Introduction

I am honoured to contribute this short review on cerebral pyruvate recycling for the special issue of Neurochemical Research dedicated to the scientific trajectory of Prof. Ursula Sonnewald. Cerebral pyruvate recycling represented a fascinating aspect of the early activity of our laboratories and became the matter of many unforgettable conversations worldwide. This topic, initially restricted to few 13C Nuclear Magnetic Resonance Spectroscopy (NMR) specialists, extended eventually to the neurochemical community and ultimately framed the applications of high resolution 13C NMR to Neurochemistry. The following paragraphs summarize some of my personal recollections on the initial events leading to the detection of the pathway, the controversies that arose later on its cellular localization in the brain and some considerations on its possible functions.

13C NMR studies of metabolism were pioneered by Eakin et al. when investigating, already in 1972, the metabolism of (1-13C) glucose in yeast preparations [1]. The results obtained, made these authors envision that “13C NMR could have numerous applications for in vivo metabolic studies”. Indeed, 13C NMR entailed important advantages that would make it later become the leading tool for metabolic studies [2]. Briefly, due to the low natural abundance of 13C (1.1% 13C), 13C NMR provided a convenient method to detect the incorporation of the stable isotope 13C, from suitable 13C enriched substrates (99.9% 13C) into metabolic intermediates and products. This wealth of information could be gathered not only in vitro, but even in vivo, making soon 13C NMR to become the dominant technique in studies of cerebral metabolism [3, 4]. However, despite its important advantages, the 13C NMR method was not devoid of limitations, its reduced sensitivity (in the mM range) representing, most probably, the largest. In fact, few scientists thought in the late 1980´s of the last century, that 13C NMR could improve the results provided earlier by the highly sensitive radioactive isotope approach. This occurred even if 13C NMR avoided the complexities of extraction, isolation and chemical degradation of the metabolites investigated, as imposed by adequate radioactive counting [5].

Even though its inherent sensitivity limitations, high resolution 13C NMR entailed the unique advantage of enabling to explore the path of groups of contiguously enriched 13C carbons, from a multiply labeled 13C enriched substrate, into metabolic intermediates and products [6]. Notably, 13C NMR could easily distinguish if a specific 13C carbon within a metabolite molecule was not vicinal to any 13C labeled neighbors, flanked by one 13C, or two 13C carbon atom neighbors [7]. Metabolites having the same chemical structure, but different distributions of 13C atoms within their carbon skeleton, were named “isotopic isomers” or “isotopomers”. Different “isotopomers” originated characteristic homonuclear 13C–13C spin coupling patterns in the 13C resonances from every carbon, revealing the relative isotopomer contributions after appropriate spectral analysis. This exceptional advantage became later instrumental for the applications of high resolution 13C NMR to neurochemistry in general and, for the detection of the cerebral pyruvate recycling system, in particular.

Cerebral Pyruvate Recycling In Vivo

Early 1989, during a postdoctoral period in Basel (CH), we explored the cerebral metabolism of the simplest molecule containing two contiguously 13C enriched carbons, namely 13C2 acetate [8]. We infused 13C2 acetate in the right jugular vein of fed rats and monitored, in vivo and in vitro, the fate of the two vicinally 13C enriched carbons. In vivo 13C NMR spectroscopy was not sufficiently developed at the time, and we could only detect in vivo, resonances from the infused substrate and 13C bicarbonate. However, the analysis of 13C resonances from glutamate and glutamine carbons in perchloric acid extracts obtained from the brain after 13C2 acetate infusion, became much more informative. Indeed, high resolution 13C NMR multiplet analysis of brain extracts yielded a wealth of metabolic information on cerebral metabolism and provided the first evidence of cerebral pyruvate recycling.

Figure 1a shows an expanded view of the 13C resonances from the C4 carbons of cerebral glutamate (resonance #9) and glutamine (resonance #8) in these extracts, and their deconvolution into the corresponding contributions of (4-13C), (4,5-13C2), and (3,4,5-13C3) isotopomers, represented by the singlet (Fig. 1b), doublet (Fig. 1c), and doublet of doublets (Fig. 1d) resonances, respectively. First, 13C incorporation was larger in glutamine C4 than in the glutamate C4 carbons, as reflected by the larger intensity of the glutamine resonances. This result was inconsistent with the strict precursor-product relationships, precluding the glutamine product to incorporate more 13C label than its glutamate precursor. The observation confirmed previous reports on the metabolism of 14C acetate, revealing the presence of “large” and “small” glutamate pools and the predominant labelling of the “small” glutamate pool and glutamine by 14C acetate [9–11]. The finding was not new then, but became very relevant for further studies, since it validated the recent 13C NMR methodology with earlier 14C results, proving that both approaches yielded similar outcomes. Indeed, the “small” and “large” glutamate pools had been reported to reflect metabolism in the glial and neuronal compartments, respectively [10, 12, 13]. The “small” glutamate pool, was associated with a “fast turnover” tricarboxylic acid (TCA) cycle, thought to occur in the glial compartment and generating most of the cerebral glutamine, a finding later supported by reports showing the predominant acetate transport to glial cells [14]. Conversely, the “large” glutamate pool was associated with a “slow turnover” TCA cycle occurring in the neuronal compartment and originating the largest proportion of cerebral glutamate [10]. It was then very surprising, to find that a high resolution 13C NMR spectrum from a rat brain extract, obtained after infusion with 13C2 acetate, could include enough information to summarize decades of previous research with radioactive isotopes, a finding supporting further applications of the 13C NMR method.

Fig. 1
figure 1

Analysis of the 13C labelling patterns of the C4 carbons from glutamate (resonance #9) and glutamine (resonance #8) from brain extracts obtained after (1,2-13C2) acetate infusion. a Experimental spectrum, b relative contributions of (4-13C) glutamate and (4-13C) glutamine to the corresponding C4 multiplets, c relative contributions of (4,5,-13C2) glutamate and glutamine, d relative contribution of (3,4,5-13C3) glutamate and glutamine, e simulated spectrum. Reproduced from ref. [8] with permission

Full size image

The C4 and C5 resonances of glutamate and glutamine in particular, became highly informative, since they reflected directly the 13C isotopomer distributions of the C2 and C1 carbons of acetyl-CoA molecules, entering the neuronal or glial tricarboxylic acid cycles, respectively. We were amazed to find, in addition to the expected glutamate and glutamine C4 doublet resonances derived directly from infused (1,2-13C2) acetate (Fig. 1a), a central singlet (full arrow), derived from monolabeled (4-13C) glutamate and (4-13C) glutamine isotopomers. Similar 13C multiplet patterns were found in the C5 carbons of glutamate and glutamine, as expected (not shown). Notably, the central singlets in glutamate and glutamine C4 and C5 resonances revealed that (2-13C) and (1-13C) acetyl-CoA had been produced during oxidative metabolism of 13C2 acetate, reentering later the TCA cycle. The mechanism by which these monolabeled forms of acetyl-CoA were originated from 13C2 acetate, disclosed the presence of the cerebral pyruvate recycling system. Since we (and others) did not observe monolabeled forms of acetate in plasma during metabolism of 13C2 acetate in fed rats or humans, we concluded that the monolabeled forms of acetate were originated mainly in the brain [8, 15]. Indeed, monolabeled (1-13C) or (2-13C) acetyl-CoA, could be produced either from (1,2-13C2) and (3-4-13C2) oxaloacetate or from (1,2-13C2) and (3-4-13C2) malate, through phosphoenolpyruvate carboxykinase (PEPCK) and pyruvate kinase (PK), or malic enzyme (ME) activities, respectively. The relative contribution of these enzymes will be addressed later.

Further analysis of the glutamate and glutamine C4 multiplets, provided information on the compartmentation of the pyruvate recycling pathway. Figure 1b–e show the complete deconvolution of the glutamate and glutamine C4 multiplets, in the contributions of individual C4 resonances from the corresponding (4-13C) (Fig. 1b), (4,5-13C2) (Fig. 1c) and (3,4,5-13C3) isotopomers (Fig. 1d). A simulated addition of all contributions (Fig. 1e) shows remarkable similarity with the experimental spectrum (Fig. 1a), supporting an adequate deconvolution. Notably, the relative contribution of (4-13C) glutamate (29%) was much larger than that of (4-13C) glutamine (8%). This finding revealed that the pyruvate recycling was mainly located in a compartment with lower (or absent) glutamine synthase activity. Otherwise the relative contributions of (4-13C) isotopomers in glutamate and glutamine should be equal, in an non-compartmented system with a homogenous distribution of the enzyme. Since glutamine synthase was shown to be exclusively located in the glial compartment [16], we proposed that the monolabeled forms of acetyl-CoA and therefore the pyruvate recycling system was mainly located in the “large” glutamate compartment, a circumstance accounting for the reduced (or absent) formation of (4-13C) glutamine from (4-13C) glutamate (Fig. 2a). In fact, it became possible later to demonstrate the preferential location of the enzymes involved in the synaptosomal fraction obtained by centrifugal fractionation of adult brain homogenates in Percoll density gradients (Fig. 2b) [17].

Fig. 2
figure 2

Early evidences on the compartmentation of the cerebral pyruvate recycling in the adult rat brain. a Cerebral tricarboxylic acid cycles of the neuronal (TCA cycle 2) and glial (TCA cycle 1) compartments during 13C2 acetate metabolism. The glial cycle metabolizes 13C2 acetate (black filled circles), while the neuronal cycle metabolizes mainly unlabeled acetyl-CoA (open circles). Glutamate, glutamine and GABA are exchanged between both TCA cycles. The different symbols (crossed circles and open circles) in OAA1 and OAA2 indicate that different 13C enrichments may occur. The absence of glutamine synthase in the neuronal compartment, precludes the synthesis of neuronal glutamine from neuronal glutamate. This accounts for the smaller contribution of monolabeled forms of acetate to glutamine labelling (Fig. 1a–b), suggesting a dominant contribution of the recycling pathway in the neuronal TCA cycle 2. Reproduced from ref [8] with permission. b Synaptosomal localization of the enzymatic activities involved in the recycling of pyruvate in the adult rat brain in a Percoll density gradient (Cruz, F. and Cerdán, S., unpublished observations). ME malic enzyme, PK pyruvate kinase, PEPCK phosphoenolpyruvate carboxykinase

Full size image

We investigated later if the cerebral pyruvate recycling system was active with other infused ketogenic or gluconeogenic substrates, as (U-13C4) hydroxybutyrate or (1,2-13C2) glucose [18]. The glutamate C4 and C5 multiplets of glutamate and glutamine observed in acid extracts of the brain after these infusions, depicted similar structures to those depicted in Fig. 1, in particular the singlets (Fig. 1b). Notably, both precursors yielded (1,2-13C2) acetyl-CoA and its recycled (1-13C) and (2-13C) acetyl-CoA isotopomers, indicating that recycling of (1,2-13C2) acetyl-CoA units, was not limited to 13C2 acetate. However, the relative contribution of the pyruvate recycling system to acetyl-CoA oxidation was smaller with (U-13C)-hydroxybutyrate (17%) or with (1,2-13C2) glucose (5%) than with 13C2 acetate (27%), suggesting that the ketogenic or gluconeogenic nature of the precursor, had an important influence on the relative contribution of the recycling pathway to acetyl-CoA oxidation. More oxidized substrates as 13C2 acetate or (U-13C4) β-hydroxybutyrate, depicted larger relative contributions of recycled, monolabeled forms of acetate, than more reduced substrates as (1,2-13C2) glucose. This suggested that the intracellular redox state of the brain, as determined by the substrate infused, played an important role in the regulation of the pyruvate recycling pathway.

Interestingly, both 13C2 acetate and (U-13C4) hydroxybutyrate originated small amounts of (2,3-13C2) lactate, derived from (3,4-13C2) oxaloacetate and malate, through PEPCK and PK or malic enzyme activities, suggesting that the (2,3-13C2) pyruvate produced intramitochondrially by these reactions, was transported into the cytosol, becoming the substrate for lactate dehydrogenase [18–21]. This finding indicated that recycling was not limited to the mitochondrial compartment hosting the TCA cycle, but disclosed an additional cytosolic component resulting in recycled lactate production.

The precise mechanism of the recycling remained uncertain, since different enzymes as PEPCK and PK or ME, could originate from the same 13C labelling pattern in the C4 (and C5) carbons of glutamate and glutamine. To overcome this difficulty, we designed an independent approach comparing the developmental profiles of these enzymes to that of the recycling pathway [17]. We measured by 13C NMR the change in relative pyruvate recycling activity in the brain of rats with increasing ages, from postnatal day 0 (P0) to postnatal day 90 (P90, Fig. 3a–b), comparing it with the developmental profile of the enzymatic activities of PEPCK (Fig. 3c), ME (Fig. 3d) and PK (Fig. 3e). The pyruvate recycling system developed in parallel with myelination, peaking at day P21 and maintaining the same activation until day P90. While PEPCK and PK maintained similar activities along development, ME activity increased in parallel to the activity of the pyruvate recycling system, suggesting that ME contributed the dominant enzymatic mechanism of this pathway.

Fig. 3
figure 3

Cerebral pyruvate recycling after infusion of 13C2 acetate in rats of increasing ages P0-P90. a High resolution 13C NMR spectra of acid extracts from rat brain during development after administration of (1,2-13C2) acetate. Arrow Glutamate C4 singlet revealing the relative contribution recycled (2-13C) acetyl-CoA. b Developmental changes in the relative activity of the cerebral pyruvate recycling system as detected in the singlet/doublet ratio of Glutamate C4, c–e Developmental changes in the enzymatic activities of PK, ME and PEPCK. Reproduced from ref. [17] with permission

Full size image

The localization of malic enzyme in brain had been extensively investigated at the time, by the Hamprecht´s and McKenna´s groups [22–24]. Brain homogenates contained 45% of the cytosolic enzyme (MEc), while astroglial cultures contained 95% of MEc, suggesting that the cytosolic enzyme was primarily located in astrocytes [23]. In contrast, inmunocytochemical staining of rat brain sections with antibodies prepared against the mitochondrial enzyme (MEm), stained primarily the neurons [22]. Moreover, ME activity was much higher in mitochondria from synaptic terminals, than in mitochondria prepared from primary cultures of cortical neurons or cerebellar granule cells [24]. In summary, both neurons and glial cells were equipped with malic enzyme, albeit with different subcellular localizations, the mitochondrial enzyme from neurons providing a slightly larger contribution to total malic enzyme activity in the adult rat brain. This subcellular distribution of malic enzymes matched well the cytosolic and mitochondrial components of the pyruvate recycling system detected in vivo, suggesting that mitochondrial MEm present in neurons could originate the mitochondrial pyruvate recycling, while the cytosolic MEc could provide the cytosolic recycling. However, it remained to be explained why the recycling detected in vivo during the metabolism of 13C2 acetate was much smaller in glutamine than in glutamate, representing the glial and neuronal compartments, since the total enzymatic activities of ME in these compartments were relatively similar, as measured by enzymatic methods.

We thought then that the relative sizes of the neuronal and glial pools of glutamate and glutamine, could underlie a differential sensitivity of 13C NMR to detect the neuronal or glial metabolisms of 13C2 acetate in vivo. Traditionally, the determination of the pool sizes of the neuronal and glial glutamate pools in vivo had remained very challenging, normally approached by mathematical modelling and immunocytochemistry [10, 12, 25, 26]. To tackle this issue experimentally, we implemented a novel strategy using both, the 13C isotope and an additional isotope, namely deuterium [27]. We infused either solutions of (2-13C, 2-2H3) acetate or mixtures of (2-13C, 2-2H3) acetate and (1-13C) glucose in the right jugular vein of rats. The presence of 13C-2H coupling patterns, only in those acetyl-CoA molecules derived from deuterated acetate, made it possible to distinguish clearly between the (2-13C, 2-2H3) acetyl-CoA molecules derived from deuterated acetate, and the (2-13C) acetyl CoA molecules derived from non-deuterated glucose. This allowed for the direct determination of the neuronal and glial pool sizes of glutamate, by solving a simple set of simultaneous equations involving mass and 13C enrichment conservation. We found that the neuronal and glial pool sizes of glutamate, were drastically different during (2-13C, 2-2H3) acetate or acetate + (1-13C) glucose infusions. Notably, the glial pool size of glutamate increased from 8.6 ± 2% during glucose and acetate infusions, to 73 ± 8% during acetate infusions, suggesting that glial processes affecting the glutamate C4 resonances would become much more difficult to detect with mixtures of glucose and acetate than with acetate as the only substrate. However, even under these unfavorable conditions, we were able to measure a small contribution of glial pyruvate recycling by detecting recycled (4-13C, 4-2H) glutamine, indicating that both neuronal and astrocytic pyruvate recycling systems could be detected in vivo with this approach. Scafidi et al. [28] confirmed these findings by detecting metabolism of the acetyl moiety from acetylcarnitine (ALCAR) both in astrocytes and GABAergic neurons.

Serres et al. [29] investigated later the effects of metabolism of mixtures of (2-13C) acetate and glucose and acetate and (1-13C) glucose, in awake or anesthetized rats, using proton observed 13C decoupled spectroscopy (POCE). Briefly, they found that different anesthetics did not induce important alterations in the activity of the pyruvate recycling system. Additionally, 13C fractional enrichments of the protonated carbons of glutamate and glutamine, as measured in the brain, were lower than those of their precursors measured in plasma. On these grounds, these authors proposed that recycled isotopomers of pyruvate in the brain could be mainly derived from peripheral tissues. However, the precise contribution of intracerebral substrates could not be established and POCE did not allow the analysis of the relative contributions of non-protonated glutamate isotopomers. Therefore, a possible underestimation of the contribution of intracerebral recycling could not be discarded. In spite of this, the experiments provided a relevant note of caution, suggesting that the contribution of peripheral tissues, to the pyruvate recycling system observed in the in vivo brain, could be larger than previously thought.

Cerebral Pyruvate Recycling Ex Vivo

The Sonnewald´s and Schouboe´s groups had implemented an extensive collection of experiments to characterize cerebral pyruvate recycling both in vivo, and in vitro, using primary cultures of neural cells enriched in neurons or astrocytes [19, 20, 30–35]. The pyruvate recycling pathway was detected initially in primary cultures of astrocytes but not in neurons [20, 32]. Briefly, astrocytes cultures incubated with (U-13C5) glutamine, showed the presence of (4,5-13C2) glutamate, an isotopomer requiring the formation of (U-13C3) pyruvate and (13C2)-acetyl-CoA and its incorporation into unlabeled oxaloacetate in the glial TCA cycle [19]. Subsequent experiments used (3-13C) glutamate incubations with primary cultures of cerebrocortical astrocytes, GABAergic cerebrocortical interneurons and co-cultures of both cell types [32]. (4-13C) glutamate derived from recycling could only be detected in astrocyte cultures, suggesting that cultivated astrocytes, but not cultivated neurons, were endowed with the pyruvate recycling capacity. Lastly, pyruvate recycling could be demonstrated, both by gas chromatography-mass spectrometry and 13C NMR, in primary cultures enriched in cerebellar neurons, confirming that the recycling pathway was active both in cell cultures from neurons and astrocytes, as well as in vivo [31], a finding consistent with the presence of mitochondrial and cytosolic malic enzymes in neurons and astrocytes, respectively [22–24].

Relevance of Cerebral Pyruvate Recycling

An important aspect relates to the physiological functions of cerebral pyruvate recycling. First, recycled carbons from substrates other than glucose, may provide a useful alternative to maintain operative the vital TCA cycle when glucose supplies are limited. This has been shown to occur under ischemic conditions, where glucose supplies to the infarcted regions decrease, and the contribution of cerebral pyruvate recycling increases, using either glutamate or glutamine as substrates [36, 37]. Second, NADPH production by malic enzyme activity might be essential for fatty acid synthesis during physiological myelin formation and thus, become a target to support re-myelinating therapies [38, 39]. Third, increased NADPH availability may become very useful in neuroprotection, modulating the GSH/GSSG ratios to protect from free radical damage and oxidative stress [40]. Finally, pyruvate recycling may underlie the neuroprotective effects of ketogenic diets in several neurodegenerative disorders [41–43]. It is also possible that neuronal and glial recycling pathways serve different cerebral functions. In fact, mitochondrial and cytosolic isoenzymes of ME could underlie important cerebral processes as redox regulation and neuroprotection or fatty acid synthesis and myelination, respectively.

To conclude, the pyruvate recycling system in the brain has come a long way since its detection in 1990. It has been demonstrated to occur in the neuronal and glial compartments, both in vivo and in vitro, with potentially important functions. However, uncertainties still remain on the relative contributions of the neuronal and glial systems to the overall process as observed in vivo, or on the precise contribution of peripheral tissues to the recycling processes detected in brain under different conditions. These aspects deserve further exploration. However, the fact that the pyruvate recycling pathway appears to occur ubiquitously in the brain, strongly suggests that it may play a vital role supporting adequate cerebral function.