|Home | About | Journals | Submit | Contact Us | Français|
Despite the long-established therapeutic efficacy of lithium in the treatment of bipolar disorder (BPD), its molecular mechanism of action remains elusive. Newly developed stable isotope-resolved metabolomics (SIRM) is a powerful approach that can be used to elucidate systematically how lithium impacts glial and neuronal metabolic pathways and activities, leading ultimately to deciphering its molecular mechanism of action. The effect of lithium on the metabolism of three different 13C-labeled precursors ([U-13C]-glucose, 13C-3-lactate or 13C-2,3-alanine) was analyzed in cultured rat astrocytes and neurons by nuclear magnetic resonance (NMR) spectroscopy and gas chromatography mass spectrometry (GC-MS). Using [U-13C]-glucose, lithium was shown to enhance glycolytic activity and part of the Krebs cycle activity in both astrocytes and neurons, particularly the anaplerotic pyruvate carboxylation (PC). The PC pathway was previously thought to be active in astrocytes but absent in neurons. Lithium also stimulated the extracellular release of 13C labeled-lactate, -alanine (Ala), -citrate, and -glutamine (Gln) by astrocytes. Interrogation of neuronal pathways using 13C-3-lactate or 13C-2,3-Ala as tracers indicated a high capacity of neurons to utilize lactate and Ala in the Krebs cycle, particularly in the production of labeled Asp and Glu via PC and normal cycle activity. Prolonged lithium treatment enhanced lactate metabolism via PC but inhibited lactate oxidation via the normal Krebs cycle in neurons. Such lithium modulation of glycolytic, PC and Krebs cycle activity in astrocytes and neurons as well as release of fuel substrates by astrocytes should help replenish Krebs cycle substrates for Glu synthesis while meeting neuronal demands for energy. Further investigations into the molecular regulation of these metabolic traits should provide new insights into the pathophysiology of mood disorders and early diagnostic markers, as well as new target(s) for effective therapies.
Mood disorders are common, chronic, severe mental illnesses that affect the lives of millions of individuals worldwide. Despite overwhelming evidence for a biological etiology and pathophysiology, such as alterations in neuronal plasticity (Manji et al. 2000; Schloesser et al. 2008) and cellular resilience (Hunsberger et al. 2009), the underlying mechanisms of the disorders remain largely unknown. Numerous methods can be used to identify putative novel targets in mood disorders; one of the most popular strategies is to study the actions of effective psychotropic medications both in vivo and in vitro.
Although its mechanism of action remains poorly understood, lithium has long been the first-line gold standard for the treatment of bipolar disorder, and as an add-on medication in unipolar and treatment-resistant depression (Manji et al. 1999). Lithium’s remarkable efficacy sparked a revolution that reshaped not only medical and scientific concept of mental illness, but also popular concepts as well. Indeed, the efforts to understand how a simple monovalent cation like lithium could exert such profound beneficial effects have led investigators to examine several signaling pathways and networks involved in mood disorders (Gould and Manji 2002; Manji et al. 1995).
During the past several years, lithium has been shown to have protective effects on neural cells in vivo and in vitro (Gould and Manji 2002; Manji et al. 1995). Underlying mechanisms that have been implicated include metabolic enzymes (Nordenberg et al. 1982), secondary signal pathways and down-stream molecular targets (Chen et al. 2000; Chen et al. 2003; Engel et al. 2009; Manji et al. 1995). One of the most interesting targets is glycogen synthase kinase-3 (GSK-3), which regulates more than 50 enzymes including the metabolism regulators glycogen synthase and pyruvate dehydrogenase (Hedgepeth et al. 1997; Hoshi et al. 1996). This work has generated interest in the effects of lithium on glucose metabolic pathways. Notably, lithium appears to stimulate glucose uptake and glycogen synthesis in cultured cells (Birch and Hughes 1989; Chen et al. 1998) and in rat brain in vivo (Basselin et al. 2006; Plenge 1976, 1982). In human subjects, a recent study using positron emission tomography (PET) showed that lithium altered relative regional cerebral glucose metabolism (rCMRglc) in some specific brain regions (Kohno et al. 2007). Furthermore, the possible indirect effects of lithium on pyruvate dehydrogenase (PDH) activity (Hedgepeth et al. 1997; Hoshi et al. 1996) implicates a possible shift in the balance between glycolysis, lactate production and mitochondrial function including the synthesis of glutamate, and concomitant anaplerotic reactions (Mason et al. 2007; Serres et al. 2008).
Besides its effect on glucose metabolism, lithium also influences brain monoamine synthesis, such as 5-hy-droxytryptophan (5HT) synthesis (Berggren 1985) and glutamate metabolism (Marcus et al. 1986) (O’Donnell et al. 2003), which may also be related to their mechanisms of action in bipolar disorder. The mechanisms underlying lithium’s effects on these biochemical metabolic pathways need further elucidation.
Although brain energy metabolic pathways have been well studied, the different metabolic profiles of astroglia and neurons, as well as details about neuron-glial cell interaction network are still being extensively explored. In general, the astroglia are highly glycolytic, and secrete significant amounts of lactate that is taken up and consumed by the neurons (Bouzier-Sore et al. 2006). The neuron consumes most of the glucose and lactate by mitochondrial oxidation to produce energy (Pellerin et al. 2007; Serres et al. 2008a; Zwingmann and Leibfritz 2003; Zwingmann et al. 2001). Astrocytes have high levels of pyruvate carboxylase (Hertz et al. 2007; Mason et al. 2007), which is one of the major anaplerotic pathways. However, the extent of anaplerosis in the neurons is largely unknown. Moreover, the astrocytes and neurons cooperate via the well-established glutamate-glutamine cycle (Hyder et al. 2006; Serres et al. 2008; Xu and Shen 2006). The importance of “symbiotic” metabolism in brain is well established, and it is plausible that neural suppressive agents such as lithium may affect the critical cycling of these metabolites either at the synapses themselves or through the specific metabolism of these compounds.
Metabolomics provides powerful tools for defining perturbations in metabolic pathways and networks in human disease and during treatment (Kaddurah-Daouk et al. 2008; Kristal et al. 2007; Lindon et al. 2007). Metabolic signatures for CNS disorders could result in the identification of biomarkers for disease, disease progression, and response to therapy, as well as new insights into mechanisms of disease (Kaddurah-Daouk 2009; Kaddurah-Daouk et al. 2008; Kaddurah-Daouk et al. 2007; Paige et al. 2007; Rozen et al. 2005). Additionally, metabolomics provides detailed biochemical knowledge about drug mechanism of action, therapeutic potential and side effects (Kaddurah-Daouk et al. 2008). In the present study, we have used the stable isotope resolved metabolomics (SIRM) approach (Fan et al. 2009) to map effects of lithium on the metabolism of isolated neurons and glial cells. This approach utilized stable isotope tracers such as 13C labeled glucose in conjunction with NMR and mass spectrometry to map biochemical transformation pathway(s) in neurons and glial cells impacted by lithium treatment. With SIRM technologies, multiple pathways can be probed simultaneously, instead of one at a time as in the past. This not only greatly facilitates the discovery of pathway(s) perturbed by lithium treatment but also enables elucidation of metabolic networks spanning different cellular compartments. These approaches could provide new insights into the pathophysiology of BPD and may be used to direct research into uncovering early diagnostic markers and novel therapeutic strategies.
Isotopes: [U-13C]-glucose, [2,3-13C]-alanine and [3-13C]-lactate were purchased from Cambridge Isotope Laboratories (Andover, MA) and were used without further purification.
Cultures of cortical neurons were prepared as previously described (Hao et al. 2004) with minor modifications. In brief, timed-pregnant (E-18) Sprague–Dawley rats (Charles River, Frederick, MD) were euthanized by CO2 inhalation, according to NIH-approved protocols for the care and use of lab animals. Embryos were removed, chilled on ice, and the cortex was microdissected under sterile conditions. Cortical cells was dissociated in Ca2+- and Mg2+-free HBSS containing 0.125% tyrosine for 15 min, triturated in Dulbecco’s Modified Eagle’s Medium (DMEM, Invitrogen, Carlsbad, CA) supplied with 10% FBS and antibiotic-antimyotics (Invitrogen), and plated in 150-mm dishes pre-coated with 5 µg/ml poly-d-lysine (Sigma, St. Louis, MO). Cells were grown at 37°C, 5% CO2 and 95% humidity, first in 10% FBS/DMEM, and one day later switched to serum-free Neurobasal medium (Invitrogen, Cat#21103. Alanine 2 mg/l; Lactate 0; Glutamate 0), plus B27 (Invitrogen). Cultures were grown in serum-free medium for 10 days before being used for experiments, and the medium was changed every 3 days. These cultures yielded >95% neurons as identified by immunocytochemistry with anti-NeuN antibody (Cat# MAB377; Millipore, Billerica, MA).
Astrocytes were isolated from cerebral cortex of postnatal day 3 Sprague–Dawley rats as previously described (Zhang et al. 2002). Briefly, the brains were removed, washed in PBS, and placed in HBSS with antibiotic-antimyotics (Invitrogen). Tissues were dissected and transferred to 15-ml tubes and triturated by mechanical dissociation with a 10-ml pipette, followed by passing through #19 needle once, #21 needle twice and finally into the #25 needle once. After being triturated well, cells were cultured in 75-cm2 flasks (at the equivalent of two brains per flask) and were maintained at 37°C in a humidified incubator maintained at 5% CO2.
After the cultures were shifted to glucose-free medium (DMEM from Sigma. Customized Neurobasal media from Invitrogen) supplied with 0.2% 13C-glucose (10.75 mM), 2 mM 13C-alanine, or 5 mM 13C-lactate, the cells were treated with or without 1.0 mM LiCl (Sigma) for 1, 2, 3 days. This is in the range of the therapeutic lithium concentration in patient blood, rodent blood and brain tissues (Jensen et al. 1996). The same concentrations of non-isotopic glucose, alanine or lactate (Sigma) were supplied to the other group for comparison. Cultured media were filtered, and the cells were washed with PBS, pelletted and immediately frozen at −80°C.
The cold pellet or medium was extracted twice with 10% cold trichloroacetic acid (TCA), followed by lyophilization as previously described (Fan et al. 2005; Fan et al. 2006). Glucose was quantified by NMR using the 13C-1αH satellite signals centered at 5.22 ppm. This accounts for 36% of the total glucose. The 13C and 12C lactate and alanine concentrations were determined by integration of the methyl peak and its satellites and normalized to the concentration of the standard DSS. From this, the amount of glucose consumed (ΔGlc) and the fraction converted to lactate plus alanine (F) could be estimated, according to Eqs. 1 and 2 (Lane and Fan 2007):
1 – F then represents the glucose carbon that enters other metabolites and macromolecules in the cell mass or otherwise not accounted for.
Abundant media components (e.g. threonine and valine) were similarly integrated and their concentrations determined as a function of sampling time to assess the utilization of essential amino acids. The concentrations of choline metabolites (choline, phosphocholine and glycerophosphocholine) were determined from the areas of the trimethyl ammonium resonance near 3.22 ppm as described previously (Fan et al. 1993).
NMR spectra were recorded at 14.1 T or 18.8 T on Varian Inova NMR spectrometer at 20°C using a 90° excitation pulse, an acquisition time of 2 s and a relaxation delay of 3 s. Under these conditions, the protons are essentially full relaxed, as determined by independent measurements of T1. For identification and determining the positional enrichment with 13C we used 1D HSQC and 2D TOCSY (Fan et al. 2005; Fan et al. 2006; Lane and Fan 2007; Lane et al. 2008). 1D HSQC spectra were recorded at 14.1 T using an acquisition time of 0.15 s and a recycle time of 1.5 s. Garp decoupling was applied during the proton acquisition period.
Metabolites were assigned based on their 1H chemical shift, TOCSY connectivity patterns and correlation with 13C, as described previously (Fan and Lane 2008; Fan 1996). All metabolites except choline and the nucleotides were quantified from GC-MS data (Fan et al. 1986; Lane et al. 2008). Positional 13C enrichment was determined from 1D or 2D proton NMR spectra as previously described (Lane and Fan 2007; Lane et al. 2008).
After NMR analysis samples were lyophilized and resus-pended in nanopure H20. An aliquot of the sample was mixed with 40% TCA to a final 10% TCA concentration. Fifty µl of 0.1 mM noreleucine (Sigma) was also added as an internal standard. Samples were lyophilized and silylated with 50 µl of acetonitrile:MTBSTFA (N-methyl-N-[tert-butyl-dimethylsilyl]trifluoroacetamide, Regis Chemical, Morton Grove, IL) (v/v 1:1) for 3 h under sonication, followed by overnight incubation. The derivatized extracts were centrifuged to remove particulates before analysis using PolarisQ GC-ion trap MSn (ThermoFinnigan, Austin, TX) as previously described (Fan et al. 2005). The metabolites were identified and quantified using XCalibur software (ThermoFinnigan).
Immunoblotting was done as previously described (Hao et al. 2004). The specific signals on the blots were detected with ECL and visualized with Kodak X-ray film. Densitometric analyses of PC and beta-actin immunoreactivities were conducted with a Kodak Image Station 440CF.
RNA isolation and reverse transcription: Total RNA was isolated from neuronal cells using Trizol reagent and PureLink RNA mini kit (Invitrogen) according to the manufacturer’s instructions. Reverse transcription reactions were performed using SuperScript III first-strand cDNA synthesis system (Invitrogen) and carried out at 50°C for 45 min followed by reaction inactivation at 70°C for 15 min.
The primers for PC and beta-actin were synthesized by Invitrogen. The sequences of the PC primers were as follows: forward primer 5′-GCGTGTTTGACTACAGTGA G-3′ and reverse primer 5′-TCTTGACCTCCTTGAACTT G-3′. Quantitative PCR was performed with iQ5 Real-Time PCR detection system (Bio-Rad Laboratories, Hercules, CA). The Real-Time PCR program included an enzyme activation step at 95°C (10 min) and 40 cycles of 95°C (30 s), 60°C (30 s). Cycle threshold (Ct) values obtained were converted into mRNA copy number using standard plots of Ct versus log copy number. Data obtained from triplicate runs for target cDNA amplification were averaged and expressed as % of control.
GC-MS analysis was performed in triplicate and %RSD (relative standard deviation) of analysis was calculated for each reported metabolite. ANOVA was used for multiple group comparisons. Comparisons between two groups were done using Student’s t-test with appropriate corrections of multiple comparisons. P < 0.05 is considered significantly different.
Astrocytes and neurons have different demands for glucose metabolism, and differ in the specific metabolites that are exchanged. We therefore measured the cellular uptake of glucose and other metabolites from the bathing medium, and release of metabolites into the media of cultured cells.
Over a period of three days, the cells consumed about one-third of the initial glucose, and converted approximately 90% of all of the consumed glucose into secreted lactate (cf. Eq. 1; Fig. 1). In contrast to glucose consumption, the uptake of essential amino acids such as Thr and Val was very low over the 3-day period (data not shown). The high conversion of glucose to lactate implies that very little of this glucose carbon was used for biosynthesis or mitochondrial oxidation, but rather for energy and presumably to supply the neurons with usable substrates. The rate of lactate release by the cells treated with lithium was initially faster than the control cells, while the glucose consumption rates were similar. This resulted in a higher fraction of glucose to lactate conversion in the lithium-treated cells than the control, indicating that lithium influences the balance between lactic fermentation and other utilization of glucose (see below).
The release of metabolites into the growth medium was also examined by GC-MS. 13C3-lactate was by far the most abundant metabolite released by astrocytes, in common with many mammalian cells. However, unlike other mammalian cells we have analyzed (Fan et al. 2005; Fan et al. 2008; Lane et al. 2009), astrocytes also released an appreciable amount of 13C-labeled citrate, Gln, and Ala into the medium (Fig. 2). Lithium enhanced the release of these labeled metabolites, particularly after 3 days of treatment. Given the “helper” status of astrocytes for neurons, the astrocyte-released lactate has been postulated to serve as a supplementary fuel source for neurons in the so-called astrocyte-neuron lactate shuttle hypothesis (Bonvento et al. 2005; Bouzier-Sore et al. 2003; Bouzier-Sore et al. 2006). Lithium stimulation of lactate release by astrocytes may help improve the energy status of neurons. The release of newly synthesized Gln is to be expected on the premise of Gln-Glu cycling between astrocytes and neurons (Zwingmann and Leibfritz 2003). However, the fate of released citrate and Ala is unknown.
The time course release of 13C-labeled metabolites by cultured neuronal cells and lithium’s effect on the release was analyzed by GC-MS. As with the astrocytes, 13C-lactate was by far the most abundant metabolite released by neuronal cells while 13C-Glu and 13C-Gln were the distant second, followed by Ala, succinate, and Asn (Fig. S1 and data not shown). Unlike the astrocytes (see above) lithium did not have a significant effect on the release of these metabolites, nor on the release of 13C-Ala and 13C-succinate (data not shown).
To understand the effects of lithium on pathways in greater detail, we next measured the intracellular fate of labeled precursors.
We used [U-13C]-glucose to trace the changes in primary metabolic pathways in astrocytes induced by lithium treatment. The 13C label incorporation into various metabolites was analyzed by a combination of 1-D and 2-D 1H TOCSY (Fig. S2) and 1H-13C-HSQC (Fig. S3) NMR experiments, which provide quantitative 13C positional enrichment data (Fan and Lane 2008) as well as GC-MS which quantified the 13C enrichment in metabolites regardless of the labeled position (13C mass isotopologues). Figure 3 illustrates 1-D 1H-13C HSQC analysis of polar extracts obtained from lithium and control-treated astrocytes. This 1-D version of the 2-D HSQC experiment detects protons that are directly attached to 13C-labeled carbons, thereby revealing 13C positional isotopomers. Lithium treatment led to a large increase in lactate peak intensity (which is proportional to concentration) with 13C label at the C-3 position (13C-3-lactate), relative to the control treatment. A smaller lithium-induced increase in the peak intensity of 13C-3-Ala, 13C-1–5′-ribose-AXP, and 13C-1′-ribose-UXP was also evident.
The same polar extracts were analyzed by GC-MS, which both confirmed and complemented the NMR analysis. Total absolute concentrations (in µ mol/g cell dry weight) of metabolites and their 13C mass isotopologues as a function of time in response to lithium treatment were determined by GC-MS (Table S1, Supplementary Materials). Relative to control, lithium-treated cells accumulated more glycolytic metabolites (lactate, Ala, αglycerol-3-phosphate or GlyOH3P), Krebs cycle intermediates (succinate, citrate, and α-ketoglutarate or α-KG), essential amino acids (Ile, Thr), neurotransmitter amino acids (Gln but not Glu) and other non-essential amino acids (Asn, Asp, Pro, Gly) after three days (Fig. 4a; Table S1 in Supplemental Materials). However, the time course of this accumulation varied among metabolites; some persisted over the 3 day period (e.g. lactate, Ala, succinate, citrate, α-KG, Phe, and Pro), whereas others either plateaued (e.g. Asp) or declined (e.g. Glu) over the three days. The concentration of some of the 13C-labeled metabolites (e.g. 13C-lactate, 13C-Ala, 13C-succinate, 13C-citrate, 13C-αKG, 13C-Asp, triply labeled Asp or 13C3-Asp) exhibited qualitatively similar time courses, while others differed (13C-Glu, 13C-Gly) (Fig. 4b). In particular, whereas the total Glu concentration decreased after three days, the 13C-labeled Glu concentration increased in response to lithium treatment. Similarly, 13C3-Asp isotopologue continued to accumulate while the increase in total Asp concentration plateaued in astrocytes in response to lithium treatment (cf. Figs. 4a, b). Such divergent behavior of labeled metabolites from that of the total illustrates the ability of tracers to discriminate the contribution of each metabolite derived from different pathways to the total metabolite pool.
A major fraction of the 13C-labeled lactate and Ala was uniformly labeled ([U-13C]-lactate or 13C3-lactate (Fig. S4) and 13C3-Ala (data not shown) as shown by the characteristic splitting pattern of 13C satellite peaks of lactate (Fan and Lane 2008; Lane et al. 2008) in the 1-D 1H NMR spectrum of astrocyte extracts. The production of 13C3-lactate and 13C3-Ala from [U-13C]-Glc implies active glycolysis and to a lesser extent the pentose phosphate pathway (PPP). The 13C labeling of Glu and glutathione (GSH) at the C-2, 3, and 4 positions (Fig. 3) and 13C-labeled citrate, α-KG, succinate, and citrate (Fig. 4b) were presumably derived from [U-13C]-Glc via the sequence of glycolysis and the Krebs cycle (Fan et al. 2009) while 13C5-ribose-AXP (Fig. 3) was synthesized from [U-13C]-Glc via PPP. Thus, lithium treatment enhanced the glycolytic pathway over the 3-day treatment period, as evidenced by the time-dependent response of 13C-labeled lactate, Ala, and GlyOH3P. The stimulation of Krebs cycle after prolonged lithium treatment was clear from the increased synthesis of the 13C-labeled Krebs cycle intermediates, citrate, succinate, αKG, Glu and Asp (Fan et al. 2009) after 3 days (Fig. 4b). There is also evidence for a lithium-induced enhancement in the anaplerotic input into the Krebs cycle via the pyruvate carboxylation reaction (PC), as indicated by a progressive increase in the synthesis of 13C3-Asp over 3-day period (Fig. 4b; Table S1), which appears to be derived from 13C3-pyruvate (end product of glycolysis) plus CO2 (Fan et al. 2009). PC is known to be active in astrocytes and contributes significantly to Glu synthesis (Hertz et al. 2007). A further enhancement in PC by lithium may have important implications in mitochondrial energetics and Glu/Gln cycling. Such enhanced PC, together with an acceleration of PPP may underlie the increased synthesis of the 13C-ribose moiety of 5′-adenine nucleotides (5′AXP) (Fig. 3).
The same [U-13C]-Glc tracer approach was applied to probe the metabolic effect of lithium on cultured neurons. The 13C isotopomers of polar metabolites extracted from lithium-treated and control neurons were profiled by 1-D 1H-13C HSQC analysis (Fig. S5). As for astrocytes, all metabolites and 13C-isotopomers were identified by 2-D 1H TOCSY and 2-D 1H-13C HSQC analysis. Neuronal cells actively transformed [U-13C]-Glc into metabolites via glycolysis (lactate, Ala), Krebs cycle (succinate, Glu, Asp, GSH, Gln), and PPP (ribose moiety of 5’AXP, 5’UTP). Based on the peak intensity (proportional to 13C isotope concentration) of protons attached to 13C in Fig. S5, lithium induced a time-dependent enhancement in the synthesis of 13C labeled lactate at C-2,3 positions, 13C labeled Ala at C-3, 13C labeled Asp at C2,3, and 13C labeled ribose moiety of 5’AXP at C1–5, relative to the control. The LiCl-enhanced synthesis of 5’AXP-ribose was also evidenced from the 13C satellite peak intensity of the 1′-ribose moiety of 5’AXP in the 1-D 1H NMR spectra of neuronal extracts (data not shown). By calibration against the internal standard DSS, the unlabeled and 13C-labeled AXP concentrations in the pooled day 1–3 neuronal extracts were, respectively, calculated to be 7.2 and 8.2 µ mol/g cell dry weight for the control treatment and 9.4 and 10.2 µ mol/g cell dry weight for the LiCl treatment.
Quantitative GC-MS analysis of the same extracts corroborated the semi-quantitative assessment of the 13C positional isotopomer levels by 1-D HSQC analysis. In addition, GC-MS provided absolute quantification of total metabolites and various 13C mass isotopologues including the monoisotopic (all 12C species), particularly for metabolites that were difficult to resolve or too low in concentration to measure by NMR (cf. Table S2). Relative to the control, lithium-enhanced accumulation of Ala reached a maximum after 2 days of treatment while lactate accumulation peaked after 1 day and diminished thereafter (Fig. 5a). This trend differed from that observed for the astrocytes. Likewise, lithium-elicited accumulation of amino acids such as Ile, Pro, Asn, Gln, Asp, Thr, Phe, and Tyr differed from the corresponding changes in astrocytes. In particular, Glu remained significantly elevated in neuronal cells while it was depleted in astrocytes after 3 days of lithium treatment (cf. Figs. 4a, ,5a).5a). Interestingly, in neurons, lithium depleted the neurotransmitter precursor Trp over time (Fig. 5a) whereas the opposite was evident in lithium-treated astrocytes, relative to control cells (Fig. 4a). Succinate accumulation (relative to the control) peaked in lithium-treated neuronal cells after 2 days. These succinate changes were much greater in extent than those observed in astrocytes. In contrast, the effect of lithium on other Krebs cycle intermediates citrate, malate, and αKG was small.
These changes in total concentrations of metabolites did not provide reliable insights into perturbation of specific pathways since each metabolite pool can be derived from multiple pathways, as indicated earlier. This is where detailed knowledge of the 13C-labeling pattern of individual metabolites is necessary (cf Table S2, Supplementary Materials). Based on the total 13C isotopomer concentrations, lactate and Glu were the main 13C sink metabolites for the [U-13C]-Glc tracer in neuronal cells (Fig. 5b; Table S2). A significant amount of 13C label from the tracer was also incorporated into Asp, Ala, and succinate. A substantial fraction of the 13C-lactate, 13C-Ala and 13C-glycerol-3-P pools were as the uniformly labeled species (i.e. 13C3-lactate, 13C3-Ala, and 13C3-glycerol-3-P) (Table S2), which indicates their synthesis from [U-13C]-Glc via glycolysis without metabolic scrambling. 13C-labeled lactate (13C2- and 13C3-lactate) can also be synthesized via the pentose phosphate pathway (PPP) but to a much lesser extent.
For the Krebs cycle metabolites (succinate, fumarate, malate, citrate, Asp, and Glu), mass isotopologues with 2, 3 and 4 13C atoms were all present and quantified by GC-MS (Table S2). The +2 isotopologue of these metabolites (i.e. 13C2-Asp, 13C2-citrate, 13C2-fumarate, 13C2-Glu, 13C2-malate, 13C2-succinate) were the highest in concentration, followed by the +3 and +4 isotopologues. The +2 isotopologue reflected the Krebs cycle activity after one turn (Fig. 6a) while the +4 isotopologue of citrate (13C4-citrate) should be derived from two turns of Krebs cycle activity (Fig. 6c). Three turns of Krebs cycle activity would generate 13C5-citrate species (cf. Fig. 6d). However, since the concentration of 13C5-citrate was negligible (data not shown), the extent to which the neurons underwent 3 turns of Krebs cycle in this experiment must have been minimal. It should also be noted that the concentration of 13C2-citrate was higher than that of 13C4-citrate, indicating that the majority of the neurons have progressed through only one turn of the Krebs cycle.
For one cycle turn, no 13C3-Asp should be produced, which was not as observed (Table S2; Fig. 5b). Thus, the production of an appreciable amount of 13C3-Asp isotopologue suggests a significant contribution of pyruvate carboxylation (PC) to the synthesis of 13C3-Asp after one cycle turn (cf. Fig. 6b). Moreover, there was a quantifiable amount of +4 isotopologue species for succinate, malate, and Asp (Table S2), which can be derived from the carboxylation of 13C3-pyruvate plus condensation with 13C2-acetyl CoA after one cycle turn (Fig. 6e) or alternatively by normal Krebs cycle activity after three turns (Fig. 6d). Since a large fraction of the neurons progressed through only one Krebs cycle turn, it is reasonable to conclude that PC was the main route of production for 13C4-Asp. Therefore, the individual isotopologues of Glu, Asp, malate, succinate, and citrate are valuable indicators for discriminating lithium’s effect between anaplerotic PC and normal Krebs cycle activity. Likewise, the +3 isotopologues of lactate and Ala revealed how lithium impacts glycolysis.
For glycolysis, despite the diminishing increase in total lactate concentrations (Fig. 5a), lithium-treated cells showed a progressive increase in the level of 13C2- and 13C3-lactate, relative to the control (Fig. 5b). This is consistent with the enhancement of 13C2- and 13C3-Ala levels elicited by lithium. As rationalized above, 13C3-lactate and 13C3-Ala are derived from [U-13C]-Glc via glycolysis and increase in their concentrations therefore indicates enhanced glycolysis by lithium. On the other hand, the +2 isotopomers of lactate and Ala are scrambled products of non-oxidative branch of the PPP. The enhanced production of 13C2-lactate and 13C2-Ala by lithium could result from a stimulation of the PPP (Fig. 5b).
With regards to the Krebs cycle products, the relative level of +2 isotopomers of Glu and citrate was not significantly altered by lithium (Fig. 5b), which indicates that lithium had at most a small effect on the 1st turn of the Krebs cycle through the Glu arm. However, a time-dependent depletion of 13C4-citrate was elicited by lithium, relative to the control treatment. This result could be interpreted as lithium-attenuation of the 2nd Krebs cycle turn and/or enhancement in diverting the newly synthesized citrate to other metabolic pathways such as lipid biosynthesis. The time course effect of lithium on the +2 isotopomers of succinate, malate, and Asp was stimulatory (Fig. 5b; Table S2), which suggests that the 1st turn of the Krebs cycle was enhanced. This is not consistent with lithium’s effect on 13C2-Glu, which could be explained by an increase in flux from oxaloacetate (OAA) to succinate (i.e. Krebs cycle in reverse, cf. Fig. 6a) and from OAA to Asp via transamination, while the forward flux of the Krebs cycle from OAA to Glu remains unaffected. The three enzymes catalyzing the conversion of OAA to succinate (malate dehydrogenase, fumarate hydratase, and succinate dehydrogenase (Hirst et al. 1996)) are freely reversible while the succinyl CoA synthetase (SCS) step is irreversible (cf. Fig. 6a), leading to the observed buildup of total and 13C-labeled succinate (Fig. 5a, b). Moreover, lithium enhanced the production of 13C3-Asp (Fig. 5b), which reflects the increased anaplerotic pyruvate carboxylation, as reasoned above. Enhanced PC would in turn increase the production of OAA, which can lead to increased flux from OAA to succinate and to Asp via mass action.
Although PC is known to be active in astrocytes, it is thought to be absent in neurons (Hertz et al. 2007). However the labeled isotopic patterns in sentinel metabolites indicate a significant flux through PC in cultured neuronal cells. This is further supported by RT-PCR measurements of PCB mRNA and Western Blot analysis of pyruvate carboxylase protein (Fig. S6, Supplementary Materials). This enzyme was present at significant levels in the primary neurons obtained from the rat cortex both in terms of mRNA and protein. However, the effect of lithium on pyruvate carboxylase was negative at the mRNA level and insignificant at the protein level. Since the absolute activity of PC depends also on the concentration of the allosteric activator acetyl CoA (Jitrapakdee et al. 2008), it is possible that lithium altered the acetyl CoA concentration, which can account for its stimulatory effect on PC.
Overall, lithium stimulated glycolysis, the anaplerotic part of the Krebs cycle and PPP in both neurons and astrocytes. Major differences in the effect of lithium on the two cell types were the time course and extent of the synthesis and release of labeled glycolytic and Krebs cycle metabolites. Lithium stimulated the production of these metabolites in neuronal cells peaked earlier than in astro-cytes, except for 13C-Glu, which showed an opposite trend. Lithium enhanced the release of 13C-lactate, Ala, citrate, and Gln by astrocytes but not by neurons. Such differential effects on astrocytes and neurons could help promote the ability of astrocytes to supply neurons fuel substrates for sustaining energy production and the Glu/Gln cycle between the two cell types.
Since lithium enhanced the release of 13C-Ala and 13C-lactate from astrocytes, the question remains whether these metabolites are utilized by neurons and if so, in which pathways they participate. To address this question, 13C-3-lactate (lactate with 13C labeled at the methyl carbon) and 13C-2,3-Ala (Ala with 13C labeled at the methyl and methine carbons) were used as tracers to probe lithium’s effect on neuronal pathways. Figure S7 in Supplementary Materials shows the 1-D 1H-13C HSQC spectral profile of neurons after 1 day of lithium treatment with either 13C-3-lactate or 13C-2,3-Ala as the tracer.
After 1 day of culture in 13C-3-lactate, the 13C abundance of 3-lactate, 3-Ala, 3-Glu, 4-Glu, 4-Gln, 2,3-succinate, 3-Asp, 2-Glu, and 2-Asp was greatly enhanced over the natural abundance (nominally 1.11%) (Fig. S7A). This means that neurons not only took up labeled lactate from the medium, but also incorporated lactate carbon into Krebs cycle metabolites. The major sinks for lactate metabolism were Glu, Asp, Ala, and succinate (cf. Table S3, Supplementary Materials). This is consistent with the view that lactate can be an energy source in neuronal cells (Bouzier-Sore et al. 2006). A similar case can be stated for neuronal cells grown in 13C-2,3-Ala, except that Glu was a less significant sink than that for lactate (Fig. S7B). Absolute quantification of the same cell extracts by GC-MS (Fig. 7; Table S3) agreed with and complemented the NMR analysis. The GC-MS analysis also showed that the intracellular levels of labeled lactate and Asp increased from day 1 to day 3 while the labeled Glu had largely disappeared by day 3 (cf. Fig. 7). The transformation of 13C-3-lactate into 13C-4-Glu, 13C-2 or 3-succinate, 13C-2 or 3-Asp represents normal Krebs cycle activity while conversion of 13C-3-lactate to 13C-2-Glu, 13C-3-Asp, 13C-1- or 4-succinate (not observable in Fig. S7A) would arise from anaplerotic PC activity (Fig. S8B, Supplementary Materials).
Lithium treatment appeared to have two notable effects on 13C-3-lactate metabolism. Figure S7A illustrates the opposite effect of lithium on the peak intensity of protons attached to 13C-4-Glu and 13C-2-Glu. The increase in 13C-2-Glu intensity indicates enhanced PC activity while the decreased 13C-4-Glu intensity could arise from attenuation of the forward Krebs cycle activity to the Glu step (cf. Fig. S8A). This opposing change in levels of 13C-positional isotopomers of Glu could not be revealed from the quantification of total 13C labeled Glu (13C-Glu in Fig. 7; Table S3) and the major isotopologues (13C1-Glu and 13C2-Glu in Fig. 7; Table S3) by GC-MS. Here again, such an effect on neuronal pathways illustrates the need for detailed knowledge of the 13C-positional isotopomers (in addition to 13C-mass isotopologues) of Glu. The stimulation of PC activity by lithium is consistent with the data acquired using [U-13C]-Glc as tracer described above.
The second significant effect of lithium was evident from the GC-MS analysis (Fig. 7). Lithium reduced the depletion of 13C2-Glu by day 3, which could result from an inhibition of oxidation of 13C-lactate-derived Glu via the forward Krebs cycle and/or decreased utilization of labeled Glu. A major sink for Glu could be the synthesis of Gln and its release to the medium. We found a significant level of 13C-labeled Gln (cf. Fig. S9, Supplemental Materials) in both control and LiCl treatment media. However, lithium did not reduce the release of labeled Gln at day 3, which is therefore not a contributing factor to the reduced depletion of intracellular 13C-Glu. In contrast, lithium significantly enhanced the production of 13C1-Asp isotopologue at day 3 (Fig. 7), which together with the increased production of 13C-3-Asp and 13C-2-Glu isotopomers (Fig. S7) could again be explained by enhanced carboxylation of pyruvate derived from lactate (cf. Fig. S8B). This lactate-linked stimulation of PC and inhibition of Glu oxidation by lithium suggests that prolonged LiCl treatment and metabolism of extracellular lactate may lead more to Glu synthesis via PC than total oxidation for energetic purpose.
In conclusion, the combined use of 13C-labeled tracers and 13C isotopomer-resolved metabolomic (SIRM) analysis by NMR and GC-MS for cultured neurons and astrocytes revealed a number of consistent yet unexpected features of neuron and glial metabolism as well as their interaction as summarized in Fig. 8. Of particular note was the presence of anaplerotic pyrvuate carboxylation in neurons. Lithium enhanced glycolysis and PPP in both cell types, as supported by the 13C-labeling patterns of lactate, Ala, and ribose moiety of nucleotides. In addition, lithium altered functioning of the Krebs cycle and stimulated the PC pathway, as revealed by the detailed 13C-labeling patterns of the Krebs cycle metabolites in both cell types. Moreover, lithium enhanced the release of fuel substrates by astrocytes, which were actively metabolized by neurons for energy and neurotransmitter production, as evidenced by the stimulated release of labeled lactate and Ala by astrocytes and their metabolism by neurons. Further investigations into the molecular regulation of these metabolic traits should provide new insights into the pathophysiology of mood disorders and early diagnostic markers, as well as new target(s) for effective therapies. Specifically, we plan to investigate the relative importance of PC and glutaminolysis in anaplerosis and release of different fuel substrates for sustaining the Glu/Gln cycle between neurons and astrocytes using other tracers such as 13C-labeled Gln. We also plan to probe how LiCl may alter the interaction and molecular regulation of the two anaplerotic pathways. The information learned from these cell studies should facilitate the design and interpretation of whole brain studies in model animals and human subjects.
NMR spectra were recorded at the JG Brown Cancer Center NMR facility, and mass spectra were obtained from the Center for Regulatory and Environmental Analytical Metabolomics (CREAM) facility at the University of Louisville. Ioline Henter of NIMH provided invaluable editorial assistance. Financial support: The study was supported in part by NIH Grant Numbers P20RR018733 from the National Center for Research Resources, 1R01CA118434-01A2 (TF, ANL, RMH), 3R01CA118434-02S1 (TF, RMH), and R24GM078233 (RKD, TF) and National Science Foundation EPSCoR grant # EPS-0447479 (TF, ANL).
Teresa W.-M. Fan, Department of Chemistry, Center for Regulatory and Environmental Analytical Metabolomics, University of Louisville, Louisville, KY 40292, USA. Department of Medicine, Structural Biology Program, James Graham Brown Cancer Center, University of Louisville, Louisville, KY 40202, USA. Department of Chemistry, University of Louisville, 2210 S. Brook St, Rm 348 John W. Shumaker Research Building, Louisville, KY 40208, USA, Email: moc.liamg@nafmwt.
Peixiong Yuan, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA.
Andrew N. Lane, Department of Chemistry, Center for Regulatory and Environmental Analytical Metabolomics, University of Louisville, Louisville, KY 40292, USA. Department of Medicine, Structural Biology Program, James Graham Brown Cancer Center, University of Louisville, Louisville, KY 40202, USA.
Richard M. Higashi, Department of Chemistry, Center for Regulatory and Environmental Analytical Metabolomics, University of Louisville, Louisville, KY 40292, USA. Department of Medicine, Structural Biology Program, James Graham Brown Cancer Center, University of Louisville, Louisville, KY 40202, USA.
Yun Wang, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA.
Anahita B. Hamidi, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA.
Rulun Zhou, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA.
Xavier Guitart, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA.
Guang Chen, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA.
Husseini K. Manji, Biomarker Laboratory, Laboratory of Molecular Pathophysiology, Mood and Anxiety Disorder Program, National Institute of Mental Health, NIH, Bethesda, MD 20892, USA. Johnson & Johnson, Titusville, NJ, USA.
Rima Kaddurah-Daouk, Department of Psychiatry, Duke University Medical Center, Box 3950, Durham, NC 27710, USA, Email: ude.ekud.cm@100uddak.