β-Cell Neogenesis During Prolonged Hyperglycemia in Rats

In the present study, tissue samples from our previous study were used (15). In brief, adult male Sprague Dawley rats were chronically catheterized (jugular vein) and infused with saline or glucose. After a 3- to 5-day recovery period, rats were infused with either 0.45% saline (n = 45) or 50% glucose in 0.45% saline (n = 60) at 2 ml/h for 0, 1, 2, 3, 4, 5, or 6 days. Then 6 h before being killed, rats were injected intraperitoneally with 100 mg/kg 5-bromo-2′deoxyuridine (BrdU; Amershan Canada, Oakville, ON, Canada). At termination, rats were anesthetized with sodium-pentobarbitol (35 mg/kg body wt, i.p.) and a cardiac puncture was performed. The pancreas was removed and the rat was killed by a diaphragm incision. The pancreas was blotted and cut into three pieces (duodenal, middle, and splenic), weighed, fixed in Bouin’s solution overnight, washed in cold water, and stored in 10% formalin until embedding in paraffin wax. All procedures involving animals were performed in accordance with the Guidelines of the Canadian Council on Animal Care.

Sections 4- to 5-μm thick were taken from each of the three pieces of pancreas and de-waxed in xylene. Endogenous peroxidase activity was blocked with a 0.3% solution of hydrogen peroxide. The sections were then washed with PBS and incubated with 10% lamb serum in PBS for 30 min at room temperature. Slides were stained with one or more primary antibodies, including insulin (guinea pig anti-porcine, 1:2000; Dako Diagnostics Canada, Mississauga, Canada); a cocktail of antibodies to glucagon, somatostatin, and pancreatic polypeptide (rabbit anti-human, 1:2000; Dako); BrdU (mouse monoclonal anti-BrdU, 1:100, Amersham Canada); and CD3 (rabbit anti-human, 1:1500, Dako). Slides stained for CD3 were incubated with citrate buffer for 30 min at 90°C for antigen retrieval before primary antibody application. Slides were incubated with primary antibodies overnight at 4°C before being washed in PBS and incubated with the secondary antibodies for 1 h at room temperature (1:500; biotinylated goat anti-guinea pig IgG, biotinylated goat anti-mouse IgG, or biotinylated goat anti-rabbit IgG; Dako). Sections were then washed in PBS, incubated with avidin-biotin complexed with horseradish peroxidase (1:1000; Vectastain Elite ABC Kit, Vector Laboratories Canada, Burlington, Canada) for 1 h, and then developed using 3,3′-diaminobenzidine tetrahydrochoride (Sigma-Aldrich Canada, Oakville, Canada). All slides were counterstained with Harris’ hematoxylin (Sigma-Aldrich).

Focal areas were defined as areas of tissue consisting of small duct-like structures with an apparent infiltration of leukocytes (Fig. 1A). These areas were clearly delineated, with defined borders usually juxtaposed to normal acinar tissue. One slide from each of the three parts of each pancreas (stained with a cocktail of antibodies to glucagon, somatostatin, and pancreatic polypeptide) were relabeled and randomized so as to blind the viewer to the treatment group. Each slide was examined using a light microscope (Olympus Model BX40; Carsen Group) at ×200 for the presence or absence of focal area tissue. To determine the cross-sectional area of focal tissue, sections were systematically sampled using an image analysis system (Northern Eclipse; Empix Imaging, Mississauga, Canada). Slides were examined in 1.0 × 1.5-mm increments, resulting in the assessment of 70 ± 5 fields/slide and 82 ± 5% of the total pancreatic section. Focal tissue area was quantified by capturing each field and hand tracing focal areas within the captured images. The thresholding option was used to quantify the cross-sectional areas occupied by tissue and to subtract unstained areas (white space) not occupied by any tissue. Focal area sizes are reported as the cross-sectional area of focal tissue as a percent of total pancreatic tissue area.

To determine the replication rate of duct cells from small, medium, large, common, and focal area ducts, sections stained with cocktail, anti-BrdU antibodies were scanned, and all duct cells were counted. BrdU incorporation was assessed for rats infused with either glucose (n = 26) or saline (n = 24) for 0–4 days. One slide from each of the three portions of the pancreas was assessed, and an average of 17,541 ± 522 duct cells/animal were counted. Duct size category was assigned based on a qualitative assessment of the number of nuclei in the duct circumference, the amount of encapsulating connective tissue, the sectioning angle of the duct, and the general morphological appearance of the duct, similar to the procedure used by Githens et al. (16). Duct cell replication rates for each duct size were determined as the percentage of nuclei stained positive for BrdU for each animal.

β-cell cluster size distribution was determined from three slides (one duodenal, one middle, and one splenic) for each rat. Slides stained with anti-insulin antibody were randomized, relabeled, and analyzed using a light microscope (Olympus; final magnification ×350) and image analysis software. All insulin-positive β -cell clusters of one or more cells were loosely traced and the size of the cluster within the trace was determined by using the thresholding option. We have used the term “β-cell cluster” rather than “islet” because clusters were identified on the basis of insulin antibody staining and did not include identification of other endocrine and nonendocrine cell types found in islets. The association of each insulin-positive cluster with acinar, duct, or focal area tissue was made according to the type of tissue bordering at least 50% of the cluster. The incidence of β-cell clusters is reported as the average across animals in a treatment group of the sum from three slides examined per animal of β-cell clusters detected for each tissue association (acinar, duct, or focal area) and size (single, small, medium, or large) category.

To establish category size cutoffs, the size of a single β-cell was determined by measuring the cross-sectional area of insulin-positive clusters with only one nuclei (Fig. 1B). Single β-cell size was determined for both acinar and duct-associated (n = 5 each) single β-cells from saline- and glucose-infused rats (n = 13; one from each day and type of infusion). Acinar-associated single β-cells were significantly larger than duct-associated single β-cells (152 ± 8 vs. 114 ± 7 μm²; P < 0.001). All β-cell clusters with areas less than or equal to the average single β-cell size (≤152 μm² for acinar, ≤114 μm² for duct) were categorized as single β-cells (Fig. 1B). All β-cell clusters with areas greater than the average single β-cell size and less than or equal to three times the average single β-cell size were considered small extra-islet β-cell clusters (Fig. 1C). The remaining insulin-positive β-cell clusters were arbitrarily divided into medium and large clusters of 15,000 μm².

The number of T-cells/mm² was determined for acinar and focal tissue in rats infused for 0, 1, 2, 3, or 4 days from slides stained with anti-CD3 antibodies. The number of CD3⁺ T-cells associated with acinar tissue were counted using a light microscope (Olympus; ×400). A T-cell was considered to be acinar associated when it was bound on all sides by acinar tissue. Intravascular T-cells were not counted. The total number of acinar-associated T-cells per animal was divided by the total area of acinar tissue determined from previous β-cell mass measurements to obtain a T-cell infiltration index for each animal. The density of T-cells in focal areas was determined using an image analysis system to measure focal tissue area and count the number of CD3⁺ T-cells in the focal areas. The T-cell density is reported as the total number of focal area T-cells divided by the total focal tissue area for each rat.

Statistical analyses were performed using SAS System for Windows (Release 6.12; SAS Institute, Cary, NC) and SYSTAT for Windows (Version 6.0; SYSTAT, Evanston, IL). One, two, or three-way ANOVA or the Wilcoxon nonparametric analysis was performed where appropriate using the following factors: infusion type (glucose and saline), day of infusion (0–6), duct size (small, medium, large, common, and focal), or β-cell cluster size (single, small, medium, and large). Five animals were used for each comparison unless otherwise reported. Because there was only one group of animals for day 0, the data from these animals were considered as day 0 for both saline and glucose time-dependent comparisons. Post hoc individual comparisons were conducted when deemed appropriate by an omnibus ANOVA measurement. Differences were considered to be significant at P <0.05. All results are expressed as means ± SE.

Pancreatic focal areas (Fig. 1A) were present in rats infused with glucose for 2, 3, or 4 days but not in any saline-infused rats (Table 1). Focal areas varied in size from 0.1 to 14.9% of the total pancreas area (Fig. 2). Focal areas were most apparent after 3 days of glucose infusion, averaging 2.9 ± 0.8% of the pancreatic tissue.

The replication rate of duct cells from small, medium, large, or common ducts was not elevated before or during the appearance of focal areas (Fig. 3). By three-way ANOVA, there was a significant effect of duct size (P < 0.0001) and length of infusion (P = 0.048), but not of treatment (P = 0.075), on duct cell replication rate. The replication rate was 7- to 20-fold greater in common duct cells than in small, medium, and large duct cells, whereas the replication rate of focal area duct cells was 2-fold greater than in the common duct (Fig. 4). Although treatment was not a significant effect by three-way ANOVA, the replication rate of small (P = 0.0016), medium (P < 0.0001), and large (P = 0.009) duct cells was significantly greater in saline- than in glucose-infused animals by one-way ANOVA (Fig. 4). The effect of length of infusion observed by three-way ANOVA was attributable to suppression of BrdU incorporation by glucose over time in medium (P = 0.0068) and large (P = 0.0272) ducts (Fig. 3), and was not attributable to any time-dependent effects in small or common ducts or in saline-infused animals.

Total β-cell mass determined by insulin antibody staining (Fig. 5) increased progressively in glucose-infused animals to an approximately twofold increase over saline-infused animals by 6 days of infusion (similar to results obtained using the cocktail of antibodies to glucagon, somatostatin, and pancreatic polypeptide [15]).

Duct-associated β-cell mass constituted ∼1% of the total β-cell mass (Fig. 1D). There was no difference, however, in the mass of β-cells associated with ducts during glucose and saline infusion (P = 0.78) (Table 2). The mass of β-cells found in focal areas (0.049 ± 0.021 mg) was similar to the mass of β-cells associated with ducts (0.055 ± 0.005 mg). The remaining 98–99% of the β-cell mass was associated with acinar tissue. The mass of acinar-associated single β-cells increased 63% in glucose- versus saline-infused animals after 4 days of infusion (0.052 ± 0.007 vs. 0.032 ± 0.003 mg, respectively) to account for <1% of the total β-cell mass (Table 3).

As with total duct-associated β-cell mass, there was no effect for type of infusion (glucose or saline, respectively) on the number of single (27.0 ± 2.7 vs. 26.4 ± 1.8), small (16.9 ± 1.6 vs. 17.8 ± 1.1), or medium (6.7 ± 0.8 vs. 6.5 ± 0.7) β-cell clusters associated with ducts (P = 0.97). Likewise, there was no effect of infusion type on the average size of single, small, or medium duct-associated β-cell clusters (P > 0.35) (Table 2).

There were focal area−associated insulin-positive clusters found in all animals with focal areas investigated for β-cell mass (day 2: n = 1, day 3: n = 4). The total number of β-cell clusters associated with focal areas ranged was 2–35 per animal. In all, 58% of insulin-positive clusters in focal areas were <152 μm² in cross-sectional area. There was also an increase in the incidence of acinar-associated single β-cells during glucose infusion as compared to saline infusion. The number of acinar-associated single β-cells was increased by 70% at 3 and 4 days of glucose infusion (Fig. 6A). The increased number was associated with a decrease in the average single β-cell size (Fig. 6B). As expected, when viewing total β-cell mass dynamics (Fig. 5), there was a significant effect of infusion type on the number of medium (P = 0.043) and large acinar-associated β-cell clusters (P < 0.0001). For large β-cell clusters, there was a significant increase in both number and average cluster size with days of glucose infusion, so that by 6 days of glucose infusion, the frequency had increased twofold and the average cluster size had increased by 30% in glucose- compared to saline-infused animals.

Acinar T-cell infiltration was elevated by ∼10-fold in glucose- as compared to saline-infused rats (P < 0.001, Fig. 7). There was an effect of day of infusion on T-cell infiltration in glucose-infused rats (P = 0.0442), peaking at 0.74 ± 0.27 T-cells/ mm² after 2 days of infusion. Along with a diffuse T-cell infiltration into the acinar tissue of glucose-infused rats, there were distinct areas with a high density of T-cells. Focal areas in glucose-infused rats had a very high density of T-cells, ∼350 times greater than in acinar tissue (151 ± 30 T-cells/ mm²) (Fig. 1).

β-cell neogenesis is known to be important during development and to contribute to β-cell regeneration after diabetogenic insults, such as 90% partial pancreatectomy and after administration of a high dosage of streptozotocin (5,11). The present study provided the first quantitative morphometric evidence that β-cell neogenesis also occurs during prolonged hyperglycemia. These data support our previous model-based conclusion that β-cell neogenesis contributes to β-cell mass expansion during chronic glucose infusion (15). The quantitative histological analyses performed in the present study suggest that neogenic β -cells may arise through differentiation of acinar tissue into focal areas and insulin-positive β-cells.

Areas of pancreatic tissue with many small duct-like structures have been called focal areas (5). These focal areas are thought to give rise to new endocrine and exocrine tissue, deduced in part from their transient appearance after diabetogenic insults (e.g., partial pancreatectomy) and before or in association with increases in β -cell mass (5,7,8). Focal areas appeared transiently in our glucose-infused rats, but were not apparent in the saline-infused animals (Figs. 1 and Table 1). Focal areas have not been previously described in glucose-infused rats, possibly because their peak appearance occurred after 3 days of glucose infusion (Table 1 and Fig. 2). In previous reports, pancreas morphology was examined after 1, 2, or 4 days of chronic glucose infusion (13,17). Although small focal areas (constituting <1% of the pancreas area) were observed in 5 of 23 animals studied after 2 or 4 days of glucose infusion, the majority of rats (56%) infused with glucose for 3 days had focal areas that constituted up to 15% of the total pancreas area (Table 1, Fig. 2).

After partial pancreatectomy, the cells of the duct-like structures in focal areas replicate at a high rate (5). Likewise, the duct-like structures thought to give rise to new β-cells after duct ligation or overexpression of IFN-γ in the β-cell are highly proliferative (7,8). Similarly, the cells of the duct-like structures in the focal areas found in our glucose-infused animals replicated at a rate that was 2-fold greater than for cells in the common duct and 15- to 40-fold greater than in cells lining the small, medium, and large ducts (Fig. 4). The high proliferative rate of these duct-like structures along with the disappearance of the focal areas suggests that the majority of the new cells differentiate into endocrine or exocrine cell types. Thus the focal areas observed in our glucose-infused rats likely contribute to an increase in β-cell neogenesis.

Single β-cells surrounded by acinar, focal, or even ductal tissue are thought to arise primarily through differentiation from the surrounding cell type (7,18), although some cells may arrive at a given location after migration from their tissue origin (19). It is also possible that single cells are derived from another cell type not visible in the plane of the section, but the appearance of single cells and small β-cell clusters has been used as an index of new β-cell formation (7,18,20). In the present study we found a high incidence of single β-cells in focal areas and in acinar tissue after 3 and 4 days of glucose infusion (Fig. 6). In support of the idea that these single β-cells represent newly formed β-cells is the finding that the average acinar-associated single β-cell size decreased concomitantly with the increased number of cells. Scaglia et al. (21) have suggested that a decrease in cell size might be indicative of newly differentiated β-cells. Whether these single β-cells arise from differentiation of duct-like cells in focal areas or from transdifferentiation of exocrine cells could not be determined from the present study. These data clearly indicate that there is an increase in the number of newly formed β-cells from these two sources after 3 days of glucose infusion leading to an increased potential for β-cell mass expansion.

In contrast to the high frequency of single β-cells in focal areas and the increased frequency of single β-cells in acinar tissue after 3 days of glucose infusion, there was no increase in the frequency of duct-associated single β-cells in glucose-infused rats. Single β-cells in ducts likely arise from replication and then differentiation of a ductal progenitor. Presumably, these cells then either replicate to form small duct-associated β-cell clusters or migrate into the surrounding tissue (19,22). Cells migrating into exocrine tissue probably develop into small β-cell clusters and new islets. Alternatively, these cells may migrate to become part of established islets (23). Thus it is possible that the rate of new β-cell formation from ductal progenitors could increase without an increase in the number of duct-associated β-cells (i.e., if migration into the surrounding tissue increases at the same rate as replication and differentiation of the duct cells). However, the suppression of duct cell replication rates by glucose (Fig. 4) and the lack of an effect of glucose on the frequency, average size, or mass of duct-associated β-cell clusters (Table 2) suggests that direct duct-to-β -cell transdifferentiation does not contribute to the increase in β-cell neogenesis in our glucose-infused rats. Although local areas of increased duct proliferation may have been missed by our sampling method, on average, duct proliferation was reduced in glucose infused rats, suggesting that this is not a quantitatively important pathway for β-cell neogenesis during glucose infusion.

These results are in direct contrast to the recent report by Bernard et al. (13). After 24 h of clamping plasma glucose at ∼22 mmol/l, they observed a high rate of BrdU incorporation into duct cells and a large number of endocrine cells within or budding from ducts in glucose- as compared to saline-infused rats (13). Whether the differences between the study of Bernard et al. (13) and the present data are attributable to differences in rat strain, the plasma glucose level achieved after 24 h (∼35 mmol/l in the present study), those investigators’ qualitative judgement versus our quantitative assessment, or some other explanation, requires clarification. That the differences in neogenic indexes are a result of differences in experimental conditions (rat strain or glucose level), however, is suggested by the finding of Bernard et al. of a 50% increase in total β-cell mass after 24 h and our finding of no increase in β-cell mass after the same length of infusion (13).

Our quantitative analysis of duct cell replication rates led us to reject our original hypothesis that the focal areas arise from ductal progenitor cells. Bonner-Weir et al. (5) demonstrated sequential waves of duct cell proliferation beginning in the common duct and progressing from large through small ducts before focal area formation in partially pancreatectomized rats. In the present study, there was no increase in duct cell proliferation before or during the time of focal area formation in glucose-infused rats (Fig. 3). In fact, glucose infusion led to a significant decrease in the proliferation rate of cells in small, medium, and large ducts (Fig. 4). Because increased proliferation of ductal progenitors likely precedes the formation of focal areas from ducts, these data suggest that the focal areas in our glucose-infused rats were not of ductal origin.

If the focal areas do not arise from the proliferation of ducts, could they arise through dedifferentiation of acinar tissue? Wang et al. (7) have suggested that acinar cell dedifferentiation is the source of focal areas in the duct ligation model of β-cell mass expansion. In this model, immune cell infiltration of the acinar tissue and upregulation of growth factors and cytokines is thought to stimulate the differentiation of acinar tissue into focal area tissue (10). In the present study, acinar T-cell infiltration was elevated in glucose-infused rats before and during focal area formation. In addition, the density of acinar T-cell infiltration was highly variable after 2 days of glucose infusion, suggesting that immune cell infiltration immediately preceded focal area formation. The T-cell infiltration index of focal areas in the glucose-infused rats was similar to that found in human pancreatitis (24). Pancreatitis is thought to result from intraductal precipitation of proteins leading to duct obstruction (25). Hyperosmotic glycemic infusions lead to decreased fluid content in the pancreas (26) and as a result may lead to duct protein precipitation and a form of duct obstruction in glucose-infused rats. Taken together, these data are also consistent with the idea that the focal areas observed in our glucose-infused rats arise through dedifferentiation of acinar tissue, rather than from proliferation of ductal precursors. To determine whether this dedifferentiation and probable redifferentiation into both endocrine and exocrine tissue is attributable to ductal obstruction caused by the infusion of the hyperosmolar glucose solution requires additional control studies with a hyperosmolar solution at normal glucose levels.

Through a detailed time-dependent morphometric analysis of pancreases from rats infused with glucose for 0, 1, 2, 3, 4, 5, and 6 days, we have been able to demonstrate that β-cell neogenesis contributes to a glucose-induced expansion of the β-cell mass. The decrease in duct cell replication and the absence of an increase in duct-associated β-cell frequency or mass all suggest that the increase in β-cell neogenesis is not from duct-associated stem cells. The increase in acinar-associated single β-cells, the time-dependent leukocyte infiltration of the acinar tissue before focal area formation, and the high frequency of focal area− associated single β-cells all suggest that acinar tissue, through both direct (acinar to β-cell) and indirect (acinar to focal area to β-cell) pathways may be responsible for the increased β -cell neogenesis during prolonged hyperglycemia.

These studies were supported by a grant from the Medical Research Council (MT10574) of Canada.

We are grateful for the technical assistance of Warren E. Fieldus in producing the image analysis algorithms and Mauro Castellarin for his assistance with morphometric assessments. Many people contributed to the experiments that provided the samples analyzed in the present report, including M. Dawn McArthur, Dan Tzur, Abha Dunichand-Hoedl, Narinder Dhatt, Jamie T. Lewis, Gerald LaChance, and Dr. Jan Przysiezniak.