The Astrocyte-Neuron Lactate Shuttle: A Challenge of a Challenge

The Journal of Cerebral Blood Flow and Metabolism is to be applauded for opening up a discussion on the astrocyte-neuron lactate shuttle hypothesis (Chih and Roberts, 2003;Pellerin and Magistretti, 2003). It is also helpful to know that some of the prepositions made by Pellerin et al. (1998) and Magistretti et al. (1999) apparently have been deleted from the present version of the astrocyte-neuron lactate shuttle hypothesis (Pellerin and Magistretti, 2003), for example, that an activity-dependent aerobic glycolysis in astrocytes, triggered by glutamate uptake and glutamine synthesis, sets the pace for total energy metabolism in brain cortex (Magistretti et al., 1999). Accordingly, the present definition of the shuttle model has been modified as consisting of two, and only two, components. These are i) enhancement of aerobic glycolysis in astrocytes in response to neuronal activation of glutamatergic synapses, reflecting a dependency of astrocytic glutamate uptake upon glycolytically derived energy, and ii) oxidation in neurons of lactate produced by astrocytes.

Does astrocytic glutamate uptake depend upon aerobic glycolysis?

The observation by Pellerin and Magistretti (1994) that uptake of glutamate by astrocytes in primary cultures causes a distinct stimulation of deoxyglucose (DG) phosphorylation and lactate production was almost immediately confirmed by Takahashi et al. (1995) in the Sokoloff laboratory, but the magnitude of the response was much smaller. However, other investigators have found that glutamate either has no effect upon DG phosphorylation and glucose use in cultured astrocytes or that it causes a decrease rather than an increase, except under anoxic conditions, suggesting that glutamate normally may be oxidized as an alternative fuel (Dienel and Cruz, 2004;Hertz et al., 1998;Liao and Chen, 2003;Peng et al., 2001;Qu et al., 2001;Swanson et al., 1990). It is known that glutamate is oxidatively degraded not only in cultured astrocytes (Hertz and Hertz, 2003;McKenna et al., 1996;Yu et al., 1982, 1992), but also in brain slices and in the brain in situ (Taylor et al., 1996;Zielke et al., 1998). Moreover, it is well established that glutamate uptake into cultured astrocytes can be fueled by either glycolytically or oxidatively derived energy (Huang et al., 1993;Swanson and Benington, 1996).

A possible explanation of the discrepancy between the results obtained by the Pellerin-Magistretti-Sokoloff groups and those reporting lack of stimulation of astrocytic glycolysis by glutamate is that the cultures used by the former groups resort to accumulating glutamate by the aid of glycolytically derived energy because they are deficient in oxidative metabolism. Many types of cultured cells, including brain cells (Dittmann et al., 1973), tend to become glycolytic and show a reduced oxidative metabolism (Guminska et al., 1969;Langvad, 1970). The predominance of glycolysis in cultured cells appears to a large extent to depend upon the culturing technique (Felder et al., 2002;Gstraunthaler et al., 1999). It can be counteracted by improved culturing conditions, especially by facilitating oxygen diffusion to the tissue (Booher et al., 1971;Kondo et al., 1997) and by not using excessive glucose concentrations (such as the 25 mM glucose used by the Pellerin-Magistretti-Sokoloff groups) in the medium (Abe et al., 2003).

The Pellerin-Magistretti group has never provided any values for oxidative metabolism in their astrocyte preparations, but Itoh et al. (2003) have shown that the cultures used by the Sokoloff group have a low rate of oxidative metabolism of [U-¹⁴C]glucose (nominal rate 0.12 nmol/[min/mg protein]). This rate is five times lower than the respiratory rate observed by the same authors in cultured neurons. A five times lower respiratory rate in astrocytes than in neurons cannot reflect the in vivo situation because it has recently been demonstrated by three groups that astrocytes, which occupy less than 30% of the volume in the brain cortex (Pope, 1978;Williams et al., 1980;Wolff and Chao, 2004), account for approximately 15% of its oxidative metabolism (Blüml et al., 2002;Gruetter et al., 2001;Lebon et al., 2002). Also, substantially higher respiratory rates (approximately 1.0 nmol/[min/mg protein]) have been reported by Hertz and coworkers in both astrocytes and cerebellar granule neurons with [U-¹⁴C]glucose as the precursor (e.g., Peng et al., 1994;Yu and Hertz, 1983) and by Lopes-Cardozo et al. (1986), using [2-¹⁴C]glu-cose, and Vicario et al. (1993) found virtually identical respiratory rates in cultures of neurons and astrocytes. Accordingly, it cannot be concluded that cultured astrocytes depend upon glycolysis for glutamate uptake. However, this does not negate the possibility that there could be situations in the brain in vivo, where glutamate uptake is fueled by glycolytically derived energy (e.g., in the most peripheral parts of both neurons and astrocytes, which are too minute to contain mitochondria).

It has been used as a powerful argument in favor of the astrocyte-neuron lactate shuttle that interference with astrocytic glutamate uptake decreases glucose utilization during in vivo activation of the brain, including its barrel cortex, regardless whether the inhibition occurs by administration of the transport inhibitors beta-D,L-threohydroxyaspartate (THA) or pirrolidine-2-4-dicarboxylate (PDC) (Demestre et al., 1997), by injection of antisense mRNA to the transporter (Cholet et al., 2001) or by the use of mutant mice without one or the other of the two astrocytic glutamate transporters GLT-1 and GLAST (Voutsinos-Porche et al., 2003). However, these observations are not conclusive. One problem is that 10-day-old rats were used for these experiments. Oxidative metabolism of glucose is not fully developed in the 10-day-old rat or mouse brain. Ketone bodies, which use the monocarboxylic acid transporter (MCT) for uptake, are an important fuel for suckling rats (Cremer, 1982;Medina et al., 1999;Nehlig, 1999), and the rate of glutamate oxidation in cultured mouse astrocytes is lower in cells corresponding to postnatal day 10 than in more mature cells (Yager et al., 1994). It would, therefore, have been preferable to use adult animals in the study by Voutsinos-Porche et al. (2003), especially since the authors indicate that the mutant mice may mature at a reduced rate.

Another caveat against the conclusion that the reduced metabolic response to brain activation must be a reflection of a diminished workload by the astrocytes is provided by a multitude of experimental observations that many facets of glutamatergic activities are impaired when glutamate uptake is reduced (e.g., Maki et al., 1994;Niederberger et al., 2003; Turecek and Trussel, 2000). Inhibition of glutamate uptake might lead to reduction of presynaptic glutamate release, activation of inhibitory metabotropic glutamate receptors, postsynaptic desensitization, or disturbance of glutamate recycling. Excitatory postsynaptic currents in the rat barrel cortex are inhibited by glutamate uptake inhibitors, and neurons in the developing neocortex are particularly sensitive to glutamate transporter function (Kidd and Isaac, 2000). As indicated in the report by Voutsinos-Porche et al. (2003), several aspects of glutamatergic transmission are unaltered in the adult mutant mice used in their work. However, others are not. In the very same GLT-1^−/−mutant as that used by Voutsinos-Porche et al. (2003), long-term potentiation (LTP) induced by tetanic stimulation is reduced by two-thirds (Katagiri et al., 2001). Development of LTP in brain slices is known to enhance Na⁺,K⁺-ATPase activity (Glushchenko and Izvarina, 1997) and thus increases energy demand. It must therefore be concluded that neither studies in cultured astrocytes nor in vivo studies in animals with glutamate transporter deficiencies prove the dependence of astrocytic glutamate transport upon aerobic glycolysis.

How much lactate is produced in activated brain in vivo, and is there a substantial lactate flux through extracellular fluid?

An essential component of the lactate shuttle hypothesis is that there must be a large, directed transfer of lactate from astrocytes to neurons in vivo, but this has never been demonstrated. It is well established that in spite of sufficient oxygen availability, activation of brain leads to an increase in intracellular and extracellular lactate from a “resting” value of approximately 1.0 μmol/g wet wt. and 1 mM, respectively, to approximately twice these values, and that lactate rapidly returns to its previous level after the activation is terminated (Dienel et al., 2002;Korf; 1996;Prichard et al., 1991). However, it is uncertain whether this lactate simply represents a larger, static pool, that is, an increased level of tissue lactate arising from the need of a higher concentration of pyruvate to stimulate pyruvate dehydrogenase activity sufficiently for TCA cycle flux to match an increased rate of glycolysis. To demonstrate a large flux from astrocytes to neurons through the enlarged lactate pool requires determination of the cells that are the main lactate producers and the main lactate consumers in the brain as well as quantification of lactate production and degradation.

Which cell type is the predominant producer of lactate in the brain in vivo?

It has not been established with certainty from which cell type(s) lactate is released during brain activation in vivo nor by which mechanism(s) the resting level is reestablished. There is evidence that at least some of the released lactate originates from astrocytes. Thus micro-dialysis performed 1 to 2 weeks after the insertion of a microdialysis probe, when the probe is surrounded by an adhering scar of reactive glia, shows a stress-induced increase in extracellular lactate, which is similar to that seen 1 to 2 days after the insertion of the probe, when access of neuronally released compounds to the probe is indicated by a large K⁺-mediated transmitter release (Korf, 1996). This observation does not, however, exclude the possibility that neurons also release lactate. Because stimulation of glucose use in response to increased energy demand is fundamental to regulation of cellular energy metabolism (see, e.g., Chih and Roberts, 2003;Hertz and Dienel, 2002), and because neurons are well equipped with glucose transporters (Dwyer et al., 2002), it is difficult to envision that glycolysis in neurons should be unaffected when neuronal energy requirements are increased. Moreover, the glycolytic enzyme hexokinase has a high activity in synaptosomes (Wilson, 1972), glycolytic ATP formation has been demonstrated in post-synaptic densities (Wu et al., 1997), and in cultured glutamatergic neurons the rate of ¹⁴ CO₂ production from labeled glucose is as high as in astrocytes and distinctly increased during K⁺-mediated depolarization (Peng et al., 1994).

In cell cultures, both neurons and astrocytes release large amounts of lactate to the medium, although the release from astrocytes (25–35 nmol/min/mg protein) is two to three times larger than the neuronal release (5–11 nmol/min/mg protein) (Dienel and Hertz, 2001;Schousboe et al., 1997;Waagepetersen et al., 2000;Walz and Mukerji, 1988). Monocarboxylate transporter (MCT)-mediated facilitated diffusion, which is driven by concentration gradients, is responsible for lactate transport in and out of cells (Halestrap and Price, 1999;Juel, 2001), and a major reason for the high rates of membrane transport in the cultured cells is the large volume of extracellular medium, which initially contains no lactate. Because of the slow rise of the extracellular lactate concentration, a concentration gradient can be maintained between even a low intracellular lactate concentration and the extracellular lactate concentration, promoting continuous release of lactate. In the brain in vivo lactate release would be delayed by rapid increase in adjacent extracellular fluid. The rate of continued release of lactate would be determined by its removal by extra-cellular diffusion, uptake into adjacent cells with a lower concentration of lactate, and under a few pathologic conditions (Dienel and Cruz, 2003) lactate release to the circulation.

Lactate production in astrocytes occurs not only during the glycolytic part of glucose degradation, but also during breakdown of glycogen (Dringen et al., 1993), which is stimulated during brain activation and can be very fast (Dienel et al., 2002;Swanson, 1992;Swanson et al., 1992). Because glycogen and activity of its degrading enzyme, glycogen phosphorylase, within brain parenchyma is virtually restricted to astrocytes (Ibrahim, 1975;Richter et al., 1996), its breakdown products must be released from astrocytes or further metabolized in astrocytes. The content of glycogen in brain, and thus also activity-induced glycogenolysis, is larger than previously recognized (Cruz and Dienel, 2002;Kong et al., 2002), and during brain activation the rate of lactate equivalents produced from glycogen exceeds that of lactate production from glucose (Dienel and Cruz, 2003, 2004). Within the activated tissue formation of lactate from glycogen, which under generally used experimental conditions is unlabeled, should therefore be expected to greatly dilute the specific activity of lactate generated from [¹⁴C]labeled glucose. However, recent experiments have shown that the specific activity of lactate produced in the brain from the administered labeled glucose is not diluted by any nonlabeled lactate (Dienel et al., 2002). This remarkable observation demonstrates that the two pools of lactate (and pyruvate), one from glycogen and one from the [¹⁴C]glucose delivered by blood, are segregated. A segregation can be explained by cytosolic “channeling” of metabolites from one enzyme to the next in a pathway. Such a channeling is known in vascular smooth muscle, where lactate formed by glycolysis is segregated from that formed by glycogenolysis (Allen and Hardin, 2000;Lynch and Paul, 1983). The channeling may not be restricted to the individual astrocyte in which lactate was produced. Rather, glycogen-derived lactate appears to be transported away from the activated area, probably by gap-junction mediated transport to other cells in the astrocytic syncytium (Medina et al., 1999). A segregation of lactate obtained by glycolysis and glycogenolysis is virtually incompatible with release of astrocytic lactate from both glucose and glycogen to a presumably nonchanneled extracellular space (a precondition for an astrocyte-neuron lactate shuttle).

Does oxidation of lactate primarily occur in neurons?

The concept that glucose-derived lactate, released from astrocytes, is a preferred substrate for neurons is a key component of the shuttle model. It has been taken as support of this hypothesis that Itoh et al. (2003) and Bouzier-Sore et al. (2003, 2004) found that unlabeled lactate profoundly reduced the production of ¹⁴CO₂ from labeled glucose in hippocampal and cortical neurons, respectively. In contrast, there was little, if any, inhibition by unlabeled glucose of production of ¹⁴CO₂ from labeled lactate. The main reason for the inhibition of neuronal ¹⁴CO₂ production from labeled glucose by unlabeled lactate was postulated by Itoh et al. (2003) to be that conversion of lactate to pyruvate converts NAD⁺ to NADH + H⁺ and therefore reduces the availability of NAD⁺ for oxidation of glyceraldehyde-3-phosphate during glycolysis. However, if this mechanism were the major reason for the reduced ¹⁴ CO₂ production from [¹⁴C]glucose in the presence of lactate, one would expect that pyruvate, which does not consume NAD⁺ before its oxidation, would be a less efficient inhibitor of ¹⁴CO₂ production from [¹⁴C]glucose than lactate. In fact, the opposite is the case, because unpublished experiments by Peng and Hertz have shown that ¹⁴CO₂ production from 7.5 mM [¹⁴C]glucose in the glutamatergic cerebellar granule neurons is reduced from 1.3 ± 0.19 to 0.29 ± 0.02 nmol/min/mg protein (n = 6), that is, by 77.1%, in the presence of 5 mM pyruvate, but only by 47.1% in the presence of 5 mM lactate (pyruvate vs. lactate:P < 0.01).

Larger inhibition of release of labeled CO₂ by addition of unlabeled pyruvate than by addition of unlabeled lactate suggests that the dilution of the specific activity of [¹⁴C]glucose-derived pyruvate/lactate by extracellular lactate or pyruvate has a greater impact than NAD⁺ availability. The strong influence of exogenous substrates on dilution of specific activity in endogenous pyruvate is to be expected because brain pyruvate levels are low, that is, 50 to 100 μM (Siesjö, 1978). Flooding of cultured cells with large amounts of unlabeled lactate, facilitated by the high MCT activities, and a rapid and reversible interchange between lactate and pyruvate (Wolfe, 1990;Wolfe et al., 1988), overwhelm the intracellular metabolite levels. Isotope dilution of glucose-derived pyruvate/lactate will cause a large reduction of the production of labeled CO₂ from glucose, regardless of whether glucose metabolism is suppressed. That a potential impairment of glucose metabolism by lactate caused by competition for NAD⁺ plays at most a minor role is also suggested by the observation by Bliss and Sapolsky (2001) that DG phosphorylation in hippocam-pal neurons is not inhibited at physiologically relevant lactate levels and that even highly abnormal lactate/glucose ratios (e.g., 0.5 mM glucose and 5 mM lactate) have only a very modest inhibitory effect (approximately 15%).

The possibility of inhibition of neuronal and astrocytic lactate oxidation by glucose has been examined experimentally by using the reverse experimental situation, that is, measurement of production of ¹⁴CO₂ from [U-¹⁴C]lactate in the presence of different concentrations of glucose. Because glucose use is governed by product inhibition of key enzymes, especially phosphofructokinase (reviewed by Hertz and Dienel, 2002), little effect of elevated glucose level on lactate metabolism would be expected as long as the extracellular glucose concentration is maintained above the level necessary to secure sufficient glucose transport into the cell. This extracellular glucose level could be 2.5 mM or perhaps slightly higher because the K_m for the neuronal glucose transporter is 2 to 3 mM (Maher et al., 1996). Raising the concentration of glucose above this level would not increase delivery of carbon to downstream metabolic pools, which is consistent with the observation that hyperglycemia does not change glucose use rates in brain in vivo (Orzi et al., 1988). This reasoning explains a modest effect of unlabeled glucose, regardless of its concentration, upon ¹⁴ CO₂ production from 2 mM [¹⁴C]lactate in the experiments by Itoh et al. (2003). The larger inhibition of ¹⁴CO₂ production from [U-¹⁴C] lactate by glucose in astrocytes observed by these authors is to be expected because of the higher rate of glycolysis in the astrocytes, leading to a higher concentration of unlabeled glucose-derived pyruvate/lactate, which accordingly can compete more efficiently with exogenous lactate. In two studies by Bouzier-Sore et al. (2003, 2004), a dilution of specific activity of glucose-derived pyruvate by addition of nonlabeled lactate must in a similar manner invalidate an otherwise ingenious approach to quantitate metabolism of glucose, exogenous lactate and glucose-derived lactate.

It is also against the notion that lactate predominantly should be oxidized in neurons that Peng et al. (1994) observed similar rates of ¹⁴ CO₂ formation from [¹⁴C]lactate (5 mM) in cultured GABAergic cerebral cortical neurons, cerebellar granule neurons, and astrocytes. However, using “trapping” of label from either [¹³C]glucose or [¹³C]lactate into glutamate as a determination of tricarboxylic acid cycle activity, Waagepetersen et al. (1998) found that incorporation of label from [¹³C]lactate into glutamate in cultured cortical astrocytes was only 50% of that in cortical neurons. However, these experiments were performed with either 0.5 mM glucose or 1 mM lactate as the metabolic substrate, and stimulation of glycogenolysis in the aglycemic lactate medium may have caused a decrease in the specific activity of intracellular lactate. This type of experiment does accordingly also not provide solid evidence that lactate should be a preferred substrate for neurons.

Publications by Bouzier et al. (2000) and by Tyson et al. (2003) claim to show that lactate is primarily oxidized in neurons in vivo, based upon the finding that intravenous injection of [3-¹³C]lactate gives rise to identical incorporation of label into the C-2 and the C-3 position of glutamate and glutamine, rather than to selective incorporation into the C-2 position, indicative of pyruvate carboxylation. Based upon this observation, the authors conclude that lactate is metabolized in a compartment expressing no pyruvate carboxylase activity, that is, a neuronal, not an astrocytic compartment. However, the oxaloacetate produced by pyruvate carboxylation is known to readily equilibrate with the symmetrical fumarate, thereby giving rise to equal labeling of C-2 and C-3 in glutamate and glutamine. For example, previous studies from the same laboratory by Merle et al. (1996) concluded, based upon a mathematical modeling of [1-¹³C]glucose metabolism in cultured astrocytes that 39% of oxaloacetate synthesized by pyruvate carboxylation equilibrates with fumarate before it is condensed with acetyl coenzyme A to form citrate, from which glutamate is produced. Sonnewald et al. (1993) found an even more extensive, and possibly quantitative equilibration between oxaloacetate and fumarate. Moreover, pyruvate carboxylation from exogenous lactate in intact brain has been demonstrated by other authors (Qu et al., 2000), although it was less pronounced than when glucose was the substrate. Finally, demonstration of pyruvate carboxylation provides only a positive identification of one part of astrocytic metabolism because dehydrogenation via acetyl coenzyme A in the astrocytic compartment is not determined. In fact, in intact brain the calculated rate of pyruvate dehydrogenation in astrocytes exceeds that of pyruvate carboxylation by a factor of almost 2 (Gruetter et al., 2001).

Another observation in intact nervous tissue used to support the astrocyte-neuron lactate shuttle is that Schwann cells in a stimulated isolated vagus nerve preparation account for as much as 78% of the total deoxyglucose phosphorylation in the nerve (Vega et al., 2003). Pellerin and Magistretti (2003) compared this value with a calculated estimate by Attwell and Laughlin (2001) of glial expenditure of energy in cortical gray matter as 5% of total energy consumption (a value that must be too low, considering that astrocytes account for 15% of the oxygen consumption in brain), and they concluded that the most likely explanation for this apparent paradox is that the remaining 73% of phosphorylated glucose were transferred to the axon in the form of lactate. However, such a comparison between a peripheral nerve and cortical gray matter is not legitimate because experiments measuring rates of oxygen consumption in crayfish giant axons and in their ensheathing glial cells have led to the estimate that in a normally functioning axonglial cell system the glial sheath accounts for 90% of the oxygen consumption by the tissue (Hargittai and Lieberman, 1991). Provided this value also applies to mammalian peripheral nerves, the Schwann cells themselves are likely to oxidize all glucose they phosphorylate. That mammalian Schwann cells probably have a high rate of oxidative metabolism is indicated by an extremely high mitochondrial density in these cells (Rydmark et al., 1998). Moreover, an inhibitor of the MCT supposed to carry lactate from the Schwann cells to the axon had no effect upon the action potential or the distribution of radioactivity after exposure to labeled deoxyglucose in the experiments by Vega et al. (2003).

Along similar lines, mobilization of the glycogen contained in a rat optic nerve sustains axonal action potential in the absence of glucose (Wender et al., 2000). It was concluded that this may be due to lactate transport from glial cells to neurons (Brown et al., 2003), although approximately 80% of the released K⁺ is likely to be initially accumulated by the glial cells, mainly by active transport mechanisms (Ransom et al., 2000). Because of the narrowness of the periaxonal space and the low resting extracellular K⁺ concentration, the relative change in extracellular K⁺ during the action potential is much larger than the changes in intraaxonal ion concentrations. That accumulation of extracellular K⁺ during energy failure leads to axonal block long before changes in intra-axonal ion concentrations affect excitability was demonstrated by Shanes (1951), who showed that axonal conduction block in squid nerve during anoxia could be reversed by simple wash with anoxic artificial sea water.

The distribution of different isotypes of MCT's between astrocytes and neurons is also cited as being in favor of directed transport of lactate from astrocyte to neuron (Pellerin, 2003). However, because lactate transport across the cell membrane occurs by facilitated diffusion, its direction is determined by concentration gradients (not only of lactate, but also of H⁺, which is cotransported), not by transporter type (Halestrap and Price, 1999;Juel, 2001). Moreover, a K_m value of the mainly neuronal MCT-2 for lactate is approximately 0.7 mM, whereas those of the mainly astrocytic MTC-1 and MCT-4 are 3 to 5 mM or higher (Bergersen et al., 2001;Broer et al., 1997, 1999;Dimmer et al., 2000;Halestrap and Price, 1999); this means that they are more responsive to the increase in the lactate concentration occurring during brain activity (Hertz and Dienel, 2004).

What conclusions, if any, can be made regarding a lactate shuttle?

The purpose of this challenge of a challenge has been to critically evaluate the prepositions of the present version of the astrocyte-neuron lactate shuttle hypothesis: i) pronounced aerobic glycolysis, especially in astrocytes, ii) stimulation of aerobic glycolysis in astrocytes triggered by glutamate uptake, and iii) oxidation of lactate primarily in neurons. There is evidence that glycolysis combined with glycogenolysis may be more pronounced in astrocytes than in neurons, especially during brain activation. Nevertheless, neurons almost certainly carry out glycolysis and they are therefore likely to contribute substantially to aerobic glycolysis and lactate production in the brain in vivo. Glutamate uptake in astrocytes can be metabolically fueled by either glycolytically or oxidatively derived energy, but there is no good evidence that it relies upon glycolytically derived energy. Because there is net production of lactate in vivo, and in most cases no release to the circulation, it is likely that lactate is also degraded in the brain, although this process may be temporally dissociated from its production, as suggested by a mismatch between the increase in glucose use and in oxygen consumption during many types of brain activation (Fox and Raichle, 1986;Fox et al., 1988). The site of lactate use is also likely to be spatially separated from its site of production. There is, however, no solid indication that lactate oxidation should primarily occur in neurons, and there is very good evidence that it is not exclusively a neuronal process. The magnitude and direction(s) of any lactate flux(es) in the functioning brain constitute two major unknowns. If the magnitude is small, then the proposed shuttle process is of little functional significance. If it is large, then its direction(s) would be of major interest, but it could at least equally well occur within a glial syncytium as from astrocytes to neurons.