Milestone Review: Metabolic dynamics of glutamate and GABA mediated neurotransmission — The essential roles of astrocytes

Since it was first generally accepted that the two amino acids glutamate and GABA act as principal neurotransmitters, several landmark discoveries relating to this function have been uncovered. Synaptic homeostasis of these two transmitters involves several cell types working in close collaboration and is facilitated by specialized cellular processes. Notably, glutamate and GABA are extensively recycled between neurons and astrocytes in a process known as the glutamate/GABA‐glutamine cycle, which is essential to maintain synaptic transmission. The glutamate/GABA‐glutamine cycle is intimately coupled to cellular energy metabolism and relies on the metabolic function of both neurons and astrocytes. Importantly, astrocytes display unique metabolic features allowing extensive metabolite release, hereby providing metabolic support for neurons. Furthermore, astrocytes undergo complex metabolic adaptations in response to injury and pathology, which may greatly affect the glutamate/GABA‐glutamine cycle and synaptic transmission during disease. In this Milestone Review we outline major discoveries in relation to synaptic balancing of glutamate and GABA signaling, including cellular uptake, metabolism, and recycling. We provide a special focus on how astrocyte function and metabolism contribute to sustain neuronal transmission through metabolite transfer. Recent advances on cellular glutamate and GABA homeostasis are reviewed in the context of brain pathology, including glutamate toxicity and neurodegeneration. Finally, we consider how pathological astrocyte metabolism may serve as a potential target of metabolic intervention. Integrating the multitude of fine‐tuned cellular processes supporting neurotransmitter recycling, will aid the next generation of major discoveries on brain glutamate and GABA homeostasis.


| INTRODUC TI ON
Precise regulation of synaptic neurotransmission is paramount for normal brain function. In the mammalian brain, glutamate and γ-aminobutyric acid (GABA) are principal neurotransmitters. While glutamate is present in high concentrations in all tissues, including the brain, GABA is restricted to the brain and some peripheral organs including the pancreas and gut (Agrawal et al., 1966;Okada et al., 1976;Roberts & Frankel, 1950;Tanaka, 1985). Due to the high cerebral concentrations, it was difficult to obtain acceptance for a neurotransmitter role of glutamate and GABA. However, detailed electrophysiological studies finally allowed the conclusion that glutamate and GABA serve as the primary excitatory and inhibitory neurotransmitters, respectively Curtis & Johnston, 1974;Krnjević, 1970;Krnjevic & Phillis, 1963). Altogether, these studies not only firmly established glutamate and GABA as central neurotransmitters but also provided the subsequent framework for understanding the roles and interactions of neurons and astrocytes in relation to neurotransmitter homeostasis, which will be the overarching topic of this review.
Synaptic function relies on a tightly regulated collaboration between neurons and glial cells (Allen & Eroglu, 2017;Semyanov & Verkhratsky, 2021). During synaptic transmission, presynaptic neurons release glutamate and GABA into the synapse, allowing binding to post-synaptic receptors hereby propagating signals. Efficient removal of synaptic glutamate and GABA is needed to obtain rapid signal transmission with high fidelity, and in the case of glutamate, to avoid harmful excitatory overactivation, known as excitotoxicity (discussed further in Section 5.2).
Astrocytes, which are the primary homeostatic glial cell of the brain (Verkhratsky & Nedergaard, 2018), are intimately associated with the synapse and take up large fractions of released glutamate and GABA (Zhou & Danbolt, 2013). However, neurotransmission is not simply governed by synaptic release and uptake of glutamate and GABA, but is also closely linked to cellular energy metabolism (Dienel, 2019;Schousboe et al., 2013). Stimulation of both glutamate (Rae et al., 2005(Rae et al., , 2006 and GABA (Nasrallah et al., , 2009) receptors generates profound metabolic effects and restoration of ion gradients related to neurotransmission accounts for the majority of the entire brain energy expenditure (Attwell & Laughlin, 2001;Yu et al., 2018). Furthermore, both glutamate and GABA are oxidized in the tricarboxylic (TCA) cycle to support neuronal and astrocytic energy metabolism. Indeed, glutamate serves as a metabolic hub ( Figure 1), connecting multiple metabolic pathways in the brain (Schousboe et al., 2014), and importantly bridges brain amino acid and carbohydrate metabolism through the TCA cycle intermediate α-ketoglutarate. To account for the loss of neuronal glutamate and GABA, caused by astrocyte uptake and cellular metabolism, astrocytes provide neurons with glutamine. Glutamine is a highly abundant cerebral amino acid and astrocyte glutamine support is imperative for brain function (Albrecht et al., 2007;. Cerebral glutamine synthesis and metabolism are catalyzed by the essential enzymes glutamine synthetase (GS) and phosphate-activated glutaminase (PAG), respectively ( Figure 1).

ANDERSEN and SCHOUSBOE
The transcellular transport of neuronal glutamate and GABA and astrocytic glutamine constitutes the basis of neurotransmitter recycling, known collectively as the glutamate/GABA-glutamine cycle ( Figure 2) (Bak et al., 2006;Hertz, 1979). The cycle is crucial to sustain neurotransmission and is a highly active metabolic flux (Shen et al., 1999), which is directly proportional to brain glucose oxidation (Sibson et al., 1998). The glutamate/GABA-glutamine cycle integrates multiple cellular processes, including metabolite release, uptake, synthesis, and metabolism in both neurons and astrocytes, all of which are essential to maintain brain glutamate and GABA homeostasis Schousboe et al., 2013).
The aim of this Milestone Review is to narrate both historical perspectives and recent advances of glutamate and GABA metabolism and recycling with emphasis on astrocytic involvement. First, we cover important features of cellular glutamate and GABA uptake and metabolism. We then review how specialized metabolic features of astrocytes allow sustained neurotransmitter recycling and metabolite transfer. Finally, we discuss these aspects in the context of excitotoxicity and neurodegeneration, ending with a focus on astrocyte metabolism as a potential therapeutic target.

| Glutamate transport
An essential feature concerning the identification of a neurotransmitter is the existence of a synaptic inactivation mechanism. Since no extracellular enzymes exist capable of degrading glutamate and GABA, the only alternative is efficient transport systems facilitating synaptic uptake. Such transport systems were first identified for glutamate in brain slices and synaptosomes (Balcar & Johnston, 1972a; F I G U R E 2 The glutamate/GABA-glutamine cycle connects neurotransmitter homeostasis and cellular energy metabolism. Glutamate, GABA, and glutamine are extensively recycled between neurons and astrocytes, a process which is coordinated and regulated by cellspecific expression of transporters and enzymes. The majority of glutamate, and a substantial fraction of GABA, released from neurons is recovered from the synapse by uptake into astrocytes. Astrocytes express the glutamate transporters GLT-1 and GLAST and the GABA transporters GAT1 and GAT3, whereas glutamatergic and GABAergic neurons express GLT-1 and GAT1, respectively. In the astrocyte, glutamate and GABA are metabolized, which supports the synthesis of the non-neuroactive amino acid glutamine, which is subsequently released. The astrocyte-derived glutamine is taken up by neurons and is converted into glutamate by phosphate-activated glutaminase (PAG) to replenish the neurotransmitter pools. Glutamine transfer is primarily mediated by the sodium-coupled neutral amino acid transporters (SNATs). Astrocytes express SNAT3 and SNAT5, whereas SNAT1, SNAT2, SNAT7, and SNAT8 are neuronal glutamine transporters. The recycling of glutamate, GABA and glutamine is known as the GABA-glutamine cycle (left) and the glutamate-glutamine cycle (right), which is also collectively called the glutamate/GABA-glutamine cycle. Neurotransmitter recycling is closely connected to energy metabolism, as glutamate, GABA, and glutamine all can undergo oxidation in the TCA cycle and hereby support energy production in both neurons and astrocytes. Furthermore, the TCA cycle intermediate α-ketoglutarate is the obligatory precursor of glutamate synthesis (Figure 1). Abbreviations not explained above: AAT: aspartate aminotransferase; GAD: glutamate decarboxylase; GDH: glutamate dehydrogenase; GABA-T: GABA transaminase; SSADH: succinic semialdehyde dehydrogenase. Balcar & Johnston, 1972b;Bennett et al. 1972;Levi & Raiteri, 1973b;Levi & Raiteri, 1973a). At this early stage, only a few studies had been performed using preparations of either bulkprepared glial cells or astrocytoma cell lines, being indicative of glutamate transport in astrocytes (Balcar et al., 1977;Faivre-Bauman et al., 1974;Hamberger, 1971;Henn et al., 1974). However, the existence of high-affinity glutamate transport in astrocytes was firmly established by studies using astrocytes in primary culture .
These studies concluded that astrocytic glutamate uptake was much more efficient than the corresponding uptake in neurons (Hertz & Schousboe, 1987), which recent studies on cloned glutamate transporters have clearly confirmed (Danbolt, 2001).

| GABA transport
In the case of GABA, a number of early studies demonstrated the presence of high-affinity transport in brain slices and synaptosomes, as well as in bulk-prepared neurons (Henn & Hamberger, 1971;Levi & Raiteri, 1973b;Schousboe, 1981). Moreover, different preparations of neurons in primary culture had provided evidence that particularly neurons of the cerebral cortex displayed efficient transport systems for GABA uptake (Larsson et al., 1981). In addition, it was also established that glioma cells, bulk-prepared glial cells, as well as astrocytes in primary culture exhibited high-affinity transport for GABA (Henn & Hamberger, 1971;Hertz, Wu, & Schousboe, 1978;Schrier & Thompson, 1974). Hence, in contrast to glutamate, bulk synaptic GABA uptake is divided between neurons and astrocytes. Based on these early cellular studies it was estimated that astrocytes accounted for 10-20% of synaptic GABA uptake, while the remainder was accumulated in presynaptic nerve endings (Hertz & Schousboe, 1987;Schousboe, 1981).
To date, four GABA transporters have been cloned: GAT1 (SLC6A1), GAT2 (SLC6A13), GAT3 (SLC6A11), and BGT1 (SLC6A12) (Borden et al., 1992;Guastella et al., 1990;Liu et al., 1993;Lopez-Corcuera et al., 1992). GAT1 and GAT3 are expressed throughout the brain (Durkin et al., 1995;Minelli et al., 1995Minelli et al., , 1996 and display the highest affinities for GABA uptake (Liu et al., 1993). Cerebral expression of GAT2 and BGT1 is low and is restricted to the leptomeninges and blood vessels (Zhou, Holmseth, Guo, et al., 2012;Zhou, Holmseth, Hua, et al., 2012), suggesting a negligible role of these transporters in synaptic GABA clearance. Early studies on cellular GAT distribution reported a preferential neuronal expression of GAT1, whereas GAT3 was located in astrocytes (Durkin et al., 1995;Minelli et al., 1995Minelli et al., , 1996. However, recent studies have uncovered widespread GAT1 expression in astrocytes Melone et al., 2015), oligodendrocytes (Fattorini et al., 2017) and microglia (Fattorini, Catalano, et al., 2020). These observations suggest that, in addition to GAT3 activity, glial GAT1-mediated GABA uptake may contribute significantly to synaptic GABA clearance and the extent of astrocyte GABA clearance may therefore have been underestimated (Andersen, Schousboe, & Wellendorph, 2023). Curiously, in the thalamus, both GAT1 and GAT3 are solely located in astrocytes (De Biasi et al., 1998), suggesting a different cellular mechanism of maintaining GABA homeostasis in this brain region. Loss of GAT1 function by genetic deletion in mice, leads to altered behavior, but does not affect longevity (Chiu et al., 2005). In contrast, no viable model of GAT3 deletion has yet been reported (MouseGenomeInformatics, 2005;Zhou & Danbolt, 2013), which may suggest that astrocyte GAT3 function is indispensable. Note that, not only transporters of glutamate and GABA but also of glutamine are essential to sustain neurotransmitter recycling between neurons and astrocytes (
Glutamate is closely linked to cellular energy metabolism via the TCA cycle intermediate α-ketoglutarate facilitated by multiple enzymes (Figure 1). Several of these reactions are catalyzed by transaminases and are thus reversible, making glutamate synthesis and metabolism a highly dynamic process. The transamination reaction of GABA-T is also in principle reversible. However, the high affinity of SSADH (Cash et al., 1978) and the temporal association between GABA-T and SSADH in a so-called metabolon within the mitochondria (Hearl & Churchich, 1984), likely facilitate efficient conversion of GABA to succinate for oxidation of the GABA carbon skeleton in the TCA cycle ( Figure 1). Metabolon formation allows channeling of substrates between enzymes, hereby increasing the pathway efficiency, which has been demonstrated for several pathways including the TCA cycle (Srere, 1987;Wu & Minteer, 2015).
Given the plethora of enzymes involved in glutamate and GABA homeostasis, multiple studies were initiated to investigate the cellular activities of these key enzymes in cultured astrocytes, as well as in cultured glutamatergic and GABAergic neurons. Collectively, these studies revealed highly compartmentalized activities of GS and GAD, being located in astrocytes and GABAergic neurons, respectively, whereas activity of GDH, AAT, PAG, and GABA-T were found in both cultures of astrocytes and neurons ( Figure 1) Hertz, Bock, & Schousboe, 1978;Juurlink et al., 1981;Kvamme et al., 1982;Larsson et al., 1985;Schousboe et al., 1979;Yu et al., 1984). The expression and activity of PAG in cultured astrocytes (Kvamme et al., 1982;Schousboe et al., 1979) were controversial at the time, due to the general notion of PAG being a neuronal enzyme (Bradford & Ward, 1976).
Although the expression and activity of PAG are higher in neurons than in astrocytes (Hogstad et al., 1988;Laake et al., 1999), active astrocyte glutamine metabolism has been functionally demonstrated (Huang & Hertz, 1995;McKenna, Tildon, et al., 1996;. Astrocyte PAG expression was recently confirmed in a subset of astrocytes in vivo (Cardona et al., 2015), which may provide these cells with a metabolic flexibility, but the exact physiological functions remain unclear .
AAT is the most active transaminase in the brain and is, together with malate dehydrogenase and several mitochondrial metabolite carriers, an important component of the malate-aspartate shuttle (MAS). The MAS facilitates transfer of reducing equivalents of NADH, generated in the cytosol by glycolysis, into the mitochondria, and is thus essential to sustain glycolytic activity McKenna et al., 2006). The MAS has also been shown to be important for synthesis of neurotransmitter glutamate. By inhibiting AAT and the dicarboxylate carrier, both of which are integral parts of the MAS, it was shown that the biosynthesis of transmitter glutamate was abolished in cultured cerebellar granule cells (Palaiologos et al., 1988), being a glutamatergic neuronal preparation (Drejer et al., 1982;Gallo et al., 1982). This study formed the basis of the "pseudo-MAS" model, in which cytosolic glutamate is derived from glutamine via PAG activity, hence linking the activity of MAS and neuronal glycolysis to neurotransmitter recycling (Hertz & Rothman, 2016). The pseudo-MAS model has recently been thoroughly dissected by Rothman and colleagues , resulting in modified models, involving a multitude of mitochondrial carriers and metabolites. These revised pseudo-MAS models may aid to explain the mechanistic basis behind the 1:1 relationship between neuronal glucose oxidation and neurotransmitter recycling (Sibson et al., 1998).

| Astrocyte glutamate metabolism
In line with the extensive uptake capacity, glutamate is an excellent oxidative substrate in astrocytes (McKenna, 2012). Early functional studies using 14 C-enriched glutamate demonstrated that a major fraction of exogenous glutamate was converted into glutamine in cultured astrocytes (Farinelli & Nicklas, 1992;Waniewski & Martin, 1986), however, a large fraction was also lost as 14 CO 2 , that is, via oxidation in the TCA cycle (McKenna, Tildon, et al., 1996;Yu et al., 1982). The advent of stable isotope tracing in combination with nuclear magnetic resonance (NMR) spectroscopy presented a major leap for neurometabolic research . NMR spectroscopy was applied to study the uptake and metabolism of [U-13 C]glutamate in cultured astrocytes (Figure 3a), confirming a significant oxidative glutamate metabolism, whereas only 30% of the glutamate was converted to glutamine .
Collectively, these studies demonstrate that astrocyte glutamate metabolism, via GDH, serves a metabolic regulator of extracellular glutamate levels.

F I G U R E 3
Application of nuclear magnetic resonance (NMR) spectroscopy to functionally explore astrocytic and neuronal metabolism. NMR spectroscopy and stable isotope tracing, that is, mapping metabolism of substrates enriched with stable isotopes, were applied to explore metabolism of cultured neurons and astrocytes. (a) Uptake and subsequent metabolism of exogenous [U-13 C]glutamate (0.5 mM) leads to extensive glutamine synthesis and release in cultured astrocytes . Furthermore, the 13 C-enriched carbon skeleton of glutamate was recovered as lactate in the media. This occurs through partial pyruvate recycling catalyzed by malic enzyme (ME) or the concerted actions of phosphoenolpyruvate carboxykinase (PEPCK) and pyruvate kinase (PK). The physiological roles of partial pyruvate recycling are not completely clear but may serve as a pathway to sustain glutamate uptake in astrocytes. (b) By analyzing the media of cultured neurons and astrocytes following metabolism of [U-13 C]glucose, a preferential astrocytic release of alanine, citrate, and glutamine was demonstrated (Sonnewald et al., 1991), whereas both neurons and astrocyte in culture released large quantities of lactate. This study prompted the notion that astrocyte-derived metabolites could play an active role for neuronal biosynthesis and function. (c) Incubating co-cultures of astrocytes and GABAergic neurons with [2-13 C]acetate, a substrate preferentially metabolized in astrocytes, showed extensive 13 C enrichment of neuronal GABA . By applying L-methionine sulfoximine (MSO), an inhibitor of the astrocytic enzyme glutamine synthetase (GS), a drastic reduction of the 13 C enrichment of GABA was observed, hereby functionally demonstrating that astrocyte-derived glutamine serves as direct precursor of neuronal GABA synthesis. This observation is now a well-established feature of the GABA-glutamine cycle ( Figure 2). Abbreviations not explained above: ALAT: alanine aminotransferase; LDH: lactate dehydrogenase; PAG: phosphate-activated glutaminase.

| Glutamate metabolism and pyruvate recycling
Further studies of [U-13 C]glutamate and [U-13 C]glutamine metabolism revealed that 13 C label in released lactate was extensively recovered from cultures of astrocytes ( Figure 3a) Waagepetersen et al., 2002;Westergaard et al., 1996), but not from neurons (Sonnewald, White, et al., 1996;. The 13 C enrichment in lactate from [U-13 C]glutamate metabolism can only occur by so-called "pyruvate recycling," in which the glutamate carbon backbone leaves the TCA cycle as either malate or oxaloacetate and is subsequently converted to pyruvate (Cerdán, 2017;Sonnewald, 2014). Pyruvate recycling may be mediated by either malic enzyme (ME) or the concerted actions of phosphoenolpyruvate carboxykinase (PEPCK) and pyruvate kinase (PK) ( Figure 3a) Cruz et al., 1998). If the generated pyruvate re-enters the TCA cycle as acetyl CoA, it is termed full pyruvate recycling, which is needed for full oxidation of the glutamate carbon skeleton (Sonnewald, 2014) and has also been documented in neurons (Olsen & Sonnewald, 2015;Olstad, Olsen, et al., 2007).
However, the generated pyruvate can also be converted into lactate and be released, a process known as partial pyruvate recycling. In line with the observed lactate generation following [U-13 C]glutamate metabolism in astrocytes (see references above), additional studies confirmed active partial pyruvate recycling in astrocytes (Håberg et al., 1998;Hassel & Sonnewald, 1995). Given that glutamate-derived lactate formation increases as the extracellular glutamate concentration rises (McKenna, Sonnewald, et al., 1996;, partial pyruvate recycling in astrocytes has been proposed to account for the mismatch between anaplerosis and cataplerosis within the brain (Sonnewald, 2014). Using a genetic GDH knock-out mouse, it was recently shown that the rates of both partial pyruvate recycling in cultured astrocytes (Nissen et al., 2015) and full pyruvate recycling in brain slices  are dependent upon GDH activity. Given that GDH may act as a metabolic safe-guard capable of elevating oxidative glutamate metabolism when glutamate concentrations rise, increased partial pyruvate recycling in astrocytes may serve as a pathway of removing excess glutamate carbons from the TCA cycle ( Figure 3a). The glutamate carbons are hereby released as lactate, avoiding intracellular glutamate accumulation, which may allow sustained astrocyte glutamate uptake. Pyruvate recycling was also increased in cultured neurons after hypoglycemia (Amaral et al., 2011) and elevated pyruvate recycling of both neurons (Pascual et al., 1998) and astrocytes (Morken et al., 2014) has been demonstrated upon ischemic insult ex vivo. These studies suggest that pyruvate recycling may provide an alternative pathway to generate acetyl CoA, and hereby maintain TCA cycle activity, when carbon supplies are limited (Cerdán, 2017). However, the physiological roles of cerebral pyruvate recycling are not yet completely clear and deserves further scientific attention.

| Neuronal and human-specific glutamate metabolism
As neuronal glutamate uptake is gaining attention (Danbolt et al., 2016;Rimmele & Rosenberg, 2016), so is the role of neuronal glutamate metabolism. As mentioned above, cultured neurons can utilize glutamate as a metabolic substrate Sonnewald, White, et al., 1996;. Deletion of the neuronal expression of GLT-1 leads to diminished glutamate utilization in the forebrain (McNair et al., 2019(McNair et al., , 2020Rimmele et al., 2021) and disturbs hippocampal glucose metabolism . Curiously, neuronal GLT-1 deletion also causes altered mitochondrial dynamics in both synapses and surrounding astrocytes, leading to increased mitochondrial density, reduced cristae distance and elevated ATP synthesis (McNair et al., 2019(McNair et al., , 2020Rimmele et al., 2021). These observations may suggest a metabolic compensation related to disrupted recruitment of mitochondria by glutamate transporters (Genda et al., 2011;Robinson et al., 2020), but the mechanism remains to be resolved. Furthermore, although the expression of GDH is concentrated primarily in astrocytes (Aoki et al., 1987;Lovatt et al., 2007;Zaganas et al., 2001), neuronal GDH activity should not be ignored. As observed for astrocytes (Bauer et al., 2012), GDH inhibition also limits synaptic glutamate uptake (Whitelaw & Robinson, 2013). Furthermore, genetic deletion of GDH hampers glutamine oxidation in cultured neurons (Hohnholt et al., 2018) and decreases glutamate metabolism in synaptosomes , underlining the metabolic importance of neuronal GDH.
Recently, species-specific differences in glutamate metabolism were explored in acutely isolated cerebral cortical slices of mice and humans (Westi et al., 2022). Overall, the direct metabolism of exogenous [U-13 C]glutamate was preserved between the mouse and human cerebral cortex, however, human brain slices exhibited higher synthesis of glutamine, and subsequent GABA formation, from metabolism of [U-13 C]glutamate (Westi et al., 2022). Since glutamine is synthesized in astrocytes, these results suggest that human astrocytes exhibits a higher capacity for glutamine synthesis, hereby supporting neuronal GABA replenishment as part of the glutamine-GABA cycle ( Figure 2). Furthermore, incubation with [U-13 C]glutamate resulted in a selective increase of the intracellular aspartate content in brain slices of mice (Westi et al., 2022), suggesting that AAT activity may be the dominant pathway of handling elevated exogenous glutamate levels in rodents. This agrees well with the expression profile of AAT (genes: Got1 and Got2), which are particularly enriched in the rodent brain compared with human (Sjöstedt et al., 2020). In addition, the human brain expresses an additional isoform of GDH: GDH2 (gene: Glud2), which is absent in rodents.
Inducing GDH2 expression in mice elevates the capacity for glutamate uptake and oxidation in astrocytes . These results not only further underline the crucial role of GDH in cerebral glutamate homeostasis but may also suggest that the combination of high GS and GDH activity of the human brain jointly acts as a metabolic defense against harmful elevations in extracellular glutamate levels. The great size and morphological complexity of astrocytes is a hallmark of the human brain (Oberheim et al., 2006(Oberheim et al., , 2009 and future efforts should be devoted to further elucidating the metabolic function of, and collaboration between, human neurons and astrocytes.

| Cellular GABA metabolism
Given the extensive presynaptic re-uptake of GABA (Figure 2), the metabolic aspects of GABA homeostasis have long been underappreciated (Andersen, Schousboe, & Wellendorph, 2023). However, GABA metabolism is essential for cerebral function, exemplified by the severe symptoms of genetic malfunction of SSADH, including developmental delay, mental retardation, and hypotonia (Malaspina et al., 2016). Experimentally, active GABA metabolism was demonstrated in brain slices by an important study of Balázs et al. demonstrating that GABA was able to elevate mitochondrial respiration and that oxidation of [1-14 C]GABA led to a substantial release of 14 CO 2 (Balázs et al., 1970). GABA is indeed able to support mitochondrial respiration, albeit at substantially lower rates than other energy substrates Cunningham et al., 1980;Ravasz et al., 2017), suggesting that GABA is a poor substrate for energy production. However, [1-14 C]GABA metabolism in rodent brain slices resulted in significant 14 C incorporation into aspartate and glutamine (Balázs et al., 1970;Berl et al., 1970).
This observation was recently confirmed, as metabolism of [U-13 C]GABA in brain slices of the human cerebral cortex revealed that GABA strongly aids to sustain astrocyte glutamine synthesis . The extensive glutamine synthesis derived from metabolism of GABA underlines that GABA oxidation in the TCA cycle indeed is an integral part of the glutamine-GABA cycle ( Figure 2) and further suggests that astrocytes are the primary cellular compartment of GABA metabolism (Andersen, Schousboe, & Wellendorph, 2023). Indeed, cultured astrocytes display highly active GABA metabolism  and elevated affinity of GABA-T for GABA conversion when compared to neurons (Bardakdjian et al., 1979;Larsson & Schousboe, 1990).
In addition, non-synaptic mitochondria exhibit higher enzymatic activities of GABA-T and SSADH in comparison to mitochondria of synaptic origin (Walsh & Clark, 1976). It has been estimated that nearly half of all released GABA from neurons is taken up and metabolized by astrocytes in the rat brain in vivo (Duarte & Gruetter, 2013). McKenna and Sonnewald further demonstrated that GABA is able to compete with glutamate for oxidation in cultured astrocytes (McKenna & Sonnewald, 2005), which is striking as astrocytic oxidation of glutamate is preferred over most alternative substrates, including glucose, lactate, and ketones (McKenna, 2012). Using selective GAT inhibitors, it was shown that inhibition of astrocyte GAT3 activity moderately decreased GABA uptake, but severely reduced GABA metabolism . In contrast, GAT1 inhibition reduced GABA uptake, but had limited effects on oxidative metabolism of GABA , which agrees well with the observation that GAT1 inhibition had no effect on TCA cycle function and GABA recycling in vivo (Patel et al., 2015).
Further insights into the physiological roles of GABA metabolism have been obtained from observing the metabolic effects of GABA-T inhibition by the compound vigabatrin (also known as γ-vinyl-GABA). In line with a highly active GABA metabolism, vigabatrin treatment rapidly elevates GABA levels (Jung et al., 1977;Pierard et al., 1999), while diminishing cerebral glutamine levels Pierard et al., 1999). Interestingly, blocking GABA-T repeatedly with vigabatrin reduced GS activity in astrocytes (Waniewski & Martin, 1995), which may be because of limited substrate availability, that is, glutamate-derived from oxidative GABA metabolism in astrocytes. Reduced cerebral glutamine levels are also observed in SSADH deficiency (Kirby et al., 2020) again indicating that astrocyte GABA metabolism is essential to sustain glutamine synthesis. Despite the evidence that astrocytes are the primary cellular compartment of GABA metabolism, neuronal GABA metabolism must not be disregarded. Neurons express both GABA-T and SSADH, and vigabatrin treatment of cultured neurons also elevates intracellular GABA levels (Gram et al., 1988).
Curiously, neurons may be more sensitive to GABA-T inhibition by vigabatrin (Gram et al., 1989;Larsson et al., 1986), which has been suggested to be mediated via a more efficient uptake mechanism . Collectively, oxidative GABA metabolism is highly active, particularly in astrocytes, and supports glutamine synthesis and the glutamine-GABA cycle. Disruption of astrocyte energy metabolism may therefore have serious consequences for neurotransmitter recycling and further investigations of cellular GABA metabolism, in both health and disease, are highly encouraged.

| Acetate can be used to study astrocyte metabolism
Early neurometabolic studies clearly demonstrated that more than one distinct pool of glutamate existed. Metabolism of radioactive 14 C-enriched acetate led to extensive 14 C incorporation in glutamine when compared to its precursor glutamate, which contrasted metabolism of 14 C-enriched glucose (Garfinkel, 1966;Nicklas & Clarke, 1969;O'Neal & Koeppe, 1966;van den Berg et al., 1966). This result can only be explained by at least two separate pools of glutamate with different turn-over rates, which led to the concept of metabolic compartmentation (Berl et al., 1968(Berl et al., , 1970van den Berg et al., 1969;van den Berg & Garfinkel, 1971;van den Berg & Ronda, 1976). The compartment of active glutamine synthesis was later identified as the astrocyte, by the seminal observation that GS expression is restricted to astroglial cells and is absent in neurons (Martinez-Hernandez et al., 1977;Norenberg & Martinez-Hernandez, 1979). The concept of cellspecific acetate metabolism also applies to 13 C-enriched acetate, leading to extensive 13 C enrichment of glutamine and citrate Badar-Goffer et al., 1990;McNair et al., 2017). The cellular specificity of acetate metabolism has recently been challenged , and although not completely selective, acetate remains a useful marker of astrocyte energy metabolism (Wyss et al., 2011). Of note, it has been suggested that the preferential astrocyte metabolism of acetate was mediated by selective transport (Waniewski & Martin, 1998). This is, however, incompatible with the fact that neuronal monocarboxylate transporters (MCTs) are capable of transporting acetate (Rae et al., 2012), which is also supported by the prominent intracellular NMR acetate signal of cultured neurons incubated with [2-13 C]acetate . Instead, the cell-specific metabolism has been proposed to be regulated at the enzymatic level by activity of acetyl CoA synthase converting acetate into F I G U R E 4 Distinct metabolic features of astrocyte are essential for brain function. Astrocyte metabolism is primed to sustain neurotransmitter recycling. (1) Astrocytes exhibit a large capacity for uptake of neurotransmitter glutamate and GABA. (2) GABA taken up by astrocytes is extensively metabolized as astrocytes display high activity of GABA transaminase (GABA-T) and succinic semialdehyde dehydrogenase (SSADH). (3) Astrocytes are the primary compartment of glutamate dehydrogenase (GDH) expression and activity, facilitating efficient glutamate oxidation in the TCA cycle. (4) Crucially, astrocytes synthesize large amounts of glutamine via glutamine synthetase (GS) activity. Glutamine is released from astrocytes and serves essential roles in restoring neuronal neurotransmitter pools as part of the glutamate/GABA-glutamine cycle ( Figure 2). Furthermore, glutamine synthesis is the primary route of cerebral ammonia fixation. (5) Astrocytes release the TCA cycle intermediate citrate which is able to regulate NMDA receptor activity. (6) To sustain the extensive biosynthesis and release of metabolites, astrocytes express pyruvate carboxylase (PC) being the primary anaplerotic enzyme of the brain. PC converts pyruvate to oxaloacetate and hereby replenish the pools of TCA cycle intermediates, which is essential to counteract the drain of metabolites. (7) Astrocytes are the primary compartment of fatty acid metabolism in the brain. Notably, acetate can be used as a selectively marker of astrocyte energy metabolism. (8) Astrocytes release lactate, which may support neuronal energy metabolism. (9) Astrocytes hold significant stores of the polysaccharide glycogen, which serves as an emergency fuel, but also facilitates multiple other functions. The functionality of the specialized features of astrocyte metabolism, including glutamate metabolism, glutamine synthesis, anaplerosis, and glycogen mobilization are all crucial to sustain synaptic function. Abbreviations not explained above: AAT, aspartate aminotransferase; AceCS, acetyl CoA synthetase; G-1-P, glucose-1-phosphate; G-6-P, glucose-6-phosphate; LDH, lactate dehydrogenase; PDH, pyruvate dehydrogenase.
acetyl CoA (Rae et al., 2012). Chemically, acetate is a short-chain fatty acid (Schönfeld & Wojtczak, 2016) and mitochondrial oxidation of fatty acids is indeed a metabolic trait of astrocytes ( Figure 4), which is further discussed in Section 5.4.

| Astrocyte metabolite release and transfer
Astrocytes synthesize and release multiple metabolites. This was illustrated by incubating cultured neurons and astrocytes with [1-13 C]glucose and subsequently quantifying the 13 C enrichment of the metabolites released to the media (Sonnewald et al., 1991) ( Figure 3b). A preferential release of lactate, alanine, citrate, and glutamine was observed from astrocytes, whereas neurons primarily released lactate (Sonnewald et al., 1991). The extensive metabolite release from astrocytes prompted the suggestion that these metabolites could be important for neuronal metabolism and biosynthesis.
Synthesis and release of lactate is a well-known metabolic attribute of astrocytes (Walz & Mukerji, 1988). It has been suggested that glutamate uptake in astrocytes stimulates lactate synthesis, which is subsequently shuttled to neurons to support oxidative metabolism (Pellerin & Magistretti, 1994). This model, linking cellular energy metabolism and neurotransmission, is known as the neuronastrocyte lactate shuttle hypothesis. The shuttle has both been supported (Mächler et al., 2016;Suzuki et al., 2011) and opposed (Diaz-Garcia et al., 2017;Lundgaard et al., 2015;Patel et al., 2014) by recent in vivo experiments, and remains a topic of debate Barros & Weber, 2018) (see further (Dienel, 2019) for a thorough discussion). Alanine is, like lactate, synthesized from pyruvate ( Figure 1). From studies of [U-13 C]lactate metabolism, it was evident that cultured neurons release more alanine than cultured astrocytes (Waagepetersen et al., 2000). In addition, astrocyte alanine uptake and accumulation was greater than in neurons (Waagepetersen et al., 2000). This led to the hypothesis than alanine transfer from neurons to astrocytes, could serve as a shuttle of the nitrogen generated by PAG activity in neurons ( Figure 1) as part of the glutamate-glutamine cycle ( Figure 2) (Waagepetersen et al., 2000). The role of alanine as a nitrogen carrier has been confirmed in brain slices (Bröer et al., 2007) but seems to occur independently of the glutamate-glutamine cycle in cultured cells (Bak et al., 2005). Alanine synthesis may furthermore serve as an important alternative route of ammonia fixation during hyperammonemia (Dadsetan et al., 2011(Dadsetan et al., , 2013. Recently, serine transfer between astrocytes and neurons has gained attention (Maugard et al., 2021). L-serine is primarily synthesized in astrocytes from the glycolytic intermediate 3-phosphoglycerate. Since astrocyte-derived L-serine, serves as a precursor for neuronal D-serine synthesis, which is a co-agonist of the NMDA receptors, transcellular serine exchange may also provide a link between energy metabolism and neurotransmission . Under specific metabolic conditions, astrocytes may also synthesize and release ketones, which will be discussed further in Section 5.4. The branched-chain amino acids (BCAAs) leucine, isoleucine and valine have also been proposed as transcellular nitrogen carriers (Sperringer et al., 2017). Indeed, the BCAAs are essential nitrogen donors for glutamate synthesis via branched-chain amino acid aminotransferase (BCAT) activity ( Figure 1) (Yudkoff, 1997). BCAT and GDH engage in a metabolon (Islam et al., 2010) and deletion of astrocytic GDH function increases BCAA metabolism (Nissen et al., 2015). This suggests that the BCAA oxidation may serve an anaplerotic role in astrocytes, which is in line with a recent report of highly active astrocytic BCAA metabolism (Salcedo et al., 2021). Citrate, being the first intermediate of the TCA cycle, is also extensively released from astrocytes (Sonnewald et al., 1991;Westergaard et al., 1994).
However, 14 CO 2 production from incubation with [ 14 C]citrate was low in cultured neurons (Westergaard et al., 1994), indicating limited neuronal uptake and metabolism of astrocyte-derived citrate, which instead suggested an extracellular role of the released citrate. Indeed, citrate was found to act as a chelating agent able to regulate NMDA-mediated neurotransmitter release by Zn 2+ chelation (Westergaard, Banke, et al., 1995). Citrate is a highly abundant metabolite in the cerebrospinal fluid, yet the complete physiological role of astrocyte-derived citrate remains to be established (Westergaard et al., 2017).
Finally, glutamine is, undoubtedly, one of the most crucial metabolites synthesized and released from astrocytes ( Figure 4).
Glutamine is synthesized selectively in astrocytes from glutamate and ammonia by GS activity (Figures 1 and 2), making glutamine synthesis the primary mechanism of cerebral ammonia homeostasis (Felipo & Butterworth, 2002;Suárez et al., 2002).
Multiple studies had shown that glutamine was an excellent precursor for neurotransmitter synthesis (Bradford et al., 1978;Reubi et al., 1978;Tapia & Gonzalez, 1978;Thanki et al., 1983;Ward et al., 1983). A glutamine-mediated coupling between astrocytes and neurons (Figure 2) was directly demonstrated by investigating [2-13 C]acetate metabolism in co-cultures of astrocytes and GABAergic neurons (Figure 3c), showing that inhibition of astrocyte glutamine synthesis severely decreased the 13 C enrichment of neuronal GABA . This observation thus directly showed that astrocyte-derived glutamine is a principal precursor of neuronal GABA synthesis, emphasizing the importance of astrocyte glutamine support. Reduced neuronal glutamate and GABA synthesis upon GS inhibition Böttcher et al., 2003;Fonnum & Paulsen, 1990;Laake et al., 1995; and concurrent disrupted neuronal signaling (Ortinski et al., 2010;Tani et al., 2014) have subsequently been confirmed.
A continuous supply of glutamine is therefore essential to sustain neurotransmission, and it has been estimated that the neuronal glutamate pool would be depleted within one minute of basal activity without external replenishment (Marx et al., 2015).
Maintaining cerebral glutamine homeostasis is a complex process, which relies on numerous cellular mechanism, including several distinctive metabolic traits of astrocytes .
Thus, glutamine synthesis, neurotransmitter transport, and energy metabolism are all integrated processes in the astrocyte.
Astrocytes furthermore display several distinct metabolic traits aiding to sustain their homeostatic functions (Figure 4). The extensive synthesis and release of metabolites from astrocytes lead to a significant drain of metabolic intermediates. Particularly, glutamine synthesis, which is derived from glutamate, and thus from the TCA cycle intermediate α-ketoglutarate (Figure 1), may deplete TCA cycle constituents in astrocytes. To counteract the loss of metabolic intermediates, anaplerotic pathways, that is, reactions capable of replenishing TCA cycle intermediates, are needed (Sonnewald, 2014). In the brain, the quantitatively most important anaplerotic pathway is pyruvate carboxylase (PC) (Patel, 1974), catalyzing the conversion of pyruvate into the TCA cycle intermediate oxaloacetate by fixation of CO 2 (Berl et al., 1962;Utter & Keech, 1963). The combined studies of Yu et al. (1983) and Shank et al. (1985) unequivocally demonstrated that this key enzyme is selectively expressed in astrocytes (Figure 4), and not in neurons, directly correlating with the capacity for CO 2 fixation of these cell types in culture (Kaufman & Driscoll, 1992). Since sufficient anaplerosis is required for de novo glutamate, GABA, and glutamine synthesis, the selective expression of PC in astrocytes makes these cells the primary regulators of neurotransmitter biosynthesis . Indeed, activity of PC is essential for sustained astrocyte glutamine synthesis and thus neurotransmitter recycling (Gamberino et al., 1997;Lapidot & Gopher, 1994). In whole brain and cerebral cortex, PC activity has been reported to account for 10-15% of the total glucose oxidation (Duarte & Gruetter, 2013;Oz et al., 2004), which may be as high as 20% in the hippocampus (McNair et al., 2022). Furthermore, PC activity correlates with the rate of astrocyte metabolism  and brain activity (Oz et al., 2004), suggesting that the anaplerotic flux through PC is needed to sustain the glutamate/GABA-glutamine cycle. However, only small, and hence non-significant, increases in PC activity were observed during cortical stimulation (Sonnay et al., 2016) and induced seizures  in anesthetized rats. Some degree of neuronal pyruvate carboxylation has been reported (as summarized in (Hassel, 2001)), but the quantitative importance of this proposed pathway in neurons remains questionable. In addition to PC activity, metabolism of odd-chain fatty acids and the BCAAs isoleucine and valine, gives rise to succinyl CoA and these two amino acids are therefore also anaplerotic substrates (Sonnewald, 2014). Intriguingly, metabolism of BCAAs is particularly active in astrocytes (Salcedo et al., 2021) and may serve as an additional pathway to sustain glutamine synthesis. Note that some metabolic aspects, which were thought to be exclusively confined to astrocytes, have been reported in other glial cell types. This includes expression of PC and GS in oligodendrocytes (Cammer, 1990;Murin et al., 2009;Tansey et al., 1991).
Glutamine synthesis in oligodendrocyte is not essential for life (Ben Haim et al., 2021;Xin et al., 2019), which is in sharp contrast to astrocyte GS function (He et al., 2010). However, hyperactive oligodendrocyte glutamine synthesis in the spinal cord may be a potential pathological mechanism of amyotrophic lateral sclerosis (Ben Haim et al., 2021) (as recently discussed in ).
Astrocytes also contain significant stores of glycogen (Figure 4), which is a branched polysaccharide of glycosyl units .
Glycogen may provide fuel during brief episodes of hypoglycemia or high neuronal activity (Brown et al., 2005;Wender et al., 2000) but is also continuously synthesized and degraded in the presence of glucose (Dienel et al., 2007). This process is known as the glycogen shunt, which is highly active in astrocytes .
Crucially, inhibition of glycogen metabolism leads to impaired learning and memory, which, however, can be rescued by external glutamine supplementation (Gibbs et al., 2006). This observation is in line with the recent report of glutamine transfer through astrocyte hemichannels, serving a crucial role for sustained glutamatergic transmission and hippocampal memory formation (Cheung et al., 2022).
The metabolic link between glycogen and glutamine is also underlined by the extensive accumulation of glycogen granules upon inhibition of astrocyte glutamine synthesis (Folbergrová et al., 1969;Phelps, 1975;Swanson et al., 1989). Genetic impairment of glycogen degradation can lead to Lafora disease (Nitschke et al., 2018), characterized by severe epilepsy and neurodegeneration, being driven by extensive astrocyte glycogen accumulation (Duran et al., 2021).
Recently, it was demonstrated that brain glycogen is not only composed of glucose but also 25% glucosamine needed for protein glycosylation , adding another important aspect, beyond intermediate metabolism, to astrocytic glycogen.

| Disturbances of the glutamate/GABAglutamine cycle are associated with seizure activity
Disruptions of glutamatergic and GABAergic neurotransmission are implicated in many different neurological diseases. Offsetting of the delicate excitatory-inhibitory balance is associated with epileptic activity (Rowley et al., 2012) and can be caused by impairments of both the metabolic enzymes and transporters related glutamate and GABA homeostasis (Figure 1 & Figure 2). Most strikingly, deletion of GS expression in astrocytes leads to early neonatal death in mice (He et al., 2010), which, in addition to seizures, is also observed in congenital human GS deficiency (Häberle et al., 2005). Selective deletion of GS in the forebrain causes epilepsy (Zhou, Dhaher, et al., 2019), whereas GS haploinssuficiency in mice leads to increased susceptibility to induced seizures (van Gassen et al., 2009). Likewise, deletions of the GABA synthesizing enzymes GAD65 and GAD67 are either lethal or associated with increased seizure activity or susceptibility (Asada et al., 1996(Asada et al., , 1997Kash et al., 1997). Malfunctions of the two enzymes involved in GABA metabolism, GABA-T and SSADH, cause severe seizures, which may lead to lethal status epilepticus (Koenig et al., 2017;Pearl et al., 2003). Selective overexpression of GDH in neurons, generating elevated cerebral glutamate levels and release, is furthermore associated with age-dependent neuronal loss (Bao et al., 2009;Michaelis et al., 2011).
In addition to the enzymes involved in glutamate and GABA metabolism, it has also been investigated to what extent manipulation of the cellular transporters for the two amino acid neurotransmitters may influence epileptiform activity. Thus, deletion of GAT1 in mice leads to increased sensitivity toward induced seizures (Chiu et al., 2005;Jensen et al., 2003), whereas human mutations in GAT1, causing dysfunctional GABA uptake, are associated with epilepsy, autism, and intellectual disability (Goodspeed et al., 2020;Mermer et al., 2021). Curiously, overexpression of GAT1 in mice, also increases susceptibility toward induced seizures Zhao et al., 2003), with subsequent impaired cognition . As previously mentioned, deletion of both global and astrocytic GLT-1 leads to seizures and reduced life span in mice (Petr et al., 2015;Tanaka et al., 1997). Seizures were also observed during EAAT3 knock-down (Rothstein et al., 1996;Sepkuty et al., 2002), but not in a full EAAT3 knock-out mouse model (Peghini et al., 1997), while reduction in GLAST expression does not seem to play any significant role in this regard (Rothstein et al., 1996). The studies above clearly illustrate that all parts of the glutamate/GABA-glutamine cycle, both the metabolic machinery and transporters, are crucial to maintain synaptic function and thus the excitatory-inhibitory balance. Epileptic activity is furthermore a common co-morbidity of several neurodegenerative diseases (Cloud et al., 2012;Gruntz et al., 2018;Vossel et al., 2013), indicative of major synaptic imbalances, which may cause toxic effects as discussed below.

| Glutamate and excitatory amino acid toxicity: Excitotoxicity
Excessive exposure of neurons to excitatory amino acids, for example, glutamate, aspartate, and NMDA, leads to cellular damage and subsequently cell death (Olney, 1969;Olney et al., 1971;Olney & Ho, 1970), a phenomenon which was later coined excitotoxicity (Olney et al., 1974). Excitotoxicity is primarily mediated by prolonged and exacerbated glutamate activation of NMDA and AMPA glutamate receptors (Lewerenz & Maher, 2015). Such receptor overstimulation initiates multiple cellular cascades including dysfunctional calcium buffering, oxidative stress, and mitochondrial dysfunction, ultimately leading to cell death (Dong et al., 2009). Acute cellular glutamate toxicity can arise from cerebral insults such as ischemic events or traumatic brain injury. That cerebral ischemia indeed leads to toxic extracellular glutamate levels was first demonstrated by the seminal study of Benveniste et al. (1984). Using a newly developed technique allowing sampling of the interstitial fluid of the brain, originally named "microdialysis" (Benveniste, 1989), it was demonstrated that a brief period of ischemia leads to excessive release of glutamate to the extracellular space in the rat brain (Benveniste et al., 1984).

Excitotoxicity can also develop over longer periods of time,
where chronic glutamate exposure leads to gradual cellular damage (Rothstein et al., 1993). Since a slowly progressing neuronal loss is a hallmark of multiple brain disorders, excitotoxicity has been suggested as a common pathological mechanism for most neurodegenerative diseases (Greenamyre, 1986;Lipton & Rosenberg, 1994).
Indeed, deletion of astrocytic glutamate transporters causes excitotoxic neurodegeneration (Rothstein et al., 1996) and decreased expression of glutamate transporters is a common observation of the diseased brain (Sheldon & Robinson, 2007). That hampered glutamate clearance, and concomitant excitotoxicity, could be a major mechanism of neurodegeneration has been investigated by studies seeking either to increase or reduce glutamate uptake capacity in animal models of human disease. Several compounds that increase astrocyte GLT-1 expression have been identified (Ganel et al., 2006;Kong et al., 2014;Rothstein et al., 2005). The most well-studied inducer of GLT-1 is the β-lactam antibiotic ceftriaxone, which has been found to reduce pathological progression in mouse models of amyotrophic lateral sclerosis (Rothstein et al., 2005), Huntington's disease (Miller et al., 2008) and Alzheimer's disease (Brymer et al., 2023;Hefendehl et al., 2016;Zumkehr et al., 2015). It should be noted that ceftriaxone also induces the transcription factor Nrf2 and expression of the cysteine-glutamate antiporter xCT (Knackstedt et al., 2010;Lewerenz et al., 2009), which may enhance the cellular antioxidant defense and hereby contribute to neuroprotective effects. Intriguingly, ceftriaxone treatment has been found not only to enhance glutamate uptake but also to elevate glutamine synthesis, transfer, and metabolism in a mouse model of Alzheimer's disease (Fan et al., 2018(Fan et al., , 2021. This may suggest that the beneficial effects of inducing GLT-1 expression, may, in part, be mediated by elevated neurotransmitter recycling between astrocytes and neurons . Alternative genetic induction of astrocyte GLT-1 expression has also been shown to slow pathological progression in models of amyotrophic lateral sclerosis  and Alzheimer's disease (Takahashi et al., 2015), supporting that the positive outcomes are related to increased GLT-1 function. This is furthermore in line with the worsened pathological states upon genetic reduction in GLT-1 expression in mouse models of amyotrophic lateral sclerosis (Pardo et al., 2006) and Alzheimer's disease (Mookherjee et al., 2011). Collectively, these studies strongly indicate a clear pathological role of diminished synaptic glutamate clearance. Intriguingly, reducing the expression of GLT-1 did not worsen the pathological phenotype of the R6/2 mouse model of Huntington's disease (Petr et al., 2013). However, a reduced expression of the neuronal glutamate transporter EAAT3 was observed in the R6/2 striatum (Petr et al., 2013). In addition, deletion of neuronal GLT-1 leads to a transcriptome profile similar to that observed in Huntington's disease (Laprairie et al., 2019), which may suggest a critical role of neuronal glutamate uptake in Huntington's disease pathology.

| Astrocyte energy and neurotransmitter metabolism in disease and neurodegeneration
As outlined in detail above, glutamate homeostasis is closely linked to the cellular energy metabolism of both neurons and astrocytes.
Thus, alterations in brain energy metabolism may also lead to harmful impairments of glutamate handling (Beal, 1992;Greene & Greenamyre, 1996;Passlick et al., 2021). Furthermore, given that astrocytes are the primary controllers of extracellular glutamate levels, excitotoxicity is closely related to astrocytic metabolic function. Declining activity in brain energy metabolism is a hallmark of several neurodegenerative diseases (Camandola & Mattson, 2017).
In Alzheimer's disease, reduced glucose metabolism, visualized by uptake of the glucose analog 18 F-fluorodeoxyglucose by positron emission tomography ( 18 F-FDG-PET), is observed during very early disease development and progressively worsen during the pathological course (Bateman et al., 2012;Gordon et al., 2018;Mosconi et al., 2008). Based on the dogma that excitatory neurons account for roughly 80% of the total glucose oxidation of the brain (Attwell & Laughlin, 2001;Yu et al., 2018), the lower 18 F-FDG-PET signals in Alzheimer's disease are often interpreted as neuronal glucose hypometabolism. However, a recent study revealed that changes in expression of glial metabolic proteins was a superior correlate with both molecular and symptomatic read-outs in Alzheimer's disease (Johnson et al., 2020). Furthermore, cerebral hypometabolism in Alzheimer's disease correlates well with the astrocytic marker 11 Cdeuterium-L-deprenyl ( 11 C-DED) (Carter et al., 2019), suggesting that astrocyte dysfunction and atrophy may underlie the observed metabolic impairments. This is supported by the observation that, during glutamate stimulation, astrocytes significantly contribute to 18 F-FDG-PET signals (Zimmer et al., 2017). Furthermore, it was recently demonstrated that both microglia and astrocytes display a substantially higher capacity of 18 F-FDG uptake than neurons (Xiang et al., 2021). When the Na + /K + -ATPase activity of astrocytes (MacAulay, 2020), which is crucial to buffer synaptic K + , is taken into account, it has been argued that astrocytes may be as energetically expensive as neurons (Barros, 2022). Finally, when the size and complexity of human astrocytes are considered (Oberheim et al., 2006), the metabolic contribution of astrocytes in the human brain may have been severely underestimated (Dienel & Rothman, 2020). Collectively, these notions suggest that altered astrocyte function and metabolism may significantly contribute to pathological changes in brain metabolism. Thus, a reevaluation of human astrocyte energetics in relation to clinical imaging is strongly warranted, both during healthy and diseased states.
During injury or disease development, astrocytes may become "reactive" and undergo highly complex adaptations (Sofroniew & Vinters, 2010). These changes are not easily defined across diseases as they are influenced by several factors including age, life-style, other pathologies and disease context (Sofroniew, 2020). Reactive astrocytes may be proliferative, display cellular hypertrophy with maintained territorial domains, become atrophic or even degenerate (Pekny et al., 2016;Verkhratsky et al., 2017). Adaptations of astrocyte gene expression translate into reversible remodeling of physical, morphological, and metabolic features, hereby altering astrocyte function (Escartin et al., 2021). During prolonged pathology these adaptations may, however, become permanent, leading to a chronic reactive state, which includes impairments of astrocyte metabolism (Iglesias et al., 2017;Kumar et al., 2021;Xiong et al., 2022). This, in turn, may lead to reduced homeostatic support of astrocytes, which may accelerate or even facilitate pathological development (Parpura et al., 2012;Verkhratsky et al., 2022).
The functional consequences of pathological changes in astrocyte metabolism, and how it affects brain function, is still poorly understood. However, short-term treatment with the cytokine TNFα, mimicking an acute inflammatory response, elevates astrocyte mitochondrial respiration and glycolytic capacity (Kabiraj et al., 2022). A sustained glycolytic activity was found to protect astrocytes against toxicity mediated by the accumulating peptide of Alzheimer's disease, amyloidβ (Aβ) (Fu et al., 2015). Furthermore, elevated acetate metabolism has been observed in Alzheimer's disease patients during early disease progression (Duong et al., 2021), which is supported by elevated 11 C-DED imaging signals (Rodriguez-Vieitez et al., 2016;Schöll et al., 2015). These observations may suggest that increased metabolism could serve as an initial transient adaptation to support astrocyte remodeling. However, during prolonged and chronic pathological insult the general metabolic capacity of astrocytes seems to decline Dematteis et al., 2020;Diaz-Castro et al., 2019;Skotte et al., 2018;Sonninen et al., 2020;van Gijsel-Bonnello et al., 2017). Given that astrocyte metabolism is directly linked to the functionality of the glutamate/ GABA-glutamine cycle, these metabolic impairments may have direct consequences for synaptic signaling and brain function . Hampered glycolytic activity may reduce efflux of lactate and serine hereby depriving neurons of important metabolites ( Figure 5). Diminished astrocyte glutamine support has been shown to directly hamper neuronal GABA synthesis in brain slices of mouse models of Alzheimer's disease Andersen, Skotte, et al., 2021) and Huntington's disease (Skotte et al., 2018), which may disrupt the excitatory-inhibitory balance. Furthermore, as GS activity is the primary pathway of ammonia fixation in the brain, reduced pathological glutamine synthesis may lead to harmful elevations of cerebral ammonia levels, which has also been reported in both Alzheimer's disease (Seiler, 2002) and Huntington's disease (Chiang et al., 2007). Finally, a lower TCA cycle activity of astrocytes may reduce the capacity for oxidation of glutamate. Together with an impaired glutamine synthesis, this may lead to an intracellular accumulation of glutamate and hereby reduce astrocyte glutamate uptake capacity ( Figure 5). Indeed, reduced GS function has been proposed to drive hampered glutamate uptake capacity and excitotoxicity in hippocampal epilepsy (Eid et al., 2004), which is in line with the intracellular accumulation of glutamate (Laake et al., 1995) and lower glutamate uptake capacity (Zou et al., 2010) of astrocytes upon pharmacological inhibition of GS. Aberrant glutamine homeostasis is implicated in many diseases and the detrimental consequences hereof have recently been reviewed in detail elsewhere . It must F I G U R E 5 Modulation of astrocyte and neuron energy metabolism via lipid and ketone supplementation. Pathological dysfunction of the energy metabolism of neurons and astrocytes may mediate or aggravate synaptic dysfunction and neurodegeneration. (1) Excessive release or impaired clearance of synaptic glutamate may lead to overstimulation, excitotoxicity, and neuronal death. (2) Diminished metabolic function of astrocytes may deprive neurons of metabolic support, including transfer of serine, lactate, ketones, and glutamine. (3) Reduced oxidative astrocyte metabolism and deficient glutamine synthesis, may lead to an intracellular accumulation of glutamate, which ultimately may impair glutamate uptake and hence contribute to excitotoxicity. (4) Ketones, for example, β-hydroxybutyrate, are a commonly applied supplementary fuel, which can either be given as dietary supplement or be generated from hepatic fatty acid metabolism. Ketones are the primary alternative fuel of the brain and are excellent substrates to support neuronal energy metabolism. (5) Astrocytes are the primary compartment of fatty acid metabolism in the brain and exogenous fatty acids support the metabolic function of astrocytes. Particularly, the medium-chain fatty acids octanoic acid (C8) and decanoic acid (C10) have been found to boost lactate, ketone and glutamine synthesis is astrocytes. (6) Targeting astrocyte metabolism, by improving TCA cycle function, glycolytic activity, and glutamine synthesis, may aid to restore activity of the glutamate/GABA-glutamine cycle, increase synaptic function, and ultimately halt pathological progression. Dietary supplementation with a combination of ketones and fatty acids, for example, via medium-chain triglycerides, will thus provide auxiliary fuels for both neurons and astrocytes. Abbreviations not explained above: AAT: aspartate aminotransferase; GDH: glutamate dehydrogenase; GS: glutamine synthetase; PDH: pyruvate dehydrogenase. be noted that metabolic impairments, not only of astrocytes but also of microglia (Aldana, 2019;Paolicelli et al., 2022) and oligodendrocytes (Lee et al., 2012;Philips & Rothstein, 2017) may play significant roles in the development of multiple cerebral diseases.

| Brain energy metabolism as a therapeutic target: Focus on astrocyte fatty acid metabolism
Given the close connection between brain energetics, neurotransmitter recycling, and synaptic function, modulation of brain metabolism as a therapeutic intervention is gaining attention in both aging and neurological disease (Augusto-Oliveira & Verkhratsky, 2021; Cunnane et al., 2011Cunnane et al., , 2020. Brain metabolism can be modulated via several mechanisms, including changes in lifestyle, hormonal treatment, or dietary supplementation (Camandola & Mattson, 2017;Cunnane et al., 2020). With regard to dietary intervention, brain ketone and lipid metabolism represent promising metabolic targets ( Figure 5) (Cunnane et al., 2016;Hertz et al., 2015;Panov et al., 2014). The ketones β-hydroxybutyrate and acetoacetate are synthesized from fatty acids in the liver and are the primary alternative substrate of the brain when glucose availability is limited (Owen et al., 1967). Dietary supplementation with ketones have proven particularly effective for the management of drug-resistant epilepsy (Lutas & Yellen, 2013;McNally & Hartman, 2012), but may also enhance cognitive function in both aging and disease (Cunnane et al., 2016).
Triglycerides of the medium-chain fatty acids octanoic acid and decanoic acid also display beneficial outcomes in neurodegenerative disease and epilepsy (Augustin et al., 2018;Croteau et al., 2018;Fortier et al., 2019;Han et al., 2021). The positive effects of medium-chain fatty acids have been hypothesized to be driven by hepatic ketone synthesis (Augustin et al., 2018), but fatty acids are also cellular energy substrates of the brain (Schönfeld & Wojtczak, 2016). Whereas ketones are primarily metabolized in neurons (Achanta et al., 2017;Pan et al., 2002), astrocytes are the primary cellular compartment of fatty acid metabolism in the brain (Eraso-Pichot et al., 2018;Fecher et al., 2019). Astrocytes oxidize both long-chain and medium-chain fatty acids more efficiently than neurons (Ebert et al., 2003;Edmond et al., 1987), which is in line with the preferential astrocytic metabolism of the short-chain fatty acid acetate as described in Section 4.1.
Both ketones and fatty acids enter cellular metabolism as acetyl CoA units to support TCA cycle function ( Figure 5). Note that odd-chain fatty acids also give rise to succinyl CoA and are hence anaplerotic (Sonnewald, 2014). Decanoic acid has furthermore been shown to exert antioxidant effects (Tan et al., 2017), increase mitochondrial biogenesis (Hughes et al., 2014), and modulate AMPA receptor signaling (Chang et al., 2016). However, astrocyte metabolism of medium-chain fatty acids also boosts mitochondrial respiration and glutamine synthesis . Glutamine, derived from metabolism of medium-chain fatty acids, was shown to support neuronal GABA synthesis , which may, in part underlie, the beneficial effects observed in epilepsy. Furthermore, octanoic acid and decanoic acid have been found to modulate astrocyte metabolite synthesis and release. Octanoic acid stimulates astrocyte ketone synthesis and release (Auestad et al., 1991;Blázquez et al., 1999;Thevenet et al., 2016), which has been proposed to be shuttled to neurons (Guzmán & Blázquez, 2001), whereas decanoic acid induces astrocyte lactate synthesis and release (Damiano et al., 2020;Lee et al., 2018;Thevenet et al., 2016). Supplementation with medium-chain fatty acids hereby supports both astrocytic and neuronal energy metabolism, via oxidation of fatty acids and ketones, respectively ( Figure 5)  . Induction of fatty acid metabolism in astrocytes have been shown to be protective during ischemia (Sayre et al., 2017), whereas supplementation with octanoic acid was able to correct Aβ-induced memory impairments in chicks (Gibbs et al., 2009), which may be mediated by elevated astrocytic metabolic support.
Lipids may also be shuttled between neurons and astrocytes.
During intense signaling activity, neurons release the lipoprotein APOE loaded with fatty acids, which is taken up and accumulated by astrocytes (Ioannou et al., 2019). The transferred fatty acids are subsequently metabolized in the astrocytes to avoid neuronal fatty acid-induced toxicity (Ioannou et al., 2019). Strikingly, the APOE4 variant is the largest genetic risk factor of sporadic Alzheimer's disease development (Strittmatter et al., 1993), which gravely enhances pathology (Shi et al., 2017;Wang et al., 2021). Expression of APOE4 disrupts the transcellular shuttling of lipids (Lin et al., 2018;Qi et al., 2021) and impairs astrocyte lipid metabolism and uptake (Farmer et al., 2021;Qi et al., 2021), which leads to intracellular lipid accumulation (Farmer et al., 2019;Qi et al., 2021;Sienski et al., 2021). Perturbed astrocyte lipid metabolism and cholesterol homeostasis have also been suggested to contribute to neurodegeneration in Huntington's disease (Karasinska & Hayden, 2011;Polyzos et al., 2019;Valenza et al., 2015). Altered astrocyte metabolism and uptake of lipids may be a crucial mechanism of neurodegeneration, as reactive astrocytes were recently reported to mediate toxic effects on oligodendrocytes via lipid release (Guttenplan et al., 2021).
Thus, a pathological rewiring of astrocyte lipid metabolism and synthesis may contribute to synaptic dysfunction and cell death. To summarize, modulation of astrocytic and neuronal energy metabolism, through lipid and ketone supplementation, may constitute an attractive therapeutic option ( Figure 5). It is important to note, that since the specific metabolic impairments vary greatly between diseases and it is doubtful that a common metabolic intervention will benefit universally. However, improving pathological brain energy metabolism, may aid to restore the metabolic collaboration between neurons and astrocytes, with the ultimate goals of improving synaptic transmission, elevating cognitive function and potentially halting disease progression.

| CON CLUDING REMARK S
Maintaining glutamate and GABA mediated neurotransmission is a complex process relying on multiple cellular components, of which astrocytes play crucial roles. Astrocyte uptake, metabolism and recycling of glutamate and GABA, as well as glutamine synthesis, facilitate the glutamate/GABA-glutamine cycle, and are all essential processes needed to maintain brain function. The specialized features of astrocyte energy metabolism are furthermore inextricably linked to their homeostatic properties and disturbances of astrocyte metabolism may have detrimental outcomes. Although many landmark discoveries on glutamate and GABA function have been uncovered since their identification as neurotransmitters, many aspects still remain to be uncovered. In particular, the functional roles of pathological astrocyte metabolism in relation to glutamate and GABA homeostasis remain elusive. What are the synaptic consequences of altered astrocyte function during pathological development? How does the complex adaptations of astrocyte metabolism affect their cellular function and homeostatic properties? Furthermore, several fundamental features of astrocyte metabolism, and how these may be manipulated to modulate disease progression, remain understudied, including the roles of glycogen, anaplerosis, and citrate release. Exploration of astrocyte metabolism as a therapeutic target is highly warranted and dysfunctional glial energetics needs to be recognized as an important aspect of many neurological diseases. Finally, it should be noted that although astrocytes serve crucial homeostatic roles in the brain, the neuronal interplay with other glial cells, particularly microglia and oligodendrocytes, in both health and disease, should not be neglected and needs to be explored further.

AUTH O R CO NTR I B UTI O N S
The authors jointly designed, wrote and revised this work.

ACK N OWLED G M ENT
JVA would like to sincerely thank the SSADH Association and the Hørslev Foundation for financial support.

FU N D I N G I N FO R M ATI O N
The SSADH Association; The Hørslev Foundation.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors have no conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable since no new data was generated for this Review article.

D ED I C ATI O N
The authors would like to dedicate this work to our past and present teachers and mentors.