Improved β-Cell Survival and Reduced Insulitis in a Type 1 Diabetic Rat Model After Treatment With a β-Cell–Selective K_ATP Channel Opener

DR-BB rats were from our own colony at the Hagedorn Research Institute (11). This study was divided into six independent experiments, with an average of 18 rats (three groups of 6 rats each) per experiment, totaling 108 rats (36 rats per group). One rat was killed because of unrelated disease early in the study. Rats were 24–33 days old at the beginning of each new experiment (day 0). They were weaned 2–3 days before the start of the experiment and randomized into one of the three treatment groups. An outline of the experimental setup is shown in Fig. 1. All experiments were conducted in accordance with Danish Legislation and with approval from the Danish Animal Experimentation Inspectorate.

On days 0–14, the rats were injected three times weekly with 2 μl/g body wt of 2.5 mg/ml poly(I:C) suspension (Sigma) and five times weekly with 300 μl of concentrated culture supernatant produced from the DS4.23 hybridoma (a gift from Jan Rozing) that secretes the rat anti-rat ART-2⁺−depleting antibody, as previously described (11). For those rats that did not exhibit severe hyperglycemia (blood glucose >20 mmol/l) until after day 14, the treatment was continued until the first day when blood glucose reached >20 mmol/l.

The rats were divided into three treatment groups: vehicle, 40 mg/kg diazoxide (Sigma), or 40 mg/kg NN414 {6-chloro-3-(1-methylcyclopropyl)amino-4H-thieno[3,2-e]1,2,4-thiadiazine 1,1-dioxide; Novo Nordisk}. These compounds were administered every 8 h by gavage from the morning of day 1 to the afternoon of day 19. The solvent was an aqueous solution of 40 mg/g glycerol, 4 mg/g gelatin, 6 mg/g sodium carboxymethylcellulose, and 1 mg/g methyl parahydroxybenzoate. Using a rat-feeding needle, 2 ml/kg of solvent alone, diazoxide, or NN414 solution was injected intragastrically.

All rats were weighed daily. Blood glucose was measured daily 1 h before the afternoon dosing using a capillary blood sample from the tail obtained with Precision Xtra (Medisense). After the onset of severe hyperglycemia, urine glucose and ketones were measured daily and the animals were treated with insulin.

To avoid complications resulting from hyperglycemia, insulin therapy was initiated at the first incidence of nonfasting blood glucose >20 mmol/l. Insulin (zinc-protracted bovine insulin; Novo Nordisk) was then administered three times daily in conjunction with the gavage. The insulin dosage was adjusted on a daily basis according to weight gain/loss, blood glucose level, glucosuria, and ketonuria; on average, diabetic rats received 12 units · kg⁻¹ · day⁻¹. The last insulin dose was given on day 21 in the morning.

On days 0, 7, and 14, a 60- to 80-μl blood sample was drawn from the tail vein and collected into EDTA-coated tubes. On day 21, plasma was obtained from the blood samples collected in connection with the intravenous glucose tolerance test (IVGTT; see below). Plasma was stored at −20°C until analysis. The samples were analyzed for insulin (day 0, 7, and 14 samples) and drug (days 7, 14, and 21) content.

Insulin analysis was carried out as a sandwich enzyme-linked immunosorbent assay. The mouse anti-insulin monoclonal antibody Hui-18 (Novo Nordisk A/S) was used as catcher, and the guinea pig anti-insulin antibody GP4042D (Novo Nordisk A/S) was used as detector antibody. Plates were blocked with wash buffer (PBS/Tween). The antibody A-TNP (raised against the hapten trinitrophyl; Novo Nordisk A/S) was included in the buffers to reduce unspecific binding, and peroxidase-conjugated goat anti−guinea pig IgG (Chemicon) was used as the secondary antibody. Finally, the color was developed with tetramethylbenzidine substrate.

NN414 and diazoxide analysis was carried out by high-turbulence liquid chromatography and tandem mass spectrometry detection using a Turboflow 2300 HTLC system for separation and an MDS/Sciex API 3000 mass spectrometer for detection.

HbA_1c was measured in 5 μl whole blood sampled from the tail vein (days 0, 7, and 14) or the periorbital plexus (day 21) in connection with the IVGTT. The blood samples were drawn into a 5-μl heparin-coated capillary tube and stored at −20°C until analyzed. The blood was mixed with 250 μl hemolysis buffer. HbA_1c was measured on a COBAS Mira by the immunoturbidimetric test, and the hemoglobin was measured by colorimetric test using an HbA_1c kit (Roche), according to the manufacturer’s instructions. The percentage of HbA_1c was then calculated.

On day 21, the rats were fasted from the morning. Diabetic rats were treated with insulin as on the previous day. After a 5- to 6-h fast, the rats were anesthetized with 1.5 μl/g body wt of a 1:1:2 solution of Dormicum (midazolam 5 mg/ml; Roche), Hypnorm (fentanyl citrate 0.315 mg/ml and fluanisone 10 mg/ml; Janssen), and sterile water. A baseline 600-μl blood sample was taken from the fully anesthetized rats. A glucose bolus (1 g/kg) was injected intravenously into the tail, and then 2 and 30 min after the glucose challenge, 600-μl blood samples from the periorbital plexus were collected in EDTA tubes containing 250 KIU Trasylol (aprotinin)/ml whole blood. Plasma was prepared and stored at −20°C for C-peptide analysis. C-peptide was determined using a radioimmunoassay for rat C-peptide (Linco) according to the manufacturer’s instructions. In some of the experiments, blood samples from untreated, nondiabetic littermates were included as controls.

The pancreata were removed en bloc with the intestines and fixed in 4% paraformaldehyde for 24 h, dissected free of surrounding tissue, and weighed. The pancreata were then fractionated by the smooth fractionator method, with f(1) = 1/4 (14,15); two paraffin-embedded blocks per pancreas were prepared, each containing 8–11 systematically randomly selected pancreatic tissue cubes. Sections (3 μm) were stained for insulin by an indirect peroxidase technique and VectorRed, and for glucagon, somatostatin, and pancreatic polypeptide combined, by the streptavidin-biotin-peroxidase technique and diaminobenzidine + Ni, lightly counterstained with Mayer’s hematoxylin and mounted with Pertex (see Rolin et al. [16] for a detailed description). Adjacent or consecutive sections were stained with hematoxylin-eosin for evaluation of mononuclear cell infiltration, insulitis, and general morphology.

The insulitis reaction was graded on a 0–4 scale, with 0 = normal islet, 1 = <33% of β-cell area infiltrated, 2 = 33–67% of β-cell area infiltrated, 3 = >67% of β-cell area infiltrated, and 4 = islet with no β-cells at all. All islets in two sections cut 250 μm apart were evaluated by two observers, blinded with respect to the origin of the sections.

Insulin-stained sections were used for estimating β-cell volume fraction, as previously described (16). In brief, stereological analyses were performed using an Olympus BX-50 microscope fitted with a video camera and monitor, a PC-controlled motorized stage, and Cast-Grid 2.0 software. The volume fractions of β-cells were estimated by the point-counting stereological technique using a ×20 objective, a grid of 4 × 64 points, and random systematic scanning of the tissue sections with step lengths of a maximum 600 × 450 μm controlled by the PC. The origin of the sections was blinded to the observer. The estimated β-cell volume was expressed as the percentage of total pancreatic volume.

The study was designed as six individual experiments, each comprising three groups with an average of six rats in each. All data shown below represent the combined results from the six individual experiments. The experimental setup is shown in Fig. 1.

The same mean plasma drug level was observed in diazoxide- and NN414-treated rats, and there was no change from day 7 to day 14 (Fig. 2A). On day 21, drug levels were below the detection limit. Although the administration of NN414 significantly inhibited plasma insulin levels (average of 59% inhibition 7 h after dosing), diazoxide administration did not (P < 0.001) (Fig. 2B). However, inhibition was higher (average of 80%) during the first 4 h after NN414 administration (K.S., C.F.G., D.L., J.B.H., J.S., H.M., unpublished observations). The fact that this dosage of diazoxide did not inhibit plasma insulin is likely explained by its having a lesser potency than NN414 (17).

As previously described for another KCO, NNC 55-0118 (18), we observed an initial stagnation in body weight development in NN414-treated rats (data not shown) after which the growth rate was similar to that of vehicle-treated rats.

NN414 treatment induced moderate hyperglycemia in rats (nonfasting blood glucose levels 4–17 mmol/l) (Fig. 3). Hyperglycemia was most pronounced on day 1 after the first NN414 dose (coinciding with the stagnation in body weight), with 30 of 35 rats exhibiting blood glucose levels 3 SDs above the mean nonfasting blood glucose of control rats. However, on the following day, only two of the rats were hyperglycemic. This observation suggested that the rats were able to adapt to the drug and compensate for the NN414-mediated inhibition of insulin secretion once the blood glucose level became sufficiently high. A modest, yet significant, increase in blood glucose in diazoxide-treated rats was observed on day 1, and, similar to NN414-treated rats, a modest decrease in blood glucose was seen on day 2 (P < 0.0001). Thus, diazoxide might have suppressed insulin secretion early in the study, although we did not see any difference in plasma insulin levels on day 7.

Insulin therapy was initiated after the first instance in which blood glucose was >20 mmol/l. Insulin therapy indirectly induces β-cell “rest” due to a lowering of blood glucose and helps to preserve the remaining β-cell function. By using a combination of insulin therapy and KCO treatment in severely hyperglycemic animals, we attempted to obtain optimal β-cell “rest” and at the same time mimic the clinical situation as much as possible. In the clinic, patients would normally be treated with insulin from the time of diagnosis (i.e., after the clinical manifestation of disease). We chose 20 mmol/l as the threshold for initiation of insulin therapy because NN414 therapy by itself did not induce blood glucose >20 mmol/l. There was no difference among the groups in time to initiation of insulin therapy (blood glucose >20 mmol/l) or in the actual dosages administered (Fig. 4). Overall, 87% of the rats had blood glucose >20 mmol/l at some point, and insulin treatment was initiated on day 14 ± 3.

On day 21, we measured glucose-induced C-peptide levels. Subsequently, all rats were killed for histological assessment of β-cell volume and insulitis. The β-cell volume as the percentage of total pancreatic volume was measured by stereology. We included five untreated, nondiabetic littermates as controls to obtain the normal β-cell volume and C-peptide levels. These five rats had an average β-cell volume of 0.49% (95% CI 0.29–0.69%). (In this study, rats having a β-cell volume >0.29% are labeled “responders” and rats with a lower β-cell volume are labeled “nonresponders.”)

In the NN414 group, 45.7% (16 of 35) were responders, which was significantly higher than the 11.1% in the vehicle group (4 of 36; P < 0.002 by χ² test). In the diazoxide group, 13.9% (5 of 36) were responders (Fig. 5A).

As a measure of β-cell function, an IVGTT was carried out under full anesthesia 48 h after the last gavage (in 26 rats per group). The rats were fasted for 5–6 h before the IVGTT. The baseline as well as 2- and 30-min blood glucose levels were measured. The average 30-min glucose-stimulated C-peptide value of the responders was 1,840 pmol/l (range 342–4,600), which was 58% of the average level of the five control rats. There was no difference in C-peptide levels between responders from the three groups. The majority of the nonresponders (77%) had 30-min C-peptide values below the detection limit (25 pmol/l), whereas the remainder all had levels <172 pmol/l. Thus, there was a good correlation between β-cell volume and function. The 30-min C-peptide values of individual rats are shown in Fig. 5B.

To assess whether NN414 treatment had an effect on insulitis, the pancreata were scored for the presence of insulitis on a 0–4 scale, with 0 indicating normal islets and 4 indicating end-stage destruction without any remaining β-cells. All responders consistently had an insulitis score <2.0, and conversely nonresponders (except one rat) had an insulitis score >2.0 (Fig. 5C). It is interesting that 15 of the 16 responders among NN414-treated rats had an average insulitis score <0.5, indicating that they were almost entirely free of insulitis. By contrast, only two of the responder rats in the vehicle group had an insulitis score <0.5. Thus, NN414 treatment clearly reduced insulitis in the responders. There was a good inverse correlation between the degree of insulitis and the level of residual β-cell function (r² = 0.78; P < 0.0001 for n = 31 with C-peptide levels higher than the detection limit) (Fig. 5D).

The finding that NN414 led to moderate hyperglycemia suggested that NN414 might also lead to increased HbA_1c. Indeed, we found a small but significant increase in HbA_1c on days 7 and 14. Nevertheless, there was no difference in the HbA_1c level among any of the groups on day 21 (Fig. 6A). The relative improvement in HbA_1c in NN414-treated rats compared with vehicle-treated rats during the last week could clearly be ascribed to the responders, who had a significantly lower HbA_1c on day 21 compared with nonresponders (Fig. 6B). Thus, the NN414-induced loss in glycemic control seen during the first 2 weeks of treatment was offset by a higher number of responding rats with significant β-cell function during the last week.

In this study, we showed that chronic suppression of insulin secretion obtained by three times daily administration of the KCO compound, NN414, led to preservation of β-cell volume, improvement in β-cell function, and reduction of insulitis in a type 1 diabetes rat model. We did not observe any effect of the reference compound, diazoxide, a finding most likely attributable to the fact that diazoxide activates the K_ATP channel with a much lower affinity than NN414 (10). Thus, it appears that a higher dosage of diazoxide is needed to obtain a beneficial effect on β-cell survival (18). In previous studies in a spontaneously diabetic strain of BB rats, partial protection was obtained with 150 mg · kg⁻¹ · day⁻¹ diazoxide (7).

At least two pathways leading to β-cell apoptosis exist; apoptosis can occur as a direct consequence of β-cell stress and high metabolic load resulting from hyperglycemia or as a result of insulitis and the resulting inflammation of the islets. Sturis et al. (19) have clearly shown that NN414 can inhibit glucose-induced apoptosis in cultured human islets. However, we were surprised to also discover that responsive rats were almost free of insulitis. The diabetes model used in this study, DR-BB rats depleted for ART-2⁺ cells and injected with poly(I:C), is a very aggressive diabetes model in which the autoimmune attack on β-cells normally leads to complete β-cell destruction and onset of clinical diabetes within 2–3 weeks (11). In this light, it is remarkable that treatment with NN414 was able to almost completely inhibit insulitis in responsive rats, with 15 of 16 responders having an insulitis score <0.5. Our results suggest a mechanism whereby NN414, in addition to its direct antiapoptotic effect, also inhibits infiltration of islets, probably by reducing secretion of granule content from the islets and thereby reducing immunogenicity and presentation of β-cell antigens. This hypothesis is supported by the finding that islet expression of GAD65, a major type 1 diabetes autoantigen, is increased during exposure to high glucose (20,21). Another explanation could be the absence of glucose-induced apoptosis and possibly necrosis in NN414-treated rats, which would otherwise lead to increased antigen uptake by islet resident APCs and increased presentation of β-cell antigens, thus stimulating the inflammatory process. In this way, treatment with NN414 might also reduce the autoimmune attack on the β-cells.

Almost half of the rats (46%) benefited from the NN414 treatment, whereas the remaining rats exhibited almost complete β-cell destruction, as was also observed in 86–89% of the vehicle- and diazoxide-treated rats. The reason for this heterogeneity is not clear at present; we did not observe any differences between responders and nonresponders in drug exposure, plasma insulin levels, or time to onset of blood glucose >20 mmol/l. The responders received slightly lower dosages of insulin during the last 8 days of treatment (data not shown). However, the explanation for this difference is most likely that insulin was administered according to the development in body weight and blood glucose, so that the lower dosage of insulin given to responders reflected their better glycemic control.

In the NN414-treated group, several substantial hyperglycemic episodes (blood glucose >20 mmol/l) were seen in all nonresponding rats, whereas responding rats exhibited only one or at the most two episodes of blood glucose >20 mmol/l (data not shown). We believe that the occurrence of blood glucose >20 mmol/l in the responding rats was due to the combined effect of NN414 (which leads to moderate hyperglycemia) and some decrease in β-cell function. Because insulin therapy was initiated after the first episode when blood glucose exceeded 20 mmol/l, the combined use of insulin and NN414 therapy was sufficient to maintain lower blood glucose. In contrast, nonresponding rats continued to exhibit substantial hyperglycemia despite insulin therapy, thus reflecting the more severe loss of β-cell function.

During the first 2 weeks, the NN414-treated rats exhibited a small but significant increase in HbA_1c compared with the other groups. This was not surprising considering the fact that NN414 induced moderate hyperglycemia. By contrast, during the last week, most NN414-treated rats received insulin treatment, which correlated with a slight decrease in blood glucose and a smaller increase in HbA_1c levels compared with the vehicle group. Both the preservation of β-cell function and the concomitant insulin therapy may have contributed to the smaller increase in HbA_1c in the NN414-treated group by lowering the blood glucose level. Thus we believe that the NN414-mediated increase in HbA_1c during the first 2 weeks could have been avoided by concomitant insulin therapy from the beginning.

Both in the present study and in an earlier study in DP-BB rats (7), diazoxide could not induce β-cell “rest” and protection similar to insulin. We have shown earlier that high dosages of insulin, involving hours of hypoglycemia, can reduce the incidence of spontaneous diabetes in DP-BB rats by ∼50% (22), a finding also demonstrated by Vlahos et al. (7). In the human Diabetes Prevention Trial, subcutaneous insulin treatment did not reduce the incidence of diabetes (23). However, the dosages of insulin used did not induce hypoglycemia and probably did not confer enough β-cell “rest.”

A small clinical trial has been conducted where recent-onset type 1 diabetic subjects were treated for 3 months with the nonselective KCO compound diazoxide (9). A significant improvement in C-peptide was seen in up to 18 months follow-up. However, because of the side effects of hypotension, edema formation, and excessive hair growth, diazoxide is not suitable for clinical use. In contrast to diazoxide, NN414 specifically activates β-cell−type K_ATP channels and has minimal cardiovascular effects in preclinical studies (10). Therefore, we suggest that a β-cell−selective KCO might be used to rescue remaining β-cell function in type 1 diabetic subjects, leading to long-term improvement in C-peptide and glycemic control.

This study was funded entirely by Novo Nordisk A/S.

We would like to thank the following people and departments at Novo Nordisk A/S: Louise Kristensen, Susanne Primdahl, and Pia Rothe for technical assistance; the Animal Unit (Site Gentofte) for help with the in vivo experiments; the Analysis Service for doing all the insulin and C-peptide assays; and Michael P. Andersen for analysis of plasma drug levels.