting and glucose-stimulated (2 g/kg) mice, which received the inhibitor, relative to the healthy NOD-scid mice. Control diabetic NOD animals that received insulin alone did not have detectable C-peptide in the blood. In mice with STZ-induced diabetes, the levels of C-peptide were insignificant independently of the AG3340 treatment ( Fig. 2A). During the 34-day period following the cessation of insulin injections, the C-peptide concentrations in fasting mice were in the range of 60 -100 pM in the serum of diabetic NOD mice that received AG3340, whereas concentrations in the range of 180 -220 and 25-30 pM were measured in healthy NOD-scid mice and in C57BL/six mice with STZ-induced diabetes, respectively (Fig. 2B). The C-peptide concentrations in the serum correlated with the combined level of the Ins1 and Ins2 gene mRNA in the pancreas. According to the Q-RT-PCR analysis, the proinsulin mRNA level in spontaneously diabetic NOD mice that received the inhibitor was 30 -40% of that detected in healthy NOD-scid mice. The proinsulin mRNA level was close to zero in mice with STZinduced diabetes.
Regeneration of ␤ Cells in Acutely Diabetic NOD Mice-To examine the functionality of the surviving and either the regenerating or the newly formed islets, the murine pancreata were excised, sectioned, and stained with antibodies to insulin and glucagon. Consistent with our earlier data (9), AG3340 caused an increase in the number of intact islets and the islets with limited peri-islet insulitis in spontaneously diabetic NOD mice. In addition, AG3340 caused a recovery of insulin-producing ␤ cells in the pancreatic parenchyma. Intriguingly, the smallsized, presumably rescued, regenerating, or newly formed, islet-like structures were predominantly observed in those spontaneously diabetic NOD animals that received the inhibitor. Similar structures were infrequent either in diabetic NOD mice that received insulin alone or in mice with STZinduced diabetes regardless of the inhibitor treatment. These islet-like structures were free from mononuclear infiltration and produced insulin and glucagon, thus providing evidence of the functional recovery of the hormone-secreting cells. In STZ-treated mice, the insulin staining was insignificant, suggesting a complete or near complete destruction of ␤ cells (Fig. 3A).
Because insulin synthesis is a surrogate measure of ␤ cell mass, we directly determined the ␤ cell mass in mice (28,42). ␤ cell mass was calculated from pancreatic sections made throughout the whole organ and stained for insulin. We used a morphometric analysis with a grid system to analyze at least 300 fields/ animal. The significant increase in the pancreatic insulin mRNA content in AG3340-treated diabetic NOD mice was accompanied by a 9-fold increase in ␤ cell mass (0.386 Ϯ 0.037 mg in insulin ϩ AG3340 mice versus 0.048 Ϯ 0.02 mg in the insulin alone NOD control; p Ͻ 0.05) (Fig. 3, B and C). In diabetic NOD mice that received the inhibitor, ␤ cell mass was equal to 40% of the mass of healthy NODscid mice.
To further support our data, an observer under double-blinding conditions counted the newly formed, infiltration-free, insulinpositive islets having the size of Յ2000 pixels (ϳ80 m diameter) (29) in the 5-m longitudinal paraffin sections prepared at 50-m intervals throughout the whole organ (4 -5 mice/group) (Fig. 3D). We selected a 2000-pixel threshold because according to our observations, the average size of the functional mature islets in the adult, diabetes-free NOD-scid mice significantly exceeded 2000 pixels. There were, on average, 35 small-sized islets per animal in diabetic NOD mice that received AG3340. In turn, only a few small-sized islets were detected both in healthy NOD-scid mice and in diabetic NOD mice that received insulin alone. These data suggest that the regenerating processes were enhanced in diabetic mice because of the presence of AG3340. and three acutely diabetic experimental NOD mice (filled symbols) were treated for 50 days with insulin. The experimental mice began receiving AG3340 injections for 30 days starting at day 20. At day 54, mice received glucose (2.0 g/kg intraperitoneally). Blood glucose was then measured at the indicated times after injection.
In general agreement with our data, TIMP-1 has been shown to inhibit cytokine-induced apoptosis in multiple ␤ cell lines (43) and, in addition, TIMP-1 overexpression in mouse ␤ cells enhanced the replication of ␤ cells in TIMP-1 transgenic mice and contributed to the regeneration of ␤ cell mass in the STZ model of IDDM (44).
The Assessment of Immunosuppression-To determine whether AG3340 specifically represses the T cell transmigration processes as opposed to causing immunosuppression, we used an adoptive transfer of diabetes model. For this purpose, NOD mice with spontaneously developed diabetes were treated with insulin and AG3340 or insulin alone. In 25 days, the total splenocyte population was isolated from the spleens of experimental and control animals. The isolated splenocytes (1.5 ϫ 10 7 ) were injected intravenously into young irradiated (725 Roentgens, 24 h in advance) syngeneic NOD recipients (six mice/group) (26). The onset of spontaneous diabetes was monitored daily by assessing urine glucose levels and verified by blood glucose measurements. Mice with blood glucose levels Ͼ450 mg/dl for three consecutive days were considered diabetic. There was no difference in the onset of diabetes among the groups, suggesting that no specific immunosuppressive effects were induced by AG3340 treatment (Fig. 4).
In conclusion, we are now confident that inhibition of T cell transmigration by the MT1-MMP antagonists including AG3340 provides a level of protection against diabetes by controlling islet-destructive autoimmunity and stimulating the FIGURE 2. Blood C-peptide levels in mice. A, AG3340 increases the blood C-peptide level in spontaneously diabetic NOD mice. Left, mice were fasted for 16 h. Right, C-peptide was measured 30 min following glucose injection (2.0 g/kg intraperitoneally). C-peptide was measured in the serum of healthy NOD-scid mice, spontaneously diabetic NOD mice, and C57BL/six mice with STZ-induced diabetes. Following the manifestation of diabetes, the last two groups of mice received insulin for 50 days, and starting on day 20, they received AG3340 injections. The level of C-peptide was measured every 4th day starting on day 40 until the end of the experiment (Fig. 1). Average value is shown for each group. B, monitoring the C-peptide blood levels in mice. C-peptide was measured every 4th day starting on day 52 (2 days after the end of insulin treatment) until the end of the treatment (Fig. 1). FIGURE 3. Islet regeneration in diabetic NOD mice. A, insulin and glucagon staining of representative pancreatic sections counterstained with hematoxylin of spontaneously diabetic NOD mice that received insulin alone and insulin ϩ AG3340 and mice with STZ-induced diabetes that received insulin ϩ AG3340. B, insulin Ins1/Ins2 mRNA level in the pancreata of healthy NOD-scid mice and in diabetic NOD mice that received insulin alone and insulin ϩ AG3340 (three mice/group each). The levels were normalized relative to the ␤-actin mRNA. C, functional ␤ cell mass in the pancreata of healthy NOD-scid mice (five mice/group) and in diabetic NOD mice that received insulin alone and insulin ϩ AG3340 (four and six mice/group, respectively). D, a number of small-sized, infiltration-free, insulin-positive islets in healthy NOD-scid mice (five mice/group) and in diabetic NOD mice that received insulin alone and insulin ϩ AG3340 (four and six mice/group, respectively). Freshly diabetic NOD mice received insulin alone and insulin ϩ AG3340 for 25 days. Animals were then sacrificed, and splenocytes were isolated, pooled for each group, and then adoptively transferred (1.5 ϫ 10 7 intravenously) into irradiated (725 Roentgens, 24 h in advance) syngeneic recipients (6 recipients/group). The recipients were monitored daily for 30 days. Animals with blood glucose levels Ͼ450 mg/dl for three consecutive measurements were considered diabetic.