Insulin Internalizes GLUT2 in the Enterocytes of Healthy but Not Insulin-Resistant Mice

Caco-2/TC7 on cells filters (3 μm HD; BD Biosciences, Pont de Claix, France) were grown in Dulbecco's modified Eagle's medium (DMEM; Gibco, Paisley, U.K.) supplemented with 25 mmol/l glucose and 10% decomplemented FCS (AbCys, Paris, France) (15). Insulin stimulation (1 mU/ml in basolateral compartment, 1 h, 37°C) was performed in serum-free media. Differentiated cells were washed with ice-cold PBS/Ca²⁺/Mg²⁺ buffer before processing.

Caco-2/TC7 cells were fixed with 4% paraformaldehyde (wt/vol) at 4°C, and antibody labeling was made on permeabilized cells with 1% Triton X-100. Antibodies were added for 1 h, and extensive washing with PBS/glycine buffer was performed between steps. Secondary antibodies were labeled with cyanin. Images were taken using a confocal microscope (LSM 510; Zeiss) and Leica DMRXA microscope with CCD camera (5MHz Micromax 1300Y; Roper Instruments) with the Metamorph software (Universal Imaging) to deconvolute Z-series. Colocalizations were analyzed with ImageJ1.37 software (http://rsb.info.nih.gov).

Cells were labeled from the BLM or BBM side with 1 mg/ml membrane-impermeant sulfo-NHSS-biotin and biotinylated proteins purified on NeutrAvidin gel columns (Pierce, Rockford, IL) as described previously (11).

Total AKT and phospho-AKT contents were evaluated on frozen cells. Western blots on cleared lysates (20 mmol/l Tris, 150 mmol/l NaCl, 5 mmol/l EDTA, 1% [vol/vol] Triton X-100, and 0.5% deoxycholate) were revealed with specific AKT and Ser⁴⁷³P-AKT antibodies (Cell Signaling Technologies).

C57bl/6J mice were fed with isocaloric 65% glucose-, fructose-, or protein-rich diets (6) or 72% fat, low-sugar (18) or standard chow (M25; Safe, Augy, France). Body weights of standard chow–and fructose-fed age-matched mice after 30 days (23.5 ± 0.5 vs. 23.10 ± 0.4 g; n = 24) and age-matched mice fed standard chow and high-fat diets after 4 months (27.4 ± 0.3 vs. 24.6 ± 0.4 g; n = 20) were similar. Gavages were performed after an 18-h fast using saline or 4 mg/g body wt glucose or fructose solutions. Intraperitoneal injections of saline (sham) or 1 mU/kg insulin (Actrapid; Novo Nordisk) were performed 20 min before gavage. Tail blood glucose (Accu-check; Roche Diagnostic) and plasma insulin (EIA; Linco Research, St. Charles, MO) were measured. The hyperinsulinemic-euglycemic clamp experiments were performed in conscious mice after a 5-h fast (19). Insulin perfusions in the jugular vein were at 0.6 mU · kg⁻¹ · h⁻¹ and 15 g/dl glucose at variable rate to maintain euglycemia (fructose diet: clamp 122 ± 17 mg/dl [n = 9], sham 135 ± 10 mg/dl [n = 9]; high-fat diet: clamp 148 ± 37 [n = 4], sham 152 ± 15 [n = 4]). Glucose infusion rates (GIRs) were calculated at the end of a 40-min clamp.

Fresh jejunal mucosa scrapings were processed for membrane preparations. Membrane fractionations in iodixanol (Optiprep Abscys) gradients performed as described previously (20) with 6 mg protein (BCA protein kit; Pierce) were laid on top of a 7.5–30% gradient, and step fractions (500 μl) were collected for density quantifications and Western blot analysis. BBM were purified using the MgCl₂/EGTA precipitation method (21).

Western blots were performed as described previously (6). Primary antibodies against rGLUT5, hGLUT5-Cterm (22), sucrase-isomaltase (23), NaKATPase-α1 (Abcam, Cambridge, U.K.), insulin receptor-β1 (Santa Cruz Biotechnology, Santa Cruz, CA), and peptide transporter-1 (PepT1) were used. For GLUT2 labels, rhGLUT2 (extracellular TM1-TM2 loop) (6), hmGLUT2 (Santa Cruz Biotechnology), and hGLUT2-Nter peptide gave similar results.

The transepithelial transfer of [³H]-3-O-methyl-glucose (3-OMG) in 25 mmol/l glucose-DMEM was measured in control and insulin-treated Caco-2/TC7 cells. The radioactivity was added at the BBM side of cells and 50-μl triplicate BLM medium samples were counted after 30 min. The results are picomoles ± SE from nine filters. Fructose uptake in jejunal tissue rings was measured as described previously (6). Uptakes are in nanomoles per milligram protein per second.

Results are given as means ± SEM (n), and ANOVA tests were performed.

The effects of insulin on enterocytes were studied in Caco-2/TC7 cells (15,24), which enabled a separation between the effects of insulin and glucose in vitro. Differentiated Caco-2/TC7 cells segregate sucrase-isomaltase in BBM and NaKATPase in BLM (Supplemental Fig. 1A, available in an online appendix at http://dx.doi.org/10.2337/db07-0928).

The β1-insulin receptor was found in BBM and BLM and in intracellular stores (XZ sections). In this experimental setting, glucose was added to the apical side and insulin to the basolateral side of the cells to mimic in vivo sugar absorption. Culture media were supplemented with 10% FCS containing insulin that could induce internalization of the insulin receptor. Glucose concentrations had not changed significantly at harvest (not shown). Insulin receptors were functional, as shown by AKT phosphorylation kinetics (Supplemental Fig. 1B).

Confocal microscopy analysis of the colocalization of GLUT2 with sucrase-isomaltase (Fig. 1A) indicated that GLUT2 was associated with BBM. GLUT2 was also abundant in subapical compartments where sucrase-isomaltase was undetectable. The quantification of GLUT2 and sucrase-isomaltase colocalization (Fig. 1A, Merge, line 1) indicated that 1 mU/ml insulin reduced BBM GLUT2 by 60% (Fig. 1D, Fluo).

To distinguish GLUT2 in intracellular membranes from BBM GLUT2, we performed surface biotinylation of proteins with impermeant NHSS-biotin. When biotin was added to the BBM side of cells, Western blotting of biotinylated proteins revealed GLUT5 and GLUT2 in BBM (Fig. 1C). NaKATPase could be biotinylated from the BLM (data not shown) but not from the BBM side of cells (Fig. 1C), indicating integrity of intercellular tight junctions. After exposure to insulin for 1 h, GLUT5, GLUT2, and NaKATPase levels in total membranes were unchanged (Fig. 1C), but BBM GLUT2 levels decreased by 80% (Fig. 1D, Biotin), whereas GLUT5 remained unchanged.

A precise deconvoluted image of a single microscopy section at mid-height of enterocytes shows association of GLUT2 with BLM identified by NaKATPase labeling (Fig. 1E). The addition of 25 mmol/l glucose increased BLM GLUT2 twofold after 15 min and fourfold after 30 min compared with sugar-deprived cells (Fig. 1F). BLM GLUT2 at 30 min contained 12% of total GLUT2 in whole cell. Insulin treatment decreased BLM GLUT2 by 45 and 34% after 15 and 30 min, respectively (P < 0.05) (Fig. 1F).

To determine the functional consequences of insulin action, we measured the transfer of the nonmetabolizable glucose analog 3-OMG from the apical to the basolateral compartment. The results revealed a 30% (P < 0.01) inhibition after a 30-min insulin treatment (Fig. 1B), which reflected the inhibition of GLUT2-mediated transport in Caco-2/TC7 cells. Thus insulin decreases BBM and BLM GLUT2 levels in enterocytes and limits the transepithelial transport of sugars by a specific internalization of GLUT2 from plasma membranes.

Mice fed a fructose-rich diet had a twofold (2.25 ± 0.9; n = 14) higher expression of GLUT2 in the jejunum compared with chow-fed mice. The maximum increase was attained after 5 days and remained stable thereafter. On the contrary, mice fed a high-fat diet expressed basal levels of GLUT2 (1.04 ± 0.3; n = 7) like chow-fed mice.

The consumption of fructose-rich and high-fat diets perturbed mice glucose homeostasis (Supplemental Table 1, online appendix). Glucose intolerance (ΔG) was found as soon as 5 days after the beginning of the fructose diet and after 3 months of fat feeding (not shown). Increased ΔG-to-ΔI ratios indicated that mice became insulin-resistant and fasting blood glucose levels were slightly increased.

To investigate insulin action in the intestine, we performed hyperinsulinemic-euglycemic clamps in conscious mice. Moderate hyperinsulinemia (2.36 ± 0.34 vs. 0.49 ± 0.05 ng/ml; n = 4) was obtained. GIRs were similar in chow- and fructose-fed mice for 15 days (67 ± 1 and 65 ± 4 mg · kg⁻¹ · min⁻¹; n = 4), indicating similar insulin sensitivity. However, GIRs were decreased at day 30 in fructose-fed mice, indicating that they were insulin resistant (36 ± 6 mg · kg⁻¹ · min⁻¹; n = 6). The GIR of mice fed a high-fat diet for 4 months was reduced even more (7 ± 4 mg · kg⁻¹ · min⁻¹; n = 6), indicating very high insulin resistance. We were then able to study insulin effects on GLUT2 trafficking by comparing sensitive and resistant mice.

Enterocyte membranes were separated according to densities (1.057–1.150 g/l) in iodixanol gradients (20). GLUT2 was found in BBM fractions containing the PepT1 marker and in BLM containing the NaKATPase marker (Fig. 2A and B). Intracellular (non-BBM, non-BLM) membranes also contained GLUT2. These results confirmed the presence of GLUT2 in BBM and intracellular GLUT2 stores in vivo.

Insulin lowered GLUT2 by 60% in PepT1 fractions, indicating removal from BBM (Fig. 2B and C) in absence of major protein degradation because total GLUT2 levels in sham and hyperinsulinemic-euglycemic clamp conditions were unchanged (Fig. 2B, inset). Accordingly, GLUT2 increased in intracellular membranes (Fig. 2C). Thus, insulin promoted GLUT2 internalization from BBM to intracellular membrane stores.

The functional significance of the insulin-dependent internalization of BBM GLUT2 with respect to sugar absorption was measured in everted jejunal rings of sham and hyperinsulinemic-euglycemic mice (Fig. 2D). Insulin reduced fructose uptake by 40% (P < 0.001), indicating that sugar transport correlated to BBM transporter abundance. We used cytochalasin B to determine the contributions of GLUT2 and GLUT5 (insensitive to cytochalasin B) to fructose uptake. Fructose uptakes were identical in hyperinsulinemic-euglycemic and sham mice exposed to cytochalasin B, thus confirming that insulin did not change localization and activity of GLUT5. Insulin specifically decreases GLUT2-dependent fructose transport.

The effect of insulin in hyperinsulinemic-euglycemic mice reduced GLUT2-dependent fructose uptake in conditions of fast. During a sugar meal, luminal sugars are high, triggering GLUT2 translocation to the BBM, and then blood glucose and plasma insulin rise. To document the impact of luminal sugar and insulin on the distribution of GLUT2 in enterocyte membranes, we force-fed insulin-sensitive mice with fructose and elicited GLUT2 trafficking into BBM (Fig. 2E–G).

Intraperitoneal insulin injections performed 20 min before the fructose test meal produced a threefold elevation of plasma insulin (range 1.2–1.7 ng/ml) to a level similar to that produced during hyperinsulinemic-euglycemic clamp. The recruitment of GLUT2 into BBM in response to luminal fructose was abolished by insulin injection, and the distribution of GLUT2 was similar to that in enterocytes of fasted mice (Fig. 2E). Interestingly, insulin effects tended to be weaker in purified BBM from mice fed 5 days with the fructose-rich compared with the glucose-rich diet (Fig. 2G). Prior insulin treatment counteracted the insertion of GLUT2 into BBM induced by luminal dietary sugar (Fig. 2E).

The membrane distribution of GLUT2 in insulin-resistant mice obtained by prolonged fructose (Fig. 3A–C) or fat feeding (Fig. 3D–F) was markedly perturbed. High levels of GLUT2 were found in BBM-PepT1 fractions (Fig. 3A and D), which represented 40% of total GLUT2 in the enterocyte membranes of mice fed fructose-rich diet for 30 days (Fig. 3C). This proportion increased to 80% in high-fat conditions (Fig. 3E). BLM GLUT2 levels were 10% (high-fructose; Fig. 3C) and 20% (high-fat; Fig. 3F) of total GLUT2 compared with 40% in insulin-sensitive mice (Fig. 2C). The functional impact of this redistribution of GLUT2 was revealed by the twofold increase in the initial slope (0–15 min) of blood glucose after an oral glucose tolerance test in resistant mice (high fat 0.30 ± 0.02 mg · dl⁻¹ · min⁻¹; high fructose 0.28 ± 0.01 mg · dl⁻¹ · min⁻¹; n = 8) compared with sensitive mice (0.13 ± 0.01 mg · dl⁻¹ · min⁻¹; n = 8; P < 0.001). Thus in insulin-resistant states, BBM GLUT2 was high and increased blood glucose entry.

In insulin-resistant mice, BBM localizations were unaffected by insulin in hyperinsulinemic-euglycemic clamp conditions. BLM GLUT2 remained <20% of total GLUT2 (Fig. 3C and F). In addition, BBM GLUT2 levels were correlated with GIR. Insulin-resistant mice exhibited a loss of control by insulin of GLUT2 membrane trafficking, leading to a permanent localization of GLUT2 in enterocyte BBM, a pathological consequence of long-term fructose or fat feeding.

We had previously reported the rapid and transient recruitment of GLUT2 into enterocyte BBM after a bolus of simple sugars, generating a threefold enhancement of uptake (6). The present study demonstrates and quantifies in vitro and in vivo the inhibition by insulin of intestinal sugar uptake, as a result of the internalization of GLUT2 from plasma membranes back into intracellular pools. This important physiological process mediated by insulin at the level of enterocytes probably reveals another mechanism by which insulin limits sugar excursion in the blood during a sugar-rich meal. Furthermore, insulin resistance provoked a loss of control by insulin of GLUT2 membrane trafficking, a pathological consequence of the long-term consumption of a fructose-rich or high-fat diet.

Insulin sensitivity and glucose tolerance are important factors in the regulation of intestinal GLUT2 trafficking. In insulin-resistant mice obtained by prolonged fructose-rich or high-fat diets, the distribution of GLUT2 was drastically altered; BBM GLUT2 was permanently high, and insulin was unable to promote GLUT2 internalization and enrichment in intracellular membranes. Insulin resistance resulted in increased intestinal sugar delivery, as reflected by higher initial rise in blood glucose after oral glucose load.

Insulin action is transduced via receptors in enterocyte plasma membranes in vivo (25,26) and in vitro in Caco-2/TC7 cells. The functional significance of BLM insulin receptors is obvious, and we speculate that enterocytes akin to epithelial renal cells undergo a rapid phosphatidylinositol (3,4,5)-trisphosphate diffusion after insulin to trigger trafficking of a BBM protein (27). The role of insulin receptors in the BBM is unclear but it creates an opportunity for oral insulin treatment (28) that might target BBM GLUT2 localization and function. Conflicting results concerning acute insulin treatment report inhibition (14) and activation (29) of SGLT1-dependent intestinal sugar absorption ex vivo. In the present work, we describe an insulin action on the high capacity GLUT2 component of fructose transport that leads to a 50% inhibition of total transport, down to GLUT5 basal transport levels. We speculate that this regulation applies to intestinal glucose absorption. In insulin-resistant animals, a permanent BBM GLUT2 might create glucose efflux from enterocytes into the lumen. This is unlikely because permanent SGLT1 recapture would operate. These results highlight the complexity of the regulation of intestinal sugar absorption by insulin to modulate but not block intestinal sugar absorption.

In adipocytes and muscle cells, insulin provokes a massive translocation of GLUT4 to the plasma membrane (rev. in 12,30). In sharp contrast, insulin internalizes BBM GLUT2 into intracellular pools. Insulin, by regulating GLUT2 and GLUT4 traffic in opposite direction, controls glucose homeostasis and limits glucose excursion in the course of digestion by increasing peripheral GLUT4-dependent glucose uptake and slowing down intestinal GLUT2-dependent sugar delivery. We therefore propose that the small intestine constitutes an early checkpoint to limit postprandial glucose excursion. We also anticipate that insulin treatment will remove GLUT2 from its permanent BBM localization in type I diabetic patients that are insulin sensitive.

Modern westernized diets and eating habits have changed, increasing the amounts of dietary fructose and fat, which are probably a nutritional basis for the obesity and type 2 diabetes pandemics (31). Intestinal adaptation to these diets may be an early event to the onset of metabolic disorders due to the rapid increase in sugar transport capacities and the alteration of insulin action in enterocytes. With time, insulin resistance and metabolic disorders (32) might be worsen by uncontrolled sugar absorption in a small intestine that is no longer responsive to attenuation by insulin. A vicious circle thus develops that equips the intestine for high transport of dietary sugar in organisms already suffering from excessive blood glucose levels. Strategies to ameliorate insulin action on intestinal function therefore constitute another target for therapeutic intervention and control of postprandial glycemia.

V.T. has received an MRT PhD fellowship. M.L.G. has received a Benjamin Delessert prize. E.S. has received an MRT PhD fellowship. A.G.B. has received a grant from FPI Spanish Ministerio de Educacion y Ciencia (BES-2004-4685). This work has received support from Institut National de la Santé et de la Recherche Médicale, University Paris 6, Action Thematique Concertée (ATC) Nutrition (ASEO22129DSA), ALFEDIAM/Merk-Lipha grants.

We thank E. Petridi and V. Ondet for their contribution and C. Lasne and E. Dussaulx for technical assistance. We are grateful to J. Chambaz for support, to G.W. Gould (Glasgow, U.K.) for antibodies, and to G.L. Kellett (York, U.K.) for antibodies and for fruitful discussions.