Novel Arguments in Favor of the Substrate-Transport Model of Glucose-6-Phosphatase Free

Yeast phosphoglucose isomerase, hexokinase, and glucose-6-phosphate dehydrogenase were from Roche Molecular Biochemicals. Dowex 1×8 chloride and borate resins were purchased from Acros. Poly(ethyleneglycol) 6,000 was from UCB (Belgium) and leupeptin and antipain from Takara (Japan). All other reagents (analytical grade) were from Acros or Sigma. [2-³H]glucose-6-phosphate was prepared by phosphorylation of [2-³H]glucose by hexokinase and purified by chromatography on Dowex 1×8 chloride and elution with an NaCl step gradient. [³²P]glucose-6-phosphate was synthesized as described by Vandercammen et al. (24). S3483 was kindly made available to us by Dr. H.J. Burger (Aventis Pharma Deutschland).

Washed liver microsomes (0.8–1.0 mg protein) prepared from overnight-fasted male Wistar rats were mixed with (final concentrations) 10 mmol/l Hepes (pH 7.1), 100 mmol/l KCl, 10 μg/ml leupeptin, 10 μg/ml antipain, 10 mmol/l octylglucoside, and 6 U yeast phosphoglucose isomerase (Roche) in a final volume of 50 μl. The mixture was frozen in dry ice for 10 min, thawed at 30°C for 10 min, and thoroughly vortexed. After two additional freezing and thawing cycles, the preparation was mixed with 1.5 ml of a buffer (buffer A) containing 10 mmol/l Hepes (pH 7.1), 100 mmol/l KCl, and 1 mg/ml defatted bovine serum albumin. After 10 min at room temperature, CaCl₂ and poly(ethyleneglycol) were added to final concentrations of 0.5 mmol/l and 2%, respectively. The microsomes were spun down by a 5-min centrifugation at 10,000g at room temperature. They were then washed three times in 1.5 ml buffer A containing 0.5 mmol/l CaCl₂ and 2% poly(ethyleneglycol). The last pellet was resuspended in 35 μl of a 10 mmol/l Hepes buffer (pH 7.1), containing 100 mmol/l KCl and 1 mmol/l dithiothreitol. Phosphoglucose isomerase assays showed that the four successive supernatants contained ∼90, ∼1, 0.2, and <0.1%, respectively, of the original phosphoglucose isomerase activity, and the final pellet contained ∼2.0–3.5%. Phosphoglucose isomerase was assayed spectrophotometrically at 30°C in a reaction mixture containing 100 mmol/l Tris (pH 7.5), 7.5 mmol/l MgCl₂, 0.25 mmol/l NADP, 5 μg/ml yeast glucose-6-phosphate dehydrogenase, and 1 mmol/l fructose-6-phosphate. One unit was the amount catalyzing the formation of 1 μmol glucose-6-phosphate per min under these conditions.

The measurements of detritiation were carried out at 30°C in a 200-μl mixture containing 20 mmol/l Hepes (pH 7.1), 100 mmol/l KCl, 1 mmol/l dithiothreitol, 0.1 mg/ml bovine serum albumin, 0.2 mmol/l glucose-6-phosphate, 100,000 cpm [2-³H]glucose-6-phosphate, and 2 μl of the microsomal preparation containing ∼10 mU phosphoglucose isomerase as well as the indicated concentrations of inhibitors. The reaction was stopped after 20 min by the addition of 0.1 ml 10% perchloric acid. After centrifugation, the supernatant was neutralized with 3 mol/l K₂CO₃ and tritiated water isolated on Dowex 1×8 columns, borate forms (25), and counted. In some experiments, an aliquot of the supernatant was also applied on Dowex 1×8 column, chloride form (1 ml of resin, in a Pasteur capillary pipette), which was washed with 2 ml water to elute [³H]water and [³H]glucose. [³H]glucose was calculated by subtracting the proportion of radioactivity eluted from the borate column from that eluted from the chloride column.

Microsomes (∼150 μg protein) were incubated at 30°C in 50 mmol/l Mes (pH 6.5), 100 mmol/l KCl, 2 mmol/l MgCl₂, 4% poly(ethyleneglycol), 1 mmol/l glucose-6-phosphate, and 100,000 cpm [³²P]glucose-6-phosphate with or without 0.5 mmol/l Pb(acetate)₂ in a final volume of 0.2 ml. At the indicated times, EDTA was added to a final concentration of 10 mmol/l, and the samples were submitted 30 s later to a short centrifugation (15 s at 10,000g). Perchloric extracts were prepared from the pellets and the supernatants. After neutralization, the extracts were diluted threefold with water and applied onto anion-exchange columns (Dowex 1×8 Cl⁻ resin), which were washed with 2 ml water. [³²P]Pi was eluted with 1.5 ml 0.5 mol/l NaCl and further isolated as a phosphomolybdate complex (26) before being counted. The chromatographic step was needed to remove poly(ethyleneglycol), which interfered in the extraction of the phosphomolybdate complex.

Microsomes (700 μg/ml) were incubated at 30°C in the presence of 50 mmol/l Hepes (pH 7.1), 100 mmol/l KCl, 10 mmol/l glucose-6-phosphate, 200 mmol/l mannitol, and 4% poly(ethyleneglycol) in a final volume of 0.4 ml. At the indicated times, vanadate was added to a final concentration of 1 mmol/l, and the samples were immediately centrifuged. The resulting supernatant was mixed with 0.2 ml 10% perchloric acid, and the pellet was resuspended in 0.2 ml 10% perchloric acid. This arrest procedure took <30 s. Mannitol 1-phosphate was measured enzymatically (23).

The method we wanted to use for the measurement of glucose-6-phosphate transport into microsomes requires the presence of phosphoglucose isomerase in the lumen of these vesicles. However, like other glycolytic enzymes, phosphoglucose isomerase is essentially present in the cytosol. Therefore, we developed a procedure to entrap purified yeast phosphoglucose isomerase into microsomes. This procedure consisted of mixing the microsomes with phosphoglucose isomerase in the presence of octylglucoside and submitting the preparation to freezing and thawing cycles. Four washings were then performed in a medium that contained CaCl₂ and poly(ethyleneglycol) to facilitate sedimentation of the microsomes.

Figure 2 illustrates this procedure applied to liver microsomes. Phosphoglucose isomerase was assayed both in the absence and presence of Triton X-100, which reveals the latent (membrane-surrounded) enzyme. The use of 10 mmol/l octylglucoside and three freezing and thawing cycles was found to be optimal for liver microsomes, resulting in the incorporation of ∼2.0–3.5% of the initial phosphoglucose isomerase, with a latency >75%. Lower degrees of incorporation were observed in the absence of octylglucoside or when no or only one freezing and thawing cycle was performed. Slightly higher degrees of incorporation were observed when 40 mmol/l octylglucoside was used instead of 10 mmol/l, but in this case, <20% of the total activity was latent.

We also evaluated the effect of such treatments on the inhibition exerted by 50 μmol/l S3483 on glucose-6-phosphatase. This inhibition amounted to 90% in untreated microsomes but to only ∼40% in the microsomes that had been freeze-thawed three times in the presence of 10 mmol/l octylglucoside, either with or without phosphoglucose isomerase. This decreased inhibition may be partly explained by membrane eversion (see discussion).

We first checked that detritiation of [2-³H]glucose-6-phosphate by purified phosphoglucose isomerase was not inhibited by 50 μmol/l S3483. Because part of the phosphoglucose isomerase in the microsomal preparation was not latent, it was of interest to use inhibitors of phosphoglucose isomerase that would be excluded from the vesicles derived from endoplasmic reticulum and therefore preferentially inhibit nonlatent phosphoglucose isomerase. Two such compounds were used, erythrose-4-phosphate and fructose-1,6-bisphosphate, both of which act as competitive inhibitors (27; E.V.S., unpublished results). Under our experimental conditions, they caused 50% inhibition of detritiation by purified yeast phosphoglucose isomerase at 8 μmol/l and 0.4 mmol/l, respectively. Their effect was not modified by the presence of 50 μmol/l S3483 (data not shown).

Figure 3 illustrates the effects of S3483, erythrose-4-phosphate and fructose-1,6-bisphosphate on the detritiation catalyzed by microsomes loaded with phosphoglucose isomerase. In this first type of experiment, glucose-6-phosphatase was inhibited by 1 mmol/l vanadate (28) to facilitate interpretation of the results. In the absence of erythrose-4-phosphate or fructose-1,6-bisphosphate, S3483 inhibited detritiation by up to ∼60% at a concentration of 50 μmol/l, whereas in the presence of these phosphate esters, the inhibition increased up to ∼90%. These results indicated that phosphoglucose isomerase entrapped into liver microsomes had become sensitive to the putative inhibitor of glucose-6-phosphate transport.

Replot of the data shown in Figs. 3A and B showed that the S3483-sensitive component of detritiation was approximately five times less sensitive to erythrose-4-phosphate and 10 times less sensitive to fructose-1,6-bisphosphate than the detritiation catalyzed by microsomes in the presence of a saturating concentration of S3483 or that catalyzed by free phosphoglucose isomerase (Figs. 3C and D). This result indicated that erythrose-4-phosphate and fructose-1,6-bisphosphate have no free access to the lumen of microsomes.

We then explored the effect of vanadate with or without fructose-1,6-bisphosphate and S3483 both on the hydrolysis and the detritiation of [2-³H]glucose-6-phosphate. If glucose-6-phosphatase and phosphoglucose isomerase competed for one single intramicrosomal glucose-6-phosphate pool, inhibition of the phosphatase might produce an increase in the detritiation rate. Vanadate, at a concentration (1 mmol/l) that inhibited glucose-6-phosphatase activity almost completely, stimulated the release of tritiated water in the absence but not in the presence of S3483 (Fig. 4), indicating that it acted only on the luminal component of detritiation. The stimulation of the S3483-sensitive component of detritiation amounted to 1.5-fold in the absence of fructose-1,6-bisphosphate and to approximately twofold in its presence. We also noted that fructose-1,6-bisphosphate had a modest stimulatory effect on glucose-6-phosphatase at all concentrations of S3483.

The goal of this second set of experiments was to quantify the proportion of the Pi formed by glucose-6-phosphatase that is initially present in the lumen of liver microsomes. To this end, we carried out incubations in the presence of Pb²⁺ to precipitate Pi in situ and hence slow down its release from microsomes. The concentration of Pb(acetate)₂ we used (0.5 mmol/l) had minimal inhibitory effect on glucose-6-phosphatase activity. When mixed with 0.1 mmol/l [³²P]Pi, it caused the instantaneous (in <30 s) and quantitative formation of a complex, which rapidly sedimented upon centrifugation (15 s at 10,000g). Addition of 10 mmol/l EDTA to such a complex caused its instantaneous (<30 s) solubilization, so that no radioactivity sedimented any longer upon centrifugation.

In the experiment shown in Fig. 5, microsomes were incubated with 1 mmol/l [³²P]glucose-6-phosphate and 4% poly(ethyleneglycol) (to facilitate sedimentation of the vesicles) either with or without 0.5 mmol/l Pb²⁺. At the indicated times, 10 mmol/l EDTA was added to the reaction mixture to dissolve all extramicrosomal precipitates, and the samples were centrifuged 30 s later. Supernatants (containing the medium) and pellets (containing the microsomes) were rapidly separated and treated with perchloric acid. After neutralization, inorganic phosphate was isolated and counted. Total formation of Pi by glucose-6-phosphatase was linear with time and was only minimally (∼10%) inhibited by Pb²⁺ (Fig. 5). In the absence of Pb²⁺, essentially all Pi was found in the medium, whereas in its presence (Fig. 5B), 80% of this product was recovered in the microsomal pellet after 5 min and 65% after 10 min.

To verify that the lead phosphate precipitate was surrounded by a membrane, we tested the effect of Triton X-100 with or without EDTA on its solubilization. As shown in Fig. 6, a combination of Triton X-100 (0.1%) and EDTA (10 mmol/l) caused the instantaneous dissolution of the precipitate, whereas these agents had minimal effects when used separately.

The problem of the primary site of release of the products of the glucose-6-phosphatase reaction was also investigated by measuring the formation of mannitol-1-phosphate from mannitol- and glucose-6-phosphate. Figure 7 shows that mannitol-1-phosphate first appeared inside the microsomes to reach a plateau after ∼5 min, whereas the appearance of this polyol-phosphate in the medium was delayed. Quite significantly, ∼80% of the mannitol-1-phosphate that had been formed at the shortest investigated time (1 min) was intravesicular. Other experiments showed that 50 μmol/l S3483 inhibited the total formation of mannitol-1-phosphate by ∼85% when glucose-6-phosphate was the phosphate donor (results not shown).

Our procedure to entrap phosphoglucose isomerase into microsomes allowed the incorporation of ∼2.0–3.5% of the initial isomerase activity. This represents ∼50% of the maximal theoretical incorporation that could be expected. The latter roughly amounts to the ratio of intravesicular water to total water in the incubation mixture, i.e., 5.1–6.4% if one assumes a microsomal protein concentration of 16–20 mg/ml in the incubation medium and an intracellular water space of 3.2 μl/mg protein (29). The high degree of enzyme incorporation indicates that phosphoglucose isomerase has been introduced in the lumen of the endoplasmic reticulum.

Part of the isomerase that was associated with the microsomes (typically 10–20% in liver microsomes prepared with our standard procedure) was not latent. This problem is most likely not caused by external phosphoglucose isomerase that remains adsorbed to the microsomes because the microsomal preparations were extensively washed after enzyme incorporation (see research design and methods). Furthermore, the proportion of isomerase that was not latent was quite high (∼85%) when the microsomes had been treated with a higher concentration (40 mmol/l) of octylglucoside. Therefore, we think that the nonlatent activity is contributed by microsomes that have incorporated phosphoglucose isomerase but are not tightly resealed because of residual detergent.

A second problem was that the loading treatment decreased (from 90 to ∼40%) the inhibition exerted by S3483 on glucose-6-phosphatase as well as the latency of mannose-6-phosphatase (data not shown). Both changes may be partly caused by the presence of leaky microsomes but also partly by eversion of the microsomal membrane. In the context of the substrate-transport model, this should lead to externalization of the catalytic site of glucose-6-phosphatase and loss of sensitivity to inhibitors of glucose-6-phosphate transport. S3483 is nonetheless expected to inhibit isomerization of glucose-6-phosphate catalyzed by such inverted vesicles because phosphoglucose isomerase is inside. The fact that S3483 inhibited more detritiation of [2-³H]glucose-6-phosphate than its hydrolysis (Fig. 4) indicates the presence of such inside-out vesicles.

Despite these problems, the phosphoglucose isomerase-loaded vesicles were a valid model to study glucose-6-phosphate entry, as further discussed below.

The observation that the inclusion of phosphoglucose isomerase into microsomes made it sensitive to inhibition by S3483 is best explained by assuming that this compound inhibits the entry of glucose-6-phosphate into microsomes. The fact that erythrose-4-phosphate and fructose-1,6-bisphosphate were 5- to 10-fold less potent to inhibit the S3483-sensitive (intramicrosomal) detritiation than that catalyzed by free phosphoglucose isomerase indicates that these compounds do not readily penetrate the intramicrosomal space. Therefore, these two pieces of evidence support the existence of a specific glucose-6-phosphate transporter that transfers its substrate from the surrounding medium to the lumen of the microsomes (and reciprocally because this transport is not energy-driven).

Previous experiments have shown that the transport of [¹⁴C]glucose-6-phosphate is apparently inhibited by vanadate, or when glucose-6-phosphatase is inactive due to mutations in its gene (18,19). These experiments were interpreted as showing that glucose-6-phosphate transport and hydrolysis are linked, i.e., that glucose-6-phosphate transport occurs concomitantly with its hydrolysis in the membrane of the endoplasmic reticulum. Quite the contrary, we show that vanadate caused a 1.5- to 2.0-fold stimulation of the detritiation of [2-³H]glucose-6-phosphate in microsomes loaded with phosphoglucose isomerase. This effect indicates that glucose-6-phosphatase and phosphoglucose isomerase compete for a single intramicrosomal substrate pool (as in the substrate-transport model) and that the glucose-6-phosphate transporter is rate-limiting under our experimental conditions.

As pointed out by Banhegyi et al. (5), the results of St-Denis et al. (18) and Lei et al. (19) can be reconciled with the substrate-transport model by considering that, when microsomes are incubated with [¹⁴C]glucose-6-phosphate, most (often close to 90%) of the radioactivity that accumulates in the vesicles (or most likely in a small fraction of them that have no rapid mechanism for glucose extrusion) corresponds to [¹⁴C]glucose, formed by hydrolysis of glucose-6-phosphate. Inhibition or inactivation of glucose-6-phosphatase decreases this formation, thereby decreasing the apparent uptake of [¹⁴C]glucose-6-phosphate.

The first piece of direct evidence for the orientation of the catalytic site of glucose-6-phosphatase toward the lumen of the endoplasmic reticulum came from the observation of luminal accumulation of lead phosphate by electron microscopy in sections of microsomes incubated with glucose-6-phosphate (22). We have now exploited the same principle to reach a quantitative assessment of the proportion of Pi that is formed inside microsomes. We found indeed that at the shortest time investigated, at least 80% of the hydrolyzed Pi can be precipitated with Pb²⁺ in membrane-bound vesicles. This intramicrosomal localization was indicated by the fact that the precipitate could be efficiently released only in the combined presence of a metal ion chelator and a detergent, whereas a simple lead phosphate precipitate only needed a chelator to be dissolved. Calculations indicate that the microsomal phosphate precipitate represented, after 10 min, ∼160 nmol Pi/mg microsomal protein. Such an amount is compatible with a luminal location (corresponding then to a concentration of ∼50 mmol/l, assuming an intravesicular water space of 3.2 μl/mg protein), but is difficult to reconcile with an intramembrane location (in a pocket in close proximity to the catalytic site, as in the conformational model). The observation that a minor part of the formed Pi was not precipitated with Pb²⁺ is most likely caused by the rate-limiting aspect of Pb²⁺ entry into microsomes as well as by the fact that a small proportion of the vesicles were not latent.

Another argument in favor of the luminal orientation of the catalytic site came from the observation that the majority (at least 80%) of the mannitol-1-phosphate that was produced from glucose-6-phosphate and mannitol was initially intramicrosomal. Under our experimental conditions, the steady-state intramicrosomal content reached 25 nmol/mg protein, corresponding to a luminal concentration of ∼8 mmol/l. Release of this phosphate ester from the vesicles at longer times indicates the presence of a transporter that slowly acts on mannitol-1-phosphate.

Although we do not provide evidence in this work that the primary site of release of glucose by glucose-6-phosphatase is the lumen of the endoplasmic reticulum, there are several reasons to think that such is the case. The first reason is that it is difficult to conceive how the catalytic site of glucose-6-phosphatase, an integral membrane protein (30), would be able to deliver some of its products (Pi, mannitol-1-phosphate) in the lumen of the endoplasmic reticulum and another product, glucose, in an intramembrane pocket. A second reason is the fact that S3483 inhibits a transporter that transfers glucose-6-phosphate to the lumen of the endoplasmic reticulum (see above), indicating that this substrate of glucose-6-phosphatase accesses the catalytic site via the luminal side of the membrane. The reason why most of the glucose that is formed from glucose-6-phosphate appears to be immediately in the medium is the same reason why it is true for Pi: transport from most vesicles is so rapid that, unless there is an efficient way of blocking it (such as Pb²⁺ in the case of Pi), released products are exported in a matter of seconds, preventing any correct assessment of their site of production. There is indeed ample evidence that a majority of the microsomal vesicles are equipped with transporters or pore-like structures that allow them to equilibrate glucose across their membrane in a matter of seconds (5,20).

This rapid equilibration of glucose across the membrane of the endoplasmic reticulum strongly argues against a role of vesicle-mediated secretion of glucose by the normal liver. Such a mechanism was postulated to account for the fact that mice that have no GLUT2 in their liver, having normal glucose output (31). An alternative explanation could be that other glucose transporters could take over the role of GLUT2, though less efficiently than the latter.

In conclusion, we have provided three new and independent arguments supporting the glucose-6-phosphatase model initially proposed by Arion et al. (2). These arguments concur with other kinetic evidence (2), with measurements of metabolite transport (5), and with topological studies of glucose-6-phosphatase (30). They agree also with the observation that GSD Ib, in which there is a defect in microsomal glucose-6-phosphate transport, is caused by mutations in a protein member of the organophosphate transporter family (13,32,33).

This work was supported by the Concerted Research Action Program of the Communauté Française de Belgique, Direction de la Recherche Scientifique; the Belgian Federal Service for Scientific, Technical and Cultural Affairs; and by the Juvenile Diabetes Foundation International.

I.G. was Fellow of the Fonds National de la Recherche Scientifique.

We thank Henri-Géry Hers, Maria Veiga-da-Cunha, and Ghislain Delpierre for helpful comments.