Oxidative and nitrosative stress underlie the pathogenesis of a broad range of human diseases, in particular neurodegenerative disorders. Within the brain, neurons are the cells most vulnerable to excess reactive oxygen and nitrogen species; their survival relies on the antioxidant protection promoted by neighbouring astrocytes. However, neurons are also intrinsically equipped with a biochemical mechanism that links glucose metabolism to antioxidant defence. Neurons actively metabolize glucose through the pentose phosphate pathway, which maintains the antioxidant glutathione in its reduced state, hence exerting neuroprotection. This process is tightly controlled by a key glycolysis-promoting enzyme and is dependent on an appropriate supply of energy substrates from astrocytes. Thus brain bioenergetic and antioxidant defence is coupled between neurons and astrocytes. A better understanding of the regulation of this intercellular coupling should be important for identifying novel targets for future therapeutic interventions.
The high vulnerability of the brain to energy and oxidative damage is a contributing factor in the aetiology of neurological and neuroinflammatory disorders [1,2]. In this tissue, neurons are the highest energy-demanding cells, and accordingly they are also the most sensitive to energy stress. For instance, when both neurons and their neighbouring astrocytes are subjected to an identical degree of mitochondrial respiratory chain damage by excess ROS (reactive oxygen species), the neurons rapidly undergo cell death, whereas the astrocytes resist . Why is there this different cellular susceptibility to energy and oxidative stress? Neurons are programmed to metabolize a considerable proportion of glucose through the PPP (pentose phosphate pathway) at the expense of a low glycolytic rate for subsequent energy generation by mitochondria. Key players in this paradigm are astrocytes, whose antioxidant and energy metabolism is coupled with the neuronal energy needs for correct brain function and survival. In the present review, we provide an overview of the key molecular players in this process and their regulation, since these may provide new clues for the treatment of neurological diseases.
REACTIVE OXYGEN AND NITROGEN SPECIES CONTRIBUTE TO NEURODEGENERATION
The evidence obtained from animal models has provided the proof of concept that excess ROS and RNS (reactive nitrogen species) are involved in the neuronal death associated with neurological diseases . Unfortunately, this notion contrasts with the high rate of inefficacy of antioxidants in clinical trials . The reasons accounting for this failure are outside the scope of the present review, but it seems plausible that inappropriate identification of a specific cellular and molecular target, as well as the lack of an adequate pharmaceutical strategy, or late onset of treatment could account for this failure . Thus recent research has aimed to gain a better understanding of the mechanism(s) through which ROS are produced within the brain, as well as how they are eliminated.
The superoxide radical (O2•−), the leading ROS, is physiologically formed by the donation of a single electron to O2, largely in complexes I and III of the mitochondrial respiratory chain via mechanisms that have been reviewed previously . Other sources of O2•− are α-ketoglutarate dehydrogenase, xanthine oxidase-catalysed hypoxanthine oxidation , the phospholipase A2-dependent cyclo-oxygenase and lipoxygenase pathways , and plasma membrane NADPH oxidase . Although O2•− can readily react with certain chemical species, an efficient SOD (superoxide dismutase) [mitochondrial Mn-SOD (manganese SOD) or cytosolic Cu/Zn-SOD (copper/zinc SOD)] inactivates it by dismutation to hydrogen peroxide (H2O2). The Fe2+-catalysed Fenton reaction can convert H2O2 into the highly reactive hydroxyl radical (•OH), although this reaction only has pathogenic relevance when Fe2+ concentrations are high, as in Parkinson's disease for example .
The formation of the above-mentioned ROS in the brain has long been known, but the identification of nitric oxide (•NO) as a brain messenger  has triggered much interest in the roles of free radicals in neurological disease. In neurons, •NO is formed by Ca2+-dependent (constitutively expressed) NOS (nitric oxide synthase) 1  and activates soluble guanylate cyclase to transduce physiological signalling through cGMP . However, excessive glutamate-receptor activation is associated with several brain pathologies, which makes •NO partially responsible for neurotoxicity [12–14] (Figure 1). Glial cells (astrocytes, microglia and oligodendrocytes), which are in close proximity to neurons, express less NOS1 (or none), but they can be induced to activate a Ca2+-independent NOS2 isoform transcriptionally, particularly under pathophysiological situations such as sepsis or ischaemia/reperfusion . The formation of •NO from NOS2 by glial cells can thus be both neuroprotective (by inactivating brain-infiltrated bacteria) or neurotoxic (by affecting key neuronal functions) (reviewed in ). The endothelial cells of the brain microvasculature express Ca2+-dependent NOS3, which responds to cGMP-dependent muscle relaxation and vasodilatation ; endothelial NOS3 can thus play a key role in neuroprotection against stroke . Excellent reviews of the different roles for •NO in brain function and dysfunction are available elsewhere [19,20].
Formation and elimination of ROS and RNS in neurons
One of the fastest reactions of O2•− is the one with •NO, forming the peroxynitrite anion (ONOO−)  (Figure 1). At physiological pH, ONOO− is decomposed into chemical species with •OH-like reactivity and nitrogen dioxide (•NO2) ; however, ONOO− can also nitrate tyrosinyl residues (forming 3-nitrotyrosine) in proteins, altering their functions [23,24]. The formation of ONOO− is best known for its role in the pathogenesis of familial ALS (amyotrophic lateral sclerosis) [25,26], in which mutations in the gene encoding Mn-SOD fail to complete proper O2•− elimination [27,28].
Different molecular mechanism(s) can account for neuronal death after the excess formation of ROS, •NO and/or ONOO− in the brain. Notably, brain mitochondrial complex IV is inhibited by •NO and ONOO−, both reversibly [29,30] and irreversibly [31,32]. The complex IV deficiency  and signs of S-nitrosylation  found in post-mortem brains from Alzheimer's disease patients support such mechanism(s). Finally, whereas physiological glutamatergic synaptic activity releases signalling •NO , excess glutamate, probably occurring in stroke and other neurological disorders, including Alzheimer's disease [35–37], causes mitochondrial energy deficiency and apoptotic neuronal death . The brain bioenergetics in response to the interaction of •NO with mitochondria has been reviewed elsewhere [39,40]. Besides these interferences with proteins, ROS react directly, but less specifically, with many neuronal constituents, including nucleic acids , lipids  and thiols , generally causing brain damage and dysfunction (Figure 1).
ANTIOXIDANT SYSTEMS AND THEIR ROLE IN NEUROPROTECTION
To guard against cell damage, the formation and elimination of ROS must be balanced. Cells are normally equipped with antioxidant scavengers and enzymes to prevent high ROS-mediated damage, but neurons express them at a very low concentration/activity [44–49]. It is therefore imperative that these cells co-operate with neighbouring astrocytes for a complete (and complex) cycle of ROS detoxification and neuroprotection. Below we attempt to decipher how this is achieved.
As mentioned above, O2•− is removed by Cu/Zn-SOD (SOD1, cytosolic) or Mn-SOD (SOD2, mitochondrial), which form H2O2. Deletion of either SOD isoform causes free-radical-mediated neuronal death , indicating the importance of efficient O2•− removal for neuronal survival. However, since SODs produce H2O2 at high concentrations and since this compound is inherently cytotoxic, its elimination is also required for maintaining neuronal survival ; this can be done via different antioxidant systems, most of which require GSH as the cofactor (Figure 1).
GSH (γ-glutamyl-cysteinyl-glycine) is the most abundant mammalian thiol-containing antioxidant . GSH is synthesized in two consecutive ATP-requiring steps, the first of which is rate-limiting and catalysed by GCL (glutamate-cysteine ligase), a heterodimeric enzyme composed of catalytic and modulatory subunits that synthesizes γGC (γ-glutamylcysteine) ; in the second step, GSH is formed by γGC binding to glycine in a GSS (glutathione synthetase)-catalysed reaction (Figure 2). GSH is a required cofactor for the detoxification of peroxides and is more abundant in astrocytes than in neurons [45,51], but the molecular mechanism explaining this difference remains elusive. Although synthesized in the cytosol, GSH can be translocated to other organelles, including mitochondria, where it contributes to defence against free-radical-mediated damage .
Astrocytes co-operate with neurons for antioxidant GSH biosynthesis
Different mechanisms account for peroxide (including H2O2) detoxification. Selenium-containing GPxs (glutathione peroxidases) catalyse the reduction of H2O2 to H2O at the expense of GSH oxidation to its disulfide form GSSG  (Figure 1). Thioredoxins catalyse the reduction of protein disulfide bonds and contribute to H2O2 detoxification by transferring reducing equivalents from NADPH(H+) to H2O2 . This thioredoxin–peroxiredoxin system is up-regulated in neurons by synaptic activity , which is critical for the intrinsic antioxidant activity and survival of neurons. Catalase is a haem-containing enzyme localized in peroxisomes that catalyses the conversion of H2O2 into H2O and O2 , but its presence is dispensable for survival since catalase-knockout mice are viable and fertile ; its expression is very low in neurons . GSTs (glutathione transferases) catalyse the conjugation of GSH with endogenous and xenobiotic electrophiles, thus representing a factor contributing to detoxification [59,60]; the GST isoform π increases in the blood of a mouse model of Parkinson's disease , suggesting a potential role for this enzyme in neuroprotection.
NQO1 [NAD(P)H quinone oxidoreductase 1], which catalyses the reduction and detoxification of highly reactive quinones , is also an O2•− scavenger . Besides acting as an electron transporter between complex I and III of the mitochondrial respiratory chain, CoQ10 (also called ubiquinone) in humans prevents lipid peroxidation and scavenges free radicals through interactions with or without α-tocopherol . Both CoQ10 and its cationic triphenylphosphonium form [MitoQ (mitoubiquinone)], which specifically accumulates in mitochondria, have been shown to exert neuroprotection in animal models of Parkinson's disease [65,66], ALS and Huntington's disease [67,68]. Thus preventing free-radical-mediated lipid peroxidation of mitochondria is neuroprotective; nevertheless, MitoQ was ineffective in a clinical Parkinson's disease trial .
Glutaredoxins are also important for the maintenance of GSH and other thiols in their reduced state. Thus dysfunction of these enzymes causes a dopaminergic neuronal death resembling that seen in Parkinson's disease [70–72]. Finally, aldehyde dehydrogenase 1 detoxifies 4-hydroxy-2-nonenol and malondialdehyde, both highly reactive by-products of lipid peroxidation .
ASTROCYTES ORCHESTRATE BRAIN ANTIOXIDANT DEFENCE
Astrocytes are key players in neurotransmission . Upon synaptic activity, these cells release purines, which activate presynaptic adenosine receptors, thus contributing to further stimulating basal synaptic activity . Furthermore, astrocytes represent the only brain cell type for glycogen storage, notably in areas of high synaptic activity ; in response to neuronal activity, astrocytes support neurons with energy precursors . In fact, astrocytic glycogen mobilization can sustain neuronal activity during hypoglycaemia [78,79]. Astrocytes can activate themselves in central nervous system pathologies, collectively known as reactive astrogliosis , which represents a defence mechanism through an as yet not fully understood signalling pathway . Thus astrocytes are critically important for normal brain function owing to their ability to actively promote neuroprotection.
Possibly the best understood molecular machinery with which astrocytes are equipped for efficient neuroprotection is the antioxidant system (Figure 2). Under resting (non-activated) conditions, these cells express a profuse battery of antioxidant enzymes, including NQO1, both the catalytic and the regulatory subunits of GCL, GPx(s), GSR (glutathione reductase), GST, as well as GSH, and vitamins C and E . The abundance of these antioxidant systems is orchestrated transcriptionally by the relatively high basal activity of Nrf2 (nuclear factor-erythroid 2-related factor 2)  (Figure 2). How Nrf2 is regulated has not yet been fully deciphered, and readers are invited to consult excellent updates addressing this transcription factor [82,83]. However, to gain a better understanding of its role in neuroprotection, we provide below a very brief description of how it works.
In the absence of oxidative stress, Nrf2 remains sequestered in the cytosol, where it is bound to Keap1 (Kelch-like ECH-associated protein 1); several models can be invoked to explain how Keap1 represses Nrf2, but the details of this are outside the scope of the present review . Nrf2 is repressed further by Keap1 binding to Cul3 (Cullin-3), which in turn binds Rbx1 (RING-box protein 1) to form a Keap1–Cul3–Rbx1 E3 ubiquitin ligase complex that targets Nrf2 for proteasomal degradation  (Figure 2). Keap1 contains at least four essential cysteine residues (at positions 23, 151, 273 and 288), the oxidation of one of them (Cys151) determining the release and stabilization of Nrf2 upon oxidative stress. Nrf2 is then translocated to the nucleus, where it forms a complex with other nuclear proteins that binds to the ARE (antioxidant response element)  to induce antioxidant and phase II detoxification enzymes . Among these are NQO1, enzymes related to glutathione metabolism (both the catalytic and regulatory subunits of GCL, GSR and GST), G6PD [G6P (glucose-6-phosphate) dehydrogenase], MRP1 (multidrug-resistance protein 1) and γGT , as well as EAAT3 (excitatory amino acid transporter 3)  (Figure 2).
ASTROCYTES EXERT ANTIOXIDANT NEUROPROTECTION
Although the regulation of Nrf2 specifically in brain cells remains to be elucidated, several recent studies have confirmed its essential neuroprotective role . Nrf2-dependent transcription prevents ROS-induced apoptosis in neurons and astrocytes after several types of insult, such as the presence of H2O2, t-butyl hydroperoxide, 6-hydroxydopamine, 3-nitropropionic acid, MPP+ (1-methyl-4-phenylpyridinium) and rotenone [87–90]. The dopaminergic neurons of the substantia nigra have been shown to be more prone to degeneration by the MPP+ Parkinsonian toxin precursor MPTP (1-methyl-4-phenyl-1,2,5,6-tetrahydropyridine) in Nrf2-null mice than in the wild-type . Nrf2 is protective against intrastriatal administration of the complex II inhibitors malonate and 3-nitropropionic acid [92,93], 6-hydroxydopamine [94,95], and also protects against cerebral ischaemia [96,97]. Moreover, the expression of Nrf2 under the control of the glial fibrillary acidic protein promoter on an Nrf2−/− background abolishes the dopaminergic neuronal death triggered by MPTP [92,95,98]. These studies elegantly highlight the importance of astrocytic Nrf2 for neuronal protection against bioenergetic and oxidative stress-mediated damage. However, whether astrocytic Nrf2-mediated induction of antioxidant systems is mandatory for neuronal protection is still controversial. Thus astrocytes treated with very mild H2O2 concentrations unable to activate Nrf2 supported neuronal protection, in co-culture, against hypoxia ; in contrast, in another study astrocytic Nrf2 activation was found to be necessary . Thus it remains unclear whether neurons necessarily require astrocyte antioxidant protection, even though the vast majority of Nrf2 activation seems to occur in astrocytes [73,101].
It has been observed that some subpopulations of neurons are more susceptible to oxidative stress damage than others [102–105]; thus, in addition to astrocyte-mediated neuroprotection, neurons rely on their own antioxidant abilities. In this context, Nrf2 has been shown to be more active in CA1 neurons, and hence its targets such as NQO1, as revealed by transcriptomics . In cell culture, in ex vivo organotypic brain slice cultures, and in acute brain slice preparations, cerebellar granule and hippocampal CA1 neurons have been shown to be significantly more sensitive to oxidative stress than cerebral cortical and hippocampal CA3 neurons . This has also been observed following glutamate excitotoxic insult, ischaemia or β-amyloid-induced neurotoxicity [107–110]. A comparative transcriptome analysis of vulnerable compared with resistant neurons revealed a higher expression of genes related to stress and the immune response, and lower expression of energy generation and signal transduction genes in the former .
Besides Nrf2, the regulation of antioxidant gene expression can be governed by PGC-1α (peroxisome-proliferator-activated receptor-γ co-activator-1α), a transcription co-activator that is important for mitochondrial biogenesis and metabolism . The neurons of the substantia nigra and hippocampus of PGC-1α-null mice are the most vulnerable against the MPTP and kainic acid models of neurodegeneration , and increased PGC-1α levels dramatically protect neural cells in culture from oxidative stress . In Huntington's disease, PGC-1α expression and function are impaired [113–115]. Finally, PGC-1α promotes the non-amyloidogenic processing of amyloid precursor protein, hence preventing the generation of neurotoxic β-amyloid peptides . Taken together, these results suggest that PGC-1α is a potential target in neurodegenerative diseases.
THE ANTIOXIDANT GLUTATHIONE GOVERNS NEURONAL SURVIVAL
The cellular antioxidant reserve governs the differential vulnerability of brain cells in human neurological diseases . Thus the GSH-dependent peroxide-detoxifying system is less efficient in neurons than in astrocytes , and this explains the high resistance of astrocytes to •NO- and ONOO−-mediated mitochondrial damage and death [47,118]. The GSH concentration is approximately half of that found in astrocytes [45–47]; this appears to be due to the ~10-fold lower protein abundance and enzymatic activity of GCL in neurons in comparison with astrocytes . Experiments in which neurons and astrocytes were incubated with identical concentrations of ONOO−  or H2O2  revealed GSH depletion in neurons, but not in astrocytes. However, the inhibition of GCL activity in astrocytes with L-buthionine sulfoximine decreases GSH and sensitizes them to ONOO− . In contrast, overexpression of GCL (catalytic subunit) increases GSH concentrations in neurons and protects them against excitotoxic-mediated damage . Thus the cellular content of GSH is critical in neuroprotection, but how is GSH metabolism co-ordinated in the brain?
Astrocytes release GSH  as well as the GSH precursors cysteine  and cysteinylglycine  into the extracellular space (Figure 2). Astrocyte-released cysteine and cysteinylglycine are both taken up by neurons and may be used for GSH biosynthesis [45,51,122] (Figure 2). Nrf2-driven GSH biosynthesis and release from astrocytes protects neurons from oxidative stress . In experimental in vivo models of Parkinson's disease and ALS, Nrf2 overexpression in astrocytes exerts neuroprotection [98,123]. Astrocytes are therefore necessary for maintaining neuronal GSH; however, they may not be sufficient. Thus neurons require their own enzymatic glutathione-synthesizing machinery to be intact for the glutathione-peroxide-detoxifying system to be efficient  (Figure 2). GCL knockdown by RNA interference in primary cortical neurons spontaneously triggers GSH depletion followed by apoptotic death, even in the presence of astrocytes in co-culture . Moreover, tyrosine hydroxylase promoter-driven expression of an antisense mRNA against GCL in transgenic mice causes the death of dopaminergic neurons , and, in vivo, stereotaxic injection of viral particles expressing a small hairpin RNA against GCL triggers dopaminergic neuronal death and motor defects , indicating that de novo endogenous GSH biosynthesis in neurons is required for their survival, even in the presence of neighbouring astrocytes.
GLUCOSE OXIDATION IN THE PPP IS COUPLED WITH GSH REGENERATION IN NEURONS
In non-stressed cells, the GSH/GSSG ratio is maintained at a value above ~20 , which is achieved by an efficient reduction of GSSG to GSH by GSR  (Figure 1). This reaction requires reducing equivalents in the form of the cofactor NADPH(H+). Therefore the ability of cells to maintain the NADPH(H+)/NADP+ ratio at a sufficiently high level is a factor that critically influences the GSH/GSSG redox status and hence the overall antioxidant status [48,127]. Different enzymatic systems are known to contribute to maintaining the NADPH(H+)/NADP+ redox status , e.g. isocitrate dehydrogenase (NADP-dependent), which is expressed both in mitochondria and the cytosol, NADP+-dependent cytosolic malate dehydrogenase (or malic enzyme)  and the PPP. In terms of efficiency, the PPP largely accounts for the vast majority of cellular NADPH(H+) regeneration .
The PPP oxidizes G6P in three consecutive steps, catalysed by G6PD, which forms 6PGL (6-phosphogluconolactone); 6PGL is then converted into 6PG (6-phosphogluconate) by a lactonase, followed by its decarboxylation into R5P (ribulose 5-phosphate) by 6PGD (6PG dehydrogenase)  (Figure 3). Two of these reactions (G6PD and 6PGD) occur at the expense of 1 mol of NADPH(H+) regenerated per mol of substrate each, and constitute the oxidative branch of the PPP. Thus this branch of the PPP can regenerate 2 mol of NADPH(H+) per mol of G6P entering the pathway, with the loss of one carbon atom as CO2. This branch is followed by a series of reactions (the non-oxidative branch) that can yield, from 2 mol of R5P, 1 mol of the glycolytic intermediates F6P (fructose 6-phosphate) and G3P (glyceraldehyde 3-phosphate)  (Figure 3). However, depending on the metabolic status of the cell, once isomerized into ribose 5-phosphate, R5P can be used for the biosynthesis of nucleotides . In neurons, the activity of the non-oxidative PPP branch seems to be very low , suggesting that most G6P entering the PPP is not recycled. This can be explained in terms of the fact that DNA repair is a particularly active process in post-mitotic neurons , which require a large pool of ADP-ribose, which in turn needs the PPP intermediate ribose 5-phosphate.
Neuronal glucose metabolism supports antioxidant GSH regeneration
The rate-limiting step of the PPP is the G6PD-catalysed reaction , and there is a large body of evidence to support the antioxidant and cytoprotective roles of the PPP. In early studies, G6PD was found to be up-regulated by GSSG . In cortical neurons, exogenous H2O2 stimulates PPP activity and regenerates NADPH(H+) and GSH, thus contributing to neuroprotection . By stimulating glutamate receptors, the endogenous oxidation of GSH to GSSG in neurons triggers an increase in the rate of glucose oxidation through the PPP, leading to NADPH(H+) and GSH regeneration and neuroprotection . At low doses, ONOO− activates G6PD, causing an enhancement of neuroprotective PPP activity in neurons . In the brains of Alzheimer's disease patients, it has long been recognized that glucose metabolism is altered  and β-amyloid causes an enhancement in the flux of glucose utilization via the PPP . Interestingly, increased G6PD levels have been found in the surviving pyramidal neurons of hippocampal slices from the post-mortem brains of Alzheimer's disease patients [138,139]. In addition, in a cybrid model of Huntington's disease, the bioenergetic dysfunction is seen as stimulated glycolysis, inhibition of the PPP and decreased levels of mitochondrial NADH(H+)/NAD+ . Together, these findings strongly suggest that, during oxidative and nitrosative stress, as well as in human brain pathologies, the consumption of glucose through the PPP is critical for maintaining the neuronal antioxidant redox status and survival. However, is this phenomenon co-ordinated with other glucose-consuming pathways, e.g. glycolysis?
THE RATE OF GLYCOLYSIS IS CONTROLLED DIFFERENTIALLY IN NEURONS AND ASTROCYTES
Measurements of the rate of [6-14C]glucose oxidation to 14CO2, which takes place mainly in the tricarboxylic acid cycle, have shown that in cortical neurons it is very low in comparison with astrocytes . Since glucose oxidation in the tricarboxylic acid cycle requires the previous conversion of glucose into pyruvate, these results suggest that the rate of glycolysis is lower in neurons than in astrocytes. This observation has been confirmed using different experimental approaches, such as determination of the rate of 3H2O formation from [3-3H]glucose, which takes place exclusively through the glycolytic enzyme aldolase, the rate of 14C-lactate formation from [U-14C]glucose  and the accumulation of intra- or extra-cellular lactate [47,135]. Thus, in comparison with astrocytes, neurons consume very little glucose through the glycolytic pathway.
The mechanism responsible for the differential activity of the glycolytic pathway in these brain cells has been studied. When complex IV of the mitochondrial respiratory chain is inhibited by •NO, astrocytes respond by stimulating AMPK (AMP-activated protein kinase), a well-established cell energy sensor , and increasing the rate of glycolysis . Moreover, AMPK is required for the activation of glycolysis, which takes place through the stimulation of PFK1 (6-phosphofructo-1-kinase) . PFK1 is potently activated by F2,6P2 (fructose 2,6-bisphosphate), which is synthesized by the bifunctional enzyme PFKFB (6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase)  (Figure 3). In the brain, the most abundant PFKFB isoform is PFKFB3, which expresses a very high kinase/bisphosphatase activity ratio . Astrocytes express PFKFB3 abundantly and AMPK is required for the activation of PFKFB3-mediated glycolysis by •NO . In contrast, inhibition of complex IV by •NO in neurons does not result in this AMPK–PFKFB3 axis and therefore they do not activate glycolysis . With the notable exception of cerebellar neurons , synaptosomes [145,146] and retinal neurons [147,148], promoting endogenous mitochondrial stress in cortical neurons by stimulating glutamate receptors [38,149] does not result in a concomitant activation of glycolysis [135,150] or glucose uptake . Thus the same stimuli that efficiently up-regulate the rate of the PPP are unable to increase that of glycolysis. Why is the glycolytic pathway so tightly controlled in neurons? Is the regulation of both glucose-consuming pathways co-ordinated to promote antioxidant defence?
GLYCOLYSIS AND THE PPP ARE CO-ORDINATED TO MAINTAIN THE NEURONAL ANTIOXIDANT REDOX STATUS
Glycolysis and the PPP are two glucose-metabolizing pathways that are connected at G6P, a common metabolite of both pathways  (Figure 3). Therefore the notion that the regulation of G6P consumption through glycolysis and the PPP is inversely correlated is a straightforward hypothesis. In cortical neurons, the protein abundance of PFKFB3 is very low when compared with that found in astrocytes ; however, overexpressing PFKFB3 in neurons increases the rate of glycolysis and concomitantly decreases that of the PPP . Moreover, co-expression of G6PD, the rate-limiting enzyme of the PPP, reverses this effect . Thus PFKFB3 protein levels appear to be a critical factor controlling both the rate of glycolysis and that of the PPP (Figure 3).
In neurons, PFKFB3 is constantly subjected to proteasome degradation after ubiquitylation by the E3 ubiquitin ligase APC/C (anaphase-promoting complex/cyclosome) and its adaptor Cdh1  (Figure 3). Thus the activity of the APC/C–Cdh1 complex is a determinant in controlling the protein levels of PFKFB3 and hence the rate of glucose consumption through glycolysis and the PPP. In neurons, the function of APC/C is strictly dependent on the presence of Cdh1, which is very abundant in these cells ; in contrast, astrocytes express very low Cdh1 levels and therefore APC/C activity is negligible in these cells . Accordingly, PFKFB3 protein levels are high and glycolysis is active in astrocytes owing to the low levels of Cdh1. In fact, overexpression of Cdh1 in astrocytes destabilizes PFKFB3 and concomitantly decreases the rate of glycolysis . Conversely, the knockdown of Cdh1 in neurons results in the inhibition of APC/C–Cdh1 activity, leading to PFKFB3 protein accumulation, shifting glucose consumption towards glycolysis at the expense of a reduction in that of the PPP. This causes an impairment of neurons to properly regenerate NADPH(H+), hence promoting oxidative stress by GSH oxidation to GSSG . Neurons are deficient in GSH and de novo GSH-synthesizing enzymes; it is therefore reasonable that they would compensate for this deficiency by promoting glucose utilization through the PPP.
This tight regulation of glycolysis and the PPP has important consequences for neuronal survival. Thus, when the activation of the rate of glycolysis is prolonged by inhibiting APC/C–Cdh1 activity, neurons undergo apoptotic death . Moreover, neuronal death can be fully reversed by incubation with the plasma-membrane-permeant GSH derivative GSH ethyl ester . Taken together, these results strongly suggest that under normal circumstance in neurons glucose is preferentially utilized through the PPP to exert antioxidant-mediated neuroprotection.
BIOENERGETICS AND ANTIOXIDANT DEFENCE COUPLING BETWEEN NEURONS AND ASTROCYTES
Neurons constantly require energy to sustain glutamatergic neurotransmission; however, they dedicate a considerable proportion of glucose to PPP oxidation to preserve the antioxidant redox status and neuroprotection (Figure 4). Since this takes place at the expense of lowering the amount of glucose entering glycolysis for energy generation, how then is energy produced in neurons? During glutamate neurotransmission, astrocytes play a key role by taking up excess glutamate from the synaptic cleft  (Figure 4); glutamate uptake is an Na+-dependent (energy-consuming) secondary co-transport phenomenon. Thus neurotransmission progresses with increased energy demands in astrocytes, which obtain energy by stimulating glycolysis and converting pyruvate into lactate [153,154] (Figure 4). Lactate is released into the extracellular space and is taken up by neurons, which convert it into pyruvate, followed by mitochondrial oxidation and energy generation [155,156] (Figure 4). This coupling mechanism is compatible with work showing that neuronal uptake and the utilization of lactate is necessary for long-term memory formation [77,157,158]. Although the astrocyte–neuronal lactate shuttle has been contested , recent assessments have provided clear support not only for its existence, but also for its prominent role [76,160]. In any case, in view of the inability of neurons to obtain energy from glucose, it seems reasonable that it must be compensated by an exogenous energy substrate, possibly derived from astrocytes.
Astrocytes and neurons couple glucose metabolism to antioxidant defence
Accordingly, the energy of neurons seems to be supported by neighbouring astrocytes, and this metabolic coupling is essential for the maintenance of the neuronal antioxidant status. Moreover, the antioxidant status of neurons also depends on astrocytes, which supply neurons with GSH precursors through Nrf2-dependent and -independent processes [99,100] (Figure 4). Thus it appears that neurons are strictly dependent on astrocytes for their antioxidant redox status and survival. However, studies in astrocyte-free neurons suggest that PPP-mediated NADPH(H+) and GSH regeneration is sufficient to exert neuroprotection . It remains to be established whether there is a redundancy in the antioxidant pathways, or whether the PPP-mediated antioxidant function represents a short-term attempt to circumvent the time-consuming transcriptional activation of an Nrf2-mediated GSH supply from astrocytes.
The brain is a particularly vulnerable tissue to bioenergetic, oxidative and nitrosative stress, although active neuroprotective machinery continuously fights against neuronal damage. Astrocytes play a pivotal neuroprotective role. These cells are self-protected by robust antioxidant equipment co-ordinated by Nrf2, and they also provide neurons with both energy substrates and antioxidant GSH precursors. However, although astrocytes are necessary for neuronal protection, recent results have suggested that neurons would have additional mechanisms that couple glucose consumption with antioxidant protection. Thus, by actively sorting glucose consumption through the PPP, neurons obtain NADPH(H+) to regenerate GSH. This is accomplished by APC/C–Cdh1 activity, which constantly degrades the glycolytic-promoting enzyme PFKFB3 in neurons. Future studies focused on understanding the physiological regulation of this pathway are planned. This would afford a huge boost to our understanding of the molecular mechanisms underlying neurological disorders, as well as our search for novel therapeutic strategies against brain oxidative damage.
amyotrophic lateral sclerosis
AMP-activated protein kinase
Kelch-like ECH-associated protein 1
multidrug-resistance protein 1
NAD(P)H quinone oxidoreductase 1
nitric oxide synthase
nuclear factor-erythroid 2-related factor 2
peroxisome-proliferator-activated receptor-γ co-activator-1α
pentose phosphate pathway
RING-box protein 1
reactive nitrogen species
reactive oxygen species
This work was supported by the FEDER (European Regional Development Fund); the Ministerio de Ciencia e Innovacion [grant numbers SAF2010-20008 and Consolider-Ingenio CSD2007-00020); the Instituto de Salud Carlos III [grant numbers PS09/0366 and RD06/0026/1008]; and Junta de Castilla y León [grant number GREX206].