A Ca²⁺-Dependent Mechanism Boosting Glycolysis and OXPHOS by Activating Aralar-Malate-Aspartate Shuttle, upon Neuronal Stimulation

Abstract

Calcium is an important second messenger regulating a bioenergetic response to the workloads triggered by neuronal activation. In embryonic mouse cortical neurons using glucose as only fuel, activation by NMDA elicits a strong workload (ATP demand)-dependent on Na⁺ and Ca²⁺ entry, and stimulates glucose uptake, glycolysis, pyruvate and lactate production, and oxidative phosphorylation (OXPHOS) in a Ca²⁺-dependent way. We find that Ca²⁺ upregulation of glycolysis, pyruvate levels, and respiration, but not glucose uptake, all depend on Aralar/AGC1/Slc25a12, the mitochondrial aspartate-glutamate carrier, component of the malate-aspartate shuttle (MAS). MAS activation increases glycolysis, pyruvate production, and respiration, a process inhibited in the presence of BAPTA-AM, suggesting that the Ca²⁺ binding motifs in Aralar may be involved in the activation. Mitochondrial calcium uniporter (MCU) silencing had no effect, indicating that none of these processes required MCU-dependent mitochondrial Ca²⁺ uptake. The neuronal respiratory response to carbachol was also dependent on Aralar, but not on MCU. We find that mouse cortical neurons are endowed with a constitutive ER-to-mitochondria Ca²⁺ flow maintaining basal cell bioenergetics in which ryanodine receptors, RyR2, rather than InsP₃R, are responsible for Ca²⁺ release, and in which MCU does not participate. The results reveal that, in neurons using glucose, MCU does not participate in OXPHOS regulation under basal or stimulated conditions, while Aralar-MAS appears as the major Ca²⁺-dependent pathway tuning simultaneously glycolysis and OXPHOS to neuronal activation.

SIGNIFICANCE STATEMENT Neuronal activation increases cell workload to restore ion gradients altered by activation. Ca²⁺ is involved in matching increased workload with ATP production, but the mechanisms are still unknown. We find that glycolysis, pyruvate production, and neuronal respiration are stimulated on neuronal activation in a Ca²⁺-dependent way, independently of effects of Ca²⁺ as workload inducer. Mitochondrial calcium uniporter (MCU) does not play a relevant role in Ca²⁺ stimulated pyruvate production and oxygen consumption as both are unchanged in MCU silenced neurons. However, Ca²⁺ stimulation is blunt in the absence of Aralar, a Ca²⁺-binding mitochondrial carrier component of Malate-Aspartate Shuttle (MAS). The results suggest that Ca²⁺-regulated Aralar-MAS activation upregulates glycolysis and pyruvate production, which fuels mitochondrial respiration, through regulation of cytosolic NAD⁺/NADH ratio.

Introduction

The mammalian brain is a prominent consumer of glucose and oxygen in the resting state. In the adult brain, neurons have the highest energy demand (Attwell and Laughlin, 2001). The restoration of the ion movements generated by postsynaptic currents and the action potentials are major neuronal ATP consumers. Neurons respond to these workloads (increases in ATP demand) through a stimulation of ATP production, which is provided through increases in glycolysis and oxidative phosphorylation (OXPHOS) (Hall et al., 2012; Mergenthaler et al., 2013; Connolly et al., 2014; Rangaraju et al., 2014; Ashrafi and Ryan, 2017; Dienel, 2019). The mitochondrial response to workloads is immediate (Llorente-Folch et al., 2013, 2015; Connolly et al., 2014); and in the case of small workloads, it takes place without a detectable decrease in cytosolic ATP levels (Baeza-Lehnert et al., 2019), in dendrites (Gerkau et al., 2019), and the presynapse (Rangaraju et al., 2014).

It has been known for a long time that the rate of respiration depends not only on ATP demand but also on Ca²⁺ (Territo et al., 2000; Denton, 2009; Glancy and Balaban, 2012; Llorente-Folch et al., 2015), providing a feedforward system to adjust OXPHOS to energy demand. In neurons, this regulatory role of Ca²⁺ was shown to be independent of its effect in ATP demand, at least under certain stimulation conditions (Llorente-Folch et al., 2013). Ca²⁺ may upregulate respiration thanks to its entry in mitochondria through the mitochondrial calcium uniporter complex (MCUC) (de Stefani et al., 2016), activating pyruvate dehydrogenase, phosphatase 1 (Fecher et al., 2019), NAD⁺-linked isocitrate dehydrogenase, and α-KGDH (Denton, 2009; Armstrong et al., 2014). However, the finding that a global MCU deletion in outbred CD1 mice did not cause gross dysfunction was at odds with the functions attributed to MCU in the control of bioenergetics (Pan et al., 2013; Wang et al., 2020, Kosmach et al., 2021).

Another way in which Ca²⁺ may operate to increase coupled respiration is through an increase in the activity of mitochondrial metabolite carriers regulated by Ca²⁺, with Ca²⁺-binding motifs facing the intermembrane space (Satrústegui et al., 2007). For example, the aspartate-glutamate exchanger Aralar/AGC1/Slc25a12, a component of the malate aspartate shuttle (MAS), is regulated by Ca²⁺, with half-maximal effects at ∼300 nm (Pardo et al., 2006; Contreras et al., 2007). In brain mitochondria (Gellerich et al., 2013) and neurons using glucose, Ca²⁺ activation of MAS drives pyruvate away from lactate and into mitochondria (Llorente-Folch et al., 2013).

In addition to its role during the response to workloads, Ca²⁺ has shown to be important in the regulation of basal respiration in many cell types (Cárdenas et al., 2010; Mallilankaraman et al., 2012; Filadi et al., 2018; Tomar et al., 2019). A constitutive transfer of Ca²⁺ from ER-to-mitochondria was required to maintain basal respiration and ATP/AMP levels. It involved inositol-3P receptors (IP3Rs) on the ER, the MCU in mitochondria, and additional proteins involved in contacts between ER and mitochondria (Rossi et al., 2019). It is unknown whether this pathway is also functional in neurons, and its specific components in ER and mitochondria, including MCU.

In this work, we have evaluated the contribution of these two systems (MCU and Aralar-MAS) as Ca²⁺-regulation mechanisms to activate mitochondrial respiration in response to workloads and in the regulation of basal respiration in neurons using glucose as fuel. A second fundamental question in this work is the influence of these Ca²⁺ regulation mechanisms in glycolysis itself.

We report that while Mcu-KD blocks the increases in mitochondrial Ca²⁺ elicited by carbachol (Cch) and NMDA, it is dispensable for regulating neuronal respiration using glucose in every condition studied: basal respiration, Cch-stimulated respiration, and NMDA-stimulated respiration. We find that neurons respond to NMDA by a rapid increase in cytosolic pyruvate and lactate production, which is blocked in the absence of Ca²⁺. Remarkably, the Ca²⁺-dependent increase in pyruvate, but not lactate, is blunted in Aralar-KO neurons, revealing a dependence of Aralar-MAS in the provision of cytosolic pyruvate for mitochondrial respiration. Moreover, we show that NMDA triggers a Ca²⁺-dependent increase in glycolysis, which also depends on Aralar. Thus, in neurons using glucose, neuronal activity is met with a Ca²⁺-dependent mechanism, which boosts simultaneously glycolysis and respiration by activating Aralar-MAS, as metabolic link between glycolysis and OXPHOS. It is possible that the Ca²⁺-binding motifs of Aralar N-terminus are involved in this activation, but additional/alternative calcium sensors cannot be excluded as the disruption of Aralar removes both domains of the protein, calcium sensor, and mitochondrial carrier.

Materials and Methods

Animals

Mice with targeted disruption of Aralar/AGC1/Slc25a12 (Aralar-KO) were described previously (Jalil et al., 2005). Aralar-KO mice on a mixed SVJ129xC57BL/6 background were used. WT and Aralar-KO E16 embryos of either sex for preparing neuronal cultures were obtained by breeding of heterozygous Aralar-KO animals.

All animal procedures were performed in compliance with animal protocols approved by institutional ethical committee (Center of Molecular Biology Severo Ochoa) and Autónoma University (CEEA-CBMSO-23/159), and in accordance with Spanish regulations (BOE 67/8509-12, 1988) and European regulations (EU directive 86/609, EU decree 2001-486), reporting followed the ARRIVE Guidelines. All efforts were made to minimize the number of animals used and their suffering.

Drugs

Cells were treated as indicated with cytochalasin B (CytB), 2,4-dinitrophenol (DNP), antimycin A (A), oligomycin (Olig), rotenone (R), caffeine, ionomycin, Cch, TTX, dantrolene, NMDA, MK-801, and CNQX all purchased from Sigma-Aldrich, also with AR-C155858 and (S)-3,5-dihydroxyphenylglycine hydrate (DHPG) from Tocris Bioscience, and ryanodine from Calbiochem. Sodium lactate, sodium pyruvate, and sodium iodoacetate (IAA) were from Sigma-Aldrich and used in the conditions indicated. 3-Desmethyl Xestospongin B (3-dmXeB) was produced by chemical synthesis and dissolved in methanol. For in vitro experiments, compounds dissolved in DMSO or ethanol, both solvents were never higher than 0.001%. Corresponding vehicle solutions were used in control experiments. Doses and concentrations of the different drugs were chosen on the basis of previous published data or preliminary experiments.

Primary neuronal culture

Neuronal cultures were obtained from E15-E16 mouse embryos, as described previously (Pardo et al., 2006; Ramos et al., 2011). Cerebral cortices were enzymatically dissociated in PBS containing 1% BSA, 0.4 mg/ml papain (Roche), and 6 mm glucose. Dissociated cells were plated on poly-L-lysine and laminin-coated glass coverslips in medium containing 20% horse serum. After 3 h, medium was completely replaced by serum-free Neurobasal medium supplemented with 2% B27, 1% Glutamax (all from Invitrogen), and 100 mg/ml penicillin-streptomycin. Neurons were maintained in these conditions until DIV 8-10 for experimental analysis. Neurons represented >80% of the total cell population (Pardo et al., 2006).

Constructs

For Mcu knockdown (KD) using shRNA, we used recombinant adeno-associated virus (rAAV) vectors with a U6 promoter driving shRNA expression that coexpress also mCherry under control of CaMKII promoter to identify infected neurons, pAAV.shRNA.CaMKI.mCherry, kindly provided by Hilmar Bading and Giles E. Hardingham (Qiu et al., 2013). The Mcu hairpin and the scrambled one, used as control, have been described previously (Qiu et al., 2013). To generate shRNA RyR2 constructs, two target sequences, shRyR2-1 and shRyR2-2, used previously (More et al., 2018 and Mu et al., 2014, respectively) were selected. RyR2-sh1, targeting sequences 5′-AGGAAGCAATGGTGGACAAGTTGGCTGAA-3′ and RyR2-sh2, 5′-CGTCCACATACTATTACTC-3′ were annealed and subcloned into the pAAV.shRNA.CaMKI.mCherry vector using standard molecular cloning techniques as described previously (Rueda et al., 2015). All constructs were verified by sequencing.

rAAV obtention

Viral particles (chimeric rAAVs containing the capsid proteins of both AAV1 and AAV2 at equal ratios) were produced for Scr, RyR2-KD, and Mcu-KD. rAAVs were purified and titrated, as described previously (McClure et al., 2011). Briefly, human HEK293 cells were cotransfected with the pAAV.shRNA.CaMKI.mCherry plasmid, the serotype-specific AAV helper plasmids, pH 21 and pRV-1, encoding rep and cap genes of AAV1 and AAV2, respectively, and the adenovirus helper plasmid (pFΔ6) by standard CaPO₄ transfection. Cells were harvested 48 h after transfection, treated with benzonase (Sigma-Aldrich) to reduce the viscosity, and cell debris were cleared by centrifugation for 15 min. The rAAV particles were purified using HiTrap heparin affinity columns (Sigma-Aldrich) and concentrated using Amicon Ultra centrifugal filters (EMD Millipore), and solutions were pooled and sterile filtered. Primary cortical neurons were infected with 10⁷ rAAV particles/μl at DIV 3. Infection efficiencies were determined at DIV 9 by analyzing the fluorescence of mCherry, they ranged from 80% to 90% of the viable neurons. rAAV infection decreased viable neurons. In infected neurons, oxygen consumption and calcium and fluorescence measurements were performed at DIV 9-10.

Transfections

Primary cortical neurons, seeded on chambered cover glass 4-well plates (Lab-Tek) at 1.5 × 10⁵ cells/well density, were transfected at DIV 7-8 using calcium phosphate protocol with 1 µg of plasmids coding fluorescent sensors for the measurement of cytosolic glucose [FLII¹²Pglu-700μδ6; Addgene #28002 (Takanaga et al., 2008)], pyruvate [Pyronic; Addgene #51308 (San Martín et al., 2014)], lactate [Laconic; Addgene #44238 (San Martín et al., 2013)], cytosolic ATP [cyt GO-ATeam1; (Rueda et al., 2015)], ATP/ADP cytosolic ratio [Perceval-HR; Addgene #49083 (Tantama et al., 2013)], and for mitochondrial Ca²⁺ determinations [4mtD3cpv; Addgene #58184 (Palmer et al., 2006)]. For the transfection of rAAV-transduced cells, neurons were seeded at higher density (2 × 10⁵ cells/well), and transfections with fluorescent sensors were performed using Lipofectamine 2000 (Invitrogen) following supplied instructions. FRET measurements were usually performed 24 or 48 h after transfection.

Mitochondrial respiration measurements

Cellular oxygen consumption rate (OCR) was measured using Seahorse XF24 Extracellular Flux Analyzer (Agilent) as previously described (Llorente-Folch et al., 2013; Rueda et al., 2015; Juaristi et al., 2019a). At DIV 9-10, cells were equilibrated with bicarbonate-free low-buffered DMEM (without pyruvate, L-lactate, glucose, glutamine, or calcium) supplemented with 2.5 mm glucose and 2 mm CaCl₂ for 1 h before XF assay. When required,1 μm BAPTA-AM (Thermo Fisher Scientific) was added to the incubation medium. OCR was determined under basal conditions and after the sequential injections of medium or NMDA at the concentrations indicated and 6 μm oligomycin, 0.5 mm DNP, and 1 μm rotenone/1 μm antimycin. Values of basal, ATP-linked, nonmitochondrial and mitochondrial uncoupled respiration were determined (Qian and Van Houten, 2010; Brand and Nicholls, 2011). Protein from each well was extracted with 0.1% NP-40 PBS solution and quantified with BCA protein assay kit (Thermo Fisher Scientific), and data were expressed as a function of protein concentration. Nonmitochondrial OCR was subtracted from OCR values to obtain basal values. OCR/ extracellular acidification rate (ECAR) ratio values were calculated under basal conditions of OCR and ECAR for each well.

Cytosolic calcium, sodium, and pH imaging

Primary cortical neurons were seeded on glass coverslips on 24-well plates at 2.0 × 10⁵, or 1.5 × 10⁵ for Cch stimulations, cells/well. At DIV 8-9, neurons were incubated during 30 min at 37°C in HEPES-buffered control salted solution (HCSS; 137 mm NaCl, 1.25 mm MgSO₄, 10 mm HEPES, 3 mm KCl, 2 mm NaHCO₃, and 1% BSA, pH 7.4) supplemented with 5 μm fura-2 AM (Thermo Fisher Scientific), 50 μm pluronic acid F-127 (Thermo Fisher Scientific), and 2.5 mm glucose without calcium (plus 100 μm EGTA) for Ca²⁺ imaging, and with 0.12 μm BCECF-AM (Thermo Fisher Scientific), 50 μm pluronic acid F-127, 2.5 mm glucose, and 2 mm CaCl₂ for pH imaging. Single-cell measurements of cytosolic Na⁺ were performed as previously described previously (Llorente-Folch et al., 2013; Juaristi et al., 2019a) using sodium-binding benzofuran isophthalate-AM (SBFI-AM, Thermo Fisher Scientific) as Na⁺ indicator. After loading, neurons were washed in HCSS with 2.5 mm glucose and 2 mm CaCl₂ during 20 min at 37°C, and then imaging was performed at 37°C as described previously (Juaristi et al., 2019a) with a Neofluar 40×/0.75 NA objective in an Axiovert 75 M microscope (Carl Zeiss). Acute additions were made as a bolus at the times indicated. ROIs were selected covering single-cell soma. Image acquisition was performed with Aquacosmos 2.5 software (Hamamatsu).

FRET experiments

For pyronic, laconic, FLII¹²Pglu-700μδ6 experiments, cells were pre-incubated 1 h in HCSS (+Ca²⁺) at 37°C; for Ca²⁺-free assays, cells were washed once in HCSS (-Ca²⁺) prior imaging. Experiments were performed in HCSS supplemented with 2.5 mm glucose and 2 mm CaCl₂ or 100 μm EGTA and the additions indicated. In pyronic and laconic experiments, 2 mm pyruvate or lactate was added at the end of the experiment to verify that the probe was not saturated. Neurons were excited at 430 nm (CFP), and emission was collected with a dual pass dichroic filter CFP-YFP (440-500 nm and 510-610 nm). Exposure time was 200 ms. Images were collected every 5 s using a filter wheel λ 10-3, Sutter Instruments (Chroma) and recorded by a Hamamatsu C9100-02 camera mounted on an Axiovert 200 M inverted microscope equipped with a 40×/1.3 oil Plan-Neofluar objective.

ROIs containing only somas were selected on each cell and analyzed using MetaMorph (Universal Imaging) and ImageJ (National Institutes of Health, https://fiji.sc/) software. Background was subtracted from YFP and CFP images; and subsequently, the ratio between CFP and YFP emissions was calculated. Rates of fluorescence increase in response to NMDA stimulation were calculated as the lineal increase in fluorescence ratio (ΔF/min) during the first 30 s after addition. Maximal fluorescence ratio increase (ΔF max) and fluorescence ratio at the end of the recording (ΔF endpoint) after NMDA stimulation were calculated. Mitochondrial calcium was analyzed on neurons expressing the 4mt-D3cpv calcium sensor as described previously (Juaristi et al., 2019a). Variations in cytosolic ATP levels and ATP/ADP ratio using cytosolic FRET probes cyt GO-ATeam1 and Perceval-HR, respectively, were determined as in Rueda et al. (2015).

Adenine nucleotide quantification

Adenine nucleotide total levels were determined by HPLC (de Korte et al., 1995). Neuronal cultures were treated with 660 mm HClO₄ and 10 mm theophylline immediately frozen and kept them overnight at −80°C. Neurons were scrapped and homogenized on ice, then centrifuged at 16,000 × g 15 min 4°C. The supernatants were neutralized using 2.8 m K₂CO₃, kept on ice 10 min, and then at −80°C for at least 2 h to allow precipitation of HClO₄. Extracts were centrifuged, and 50 µl of the supernatants was injected for adenine nucleotide determination. To determine the identity of each peak and quantify the amount of the species of interest, standard curves were constructed by plotting peak heights versus known concentrations of ATP, ADP, and AMP.

qRT-PCR

RNA from primary cortical neurons was isolated using Trizol reagent (TRIzol; Sigma-Aldrich) following the manufacturer's instructions. Retrotranscription reactions were performed using the iScript cDNA Synthesis kit (Bio-Rad PN170-8891) following the manufacturer's instructions. Thermal conditions consisted of the following steps: 5 min × 25°C, 20 min × 46°C, and 1 min × 95°C. cDNA was amplified with Fast SYBR MasterMix probe (Thermo Fisher Scientific) in the ABI Prism 7900HT sequence detection system (Thermo Fisher Scientific) at the Genomics and Massive Sequencing Facility. The primers used for amplifying the target genes were as follows: mouse RyR2 (5′-GAATTCATCATGGATACTCTACC-3′; 5′-GTCATGCACATTATCTTCTGCAT-3′). Mouse β-actin (5′-CTAAGGCCAACCGTGAAAAG-3′; 5′-ACCAGAGGCATACAGGGACA-3′) and mouse GAPDH (5′-TGCGACTTCAACAGCAACTC-3′; 5′-GGATAGGGCCTCTCTTGCTC-3′) were used as housekeeping genes. ValidPrime expression (Tataa Biocenter, A106S10) was checked in all the samples, and necessary corrections were done as described by Laurell et al. (2012). The relative expression of genes was calculated using the 2^-ΔΔCt Livak comparative method (Schmittgen and Livak, 2008).

Western blotting and antibodies

To examine protein levels, total protein extracts were prepared in RIPA buffer (125 mm NaCl, 25 mm Tris-Cl, pH 7.4, 1 mm EGTA, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, and Complete EDTA-free protease inhibitor, Roche); 22.5 µg (for RyR2 analysis) or 40 µg (for MCU and AMPK analysis) of total proteins was loaded. Proteins were separated by SDS-PAGE electrophoresis, and transferred onto PVDF or nitrocellulose membranes (for RyR2 or other analysis, respectively) by wet electrophoretic transfer. Blots were blocked 1 h at room temperature with 5% nonfat dry milk in TBS-Tween (0.5 m Tris, 1.5 m NaCl, 0.01% Tween) solution and incubated at 4°C with primary antibodies. For phospho-antibodies, 5% BSA in TBS-Tween was used as blocking solution. Secondary antibodies were incubated 1 h at room temperature. The following antibodies were used: anti-MCU (1:1000, Sigma-Aldrich), anti-RYR2 (1:500, Thermo Fisher Scientific), anti-GAPDH (1:1000, Merk), anti-AMPKα (1:1000, Cell Signaling Technology), anti-p-AMPK T172 (1:1000, Cell Signaling Technology), and anti-β-F₁ATPase (1:5000, kindly provided by Jose M. Cuezva). Secondary HRP-conjugated antibodies were purchased from Bio-Rad and used at 1:5000 dilution. The signal was detected by chemiluminescence with detection solution (75 mm Tris, pH 8.8, 2.5 mm luminol, 0.40 mm p-coumaric acid, 0.01% H₂O₂). The quantification of the bands was performed with ImageJ (National Institutes of Health) software. We used 661W cells (expressing RyR2) as positive control.

Quantification and statistical analysis

Statistical details of experiments can be found in the figure legends. All results are representative of at least three independent experiments unless otherwise specified and are presented as mean ± SEM. Significance was calculated by ANOVA (one-way or two-way), post hoc Bonferroni test, or t test. All statistical tests were plotted either with Sigma Plot 11.0 software or GraphPad Prism version 6 or 7 software.

Results

Control of basal neuronal respiration by ER calcium: the role of ER Ca²⁺ stores and MCU

In 2010, a constitutive transfer of Ca²⁺ from ER to mitochondria was shown to be required to maintain basal respiration and ATP/AMP levels, and to block macroautophagy via an AMPK-dependent pathway (Cárdenas et al., 2010; Mallilankaraman et al., 2012).

We have examined whether this constitutive Ca²⁺ transfer pathway plays a similar role in neurons, and the role of MCU in this process. As neurons express two types of Ca²⁺ release channels in the ER, IP3R, and ryanodine receptors (RyRs) (Zalk et al., 2007; Galeotti et al., 2008; Egorova and Bezprozvanny, 2018), we have analyzed the role of both. To study the role of ER efflux via IP3R, we used a membrane-permeable selective blocker of all types of IP3R, analog of Xestospongin B (XeB), a macrocyclic bis-1-oxaquinolizidine alkaloid originally extracted from the marine sponge Xestospongia exigua (Gafni et al., 1997; Jaimovich et al., 2005). This analog 3-dmXeB has been produced by chemical synthesis and exerts similar actions as XeB on IP3R (Podunavac et al., 2021).

Figure 1A shows that 60 min preincubation of cortical neurons with 5 μm 3-dmXeB substantially reduces the Ca²⁺ transient induced by 200 μm DHPG, an agonist of Group I metabotropic glutamate receptors (Niswender and Conn, 2010) consistent with the ability of XeB to block IP3R (Jaimovich et al., 2005; Cárdenas et al., 2010); 10 μm 3-dmXeB was also inhibitory but induced cellular damage and was not used any further. 3-dmXeB did not modify the basal respiration rate or OCR/ECAR ratio in the same cortical neurons, as determined with a Seahorse extracellular flux analyzer (Fig. 1B–D), indicating that Ca²⁺ efflux from the ER through IP₃R does not play a relevant role in the control of basal respiration.

Figure 1.

RyRs control neuronal mitochondrial respiration. A, 5 and 10 μm 3-dmXeB substantially reduces the Ca²⁺ transients induced by 200 μm DHPG, an agonist of Group I metabotropic glutamate receptors in cortical neurons. Neuronal cultures were preincubated with 3-dmXeB 60 min before the experiment was started. B, C, Incubation with 5 μm 3-dmXeB does not change basal respiration in cortical neurons. 3-dmXeB was added to the culture 60 min before the experiment was started. Mitochondrial function was determined through sequential addition of 6 μm oligomycin (Olig), 0.5 mm DNP, and 1 μm/1 μm antimycin A/rotenone (A/R) at the indicated time points. Data are mean ± SEM; n = 7. C, Basal respiration corrected for nonmitochondrial respiration. Dots represent individual data. Error bars indicate mean ± SEM; n = 14 or 15. D, Basal OCR/ECAR ratio values. Data are mean ± SEM; n = 8-17. E, qPCR analysis of the cDNA obtained from neuronal cultures transduced with rAAV containing two different RyR2-directed (RyR2-KD1 or RyR2-KD) or Scrambled (Scr) shRNA using specific oligos; values were normalized to RyR2 mRNA in Scr neurons. Data are mean ± SEM, 5-9 wells (dots) from two independent experiments. One-way ANOVA: **p ≤ 0.01, ****p ≤ 0.0001, post hoc Bonferroni test. F, Western blot analysis of RyR levels in neurons transduced with rAAV containing RyR2-directed (RyR2-KD) or Scrambled shRNA (Scr). GAPDH was used as loading control. G, Quantitative analysis from Western blot of RyR2/GAPDH. Data are mean ± SEM; n = 3. *p = 0.049 (two-tailed t test). H, RyR2 silencing decreases caffeine-induced [Ca²⁺]_i signal, which is blocked by ryanodine or dantrolene (see Extended Data Fig. 1-1A,B). I–K, RyR2 silencing decreases basal and oligomycin-sensitive OCR levels in neurons. I, Data are mean ± SEM; n = 7 or 8. J, Data are mean ± SEM; n = 14-16. **p = 0.0037. K, Data are mean ± SEM; n = 20-24. **p = 0.0017 (t test). L, Representative Western blot analysis of p-AMPK and AMPK levels. M, Quantitative analysis from Western blot of p-AMPK/AMPK levels. Dots represent individual data. Error bars indicate mean ± SEM; n = 9. **p = 0.0069 (t test). N, RyR2 silencing decreases basal OCR/ECAR ratio values. Data are mean ± SEM; n = 8-17, from 3 independent experiments and does not change ATP, ADP, or AMP levels (see Extended Data Fig. 1-1C–G). **p = 0.0041 (two-tailed t test).

Figure 1-1

Caffeine-discharge ER-Calcium and Adenine nucleotide content in RyR2-KD neurons. A-B: 15 mM Caffeine (Caff.) discharges ER calcium by activating RyR in cortical neurons at 9 DIV. Inhibition of the response to caffeine by preincubation with 50 µM ryanodine (Ryan.) (A) or 10 µM dantrolene (Dandr.) (B) during 15 or 30 min. C-G: Adenine nucleotide content, measured by HPLC in neurons transduced with rAAV containing RyR2-directed (RyR-KD) or Scrambled shRNA (Scr). Mean ± SEM (bars), 5 independent experiments (dots). fura-2 AM Download Figure 1-1, TIF file ^{(299.2KB, tif)} .

RyRs, but not MCU, are involved in the maintenance of basal respiration in cortical neurons

In neurons, ER Ca²⁺ mobilization via activation of RyR is involved in calcium-induced calcium release of CICR (Shmigol et al., 1995; Solovyova et al., 2002; Patel et al., 2009). As observed (see Extended Data Fig. 1-1A), 15 mm caffeine can discharge the ER Ca²⁺ store, as shown previously (Solovyova et al., 2002), resulting in an increase in cytosolic Ca²⁺. Preincubation with RyR inhibitors, such as ryanodine (50 μm; Extended Data Fig. 1-1A,B) (Solovyova et al., 2002) or dantrolene (10 μm; Extended Data Fig. 1-1A,B) (Liu et al., 2009; Oulès et al., 2012) was able to block the release of Ca²⁺ caused by caffeine, confirming that these compounds block RyR in cortical neurons.

To study the role of RyR in providing a constitutive Ca²⁺ flow regulating respiration in cortical neurons, we have silenced RyR2, the major RyR isoform in brain cortex and hippocampus (Zalk et al., 2007; Galeotti et al., 2008), using selected sequences used previously (Mu et al., 2014; More et al., 2018). Neurons were infected with two rAAV-containing vectors encoding RyR2-directed small hairpin RNA (shRNA-1 and shRNA2), or nontargeted control sequence (Scrambled, Scr). Silencing was verified by qPCR (Fig. 1E) and Western blot analysis, indicating a decrease in RyR2 protein levels of ∼50% with shRNA2. (Fig. 1F,G). RyR2 silencing with shRNA2, RyR2-KD, resulted in a marked decrease in caffeine-induced Ca²⁺ release (Fig. 1H).

A significant decrease in basal respiration was observed in the RyR2-silenced neurons (RyR2-KD, Fig. 1I,J), which affected oligomycin-sensitive respiration (Fig. 1K) supporting a constitutive flow of ER Ca²⁺ to mitochondria involving, at least partially, RyR2 (Fig. 1I).

Failure to maintain ER to mitochondria Ca²⁺ flux causes an increase in autophagy triggered by activation of AMPK (Cárdenas et al., 2010; Mallilankaraman et al., 2012). Indeed, AMPK phosphorylation status increased in RyR2-KD neurons (Fig. 1L,M). This was accompanied by an increase in ECAR and a decrease in the OCR/ECAR ratio (Fig. 1N), suggesting a higher glycolytic flux in response to AMPK activation. Indeed, when evaluating the adenine nucleotide levels present in these cells, RyR2 silencing did not lead to any significant changes (Extended Data Fig. 1-1C–G) in agreement with the robust energetic stability in neurons (Baeza-Lehnert et al., 2019).

We have next studied the involvement of the MCUC in this constitutive pathway. To this end, we silenced Mcu, the Ca²⁺ transporter component of MCUC (de Stefani et al., 2016). Neurons were transduced with rAAV containing Mcu-directed shRNA, or nontargeted control sequence (Scrambled, Scr) (Qiu et al., 2013). Analysis by Western blot shows a 65 ± 10% decrease in MCU protein level (Fig. 2A,B). Figure 2C shows that the basal FRET ratio of the mitochondrial Ca²⁺ probe 4mtD3cpv in Mcu-KD neurons is very similar to that of Scr neurons. Indeed, unchanged resting Ca²⁺-mit levels were observed for other tissue-specific MCU deletions, as muscle- (Gherardi et al., 2019), cardiomyocyte- (Kwong et al., 2015), or neuronal-specific Mcu-KD (Qiu et al., 2013; Ashrafi et al., 2020). Silencing of Mcu had no effect on basal respiration with glucose as the main substrate compared with neurons transduced with adenovirus carrying Scrambled sequences (Fig. 2D,E). Similarly, neuron-specific KD of MCU did not cause changes in basal respiration in cortical neurons (Nichols et al., 2018). As MCU is clearly able to take up Ca²⁺ in mitochondria (see below), we conclude that it is not required as part of the ER to mitochondria Ca²⁺ flow maintaining basal respiration in neurons using glucose.

Figure 2.

Basal mitochondrial Ca²⁺ and respiration do not change in Mcu-KD cortical neurons. A, Representative Western blot analysis of MCU levels in neurons transduced with rAAV-containing MCU-directed (Mcu-KD) or Scrambled shRNA (Scr). βATPase was used as loading control. B, Quantitative analysis from Western blot of MCU/ βATPase. Dots represent individual data. Error bars indicate mean ± SEM; n = 3-5. **p = 0.0072 (t test). C, Basal ratio of 4mtD3cpv probe reflecting basal mitochondrial Ca²⁺ levels in Scr and Mcu-KD neurons. Dots represent individual data. Error bars indicate mean ± SEM; n = 19-21 neurons per condition from 5 independent platings. p = 0.2335 (t test). D, E, Basal respiration in Scrambled and Mcu-KD neurons expressed as OCR (nmol O₂/min/mg protein). Dots represent individual data. Error bars indicate mean ± SEM; n = 26 per condition from 5 independent platings. p = 0.4193 (t test).

Therefore, our results show that a pathway for Ca²⁺ flow from ER to mitochondria that controls basal respiration exists in cortical neurons, but it is not dependent on IP3R on the ER or Ca²⁺ entry in mitochondria along MCU.

The neuronal respiratory response to activation of acetylcholine receptors by Cch depends on Aralar-MAS but not MCU

Having shown that ER to mitochondria calcium flow via IP3R does not control neuronal respiration under basal conditions, we have investigated its role in neuronal stimulation by Cch (Delmas et al., 2002, 2004). The spontaneous Ca²⁺ activity of the neuronal network is synaptically driven (Bacci et al., 1999; Opitz et al., 2002; Young et al., 2005), as it is blocked by TTX and ionotropic glutamate receptor inhibitors (Extended Data Fig. 3-1A,B), and is enhanced by Cch with an increase in the amplitude of the Ca²⁺ transients (Fig. 3A,B). The Cch-induced workload was fully Ca²⁺-dependent with no changes in Na⁺-cyt (Fig. 3C) and accordingly, caused a stimulation of neuronal respiration only in the presence of extracellular Ca²⁺ (Fig. 3D,E).

Figure 3.

Cch stimulation of mitochondrial respiration depends on calcium and the ARALAR-MAS pathway. A, fura-2 AM [Ca²⁺]i signals in neurons in HCSS medium containing 2 mm CaCl₂, on addition of 250 μm Cch where indicated. The figure shows a representative experiment; each trace corresponds to a single neuron from the same recording field. Spontaneous Ca²⁺ oscillations are blocked by inhibitors of ionotropic glutamate receptors (Extended Data Fig. 3-1A,B). B, Quantification of peak amplitude as ΔRatio (F₃₄₀/F₃₈₀) ± SEM comparing basal spontaneous Ca²⁺ oscillations to Cch-enhanced Ca²⁺ oscillations. Data were obtained from 4 independent experiments. Means were compared using one-tailed t test (p = 0.0064). C, SBFI [Na⁺]i signals in neurons on addition of 250 μm Cch where indicated. Data are mean ± SEM: 3 experiments, n = 96. D, Stimulation by 250 μm Cch of OCR expressed as percentage of basal OCR in the absence of Ca²⁺ (–Ca²⁺) or presence of 2 mm Ca²⁺ (+Ca²⁺) in the incubation medium. E, Quantification of percentage of respiratory stimulation (OCR % 3 min after Cch addition). Data were obtained from 6 independent experiments (n = 6). Means were compared using one-tailed t test (p ≤ 0.0001). The effect of Cch on Ca²⁺ oscillations was the same in MCU-KD or in Aralar-KO neurons (Extended Data Fig. 3-1C–H). F, 4mtD3cpv mitochondrial Ca²⁺ signals in Scrambled or Mcu-KD neurons on addition of 250 μm Cch where indicated. Data are normalized to the initial values and are expressed as mean ± SEM. Data were obtained from 3 independent experiments (n = 8-11 cells). G, Cch (250 μm) stimulation of OCR expressed as percentage of basal OCR in Scrambled and Mcu-KD neurons. H, Percentage of respiratory stimulation (OCR % 3 min after Cch addition) in neurons infected with Scr and Mcu rAAVs. Data were obtained from 3 independent experiments (n = 10-12). I, Cch (250 μm) stimulation of OCR expressed as percentage of basal OCR in WT and Aralar-KO neurons. J, Percentage of respiratory stimulation (OCR % 3 min after Cch addition) in WT and Aralar-KO neurons. Data were obtained from 6 independent experiments (n = 14-21). Means were compared using one-tailed t test (p = 0.0063). K, Cch (250 μm) stimulation of OCR in WT and Aralar-KO neurons in the presence of 2 mm pyruvate. L, Percentage of respiratory stimulation (OCR % 3 min after Cch addition). Data were obtained from 3 independent experiments (n = 11). Means were compared using one-tailed t test (p = 0.53).

Figure 3-1

Cch induced cytosolic Ca²⁺ signals. Effects of MCU-KD and Aralar-KO. A. Mouse cortical neurons plated at high density and cultured for 8-9 days showed spontaneous Ca²⁺ activity, consisting of repetitive Ca²⁺ transients occurring at the same time in all neurons. These Ca²⁺ transients were blocked by treatment with the AMPA/Kainate receptor antagonist CNQX (10 μm), the NMDAR antagonist MK-801 (10 μm), or the Na⁺ channel blocker TTX (1 μm) (results not shown). B. Cch addition in the presence of TTX, which blocks other responses except those due to Ca²⁺ mobilization by IP3R, resulted in a weak response. C-D: fura-2 AM [Ca²⁺]i signals in Scr (C) or Mcu-KD (D) neurons in HCSS medium containing 2 mM CaCl₂ and 2.5 mM glucose upon addition of 250 μm Cch where indicated. Representative experiments, each trace corresponds to a single neuron from the same recording field. E: Quantification of peak amplitude as ΔRatio (F₃₄₀/F₃₈₀) ± SEM. Data were obtained from 3 independent experiments. F-G: fura-2 AM [Ca²⁺]i signals in WT (E) and Aralar-KO (G) neurons in HCSS medium containing 2 mM CaCl₂ and 2.5 mM glucose upon addition of 250 μm Cch where indicated. Representative experiments, each trace corresponds to a single neuron from the same recording field. H: Quantification of peak amplitude as ΔRatio (F₃₄₀/F₃₈₀) ± SEM. Data were obtained from 5 independent experiments. Download Figure 3-1, TIF file ^{(405.7KB, tif)} .

To study the impact of MCU signaling on the Cch-induced increase in respiration, we silenced Mcu and studied its effect on Cch-stimulated respiration. Downregulation of MCU protein resulted in drastic block in mitochondrial Ca²⁺ signals, as Cch-driven mitochondrial Ca²⁺ rise was dramatically decreased in Mcu-silenced neurons (Fig. 3F), while the increase in cytosolic Ca²⁺ transients evoked by Cch were the same in both Scrambled (Scr) or Mcu-KD neurons. No differences were found in amplitude or number of peaks after Cch addition (Extended Data Fig. 3-1C–E).

As described, basal respiration rates were similar in Scrambled and Mcu silenced neurons (Fig. 2D,E). Strikingly, although Cch-induced increase in matrix Ca²⁺ did not occur under Mcu silencing conditions (Fig. 3F), the stimulation of mitochondrial respiration by Cch was not affected: 114 ± 2% in Scrambled and 113 ± 1% (p = 0.71, t test) in Mcu-silenced neurons over basal levels, respectively (Fig. 3G,H).

Having shown that MCU is not required for Cch stimulation of respiration, we investigated the role of Aralar. We first verified that cytosolic Ca²⁺ signals induced by Cch in cortical neurons derived from WT and Aralar-KO mice were the same (Cch enhanced Ca²⁺ oscillations to the same level; Extended Data Fig. 3-1F–H). Next, we studied Cch stimulation of respiration in the two genotypes finding that Cch stimulation of respiration is reduced by ∼40% in the absence of Aralar (WT neurons: 118 ± 1% and Aralar-KO neurons: 111 ± 1% increase over basal levels, p = 0.0063, t test; Fig. 3I,J). It is likely that the remaining stimulation of respiration is due exclusively to Ca²⁺-stimulated workload (i.e., Ca²⁺-induced ATP consumption), not to Ca²⁺ signaling. Although much smaller than that dependent on Na⁺ entry, Ca²⁺-stimulated workload plays also a role in neuronal energy consumption (Llorente-Folch et al., 2013; Díaz-García et al., 2021). The lack of Aralar also caused a pronounced decrease in DNP-stimulated respiration (Fig. 3I).

Activation by extramitochondrial Ca²⁺ of Aralar-MAS results in an increase in NADH production in neuronal mitochondria, which is blocked in Aralar-KO neurons, and the presence of exogenous pyruvate in Aralar-KO neurons overcomes this block (Pardo et al., 2006). It has been proposed that Ca²⁺ activation of Aralar functions as a “gas pedal” to increase pyruvate formation (Gellerich et al., 2009, 2012, 2013). Indeed, Cch-stimulated respiration in Aralar-KO neurons was fully restored in the presence of 2 mm pyruvate (Fig. 3K,L), and this also restored DNP-stimulated respiration (Fig. 3G). Therefore, we conclude that the lack of Aralar results in a limitation in substrate (pyruvate) supply to mitochondria, which prevents full Cch-induced stimulation of OCR.

Role of MCU and Aralar-MAS in Ca²⁺-dependent stimulation of neuronal respiration triggered by NMDA

Activation of NMDARs leads to Na⁺ and Ca²⁺ entry in cortical neurons (Rueda et al., 2015). In order to restore the electrochemical gradients, Na⁺/K⁺-ATPases, plasma membrane and ER Ca²⁺ ATPases, and Na⁺/Ca²⁺ exchangers are activated, leading to stimulated ATP use and to the upregulation of mitochondrial respiration to meet the energetic demands. In the absence of extracellular Ca²⁺, 100 μm NMDA evoked a smaller increase in respiration than in the presence of calcium (Rueda et al., 2015). As the responses to NMDA were much larger than those elicited by Cch, we have switched to NMDA as agonist to analyze in more detail the role played by Aralar and MCU and other metabolic pathways in the responses.

The Ca²⁺ dependence of the OCR response in cortical neurons is also observed at lower NMDA concentrations, 25 μm NMDA (Fig. 4A,B) larger responses in the presence than in the absence of Ca²⁺. The response to NMDA entails increased respiratory substrate utilization, which probably limits the maintenance of the OCR beyond the initial response (Fig. 4A). In addition, the maximal respiratory response was lower than in nonstimulated neurons (compare the responses to the uncoupler in Fig. 1B with those in Fig. 4A). Thus, after correction for nonmitochondrial respiration, uncoupler-stimulated respiration was 280.7 ± 9.6% or 83.2 ± 5.9% of basal respiration in unstimulated or NMDA stimulated neurons, respectively (p < 0.001), indicating a limitation in substrate supply to maintain the response to the uncoupler after stimulation by NMDA. The limitation of substrate supply after glutamate stimulation of neuronal OCR has been confirmed experimentally, by showing that the addition of other biofuels at the time of uncoupler addition restores maximal OCR to control values (Laird et al., 2013). Thus, in agreement with the larger NMDA-OCR stimulation in the presence than absence of Ca²⁺, maximal respiration was larger in Ca²⁺-free media, in which the drain of substrates during NMDA stimulation is lower (Fig. 4A).

Figure 4.

NMDA stimulation of OCR and cytosolic pyruvate and lactate formation. A, 25 μm NMDA-stimulated respiration in cortical neurons incubated in the presence of 2 mm Ca²⁺ (+Ca²⁺) or without calcium (–Ca²⁺; plus 100 μm EGTA), expressed as percentage of basal values (OCR %). B, NMDA-stimulated OCR (%). Data are mean ± SEM (bars); n = 9-16 (dots) from two different experiments. One-way ANOVA: ***p ≤ 0.001, post hoc Bonferroni test. C, 25 and 5 μm (Extended Data Fig. 4-1A,B) NMDA-induced changes in mitochondrial Ca²⁺ (Ca²⁺-mit) in Scrambled (Scr) and Mcu-silenced (Mcu-KD) neurons transfected with 4mtD3cpv probe. D, Quantification of maximum Ca²⁺-mit increment (ΔF max) after NMDA stimulation. Data are mean ± SEM (bars); n = 10 or 11 from 3 independent experiments. ***p = 0.0006 (t test). NMDA-induced cytosolic Ca²⁺ signals were the same in Scrambled and MCU-KD neurons (Extended Data Fig. 4-2A–D). E, 25 and 5 μm (Extended Data Fig. 4-1C,D) NMDA-stimulated respiration in Scrambled and Mcu-KD neurons, expressed as percentage of basal values (OCR %). F, Quantification of NMDA-stimulated OCR (%). Data are mean ± SEM (bars); n = 30-33, 5 independent experiments. *p = 0.0173 (t test). G, 25 and 5 μm (Extended Data Fig. 4-1E,F) NMDA-stimulated respiration in WT and Aralar-KO neurons, expressed as percentage of basal values (OCR %). H, Quantification of NMDA-stimulated OCR (%). Data are mean ± SEM (bars); n = 9-16 (dots) from 2 independent experiments. ****p < 0.0001 (t test). Cytosolic and mitochondrial Ca²⁺ signals were the same in WT and Aralar-KO neurons (Extended Data Fig. 4-2E–L). I, J, Effect of NMDA in the presence (+Ca²⁺) or absence (–Ca²⁺) of calcium on cytosolic pyruvate (Pyr) and lactate (Lac) in neurons transfected with pyronic (I) and laconic (J) probes. K, L, Dose-dependent effect of 5 and 25 μm NMDA induction of Pyr (K) and Lac (L) production. Data are mean ± SEM; n = 5-15 neurons per condition from 2-6 independent experiments. M, Intracellular pH variation determined in BCECF-AM-loaded neurons after 5 mm lactate (Lac) acute addition with or without 1 min pre-incubation with 1 μm MCT1/2 inhibitor AR-C155858 (AR-C1). Data are mean ± SEM; n = 6 per condition from 2 independent experiments. N, Lactate changes induced by 25 μm NMDA in neurons to which 1 μm AR-C1 or medium (Veh) was pre-added. O, Quantification of maximum fluorescence change (ΔF max) in Lac after Veh/1 μm AR-C1 and 25 μm NMDA. P, Lactate changes induced by 25 μm NMDA in neurons in the presence (Lact) or absence (Ctr) of extracellular 2 mm lactate. Q, Pyruvate changes induced by 25 μm NMDA in neurons pre-incubated 30 min with HCSS (Veh), 0.5 mm IAA, or IAA + 1 μm AR-C1 (IAA + AR-C1). Data are mean ± SEM; n = 8-20 per condition from 4-8 independent experiments. The 4mtD3cpv, pyronic, laconic, and BCECF-AM experiments were performed in HCSS, 2.5 mm glucose, and 2 mm Ca²⁺ or 100 μm EGTA (–Ca²⁺).

Figure 4-1

Changes in mitochondrial Ca²⁺ and OCR upon stimulation with 5 μm NMDA. A: Changes in mitochondrial Ca²⁺ (Ca²⁺-mit) in Scrambled (Scr) and Mcu-silenced (Mcu-KD) neurons transfected with 4mtD3cpv probe, after 5 μm NMDA stimulation in 2.5 mM glucose and 2 mM Ca²⁺ HCSS; B: Quantification of maximum Ca²⁺-mit increment (ΔF max) after NMDA stimulation. Data are mean ± SEM (bars), n = 8-9 (dots) from three independent experiments; T- test *p = 0.0365. C: 5 μm NMDA-stimulated respiration in Scrambled and Mcu-KD neurons maintained in 2.5 mM glucose and 2 mM Ca²⁺ DMEM, expressed as percentage of basal values (OCR %); D: Quantification of NMDA-stimulated OCR (%); mean ± SEM (bars), n = 14-15 (dots) from 5 independent experiments; T test *p = 0.0339. E: 5 μm NMDA-stimulated respiration in wild type and Aralar-KO neurons maintained in 2.5 mM glucose and 2 mM Ca²⁺ DMEM, expressed as percentage of basal values (OCR %); F: Quantification of NMDA-stimulated OCR (%); mean ± SEM (bars), n = 4-6 (dots) from 2 independent experiments; T test ***p = 0.0007. Download Figure 4-1, TIF file ^{(311.5KB, tif)} .

Figure 4-2

NMDA-induced Cytosolic and mitochondrial Ca²⁺ signals in Scr, Mcu-KD, Wild type and Aralar-KO neurons. A-D: Changes in cytosolic Ca²⁺ (Ca²⁺-cyt) after 5 (A, C) and 25 μm NMDA (B, D) stimulation in Scr (A, B) or Mcu-KD (C, D) neurons. E-I: Changes in cytosolic Ca²⁺ (Ca²⁺-cyt), after 5 (E, G) and 25 μm (F, H) NMDA in wild type (E, I) and Aralar-KO (G, H) Fura2-AM loaded neurons. Individual cell recordings (gray) and average (black) are shown. I: Quantification of peak amplitude as ΔRatio (F₃₄₀/F₃₈₀) ± SEM (bars) comparing 5 and 25 μm NMDA stimulus; two-way ANOVA, **p ≤ 0.01, ***p ≤ 0.001, post hoc Bonferroni test. Recordings from at least 50 cells/condition from 4-7 wells (dots) and two independent experiments. J-L: Changes in mitochondrial Ca²⁺ (Ca²⁺-mit), in wild type (J) and Aralar-KO neurons (K) transfected with 4mtD3cpv probe, after 5 and 25 μm NMDA stimulation; L: maximum Ca²⁺-mit increment (ΔF max) after NMDA stimulation. Mean ± SEM (bars), 5-10 cells (dots) from two independent cultures; two-way ANOVA, *p ≤ 0.05, post hoc Bonferroni test. All cells were maintained in 2.5 mM glucose and 2 mM Ca²⁺ HCSS. Download Figure 4-2, TIF file ^{(351.2KB, tif)} .

To study the contribution of MCU and Aralar-MAS as Ca²⁺ regulation mechanisms in NMDA stimulation of respiration, we started by silencing Mcu in cortical neurons, as described above. NMDA-stimulated increase in matrix Ca²⁺ was clearly smaller in Mcu-KD than Scrambled neurons both with 25 μm NMDA (Fig. 4C,D) and 5 μm NMDA (see Extended Data Fig. 4-1A,B). However, Mcu silencing did not block completely NMDA-stimulated Ca²⁺ entry in mitochondria as reported previously (Qiu et al., 2013) in agreement with the incomplete block of Ca²⁺ uptake in isolated brain mitochondria from global Mcu-KO mice (Hamilton et al., 2018; Szibor et al., 2020). This diminished Ca²⁺ transport into mitochondria was not translated into an increase in cytosolic Ca²⁺ in Mcu-silenced neurons; rather, 5 and 25 μm NMDA caused a similar increase in cytosolic Ca²⁺ in Mcu-silenced and Scrambled neurons (Extended Data Fig. 4-2A–D).

Figure 5.

NMDA-induced changes in cytosolic pyruvate and lactate, in WT and Aralar-KO primary neuronal cultures. A, D, Cortical neurons from WT and Aralar-KO mice transfected with pyronic (A) and laconic (D) probes were stimulated with 25 μm NMDA in 2.5 mm glucose HCSS, in the presence of 2 mm Ca²⁺ (+Ca²⁺) or 100 μm EGTA (–Ca²⁺). FRET changes reporting cytosolic pyruvate or lactate levels are shown. B, E, Quantification of maximum fluorescence change (ΔF max) in pyruvate (B) and Lac (E) after NMDA stimulation in the presence of Ca²⁺. C, F, Quantification of velocity of increase of pyruvate (C) and Lac (F) (ΔF/min) in the presence of Ca²⁺ as the increment of fluorescence ratio during the first 30 s after stimulation. n = 10-22 per condition from 5-8 independent experiments for all pyruvate and Lac assays; 5 mm pyruvate or Lac was added 3 min after NMDA as a control. Data are mean ± SEM. *p = 0.0133 (t test). In the absence of Ca²⁺, pyruvate and Lac output was the same in Aralar-KO and control neurons (A,D). MCU-KD or Scrambled neurons had similar increases in pyronic or laconic fluorescence on NMDA exposure (Extended Data Fig. 5-1).

Figure 5-1

Effects of NMDA stimulation on cytosolic pyruvate and lactate production in Mcu-silenced neurons. Cytosolic pyruvate (Pyr) (A, C) and lactate (Lac) (B, D) levels of Scr and Mcu-KD neurons transfected with Pyronic and Laconic probes, respectively, upon neuronal stimulation with 5 μm (A, B) or 25 μm NMDA (C, D). Neurons were incubated in 2.5 mM glucose and 2 mM Ca²⁺ HCSS. Data are mean ± SEM, n = 7-15 neurons per condition from 2 independent cultures. Download Figure 5-1, TIF file ^{(293.6KB, tif)} .

Remarkably, 5 μm (Extended Data Fig. 4-1C,D) and 25 μm (Fig. 4E,F) NMDA-stimulated OCR was not decreased by Mcu-silencing as expected if matrix Ca²⁺ played a role in stimulation of respiration; indeed, the stimulation was slightly larger compared with Scrambled neurons. The lack of inhibition of NMDA-stimulated respiration by Mcu silencing cannot be because of incomplete inhibition of Ca²⁺ entry in mitochondria, as that would result in small inhibition or no inhibition of NMDA-stimulated OCR, but not in an even larger response. Indeed, Mcu-KD in cultured neurons was effective in preventing glutamate excitotoxicity (Qiu et al., 2013). Therefore, the results clearly indicate that cortical neurons using glucose do not require MCU and matrix Ca²⁺ to upregulate respiration, in response to NMDA-induced increase in workload.

To explore the role of Aralar, we have studied the respiratory response to 5 and 25 μm NMDA. Stimulation of respiration by 5 μm (Extended Data Fig. 4-1E,F) and 25 μm (Fig. 4G,H) NMDA was strikingly blunted in Aralar-KO compared with WT, as found in response to 50 μm glutamate (Llorente-Folch et al., 2016), whereas the absence of Aralar did not change the NMDA-dependent rise in cytosolic (Extended Data Fig. 4-2E–I) and mitochondrial Ca²⁺ (Extended Data Fig. 4-2J–L) at either of the two NMDA concentrations. Therefore, the presence of Aralar-MAS and not MCU has a critical role in Ca²⁺ regulation of respiration in response to NMDA.

Activation of NMDARs increases cytosolic pyruvate and lactate production in a Ca²⁺-dependent way

The effect of the lack of Aralar on Cch-stimulated respiration is rescued by exogenous pyruvate (Fig. 3K,L) (Llorente-Folch et al., 2013, 2016). Therefore, we have analyzed the impact of the presence or absence of Aralar in cytosolic pyruvate and lactate levels on NMDA stimulation.

To this end, cortical neurons were transfected with pyronic or laconic FRET nanosensors (San Martín et al., 2013, 2014) to register pyruvate or lactate levels, respectively, as described previously (Juaristi et al., 2019a). The addition of 25 μm NMDA to neurons in the presence of Ca² resulted in a prominent increase in both cytosolic pyruvate and lactate levels. The increase in both metabolites was drastically reduced in a Ca²⁺ free medium (Fig. 4I,J). NMDA-induced pyruvate and lactate production was dependent on NMDA concentration (Fig. 4K,L).

We next addressed the origin of lactate and pyruvate. Although purity of primary neuronal cultures is high, glial contamination can occur; and the surrounding astrocytes could provide lactate taken up by neurons.

In order to elucidate the origin of these metabolites, bilateral transport across plasma membrane was blocked through pharmacological inhibition of neuronal monocarboxylate transporter 2 (MCT2) with AR-C155858 (AR-C1), a potent inhibitor of both MCT1 and MCT2 (Ovens et al., 2010). MCT2 catalyzes the cotransport of a monocarboxylate (i.e., pyruvate, lactate, β-hydroxybutyrate) with a proton (Alvarez et al., 2003). To demonstrate the inhibitory effect of AR-C1 in our conditions, we have used the acidification associated with MCT2 activity as readout. Acute addition of 5 mm lactate (Lac) caused the expected intracellular acidification, as indicated from the drop of F₄₉₀/F₄₅₀ fluorescence ratio in BCECF-AM loaded neurons (Fig. 4M). This was prevented by 1 min preincubation with 1 μm AR-C1 revealing the effectiveness of AR-C1 as MCT2 blocker. Acute inhibition of neuronal MCT2 with 1 μm AR-C1 leads to an accumulation of lactate per se, that further increases with 25 μm NMDA stimulation (Fig. 4N); thus, the total increase in cytosolic lactate after AR-C1 + NMDA is similar to that caused by NMDA alone (Fig. 4N,O). These results show that lactate arises from neuronal synthesis rather than by capture from the external media, as shown in activated brain neurons (Díaz-García et al., 2017). It should be noted that neuronal cultures have a relatively large lactate production even in the basal state (Juaristi et al., 2019a, 2019b), possibly larger than that of brain neurons, likely limited by astrocyte lactate buildup in the extracellular space. However, regardless of the actual basal state, NMDA clearly stimulates lactate production even in the presence of 2 mm lactate in the media (Fig. 4P).

In the case of pyruvate, 1 μm AR-C1 also increased rather than decreased pyruvate accumulation (results not shown), indicating an internal origin of pyruvate. This origin is glycolytic, since 30 min pre-incubation with 0.5 mm IAA, an inhibitor of GAPDH, prevented neuronal generation of pyruvate (Fig. 4Q), and partially inhibited lactate production (results not shown), confirming its glycolytic origin. Together, our results show that neurons stimulate glycolytic production of pyruvate and lactate on 25 μm NMDA action in a Ca²⁺-dependent manner.

NMDA-induced pyruvate production drops in the absence of Aralar-MAS

NMDA (25 μm)-induced pyruvate production decreased in Aralar-KO neurons with the maximum pyruvate increase reduced by 44% compared with WT (Fig. 5A,B; 0.059 ± 0.007 and 0.033 ± 0.006 for WT and Aralar-KO). Pyruvate production in Ca²⁺-free media was largely abolished in Aralar-KO, and WT neurons (Fig. 5A–C). However, NMDA-induced lactate production was not affected by the absence of Aralar (Fig. 5D), as reflected in the same maximum increase (Fig. 5E; 0.093 ± 0.002 and 0.086 ± 0.002 for WT and Aralar-KO), and in the same velocity of lactate production (Fig. 5F; 0.063 ± 0.002 and 0.046 ± 0.001 for WT and Aralar-KO). Ca²⁺ dependence of lactate production on 25 μm NMDA stimulation was also observed in Aralar-KO neurons (Fig. 5D–F).

It should be noted that, following the application of a smaller workload, short theta bursts (40 pulses in 11 s) changes in lactate and pyruvate were not detected in hippocampal neurons (Baeza-Lehnert et al., 2019) probably because of the small dynamic range of the probe, especially around basal metabolite levels (San Martín et al., 2014; Juaristi et al., 2019a). This may explain the variability in stimulation-induced increase in pyronic signal in WT and Aralar-KO cultures found in this study. However, studies in brain neurons found increases in lactate and pyruvate after electrical stimulation (Díaz-García et al., 2017, 2021; Díaz-García and Yellen, 2019) or by arousal-evoked cortical activity (Zuend et al., 2020).

Together, these results suggest that Aralar is required to stimulate mitochondrial respiration in response to an NMDA-induced workload in the presence of Ca²⁺ by providing pyruvate to mitochondria. The Ca²⁺-dependent stimulation of pyruvate generation is consistent with the following: (1) the stimulation of MAS by Ca²⁺ conferred by Aralar in isolated brain mitochondria (Pardo et al., 2006; Contreras et al., 2007); (2) the stimulation by Ca²⁺ of respiration on glutamate + malate of brain mitochondria (Gellerich et al., 2009, 2012, 2013; Szibor et al., 2020); and (3) the observed recovery of Aralar-KO neuronal respiration by external pyruvate administration during stimulation (Fig. 3L) (Pardo et al., 2006; Llorente-Folch et al., 2013). On the other side, Mcu-KD had no influence on NMDA-induced production of pyruvate and lactate (Extended Data Fig. 5-1A–D).

Activation of NMDARs activates glucose uptake and glycolysis in a Ca²⁺-dependent way

We wondered about the origin of glucose fueling this process and its stimulation by NMDA. To monitor cytosolic glucose, we have used the glucose FRET nanosensor FLII¹²Pglu-700μδ6 (Takanaga et al., 2008), which is pH-independent, thanks to the replacement of eYFP by citrine (Takanaga et al., 2008; Bittner et al., 2010). Indeed, NMDA induces a rapid cytosolic acidification (Fig. 6A, see comparison with that caused by acute addition of HCl to BCECF-loaded neurons) as shown previously (Wu et al., 1999; Rathje et al., 2013). However, the response to NMDA observed with FLII¹²Pglu-700μδ6 was an upward deflection of the FRET ratio consistent with an increase in glucose levels (Fig. 6B). In contrast, HCl addition did not cause any increase in FRET ratio, but rather a gradual decrease probably because of side effects and probe quenching. Therefore, the FLII¹²Pglu-700μδ6 fluorescence changes observed in our experiments are because of changes in cytosolic glucose levels.

Figure 6.

25 μm NMDA-induced changes in cytosolic glucose levels. A, B, Changes in cytosolic pH (A) or intracellular glucose (B) in BCECF-loaded or FLII¹²Pglu-700μδ6 transfected neurons after the acute addition of 25 μm NMDA or 5 mm HCl. Data are mean ± SEM; recordings from 5 or 6 cells per condition. C, Representative scheme Ca²⁺-regulated Aralar-MAS activation upregulates glycolysis and pyruvate production, which fuels mitochondrial respiration, through regulation of cytosolic NAD⁺/NADH ratio. Image created with www.BioRender.com. D, F, 25 μm NMDA induced changes in glucose levels in WT (D) and Aralar-KO (G) neurons, with or without calcium or with a 20 s pre-incubation with 50 μm CytB or 0.5 mm IAA. E, H, Velocity of increase of glucose levels during the first 30 s after stimulation (ΔF/min) in WT (E) and Aralar-KO (H) neurons. F, I, FRET ratio change between pre-addition of NMDA and the end of the recording (ΔF endpoint) in WT (F) and Aralar-KO (I) neurons. Data are mean ± SEM; recordings from 7 to 17 cells per condition from 3 to 5 independent experiments. One-way ANOVA: *p ≤ 0.05, **p ≤ 0.005, ***p ≤ 0.001, post hoc Bonferroni test. ΔF at the endpoint after NMDA addition is higher in Aralar-KO than in WT neurons. Two-way ANOVA: **p ≤ 0.005, post hoc Bonferroni test. Assays using BCECF and FLII¹²Pglu-700μδ6 probes were performed in 2.5 mm glucose and 2 mm Ca²⁺ (+Ca²⁺) or 100 μm EGTA (–Ca²⁺) HCSS.

Neuronal stimulation is expected to decrease intracellular glucose levels, because of activation of glycolysis to cope with ATP demand. However, the results depicted in Fig. 6D (black traces) show that FLII¹²Pglu-700μδ6 fluorescence ratio increases immediately after 25 μm NMDA stimulation. The rise in cytosolic glucose levels is maintained during a short period, and then steadily declines, suggesting an initial upregulation of glucose uptake or synthesis which is subsequently consumed.

To determine whether the increase in glucose obtained after NMDA exposure arises from uptake from the external medium, cortical neurons were treated with 50 μm CytB, an inhibitor of plasma membrane glucose transporters (Basketter and Widdas, 1978), 30 s before the addition of 25 μm NMDA (Fig. 6D, purple traces). NMDA-induced increases in citrine-eYFP/eCFP fluorescence ratio in terms of slope (Fig. 6E) and at the end of the experiment (ΔF endpoint; Fig. 6F) are completely abolished in the presence of CytB, indicating that the increase in intracellular glucose arises from uptake from extracellular medium.

The following results clarify that the subsequent steady decline in cytosolic glucose is because of glycolysis: (1) NMDA-stimulated glucose consumption still continues in the presence of CytB (Fig. 6D, purple traces); and (2) when 0.5 mm IAA was administered before NMDA addition (Fig. 6D, blue traces), glucose levels dramatically increased consistent with a clear inhibition of glucose consumption by glycolysis. By using mild electrical stimulation rather than bath application of NMDA, such an increase in cytosolic glucose (recorded with the same probe) and subsequent decrease was not observed (Baeza-Lehnert et al., 2019), possibly because increased uptake was balanced with increased glucose consumption.

Surprisingly, the NMDA-dependent rise in cytosolic glucose completely disappears in the absence of Ca²⁺, as shown by the block in the initial increase in glucose levels (Fig. 6D,E, gray traces), and the NMDA-dependent decay in cytosolic glucose is also slower (ΔF endpoint; Fig. 6F). Together, these results indicate that activation of NMDARs in cortical neurons upregulates glucose uptake from the external medium and its immediate consumption in glycolysis, both processes being dependent on Ca²⁺.

NMDA stimulation of glycolysis, but not glucose uptake, is strongly inhibited in Aralar-KO neurons

We next studied the involvement of Aralar in the upregulation of glucose uptake and glycolysis. Strikingly, NMDA-induced a much higher increase in glucose levels in Aralar-KO than in WT neurons (Fig. 6F,I; +Ca²⁺ in WT and Aralar-KO; p = 0.0009, unpaired t test), with no differences in velocity of increase between genotypes (Fig. 6E,H). NMDA-elevated glucose levels in Aralar-KO neurons were maintained throughout the recording interval instead of being immediately consumed as in WT neurons (Fig. 6G,I, black traces). NMDA-induced increase in glucose was blocked by incubation with 50 μm CytB (Fig. 6G,H, purple traces) and in Ca²⁺-free medium (Fig. 6G, gray traces), confirming that it was because of uptake from the external medium and was Ca²⁺-dependent also in Aralar-KO neurons. Thus, glucose uptake in neurons seems to be a Ca²⁺-dependent process, but independent from Aralar.

Unlike WT neurons (Fig. 6D), Aralar-KO neurons failed to consume cytosolic glucose in the absence or presence of CytB during the recording interval (Fig. 6G, compare black and purple traces). The very low cytosolic glucose consumption in Aralar-KO neurons was confirmed in incubations with 0.5 mm IAA, which caused only a small increase in glucose accumulation (Fig. 6I) compared with its effects in WT neurons (Fig. 6F). Together, the results reveal that (1) NMDA-induced glucose uptake is Ca²⁺-dependent in cortical neurons and is not affected by the lack of Aralar, and (2) NMDA-induced stimulation of glycolysis is also Ca²⁺-dependent and requires Aralar (Fig. 6C).

Basal glucose consumption and lactate production are the same in WT and Aralar-KO neurons (0.40 ± 0.08 and 0.43 ± 0.08 μmol glucose/mg/h, and 0.42 ± 0.05 and 0.44 ± 0.07 μmol lactate/mg/h, respectively) (Juaristi et al., 2017). In the presence of Ca²⁺, NMDA-induced stimulation of glycolysis is much smaller in Aralar-KO than in WT neurons, both in the absence (compare the glucose disappearance rate in WT and Aralar-KO in Fig. 6D,G, black traces) or presence of CytB (compare purple traces in Fig. 6D,G). This is followed by a smaller stimulation of pyruvate formation (Fig. 5A) and respiration (Fig. 4G,H), while lactate formation was not significantly altered (Fig. 5D). This entails that the fate of the small NMDA increase in glycolysis in Aralar-KO neurons is largely lactate, with a diminished pyruvate formation and respiration.

Ca²⁺ dependence of NMDA stimulation of glycolysis and OXPHOS: workload and/or signaling

The Ca²⁺ dependence of NMDA stimulation of glycolysis-pyruvate formation-respiration may be because of a feedback mechanism arising as a consequence of Ca²⁺-induced increased workload caused by NMDA exposure in the presence of Ca²⁺ and also to feedforward Ca²⁺ signaling on downstream targets. This dual role of Ca²⁺ is well known in the case of the muscle metabolic response to contraction (Balaban, 2009; Richter and Hargreaves, 2013; Hargreaves and Spriet, 2020). To tell apart these possibilities, we have measured the changes in intracellular Na⁺, Ca²⁺, ATP, and ATP/ADP ratio on addition of 25 μm NMDA to cortical neurons in the presence or absence of external calcium. As shown in Figure 7, the increase in cytosolic Ca²⁺ was abolished in Ca²⁺-free medium, and the increases in Na⁺ triggered by NMDA were the same in the presence or absence of Ca²⁺ in the external medium. Thus, the combined workload (Na⁺ + Ca²⁺ extrusion) is higher in the presence than in the absence of Ca²⁺. Accordingly, the fall in cytosolic ATP and in the ATP/ADP ratio is larger in the presence than in the absence of Ca²⁺ (Fig. 7). These results with 25 μm NMDA differ from those obtained with 100 μm NMDA (Rueda et al., 2015) and indicate that, in the presence of Ca²⁺, NMDA induces a larger workload than in its absence. We conclude that the larger fall in the ATP/ADP ratio could upregulate respiration, pyruvate supply, and glycolysis simply via increased ATP demand (i.e., via a feedback mechanism).

Figure 7.

Dependence of external calcium or BAPTA-AM of the changes in cytosolic Ca²⁺, Na⁺, ATP, and ATP/ADP ratio and OCR induced by 25 μm NMDA in neocortical neuronal cultures. A, E, Cytosolic Ca²⁺ measurements in fura-2-loaded neurons; B, F, cytosolic Na⁺ in SBFI-loaded neurons; C, G, cytosolic ATP measurement in GO-ATeam-transfected neurons; D, H, cytosolic ATP/ADP ratio in Perceval-transfected neurons. A–D, In the presence of Ca²⁺. E–H, In nominally Ca²⁺-free media. The results correspond to representative experiments. I, K, OCR response to 25 μm NMDA in wt and Aralar KO neurons incubated with Ca²⁺ and in presence or absence of 1 μm BAPTA-AM. Data are mean ± SEM; n = 4-9. J, Histogram represents the decrease (***p < 0.001, n = 9, unpaired t test) in NMDA-induced stimulation of respiration in wt neurons in the presence of BAPTA-AM. K, In Aralar-KO neurons, the decrease was not significant.

To investigate whether Ca²⁺ signaling could also play a role in the effects of Ca²⁺ in addition to that on workload, we have used BAPTA-AM, a rapid intracellular Ca²⁺ chelator. BAPTA loading blunts cytosolic Ca²⁺ signals (Llorente-Folch et al., 2013) and hence Ca²⁺ signaling, but does not reduce Ca²⁺ inflow through NMDARs (Legendre et al., 1993) or voltage-dependent Ca²⁺-channels (Adler et al., 1991), thus maintaining the Ca²⁺-dependent fraction of the global workload. In the presence of Ca²⁺, incubation with 1 μm BAPTA-AM diminishes the OCR response to NMDA in control neurons (Fig. 7I,J), suggesting that Ca²⁺ signaling, and not only Ca²⁺-dependent workload, contributes to upregulation of OCR. Interestingly, BAPTA-AM had no detectable effect in Aralar-KO neurons (Fig. 7K), suggesting that the absence of Aralar blocks the Ca²⁺ signaling mechanism present in control neurons. Together, the results suggest a major role of Ca²⁺-dependent workload in the metabolic response to NMDA (i.e., a feedback effect) and also a signaling role of calcium relying on Aralar-MAS in boosting this response. Although the Ca²⁺-binding N-terminus of Aralar is likely the Ca²⁺ target involved in NMDA-dependent MAS activation by Ca²⁺, we cannot exclude that other Ca²⁺ sensors play a role, as the disruption of Aralar removes both protein domains, Ca²⁺ sensing and mitochondrial carrier. A rigorous study of this possibility will require selective inactivation of the Ca²⁺-binding domain while maintaining an intact carrier domain.

Discussion

Constitutive Ca²⁺ flow from ER to mitochondria maintains basal respiration in neurons depends on RyR2, and not IP3R, and does not require MCU

The constitutive Ca²⁺ flow from ER to mitochondria has been shown to regulate basal respiration in many cell lines (Cárdenas et al., 2010; Filadi et al., 2018), hepatocytes (Tomar et al., 2019), and skeletal muscle (Diaz-Vegas et al., 2018). We find that such a pathway also operates in neurons, with RyR2, but not IP3R, as source of ER Ca²⁺. This would seem surprising since IP3R are in close apposition to mitochondria, and IP3R1 is abundant in brain neurons (Egorova and Bezprozvanny, 2018). However, the role of RyR in controlling basal respiration is not new; RyR and IP3R control basal respiration in skeletal muscle (Diaz-Vegas et al., 2018). Indeed, RyR are also close to mitochondria in neurons (Jakob et al., 2014), facilitate a rapid Ca²⁺ transfer from ER/SR to the matrix (Hajnóczky et al., 2002), participate in MAM formation in brain (Ma et al., 2017; Völgyi et al., 2018), and are positioned in MCU enriched nanodomains in heart mitochondria (de la Fuente et al., 2018), therefore in an adequate position to participate in a constitutive Ca²⁺ flow to mitochondria.

We find that Mcu-KD has no influence on neuronal basal respiration. It could be argued that this is because of compensatory adaptations to MCU deletion masking its true role. By using shRNAs to acutely silence Mcu in neurons, our approach is not expected to cause compensatory effects; therefore, the lack of effect of Mcu-KD in basal respiration (this work and Nichols et al., 2018) shows that MCU is not part of the ER to mitochondria Ca²⁺ flow regulating basal energetics in neurons using glucose. The Ca²⁺ sensor on mitochondria remains to be identified.

Stimulation of mitochondrial respiration by agonists: stimulation by Cch requires Aralar-MAS but not MCU

Our results clearly show that, in cortical neurons with spontaneous Ca²⁺ activity, Cch induces an increase in Ca²⁺ oscillations and an increase in neuronal respiration. Cch-induced Ca²⁺ oscillations reached mitochondria, resulting in an increase in matrix Ca²⁺. Although matrix Ca²⁺ signals were definitely reduced in Mcu-silenced neurons, Mcu-KD did not affect Cch-stimulated respiration.

In contrast, Cch stimulation of respiration was blunted in Aralar-KO neurons. This reflects a limitation in substrate supply to mitochondria, as pyruvate addition abolished the differences in Cch-stimulated OCR between Aralar-KO and WT neurons. As neither MCU nor Aralar deficiency altered Cch-induced cytosolic Ca²⁺ oscillations, these results suggest that the Aralar-MAS pathway, but not MCU-driven Ca²⁺ entry in mitochondria, is involved in boosting respiration by Cch-enhanced Ca²⁺ oscillations. The mechanism whereby Aralar-MAS regulates respiration in response to Cch becomes apparent from the analysis of the neuronal response to NMDA.

Stimulation of mitochondrial respiration by agonists: NMDA stimulation in cortical neurons causes Ca²⁺-dependent increases in glucose uptake, glycolysis, pyruvate and lactate formation, and respiration

We have shown that NMDA exposure results in a Ca²⁺-dependent stimulation of respiration, and Ca²⁺-dependent increases in cytosolic pyruvate and lactate, both of which arose from glycolysis.

NMDA also triggers a Ca²⁺-dependent disappearance of cytosolic glucose blocked by inhibitors of glycolysis, showing that neurons respond to NMDA with a Ca²⁺-dependent stimulation of glycolytic glucose utilization. Additionally, NMDA stimulates Ca²⁺-dependent glucose uptake.

It makes sense that Ca²⁺ activates at the same time respiration, substrate supply to mitochondria, glycolysis, and glucose uptake in response to an increase in workload. As neurotransmission is linked to Ca²⁺ signals, in addition to a direct feedback response to the increased workload, Ca²⁺ regulation of all these steps can be taken as a feedforward mechanism to increase glycolysis and mitochondrial function on initiation of the workload itself. But does this occur through a single mechanism or are there several Ca²⁺-dependent individual steps controlling the fate of glucose in neurons up to its combustion in mitochondria? Based on our results, we suggest two Ca²⁺-dependent steps: one regarding glucose uptake and another one involving Aralar-MAS driven glycolysis and OXPHOS regulation. Our results show that glucose uptake has a specific Ca²⁺ regulation system independent of glycolysis. An increase in glucose uptake on synaptic activity or NMDA stimulation has been observed in both cerebellar and cortical neuronal cultures (Ferreira et al., 2011; Connolly et al., 2014). It required GLUT3/Slc2a3 translocation to the plasma membrane, via activation of AMPK (Connolly et al., 2014) or nNOS and cGMP-dependent protein kinase (Ferreira et al., 2011). Ashrafi and Ryan (2017) have found that, similarly to muscle, activity at synapses triggers the rapid (seconds) translocation of GLUT4 to the plasma membrane, driven by AMPK. As CaMKKβ is involved in AMPK activation by depolarization in neurons (Kawashima et al., 2012), it is possible that these mechanisms provide Ca²⁺ sensitivity to the NMDA-dependent increase in glucose uptake found in this study.

Regarding the other steps beyond glucose uptake, in one possible scenario there may be at least two. One is Ca²⁺ regulation of glycolysis, and the other is Ca²⁺ activation of mitochondrial dehydrogenases (Denton, 2009) which may boost the activity of the tricarboxylic acid cycle, and the use of pyruvate when Ca²⁺ enters the matrix. Ca²⁺ stimulation of glycolysis may involve a Ca²⁺-dependent activation of PFKFB3 (Rider et al., 2004). PFKFB3 is activated by AMPK (Almeida et al., 2004), which in turn is phosphorylated by CAMKKβ in response to rise in cytosolic Ca²⁺ (Sanders et al., 2007; Xiao et al., 2011). However, expression of PFKFB3 in neurons is low because of continuous degradation (Herrero-Méndez et al., 2009), and the Yellen group has shown that AMPK is not involved in regulation of glycolysis in stimulated neurons (Díaz-García et al., 2021), suggesting that this mechanism does not operate in response to acute NMDA stimulation. On the other hand, Ca²⁺ activation of mitochondrial dehydrogenases requires MCU.

In another scenario, glycolysis and OXPHOS are Ca²⁺-modulated by the same mechanism, Aralar-MAS. The results from this study support this second possibility.

Ca²⁺ dependence of NMDA stimulation of glycolysis, pyruvate formation, and respiration depends on Aralar-MAS and not MCU

Glycolysis can only proceed if there is sufficient supply of NAD⁺. The oxidation of glucose requires NAD⁺ (2 molecules for each glucose) that becomes NADH in the GAPDH reaction. Therefore, to sustain glycolysis, NAD⁺ needs to be continuously regenerated either by lactate dehydrogenase with conversion of pyruvate to lactate, or by NADH shuttle systems of which MAS is prevalent in neurons (Juaristi et al., 2019a). MAS diverts pyruvate away from lactate and into OXPHOS and is regulated by cytosolic Ca²⁺ (Pardo et al., 2006; Contreras et al., 2007; Gellerich et al., 2012, 2013, 2020; Llorente-Folch et al., 2013, 2015; Rueda et al., 2014), providing a potential feedforward mechanism linking early Ca²⁺ signals arising from synaptic activity to both glycolysis and OXPHOS, as suggested previously (Ashrafi and Ryan, 2017).

The results from the present study indicate that, in neurons, using glucose NMDA triggers a substantial Ca²⁺-dependent workload met by a Ca²⁺-regulated response in which Ca²⁺ boosts glycolysis and OXPHOS by means of Ca²⁺ upregulation of Aralar-MAS. In contrast, Mcu-KD neurons were unaffected in terms of NMDA stimulation of respiration or pyruvate production. Although NMDA-induced Ca²⁺ entry in mitochondria was only partially blocked by MCU-KD, this indicates that MCU-induced Ca²⁺ entry in mitochondria and activation of mitochondrial dehydrogenases (McCormack and Denton, 1993) are largely dispensable in the stimulation of respiration of neurons using glucose as also clearly shown by Díaz-García et al. (2021) and Szibor et al. (2020).

In conclusion, it is important to emphasize that the reliance on Aralar-MAS as Ca²⁺ regulation switch connecting neuronal activity to metabolic response rather than MCU is limited to the use of glucose as substrate. Using a mixture of lactate and pyruvate instead of glucose, Ashrafi et al. (2020) have found that MCU is essential for synaptic vesicle endocytosis and the maintenance of ATP homeostasis after electrical stimulation in nerve terminals. By providing external pyruvate and not glucose, Aralar-MAS function is bypassed, allowing a role of MCU and possibly matrix Ca²⁺ activation of NAD⁺-linked isocitrate dehydrogenase, alpha-KGDH and pyruvate dehydrogenase in the control of respiration. It is likely that matrix pyruvate levels need to increase beyond those present in neurons using glucose to allow a functional role of Ca²⁺-dependent activation of mitochondrial dehydrogenases in upregulating neuronal respiration.