Building a Bridge Between NMDAR-Mediated Excitotoxicity and Mitochondrial Dysfunction in Chronic and Acute Diseases

Abstract

Glutamate is the major excitatory neurotransmitter in the brain, and it is widely accepted to play a role in synaptic plasticity and excitotoxic cell death. Glutamate binds to several receptors, including ionotropic N-methyl-d-Aspartate receptor (NMDAR), which is essential in synaptic plasticity and excitotoxicity. This receptor is a calcium channel that is located in synaptic and extrasynaptic sites, triggering different signalling cascades in each case. The calcium entry through extrasynaptic NMDARs is linked to calcium overload in the mitochondria in neurons in vitro. The mitochondria, besides their role in ATP production in the cell, participate in calcium homeostasis, acting as a buffering organelle. Disruption of mitochondrial calcium homeostasis has been linked to neuronal death either by triggering apoptosis or driven by the opening of the mitochondrial transition pore. These cell-death mechanisms contribute to the pathophysiology of diverse diseases such as neurodegenerative Alzheimer’s disease or Parkinson’s disease, and acute neuropathological conditions such as stroke or traumatic brain injury. In this review, we will address the available evidence that positions the mitochondria as an essential organelle in the control of calcium-mediated toxicity, highlighting its role from the perspective of specific NMDAR signalling microdomains at the level of the central synapse.

Introduction

Glutamate is the main excitatory neurotransmitter in the mammalian brain, and it is essential in synaptic plasticity as a substrate for learning and memory processes. It is packaged into synaptic vesicles in the presynaptic terminals and released to the synaptic cleft, where it binds to glutamate receptors at the postsynaptic density and astrocytes (Perea et al. 2009; Platt 2007; Traynelis et al. 2010). There is a wide variety of glutamate receptors, including ionotropic and metabotropic glutamate receptors. Ionotropic glutamate receptors mediate fast excitatory transmission, and there are three types: α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) receptor, kainic acid (KA) receptor, and N-methyl-d-Aspartate (NMDA) receptor (Platt 2007; Zhu and Gouaux 2017).

Ionotropic glutamate receptors are non-selective ion channels permeable to cations and inserted into the postsynaptic density anchored to scaffold proteins. AMPA and KA receptors are predominantly permeable to Na⁺ and K⁺, depolarizing the membrane once activated. NMDA receptors are also permeable to Na⁺ and K⁺, but importantly, they are permeable to Ca²⁺, linking synaptic activity with several signalling pathways (Willard and Koochekpour 2013; Zhu and Gouaux 2017).

The influx of calcium into the postsynaptic neuron is buffered, in part, by the mitochondria (Todorova and Blokland 2017), the double-membrane organelles responsible for energy production (Picard and McEwen 2014). Mitochondria play critical roles in calcium homeostasis, energy production in the form of ATP, and redox homeostasis and are indispensable in synaptic transmission and synaptic plasticity in health and disease (Picard and McEwen 2014).

Many neurodegenerative diseases have in common dysregulation of excitatory neurotransmission and mitochondrial dysfunction such as Alzheimer’s disease (Wang and Reddy 2017; Wang et al. 2015a, b), schizophrenia (Phensy et al. 2017), and traumatic brain injury (Arundine and Tymianski 2004; Chen et al. 2016). The mechanisms between mitochondrial dysfunction and synaptic failure are partially understood; however, only in vitro models have established a functional link between NMDAR-mediated excitotoxicity and mitochondrial dysfunction. Here, we reviewed this possible functional interaction from NMDAR and mitochondria function in the central nervous system to pathology.

NMDA Receptor: Localization and Signalling Pathways

NMDA receptors are heterotetramers found in postsynaptic neurons mediating fast excitatory transmission and synaptic plasticity (Vyklicky et al. 2014). They consist of two obligatory GluN1 subunits and two other subunits, which may be GluN2A, GluN2B, GluN2C, GluN2D, GluN3A or GluN3B (Traynelis et al. 2010; Ulbrich and Isacoff 2008). GluN1 can vary between eight spliced alternative forms, and GluN2A and GluN2B are the most common subunits found in NMDAR in the synapse of cortical and hippocampal neurons (Vyklicky et al. 2014). This ionotropic glutamate receptor is permeable to Na⁺, K⁺, and Ca²⁺, and it is basally blocked by a Mg²⁺ ion. This receptor needs a subtle depolarization (given by the activation of AMPA receptors), which removes the Mg²⁺, permitting the ion fluxes (Cerpa et al. 2016). In addition, the activation of this receptor requires a co-agonist. Glycine and serine are two described co-agonists for NMDA receptors with different binding affinities and functions (Papouin et al. 2012; Vyklicky et al. 2014).

NMDA receptors are a major player in synaptic plasticity and a central modulator in long-term potentiation (LTP) and long-term depression (LTD) in hippocampal synapses. This receptor mediates complex processing in the brain, and it is involved in learning and memory (Luscher and Malenka 2012; Volianskis et al. 2015). This receptor has been described as a critical player in glutamate-mediated toxicity or excitotoxicity, where neurons degenerate in response to excessive glutamate stimulation (Mehta et al. 2013). Thus, NMDAR is implicated in health and disease, with a dual role in synaptic plasticity and neurodegeneration, and is becoming an interesting target of modulation (Hardingham and Bading 2010).

NMDAR is found in the postsynaptic density, anchored to scaffold proteins of the Membrane-associated guanylate kinases (MAGUKs) family, specially PSD-95 and PSD-93 (Frank and Grant 2017). NMDARs are also found in extrasynaptic sites that are stable and functional (Hardingham et al. 2002; Papouin and Oliet 2014; Scheefhals and MacGillavry 2018). GluN2A subunit of NMDAR is preferentially anchored to PSD-95, while GluN2B subunit of NMDAR is preferentially anchored to SAP102 (Sans et al. 2000). NMDA receptors can move from one destination to another through lateral diffusion, a process that is regulated by phosphorylation and interaction of the receptor with other proteins (Carvajal et al. 2016). Kinases and phosphatases modulate the phosphorylation state of the NMDARs such as Fyn, PKA, PKC, CaMKII, and CKII kinases, and phosphatases such as STEP (Qiu et al. 2011). For example, GluN2B has three phosphorylation sites for Fyn kinase, which are hallmarks of receptor localization. Phosphorylation of tyrosine 1472 is enriched in the receptor in synaptic localization, preventing endocytosis (Trepanier et al. 2012), and phosphorylation of tyrosine 1336 in the same receptor subunit is enriched in extrasynaptic receptors (Goebel-Goody et al. 2009). Alternatively, scaffold proteins and their phosphorylation state are also determinants of localization of the receptor. NMDARs interact with PDZ domains of scaffold proteins such as PSD-95, PSD-93, and SAP102. The phosphorylation of serine 1480 in the C-terminus of the GluN2B subunit by casein kinase II and CaMKII complex (Sanz-Clemente et al. 2013) disrupts the interaction with PDZ domain of PSD-95 and SAP102 scaffold proteins, reducing the synaptic surface expression of the receptor (Chung et al. 2004). Moreover, the p-Ser1480 of GluN2B increases the extrasynaptic content of the receptor in a protein complex with protein phosphatase 1 (PP1) (Chiu et al. 2019). In this context, SAP102 has been identified as a molecular component important in the internalization of the receptor through endocytosis, interacting with GluN2B through no-PDZ domains (Chen et al. 2012). Interestingly, during stroke, it has been demonstrated that death-associated protein kinase 1 (DAPK1) associates with GluN2B C-terminal tail and phosphorylate it at Ser1303. This phosphorylation increases the conductance of the extrasynaptic NMDAR, increasing calcium influx and cell death (Tu et al. 2010). However, more recent evidence has pointed out that excitotoxicity through NMDAR occurred independently of this phosphorylation and DAPK1 (McQueen et al. 2017). Regarding GluN2A, a phosphorylation site increases the synaptic localization of GluN2A-containing NMDARs. The Ser1048 is phosphorylated by the dual specificity tyrosine-phosphorylation-regulated kinase 1A (DYRK1A) and hinders the internalization of these receptors (Grau et al. 2014). Another important kinase is the glycogen-synthase kinase 3-β (GSK-3β), where it could inhibit the surface expression of GluN2A-containing NMDAR. The inhibition of GSK-3β by lithium increases currents through GluN2A NMDARs (Monaco et al. 2018).

Moreover, these two pools of NMDA receptors had been described to induce differential responses in hippocampal neurons after activation (Hardingham and Bading 2010). Synaptic NMDA receptor activation is related to the transcription of pro-survival genes and anti-apoptotic genes (Hardingham et al. 2001). Synaptic NMDAR activation leads to calcium influx into dendritic spines, with the consequent calcium release from internal stores, generating calcium signals to the nucleus. In the nucleus, calcium can activate kinases such as Ca²⁺-Calmodulin kinase IV (CaMKIV) (Hardingham et al. 2001). This calcium influx triggers calcium signalling into the nucleus, activating the transcription factor CREB (Hardingham et al. 2001) and inhibiting the transcription factor FOXO3a (Dick and Bading 2010). This transcription induction results in the expression of the neurotrophic factor brain-derived neurotrophic factor (BDNF) and its release in the brain (Hardingham et al. 2001). Synaptic activation of NMDAR also increases MAPK ERK1/2 activity (Ivanov et al. 2006; Zheng et al. 2010) and Akt activity to promote the inhibition of FOXO (Dick and Bading 2010). In contrast, the activation of extrasynaptic NMDAR leads to neurodegenerative processes, inhibiting pro-survival gene transcription and enhancing pro-apoptotic signalling. Extrasynaptic NMDAR activation leads to the shut-off of the CREB pathway (Hardingham and Bading 2002; Hardingham et al. 2002), activation of the FOXO transcription factor (Dick and Bading 2010), inhibition of ERK1/2 signalling (Ivanov et al. 2006), and activation of calpain and subsequent cleavage of STEP tyrosine phosphatase (Xu et al. 2009). Interestingly, the activation of calpains is selective between synaptic and extrasynaptic NMDA receptors mediating neuroprotection or neurodegeneration. Synaptic NMDAR leads to µ-calpain activation and the cleavage of the PH domain and Leucine-rich repeats Protein Phosphatase 1 (PHLPP1) with the sequential activation of Akt and ERK1/2 pathways. Extrasynaptic NMDAR, on the other hand, leads to the activation of m-calpain and the subsequent cleavage of STEP and activation of pro-apoptotic pathways (Wang et al. 2013). Despite notable differences in signalling pathways activated downstream of NMDA receptors in different locations, whether extrasynaptic NMDA receptors are the leading players in excitotoxic cell death is still debated. It has been proposed that the activation of both synaptic and extrasynaptic NMDA receptors are crucial in cell death, and even more, specific synaptic NMDAR activation leads to cell death in hypoxia (Papouin and Oliet 2014; Wroge et al. 2012; Zhou et al. 2013).

In normal conditions, extrasynaptic NMDA receptors comprise a vast proportion of NMDA receptors found in hippocampal CA1 pyramidal neurons, corresponding to 36%, which is stable (Harris and Pettit 2007). Indeed, the activation of the cell-death programme develops upon the activation of both synaptic and extrasynaptic NMDA receptors (Stanika et al. 2009). Moreover, the subunit composition of these pools of NMDA receptors is not very different. It has been proposed that extrasynaptic NMDA receptors are enriched in the GluN2B subunit. However, this subunit is also found in synaptic NMDA receptors (Groc et al. 2006). Extrasynaptic NMDA receptors are responsible for physiological tonic NMDA-mediated currents not sensible to tetrodotoxin (TTX). These currents rely on extracellular glutamate and co-agonist concentration (Le Meur et al. 2007), regulating dendritic inputs and neuronal excitability (Papouin and Oliet 2014; Riebe et al. 2016). The activation of extrasynaptic NMDA receptor by glutamate release by astrocytes generates slow-inward currents (SICs), which are believed to participate in the talk between neurons and glia (Angulo et al. 2004; Fellin et al. 2004). However, this question is still open, and the glia–neuron interaction through SICs had not been well characterized. In synaptic plasticity, the extrasynaptic NMDA receptor had been involved in long-term depression (LTD) as well as synaptic NMDA receptors (Liu et al. 2013,2004; Papouin et al. 2012). On the other hand, long-term potentiation (LTP) had only been related to synaptic NMDA receptor function (Liu et al. 2004; Papouin et al. 2012; Volianskis et al. 2015), although recent evidence suggests a role for extrasynaptic NMDA receptors in LTP. Interestingly, extrasynaptic NMDARs containing the subunit GluN2A are responsible for this type of LTP, and GluN2B-containing NMDARs are not necessary for this type of plasticity (Yang et al. 2017). A requirement of LTP and LTD is the autophosphorylation of CaMKII. In LTP, phosphorylated CaMKII associates with GluN2B subunit of NMDAR but this binding is not necessary for LTD, despite autophosphorylation (Barcomb et al. 2016; Halt et al. 2012). In LTD, calcineurin dephosphorylates DAPK1, which impedes the binding of CaMKII to GluN2B (Goodell et al. 2017). Importantly, in pathology, an imbalance between synaptic and extrasynaptic NMDA receptor activity has been reported, principally mislocalization and defects of trafficking (Gladding and Raymond 2011), and increased tonic NMDA receptor-mediated currents (Papouin and Oliet 2014). This issue will be reviewed in the last section of this article.

Mitochondrial Calcium and Its Role in the Central Nervous System

Mitochondria are dynamic double-membrane organelles known as the powerhouse of the cell. Mitochondria contain an outer membrane, an intermembrane space, an inner membrane, and the mitochondrial matrix. The functions of mitochondria comprise ATP production through oxidative phosphorylation (OXPHOS) coupled to electron transport chain (ETC), redox and calcium homeostasis, and central metabolic pathways of the cell such as the tricarboxylic acid cycle and β-oxidation of fatty acids. The activity of the ETC leads to the generation of pH and ion gradient leading to mitochondrial membrane potential, characteristic of this organelle, and a marker of correct function (Tzameli 2012).

Calcium homeostasis is an important function of the mitochondria (Fig. 1); this ion is a regulator of its metabolic enzymes, and it is essential in neuron physiology (Picard and McEwen 2014). The outer mitochondrial membrane is intrinsically permeable to ions, and the main pathway of calcium entry is the voltage-dependent anion channel (VDAC), a widely expressed porin (Bathori et al. 2006; Shoshan-Barmatz and De 2017). The inner mitochondrial membrane, on the other hand, is particularly impermeable to ions, and the main route of calcium entry is the complex known as the mitochondrial calcium uniporter (MCU). It is composed of MCU protein (previously known as Ccdc109a), MICU1, MICU2, MICU3, MCUb, and EMRE (De Stefani et al. 2016; Liu et al. 2016; Williams et al. 2013; Xu et al. 2016). Other mitochondrial calcium influx routes are the mitochondrial ryanodine receptor (mRyR) and Ram protein (Xu et al. 2016). The calcium efflux is mediated by exchangers that use H⁺ or Na⁺ electrochemical gradients (De Stefani et al. 2016). The best-characterized exchanger is the Na⁺-dependent exchanger called NCLX (Palty and Sekler 2012; Palty et al. 2010). This protein can also remove calcium from the mitochondria in a Li⁺-dependent manner (De Stefani et al. 2016; Kostic and Sekler 2019). Calcium buffering is one of the processes that link mitochondria and endoplasmic reticulum (ER) functions, forming functional structures called mitochondrial associated membranes (MAMs) (Marchi et al. 2017), which regulate many biological functions. The connection between the ER and mitochondria allows the rapid influx of calcium into the mitochondria and regulates ATP production, as well as the regulation of mitochondrial dynamics, lipid biosynthesis, and cell survival (Herrera-Cruz and Simmen 2017; Marchi et al. 2017). Mitochondrial membrane potential, VDAC, and MCU complex are essentials in the calcium movement from the ER to mitochondria, participating in calcium homeostasis (Filadi et al. 2015; Hajnoczky et al. 2002). VDAC and MCU localize near ryanodine receptors (RyR) and IP₃ receptors (IP₃R) in the ER, forming hotspots of high calcium concentration (more than 10 µM). Indeed, VDAC and IP₃R show physical interaction in the MAMs, mediated by the chaperone GRP75 (Marchi et al. 2017). Many other proteins have been implicated in the tethering of these two organelles, such as Mitofusin 2 (Mfn2), Vesicle-associated membrane protein-associated protein B/C (VAPB), Protein tyrosine phosphatase interacting protein 51 P (TPIP51), MICOS complex, and phosphofurin acidic cluster sorting protein 2 (PACS2), among others (Herrera-Cruz and Simmen 2017). Mfn2 is a GTPase responsible of fusion of the outer mitochondrial membrane along with Mitofusin 1 (Mfn1) (Filadi et al. 2018). It is a well-established tether between ER and mitochondria, and the KO of Mfn2 increases the contacts between these two organelles, increasing the efficacy of calcium transfer across IP₃R. Mfn2 acts, therefore, as an antagonist tether that prevents the excessive contacts that could sensitize mitochondria to calcium overload (Filadi et al. 2015). Indeed, the mitochondrial morphology is a regulator of calcium homeostasis in mitochondria and ER. The expression of a dominant negative of Drp1 increases mitochondrial calcium uptake and calcium retention capacity, and opposed effects are seen in Mfn2 knock-down cells (Kowaltowski et al. 2019).

Fig. 1.

Mitochondrial calcium handling and function. Mitochondria structure indicating outer mitochondrial membrane (OMM), intermembrane space (IE), inner mitochondrial membrane (IMM), and mitochondrial matrix (Mx). The main mitochondrial function is ATP production through ETC and ATP synthase. Mitochondrial calcium handling includes the porin VDAC, the complex channel MCU, the exchanger NCLX, and the unspecific channel mPTP. Mitochondrial calcium handling includes functional and structural interactions with endoplasmic reticulum (ER) through tethers such as Mfn2 and IP₃ receptors (IP₃R). The mitochondrial motility through the microtubule cytoskeleton is regulated by the calcium-binding protein Miro, adaptor protein Milton, and molecular motors kinesin (shown) and dynein (not shown). Cartoon made using Servier Medical Art (https://smart.servier.com/) and ©BioRender – biorender.com

Calcium overload in the mitochondria is produced by excessive cytoplasmic calcium. It triggers the formation of a non-selective pore across the mitochondrial membrane known as the mitochondrial permeability transition pore (mPTP) (Chinopoulos 2018). This channel permits the flux of molecules outside the mitochondria up to 1.5 kDa, including calcium (Chinopoulos 2018). The molecular identity of this channel is still controversial; molecular components such as adenine-nucleotide transporter (ANT) (Doczi et al. 2016) and cyclophilin D (CypD) have been described as modulators (Baines et al. 2005). Indeed, cyclosporine A (CsA), an inhibitor of CypD, suppresses the assembly of the complex (Chinopoulos 2018; Waldmeier et al. 2003). The core component of the pore is proposed to be ATP synthase (F_oF₁) (Carraro et al. 2014; Giorgio et al. 2013), and VDAC, ANT, and CypD have been proposed as partner components (Halestrap 2009). The mPTP has been reported in many pathological conditions, including neurodegenerative diseases, principally via its role in apoptosis and cell death triggers in response to external signals such as calcium overload and ROS production (Kalani et al. 2018).

Mitochondrial calcium buffering plays an important role in the central nervous system. Mitochondria are found in growth cones, axon terminals, and dendrites in neurons (Mattson 2007). Mitochondrial calcium buffering in the presynaptic compartment cushions the calcium overload to produce neurotransmitter vesicle exocytosis and produce the ATP needed for pumps in the presynaptic terminal (Billups and Forsythe 2002; Mattson 2007). The calcium influx through voltage-gated calcium channels on the presynaptic terminals, a consequence of an action potential, is, in part, buffered by mitochondria and MCU function (Billups and Forsythe 2002). On the other hand, in the postsynaptic compartment, the calcium release in dendrites is primarily from ER stores in response to synaptic activity, and mitochondria form many contacts participating in calcium buffering (Hirabayashi et al. 2017). The dendritic mitochondria are determinants in synaptic plasticity participating in post-tetanic potentiation (Tang and Zucker 1997) and the plasticity of dendritic spines (Li et al. 2004). Moreover, the mitochondrial cycles of fission and fusion are critical in synaptic plasticity (Bertholet et al. 2016). Importantly, the fission protein Drp1 is needed for dendritic spine morphogenesis (Bertholet et al. 2016), hippocampal CA1 synaptic function (Shields et al. 2015), and brain development (Ishihara et al. 2009).

The positioning of mitochondria contributes to their synaptic functions (Sheng and Cai 2012). The mitochondria move through dendrites and axons using the microtubule cytoskeleton and associated motors, kinesin, and dynein. A protein complex mainly mediates the mitochondria motility, formed by the motor molecule kinesin-1 heavy chain (Kif5), or dynein, and Miro protein (RhoT1 and RhoT2) in the outer mitochondrial membrane and Milton (TRAK1 and TRAK2), the bridge between kinesin and Miro (Schwarz 2013). In presynaptic terminals, calcium is a major regulatory signal to mitochondria localization through the Miro1 trafficking protein (Vaccaro et al. 2017). Calcium influx through the NMDA receptor or voltage-gated calcium channels recruited mitochondria in places of high synaptic activity, and the sensor of this calcium signal is the protein Miro. Miro has two calcium-binding EF-hands, and the calcium-binding to Miro permits the disassembly of the complex Kif5-Miro-Milton, stopping the movement along microtubules (Sheng 2014).

The presynaptic and postsynaptic mitochondria not only differ in function, localization, and regulation but also in structure. Presynaptic mitochondria are smaller and “darker” than postsynaptic mitochondria in electron microscopy, suggesting a more active energy production and calcium buffering than postsynaptic mitochondria. This last pool of mitochondria is bigger and less intense (Freeman et al. 2017) in the electron microscopy.

The role of mitochondria in calcium homeostasis is widely accepted in other physiological and pathological conditions (Mammucari et al. 2018). For example, it has been widely studied in skeletal muscle. In MCU KO mice, considering their mild phenotype, the organ most affected is the skeletal muscle (Pan et al. 2013). In fact, the overexpression or silencing of the MCU leads to hypertrophy or atrophy of the muscle (Mammucari et al. 2015). Moreover, mitochondrial calcium uptake is mutually related with other normal cellular pathways in the skeletal muscle (Gherardi et al. 2019) and the response of the muscle to eccentric exercise (Rattray et al. 2013). The heart is another organ in which mitochondrial calcium homeostasis plays an important role. During contraction events in the heart, mitochondria play a crucial role in regulating the amplitude of calcium rise in the cytosol (Drago et al. 2012; Wust et al. 2017), and this is regulated by the stoichiometry of the complex, as well as in other organs such as liver (Paillard et al. 2017). In addition to its physiological role in the heart, the calcium handling by the mitochondria participates in the cell-death pathways triggered by calcium overload, possibly at least partially by mitochondrial connexin-43 (Gadicherla et al. 2017). Other examples where calcium handling by the mitochondria is essential include cell metabolism (Rossi et al. 2019), neurovascular and neurometabolic coupling (Kannurpatti 2017), and cancer (Sterea and El Hiani 2020; Vultur et al. 2018). These other functions highlight the role of the mitochondrial calcium in homeostasis and pathological conditions, and the calcium increase in neurons in response to glutamate is not the exception.

Excitotoxicity: Calcium Linking NMDAR and Mitochondria

The accumulation of extracellular glutamate is toxic to neurons and is called excitotoxicity (Mehta et al. 2013). The neuronal death caused by overactivation of glutamate receptors is mediated by elevated intracellular calcium concentration. Mitochondria, on the other hand, are a critical regulatory calcium homeostasis organelle, widely located in dendrites as previously mentioned (Kasthuri et al. 2015; Wu et al. 2017). In the late 1990s, the first evidence that linked both processes emerged. In mixture cultured cells of glial and hippocampal neurons, NMDA treatments (excitotoxic stimulus) increase intracellular calcium concentration, and this rise is enhanced in the presence of mitochondrial function inhibitors (Schinder et al. 1996; Stout et al. 1998).

Moreover, prolonged NMDA treatment disrupts the mitochondrial membrane potential, an important marker of correct organelle function (Schinder et al. 1996). Measuring both intracellular calcium and mitochondrial membrane potential in the same cells, sustained increases in calcium concentration correlate with sustained mitochondrial depolarization (Vergun et al. 1999). Acute mitochondrial depolarization with carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), which dissipates proton gradients, impedes neuronal death triggers by excitotoxic glutamate treatment (Stout et al. 1998). These first experiments suggest that the calcium influx into mitochondria is important in the cell-death mechanism by overactivation of glutamate receptors. The inhibition of calcium entry by acute membrane depolarization with FCCP or mitochondrial function inhibitors explains the increase in intracellular calcium concentration and cell survival. The mitochondrial depolarization product of glutamate excitotoxicity could be avoided with cyclosporine A, an inhibitor of mPTP opening. Cyclosporine A also avoids cell death, suggesting the role of mitochondrial permeability transition pore in cell death by glutamate receptor overactivation (Nieminen et al. 1996; Schinder et al. 1996; Vergun et al. 1999).

Interestingly, not all mitochondria are damaged under NMDA treatment conditions. In fact, a proportion of mitochondria suffer reversible calcium overload by NMDA stimulation, which agrees with structural changes, swelling, and release of cytochrome c to cytosol, a known apoptogenic factor. However, there is a proportion of mitochondria that do not suffer calcium overload and lack of structural changes and cytochrome c release (Pivovarova et al. 2004). This evidence supports a mechanism of excitotoxic-induced cell death that involves mitochondrial calcium overload, mPTP opening, and apoptosis through the release of cytochrome c to the cytosol. Indeed, acute glutamate excitotoxicity in differentiated motor neurons in vitro showed transient opening of the mPTP, which is abolished blocking calcium influx into mitochondria through MCU blockage (Ru360) or silencing (shRNA). Moreover, 24 h of glutamate treatment increases the open probability of the mPTP leading to cell death (Liu et al. 2015).

In more detail, the activation of synaptic NMDA receptors through a pharmacological approach in cultured neurons increases intracellular calcium concentration, but it does not alter the mitochondrial membrane potential. However, when both synaptic and extrasynaptic NMDA receptors are activated by extracellular glutamate, the intracellular calcium concentration rises and disrupts the mitochondrial membrane potential. The activation of extrasynaptic NMDA receptors by a pharmacological approach also disrupts the mitochondrial membrane depolarization (Hardingham et al. 2002), indicating a population of NMDA receptors responsible for mitochondrial dysfunction under excitotoxic injury (Fig. 2). The use of inhibitors of GluN2B-NMDA receptor subunits enriched at extrasynaptic sites attenuates the mitochondrial depolarization, cytosolic and mitochondrial calcium overload, and causes mitochondrial structural changes in neurons in vitro (Stanika et al. 2009).

Fig. 2.

NMDA receptor activity and mitochondrial damage. Excessive extracellular glutamate activates both synaptic and extrasynaptic NMDA receptors. Synaptic NMDA receptor activity and calcium influx into the dendritic spine cooperate with correct mitochondrial function and send pro-survival signals to the nucleus. Extrasynaptic NMDA receptor activity and calcium influx into dendrites activate pro-death pathways and suppress pro-survival signals. Moreover, calcium influx is buffered by the mitochondrial overload. This phenomenon permits the opening of the mPTP and mitochondrial depolarization. Cartoon made using Servier Medical Art (https://smart.servier.com/)

Calcium transporter and channels in the mitochondria are in direct association with calcium derived from glutamate receptors. Using an in vitro approach with pharmacological agents, the calcium enters the cell by glutamate receptor flux to the mitochondrial matrix via the mitochondrial calcium uniporter and effluxes through the NCLX exchanger, indicating a mitochondrial calcium processing of calcium influx via glutamate receptors (Strokin and Reiser 2016). Under excitotoxic insult, the overexpression of MCU in hippocampal neurons potentiates mitochondrial membrane depolarization and cell death. At the same time, the silencing of MCU ameliorates the mitochondrial membrane depolarization and cell death, supporting observations that the influx of calcium to mitochondria is detrimental under excitotoxic glutamate or NMDA insults (Qiu et al. 2013). The inhibition of mitochondrial calcium uptake triggered by low calcium concentration (and not high calcium concentration) with TG-2112x was shown to protect cultured neurons against excitotoxicity without altering mitochondrial membrane potential (Angelova et al. 2019). The mitochondrial division inhibitor 1 (mdivi-1), a known inhibitor of fission protein Drp1, is also protective against excitotoxicity. However, the mechanism by which mdivi-1 protects neurons against NMDA correlates with decreasing intracellular calcium overload and calpain activation more than preventing mitochondrial fragmentation (Ruiz et al. 2018). In HT-22 neuronal cell line, the downregulation of Drp1 protects against cell death induced by glutamate excitotoxicity and ameliorates the mitochondrial damage, decreasing apoptotic markers and enhance ATP production (Zhang et al. 2014). Regarding mitochondrial dynamic machinery, the calcium-induced activation of calpain induced the cleavage of Mfn2, contributing to mitochondrial fragmentation and dysfunction (Wang et al. 2015a, b). The overexpression of Parkin was also observed to be neuroprotective against glutamate excitotoxicity in cultured retinal ganglion cells (RGC), reducing cell death and stabilizing membrane potential (Hu et al. 2017). In cortical neurons, excitotoxicity induces the localization of Parkin in the mitochondria and in ER, including MAMs, possibly regulation calcium flux between both organelles (Van Laar et al. 2015).

The synaptic activity, as previously mentioned, did not disrupt mitochondrial membrane potential (Hardingham et al. 2002), and it is even protective against glutamate excitotoxicity. The excitotoxic insult induces mitochondrial dysfunction, which is accompanied by increased oxidative stress (Rivero-Segura et al. 2019) and a consequent mitochondrial contraction. Pyruvate has been indicated as a neuroprotective factor against excitotoxicity, enhancing mitochondrial oxygen consumption (Kumagai et al. 2019), as well as dissociating DAPK1 complex from NMDAR, avoiding intracellular calcium overload (Tian et al. 2014). Synaptic activity induced by gabazine (GABA-A receptor antagonist increases the rate of action potentials) protects mitochondria from contraction and oxidation against excitotoxicity. This effect could be mediated by the synaptic induction of the transcription factor Npas4 that mediates the transcriptional repression of MCU (Depp et al. 2018), avoiding mitochondrial calcium overload.

NMDAR and Mitochondria in Chronic Pathology

Altered NMDA receptor activity and mitochondrial dysfunction have been implicated in neurodegenerative disorders (Bading 2017). Moreover, neurodegeneration has been at least partially linked to extrasynaptic NMDAR. The enhanced activity of these receptors could be due to failure in glutamate uptake, increased presynaptic glutamate release, gliotransmission, increases in number, or stability of extrasynaptic NMDARs, among others (Parsons and Raymond 2014). On the other hand, mitochondrial dysfunction has been described as a central player in neurodegenerative disorders through reactive oxygen species (ROS) production, electron transport chain (ETC) dysfunction, altered calcium homeostasis and mPTP opening, and decreased ATP production (Briston and Hicks 2018) (Fig. 3).

Fig. 3.

Mitochondrial alterations in pathology. Many neurodegenerative diseases linked to NMDAR dysfunction also display mitochondrial dysfunction. Markers of mitochondrial failure include calcium overload, mPTP opening, depolarization of mitochondrial membrane potential, and decreased ATP production. Specific mitochondrial alterations have been associated with different diseases, as indicated in the figure. Chronic disease is in red text and acute pathologies are described in blue text. Many other alterations have been found associated with each pathology, and those described are just examples. Cartoon made using Servier Medical Art (https://smart.servier.com/) and ©BioRender – biorender.com

Alzheimer’s disease (AD), the most common form of dementia in the older population, has been described as synaptic alterations as a consequence of amyloid-beta (Aβ) toxicity. Aβ impaired evocation of LTP in neurons and increased Ca²⁺ concentration through GluN2B-containing NMDA receptors. LTP deficiency could be attenuated using GluN2B selective antagonists, as well as inhibition of p38 or calpain, proteins downstream extrasynaptic NMDA receptor signalling (Hynd et al. 2004; Zhang et al. 2016). Excitotoxic consequences of Aβ are partially mediated by increasing extracellular glutamate concentration. Some explanation includes decreased glutamate uptake from the synaptic cleft, correlated with impaired function of excitatory amino acid transporter 2 (EAAT2) in perisynaptic astrocytes. The decrease in the recycling of glutamate (Wang and Reddy 2017) and astrocytic glutamate release, which activate extrasynaptic NMDA receptors (Talantova et al. 2013), are also possible mechanisms. In brain tissue and animal models, selective inhibition of GluN2B and selective enhancement of the activity of GluN2A NMDA receptors partially restored synaptic functions on Aβ-treated brain slices and animals (Huang et al. 2017), and electrophysiological recordings have indicated increased tonic currents mediated by extrasynaptic NMDA receptors (Talantova et al. 2013). The other protein hallmark of AD is tau and neurofibrillary tangles. Extrasynaptic NMDA receptor activity promotes tau overexpression in neuronal cultures, and tau ablation in mice is protective against activation of extrasynaptic NMDA receptors (Sun et al. 2016). Moreover, tau seems to be required for extrasynaptic NMDAR function, since the absence of tau practically caused extrasynaptic NMDAR-mediated currents to disappear (Pallas-Bazarra et al. 2019), supporting the protective role of tau ablation in AD models.

In mouse models of AD, APP/PSEN1 transgenic mice and in vitro models have indicated mitochondrial dysfunction through mitochondrial membrane depolarization and reduced oxygen consumption (Dixit et al. 2017; Guo et al. 2017; Wang et al. 2015a, b). Several reports have indicated the presence of Aβ into the mitochondria, bound to CypD, probably partially explaining the OXPHOS deficiency in AD patients and models (Swerdlow 2018). On the other hand, intracellular calcium dysregulation is also evident in AD (Supnet and Bezprozvanny 2010). Aβ aggregates in the plasma membrane form pores that let calcium influx into the cell. This aspect, besides NMDA receptor dysfunction, increases intracellular calcium. The ER is the major calcium store into the cell, and presenilin mutations also altered ER function, leading to release calcium to the cytosol. This calcium could explain mitochondrial dysfunction and mPTP opening, leading to neurodegeneration (Supnet and Bezprozvanny 2010). VDAC is a component in calcium homeostasis machinery in mitochondria. VDAC1-deficient mice (VDAC1+/−) showed downregulation of AD-related genes such as PSEN1, PSEN2, BACE1, tau, and AβAPP, suggesting mitochondrial calcium as a possible target in AD (Manczak et al. 2013). In neurons, Aβ and altered NMDA receptor function are linked with mitochondrial dysfunction through dyshomeostasis of mitochondrial calcium and, importantly, GluN2B-NMDA receptor calcium influx (Ferreira et al. 2015). Vast data indicate both NMDA receptor dysfunction and mitochondrial impairment in AD patients, animal models, and culture.

Huntington’s disease (HD) is a neurodegenerative disorder caused by many pathological CAG repeats in the huntingtin gene. The NMDA receptor had been described as a central component of neurodegeneration related to its enhanced activity. Hyperactivation of glutamate receptor and signalling pathways impair calcium homeostasis and result in mitochondrial dysfunction (Fan and Raymond 2007). Mouse models of HD have shown impaired LTP, probably by a lower release of glutamate from presynaptic terminals after high-frequency stimulation, and basal NMDA receptor activity is altered, as indicated by enhanced sensitivity to NMDA and enhanced NMDAR currents (Fan and Raymond 2007). HD models have revealed an enhanced extrasynaptic NMDA receptor function by an increase in the number of extrasynaptic receptors and increased extracellular glutamate concentration (Levine et al. 2010). Importantly, the enhanced extrasynaptic NMDA receptor activity occurred earlier than symptoms in mouse models (Milnerwood et al. 2010), indicating the contribution of altered NMDA receptor function in the aetiology of HD. Mitochondrial function is also impaired in HD, where alterations in ATP production and bioenergetics have been described. Other hallmarks of mitochondria dysfunction in HD include mitochondrial depolarization, altered calcium handling, mitophagy, mitochondrial dynamics, oxidative stress, defects in mitochondrial biogenesis, and enhanced mitochondrial-dependent apoptosis (Carmo et al. 2018). Decreased ATP production, indeed, has been linked to mitochondrial calcium dysregulation provoked by altered NMDA receptor function (Seong et al. 2005), as described by Zeron in cultures from transgenic mice (Zeron et al. 2004). Calcium dysregulation leads to the opening of mPTP in neurons of HD mouse models, and cyclosporine A is effective in restoring mitochondrial calcium handling and correct function (Quintanilla et al. 2017). Oxidative damage and reduced mitochondrial biogenesis are responsible for the altered functioning of peroxisome proliferator-activated receptor γ (PPAR-γ) and its co-activator PGC-1α. PGC-1α binds to mutant Huntingtin, reducing its activity and decreasing the transcription of its target genes (Johri et al. 2013).

Parkinson’s disease (PD) is the second most prevalent neurodegenerative disorder, and among its characteristics are tremors, akinesia, bradykinesia, and rigidity. It is caused by loss of dopaminergic neurons in the substantia nigra pars compacta, and the formation of protein aggregates called Lewy bodies, where α-synuclein plays a central role (Jagmag et al. 2015). In the pathophysiology of PD, the loss of dopamine in the basal ganglia produces abnormalities in glutamate transmission, including alterations in NMDA glutamate receptor in the striatum (Bagetta et al. 2010). Indeed, membrane and intracellular levels of the NMDAR subunit are altered in a rat model of 6-hydroxydopamine neurodegeneration in the substantia nigra. GluN1 levels were increased at the surface and reduced in the intracellular compartment. At the same time, GluN2B was increasing only in the surface and total fraction, indicating a redistribution of the receptor in the striatum as a result of dopamine neuron degeneration (Gan et al. 2014). NMDA receptor dysregulation has also been implicated in the dyskinesia produced by the levodopa treatment in PD patients (Ahmed et al. 2011). Indeed, NMDA receptor antagonism has shown positive results in PD models (Nouhi et al. 2018; Vanle et al. 2018), including non-human primates (Bhattacharya et al. 2018).

Mitochondrial dysfunction has been described as relevant in the pathophysiology of PD. Indeed, toxin-related parkinsonism is associated with chemicals that produce mitochondrial dysfunction, such as rotenone and MPTP. Additionally, genetic factors associated with PD are related to genes that play roles in the mitochondria. Mutations in the LRRK2 gene, which encodes a cytoplasmic kinase and is also associated with the outer mitochondrial membrane, and mutations in Parkin and PINK1, both proteins involved in mitophagy and mitochondrial quality control, have been associated with familial PD, in addition to mutations in α-synuclein and tau (Exner et al. 2012; Jagmag et al. 2015). Altogether, sporadic and familial PD share defects in the mitophagy process, dysregulation of mitochondrial dynamics, altered mitogenesis, OXPHOS impairment, oxidative stress, calcium imbalance, and defects in mitochondrial trafficking. Thus, mitochondrial dysfunction is a central player in the pathophysiology of PD (Gonzalez-Casacuberta et al. 2019). New insights in the discovery of molecular mechanisms underlying PD showed that α-synuclein protein aggregates could induce mitochondrial dysfunction through the actin cytoskeleton, mis-localizing Drp1 protein, and the interaction of protein aggregates with spectrin (Ordonez et al. 2018). In addition, Miro1 has been shown to accumulate in depolarized mitochondria of PD patients’ fibroblasts. Indeed, a small molecule able to bind to an atypical Rho GTPase domain of Miro1 was capable of restoring Miro1 defects in fibroblasts from patients and animal models (Hsieh et al. 2019). Finally, the reestablishment of mitochondrial function using a light-inducible proton-motive force pump mitigates the loss of dopaminergic neurons and motor impairment in a fly model of PD (Imai et al. 2019). All these strategies support the idea that mitochondrial dysfunction is crucial in the development of PD in human patients and animal models, and pharmacological reestablishment of mitochondrial function is a therapeutic target in development.

NMDAR and Mitochondria in Acute Pathology

Brain ischaemia is a neurodegenerative condition produced by the interruption of blood flow to the brain, which can trigger neuronal damage. The release of K⁺ and glutamate are possible initiated factors for cell death, triggering overactivation of glutamate receptors, including NMDAR. In a model of oxygen and glucose deprivation (OGD), the inhibition of the astrocytic cystine/glutamate antiporter reduced neuronal death, suggesting that glutamate release to extrasynaptic sites and activation of extrasynaptic NMDA receptors are responsible for neuronal death (Soria et al. 2014). In a cellular model of OGD, extrasynaptic NMDA receptor signalling was upregulated and induced the expression of Clca1, a putative calcium-activated chloride channel, part of the pro-death programme trigger by extrasynaptic NMDA receptors (Wahl et al. 2009). Mitochondrial dysfunction in ischaemic injury occurs as the product of reperfusion when intracellular calcium overload happened. As a consequence, mitochondrial calcium is also dysregulated, leading to mPTP opening and permeabilization of the inner mitochondrial membrane. Oxidative stress products of excessive ROS production by impairment of ETC and OXPHOS lead to decreased ATP production and mitochondrial membrane depolarization (Bakthavachalam and Shanmugam 2017; Liu et al. 2009). Under ischaemic damage, melatonin and rapamycin have been used to reduce mitochondrial dysfunction and cell death, mechanisms which include activation of SIRT1 and mitophagy (Li et al. 2014; Yang et al. 2015), suggesting the role of mitochondrial dysfunction in neurodegeneration. Interestingly, the overexpression of Opa1, a mitochondrial fusion machinery protein and regulator of cristae structure, protects the brain against ischaemic damage, indicating the role of cristae structure in mitochondrial dysfunction (Varanita et al. 2015).

Traumatic brain injury (TBI) is brain damage as a result of external forces such as direct hits, acceleration, or penetrating objects; among its leading causes are falls, vehicle accidents, and contact sports (Blennow et al. 2016; Coronado et al. 2015; Maas et al. 2008). It shows variable symptoms, including headache, vomiting, memory loss, or even loss of consciousness (Blennow et al. 2016). TBI has been indicated as a risk factor in the development of neurodegenerative diseases such as Alzheimer's and dementia (Gardner et al. 2014; Gupta and Sen 2016). TBI induces the increase in extracellular glutamate levels (Chamoun et al. 2010; Hinzman et al. 2012), as well as alterations in synaptic plasticity such as LTP in the hippocampus (Schwarzbach et al. 2006), cognitive impairment in spatial learning tasks (Acosta et al. 2013; Lloyd et al. 2008), oxidative stress (Ohta et al. 2013; Wu et al. 2006), and neuroinflammation (Acosta et al. 2013; Lloyd et al. 2008). All this evidence suggests that TBI may cause alterations in NMDAR function and signalling. However, it is still unknown if there is an imbalance between the activity of synaptic and extrasynaptic NMDA receptors during TBI. Our results indicate this mechanism (data not published).

Mitochondrial impairment has been observed in animal models of TBI. Hippocampal and cortex mitochondria show OXPHOS deficiency early in the time course after the trauma, and this bioenergetic failure occurred in TBI models of different severities (Gilmer et al. 2009; Hubbard et al. 2019). Moreover, the modulation of mitochondrial function and homeostasis using pharmacological tools has shown an improvement in cognitive performance and a decrease in cell death. Such is the case of the mitochondrial antioxidant MitoQ, which could develop its effect through the modulation of the transcription factor Nrf2 (Zhou et al. 2018), the inhibitor of mPTP, NIM811 (Readnower et al. 2011), and the inhibition of the mitochondrial calcium uniporter (Zhang et al. 2019). Evidence indicates that synaptic mitochondria in particular are more susceptible to damage than non-synaptic mitochondria after TBI, expressing dysfunction of the ETC and oxidative damage in biomolecules (Hill et al. 2018; Kulbe et al. 2017). Interestingly, the more damaged synaptic mitochondria are protected when animals were treated with cyclosporine A (Kulbe et al. 2017), indicating the role of mPTP in mitochondrial dysfunction after TBI.

Many other neurodegenerative diseases or conditions such as amyotrophic lateral sclerosis (ALS) (Carri et al. 2017; Cozzolino and Carri 2012; Paul and de Belleroche 2014; Spalloni et al. 2013) and alcohol consumption (Bonet-Ponce et al. 2015; Carpenter-Hyland et al. 2004; Pian et al. 2010; Tapia-Rojas et al. 2018, 2019) share NMDA receptor-mediated cell damage and mitochondrial dysfunction, with calcium as the central mediator between both processes. However, in vitro approaches are the favourite route to demonstrate a functional relationship between NMDAR and mitochondrial dysfunction involving calcium homeostasis. Brain slices or organotypical approaches are less used, probably due to technical challenges that they represent in time and space resolution. Although some reports presented here demonstrate that calcium influx through NMDA receptors, and probably predominantly extrasynaptic NMDA receptors, is the primary source of calcium that triggers mitochondrial dysfunction in cultured neurons, it has not been demonstrated in pathological tissues, suggesting that both mechanisms are in a close relationship only using indirect methodologies.

Conclusions

Glutamate neurotransmission is essential in synaptic plasticity and cell death through the dual role of NMDA receptors. These receptors are found in both synaptic and extrasynaptic membranes, and they are permeable to calcium, which triggers different signalling pathways depending on their location, driving the cell fate to survive or die. Mitochondria are organelles widely present in dendrites that participate in calcium homeostasis. Excitotoxic insults and activation of extrasynaptic NMDA receptors depolarize the mitochondrial membrane and open the permeability transition pore, which is dependent on the calcium influx into mitochondria. In vitro neurons have established the basis of this cell-death mechanism, and it is proposed to participate in many neurodegenerative disorders in which NMDA receptor and mitochondrial dysfunction are observed, such as AD, HD, PD, ischaemic-reperfusion damage, or TBI. A few reports have indicated that activation of extrasynaptic NMDA receptors, or both synaptic and extrasynaptic NMDA receptors, is coupled to mitochondrial dysfunction through calcium influx into mitochondria and triggers the mPTP in brain slices or pathological tissue; however, it remains to be seen whether this relationship known in vitro is a possible mechanism in vivo. Further investigation and new experimental approaches or designs are needed to explore this interaction in a complex tissue environment.

Author contributions

RGM and WC conceived the project. RGM performed the literature review and wrote the manuscript. WC supervised the work and edited the final version.

Funding

This work has been supported by grants Fondo Nacional de Desarrollo Científico y Tecnológico (FONDECYT) 1190620, Center for Excellence in Science and Technology (AFB 170005, PFB 12/2007), and Sociedad Química y Minera de Chile (SQM) for special grant “The role of lithium in human health and disease” to WC and pre-doctoral fellowship to RGM.

Compliance with Ethical Standards

Conflict of interest

The authors declare that they have no competing interests.