These objections do not negate an important quantitative association between glucose utilization and glutamine formation (2), but the association could not be as simple as Magistretti et al. suggest. There is little doubt that transmitter glutamate in the brain is continuously synthesized from glucose in astrocytes, transferred to neurons (3), probably by way of glutamine (4, 5), and eventually degraded to CO₂ and water (6). Interruption of this metabolic exchange leads to an impairment of glutamatergic neurotransmission (7), a depletion of neuronal glutamate (4, 8) and an inability to consolidate memories (9).

The need for glycolytically derived energy to fuel glutamate uptake (10) is central to their thesis (1). This requirement, however, has been shown in astrocytes cultured with 25mM glucose; astrocytes that are cultured under physiological conditions readily use oxidative metabolism to fuel both glutamate uptake (11, 12) and glutamine formation (12, 13). Furthermore, the statement that glutamatergic neurons depend almost entirely on lactate from astrocytes as their source of energy is difficult to reconcile with the high density of glucose transporters found on these neurons (14).

We agree that neurons and glia do share the same table, but the menu is more varied, and the etiquette more subtle than envisaged Magistretti and his colleagues.

Leif Hertz
Department of Pharmacology,
University of Saskatchewan,
Saskatoon S7N 0W0, Canada
Stephen R. Robinson
Department of Psychology,
Monash University,
Clayton, Victoria 3168, Australia
E-mail: stephen.robinson{at}sci.monash.edu.au

REFERENCES

1. P. J. Magistretti, L. Pellerin, D. L. Rothman, R. G. Shulman, Science 283, 496 (1999) [Free Full Text] .
2. N. R. Sibson, et al., Proc. Natl. Acad. Sci. U.S.A. 95, 316 (1998) [Abstract/Free Full Text] .
3. L. Peng, et al., Dev. Neurosci. 15, 367 (1993) [Web of Science] [Medline] ; A. Lapidot and A. Gopher, J. Biol. Chem. 269, 27198 (1994) [Abstract/Free Full Text] ; T. Aureli, M. E. Di Cocco, M. Calvani, F. Conti, Brain Res. 765, 218 (1997) [CrossRef] [Web of Science] [Medline] ; S. R. Robinson et al., in Glial Contributions to Behaviour, P. Laming, Ed. (Cambridge Academic Press, Cambridge, 1998), pp. 63-82.
4. D. V. Pow and S. R. Robinson, Neuroscience 60, 355 (1996) .
5. B. Hassel, H. Bachelard, P. Jones, F. Fonnum, U. Sonnewald, J. Cereb. Blood Flow Metab. 7, 1230 (1997) .
6. U. Sonnewald, N. Westergaard, G. Unsgard, S. B. Petersen, A. Schousboe, J. Neurochem. 1, 1179 (1993) ; J. M. Pasqual, F. Carceller, J. M. Roda, S. Cerdan, Stroke 29, 1048 (1998) [Abstract/Free Full Text] .
7. J. D. Rothstein and B. Tabakoff, J. Neurochem. 43, 1438 (1984) [CrossRef] [Web of Science] [Medline] ; D. O. Keyser and T. C. Pelmar, Glia 10, 237 (1994) [CrossRef] [Web of Science] [Medline] .
8. J. L. Laake, T. A. Slyngstad, F.-M. S. Haug, O. P. Ottersen, J. Neurochem. 65, 871 (1995) [Web of Science] [Medline] .
9. L. Hertz, et al., Neurosci. Biobehav. Rev. 20, 537 (1996) [CrossRef] [Web of Science] [Medline] ; M. E. Gibbs, et al., Cognit. Brain Res. 4, 57 (1996) [CrossRef] [Medline] .
10. L. Pellerin and P. J. Magistretti, Proc. Natl. Acad. Sci. U.S.A. 91, 10625 (1994) [Abstract/Free Full Text] ; S. Takahashi, B. F. Driscoll, M. J. Law, L. Sokoloff, 92, 4616 (1995) .
11. L. Hertz, et al., Dev. Neurosci. 20, 339 (1998) [CrossRef] [Web of Science] [Medline] .
12. I. J. Bakken, L. R. White, G. Unsgard, J. Aasly, U. Sonnewald, J. Neurosci. Res. 51, 636 (1998) [CrossRef] [Web of Science] [Medline] .
13. U. Sonnewald, H. Akiho, K. Koshiya, A. Iwai, ibid., p. 103.
14. F. Maher, T. M. Davies-Hill, I. A. Simpson, Biochem. J. 315, 827, (1996); R. L. Leino, D. Z. Gerhart, A. M. van Buren, A. L. McCall, L. R. Drewes, J. Neurosci. Res. 49, 617 (1997) [CrossRef] [Web of Science] [Medline] ; S. J. Vannucci, et al., Dev. Neurosci. 20, 369 (1998) [CrossRef] [Web of Science] [Medline] .

Magistretti et al. (1) offer an intriguing explanation of the complex link between activation and metabolism in the brain. There are a number of perplexing issues, however, that they did not address.

First, neurons have efficient mechanisms for the uptake of glucose. Thus, it is not immediately apparent why astrocyte-derived lactate would be the preferred substrate of the stimulated neuron, or indeed why astrocytes completely out-compete neurons for glucose in the model suggested. Furthermore, the brain is predominantly an aerobic organ. As the trichloroacetic acid (TCA) cycle produces 34 ATP during complete oxidation of glucose for every 2 ATP produced by glycolysis, if the stoichiometry suggested by Magistretti et al. is correct, this would seem to preclude any oxidative metabolism by astrocytes involved in ion homeostatis. However, it is well documented that astrocytes can oxidize acetate, glutamate and glutamine in the TCA cycle (for example, 2). While it has been suggested that the process of glycolysis is coupled to Na⁺/K⁺ ATPase activity (3), I know of no direct evidence that the ATPase relies exclusively on glycolytic ATP and cannot use ATP generated in the mitochondria. Consequently, a small amount of oxidative metabolism may account for the energy demands of ion homeostatis.

The efficiency of glutamate uptake by astrocytes may be overemphasized in this model. Although astrocytes have a prolific ability at uptaking glutamate, neurons themselves have efficient mechanisms at uptaking glutamate both post- and pre-synaptically through the EAAC-1 carrier (4). Furthermore, the brain also exports lactate (5), and astrocytes can make use of glycogen as a substrate. This complicates any such coupling between lactate production and glutamate-glutamine cycling, and thus the link between metabolism and function.

So perhaps the biggest intrigue to this thesis may be why glutamine-glutamate cycling appears to be so tightly coupled to cerebral metabolism. The brain is still a mysterious box.

Julian L. Griffin
Department of Biomedical Sciences,
Imperial College,
South Kensington,
London SW7 2AZ, United Kingdom
E-mail: j.griffin{at}ic.ac.uk

REFERENCES

1. P. J. Magistretti, L. Pellerin, D. L. Rothman, R. G. Shulman, Science 283, 496 (1999) .
2. U. Sonnewald, N. Westergaard, S. B. Petersen, G. Unsgard, A. Schousboe, J. Neurochem. 61, 1179 (1993) [Web of Science] [Medline] .
3. R. W. Mercer and P. B. Dunham, J. Gen. Physiol. 78, 547 (1981) [Abstract/Free Full Text] .
4. R. W. Albers, G. J. Siegel, W. L. Stahl, in Basic Neurochemistry, G. J. Siegel, B. W. Agranoff, R. W. Albers, P. B. Molinoff Eds. (New York, Raven Press, ed. 5, 1994), pp. 66-68.
5. E. L. Gibbs, W. G. Lennox, L. F. Nims, F. A. Gibbs, J. Biol. Chem. 144, 325 (1942) [Free Full Text] .

Response: Hertz and Robinson agree with us about the importance of the quantitative association between glucose utilization and glutamine formation (1). They then suggest, however, that the observed stoichiometry and its explanation in terms of glutamate neurotransmitter cycling has led us to neglect other neurochemical reactions. That we did not mention, in our short Perspective, the energetic requirements of neurons releasing other neurotransmitters and the astrocytes, should not imply that we deny the well-established activities. Rather, on the basis of our measurements, in vivo glutamatergic systems account for the majority of energy and consumption. Within our present experimental accuracy, up to 15% of the increase in glucose oxidation induced by cortical activity may take place in other neurotransmitter systems and in glia cells. We hope to take advantage of the capability of in vivo NMR to measure other amino acid neurotransmitters in order to quantify the specific energetic requirements of these systems.

Similarly, the existence of neuronal glucose consumption is not denied by our model. The observed ratio between the rate of glucose oxidation and glutamate-glutamine neurotransmitter cycle is, however, consistent with the stoichiometry predicted by coupling glutamate uptake in the uptake of astrocyte to glucose.

Stimulation of aerobic glycolysis in brain slices by excitatory amino acids and their transportable analogs has long been known (for example, 2); thus, it is not an observation related to culture conditions. In addition, ultrastructural studies (3) have clearly indicated that the astrocytic profiles that ensheath synaptic contacts (thus, where glutamate reuptake occurs) do not contain mitochondria.

Herz and Robinson propose that the majority of glutamate used as a neurotransmitter is oxidized and must be immediately replaced by anaplerotic synthesis. Although glutamine oxidation may be significant in isolated astrocytes, direct ¹⁵N and ¹³C NMR measurements of astrocytic anaplerosis in vivo support the glutamate-glutamine cycle, in which transmitter glutamate is recycled as opposed to oxidized, as accounting for ~90% of glutamine synthesis in vivo. The glutamine synthesized by anaplerosis is consistent with the amount necessary for ammonia detoxification as established by arteriovenous-difference studies (4).

In summary, our results provide room for energy roles in vivo for several of the neurotransmitter systems. Processes that reflect the activity of glutamate release, including pre- and post-synaptic ion pumping and glutamate vesicularization, are the major energy-consuming pathway. The alternative metabolic pathways proposed by Herz and Robinson, derived from the study of cells, cannot be taken to disagree with our explanation until rates they have calculated are quantified in vivo.

We would like to thank Griffin for pointing out several issues that are important, but could not be addressed within the context of a short Perspective. Griffin points out quite correctly that the applicability of this model in vivo could be weakened by flux through several alternate pathways. These pathways certainly have a physiological role in the brain and are not excluded by the model proposed. The combined cellular and in vivo evidence, however, supports the thesis that neuronal glutamate release is coupled to glial cell glucose consumption, and that this is the major flux of cortical oxidative glucose consumption in vivo.

Because neurons express the glucose transporter GLUT3 and hexokinase, they take up and use glucose as an energy substrate. A number of facts, however, indicate that astrocytes could be a privileged site of glucose uptake. All capillaries are covered by astrocytic processes, the end-feet, which express the glucose transporter GLUT1 (5). Thus, glucose, in leaving the circulation to enter the brain parenchyma, is likely to first encounter an astrocytic membrane. The point made in our model is that glutamatergic synaptic activity stimulates glucose uptake into astrocytes. In the absence of this activity, some glucose is most likely taken up directly by neurons. Within the certainty of the present NMR data, between 15 and 30% of cortical neuronal glucose oxidation under basal conditions is the result of other processes.

A second point is that neurons do use lactate. Experiments performed in the superior cervical ganglion, in the retina, in hippocampal slices, or in cerebral cortical neurons have indicated that lactate may in fact represent a substrate preferred over glucose (see 6 for references; also 7). If the high rate of the glutamate-glutamine cycle in vivo is coupled to astrocytic nonoxidative glycolysis, as we propose, then the neurons will have sufficient lactate available to meet approximately 80% of their energetic requirements. The export of this high-lactate flux from the brain, as opposed to oxidation in the neurons, would be wasteful of glucose. The production of glucose by the liver from lactate is energetically costly. The notion that a significant reuptake of glutamate occurs in neurons does not take into account most recent data. Evidence obtained by immunohistochemistry of glutamate transporters at the electronmicroscopic level shows that neuronal glutamate transporters are mostly localized on the soma or on post-synaptic profiles, but not on pre-synaptic terminals. In contrast, the glial glutamate transporters are concentrated on the astrocytic processes that ensheath synapses (8). In addition, the coupling between glutamate release from terminals and reuptake by astrocytes is such that the Na⁺ current, which is associated with glutamate uptake measured in astrocytes, provides a faithful reflection of glutamatergic synaptic activity (9).

The relative role of glutamate-stimulated glucose utilization and glycogenolysis triggered by neurotransmitters such as VIP or noradrenaline is still a matter of study (6). Most likely, the two mechanisms are complementary, occurring in a temporal sequence in relation to synaptic activity. Griffin rightly points out that astrocytes, in addition to glycolysis, do have oxidative metabolism. This is certainly the case. We have observed, for example, that blocking oxidative phosphorylation in astrocytes with azide results in a 3-fold increase in lactate release. This means that there is approximately a 60% reserve of oxidative phosphorylation.

With regard to the nature of the energy-yielding pathway that fuels the activity of the pump, there is little doubt that glycolysis is important. Under "basal" conditions, there is indeed a significant contribution by oxidative phosphorylation (10); however, the point made by our model is that, during activation by glutamate, the additional energy needs for glutamate transport and glutamine synthesis are met essentially by increased glycolysis. An explanation for the coupling of glutamate transport to astrocytic glycolysis may be the requirement to have rapid bursts of ATP production to clear glutamate out of the synaptic cleft within a few milliseconds after vesicular release. Because glycolytic enzymes, which generate ATP, are in close proximity to the Na/K ATPase on the glial membrane, the ATP required by the pump may be generated without a transient drop in the local energy charge, which would occur if distant mitochondria were the main ATP source.

Finally, there are many possible pathways for glucose uptake and oxidation. The close to 1:1 ratio between increases in the rate of the glutamate-glutamine cycle and glucose oxidation is strong evidence that our model describes what is occurring in vivo under physiological conditions. The two ATP molecules generated by glycolysis of one glucose molecule are the precise number needed to fuel the energy requirements for astrocytic glutamate uptake and conversion to glutamine. While further experiments are necessary to test this proposed mechanism, it seems unlikely that this stoichiometric ratio between these processes is just a coincidence. While our results are within the accuracy of the NMR measurement, there is room for increases in other energetic processes not coupled to glutamate release; these would account for no more than 15% of the total increase in energy production.

Pierre J. Magistretti
Luc Pellerin
Robert G. Shulman
Douglas L. Rothman
Institut de Physiologie,
Université de Lausanne,
CH-1005 Lausanne, Switzerland
E-mail: pierre.magistretti{at}iphysiol.unil.ch

REFERENCES

1. P. J. Magistretti, L. Pellerin, D. L. Rothman, R. G. Shulman, Science 283, 496 (1999) .
2. A. Weil-Malherbe, Biochem. J. 32, 2251 (1938) .
3. A. Rohlmann and J. R. Wolff, in Gap Junctions in the Nervous System, D. C. Spray and R. Demierzel, Eds., (Landes, Austin, TX, 1996), pp. 175-192.
4. N. R. Sibson, et al., Proc. Natl. Acad. Sci. U.S.A. 95, 316 (1998) .
5. S. Morgello, et al., Glia 14, 43 (1995) [CrossRef] [Web of Science] [Medline] .
6. M. Tsacopoulos and P. J. Magistretti, J. Neurosci. 16, 877 (1996) [Free Full Text] .
7. A. Schurr, et al., J. Neurosci. 19, 34 (1999) [Abstract/Free Full Text] .
8. J. D. Rothstein, et al., Neuron 13, 713 (1994) [CrossRef] [Web of Science] [Medline] .
9. D. E. Bergles and C. E. Jahr, J. Neurosci. 18, 7709 (1998) [Abstract/Free Full Text] .
10. I. A. Silver and M. Erecinska, Glia 21, 35, 1997.