Normal hepatic glucose production in the absence of GLUT2 reveals an alternative pathway for glucose release from hepatocytes

Abstract

Glucose production by liver is a major physiological function, which is required to prevent development of hypoglycemia in the postprandial and fasted states. The mechanism of glucose release from hepatocytes has not been studied in detail but was assumed instead to depend on facilitated diffusion through the glucose transporter GLUT2. Here, we demonstrate that in the absence of GLUT2 no other transporter isoforms were overexpressed in liver and only marginally significant facilitated diffusion across the hepatocyte plasma membrane was detectable. However, the rate of hepatic glucose output was normal. This was evidenced by (i) the hyperglycemic response to i.p. glucagon injection; (ii) the in vivo measurement of glucose turnover rate; and (iii) the rate of release of neosynthesized glucose from isolated hepatocytes. These observations therefore indicated the existence of an alternative pathway for hepatic glucose output. Using a [¹⁴C]-pyruvate pulse-labeling protocol to quantitate neosynthesis and release of [¹⁴C]glucose, we demonstrated that this pathway was sensitive to low temperature (12°C). It was not inhibited by cytochalasin B nor by the intracellular traffic inhibitors brefeldin A and monensin but was blocked by progesterone, an inhibitor of cholesterol and caveolae traffic from the endoplasmic reticulum to the plasma membrane. Our observations thus demonstrate that hepatic glucose release does not require the presence of GLUT2 nor of any plasma membrane glucose facilitative diffusion mechanism. This implies the existence of an as yet unsuspected pathway for glucose release that may be based on a membrane traffic mechanism.

The liver plays a central role in maintaining glucose homeostasis. In the postabsorptive or fasted states, production of glucose by liver is activated by a low insulin to glucagon ratio to meet the body needs. In the absorptive state, when plasma glucose and insulin levels are elevated, this mechanism is shut off and the liver stores glucose. A defect in the inhibition by insulin of hepatic glucose output is a characteristic of diabetes mellitus and is responsible for maintaining hyperglycemia (1, 2). The enzymatic mechanisms by which glucose is produced in liver through glycogenolysis or gluconeogenesis have been extensively studied (3, 4). The last step in the production of glucose from both pathways is the hydrolysis of glucose-6-phosphate to glucose and phosphate. This reaction is catalyzed by glucose-6-phosphatase, an enzyme complex whose hydrolytic active site is located in the lumen of the endoplasmic reticulum (5–7). This topological organization requires glucose-6-phosphate, produced in the cytosol, to be transported into the endoplasmic reticulum for hydrolysis. It has been assumed that glucose was then transported back into the cytosol for its eventual release out of the hepatocytes by facilitated diffusion through the plasma membrane glucose transporter GLUT2 (8).

We previously described the establishment and characterization of GLUT2^−/− mice (9). These mice are hyperglycemic, relatively hypoinsulinemic, have markedly elevated plasma glucagon levels, severe glycosuria and die around weaning. Because before weaning most of the glucose is produced by gluconeogenesis (10), the combined presence of hyperglycemia and glycosuria suggested that, even in the absence of GLUT2, hepatic glucose production proceeded efficiently. We therefore investigated the uptake, efflux and neosynthesis of glucose by liver and isolated hepatocytes from GLUT2^−/− mice. Our data indicate that hepatic glucose production does not need the presence of GLUT2 in the plasma membrane nor of any other facilitated diffusion mechanism for glucose. This suggests the existence of an alternative mechanism for hepatic glucose output that may be based on a membrane traffic mechanism.

MATERIALS AND METHODS

Animals and Analytical Techniques.

GLUT2^−/− mice were from our own colony (9). As control animals, we used wild-type or heterozygous mutant mice as no difference in any parameters could be observed between both types of mice. RNA was extracted by the acid guanidinium-thiocyanate/phenol-chloroform method (11) from total livers. Poly(A)⁺ mRNA was purified on oligo(dT)-cellulose (Collaborative Research). The cDNA probes were a 1.4-kb fragment of the rat GLUT1 cDNA (12), a 300-bp PvuII-EcoRI fragment of the mouse GLUT2 cDNA (13), a 1.4-kb XbaI-SmaI fragment of the GLUT3 cDNA (gift of F. Brosius, Michigan St. Univ.), a 2-kb HindIII-XbaI fragment of the GLUT4 cDNA (14), and a 3-kb PvuI-XhoI fragment of the rat SGLT1 cDNA (15). Glucagon injection tests were performed on 15 day-old mice. Glucagon was administered at 0.15 μg/10 g of body weight. Blood glucose levels were measured from tail vein samples with a glucometer (Bayer).

Hepatocyte Preparation.

Livers of 15-day-old mice were perfused through the vena cava with a buffer consisting of: 140 mM NaCl, 2.6 mM KCl, 0.28 mM Na₂HPO₄·2H₂O, 5 mM glucose, and 10 mM Hepes (pH 7.4). The perfusion was first for 5 min with the buffer supplemented with 0.1 mM EGTA and then for 15 min with the buffer containing 5 mM CaCl₂ and 0.2 mg/ml collagenase type 2 (Worthington). The isolated hepatocytes were then washed and suspended in a small volume of DMEM (GIBCO) without glucose and pyruvate and counted.

3-O-Methyl-d-Glucose (3OMG) Uptake and Efflux.

Hepatocytes were preincubated for 2 hr at 37°C in DMEM without glucose but containing pyruvate 1 mM and lactate 10 mM, then washed once with Krebs-Ringer buffer containing 10 mM HEPES, pH 7.4, and 1% BSA. Uptake was initiated by diluting 20 μl of the cell suspension (25·10⁶ cells/ml) with 230 μl of 3OMG at concentrations ranging from 0.1 to 20 mM and containing 1.25–5 μCi 3-O-methyl-d-[1-³H]glucose per assay (1 Ci = 37 GBq). Uptake was stopped after 2 min for the mutant hepatocytes and after 15 sec for the control hepatocytes with 1.5 ml of ice-cold PBS (10 mM Na₂HPO₄·2H₂O/138 mM NaCl/2.7 mM KCl/1.76 mM KH₂PO₄, pH 7.4) containing 1 mM HgCl₂ (stop solution) as described (16). After two washings in stop solution, the cells were lysed in 0.1% SDS, an aliquot was kept for protein determination and the rest was used for determination of radioactivity. For efflux measurements, hepatocytes were incubated with Krebs-Ringer buffer containing 10 mM Hepes (pH 7.4)-BSA containing 20 mM 3-O-methyl-d-[³H]glucose (2.5 μCi per test) at 5·10⁶ cells/ml for 75 min at room temperature. To initiate efflux, 5·10⁵ cells were pelleted and rapidly resuspended in 500 μl of Krebs-Ringer buffer containing 10 mM Hepes (pH 7.4)-BSA. Efflux was stopped by addition of stop solution. Washings, protein determination, and scintillation counting were performed as above.

Release of Neosynthesized Glucose.

Hepatocytes were incubated for 2 hr at 37°C with shaking in DMEM without glucose, but in the presence of 1 mM pyruvate, 10 mM lactate and 250 μM 3-isobutyl-1-methylxanthine. They were then pelleted, lysed in 0.1% SDS in PBS, and the protein content was determined. The glucose content of the cell lysate and the supernatant were measured by the glucose oxidase method (100 TRINDER kit, Sigma). For pulse-labeling experiments, hepatocytes (7.5 × 10⁵) were incubated with shaking at 37°C or 12°C in 500 μl of DMEM containing 1 mM pyruvate, 250 μM 3-isobutyl-1-methylxanthine, and 0.05 μCi of [¹⁴C]pyruvate. Incubations were stopped by placing the cells on ice followed by centrifugation at 4°C. The supernatants were removed and the cells lysed in 0.2% of sodium deoxycholate. [¹⁴C]Glucose was separated from charged metabolites by passage of lysates or supernatants on anion- and cation-exchangers (Dowex AG1-X8 and 50W-X8, respectively, from Bio-Rad) (17). For cytochalasin B inhibition tests, the cell suspensions were incubated with 50 μM of the inhibitor for 10 min at room temperature and then transferred at 37°C or 12°C in the medium containing [¹⁴C]pyruvate for 1 hr. For brefeldin A (5 μg/ml), monensin (1 μM), and progesterone (10 μg/ml) inhibition tests, cells were first incubated for 1 h at 12°C in the gluconeogenic medium containing [¹⁴C]pyruvate and the inhibitors were added 10 min before the end of the incubation. The cells were then transferred on ice, washed twice and incubated for 15 min at 37°C in DMEM without glucose but containing 2 mM pyruvate and 250 μM 3-isobutyl-1-methylxanthine and in the presence of the inhibitors.

Glucose Turnover Studies.

Rat insulin promoter (RIP)GLUT1 × GLUT2^−/− mice were generated by crossing transgenic mice expressing the GLUT1 glucose transporter specifically in their pancreatic β cells with GLUT2^+/− mice. The RIPGLUT1 × GLUT2^+/− mice were then intercrossed to generate RIPGLUT1 × GLUT2^−/− mice. These mice were as glycosuric as the GLUT2^−/− mice and their isolated hepatocytes were also unable to take up glucose. Adult RIPGLUT1 × GLUT-2^−/− and RIPGLUT1 × GLUT-2^+/? littermates were anaesthetized with intraperitoneal injection of pentobarbital (65 μg/g of body weight). They were equipped with an indwelling catheter placed into the vena cava through the femoral vein as described (18, 19). After a 2 day recovery period, the mice were fasted for 5 hr before measurement of glucose turnover. This was initiated by a first 1.5-μCi bolus injection of HPLC-purified d-[3-³H]glucose (Amersham) followed by continuous infusion of the tracer at a rate of 5 μCi⋅kg⁻¹⋅min⁻¹ for 120 min. At 80, 90, 100, 110, and 120 min of infusion, 5 μl of blood were sampled through the tail vein, deproteinized and the specific activity of d-[3-³H]glucose in the supernatant was analyzed (20). Under these conditions the rate of glucose turnover in the blood represents the rate of hepatic glucose production that is calculated from the specific activity of d-[3-³H]glucose in the blood. At completion of the experiment, the mice were killed, the liver removed and immediately weighted. Total liver protein content was determined by the BCA protein assay.

RESULTS AND DISCUSSION

As a first confirmation that glucose could be released from liver we performed intraperitoneal glucagon challenges in 15 days-old mice. Fig. 1A shows that the glycemia of control or GLUT2^−/− mice were increased with similar rapid kinetics after glucagon injection. Fasting hepatic glucose output was then calculated after constant in vivo infusion of [3-³H]glucose and measuring the isotopic dilution of the marker. To perform these measurements, we used GLUT2^−/− mice that had been crossed with transgenic mice expressing GLUT1 in their β cells under the control of the RIP. These mice live to the adult age and can breed (unpublished observations). They were chosen because they can reach a size that is amenable for catheter implantation and in situ metabolic measurements. The measured hepatic glucose production from GLUT2^−/− × RIPGLUT1 mice is presented in Table 1. It is similar to the production of glucose by control mice when normalized to body weight but is lower when normalized by liver weight, because homozygous mutant mice have larger liver than control mice (1.66 ± 0.06 g, n = 7, vs. 1.36 ± 0.12 g, n = 4, P < 0.05). Finally, when hepatocytes freshly isolated from 15 days old GLUT2^−/− mice were incubated in the presence of gluconeogenic precursors, total glucose output measured over a 2-hr period was not different between control and GLUT2-negative hepatocytes (395 ± 87 pmol/mg protein/2 hr vs. 297 ± 21 pmol/mg protein/2 hr; mean ± SEM, n = 9) (Fig. 1B).

Figure 1.

Hepatic glucose output proceeds unimpaired in the absence of GLUT2. (A) Control (+/?) or GLUT2^−/− mice were injected i.p. with saline or glucagon and the blood glucose levels were measured at the indicated times. Glucagon induced a marked increase in glycemia in both control and GLUT2^−/− mice. Results are expressed as mean ± SEM, n = 4–6. ∗, Different from glycemia at t = 0, P < 0.05. (B) Glucose output from freshly isolated hepatocytes of control (+/?) or GLUT2 ^−/− mice. Isolated hepatocytes were incubated for 2 hr at 37°C in DMEM without glucose but in the presence of 1 mM pyruvate and 10 mM lactate. The glucose released in the medium was not different between GLUT2^−/− and control hepatocytes. The data are the mean ± SEM of nine experiments.

Table 1.

Endogenous glucose production

	mg/kg body weight/min	pmol/μg liver protein/min
Control (n = 4)	29.0  ±  1.8	24.2  ±  2.3
GLUT2−/−(n = 7)	24.9  ±  1.2	16.9  ±  0.7*

In vivo assessment of endogenous glucose production in awake unrestrained adult, age and sex-matched littermate mice. Control: GLUT2^+/? × RIPGLUT1; GLUT2^−/−: GLUT2^−/− × RIPGLUT1. 

^*Significantly different from control for P < 0.05. 

To determine whether the observed glucose output was due to the reexpression of another glucose transporter isoform, we first analyzed the expression of the known functional glucose transporters. This was performed by Northern blot analysis of twice oligo(dT)-selected liver mRNA using as control total RNA from the indicated tissues. Fig. 2A shows that GLUT1 was expressed at low level in control and was not overexpressed in mutant mouse liver; GLUT2 was not detectable; GLUT3 was present at a very low level; GLUT4, GLUT5, SGLT-1 and SGLT-2 (not shown) were not detected. We then measured 3OMG uptake using freshly isolated hepatocytes. Fig. 2B shows the rate of 3OMG uptake in the presence of increasing substrate concentrations. Control hepatocytes take up 3OMG with a calculated V_max of 42 pmol/min/μg protein and a K_m of 14 mM whereas in the absence of GLUT2, V_max was 1.6 pmol/min/μg protein with a K_m of 7 mM. To evaluate the capacity for substrate efflux, hepatocytes were then preloaded over a 75-min period at 22°C with 20 mM radioactive 3-O-methylglucose. In these conditions, accumulation of the substrate reached the same level in control and GLUT2-negative hepatocytes and were at equilibrium with the external concentration. After rapid washing and incubation in nonradioactive medium, efflux of substrate from hepatocytes was measured. It was very fast and reached equilibrium in ≈10 sec with control hepatocytes whereas efflux was extremely slow from the GLUT2-negative hepatocytes with ≈25% of the accumulated substrate being released in 15 min.

Figure 2.

Absence of glucose facilitated diffusion transport mechanism in GLUT2^−/− hepatocytes. (A) The expression of glucose transporter isoforms was assessed by Northern blot analysis using 5 μg of twice oligo(dT)-cellulose-selected liver poly(A)⁺ RNA from control or GLUT2^−/− mice. For control tissues, 5 μg of total RNA was used. B, Brain; H, heart; I, intestine; K, kidney. SGLT1, Na⁺/glucose cotransporter 1. (B) 3OMG uptake by hepatocytes freshly isolated from control (+/?) and GLUT2^−/− mice. Uptake by control cells proceeded with high K_m (14 mM) and high V_max (42.6 pmol/μg protein/min). In the absence of GLUT2, K_m was 7 mM and V_max was 1.6 pmol/μg protein/min. (Inset) Kinetics of uptake of GLUT2^−/− hepatocytes with an expanded scale. (C) 3OMG efflux from preloaded control (+/?) and GLUT2^−/− hepatocytes. Hepatocytes were preloaded with 3-O-methyl-[³H]glucose for 75 min. Efflux was initiated by rapid washing and dilution with cold medium. The data represent the mean ± SEM of three to eight experiments for each time point.

The above experiments indicated, (i) that in the absence of GLUT2, hepatic glucose output proceeded at a normal rate; (ii), that only extremely low glucose facilitated diffusion activity remained; (iii) that this remaining activity could not account for the rate of hepatic glucose production measured in vivo [16 pmol/min/μg protein) for output (Table 1) vs. 1.6 pmol/min/μg protein for uptake and a >100-fold reduction in the rate of 3OMG efflux in the absence of GLUT2]. Thus, in the absence of GLUT2, hepatic glucose release must proceed by a mechanism distinct from facilitated diffusion through the plasma membrane.

As a first step toward the characterization of this alternate pathway, we set up a protocol for pulse-labeling glucose neosynthesized from [¹⁴C]pyruvate by the gluconeogenic pathway in isolated hepatocytes. This protocol further allowed to quantitatively measure intracellular and secreted glucose. We showed that in our experimental conditions there was a linear accumulation of radioactive glucose in the medium for at least 1 hr. We thus used this labeling protocol to first perform experiments at 37°C or 12°C. Fig. 3A shows that the rate of glucose neosynthesis by control and GLUT2-negative hepatocytes was identical at both 37°C and 12°C although, at low temperature, total glucose production was lower than at 37°C. At this latter temperature, >95% of the newly synthesized glucose was released from control and ≈75% from GLUT2-negative hepatocytes (Fig. 3B). When the temperature was lowered to 12°C, differences in glucose release became striking: ≈80% of glucose was still released from control hepatocytes but only 20% from GLUT2-negative cells. This indicated that the mechanism by which GLUT2-negative hepatocytes released glucose was temperature-sensitive. To determine whether two pathways coexisted in control hepatocytes, one temperature-sensitive and the other GLUT2-dependent, we evaluated the effect of cytochalasin B, an inhibitor of facilitated diffusion (21), on glucose release at low and high temperatures (Fig. 3C). At 37°C, cytochalasin B inhibited glucose release only slightly (≈12%). However, at 12°C, a 50% inhibition was observed. This was in agreement with the hypothesis that two pathways coexisted in normal hepatocytes with the cytochalasin B-sensitive pathway evidenced only when the low temperature-sensitive pathway was blocked. That the cytochalasin B-sensitive pathway depended on facilitated diffusion through GLUT2 was indicated by the absence of cytochalasin B-inhibitory effect on glucose release from GLUT2-negative hepatocytes (Fig. 3C).

Figure 3.

Release of pulse-labeled glucose from control and GLUT2^−/− hepatocytes. Isolated hepatocytes were incubated in glucose-free medium containing 1 mM [¹⁴C]pyruvate. Newly synthesized [¹⁴C]glucose in the cell lysate or culture medium was then separated from gluconeogenic intermediates by ion exchange chromatography. (A) Total neosynthesis of [¹⁴C]glucose was similar in both control (+/?) and GLUT2^−/− hepatocytes at 37°C or at 12°C. (B) At 37°C the release of glucose from control (+/?) hepatocytes was >95% whereas it was ≈75% in GLUT2^−/− hepatocytes. At 12°C, ≈80% of newly synthesized [¹⁴C]glucose was released from control hepatocytes. Strikingly however, only 20% of newly synthesized glucose was released from GLUT2^−/− hepatocytes. (C) Inhibition of glucose release by cytochalasin B. The release of glucose from control (+/?) hepatocytes pulse-labeled for 1 hr with [¹⁴C]pyruvate was not inhibited when the experiment was carried out at 37°C but a ≈50% inhibition was observed at 12°C. No inhibition of glucose release from GLUT2^−/− hepatocytes could be observed either at 37°C or 12°C, in agreement with the absence of facilitated diffusion. (D) Glucose release from GLUT2^−/− hepatocytes was not sensitive to the intracellular transport inhibitors brefeldin A (5 μg/ml) and monensin (1 μM). Hepatocytes were pulse-labeled for 1 hr with [¹⁴C]pyruvate, washed and then chased for 15 min at 37°C. Brefeldin A, monensin or diluent (C) were added 10 min before the end of the pulse and included in the washing and chase solutions. No inhibition of release could be observed. (E) Glucose release from GLUT2^−/− hepatocytes pulse labeled as in D could be markedly reduced when the experiments were performed in the presence of progesterone (10 μg/ml). ∗, Significantly different from control with P < 0.05.

The low-temperature sensitive pathway may be a membrane vesicle traffic-based pathway originating from the endoplasmic reticulum, in the lumen of which hydrolysis of glucose-6-phosphate takes place. We therefore evaluated whether glucose produced at 12°C in GLUT2-negative hepatocytes could be prevented from being released during a chase period at 37°C in the presence of the intracellular traffic inhibitors brefeldin A (22, 23) and monensin (24). GLUT2-negative hepatocytes were thus labeled for 1 hr at 12°C with [¹⁴C]pyruvate to accumulate intracellular glucose. At the end of a 15 min chase period at 37°C, intracellular and released glucose were measured. Fig. 3D shows that ≈50% of the newly synthesized glucose was released in control chase conditions and that neither brefeldin A nor monensin inhibited this release. We therefore tested the effect of progesterone on glucose release. Indeed, this substance has been shown to inhibit the transport of newly synthesized cholesterol from the endoplasmic reticulum to the plasma membrane (25, 26) by a pathway insensitive to brefeldin A and monensin (26, 27) but possibly involving caveolae (25, 28). Fig. 3E shows that progesterone led to a 50% inhibition of glucose release.

Our data thus demonstrated that, in the absence GLUT2, no compensatory expression of any known glucose transporter was induced and that only very low levels of facilitated diffusion of 3OMG through the plasma membrane remained detectable. These transport experiments were carried out with 3OMG to assess diffusion through the plasma membrane independently of further metabolism of the substrate. That the use of 3OMG, instead of glucose, would prevent the detection of an as yet unknown transport activity, specific for glucose and unable to carry 3OMG, seems unlikely as all the facilitated or Na⁺-dependent glucose transporters known so far can use 3OMG as substrate (29, 30). In the absence of GLUT2, the rate of hepatic glucose production was however not impaired. These observations therefore indicated the existence of a facilitated diffusion-independent mechanism for glucose release. We demonstrated that this pathway was temperature-sensitive, not inhibitable by cytochalasin B, brefeldin A or monensin but inhibitable by progesterone. These data are compatible with the hypothesis that a direct membrane traffic mechanism, from the endoplasmic reticulum to the plasma membrane, is responsible for glucose release in the absence of GLUT2. Alternative mechanisms may nevertheless be possible such as transport of glucose across the plasma membrane through nonspecific membrane carriers, like those belonging to the multidrug resistance transporter family (31). Finally, our data also indicated that in control hepatocytes, two pathways for glucose release coexist. The temperature sensitive pathway and one relying on diffusion of glucose through the plasma membrane transporter GLUT2. This latter pathway probably requiring glucose to diffuse from the lumen of the endoplasmic reticulum back into the cytosol.

A GLUT-2-independent pathway for hepatic glucose release may also exists in humans. This is supported by investigations of Fanconi-Bickel patients who lack functional GLUT2 due to mutations in both alleles of this gene (32). These patients however respond to glucagon or epinephrine challenges by an increase in glycemia (33) and markedly elevated glycosuria (34) in the 2 hr after glucagon injection thereby indicating that hepatic glucose production can also be strongly and acutely stimulated in the absence of GLUT2 in humans.

Glucose output is an essential component of glucose homeostasis and its dysregulation is a key pathogenic event in diabetes mellitus. The present results describe a so far unsuspected mechanism operating in the last step of glucose release. It will be important to further define its molecular nature. As hepatic glucose output is tightly regulated by hormones, in particular glucagon and insulin, it will also be important to determine whether this mechanism is under hormonal control.

ABBREVIATIONS

3OMG

3-O-methyl-d-glucose

RIP. rat insulin promoter.