Brain Energy Metabolism

Pierre J. Magistretti, Luc Pellerin, and Jean-Luc Martin

—W. L. ANDRIEZEN (1)

Glucose is the obligatory energy substrate for brain and it is almost entirely oxidized to CO2 and H2O. This simple statement summarizes, with few exceptions, over four decades of careful studies of brain energy metabolism at the organ and regional levels, extensively reviewed elsewhere (e.g., 10, 60, 61). To reflect the focus of this book, and to include recent observations made in several laboratories including our own, we provide in this chapter a key for reinterpreting brain energy metabolism with a cellular perspective. This key relies primarily on the cytological relationships and chemical interactions among the various cell types of the brain. The view that emerges from this cellular and molecular analysis is a cell-specific sequence of processes that eventually leads to the almost complete oxidation by the brain of blood-borne glucose, which is in accordance with the introductory statement. The proposed model relies on already available data; it can further be tested experimentally, and it provides an explanation for some recent unexpected data obtained by positron emission tomography (PET) and functional magnetic resonance imaging (MRI) studies in humans (14, 40, 49).

Although the brain represents only 2% of the body weight, it receives 15% of the cardiac output, 20% of total body oxygen consumption, and 25% of total body glucose utilization. With a global blood flow of 57 ml/100 g·min, the brain extracts approximately 50% of oxygen and 10% of glucose from the arterial blood. Hence, the glucose utilization of the brain, as assessed by measuring the arterial–venous difference (22), is 31 mmol/100 g·min. Oxygen consumption is 160 mmol/100 g·min; because CO2 production is almost identical, the respiratory quotient (RQ) of the brain is nearly 1, indicating that carbohydrates are the substrates for oxidative metabolism (60). Given a theoretical stoichiometry of 6 mmol of oxygen consumed for each mmole of glucose, glucose utilization by the brain should in theory be 26.6 mmol/100 g·min. As indicated earlier, the measured glucose utilization is 31 mmol/100 g·min, indicating that an excess of 4.4 mmol/100 g·min of glucose follows other metabolic fates. Glucose can produce metabolic intermediates, such as lactate and pyruvate, which do not enter necessarily in the tricarboxylic acid cycle but rather can be released and removed by the circulation. Glucose can be incorporated into lipids, proteins, and glycogen, and it is also the precursor of certain neurotransmitters such as g-aminobutyric acid (GABA), glutamate, and acetylcholine (10, 60).

Numerous studies have been performed to identify molecules that could substitute for glucose as an alternative substrate for brain energy metabolism. Among the vast array of molecules tested, mannose is the only one that can sustain normal brain function in the absence of glucose (59). Mannose crosses the blood–brain barrier and in two enzymatic steps is converted to fructose-6-phosphate, a physiological intermediate of the glycolytic pathway. However, mannose is not normally present in the blood and cannot therefore be considered a physiological substrate for brain energy metabolism. Lactate and pyruvate can sustain synaptic activity in vitro (36, 55). Because of their limited permeability across the blood–brain barrier, they cannot substitute for plasma glucose to maintain brain function (43). However, if formed inside the brain parenchyma, they are useful metabolic substrates for neural cells (66). Under particular conditions, such as starvation, diabetes, or in breast-fed neonates, plasma levels of the ketone bodies acetoacetate and D-3-hydroxybutyrate increase markedly (41). Under these conditions, acetoacetate and D-3-hydroxybutyrate can be used by the brain as metabolic substrates (41).

As a corollary to these studies, steady-state arterial–venous (A–V) differences provide indirect evidence that a substance can be either used as a substrate by the brain (a positive A–V difference) or produced by the brain (a negative A–V difference) from a particular substrate such as glucose. Thus, in ketotic states, positive A–V differences have been measured for acetoacetate and D-3-hydroxybutyrate, indicating net utilization under these particular conditions. Net release of lactate and pyruvate (a negative A–V difference) is occasionally measured in normal individuals and more frequently in aged subjects or during convulsions (21).

Whole-organ studies, which allowed the determination of the substrate requirements for the brain, failed to provide the appropriate level of resolution to appreciate two major features of brain energy metabolism: (a) its regional heterogeneity and (b) its tight relationship with the functional activation of specific pathways. The autoradiographic 2-deoxyglucose method (2-DG) developed by Sokoloff and colleagues afforded a sensitive means to measure local rates of glucose utilization (LCMRglu) with a spatial resolution of approximately 50 to 100 mm (61). The method is based on the fact that tracer amounts of radioactive 2-DG are taken up by glucose transporters and phosphorylated by hexokinase with kinetics that are similar to those for glucose; however, unlike glucose-6-phosphate, 2-deoxyglucose-6-phosphate cannot be metabolized further and therefore accumulates intracellularly, thus providing, after appropriate corrections (61), an accurate measurement of the amount of glucose utilized. For a detailed description of the method, the reader is referred to the original articles by Sokoloff and colleagues (60, 61). Using this method, LCRMglu have been determined in virtually all structurally and functionally defined brain structures in various physiological and pathological states including sleep, seizures, and dehydration, and following a variety of pharmacological treatments (10). Furthermore, the increase in glucose utilization following activation of pathways subserving specific modalities, such as visual, auditory, olfactory, or somatosensory stimulations, as well as during motor activity, has been revealed in the pertinent brain structures (10).

Basal glucose utilization of the grey matter as determined by the 2-DG technique varies, depending on the brain structure, between 50 and 150 mmol/100 g wet weight·min in the rat (61). If a protein content of 10% of wet weight is assumed, a value of 5 to 15 nmol/mg protein·min is obtained. These values are approximately 50% lower in the primate brain (10). Physiological activation of specific pathways results in a 1.5 to 3-fold increase in lCMRglc as determined by the 2-DG technique (38).

With the advent of PET and the use of positron-emitting isotopes such as 18F, local glucose utilization has been studied in humans with 2-(18F)fluoro-2-deoxyglucose (52) (see Methodological Issues in the Neuropathology of Mental Illness). Similarly, local oxygen consumption and changes in blood flow can be studied in humans by PET using 15O2 and H215O (15, 53). Earlier studies had already demonstrated changes in local cerebral blood flow during activation of relevant brain regions by specific modalities (27).

In summary, changes in local brain energy metabolism can now be studied in humans with PET by monitoring alterations in glucose utilization, oxygen consumption, and blood flow during activation of specific areas. As we discuss below, recent studies in which these three parameters have been analyzed during activation of a given modality have yielded unexpected results, which suggest provocative hypotheses. Thus, an uncoupling between glucose uptake and oxygen consumption was observed during activation, since the increase in blood flow and in glucose utilization in the activated cortical area was not matched by an equivalent increase in oxygen consumption (13, 14). This observation raises the puzzling possibility that, at least during the early stages of activation, the increased energy demand is met by glycolysis rather than by oxidative phosphorylation.

Glycolysis (Embden-Meyerhof pathway) is the metabolism of glucose to pyruvate and lactate (Fig. 1). It results in the net production of only 2 mol of adenosine triphosphate (ATP)/mol of glucose as well as in the regeneration of reducing equivalents (the oxidized form of nicotinamide-adenine dinucleotide, NAD+) through the conversion of pyruvate into lactate. Alternatively, pyruvate can enter the tricarboxylic acid (TCA) cycle (or the Krebs cycle) and produce 30 mol of ATP/mol of glucose via the mitochondrial oxidative phosphorylation cascade (Fig. 2). The energetic value of oxidative phosphorylation over glycolysis is thus obvious. The only positive A–V differences consistently observed in the human brain are those of glucose and oxygen (except for ketotic states), and the respiratory quotient of the brain is virtually 1; therefore, the question of whether glycolysis or oxidative phosphorylation play a significant role in brain energy metabolism seems superfluous. As noted earlier, at most an excess of 5 mmol/100 g·min (i.e., 20% of total utilized glucose) is not oxidized completely to CO2 and H2O, and only a portion of it will yield pyruvate and lactate. Therefore, based on these whole-organ studies (i.e., A–V differences), at best less than 20% of glucose may eventually be utilized glycolytically. However, an array of recent data, obtained from studies both in vitro and in vivo raise the provocative question of a key role of glycolysis in brain energy metabolism. Consistent with the PET studies indicating an uncoupling between glucose uptake and oxygen consumption during activation (13, 14), rises in lactate have been monitored by 1H MRI spectroscopy in the primary visual cortex of humans following appropriate photic stimulation (49, 54). Lactate levels are increased in the rat somatosensory cortex following forepaw stimulation (68). When lactate was measured in vivo by microdialysis in freely moving rats, similar increases in hippocampus and striatum following somatosensory stimulation were demonstrated (11). Interestingly, the rate of lactate clearance from the extracellular space was markedly slowed in the presence of tetrodotoxin, a specific blocker of the neuronal voltage-sensitive sodium channels responsible for the generation of action potentials (11). This latter observation implies that during activation, lactate may normally be taken up by neurons as an energy fuel. It should be remembered that, after conversion to pyruvate, lactate can enter the TCA cycle with the potential to generate a total of 36 mol of ATP/mol of glucose (Fig. 2 ).

These in vivo data reveal a previously unrecognized prevalence of glycolysis over oxidative phosphorylation during activation. In fact, estimates of enzyme capacities indicate that glucose oxidation is already nearly maximal under basal conditions, implying that the activation-induced increases in energy demands are to be met primarily by glycolysis (69). One of the possible roles for activation-induced glycolysis may be to provide ATP to fuel energy-dependent ion transport, in particular the Na+/K+-ATPase, which represents the main energy-consuming process in neural cells (58). In fact, for the activity of the Na+/K+-ATPase, a preferential role of ATP derived from glycolysis has been recognized in various tissues (44, 50), including the brain (29). Other energy-consuming processes in the nervous system appear to preferentially use glycolytically derived ATP (51).

In summary, the analysis of brain energy metabolism at the regional level, afforded by the autoradiographic 2-DG method and by the development of PET-based analyses of glucose utilization, oxygen consumption, and blood flow, have clearly established a relationship between functional activity ("brain work") and energy metabolism. The PET and MRI studies in humans (13, 14, 40, 49, 54) have also revealed a previously unexpected role of glycolysis during activation of discrete and functionally defined areas, in the face of indisputable evidence from whole-brain steady-state A–V differences indicating that glucose is almost entirely oxidized to CO2 and H2O. How can these apparently opposite results be reconciled? As we discuss below, in vitro analyses of brain energy metabolism at the cellular level, and in particular of the flux of metabolic substrates between neurons and astrocytes, provide clues that may be useful in resolving this controversy.

Most of the information on energy metabolism at the cellular level has been obtained from cellularly homogeneous, purified preparations enriched in astrocytes, neurons, or vascular endothelial cells. Currently, most studies are conducted in primary cultures prepared from neonatal or embryonic rodent brain tissue. A cautionary note is always necessary when attempting to extrapolate results obtained in vitro to in vivo situations. It is, however, generally accepted that important insights may be gained from these cellularly homogeneous preparations.

Brain energy metabolism is often considered to reflect predominantly, if not exclusively, neuronal energy metabolism. However, it is now clear that other cell types, namely neuroglia and vascular endothelial cells not only consume energy but also can play an active role in the flux of energy substrates to neurons. First, there is a quantitative consideration; although it is arduous to provide a definitive ratio between neurons and nonneuronal cells given the variability in figures obtained in various species, brain areas, and developmental ages using methods that are not easily comparable, it is clear that neurons contribute at most 50% of cerebral cortical volume (23). In addition there is clear evidence indicating that the astrocyte-to-neuron ratio increases with increasing brain size (67); this is an important consideration when approaching the study of the cellular bases of brain energy metabolism in humans. It is therefore clear that glucose reaching the brain parenchyma through the circulation should provide energy substrates to a variety of cell types, only a portion of which are neurons.

Some structural relationships between astrocytes and other elements of nervous tissue are of particular relevance in this discussion of brain energy metabolism at the cellular level. Astrocyte processes are wrapped around synaptic contacts, whereas particular astrocytic profiles, the end-feet, surround intraparenchymal capillaries, and provide a cellular zone interposed between the bloodstream and other elements of the brain parenchyma (46) (Fig. 3). This latter structural feature has long been suggested as evidence indicating a role of astrocytes in the transit of substances from blood to other brain cells (1) (Fig. 4). A review of the physiological functions of astrocytes, which are only beginning to be elucidated, is beyond the scope of this article; however, their relationship to the structural features outlined above provides a background to the discussion of metabolic fluxes between cell types of the brain.

Two well-established functions of astrocytes are to maintain extracellular K+ homeostasis (3, 46 ) and to ensure the reuptake of neurotransmitters (17, 46). Neuronal activity results both in increases in extracellular K+ concentration and, at least at excitatory synapses, in augmented glutamate levels in the synaptic cleft (for the purpose of this discussion it should be borne in mind that glutamate is the major excitatory neurotransmitter in the brain, see ref. 12). One of the proposed mechanisms for the clearance of K+ from the extracellular milieu is by spatial buffering through the astrocytic syncytium (3). The potassium ion can also accumulate in astrocytes through inwardly rectifying K+ conductances (3). The activity of the Na+,K+-ATPase, which by hydrolyzing ATP to adenosine diphosphate (ADP) extrudes 3Na+ against 2K+, has also been shown to contribute to K+ homeostasis (24). In this latter case, maintaining K+ homeostasis is an energy-consuming process. Accordingly, 2-DG uptake into primary astrocyte cultures is markedly inhibited by ouabain, an inhibitor of the Na+,K+-ATPase (5).

On the other hand, recent results in our laboratory have indicated that glutamate also increases 2-DG uptake in cultured astrocytes (Table 1), with an EC50 of approximately 100 mM. This effect is blocked by the specific glutamate uptake inhibitor THA and by ouabain. This latter finding indicates that glutamate uptake into astrocytes, which occurs through a cotransport with Na+, results in the activation of Na+,K+-ATPase, probably by an increase in the intracellular concentration of Na+ (24). In fact, there appears to exist in astrocytes a large reserve of Na+,K+-ATPase activity, which can be stimulated by Na+ entry (24), resulting in a two- to threefold increase in 2-DG uptake (71). In the present context, it is important to note that the main mechanism that links energy metabolism and functional activity as determined by 2-DG uptake is represented by the activation of the Na+,K+-ATPase (35).

In vitro observations indicate (a) that astrocytes utilize glucose (5, 48) at a basal rate of 5 to 10 nmol/mg·min (5, 71, 72), a rate similar to that determined in whole-animal studies with the 2-DG autoradiographic method (61); (b) that two physiological functions of astrocytes linked to increased synaptic activity, that is, K+ and glutamate clearance from the extracellular space, markedly stimulate in a ouabain-sensitive manner 2-DG uptake; and finally (c) that the activation of Na+,K+-ATPase represents the coupling mechanism between the increase in glucose utilization and functional activity of the nervous tissue, raise the possibility that glucose utilization, as determined by PET in humans and by autoradiography in laboratory animals, reflects primarily the uptake of glucose by astrocytes rather than by neurons.

These considerations, based on in vitro studies and on the actual cytological relationships between astrocytic and neuronal processes, are in fact substantiated by results obtained in animal or human studies with the 2-DG method. It is now well established that the increases in 2-DG uptake linked to functional activation occur in the neuropil, that is, in regions that are enriched in axon terminals, dendrites and synapses ensheathed by astrocytic processes and not where neuronal perikarya are located (18). For example, when the sciatic nerve of anesthetized rats is stimulated, a frequency-dependent increase in 2-DG uptake occurs in the dorsal horn of the spinal cord (where afferent axon terminals make synaptic contacts with second order neurons) but not in the dorsal root ganglion, where the cell body of the sensory neurons is localized (18). As another example, increases in glucose utilization in the well-laminated monkey primary visual cortex elicited by appropriate visual stimuli are most pronounced in layer IV, which is poor in perikarya but in which the terminals of axons projecting from the lateral geniculate engage in synaptic contacts (20). In addition, activation studies of specific functional pathways using PET determination of cerebral blood flow indicate that the increases in energy demands occur in the projection areas, that is, where axon terminals are found (73).

The resolution of the 2-DG autoradiographic method does not allow us to determine whether the increase in 2-DG uptake occurs in axon terminals, dendrites, or the astrocytes that surround these elements. However, the observations on glucose utilization as determined by the 2-DG autoradiographic method, taken together with the fact that a rate of 2-DG uptake very similar to that observed in vivo can be demonstrated in pure astrocyte preparations is strongly suggestive of the fact that a large proportion of the glucose uptake that occurs during activation of modality-specific circuits is localized in astrocytes rather than in neurons. Other in vitro observations support this view. Reports have indicated that the rate of glucose uptake in cultured neurons is lower than in astrocytes (16, 30). In addition, glucose as the sole energy substrate cannot support neuronal survival in vitro (39, 56); other substrates such as pyruvate, glutamine, and lactate are needed (36, 39, 56).

To summarize, in vivo and in vitro data indicate that glucose utilization occurs at synaptic sites, not at neuronal perikarya, and that astrocytes are the likely cells where glucose uptake occurs during activation.

If glucose is taken up predominantly by astrocytes and if glucose alone cannot support the survival of neurons in vitro, then energy substrates other than glucose must be released by astrocytes. As indicated earlier, lactate and pyruvate are adequate substrates for brain tissue in vitro (36, 55, 66). In fact, synaptic activity can be maintained in cerebral cortical slices with only lactate or pyruvate as a substrate (36, 55). Thus a metabolic compartmentation whereby glucose taken up by astrocytes is metabolized glycolytically to lactate or pyruvate (Fig. 5), which are then released in the extracellular space to be utilized by neurons, is consistent with the available biochemical and electrophysiological observations. In particular, in vitro studies indicate that quantitatively lactate is the main metabolic intermediate released by astrocytes at a rate of 15 to 30 nmol/mg of protein·min (9, 70). This rate of release correlates well with the rate of glucose uptake by the grey matter (61) or by astrocytes in culture (71, 72), which is between 5 and 15 nmol/mg of protein·min. Other, quantitatively less important intermediates released by astrocytes are pyruvate (approximately 10 times less than lactate), a-ketoglutarate, citrate, and malate (56, 57, 62). Furthermore, fluxes of endogenous lactate between astrocytes and neurons have been quantified in vitro (26), and an avid lactate uptake has been demonstrated in neurons (8, 26). In addition, cytological evidence also supports this view. First, immunohistochemical data on the localization of the glucose transporter (GT) protein and of the pyruvate dehydrogenase (PDH) complex indicate that GT is primarily localized at the neuropil, that is, where axon terminals, dendrites, and the astrocytic processes that ensheath them are localized, whereas PDH [the enzyme that catalyzes the entry of pyruvate in dynamic equilibrium with lactate through the action of lactate dehydrogenase (Fig. 1) in the TCA cycle] is localized predominantly in the neuronal perikaryon (2, 37). In the adult rat cerebellum, an analogous differential distribution of key enzymes of energy metabolism exists (19). Thus, although Purkinje cells contain a high PDH activity, Bergman glia surrounding them is enriched in hexokinase, implying a high glucose phosphorylating activity (19).

The functional cytology of astrocytes provides a second set of arguments to support the existence of such metabolic compartmentation and trafficking. Astrocytes derive their name from the numerous processes that emerge from their cell-body conferring on them a star-shaped morphology. In addition, astrocytes are connected between them through gap junctions, thus giving rise to what is called the astrocytic syncytium (46). A critical role has been shown for such an astrocytic syncytium in the spatial buffering of K+ (3), whereby potassium ions flow through the syncytium from sites of high K+ extracellular concentration (Ke) brought about by increased neuronal activity, to sites where Ke is lower, thus maintaining Ke within the physiological range. The astrocytic syncytium appears to be equally suited to maintain energy metabolism homeostasis. Thus, specialized astrocytic processes, the end-feet surround intraparenchymal microvessels, which are the source of glucose; whereas other astrocytic processes surround the synapses, where the energy-consuming uptake of glutamate, and to a lesser extent of K+, coupled to synaptic activity occurs. Release of lactate and pyruvate is likely to occur at yet other sites, that is, at astrocytic processes that surround neuronal perikarya, where PDH, the enzyme complex for which pyruvate (or lactate through the action of LDH) is a substrate, is preferentially localized (2, 37) (Fig. 6).

These observations on the metabolic trafficking between astrocytes and neurons as well as on the cellular compartmentalization of enzymes that regulate glucose uptake and glycolysis versus oxidative phosphorylation support the fact that glucose, taken up by astrocytic processes is metabolized glycolytically to lactate and pyruvate, which are then released as substrates for oxidative phosphorylation in neurons (Fig. 5).

Local energy metabolism, in particular blood flow, has been viewed until recently as being regulated by products that accumulate following neuronal activity, such as adenosine and K+, as well by changes in extracellular pH and pCO2 (25, 45). This mechanistic view of events implies a post-hoc regulation of energy metabolism, that is, adenosine, K+, pCO2, and pH are the coupling factors between neuronal activity and blood flow (25, 45). One of the lines of research of our laboratory during the last 10 years has been aimed at testing the hypothesis that energy metabolism may be controlled by specific neurotransmitter systems that function in parallel with, or possibly even in anticipation of, those that act by releasing fast-acting neurotransmitters, such as glutamate and GABA, and convey rapid, point-to-point information within the brain. Experimental evidence indicates that the neuronal systems that contain norepinephrine (NE) and vasoactive intestinal peptide (VIP) may play such a homeostatic function within the cerebral cortex.

Within the brain, glycogen is primarily stored in astrocytes, although ependymal and choroid plexus cells, as well as certain large neurons in the brainstem contain the polysaccharide (see ref. 34 for review). Glycogen levels in brain are low compared to liver and muscle; however, the glycogen turnover rate is very rapid; its synthesis and breakdown are regulated by the two key enzymes glycogen phosphorylase and synthase (34). Glycogen levels are tightly coupled to synaptic activity as illustrated by the fact that during anesthesia they rise sharply (47); furthermore, reactive astrocytes, which develop in areas in which neuronal activity is decreased or absent as a consequence of injury, contain high amounts of glycogen (34 and the refs. therein). Approximately a decade ago, we showed that VIP, a then recently discovered peptide contained in a homogeneous population of bipolar, radially oriented neurons (33), could promote a cyclic adenosine monophosphate (cAMP) dependent glycogenolysis in mouse cerebral cortical slices (31). In view of the morphology and arborization pattern of VIP-containing neurons (Fig.7), we proposed that these cells could regulate the availability of energy substrates locally, within cortical columns (31, 33). A similar effect had been previously described for NE, serotonin, and histamine (34). The noradrenergic system is organized according to principles strikingly different from those of VIP neurons: The cell bodies of NE-containing neurons are localized in the locus coeruleus in the brainstem from which axons project to various brain areas including the cerebral cortex. Here, they enter the rostral end and progress caudally with a predominantly horizontal trajectory, across a vast rostrocaudal expanse of cortex (33). Given these morphological features, we suggested that, in contrast to VIP-containing intracortical neurons, the noradrenergic system could regulate energy homeostasis globally, spanning across functionally distinct cortical areas (31, 33) (Fig.7).

The glycogenolytic effect of VIP and NE is exerted in astrocytes, as indicated by studies in primary astrocyte cultures (32, 63) as well as by the fact that glycogen is primarily localized in this cell type (46). Thus VIP and NE promote a concentration- and time-dependent glycogenolysis in astrocytes, with EC50 of 3 and 20 nM, respectively (32, 63). The effect of NE is mediated by both b and a1 receptors. The initial rate of glycogenolysis is between 5 and 10 nmol/mg protein·min (63), a value that is remarkably close to glucose utilization of the grey matter as determined by the 2-DG autoradiographic method (61). This correlation indicates that the glycosyl units mobilized in response to the two glycogenolytic neurotransmitters can provide quantitatively adequate substrates for the energy demands of the brain parenchyma.

Another action of NE on energy metabolism is the marked stimulation of 2-DG uptake in primary astrocyte cultures (72). This action is functionally coordinated with glycogenolysis, because the same extracellular signal (NE) results in an increased availability of glycosyl units for ATP production in astrocytes. In contrast to NE, VIP does not influence glucose uptake by astrocytes (72).

Glycogenolysis, revealed by a newly developed autoradiography technique for glycogen, has also been demonstrated in vivo following physiological activation of a modality-specific pathway (64). Thus, repeated stimulation of the vibrissae resulted in a marked decrease in the density of glycogen-associated autoradiographic grains in the somatosensory cortex of rats (barrel field) as well as in the relevant thalamic nuclei (64). These observations indicate that the physiological activation of specific neuronal circuits results in the mobilization of glial glycogen stores.

As noted earlier, systemic administration of general anesthetics increases brain glycogen levels (47). Interestingly however, they do not increase the glycogen content of cultures exclusively containing astrocytes (65); this observation indicates that the in vivo action of general anesthetics on astrocytic glycogen is due to the inhibition of neuronal activity, further stressing the existence of a tight coupling between neuronal activity and astrocytic glycogen.

In view of the foregoing, it appears that astrocytic glycogen represents a metabolic buffer under the dynamic control of neuronal activity, which can be mobilized in the early stages of activation. Evidence supporting such a role of astrocytic glycogen has been provided by the hippocampal slice preparation. Electrical stimulation of the slice resulted in an immediate and marked increase in NADH fluorescence, an index for the activation of glycolysis (28). This increase in NADH fluorescence was observed in a well-oxygenated medium containing adequate supplies of glucose and occurred at the onset of synaptic activity. However, the signal disappeared when the glycogen content of the slices was depleted by dibutyryl cAMP (28). This observation further suggests that an activation of glycogenolysis occurs at the onset of synaptic activity.

What is the metabolic fate of the glycosyl units mobilized from glycogen? As noted earlier, lactate is the main metabolic intermediate released by astrocytes (9, 70). No glucose is released from astrocyte cultures, even when glucose is absent from the medium (9), consistent with the view that brain glucose-6-phosphatase activity is very low or not measurable (61). Blockade of oxidative phosphorylation by azide or cyanide doubles the release of lactate by astrocytes (9) indicating that part of the glycosyl units mobilized from glycogen are oxidized by astrocytes rather than being exported as lactate.

The foregoing observations support the notion that energy metabolism at the cellular level is regulated by specific neurotransmitter systems, such as those containing VIP and NE. These neurotransmitter systems could be viewed as the CNS counterparts of the autonomic nervous system that maintains cellular homeostasis in peripheral tissues. At this stage, it appears that astrocytes are the preferred targets for the homeostatic functions exerted by VIP and NE. However, immunohistochemical and more recently ultrastructural evidence provides strong support for the existence of interactions between various neurotransmitter systems, including those containing VIP, NE, serotonin, and acetylcholine, with endothelial cells of intraparenchymal microvessels in the rodent cerebral cortex (7, 10). In further support of the existence of interactions between certain neurotransmitter systems and vascular endothelial cells, specific recognition sites for a variety of neurotransmitters including VIP, NE, serotonin, and acetylcholine, have been demonstrated in various preparations of intraparenchymal microvessels (42). In addition, most of these recognition sites represent functional receptors, because they are coupled to conventional signal transduction pathways (42).

There is little doubt that in the coming years the interactions between neurotransmitters and nonneuronal cells of the brain will provide new insights into brain functions and open fertile areas for pharmacological developments.

The preceding sections of this chapter have attempted to provide an overview of brain energy metabolism at the organ, regional, and cellular levels. The claim is not to have been exhaustive, rather the challenge has been to try to provide an integrated perspective for well-accepted experimental evidence spanning from in vivo studies to cellular and molecular analyses. As a means to provide a synthesis of the features that we consider salient and conceptually novel, we should look at the fate of a molecule of glucose from blood to neurons (Fig. 5).

Glucose is avidly taken up by astrocytes in vitro, at a rate that is similar to brain glucose utilization as determined with the 2-DG autoradiographic method. In vivo, the activation-induced increase in glucose uptake is visualized in the neuropil, that is, where synapses ensheathed by astrocytes are present, and not at the level of the neuronal perikarya. Furthermore, astrocyte end-feet surround intraparenchymal blood vessels. From this and other evidence previously reviewed, it can be inferred that most of the activation-induced glucose uptake in the brain parenchyma, notably as visualized by the PET (18F)2-DG technique, occurs in astrocytes. In vivo as well as in vitro studies indicate that the physiological stimulation of a given brain region triggers a rapid activation of glycogenolysis (proven to be exclusively astrocytic) and glycolysis, which in turn result in the release of lactate.

Lactate, which has the potential of providing 36 ATP molecules through oxidative phosphorylation can be taken up and oxidized by neurons (actually 1 mol of lactate yields 18 mol of ATP; however, since 1 mol of glucose provides 2 mol of lactate, the overall ATP flux from glucose-derived lactate is 36 mol). Supporting this view is the fact, among other evidence, that synaptic activity in vitro can be maintained when lactate is the only metabolic substrate present. The bookkeeping of energy metabolism provided by the A–V differences showing complete oxidation of glucose is consistent with this view: Glucose enters the brain from the arterial side and is predominantly taken up by astrocytes, which then transform it to lactate. Lactate exchange occurs with neurons, which oxidize it to CO2 and H2O drained by the venous blood. The increase in lactate levels measured in vivo by 1H MRI spectroscopy or by microdialysis as well as the uncoupling between oxygen consumption and glucose uptake revealed by PET at the early stages of activation, can be interpreted in light of the proposed cellular compartmentation of lactate fluxes between astrocytes and neurons. Thus, because phosphofructokinase (PFK) activity is one of the rate-limiting steps for glycolysis (Fig. 1 ), activation-induced glycolysis implies a disinhibition of PFK and an increase in the levels of the metabolic intermediates that are downstream to it, notably lactate. A delay in the entry of lactate in the TCA cycle is likely to occur, because it takes place in another cell type (neurons), resulting in a temporary overflow of lactate linked to an absence of increase in oxygen consumption (uncoupling).

Several questions remain open to further investigation. To list a few: What are the molecular mechanisms of the coupling between neuronal activation and astrocytic glycolysis? Because ATP inhibits PFK activity, a decrease in the energy charge of astrocytes following activation is a likely possibility. What is the relative role of other metabolic substrates (such as citrate, a-ketoglutarate, or malate) that have been shown to be shuttled from astrocytes to neurons? Glycolysis and oxidative phosphorylation are not strictly compartmentalized between astrocytes and neurons, respectively. Clearly, some glucose oxidation occurs in astrocytes, and moderate release of lactate can be demonstrated from cultured neurons; mechanisms that regulate the relative activity of these two metabolic pathways in neurons and astrocytes are likely to exist and still need to be elucidated. The flux of metabolic substrates within the astrocytic syncytium provides yet another level of regulation to be studied (Fig.6 ).

One can hope that, in addition to offering an accurate view of physiological brain energy metabolism, further elucidation of the cell-specific molecular mechanisms of energy metabolism regulation will provide useful clues to probe, with cellular and molecular specificity, the physiopathological processes that underlie the expression of certain neurological and psychiatric disorders.

Research in the laboratory of PJM is supported by a grant from Fonds National Suisse de la Recherche Scientifique (31-26427.89). The authors are grateful to Ms. M. Emch for expert secretarial help.

P. J. Magistretti, L. Pellerin, and J.-L. Martin: Institut de Physiologie, Faculté de Médecine, Université de Lausanne, CH 1005 Lausanne, Switzerland

published 2000