Brain Contains a Functional Glucose-6-Phosphatase Complex Capable of Endogenous Glucose Production*

Glucose is absolutely essential for the survival and function of the brain. In our current understanding, there is no endogenous glucose production in the brain, and it is totally dependent upon blood glucose. This glucose is generated between meals by the hydrolysis of glucose-6-phosphate (Glc-6-P) in the liver and the kidney. Recently, we reported a ubiquitously expressed Glc-6-P hydrolase, glucose-6-phosphatase- (cid:1) (Glc-6-Pase- (cid:1) ), that can couple with the Glc-6-P transporter to hydrolyze Glc-6-P to glucose in the terminal stages of glycogenolysis and gluconeogenesis. Here we show that astrocytes, the main reservoir of brain glycogen, express both the Glc-6-Pase- (cid:1) and Glc-6-P transporter activities and that these activities can couple to form an active Glc-6-Pase complex, suggesting that astrocytes may pro-vide an endogenous source of brain glucose. Glc- 6-P, glucose-6-phosphate; Glc-PT, glucose-6-phosphate transporter; hGlc-6-Pase, human Glc-6-Pase; mGlc-6-Pase, mouse Glc-6-Pase; Ad, adenovirus; ER, endoplasmic reticulum; GFAP, glial fibrillary acidic protein; phosphohydrolase activity was assayed using three independ- ent brain microsomal preparations and performed with two measurements, namely one on disrupted microsomes to measure total phospha- tase activity and the other on disrupted microsomes pre-incubated at pH 5.0 and 37 °C for 10 min to measure the residual acid-resistant non-specific phosphatases. The difference between these measurements reflects the acid-sensitive Glc-6-Pase specific activity. Data are present as the mean (cid:3) S.E. The concentration of vanadate for 50% inhibition of Glc-6-P phosphohydrolase activity ( (cid:7) vanadate (cid:8) 0.5 ) was determined in deoxycholate-disrupted microsomes as described previously (14).

Most tissues and organs in the body are not thought to be able to generate endogenous glucose. Therefore, between meals these tissues depend on glucose generated predominantly in the liver and kidney and distributed via the blood. The liver and the kidney are the primary organs responsible for interprandial blood glucose homeostasis. This homeostasis is dependent upon the activity of the glucose-6-phosphatase (Glc-6-Pase) 1 complex, which is comprised of a glucose-6-phosphate transporter (Glc-6-PT) and a Glc-6-Pase catalytic unit (reviewed in Refs. 1 and 2). The Glc-6-PT is a single copy gene (3)(4)(5) that is expressed ubiquitously (6). In contrast, until recently, Glc-6-Pase activity was considered restricted solely to the liver, kidney, and intestine (1,2). However, there is evidence that endogenous glucose production does occur outside of these tissues. In the genetic disease glycogen storage disease type Ia (1,2), where the liver/kidney/intestine Glc-6-Pase-␣ (G6PC) (7,8) activity is deficient, patients are still capable of producing glucose (9 -11), implying there are alternative pathways for endogenous glucose production. This led to the discovery of a second Glc-6-Pase activity, now called Glc-6-Pase-␤ (G6PC3 or UGRP) (12)(13)(14), which is widely expressed (12).
Glc-6-Pase-␣ (15), Glc-6-Pase-␤ (16), and Glc-6-PT (17) are co-localized in the membrane of the endoplasmic reticulum (ER), embedded by multiple transmembrane domains. Glc-6-Pase-␣ is a 357-amino acid phosphohydrolase (7,8) expressed primarily in the liver, kidney, and intestine (18,19). Glc-6-Pase-␤ is a 346-amino acid phosphohydrolase (13,14) expressed ubiquitously (12). Both exhibit similar active site structures, form similar covalently bound phosphoryl enzyme intermediates during catalysis (16,20), and exhibit similar kinetic properties (14). In the active Glc-6-Pase complex, Glc-6-PT transports cytoplasmic glucose-6-phosphate (Glc-6-P) across the ER membrane into the lumen where Glc-6-Pase, with its active site inside the lumen, hydrolyzes intra-lumenal Glc-6-P to glucose and phosphate (reviewed in Refs. 1 and 2). Alone, neither Glc-6-PT nor Glc-6-Pase-␣ have significant microsomal Glc-6-P transport activity, but when co-expressed they couple to significantly increase the activity of the Glc-6-Pase⅐Glc-6-PT complex (3). We have recently shown that Glc-6-Pase-␤ also has the ability to couple functionally with Glc-6-PT (14). The discovery of Glc-6-Pase-␤ implies that non-hepatic tissues may be capable of endogenous glucose production through the activity of a Glc-6-Pase-␤⅐Glc-6-PT complex. This is of particular interest, because the liver and kidney are not the sole sites of glycogen storage in the body. Indeed, the muscle is the major reservoir of body glycogen (21,22), and the brain, which expresses significant levels of both the Glc-6-PT (6) and Glc-6-Pase-␤ (12,13), also stores glycogen. Moreover, glucose export from the brain has been demonstrated in children undergoing elective cardiopulmonary bypass surgery for congenital heart disease (23), suggesting that the brain may be capable of endogenous glucose production. In the brain, the primary sites of glycogen storage are the astrocytes. Astrocytes are the most abundant glial cells in the central nervous system, responsible for regulating the external neuronal environment, responding to injury, and modulating neuronal growth and maturation (reviewed in Refs. 24 and 25). An acid-labile Glc-6-Pase-like activity (26) has been reported in astrocytes. However, the role of astrocyte glycogen in glucose production is controversial because the presence of a functional Glc-6-Pase complex has never been demonstrated in the brain. In this study, we show that brain astrocytes possess an active Glc-6-Pase-␤⅐Glc-6-PT complex that can hydrolyze Glc-6-P to glucose, suggesting that astrocyte glycogen can be converted to glucose and may be a source of alternative energy in the neurons.
Glc-6-P uptake measurements were performed as described (3,14). Briefly, microsomes (40 g) were incubated in a reaction mixture (100 l) containing 50 mM sodium cacodylate buffer, pH 6.5, 250 mM sucrose, and 0.2 mM [U-14 C]Glc-6-P (50 Ci/mol). The reaction was stopped at the appropriate time by filtering immediately through a nitrocellulose membrane (BA85; Schleicher & Schuell) and washing with an ice-cold solution containing 50 mM Tris-HCl, pH 7.4, and 250 mM sucrose. The radioactivity measured within the microsomes represents both the translocated substrate, [U-14 C] Glc-6-P, and its hydrolytic product, [U-14 C]glucose. Microsomes permeabilized with 0.2% deoxycholate to abolish Glc-6-P uptake were used as negative controls. Two to three independent experiments were conducted, and at least three Glc-6-P uptake studies were performed for each microsomal preparation. Statistical analysis using the unpaired t test was performed with the GraphPad Prism program (GraphPad Software, San Diego, CA). Data are presented as the mean Ϯ S.E.
Glc-6-Pase-␣ Ϫ/Ϫ and Glc-6-PT Ϫ/Ϫ Mice-Mice deficient in Glc-6-Pase-␣ (28) and Glc-6-PT (29) have been described. All animal studies were conducted under an animal protocol approved by the NICHD Animal Care and Use Committee. To maintain viability of the Glc-6-Pase-␣ Ϫ/Ϫ and Glc-6-PT Ϫ/Ϫ mice, glucose therapy consisting of intraperitoneal injection of 25-100 l of 15% glucose every 12 h was initiated on the first post-natal day (29). Weaned mice were also given unrestricted access to mouse chow (Zeigler Bros., Inc., Gardners, PA). Microsomes were isolated from the brain and liver of 6 -7-week-old mice essentially as described (8,30). Each microsomal preparation represents one individual mouse, and at least three independent microsomal preparations were used for each assay.
Northern Blot and Western Blot Analyses-Total RNA was isolated by the guanidinium thiocyanate/cesium chloride method, fractionated by electrophoresis through 1.2% agarose gels containing 2.2 M formaldehyde, and transferred to a Nytran membrane by electroblotting. The filters were hybridized to a uniformly labeled mGlc-6-Pase-␤, mGlc-6-PT, or ␤-actin riboprobe.
For Western blot analysis, cell lysates or microsomal proteins were separated by electrophoresis through a 12% polyacrylamide-SDS gel and blotted onto polyvinylidene fluoride membranes (Millipore Co.). The membranes were incubated either with a monoclonal antibody against the FLAG epitope (Scientific Imaging Systems, Eastman Kodak), a polyclonal antibody against hGlc-6-Pase-␤ (22), or a rabbit anti-glial fibrillary acidic protein (GFAP) antibody (Affinity BioReagents, Inc., Golden, CO). The antigen-antibody complex was visualized as described previously (14,22).
Mouse Astrocytes in Primary Culture-Mouse astrocytes were prepared using the method of Pousset et al. (31). Briefly, brains were dissected from 2-3-day-old wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , or Glc-6-PT Ϫ/Ϫ pups, and the meningeal tissues were removed. The cortices were suspended in phosphate-buffered saline and flushed several times with a fire-polished Pasteur pipette. The resulting cell suspension was passed through a sterile nylon sieve (70-m pore size; Falcon) to remove clumps and centrifuged at 1,200 rpm for 5 min to collect the cells. The cell pellet was resuspended in Dulbecco's modified minimal essential medium containing 20% heat-inactivated fetal bovine serum, seeded at a density of 9 ϫ 10 4 cells/6-cm dishes, and then incubated at 37°C in a humidified 5% CO 2 , 95% air atmosphere. Under these conditions, neurons do not survive the mechanical dissociation, and the low plating density prevents oligodendrocyte proliferation (31). After 7 days in culture, the cells were incubated with fresh Dulbecco's modified minimal essential medium containing 10% fetal bovine serum, and this medium was changed weekly. After 21 days in culture, one set of cultures was fixed in 4% paraformaldehyde to stain for marker proteins. The second set of cultures was used to prepare cell lysates and microsomes for phosphohydrolase and Glc-6-P uptake assays and Western blot analyses.
The purity of the astrocytes was determined by staining for GFAP (32). The cells were fixed for 10 min in 4% paraformaldehyde, incubated for 30 min at room temperature in TST buffer (0.05 M Tris-HCl, pH 7.5, 0.15 M NaCl, and 0.1% Triton-X-100) containing 1% bovine serum albumin (BSA) and 10% horse serum (TST-BSA), and then incubated overnight at 4°C with a rabbit anti-GFAP antibody at 4 g/ml in TST-BSA. Following three washes with TST buffer, the cells were then incubated with a biotinylated goat anti-rabbit IgG for 30 min, and the antigen-antibody complex was visualized with the Vectastatin Elite ABC kit (Vector Laboratories, Burlingame, CA). Replicates omitting the primary antibody or substituting the primary antibody with a preimmune rabbit serum were used as controls. A mouse monoclonal antibody against fibronectin (Sigma) and a rat monoclonal antibody against myelin basic protein (Sigma) were used to measure contamination of the cultures by fibroblasts and oligodendrocytes, respectively.
In contrast to nonspecific phosphohydrolases, Glc-6-Pase has a characteristic acid-labile profile (34). Therefore, to measure the specific Glc-6-P hydrolase activity in the brain, we performed two measurements, one on disrupted brain microsomes to measure total phosphatase activity and the other on disrupted microsomes pre-incubated at pH 5.0 and 37°C for 10 min to measure the residual acid-resistant nonspecific phosphatases. The difference between these measurements reflects the acid-sensitive Glc-6-Pase-specific activity. The acid-sensitive Glc-6-P hydrolase activity in microsomes isolated from the brain of wild-type mice is 10.07 Ϯ 0.84 nmol/mg/min (Table I), which represents 60% of the total phosphohydrolase activity. The brain Glc-6-P hydrolase activity is ϳ5% of the activity of the liver at 217.2 Ϯ 6.0 nmol/mg/min (22). A previous report (26) ascribed 40% of the activity to acid-sensitive Glc-6-Pase when assayed at pH 6.5 and 30°C. Our findings are reasonably consistent with this report, given that the initial finding was assayed at a temperature suboptimal for Glc-6-Pase-␤ (Fig. 1B).
Glc-6-Pase-␣ occurs in the brain and is more readily measured at the higher temperature we used to assay Glc-6-Pase-␤, we examined the Glc-6-P hydrolytic activity of brain microsomes from Glc-6-Pase-␣ Ϫ/Ϫ mice. The Glc-6-Pase-␣ Ϫ/Ϫ brain has an activity identical to that of the wild-type brain, implying that Glc-6-Pase-␣ is not expressed in the brain (Table I). As expected, Glc-6-PT Ϫ/Ϫ brain microsomes have an identical activity as that of the wild-type brain (Table I). To further support the identification of the acid-labile brain activity as Glc-6-Pase-␤, the vanadate sensitivity (Table I) of the Glc-6-Pase in the brain was examined. The hepatic hGlc-6-Pase-␣ activity is inhibited 50% at a vanadate concentration of 3.1 mM ([vanadate] 0.5 ϭ 3.1 mM) (14), whereas the hGlc-6-Pase-␤ has a sensitivity of [vanadate] 0.5 equal to 1.4 mM (14). Consistent with this finding, the mouse brain Glc-6-Pase activity is similar to the hGlc-6-Pase-␤ activity, with a [vanadate] 0.5 of 1.3-1.34 mM (Table I).
Brain Astrocytes Possess an Active Glc-6-Pase-␤⅐Glc-6-PT Complex-In the brain, astrocytes constitute Ͼ50% of the cell mass and are the only cells that contain substantial amounts of glycogen (reviewed in Ref. 35). If astrocytes are a site of expression of Glc-6-Pase-␤ and Glc-6-PT, they may be capable of providing an endogenous source of glucose for the brain. To examine this possibility we isolated primary mouse astrocytes from 2-3-day-old wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , and Glc-6-PT Ϫ/Ϫ pups. After 21 days in culture, nearly all of the cells from wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , or Glc-6-PT Ϫ/Ϫ pups have the characteristics of astrocytes. Immunocytochemical analysis (Fig.  4A) shows that nearly all the cells stained positively for GFAP, a marker for brain astrocytes (32), and negatively for fibronectin and myelin basic protein, which are markers of fibroblasts and oligodendrocytes, respectively (31). Western blot analysis confirmed that the primary astrocytes from wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , or Glc-6-PT Ϫ/Ϫ pups express GFAP (Fig. 4B) but not fibronectin or the myelin basic protein (data not shown). Importantly, primary astrocytes from wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , and Glc-6-PT Ϫ/Ϫ pups express the Glc-6-Pase-␤ protein in their microsomes (Fig. 4B), strongly suggesting that brain astrocytes may be capable of hydrolyzing Glc-6-P to glucose via a functional Glc-6-Pase-␤ complex. To examine if this complex is functional, we assayed the astrocyte microsomes for Glc-6-P uptake and hydrolysis.
The acid-sensitive Glc-6-P hydrolase activities in microsomes isolated from the astrocyte primary cultures of wildtype, Glc-6-Pase-␣ Ϫ/Ϫ, and Glc-6-PT Ϫ/Ϫ mice are similar, ranging from 15.0 to 15.7 nmol/mg/min (Fig. 4C) and accounting for ϳ7% of the activity of the liver (22). This is ϳ50% higher than the specific activities in microsomes isolated from the whole brain homogenates (Table I), suggesting, based on cell mass, that the astrocytes are the primary source of the brain activities.
A comparison of Glc-6-P uptake in primary astrocytes shows that Glc-6-P uptake activities in the intact microsomes of wildtype and Glc-6-Pase-␣ Ϫ/Ϫ mice are identical (Fig. 4C). In contrast, the Glc-6-PT Ϫ/Ϫ microsomes show a markedly reduced Glc-6-P accumulation (Fig. 4C), again consistent with the absence of Glc-6-PT in these mice (29). DISCUSSION The mammalian brain has no energy reserves, and a constant supply of oxygen and glucose is required for its function and survival. Blood glucose homeostasis between meals is maintained by balancing the blood glucose uptake by tissues with the release of Glc-6-P-derived glucose to the blood by the liver and kidney. Therefore, the dephosphorylation of Glc-6-P to glucose, catalyzed by the Glc-6-Pase complex in the terminal step of both gluconeogenesis and glycogenolysis, is a key control point for interprandial glucose homeostasis (1,2). Until very recently, specific Glc-6-Pase activity has not been detected outside of the liver, kidney, and intestine. This has led to the current view that interprandial glucose homeostasis depends solely upon these organs. The recent recognition of Glc-6-Pase-␤ (13,14) and its universal expression profile (12,33) has raised the possibility that glucose recycling might also occur to some extent in many tissues. The tissues that immediately became of interest to this hypothesis are those with substantial reserves of glycogen that could be converted to glucose via glycogenolysis and that also express the higher levels of Glc-6-Pase-␤ and Glc-6-PT. Of particular interest are the skeletal muscle (21,22), the brain, and the heart.
The Glc-6-Pase found in mouse brain in this study and the in vitro expressed mouse and human Glc-6-Pase-␤ proteins have identical kinetic characteristics. All are acid-labile, readily inhibited by vanadate with [vanadate] 0.5 of 1.3-1.4 mM, and all share optimal assay conditions of pH 6.5 and 37°C, differing from those of Glc-6-Pase-␣ at pH 6.5 and 30°C. These characteristics, along with the RNA and protein expression profiles, support the conclusion that the acid-labile Glc-6-P hydrolytic activity in brain is Glc-6-Pase-␤. Of particular significance, the brain Glc-6-Pase is capable of coupling to the ubiquitous Glc-6-PT to create a functional Glc-6-Pase complex capable of converting Glc-6-P to free glucose. In refining the site of this activity, we have shown that astrocytes, the main site of glycogen storage in the brain, contain a functional Glc-6-Pase-␤⅐Glc-6-PT complex in their microsomes, suggesting they are capable of hydrolyzing Glc-6-P to glucose and might be a source of endogenous brain glucose.
Whereas the Glc-6-P hydrolase activity in the brain (or as-TABLE I Brain acid-sensitive Glc-6-P phosphohydrolase activity and inhibition by vanadate Glc-6-P phosphohydrolase activity was assayed using three independent brain microsomal preparations and performed with two measurements, namely one on disrupted microsomes to measure total phosphatase activity and the other on disrupted microsomes pre-incubated at pH 5.0 and 37°C for 10 min to measure the residual acid-resistant non-specific phosphatases. The difference between these measurements reflects the acid-sensitive Glc-6-Pase specific activity. Data are present as the mean Ϯ S.E. The concentration of vanadate for 50% inhibition of Glc-6-P phosphohydrolase activity (͓vanadate͔ 0.5 ) was determined in deoxycholate-disrupted microsomes as described previously (14). trocytes) represents only ϳ5-7% of the activity in the liver, the brain (or astrocyte) Glc-6-Pase-␤⅐Glc-6-PT complex has ϳ30 -32% of the Glc-6-P transport activity of the liver Glc-6-Pase-␣⅐Glc-6-PT complex. This is consistent with the previous reports that Glc-6-P translocation by Glc-6-PT is the rate-limiting step in Glc-6-Pase catalysis (19,36). Our results indicate that despite the lower hydrolase activity in the brain, the overall activity of the Glc-6-Pase⅐Glc-6-PT complex in the brain is still sufficient to convert glycogen to glucose at a significant rate. The potential role of brain glycogen in glucose homeostasis has been of significant interest. One concern that is expressed about the role of non-hepatic contributions to glucose homeostasis is based on kinetics. Glc-6-P is the intracellular form of extracellular glucose, trapped within the cell by the charged phosphate that prevents ready diffusion across the lipid cell membrane. The ability of the liver to maintain blood glucose homeostasis between meals is believed to stem, in part, from the balance between the Glc-6-P phosphohydrolase and glucose kinase activities present in the liver. The Glc-6-Pase-␣ activity in the liver is high, which allows for the rapid conversion of Glc-6-P derived from gluconeogenesis or glycogenolysis to glucose. Of the two glucose kinase activities, glucokinase and hexokinase, glucokinase predominates in the liver. Because glucokinase has the higher K m (ϳ20 mM) (37,38) compared with that of hexokinase (K m , ϳ50 M) (24,38), Glc-6-Pasemediated glucose release to the blood can compete with rephosphorylation by glucokinase. On the other hand, most other tissues, including the kidney, express the low K m hexokinase. This might imply that in non-hepatic tissues, Glc-6-Pase-derived glucose is more likely to be rephosphorylated than released into the circulation. There is one clear inconsistency with this explanation, namely the observation that the kidney does play a proven role in blood glucose homeostasis, especially during prolonged starving (reviewed in Refs. 39 and 40). One possible explanation is that the kinetic reasoning does not consider the subcellular compartmentalization of the different activities. The fate of cytoplasmic glucose, exposed to the glucokinase and hexokinase activities, might be very different from that of glucose produced within the ER lumen by the Glc-6-Pase complex. Within the lumen, glucose is protected from the glucose kinase activities and may be released directly via the ER/Golgi pathway (2) where it is promptly taken up and re-phosphorylated to Glc-6-P by adjacent cells. Indeed, using GLUT2-deficient cells, it was demonstrated that hepatic (41) or intestinal (42) glucose release does not require the presence of the plasma membrane glucose facilitative diffusion mechanism, implying the existence of an alternative pathway for glucose release via a membrane traffic mechanism. This finding might explain why the cytoplasmic pathways of gluconeogenesis and glycogenolysis evolved a final step dependent upon a compartmentalized Glc-6-Pase enzyme complex.
The finding of a functional Glc-6-Pase complex in the brain astrocytes is consistent with a number of earlier observations and the known roles of astrocytes within the brain. It has been shown that astrocytes are the primary source of glycogen storage in the brain, containing sufficient glycogen reserves to buffer the majority of the glucose supply deficit for Ͼ100 min of hypoglycemia (24). It is also known that sensory stimulation causes a breakdown of astrocyte glycogen, whereas anesthesia FIG. 3. The Glc-6-Pase-␤ complex in the brain is functional. Glc-6-P uptake assays were performed using microsomes isolated from the brain or liver of wildtype, Glc-6-Pase-␣ Ϫ/Ϫ , and Glc-6-PT Ϫ/Ϫ mice as described under "Experimental Procedures." A, time course of Glc-6-P uptake activity in intact microsomes isolated from the brain of wild-type (E), Glc-6-Pase-␣ Ϫ/Ϫ (•), or Glc-6-PT Ϫ/Ϫ () mice. B, microsomal Glc-6-P uptake activity. The radioactivity accumulated in the lumen of the ER consists of both [U- 14  Mouse astrocytes were isolated from 2-3-day-old wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , or Glc-6-PT Ϫ/Ϫ pups and plated at a density of 9 ϫ 10 4 cells/6-cm dish. After 21 days in culture, the cells were stained for GFAP (a marker for astrocytes), fibronectin (a marker for fibroblasts), or myelin basic protein (a marker for oligodendrocytes) as described under "Experimental Procedures." The data shown are for astrocytes obtained from wild-type mice at magnifications of ϫ1 (upper panels) and ϫ50 (lower panels). Similar results were obtained with primary astrocytes from Glc-6-Pase-␣ Ϫ/Ϫ and Glc-6-PT Ϫ/Ϫ pups. B, Western blot analysis. Glc-6-Pase-␤ in microsomes isolated from primary astrocytes of wildtype, Glc-6-Pase-␣ Ϫ/Ϫ , or Glc-6-PT Ϫ/Ϫ mice were detected using anti-hGlc-6-Pase-␤ (22). The lane marked mGlc-6-Pase-␤ represents protein from Ad-mGlc-6-Pase-␤-infected COS-1 lysates. GFAP in cell lysates was detected using a rabbit anti-GFAP antibody. C, astrocyte phosphohydrolase and Glc-6-P uptake activities. Glc-6-P hydrolytic and uptake assays were performed using microsomes isolated from primary astrocytes of wild-type, Glc-6-Pase-␣ Ϫ/Ϫ , and Glc-6-PT Ϫ/Ϫ mice as described under "Experimental Procedures." The results are given as mean Ϯ S.E. and hibernation increase glycogen stores in the brain (35). These findings suggest a role of astrocyte glycogen metabolism and neuronal activity. The known roles of astrocytes, not only in structural support and regulation of the external chemical environment of the neurons but also in their roles directly supporting neuronal maturation and neurogenesis, would be consistent with an ability to regulate glucose release.
Although glucose export from the brain has been demonstrated in children undergoing elective cardiopulmonary bypass surgery for congenital heart disease (23), and we have shown here that astrocytes express an active Glc-6-Pase-␤⅐Glc-6-PT complex, there is very limited evidence that confers a glucose export capacity on astrocytes per se. Under most conditions in vitro, lactate export predominates. Primary cultures of astrocytes convert their glycogen into extracellular lactate upon glucose deprivation (38), and glycogen-derived lactate supports intact rodent optic nerve function in vitro during aglycemia and increased metabolic demand (43). Moreover, neuronal activation leads to an increase in lactate that can be used as a metabolic substrate by neurons, suggesting that energy is transferred from astrocytes to axons in the form of lactate (reviewed in Refs. 44 and 45).
Even though the proposal that astrocytes have the ability to convert glycogen to glucose can explain the source of glucose reported to be released from the brain under certain conditions of cardiopulmonary bypass (23), we do not know under what conditions the astrocyte glycogen is cycled into glucose in vivo. There are many fates of Glc-6-P in the cell, depending on the physiological status of the cell and its environment. These include lactate and/or energy production via glycolysis, entry into the pentose phosphate pathway, and glucose production via the Glc-6-Pase-␤ complex in the ER. Because the level of glycogen storage in astrocytes is low compared with that in the liver and muscle, we do not consider glucose production to be a primary pathway in astrocytes but rather a pathway that may be used during periods of stress or metabolic restriction to buffer the brain function. The relative importance of the production and use of lactate and glucose in the brain remains a question for careful experimental scrutiny.
In conclusion, we have demonstrated that there is a functional Glc-6-Pase activity in the mouse brain that is present in astrocytes. The activity is independent of Glc-6-Pase-␣ and has the expected characteristics of Glc-6-Pase-␤. The activity can couple functionally with Glc-6-PT to produce an active Glc-6-Pase complex, and the complex can metabolize Glc-6-P to glucose and phosphate. These results can explain previous reports of the release of glucose from the brain and suggest that there is reason to think beyond simple lactate energy in the brain. The biological significance of this finding awaits more detailed studies.