In Vivo Role of Focal Adhesion Kinase in Regulating Pancreatic β-Cell Mass and Function Through Insulin Signaling, Actin Dynamics, and Granule Trafficking

RIPcre⁺ mice (11) were bred to fak^fl/fl mice (a gift of Dr. Louis Reichardt, University of California, San Francisco, CA) to generate RIPcre⁺fak^+/fl, which were then intercrossed to generate β-cell–specific FAK knockout mice (RIPcre⁺fak^fl/fl). RIPcre⁺fak^+/+ littermates were used as controls, and both male and female mice were used in similar number for all experiments, with the exception of streptozotocin (STZ) experiments, where only male mice were used to avoid potential confounding effects of estrogen in β-cells (12). Genotypes for cre and fak genes were determined by PCR using ear clip DNA. All mice were maintained on a C57BL/6 background and housed in a pathogen-free animal facility with a 12-h light-dark cycle and fed standard irradiated rodent chow (5% fat; Harlan Tekad, Indianapolis, IN). All animal experimental protocols were approved by the Ontario Cancer Institute Animal Care Facility.

Glucose tolerance tests (GTTs) were performed on overnight-fasted (14–16 h) mice by intraperitoneal injection of glucose (1 g/kg of body weight), and blood glucose levels were measured at 0, 15, 30, 45, 60, and 120 min after glucose injection by glucose meter (Precision Xtra; Abbott Laboratories). Insulin tolerance tests were performed on overnight-fasted mice using human recombinant insulin (Novolin R; Novo Nordisk) at a dose of 1 unit/kg of body weight, and blood glucose levels were measured at 0, 15, 30, 45, and 60 min. The mice that were used were 4–8 and 12–18 weeks old.

The pancreas was isolated from 4–8- and 12–18-week-old mice as described in previous studies (13–15). Paraffin-embedded sections at three levels 150 μm apart were immunostained for insulin, Ki67 (DAKO), glucagon (Cell Signaling), and GLUT2 (Millipore). Immunofluorescent images were obtained by a Zeiss inverted fluorescent microscope (Advanced Optical Microscopy Facility, Toronto, Ontario, Canada). Immunohistochemically stained pancreatic sections for insulin or glucagon were scanned by ScanScope ImageScope system at ×20 magnifications and analyzed with ImageScope version 9.0.19.1516 software (Aperio Technologies, Vista, CA) for β- and α-cell area. Cell mass was calculated by β- or α-cell area multiplied by whole pancreas weight. Ki67-positive cells were manually counted on immunohistochemically stained pancreatic sections as percentages of total islet cells (∼250 islets were counted from each animal). Pancreatic sections were stained with hematoxylin and eosin (H&E) and imaged by light microscopy (Leica Microsystems, Inc.).

Male mice (6–8 weeks) were injected intraperitoneally with STZ (40 mg/kg of body weight) for three consecutive days and then killed for pancreas isolation. β-Cell apoptosis was assessed by transferase-mediated dUTP nick-end labeling (TUNEL) assay (Roche Biochemicals) according to the manufacturer’s protocol and imaged by a Zeiss inverted fluorescent microscope (Advanced Optical Microscopy Facility).

Protein lysates of isolated islets, liver, muscle, and hypothalami were isolated from 4–8-week-old mice, separated by SDS-PAGE, and immunoblotted with antibodies for FAK, IR, IRS2, pIRS1/2, p27, phospho-paxillin (Tyr 118), B-cell lymphoma-extra large (Bcl-xL), cyclin-dependent kinase 5 (CDK5), talin (Santa Cruz Biotechnology), phospho-IR (Tyr 1158/1162/1163) (BioSource), paxillin (BioLegend), Bcl-2 (Calbiochem), phospho-Akt (Ser 473), Akt, p53, phospho-extracellular signal–related kinase 1/2 (phospho-ERK1/2) (Thr202/Tyr 204), ERK1/2, pancreatic and duodenal homeobox 1 (PDX-1), cleaved caspase 3, cyclin D1, and glyceraldehyde-3-phosphate dehydrogenase (Cell Signaling) as previously described (14–16). The signal densities of Western blots were quantified by Quantity One software (BioRad).

Glucose-stimulated insulin secretion was measured on overnight-fasted 4–8-week-old mice after intraperitoneal injection of glucose (3 g/kg of body weight), from saphenous vein blood samples at 0, 2, 10, and 30 min after glucose injection. Pancreatic islets were isolated from 4–8-week-old mice, and 10 similar-sized islets per mouse were handpicked under a dissecting microscope (Leica Microsystems, Inc.). Islets were incubated overnight in RPMI 1640 media without glucose (Gibco), and 2.5 mmol/L or 15 mmol/L glucose-containing media stimulation for 30 min and then acid/ethanol extraction was performed for insulin content as previously described (15,16). Serum and media samples were assayed for insulin by ELISA (Crystal Chem, Downers Grove, IL).

To detect F-actin, cells were fixed with Z-FIX (Anatech Ltd., Battle Creek, MI) and stained with Alexa Fluor 488–conjugated phalloidin (Invitrogen). β-Cells were identified by insulin immunostaining (Santa Cruz Biotechnology). Cell images were captured with a Zeiss AxioCamHRm and acquired with AxioVision 4.8 imaging software (Carl Zeiss MicroImaging). Data were analyzed using ImageJ software (version 1.41o; NIH) by averaging the two peak-intensity line scans after image background subtraction.

For intracellular Ca²⁺ measurements, islets were incubated for 45 min with 3 μmol/L Fura-2-AM (Fura-2-acetoxymethyl ester) (Invitrogen) and 0.06% pluronic acid (Invitrogen) in an extracellular calcium imaging solution as previously described (17). Islets were then imaged in fresh imaging solution with 0.5 mmol/L glucose and without Fura-2-AM or pluronic acid at 37°C with constant bath perfusion. Glucose (11 mmol/L) and KCl (20 mmol/L) were increased as indicated, and NaCl concentration was reduced as required. Fluorescence recordings were obtained every 5 s. Images were analyzed with Image Pro Plus version 6.2 (Media Cybernetics) or Ratio Cam software (Metamorph).

For localization of phosphorylated paxillin detection, cells were washed with PBS and fixed in 2% paraformaldehyde. Cells were coimmunostained using antiphospho-paxillin and either anti–synaptosomal-associated protein 25 (anti-SNAP25) or anti–syntaxin 1 (Sigma-Aldrich). Images were captured and colocalization coefficient analyses were performed using a laser scanning confocal imaging system (LSM510) equipped with LSM software (Carl Zeiss, Oberkochen, Germany).

Standard, whole-cell technique with sine+DC lockin function of an EPC10 amplifier and Patchmaster software (HEKA Electronics, Lambrecht/Pfalz, Germany) were used with experiments performed at 32–35°C as previously described (17). β-Cells were identified by size and the presence of a voltage-gated Na⁺ current that inactivated at approximately –90 mV (except when tetrodotoxin is present).

Pancreatic islets were isolated from 4–8-week-old mice by collagenase digestion and cultured in RPMI 1640 medium without glucose for 1 h. Islets were then stimulated with 15 mmol/L glucose-containing medium or saline for 2 min and fixed in Karnovsky-style fixative at 4°C for overnight fixation as previously described (18). Sections were imaged by a Hitachi H7000 transmission electron microscope at an accelerating voltage of 75 kV.

Data are presented as means ± SEM and were analyzed by two-tailed, independent-sample Student t test or one-way ANOVA, as appropriate. P values <0.05 were considered as statistically significant.

FAK was efficiently deleted in pancreatic islets of RIPcre⁺fak^fl/fl mice as confirmed by Western blot (Fig. 1A). A minimal residual expression of FAK protein in islets likely represents expression in non–β-cells. Insulin promoter activity has also been shown to be present in the hypothalamus (19); therefore, FAK expression levels were also examined in isolated hypothalami, in addition to other metabolic tissues including the liver and skeletal muscle. We confirmed that FAK expression was not diminished in these tissues in RIPcre⁺fak^fl/fl mice compared with RIPcre⁺fak^+/+ littermates (Fig. 1A). RIPcre⁺fak^fl/fl mice were healthy and born at the expected Mendelian ratio. Body weight was similar between RIPcre⁺fak^fl/fl and RIPcre⁺fak^+/+ littermates (Fig. 1B).

We next assessed the effects of FAK deletion in the pancreatic β-cells on glucose homeostasis. Random blood glucose levels, measured in 4–8- and 12–18-week-old mice, showed an increase in RIPcre⁺fak^fl/fl compared with their RIPcre⁺fak^+/+ littermates in both age-groups (Fig. 1C). We then performed GTTs, which demonstrated similar fasting blood glucose levels but impaired glucose tolerance in RIPcre⁺fak^fl/fl mice compared with RIPcre⁺fak^+/+ controls in both age-groups (Fig. 1D). To investigate whether the impaired glucose tolerance was due to changes in peripheral insulin sensitivity, ITTs were performed. The results showed similar blood glucose lowering after insulin injection between RIPcre⁺fak^+/+ and RIPcre⁺fak^fl/fl mice in both age-groups (Fig. 1E), which suggests that the impaired glucose tolerance in RIPcre⁺fak^fl/fl mice was not due to changes in peripheral insulin sensitivity but rather due to defects in β-cell mass and/or function. These results show that FAK in β-cells plays an essential role in regulating glucose homeostasis.

β-Cell area was measured by morphometric assessment on insulin-immunostained pancreatic sections. RIPcre⁺fak^fl/fl mice showed reduced β-cell area relative to pancreatic area and β-cell mass compared with RIPcre⁺fak^+/+ controls in both age-groups (Fig. 2A and B). Diminished β-cell mass in RIPcre⁺fak^fl/fl mice could arise from changes in β-cell proliferation or viability. Immunostaining for Ki67, a marker of proliferation, on pancreatic sections showed a decreased number of Ki67-positive cells in islets of RIPcre⁺fak^fl/fl mice compared with littermate RIPcre⁺fak^+/+ controls (Fig. 2C). We next examined for apoptosis and observed an increase in cleaved (activated) caspase 3 expression in isolated islets and TUNEL-positive, insulin-immunostained cells in pancreatic sections of RIPcre⁺fak^fl/fl mice compared with RIPcre⁺fak^+/+ controls (Fig. 2D and F), indicating increased β-cell apoptosis under basal conditions. This, combined with reduced β-cell proliferation in islets of RIPcre⁺fak^fl/fl mice, is likely responsible for the decreased β-cell mass. In order to assess whether deletion of FAK in β-cells would also render these cells more susceptible to apoptotic stimuli, we challenged these mice with multiple low doses of STZ. Indeed, RIPcre⁺fak^fl/fl mice were more susceptible to hyperglycemia than RIPcre⁺fak^+/+ littermates (Fig. 2E). An increase in random blood glucose levels of RIPcre⁺fak^fl/fl mice persisted post–STZ injections and remained significantly higher than RIPcre⁺fak^+/+ controls throughout the 14-day period post–STZ injections. This was likely due to an increase in susceptibility to β-cell apoptosis, as evidenced by increased TUNEL-positive β-cells in the pancreas of RIPcre⁺fak^fl/fl mice compared with RIPcre⁺fak^+/+ littermate controls (Fig. 2F).

Given the role of FAK in integrin signaling and cell adhesion, linking communication between the ECM and the intracellular actin cytoskeleton, we next examined islet architecture in RIPcre⁺fak^fl/fl mice. Murine islets are characterized by a spherical structure containing insulin-producing β-cells in the core and other endocrine cells (α, ε, δ, and pancreatic polypeptide cells) in the periphery (20). By H&E staining and insulin-glucagon fluorescent coimmunostaining, islet architecture appeared intact in RIPcre⁺fak^fl/fl mice (Fig. 3A), and α-cell area and mass were similar between RIPcre⁺fak^fl/fl and RIPcre⁺fak^+/+ mice (Fig. 3B). Given the role of FAK in integrin and insulin/PI3K signaling, we next assessed expression of signaling mediators in these pathways in isolated islets under basal conditions. Insulin and integrin signaling are regarded as critical molecular pathways in regulating pancreatic β-cell proliferation, viability, and function (21,22). We observed that insulin signaling was suppressed in the islets of RIPcre⁺fak^fl/fl mice, as evidenced by a significant reduction in the phosphorylation levels of IR, IRS1/2, and Akt (Fig. 3C). Some target proteins of this signaling pathway relevant for cell cycle and apoptosis in β-cells (23–25), cyclin D1, p53, p27, and PDX-1, were also affected (Fig. 3D). In addition, reduced phosphorylation of ERK1/2, the main effector of integrin signaling, was also found in islets of RIPcre⁺fak^fl/fl mice. Interestingly, studies have shown that ERK is required for cell growth through regulating CDK5 expression and activation (26). Furthermore, CDK5/ERK signaling has been shown to activate Bcl-2/Bcl-xL expression to prevent apoptosis in neuronal cells (27). Here we found reduced CDK5, Bcl-2, and Bcl-xL expression levels in RIPcre⁺fak^fl/fl islets (Fig. 3D), suggesting that FAK plays a critical role in β-cell survival through regulating multiple signaling pathways that are involved in cell proliferation and apoptosis.

Insufficiency in either β-cell mass and/or function can lead to impaired glucose tolerance. Thus, we next examined whether FAK is also essential for β-cell function. Random plasma insulin level was decreased in RIPcre⁺fak^fl/fl mice (Fig. 4A) but similar insulin content was maintained within similarly sized islets (Fig. 4B). To further investigate β-cell function, we measured glucose-stimulated insulin secretion and found reduced insulin secretion in response to glucose in vitro (Fig. 4C). Phase I and phase II insulin secretion were also suppressed in vivo in RIPcre⁺fak^fl/fl mice (Fig. 4D). In response to glucose, multiple critical steps take place in β-cells, starting with glucose transport through GLUT2, followed by glucose metabolism and generation of ATP, resulting in closing of ATP-sensitive potassium (K_ATP) channels. Subsequent cell membrane depolarization activates voltage-dependent calcium channels, causing the influx of calcium ions that in turn stimulates insulin granule trafficking and exocytosis.

To investigate which aspects of glucose-stimulated insulin secretion were affected by absence of FAK, we first examined levels of GLUT2. Expression and distribution of GLUT2 in β-cells was similar between those of RIPcre⁺fak^fl/fl and RIPcre⁺fak^+/+ mice (Fig. 4E). Furthermore, intracellular Ca²⁺ homeostasis, as evidenced by baseline Ca²⁺ levels and response to glucose (11 mmol/L), was similar between RIPcre⁺fak^+/+ and RIPcre⁺fak^fl/fl islets (Fig. 5A and B). There was, however, an increase in the Ca²⁺ response to KCl in the RIPcre⁺fak^fl/fl islets, as shown in Fig. 5A. This was due to an increased density of voltage-dependent Ca²⁺ current in β-cells from RIPcre⁺fak^fl/fl islets (Fig. 5C and D). This increased activity in Ca²⁺ channels in response to KCl may reflect a compensatory mechanism to overcome insufficient insulin secretion. These results suggest that glucose sensing and Ca²⁺ response to glucose were normal in FAK-deficient β-cells.

The normal Ca²⁺ response to glucose stimulation coupled with impaired insulin secretion in RIPcre⁺fak^fl/fl islets indicates that the functional defect in the islets might be due to impaired insulin granule trafficking. Since FAK is known to regulate the structure and function of F-actin (2), which is a key determinant controlling the access of insulin granules to the plasma membrane (28), we examined actin regulation in the β-cells of RIPcre⁺fak^fl/fl mice. At low glucose concentration (2.8 mmol/L), F-actin density was normally distributed in β-cells of RIPcre⁺fak^+/+ and RIPcre⁺fak^fl/fl mice (Fig. 6A). Importantly, although glucose stimulation decreased cortical actin density in RIPcre⁺fak^+/+ β-cells, there was no significant effect of glucose on cortical actin density in β-cells of RIPcre⁺fak^fl/fl mice. To further investigate this absent effect, we next examined paxillin and talin, which are focal proteins that regulate actin dynamics. These proteins can associate with FAK and form a linkage between integrin and F-actin filaments (29). We observed reduced expression levels of talin in RIPcre⁺fak^fl/fl islets (Fig. 6B). Paxillin can be directly activated by FAK and in our RIPcre⁺fak^fl/fl islets; we showed suppressed paxillin activity as evidenced by reduced phosphorylation at tyrosine 118 in both basal and glucose-stimulated conditions (Fig. 6C). In response to glucose, phospho-paxillin has been shown to colocalize with t-SNARE proteins, SNAP-25 and syntaxin 1, in primary β-cells (10). Consistent with the reduced phospho-paxillin expression by Western blot, RIPcre⁺fak^fl/fl β-cells showed reduced phospho-paxillin, which was colocalized with SNAP-25 or syntaxin 1 at the plasma membrane (Fig. 6D). These results demonstrate that the defect in glucose-stimulated insulin secretion is likely due to impaired glucose-dependent depolymerization of cortical actin through defective focal protein dynamics in β-cells of RIPcre⁺fak^fl/fl mice.

The actin-formed dense web beneath the plasma membrane blocks insulin secretion by providing a barrier for the movement of insulin granules under low glucose concentrations (30,31). Upon glucose stimulation, actin undergoes depolymerization, allowing the granules to dock and fuse with the plasma membrane, resulting in insulin release. Our results in Fig. 6 show that cytoskeleton actin depolymerization was impaired in FAK-deficient β-cells, which suggests that insulin granule trafficking might be hampered due to defective actin dynamics. To assess insulin granule trafficking, we examined insulin granule localization in the RIPcre⁺fak^fl/fl β-cells by electron microscopy. A previous study indicates that insulin granules that reside within 0.2 μm of the plasma membrane can be morphologically considered docked granules (32). We found that under basal conditions, there were fewer insulin granules near the plasma membrane (<0.2 μm) in the β-cells of RIPcre⁺fak^fl/fl mice compared with RIPcre⁺fak^+/+ mice (Fig. 7A and B). Upon glucose stimulation, the number of membrane-associated granules in RIPcre⁺fak^+/+ β-cells increased significantly as expected, whereas in FAK-deficient β-cells, this failed to occur. Consistent with the notion that knockout of FAK results in fewer docked insulin granules at the plasma membrane of β-cells, we observed a reduced exocytotic response in single β-cells from RIPcre⁺fak^fl/fl islets in response to a series of 10 membrane depolarizations from −70 to 0 mV (Fig. 8A and B), demonstrating that FAK may regulate actin dynamics in vivo to control insulin granule trafficking and insulin secretion in pancreatic β-cells.

Null mutation of FAK causes embryonic lethality in the early developmental period, supporting the critical role of FAK in multiple cellular processes (33). To gain insight into the in vivo role of FAK in pancreatic β-cell homeostasis, we developed mice with β-cell–specific deletion of FAK using the Cre-loxP recombination system. In this study, we show that these mice have reduced β-cell mass and exhibit a defect in glucose-stimulated insulin secretion, which together result in impaired glucose tolerance. The reduction in β-cell mass results from a concomitant downregulation of proliferation and upregulation of apoptosis, whereas insufficient insulin secretion appears to be due to the impaired control of actin dynamics and reduced insulin granule trafficking to the plasma membrane.

FAK coordinates signals from growth factor receptors, such as those for insulin or epidermal growth factor, to regulate cell growth and survival. As such, disruption of FAK activates p53, the cell cycle progression inhibitor, in endothelial cells during mouse embryogenesis (34). Cell cycle–regulating proteins p27 and cyclin D1 were shown to be required for FAK-dependent cell cycle progression in glioblastoma cells (24). FAK has also been identified as a critical mediator of insulin signaling (7–9), participating in IR phosphorylation (35) and activating PI3K/Akt (36,37) through phosphorylating the p85 subunit of PI3K, suggesting that FAK might associate with IRS/PI3K to activate insulin signaling and deletion of FAK in β-cells may directly attenuate insulin/Akt signaling, as shown in this study.

Genome-wide association studies have identified CDK5 regulatory subunit–associated protein-like 1 (CDKAL1) as one of the genes associated with the development of type 2 diabetes (38). A recent report indicates that siRNA knockdown of CDK5 expression in a rat insulinoma cell line (INS 832/13) leads to enhanced apoptosis through decreased activation of FAK, resulting in attenuation of the PI3K/Akt survival pathway (39). ERK is required for CDK5 expression (26) and the CDK5/ERK pathway is responsible for upregulation of antiapoptotic protein Bcl-2 and Bcl-xL (27). Consistent with these observations, FAK deletion in β-cells leads to increased susceptibility of apoptosis, suggesting FAK/ERK/CDK5 signaling to be crucial for β-cell survival. Furthermore, decreased phosphorylated FAK^Ser732 expression was observed in pancreas sections of humans with type 2 diabetes (39), which confirms that the activity of FAK is important for β-cell survival and may play a causal role in the development of type 2 diabetes. The results presented here support the notion that reduced FAK activity in β-cells can play a role in diabetes pathogenesis.

During the process of glucose-stimulated insulin secretion, pancreatic β-cells have been observed to undergo focal adhesion remodeling similar to the events that occur during cell migration (10). Upon glucose stimulation, paxillin is phosphorylated by activated FAK, and both of these activated focal proteins migrate to the plasma membrane at the newly formed filopodia in primary β-cells (10). Interestingly, phospho-paxillin has been shown to connect with F-actin at focal adhesions (40). Together, actin dynamics is a complex process that is critical for glucose-stimulated insulin secretion (4,31). Here, we show that FAK can regulate focal protein dynamics, which in turn can control cortical F-actin depolymerization in response to glucose. Additionally, phosphorylation of paxillin has also been considered to be important in globular actin/F-actin transition (41). In this study, we found reduced phospho-paxillin levels in basal and glucose-stimulated conditions in RIPcre⁺fak^fl/fl islets and decreased colocalization of phospho-paxillin with plasma membrane t-SNARE proteins SNAP-25 and syntaxin 1 in RIPcre⁺fak^fl/fl β-cells. These SNARE proteins are thought to be involved in regulating insulin granule fusion to the plasma membrane (42) and are linked to the actin cytoskeleton (43). Thus, in response to glucose, FAK phosphorylates and activates paxillin, which localizes with t-SNAREs to regulate cortical F-actin depolymerization, in turn facilitating insulin granule trafficking and exocytosis.

Talin acts as the key component for integrin-actin linkage and focal adhesion assembly (44). Talin can directly bind to integrin cytoplasmic tails and FAK by its NH₂-terminal head domain and actin filament by either NH₂- or COOH-terminal rod domains (29). Downregulation of talin expression by siRNA in HeLa cells slows the kinetics of cell spreading and prolongs the process time of β1 integrin maturation (45), which can upregulate Bcl-2 expression (46), and has been identified as a critical factor for maintain β-cell survival and function (21). These studies suggest that talin may play an important role not only in cytoskeleton remodeling but also in cell proliferation. The talin expression level can be modulated by mechanical stimuli through nitric oxide synthase activity (47), which has been shown to be induced by the integrin/FAK/Src/ERK pathway to regulate cell migration (48,49). Accordingly, we observed reduced talin expression in the RIPcre⁺fak^fl/fl islets, suggesting that FAK is essential in regulating talin expression, which in turn can contribute to impaired β-cell function by affecting actin dynamics and reducing β-cell viability.

Altogether, in this study, we show for the first time that FAK is required for the maintenance of both pancreatic β-cell mass and function in vivo, such that in its absence, glucose homeostasis is perturbed. We show that deletion of FAK in β-cells results in impaired cell proliferation, survival, and function. The reduced insulin exocytosis in the absence of FAK is likely due to defects in actin dynamics through impaired focal protein dynamics, resulting in insufficient insulin granule trafficking. We demonstrate in vivo that FAK has a critical dual role in regulating both β-cell viability and cell function, and may be a potential therapeutic target for type 2 diabetes.

This work was supported by Canadian Institutes of Health Research (CIHR) operating grants MOP-201188 and MOP-191501 to M.W. and a Canadian Diabetes Association (CDA) grant-in-aid to P.E.M. E.P.C. is supported by the CDA Doctoral Student Research Award. M.C. is supported by a fellowship from Alberta Innovates-Health Solutions (AI-HS). C.T.L. is supported by the Eliot Phillipson Clinician Scientist Training program and the Banting and Best Diabetes Centre (BBDC) postdoctoral fellowship. S.Y.S. is supported by the BBDC–Novo Nordisk Studentship and the Natural Sciences and Engineering Research Council Alexander Graham Bell Canada Graduate Scholarship. C.H. is supported by a studentship from CIHR and AI-HS. P.E.M. is an AI-HS Scholar and holds the Canada Research Chair in Islet Biology.

No potential conflicts of interest relevant to this article were reported.

E.P.C. generated and analyzed research data and prepared the manuscript. M.C., S.A.S., D.C., X.Q.D., C.H., A.F.S., and D.Z. generated research data. C.T.L. and S.Y.S. generated research data and edited the manuscript. H.Y.G., P.E.M., and M.W. designed experiments, supervised students, contributed to discussion and interpretation of the data, and reviewed and edited the manuscript. M.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

This study was presented as a guided audio poster at the 71st Scientific Sessions of the American Diabetes Association, San Diego, California, 24–28 June 2011.

The authors thank Dr. Shun Yan Lu (Ontario Cancer Institute, Toronto, Ontario, Canada) and Nancy Smith (University of Alberta) for their technical assistance.