Role of the forkhead protein FoxO1 in β cell compensation to insulin resistance (original) (raw)
To study the mechanism of compensatory β cell mass expansion in insulin resistance, we generated diabetic mice with mutations affecting insulin signaling and β cell replication. In the first experimental cross, we intercrossed Insr transgenic knockout mice, in which Insr is expressed only in liver, brain, and β cells (henceforth, L1) (26), with transgenic mice expressing a constitutively nuclear FoxO1 in liver (henceforth, 307) (25). The mutation replaces a key phosphorylation site (Ser253Ala), thereby preventing phosphorylation-dependent nuclear exclusion (25). The expectation was that the FoxO1 transgene would impair insulin action in liver and cause a compensatory increase in β cell mass (Table 1).
In the second experimental cross, we intercrossed L1 mice with transgenics expressing constitutively nuclear FoxO1 in β cells, pancreatic ducts, and liver (henceforth, 305). The expectation here was that, in addition to impairing liver insulin action, expression of the mutant FoxO1 in β cells would curtail expansion of β cell mass by preventing β cell replication (Table 1). The mutant transgene is not expressed in the exocrine compartment (25).
Figure 1 summarizes expression patterns of Insr and FoxO1 transgenes in the 3 mouse strains. We detected Insr in liver and brain, but not in muscle extracts from L1 mice. As negative controls, we show liver extracts from Insr–/– mice and brain or muscle extracts from L2 transgenic knockout mice, which express Insr only in liver and β cells (26) (Figure 1A). The mutant FoxO1 transgene was tagged with a c-Myc epitope to distinguish it from the endogenous protein (25). We detected expression of the c-Myc–FoxO1 fusion protein by immunoprecipitation with anti–c-Myc antiserum followed by immunoblotting with anti-FoxO1 antiserum in liver of lines 307 and 305, but not in WT mice (Figure 1B). We also detected expression of c-Myc–FoxO1 fusion protein in β cells of 305 mice, but not 307 or WT mice, by immunohistochemistry of pancreatic sections (25). It should be noted that the transthyretin-driven transgene is expressed in approximately 80% of all β cells (Figure 1C).
Detection of c-Myc–FoxO1 fusion proteins. (A) Insr Western blot. We detected Insr by Western blotting of liver, brain, and muscle extracts of WT mice. L1 mice express Insr in liver and brain, but not muscle. Insr–/– mice do not express Insr in liver while L2 mice are Insr transgenic knockouts with Insr expression limited to liver and β cells (26). (B) FoxO1 Western blot. We performed immunoprecipitations with anti–c-Myc antiserum and Western blotting with anti-FoxO1 antiserum (upper panel). As a control, we show immunoblotting with anti-FoxO1 (middle panel) and antitubulin antisera (lower panel). (C) Pancreatic immunohistochemistry. We performed immunostaining of pancreatic sections from WT, 307, and 305 mice with anti–c-Myc antiserum to detect transgene-encoded FoxO1 (left panels, red). DNA counterstaining with DAPI is shown in right panels. We show representative sections to illustrate the difference between transgene-positive and transgene-negative cells, indicated by the yellow arrows. Magnification, ×40 (top 3 rows); ×100 (bottom row).
Early-onset diabetes in double-mutant mice. Transgenic mice carrying the various mutations were born with the expected Mendelian frequency. At 2 weeks of age, all mutants showed an approximately 20% decrease in body weight compared with WT mice (Table 2). Metabolic analyses indicated that mean glucose levels of single and double transgenics did not differ from WT mice. However, approximately 14% of L1/307 and approximately 27% of L1/305 double-transgenic mice had glucose levels in the diabetic range (>200 mg/dl) (Figure 2A). L1 mice showed a 7-fold increase in insulin levels compared with WT mice (Figure 2B) (26). The 307 and 305 mice had insulin levels similar to WT mice. In contrast, both double mutants showed hyperinsulinemia. L1/307 had the same levels as L1 mice whereas L1/305 mice had approximately 55% lower levels than L1 mice. It should be noted that mean glucose levels and percentage of diabetic mice were higher in L1/305 than in L1/307 mice. Therefore, L1/305 mice have a more severe metabolic derangement than L1/307 mice. We also measured β cell mass using immunohistochemistry on pancreatic sections. We failed to detect significant differences among the various genotypes (Figure 2C). However, since the mutant mice are growth retarded, one would predict that, in order to achieve the same β cell mass as WT mice, β cell replication rates should be increased. Indeed, when we measured proliferation as the percentage of Ki67-positive β cells, we detected an approximately 2-fold increase in all mutant mice (Figure 2D).
Metabolic data in 2-week-old mice. We measured glucose (A) and insulin levels (B) in random-fed mice. (C) β cell mass. We determined the percentage of the area of pancreas occupied by β cells on pancreatic sections stained with anti-insulin antiserum. We scored at least 3 animals per genotype and 4 sections per animal spaced 80 μm apart. We then determined β cell mass by multiplying the area for pancreatic weight. (D) We measured the percentage of Ki67-positive cells (Ki67-labeling index) in 3 sections for each of 3 animals per genotype. *P < 0.05 versus WT.
Body weights of 2-week-old mice
At 8 weeks of age, both L1/307 and L1/305 mice had developed overt diabetes whereas 307 and 305 mice showed slight increases in glucose levels (25) (Figure 3A). Most of the double-mutant mice succumbed to diabetes between 4 and 12 months of age. Therefore, to avoid the confounding effects of chronic hyperglycemia, we carried out metabolic measurements in 8-week-old mice. L1 mice showed a 25-fold increase in insulin levels with normal glucose levels, indicative of a state of compensated insulin resistance. L1/307 and L1/305 mice had greater than 50-fold higher insulin levels than WT mice, but this increase failed to prevent diabetes (Figure 3B).
Metabolic data in 2-month-old mice. We measured glucose (A) and insulin levels (B) in random-fed 2-month-old mice. *P < 0.01 versus WT, 305, and 307.
We have previously shown that L1 mice have normal hepatic insulin signaling and gene expression (26) while 307 and 305 mice have increased levels of gluconeogenic genes (25). Measurements of hepatic insulin signaling and gene expression in double-mutant mice L1/305 and L1/307 showed the expected metabolic abnormalities, with increased expression of glucogenetic enzymes (Pck1 and G6pc) accompanied by normal Insr content and phosphorylation (data not shown).
Expression of the FoxO1 transgene in β cells prevents expansion of β cell mass. We next examined islet morphology (Figure 4, A and B) and β cell mass (Figure 4C) in 8-week-old mice. L1 mice showed a 2.4-fold increase in β cell mass whereas neither 307 nor 305 transgenics showed significant differences compared with WT mice. L1/307 double transgenics showed a 3.4-fold increase in β cell mass while L1/305 mice showed a 60% increase compared with WT but an approximately 65% decrease compared with single L1 transgenics (Figure 4C). These data are consistent with the hypothesis that hepatic expression of the FoxO1 transgene causes diabetes by augmenting insulin resistance while β cell expression of the FoxO1 transgene does so by curtailing expansion of β cell mass. The relative increase of β cell mass in L1/305 compared with WT or 305 mice (Figure 4C) can be accounted for by increased replication of the 20% of β cells that do not express the FoxO1 transgene (Figure 1C).
Pancreatic histology and morphometry. (A) Insulin immunohistochemistry. (B) Glucagon immunohistochemistry. Magnification, ×10 (top panels); ×40 (bottom panels). (C) Measurements of β cell mass. We determined β cell mass as indicated in Methods, following immunohistochemistry with anti-insulin antiserum. We scored at least 3 animals per genotype and 4 sections per animal spaced 80 μm apart. *P < 0.01 versus WT, 305, and 307.
To examine why β cell mass failed to increase in L1/305 mice, we measured rates of β cell proliferation and apoptosis. Only occasional apoptotic cells (<0.001% of β cells) were present in all mice analyzed, and we did not detect differences among the various genotypes. Proliferating β cells, as measured by Ki67 incorporation, were rare in WT, 307, L1, and L1/307 mice (<0.01% of β cells) but were completely absent in 305 and L1/305 mice, indicating that proliferation was effectively prevented by the FoxO1 transgene. It should be noted that, although L1/305 mice had smaller β cell mass than L1/307 mice, their circulating insulin levels were similar, suggesting that they compensated for peripheral insulin resistance by increasing insulin secretion. These data indicate that, while the FoxO1 transgene inhibits β cell proliferation, it does not adversely affect insulin secretion. Indeed, we have recently reported that FoxO1 promotes expression of the insulin gene transcription factors NeuroD and MafA (27).
Effects of the FoxO1 transgene on β cell neogenesis. There is continuing controversy on the role of neogenesis versus proliferation in β cell turnover (4). We explored this issue in our experimental crosses. The rationale was that, since the FoxO1 transgene is expressed from the transthyretin promoter, it should affect proliferation not only of terminally differentiated β cells but also of duct cells, a site of transthyretin expression (28). We obtained 2 surrogate measures of β cell “neogenesis” by counting β cell singlets/doublets and β cells abutting the lumen of pancreatic ducts (4, 21, 22). While neither measure can be unequivocally assumed to represent neogenesis, certain testable predictions can be made.
If singlets/doublets arise from undifferentiated progenitors in response to insulin resistance, their number should increase in L1/305 transgenics to compensate for the failure to increase β cell mass through replication because the FoxO1 transgene is not expressed in progenitor cells. If they arise from replication, their number should decrease in L1/305 transgenics in proportion to the overall decrease in β cell mass because the FoxO1 transgene prevents their replication. Indeed, the number of singlets/doublets remained constant throughout the various genotypes, except in L1/305 double mutants. The latter showed a 65% decrease compared with WT mice and a 55% decrease compared with L1 mice (Figure 5A). These data are consistent with the notion that β cell singlets/doublets arise from replication of existing β cells (4).
Measurements of β cell singlets/doublets and ductal β cells. (A) We measured the number of extra-islet β cell singlets and doublets and normalized the number by the area of pancreas examined. At least 4 independent sections were scored for each of 3 animals of each genotype. *P < 0.01 versus WT. (B) For measurements of duct-associated β cells, we scored insulin-positive cells that could be unequivocally located within the lumen of a pancreatic duct and were not part of an islet. The number was normalized by section area. At least 4 independent sections were scored for each of 3 animals of each genotype. *P < 0.01 versus WT, 305, and 307. (C) Duct cell proliferation. We determined the Ki67-labeling index of duct cells, identified based on their anatomical location. We scored 6 mice per genotype and 3 sections for each animal.
With respect to the ductal origin of β cells, the prediction was that, if duct epithelia are a site of β cell neogenesis, the number of periductal β cells should decrease in mice expressing the FoxO1 transgene, as the ductal promoter-driven transgene will curtail expansion of this cell population. If β cells arise from other cell types, residing by happenstance in proximity to ducts, their number should increase to compensate for peripheral insulin resistance. In L1 mice, despite the increase in β cell mass, we failed to detect an increase in the number of duct-associated β cells. Unexpectedly, we saw a decrease in 307 mice, which, however, do not express the mutant transgene in ducts (25). The 305 transgenics had numbers similar to those of WT and L1 mice. Both L1/307 and L1/305 transgenics showed a decrease in periductal β cells compared with L1 mice (Figure 5B). As a control, we measured overall proliferation rates in pancreatic ducts and detected no differences among the various genotypes (Figure 5C). The decrease of periductal β cells in L1/305 could be construed as consistent with the ductal cell hypothesis. However, because a similar decrease was present in L1/307 mice, which do not express the transgene in ducts, this finding cannot be attributed solely to the effects of the FoxO1 transgene. The conclusion of these experiments is that the reduction of β cell mass in L1/305 transgenics is largely, if not exclusively, due to impaired β cell replication in adult mice. If there is a contribution from extra-islet β cells (ductal or otherwise), it is quantitatively small. Our data also caution against a quantitative interpretation of periductal β cells until more specific lineage markers are employed to address this question.
A paracrine mechanism of β cell expansion. The failure to increase β cell mass in L1/305 transgenics may be due to the complex genetic makeup of these mice, which lack Insr in many cell types (26). To obtain independent evidence for the role of β cell proliferation in insulin-resistant states, we developed a model of β cell hyperplasia by ectopic growth factor expression in the pancreas. To this end, we generated elastase-IGF2 transgenic mice (henceforth, El-IGF2) (29). We chose IGF2 because it binds with equal affinity to the 2 receptors of the insulin/IGF family, Insr and Igf1 receptor (30). Moreover, since IGF2 is not expressed in adult rodents (31), changes in IGF2 levels are readily detectable. The elastase promoter was chosen because it is expressed in exocrine pancreas and ducts, thus enabling transgenically produced IGF2 to act in a paracrine fashion on β cells in view of the anatomical contiguity of the exocrine/duct and endocrine compartments (32). El-IGF2 transgenic mice showed elevated IGF2 levels in whole pancreas extracts (Figure 6, A and B) without detectable circulating IGF2 in plasma (data not shown). Four-week-old El-IGF2 mice had a slight decrease in glucose levels and a 2-fold increase in insulin levels, associated with a 2-fold increase in Ki67 labeling of β cells (Figure 6, C–E). Total pancreas weight (Figure 6F) and β cell mass (Figure 6G) were similar to WT mice at this age. In age-matched El-IGF2/305 mice, we observed whole-pancreas IGF2 (Figure 6, A and B), plasma glucose, and insulin levels similar to those of El-IGF2 mice (Figure 6, C and D). In contrast, we detected trends toward decreased Ki67-positive β cells (Figure 6E) and β cell mass (Figure 6G), with normal pancreas weight (Figure 6F). Because the El-IGF2/305 mice do not develop diabetes, we could also measure the long-term effects of the FoxO1 transgene in 6-month-old animals (Figure 7). At this age, we observed normal glucose levels across all genotypes and only modest increases of insulin levels (<2-fold) in all transgenics. β cell mass increased approximately 2-fold in El-IGF2 mice but decreased approximately 30% in 305 mice and approximately 40% in El-IGF2/305 double-transgenic mice.
IGF2 levels and metabolic data in 4-week-old mice. (A) We measured mean ± SEM IGF2 levels using an ELISA assay in acid-ethanol extracts from whole pancreata. *P < 0.01 versus WT and 305. (B) IGF2 levels in individual transgenic mice and WT controls. (C) Glucose and (D) insulin levels in random-fed mice. (E) β cell replication. We determined the labeling index of pancreatic β cells by double immunohistochemistry with anti-Ki67 and anti-insulin antibodies. We scored at least 3 animals per genotype and 4 sections per animal. (F) Pancreas weight in El-IGF2, 305, and WT mice in 6 mice for each genotype. (G) We determined β cell mass as indicated in Methods in 6 mice for each genotype. IGF2, El-IGF2.
Metabolic data in 6-month-old mice. We measured glucose (A) and insulin levels (B) in random-fed mice. (C) Pancreatic morphometry. We determined β cell mass as indicated in Methods. We scored at least 3 animals per genotype and 4 sections per animal. *P < 0.05 versus WT.