Effects of first-line diabetes therapy with biguanides, sulphonylurea and thiazolidinediones on the differentiation, proliferation and apoptosis of islet cell populations

All experiments, carried out under the UK Animals (Scientific Procedures) Act 1986 and EU Directive 2010/63EU, were approved by the University of Ulster Animal Welfare and Ethical Review Body. Animals were maintained in environmentally controlled rooms at 22 ± 2 °C with a 12 h dark and light cycle and given ad libitum access to standard rodent diet (10% fat, 30% protein and 60% carbohydrate; Trouw Nutrition, Northwich, UK) and water.

Following the removal of pancreatic tissue, samples were cut longitudinally and fixed with 4% PFA for 48 h at 4 °C. Fixed tissues were embedded and processed for antibody staining as described [30]. Tissue Sects. (7 μm) were blocked with 2% BSA and incubated with respective primary antibodies overnight at 4 °C, and, subsequently, with appropriate secondary antibodies (Table S2). To stain nuclei, a final incubation was carried out at 37 °C with 300 nM DAPI (Sigma-Aldrich, D9542). To assess cell proliferation and/or apoptosis, co-staining of mouse anti-insulin (Abcam, Cambridge, UK; 1:1000; ab6995) or guinea pig anti-glucagon (PCA2/4, 1:200; raised in-house) with rabbit anti-Ki-67 (1:200; Abcam ab15580) or TUNEL reaction mixture (Roche Diagnostics Ltd, UK) was used. YFP, indicating the α-cell lineage, was detected by with a rabbit anti-GFP antibody (1:1000; Abcam, ab6556) (Table S2), which is reactive against all variants of Aequorea Victoria GFP, including YFP. The slides were imaged on an Olympus BX51 microscope, equipped with a 40x/1.3 objective. We aimed to include all the islets visible on the slide in the morphometry analysis, independently of their localisation in relation to other pancreatic structures, with at least 50 cells analysed within each islet cross-section in the per-cell studies (Figs. 2B, 3). The multichannel fluorescence was recorded using DAPI (excitation 350 nm/emission 440 nm), FITC (488/515) and TRITC (594/610) filters and a DP70 camera controlled by Cell^F software (Olympus, UK). Images were analysed using ImageJ software. All counts were determined in a blinded manner with 60–150 islets analysed per treatment group, as indicated in the figure legends. The non-stained cells visible in the middle of the islet were not excluded from the computation of the islet area.

Statistical analysis was performed using PRISM 5.0 (GraphPad, USA) or R. Values are expressed as mean ± SEM. Comparative analyses between experimental groups were carried out using independent-samples Student's t test or (for > 2 samples) a one-way ANOVA with Bonferroni’s post hoc. The difference between groups was considered significant for p < 0.05.

The treatment with STZ resulted in a progressive diabetic phenotype in the mice, which was reflected by the elevation of blood glucose concentration (Fig. 1B). Non-fasting blood glucose increased in the STZ-treated mice from 8.2 ± 0.4 mM (end of the STZ treatment) to 32.6 ± 0.4 mM 14 days afterwards (7.6 ± 0.7 and 8.4 ± 0.6 mM, respectively, in the control group).

As designed, 10-day administration of rosiglitazone, tolbutamide or metformin had no significant impact on glycaemia (30.8 ± 0.7, 31.3 ± 0.4, 30.0 ± 0.1 mM, respectively) (Fig. 1B). The onset of hyperglycaemia coincided with the 9% decrease in the body weight from 19.2 ± 0.4 g at the end of the STZ treatment (20.7 ± 0.3 g in the control group, n.s.) to 17.4 ± 0.4 g after 14 days (20.5 ± 0.3 g in the control group, p < 0.05) (Fig. 1C). 10-day administration of rosiglitazone and tolbutamide had no statistically significant impact on body weight (17.9 ± 0.6, 17.4 ± 0.3 g, respectively, p < 0.05 vs control), whereas metformin tended to exacerbate (16.7 ± 0.6 g, p < 0.05 vs control) the weight loss (Fig. 1C).

The effects of STZ treatment on the intake of food or fluid by the experimental animals were palpable 4 days post its cessation (day 0, Fig. 1D,E) and were progressively elevating from that point. Both food and fluid intake were significantly attenuated after 4 days of treatment with metformin (Fig. 1D,E), coincident with the decrease in the body weight (Fig. 1C). Neither of the remaining two OHA influenced fluid intake (Fig. 1E); however, rosiglitazone and, at one point, tolbutamide significantly attenuated the intake of food (Fig. 1D).

As a result of the STZ treatment, the non-fasting terminal plasma insulin levels that were measured on day 10 were substantially decreased (0.16 ± 0.06 vs 0.95 ± 0.04 ng/mL in STZ-treated and control groups, P < 0.01), whereas the differences between corresponding glucagon levels did not attain statistical significance (0.19 ± 0.07 vs 0.32 ± 0.11 ng/mL) (Fig. 1F). Whilst none of the OHA elevated insulin levels (Fig. 1F), metformin induced a significant decrease of plasma glucagon levels, on the STZ-treatment background (0.15 ± 0.03 vs 0.32 ± 0.11 ng/mL in the control group, p < 0.05) (Fig. 1F).

In line with the effect on plasma hormone levels (Fig. 1F), STZ substantially decreased pancreatic content of insulin (27.5 ± 9.9 vs 109.2 ± 8.0 ng/mg of tissue in control, p < 0.05), without any appreciable effect on the glucagon content (Fig. 1G). Following subsequent rosiglitazone treatment, the glucagon content was substantially decreased (13.3 ± 3.8 vs 22.7 ± 4.2 ng/mg of tissue in control, p < 0.05), whereas tolbutamide or metformin had no effect on this parameter (Fig. 1G).

We did not detect any significant alteration in the islet number, in response to any treatments (red in Fig. 2A). At the same time, the observed decrease in plasma and pancreatic insulin (Fig. 1F,G) coincided with the decrease in the average cross-section area of islets in the STZ-treated mice (black in T Fig. 2A). The OHA therapy that followed the STZ treatment resulted in a mild increase in this metric (black in Fig. 2A).

The STZ treatment produced a significant reduction in the relative β-cell area (red/insulin + in Fig. 2B,C) and, respectively, an increase in the relative α-cell area (black/glucagon + in Fig. 2B,C). Remarkably, a 10-day oral administration of metformin, but not rosiglitazone or tolbutamide, counter-acted the effects of the STZ treatment, resulting in small but significant differences in the percentage of β-cells (55 ± 2% vs 48 ± 2% in STZ mice, p < 0.05) and α-cells (44 ± 2% vs 51 ± 2% in STZ mice, p < 0.05) (Fig. 2B,C). Interestingly, the islets from the STZ-treated animals contained a palpable fraction of cells that did not express insulin or glucagon (Fig. 2C): we need to stress that, among other types, islets contain significant numbers of vascular endothelial cells [35, 36], which are likely to contribute to this phenomenon.

In line with the report of STZ inducing β-cell apoptosis, when used in small repeated doses [37], we observed a sixfold (2.2 ± 0.1 vs 0.4 ± 0.1% in control mice, p < 0.05) increase in the percentage of β-cells expressing an apoptosis marker, TUNEL, in STZ-treated mice (red, Fig. 3A, Figure S1A). The metformin therapy tended to attenuate the β-cell apoptosis, whereas tolbutamide further increased the expression of TUNEL by β-cells (3.2 ± 0.3% vs 2.2 ± 0.1 in the STZ-treated mice, p < 0.05) (red, Fig. 3A). Although the STZ treatment per se has not affected the apoptosis of α-cells (black, Fig. 3A), metformin administered to the STZ-treated animals mildly increased this characteristic (0.5 ± 0.1% vs 0.4 ± 0.1% in the STZ-treated group, p < 0.05) (black, Fig. 3A, Figure S1A).

The pro-apoptotic effect of the STZ treatment did not affect the percentage of proliferating β-cells, as was assayed via Ki-67 staining (red, Fig. 3B, Figure S1B). This metric was increased by subsequent treatment with rosiglitazone or metformin (4.0 ± 0.2% and 2.5 ± 0.4% respectively, vs 1.3 ± 0.1% in the STZ-treated group, p < 0.05). The STZ treatment produced a fivefold increase in the fraction of proliferating α-cells (black in Fig. 3B, Figure S1B), which was not affected by any of the OHA (black in Fig. 3B).

The key feature of the animal model used in this study, the Glu^CreERT2; ROSA26-eYFP mice, is the tissue-specific targeting (pancreatic α-cells) and the inducible nature of the expression of YFP. When co-detected with anti-glucagon antibodies, 20 days post-induction of the targeted YFP expression, the islets from these mice had only a small fraction of YFP + cells that did not express glucagon (0.3 ± 0.1%) (black, Fig. 3C, Figure S1C). The YFP + cell percentage was increased almost threefold after the STZ treatment (0.8 ± 0.4%, p < 0.05 vs control) and further potentiated by tolbutamide (2.7 ± 0.8%, p < 0.005 vs control, p < 0.01 vs STZ) but not rosiglitazone or metformin (0.9 ± 0.4%, p < 0.05 vs control and 1.2 ± 0.4%, p < 0.01 vs control, respectively) (black, Fig. 3C, Figure S1C). Of note, we were unable to detect YFP in almost half of glucagon⁺ cells, which we believe to reflect a technical feature of the anti-GFP antibody staining (red in Fig. 3C).

The percentage of the YFP⁺insulin⁺ cells was low in the experimental animals with α-cell-specific targeting of YFP (0.6 ± 0.1% in the control group). The treatment with STZ however triplicated this number (1.7 ± 0.1%) (red, Fig. 3D, Figure S1D). Neither of the OHA was able to further enhance the commitment of the YFP + cells towards the insulin lineage (red, Fig. 3D). At the same time, the administration of each of the OHA, following the STZ treatment, increased the percentage of bi-hormonal (insulin + glucagon +) cells (black, Fig. 3D, Figure S1D). The size of this small cell subpopulation was unaffected by the STZ treatment (0.27 ± 0.01% vs 0.25 ± 0.02% in the control group), whereas subsequent rosiglitazone (0.33 ± 0.02%), tolbutamide (0.31 ± 0.02%) and metformin (0.38 ± 0.03%) administration substantially expanded it (black, Fig. 3D).

The diabetes model and the OHA dosage were designed to resolve the direct effects of the OHA on pancreatic islet cell plasticity [39]. We have opted for repeated injections of small doses of STZ [9, 40] over high-fat diet or leptin receptor deficiency animal models of diabetes to enable cell lineage tracing and rule out any indirect effects, mediated by changes in insulin sensitivity or blood glucose, that may impact the islet plasticity. The model animals displayed stably elevated glycaemia and reduced body weight (Fig. 1B,C), whereas the three treatments, at the doses chosen, affected only food and fluid intake (Fig. 1D,E). Notably, the doses used compare well with daily human recommended doses, given the differences in the pharmacokinetics of the three drugs in the mouse and human systems [41‐43].

A side effect of the OHA therapy, lowering of the systemic glucagon in response to metformin (Fig. 1F), is unlikely to reflect the depletion of the α-cell population due to its transdifferentiation or apoptosis, as α-cells are well in excess, in rodent islets [35, 44]. The phenomenon could have stemmed from the elevation of circulating GLP-1 levels, reported to be induced by metformin [45]. Another possible explanation for this effect is the activation of the intra-islet GLP-1 secretion system [46‐48], under the conditions of the STZ treatment [9, 49]. The likely mechanism for that is the acquisition of the proconvertase PC1/3 activity by α-cells [46], with a subsequent shift in the α-cell secretory output from glucagon to GLP-1. In line with the reported cytostatic effect of metformin [50] that, in our hands, stimulated apoptosis in α-cells (but, surprisingly, given earlier reports [51], not in β-cells, Fig. 3A), the elevation of intra-islet and circulating GLP-1 could explain partial recovery of the ratio of β- and α-cells (Fig. 2B), presumably by upregulating β-cell proliferation [9, 30]. Notably, rosiglitazone, reported to increase the β-cell mass by reversing the apoptosis [52], was not efficient in doing so in our model (Fig. 3A). In our hands, it induced a fourfold increase of the proliferating β-cells, in line with the previous reports [52].

The STZ-induced β-cell injury per se resulted in a detectable expression of insulin and a loss of expression of glucagon by YFP⁺ cells (Fig. 3C), reflecting the α-cell population before the STZ treatment. The fact that none of the OHA affected the co-expression of insulin and YFP on a per-cell basis suggests the lack of a role in regulating α-/β-cell transdifferentiation (Fig. 3D).

In the present study, a small but detectable number of bi-hormonal [8] cells was evident after the STZ treatment, followed by administration of metformin (Fig. 3D). This effect can be explained by non-pancreatic signals [53] that may induce α-cell transdifferentiation. On the contrary, the de-differentiation of α-cells by tolbutamide (Fig. 3C), likely to result from a direct effect on the K_ATP channels, was not associated with any β-cell phenotype.

Since sulphonylureas are actively prescribed for diabetes, further elucidation of their global effects on islet function is highly relevant. No previous study has reported, to our knowledge, on the effects of this drug class on islet cell transdifferentiation. Our data with tolbutamide are important in revealing that not only does the sulphonylurea lack beneficial effects on islet plasticity (unlike the two other classes of OHA) but that it exerts adverse effects on β-cell health and apoptosis. This can be viewed as a significant limitation of first-line sulphonylurea therapy and would suggest that incretin mimetics which have recently shown to have positive effects on β-cell transdifferentiation, apoptosis and proliferation [9, 29] would be a better therapeutic option for direct β-cell actions.