Simultaneous Single Neuron Recording of O₂ Consumption, [Ca²⁺]_i and Mitochondrial Membrane Potential in Glutamate Toxicity

Abstract

To order the cellular processes in glutamate toxicity, we simultaneously recorded O₂ consumption, cytosolic Ca²⁺ concentration ([Ca²⁺]_i) and mitochondrial membrane potential (mΔψ) in single cortical neurons. O₂ consumption was measured using an amperometric self-referencing platinum electrode adjacent to neurons in which [Ca²⁺]_i and mΔψ were monitored with Fluo-4 and TMRE⁺, respectively using a spinning disk laser confocal microscope. Excitotoxic doses of glutamate caused an elevation of [Ca²⁺]_i followed seconds afterwards by an increase in O₂ consumption which reached a maximum level within 1 to 5 min. A modest increase in mΔψ occurred during this time period, and then, shortly before maximal O₂ consumption was reached, the mΔψ, as indicated by TMRE⁺ fluorescence, dissipated. Maximal O₂ consumption lasted up to 5 min and then declined together with mΔψ and ATP levels, while [Ca²⁺]_i further increased. mΔψ and [Ca²⁺]_i returned to baseline levels when neurons were treated with an N-methyl-D-aspartate receptor antagonist shortly after the [Ca²⁺]_i increased. Our unprecedented spatial and time resolution revealed that this sequence of events is identical in all neurons, albeit with considerable variability in magnitude and kinetics of changes in O₂ consumption, [Ca²⁺]_i and mΔψ. The data obtained using this new method are consistent with a model where Ca²⁺ influx causes ATP depletion, despite maximal mitochondrial respiration, minutes after glutamate receptor activation.

Introduction

Neurons are electrically excitable cells which produce relatively large amounts of ATP (and consume correspondingly large amounts of O₂) to drive the ion pumps that restore intracellular Na⁺ and Ca²⁺ levels after synaptic activity and action potentials (Mattson and Liu, 2002). Glutamate, the major excitatory neurotransmitter in mammals, is responsible for fast synaptic transmission and also plays fundamental roles in long-lasting changes in synaptic strength in processes such as learning and memory (Thiagarajan et al., 2007). Excessive activation of glutamate receptors can result in irreversible disruption of ion homeostasis and cell death, a process called excitotoxicity (Mattson, 2003). Many of the physiological effects of glutamate on neurons are mediated by Ca²⁺ influx through plasma membrane ligand-gated channels (primarily NMDA receptors) and voltage-dependent channels, and excessive Ca²⁺ influx is believed to play a major role in excitotoxicity. However, the exact mechanisms during the process of excitotoxicity leading to cell death and their relationships to intracellular Ca²⁺ concentrations and mitochondrial respiration in individual neurons are still a matter of debate. To further clarify the sequence of cellular processes that occur, we present a method for the simultaneous recording of O₂ consumption, cytosolic calcium concentration ([Ca²⁺]_i) and mitochondrial membrane potential (mΔψ) in single neurons. To measure single neuron O₂ consumption we used a self referencing method (Land et al., 1999; Smith et al., 2007); a small surface (2-3 μM) Whalen-style platinum electrode (Whalen et al., 1967) that measures the local O₂ gradient generated by a neuron via O₂ consumption. For simultaneous imaging of [Ca²⁺]_i and mΔψ we use a spinning disk confocal microscope that minimizes laser exposure via a reduction in scanning time. Using this approach we were able to clearly dissect the order of events in a single neuron with high temporal and spatial resolution, demonstrating the advantages of this method over other methods such as using isolated mitochondria, a cell respirometer (Jekabsons and Nicholls, 2004) or a non-self referencing small surface electrode, which has been used for large Purkinje neurons, albeit without simultaneous imaging (Hayakawa et al., 2005).

Methods

Cortical neuron cell cultures and experimental treatments

Cortical tissue was removed from the brains of embryonic day 18 rat pups (Sprague-Dawley) and digested with 0.05% trypsin for 20 min at room temperature followed by addition of soybean trypsin inhibitor (0.52mg/ml) and trituration through a sterile pasteur pipet. Neurons were plated at a density of 50,000/cm² in 35 mm glas bottom culture dishes (Mattek, Ashland, MA) containing culture medium consisting of Neurobasal medium (Invitrogen, Carlsbad, CA) supplemented with 73.5 μM glutamax and B27 (Invitrogen). Experiments were performed in neurons that had been maintained in culture for 14 days. For exposures to glutamate and FCCP the cells were washed in artificial cerebrospinal fluid (ACSF) (in mM: NaCl 120, KCl 3.1, KH₂PO₄ 0.4, Tris pH 7.4 20, NaHCO₃ 5, NaSO₄ 1.2, CaCl₂ 1.3, glucose 15) and then exposed to the indicated concentrations of glutamate in ACSF for the indicated time periods. For viability studies cells were washed again in ACSF and returned to conditioned culture medium plus 10μM MK801 (Tocris, Ellisville, MO) to block any ongoing activation of NMDA receptors. For oxygen measurements, neurons were plated into 35mm poly-d-lysine-coated coverslip-mounted plastic dishes (Mattek, Ashland, MA).

Cell viability

For viability studies cells were grown in 48 well plates. Before and 24 h after glutamate exposure, cells were counted in the same visual field of one well by hand through a phase-contrast microscope using a 20x objective. The percentage of surviving neurons was calculated for all conditions. Two visual fields per one well were chosen and 4 to 6 wells per condition were counted. Experiments were done in triplicate. The same experiments were performed using a resazurin reduction assay, in which 10 μM resazurin was added to the culture medium 24h after exposure to glutamate for 1h and subsequent resazurin reduction was measured in a plate reader with 560/590nm ex/em. Resazurin is the active compound of the commercial Alamarblue assay. Resazurin reduction assays showed similar results as counting neurons, except for a slightly higher relative viability values after glutamate treatment (data not shown). This is because resazurin is not neuron specific and also reduced by non neuronal cells.

Determination of ATP levels

Neurons in 48 well plates were washed once in ACSF and incubated with 100 μL ACSF plus glutamate for the indicated time periods. At the end of the experiment 100 μl of the luciferase/luciferin/detergent solution of the CellTiter Glow kit (Promega, Madison, WI) was added to each well. After a 10 min incubation period 50 μl were transferred into a well of a black 96 well plate. For ATP to ADP ratio measurements the kit from Biovision (Mountain View, CA) was used. ATP and ADP measurements were performed simultaneously in duplicate wells of 48 well plates. Due to a different luciferase preparation, the absolute luminescence values in the Biovision kit were about an order of magnitude lower than in the Promega kit. Luminescence was determined as relative luminescent units using a HTS 7000 plus plate reader (Perkin Elmer). Background levels were determined by analysis of ACSF from cell-free wells, and were subtracted from the raw data. Experiments were done in tripiclate.

Determination of O₂ consumption with a self-referencing O₂ electrode

All self-referencing instrumentation, amplifiers as well as manipulators, are products of the BioCurrents Research Center (MBL, Woods Hole, MA). O₂ consumption was recorded according to the self referencing method (Land et al., 1999; Smith et al., 2007). Briefly, a 2-3 μm oxygen microsensor containing a recessed-platinum electrode was prepared as in Jung et al. (1999). This amperometric probe was polarized with -0.6V, at which voltage the reduction of oxygen at the probe tip was diffusion-limiting, and therefore [O₂] in the solution could be assessed. The circuit was completed by a reference electrode (3M KCl / 3% agar) and both were immersed into the 35 mm glass bottom dish. With the help of a micromanipulator and three-axis stepper motor-driven control the platinum electrode was placed 5 μm perpendicular to the soma of a typical large pyramidal neuron. Smaller neurons with an interneuron morphology were not used for this study. Software (IonView; BioCurrents Research Center, MBL) was used to translate the electrode over a distance of 10 μm at 0.3Hz (40 μm/s), with the electrode remaining in the near position for 1s, and staying in the far position for 1 s. A current (pA) was recorded in both the near and the far position and the IonView software calculated the differential current (fA) in real-time. As the O₂ concentration near the neuron (O₂ consumer or O₂ sink) is smaller than in the far position, the delta current between near and far pole reflects the difference in O₂ concentration and thus measures the O₂ gradient surrounding the neuron. Each electrode was tested against with small, circular (2-3 μm diameter) O₂ sources, before being used in experiments with neurons. As a general rule, more sensitive electrodes would correspondingly generate more electrical noise and less sensitive electrodes would give more stable signals with time. Electrode motion and continuous recording of the delta current (fA) were controlled by the IonView software and data were exported into Excel. Electrode drift was controlled for by manually tracing the tip of the electrode on the monitor (ImageSuite; Perkin Elmer). Recordings were performed at 37°C. Temperature control was achieved by covering the complete electrophysiology work station with isolative wool and aluminum plates. Inside the workstation two heaters were maintained at 37°C using a proportional integrative derivative (PID) controller. At the time of recording the workstation was completely covered by aluminum plates, reducing airflow and aberrant temperature gradients. Glutamate and other agents were prepared as 11x stock solutions and added into the 35 mm glass bottom dish with the help of a syringe pump (Braintree Scientific Inc.) and two-way solenoid valve controlled by customized software (BioCurrents Research Center, MBL). This prevented erroneous temperature fluctuations at the microscope stage.

ΔI [fA] was converted to flux J [pmol*cm^-2*s^-1] using the following equations. J=-D*ΔC/Δr, where D is the diffusion coefficient of O₂ in H₂O (0.0000268 cm²*s^-1 at 37°C), ΔC is the concentration difference and Δr is the distance of electrode translation (20 μM) (Jung et al., 1999). ΔC was calculated from ΔI by the following equation ΔC=[O₂]*I_tot^-1*ΔI. [O₂] is the concentration of O₂ in air saturated water at 37°C (6.7 mg/l or 0.42 mmol/l) ΔI is the difference in current measured between far and near point and I_tot is the total current in the system (which is orders of magnitude higher than ΔI). To calculate O₂ consumption of a single neuron the following equation was used Conc[O₂]=J*SA, where J is flux and SA is the surface area (in cm²). To calculate a neurons surface area it was assumed that the neuron is a half spheroid and the typical dimensions of a typical cortical neuron (30 μm length, 12μm width and 7μm) depth were determined by the micromanipulator SA= 4π(30+12+7)/3+x+Δr/2)²/2 (Jung et al., 2001), where x is the distance the electrode tip is placed from the neuron (5μm) and Δr is the distance of electrode translation (20 μm).

In this system air saturated aqueous solutions were used, which are hyperoxic compared to in vivo tissue, a feature common to all in vitro meaurements of O₂ consumption. Since the bath solution is static and not perfused (but in contact with room air) a small decreases in total O₂ concentration took place during phases of maximal O₂ consumption, but not under baseline conditions, as equilibration with room air was sufficient to compensate for neuronal O₂ consumption. This decrease was monitored through the total current that is recorded by the small surface platinum electrode. The total decline in current was usually between 1 - 10% over a time period of minutes. This change in O₂ concentration / current is controlled for by the self referencing mechanism, which is working at a 0.3 Hz rate.

Calcium imaging and determination of mΔψ

For calcium imaging, neurons on 35 mm glass bottom dishes were loaded with 1 μM fluo-4 AM (Invitrogen) for 20 min at 37°C in Neurobasal and then washed with ACSF followed by ACSF plus 5 nM tetramethylrhodamine ethyl ester (TMRE⁺, Invitrogen) for 20 min at 37°C. Cells were then placed on the stage of a Zeiss Axiovert microscope custom-equipped with a Perkin Elmer spinning pinhole disk and a laser for confocal microscopy. The spinning disk was placed under the microscope outside the electrophysiology cage. Fluo-4 fluorescence was monitored with a 488 nm laser and TMRE⁺ fluorescence with a 568 nm laser. Images were acquired every 5s and analyzed with Image-Suite software (Perkin Elmer). The use of a spinning disk confocal microscope system reduced exposure time to laser light and thus likely reduced free-radical formation by dye bleaching. After an initial recording period of ~3 min, to check signal stability, glutamate or other agents were applied as described above. Fluorescence intensity data were thus adjusted for background.

Simultaneous recording of plasma membrane potential and mΔψ

Cortical neurons on 35 mm glass bottom dishes were loaded with 1 μM of the plasma membrane potential (pΔψ) sensitive dye bis-(1,3-dibutylbarbituric acid)trimethine oxonol (DiBAC₄, Invitrogen)and 10 nM TMRE⁺ for 20 min in ACSF at 37°C. Increased DiBAC₄ fluorescence indicates depolarized pΔψ, whereas decreased TMRE⁺ indicates depolarized mΔψ. Cells were then placed on a confocal microscope and fluorescence at 488 nm and 568 nm laser excitation was acquired every 5s.

Immunoblots

Cell lysates were prepared from neuron cultures 24h after glutamate exposure or 24h after treatment with 1 μM staurosporine (Sigma) was started. Standard immunoblot procedures were performed. A 1:1000 dilution of mouse anti α-spectrin (= fodrin) antibody (Chemicon, Temecula, CA) was used.

Genomic DNA agarose gel electrophoresis

Genomic DNA was prepared from cultured neurons 24h after glutamate exposure or 24h after treatment with 1 μM staurosporine (Sigma, St. Louis) was started. The Qiaquick gel extraction kit (Qiagen, Hilden, Germany) was used. 100 μl of lysis buffer were used for ~10⁶ neurons on a 60 mm dish to lyse cells (three 60 mm dishes per condition). DNA extraction was performed according to the manufacturer’s instructions. DNA was separated on a 1.5% agarose gel and stained with ethidium bromide according to standard procedures.

Data Analysis

To assess sustained maximal O₂ consumption after addition of glutamate or FCCP, a time period of 1min or longer was selected to reflect a steady plateau phase of O₂ consumption. In addition, the single highest value of flux J obtained after addition of FCCP or glutamate was selected for peak value comparisons. Conversely after oligomycin and rotenone addition the reduction of respiration was assessed for a time period of 1min or longer during a steady state of reduced respiration. Significance was calculated by two-tailed t-test for pair comparisons and ANOVA/Tukey for multiple comparisons

Results

Glutamate dose dependently induces NMDA receptor-mediated necrotic death of cortical neurons

All experiments were performed in primary embryonic rat cortical cells that had been maintained in culture for 14 days. The neurons in these cultures possess elaborate dendritic and axonal arbors and abundant glutamatergic synapses (Yao et al., 2003). Before beginning analyses of O₂ consumption, [Ca²⁺]_i and mΔψ, we characterized the mode of cell death and performed a dose response and time response assay in our model of glutamate exctitotoxicity. Viability was assessed by counting neurons before and 24 h after treatment in the same visual field; 15 min of exposures to 33 and 100 μM glutamate caused ~75% of neurons to die, whereas no significant cell death was caused by 1 or 5 μM glutamate (Fig. 1a). Maximal death of neurons occurred with a 15 min exposure to 33 μM glutamate, longer incubations did not cause additional cell death. Treatment of neurons with the NMDA receptor antagonist MK-801 completely prevented glutamate-induced cell death, while the AMPA receptor antagonist CNQX did not protect the neurons (Fig. 1a). To characterize the mode of excitotoxic neuronal death, we exposed neurons to 33 μM glutamate or 1 μM staurosporine (an agent known to induce apoptosis) and then genomic DNA was isolated and separated on an 1.5% agarose gel. Staurosporine caused a typical DNA ladder, indicative of apoptosis, whereas glutamate caused DNA degradation with high molecular weight fragments, indicative of necrosis (Fig. 1b). The excitotoxic death of neurons involved Ca²⁺-mediated processes because immunoblots of cell lysates showed that glutamate caused accumulation of the 145 kD breakdown product of α-spectrin (fodrin), indicative of calpain activation. In contrast, staurosporine caused accumulation of the 115 kD α-spectrin breakdown product, indicative of caspase 3 activation (Fig. 1c).

Figure 1. Glutamate induces necrosis in a dose- and time-dependent manner by a mechanism involving NMDA receptor activation.

Viability was tested after glutamate exposure (a). The left panel depicts concentration dependent toxicity of a 15 min glutamate exposure. The mid panel shows time dependent toxicity of 33 μM glutamate and the right panel shows the effects of AMPA receptor antagonist CNQX (10 μM) and NMDA receptor MK 801 (10 μM). Viability was assessed by counting neurons in the same visual field before and 24 h after glutamate exposure and is expressed as percentage of surviving neurons. Genomic DNA from control, glutamate treated and staurosporine treated cortical neurons was separated on a 1.5% agarose gel to check for the presence of an apoptotic “ladder”(b). Immunoblot of α spectrin (fodrin) (at 240 kD) and its (145 kD and 115 kD) breakdown products 24 hr after glutamate or staurosporine or in control conditions. (c). (ANOVA/tukey; ns = non significant; *** p<0.001; n=4).

Cellular ATP levels and the ATP/ADP ratio drop rapidly in cortical neurons in response to glutamate

After exposure to 33 μM glutamate for 1 min, total ATP levels dropped to 78% of control levels and further declined to 47% within 15 min (Fig. 2a). A longer exposure for 30 min caused no further decline. The glutamate-induced decline in ATP levels was concentration-dependent with no significant decrease with 1 μM glutamate, and decreases to 57% and 47% of control levels with 10 μM and 33 μM glutamate, respectively (Fig. 2b). Higher glutamate concentrations of 50 and 100 μM caused no further decline of ATP levels during a 15 min exposure. A 15 min exposure to 33 μM glutamate caused a significant increase in ADP levels to 191% of control levels (Fig. 2c) and as a result the ATP/ADP ratio dropped from 2.92 to 0.67, a reduction to 23% of the control ATP/ADP value (Fig. 2d).

Figure 2. Glutamate causes ATP concentration and ATP/ADP ratio to decrease.

The effect of a 15 min exposure of 1, 10, 33, 50 and 100 μM glutamate on ATP concentration tested (a). Time dependent decrease in ATP concentration after 1, 5, 10, 15 and 30 min exposure to 33 μM glutamate (b). Changes in ATP and ADP concentration(c) and ATP/ADP ratio (d) after a 15 min exposure to 33 μM glutamate.

O₂ consumption in single and multiple cortical neurons under baseline conditions and after blockade of mitochondrial electron transport and ATP synthesis

O₂ consumption in cortical neurons was assessed using a platinum electrode that was placed 5 μm from a single cortical neuron or a cluster of neurons (Fig. 3a,b,c), and then moved rapidly back and forth over a 20 μm distance between two points located 5 and 25 μm away from the cell soma. When the electrode is polarized (-0.6V), the difference in current (ΔI [fA]) between the near and the far position of the platinum electrode represents the gradient of the O₂ concentration around the neuron, a reflection of O₂ consumption. Thus an increase in ΔI reflects an increase in O₂ consumption and a decrease in ΔI reflects a decrease in O₂ consumption. We then converted ΔI into flux J [pmolO₂*cm^-2*s^-1]. Initial measurements were recorded from small clusters of neurons, as these current recordings had a higher signal-to-noise ratio than single neurons. Also, recordings from clusters displayed less variability in terms of responsiveness to reagents. In an initial series of experiments, after a stable oxygen flux was recorded from a cluster of neurons, oligomycin was applied to block mitochondrial ATP synthesis. O₂ consumption dropped to 21.8 +/- 2.1% (n=6) of the basal level (Fig. 3c); further addition of rotenone (to block complex I of the electron transport chain (ETC)) reduced O₂ consumption to 11.8% +/- 2.1% (n=6; Fig. 3c). This result indicated that an average of 88.2% of the O₂ consumption of cortical neurons is required to drive the ETC, but only 78.2% is required to generate ATP. The difference of 10% is likely used for mitochondrial uncoupling and calcium buffering (Porter et al., 1999; Murchinson et al., 2004). The remaining 11.8 % of neuronal oxygen consumption is therefore not ETC related and may include other oxygen sinks such as peroxisomes or additional oxygen consuming reactions (e.g. detoxification in mitochondria).

Figure 3. The self referencing O₂ electrode measures O₂ consumption adequately in single neurons and neuron clusters.

A layout shows the principle of the self referencing electrode. The reference electrode (blue) is in the same dish with the O₂ electrode (green) which is moving between the near and the far position, measuring current at both extremes. The system is set up on top of a spinning disk confocal microscope (a). Transmitted light microscope pictures (b) show a self referencing electrode next to a single cortical neuron (left) and a cluster of cortical neurons (right). The neuron to the left is also stained with TMRE⁺ to depict mitochondria. The electrode is shown as an overlay of both the near and the far position. Scalebar is 20 μM. O₂ consumption in neuron clusters (c) and single neurons (d), after addition of Oligomycin and rotenone (indicated by red bars). The steady phase after oligomycin and rotenone addition is indicated by an orange bar and the average flux value for this period is noted underneath. A slower decline of O₂ consumption in neuron cluster likely reflects slower diffusion of oligomycin to the center of a neuron cluster. Bckgr denotes the background signal recorded 400 μm above the neuron layer, where no oxygen gradient is to be expected. Black bar denotes the time scale. Average reduction in O₂ consumption after oligomycin and rotenone treatment in neuron clusters and single neurons. (** p< 0.01, t-test). Two way ANOVA finds a significant effect of drug treatment (p<0.0001) but not of cluster or single neuron recording (p=0.25).

We next measured the O₂ consumption of individual neurons (Fig. 3b,d,e). The decline of O₂ consumption in single neurons after addition of oligomycin was faster than in neurons clusters, likely reflecting the faster diffusion of oligomycin into an entire neuron as compared to an entire neuron cluster. Individual neurons usually had a flux between 1 to 5 pmol O₂*cm^-2*s^-1, which appeared to be related to their size (larger neurons exhibited higher O₂ consumption rates). The average O₂ consumption rate was 12 ± 2.1 fmol O₂*min^-1 per neuron (sem, n=8). O₂ consumption dropped to 20.1% ± 3.0% (n=5) of the baseline level after addition of oligomycin, and further declined to 11.2% ± 2.1% (n=5) after rotenone addition (Fig. 3e). The data indicated that results obtained in recordings from small neuron clusters and single neurons are similar. Previous data published using two O₂ electrodes to record from large populations (>10⁶) of cerebellar granule cells showed that 89% of the basal O₂ consumption is used for the ETC and 61% for ATP synthesis (Jekabsons et al., 2004). The latter results are similar to those described here, and validate our method. However, although total mitochondrial O₂ consumption is almost identical, cerebellar granule neurons use less O₂ for ATP synthesis, possibly indicating higher amounts of glycolysis and mitochondrial Ca²⁺ buffering in these neurons (Budd and Nicholls, 1996).

Activation of glutamate receptors increases neuronal O₂ consumption to near maximal levels

In order to assess the maximal respiratory capacity of neurons, O₂ consumption was measured during exposure to the uncoupling agent carbonyl cyanide p-trifluoromethoxy-phenylhydrazone (FCCP), that was added at increasing concentrations (Fig. 4a,b). The optimal dose of FCCP to induce maximal O₂ consumption varied considerably among individual neurons, ranging from 100 nM to 1 μM. Once maximal O₂ consumption was induced it could only be sustained for a limited time (usually around 5 min or less for both clusters and single neurons) and would then progressively decline to or near zero during the next 5-30 min (Fig. 4a,b), during which time the appearance of the neurons under phase-contrast optics appeared normal (data not shown). Additional increases of FCCP concentration (above that required to induce maximal O₂ consumption) shortened the time of sustained maximal O₂ consumption and accelerated the subsequent decline.

Figure 4. FCCP and glutamate cause maximal O₂ consumption in cortical neurons.

Traces of recordings of O₂ consumption in neuron clusters (a, c, e) and single neurons (b, d, f) after addition of FCCP (a + b), glutamate (c +d) and FCCP after glutamate (e + f). Addition of drug and concentration are indicated by brown (FCCP) and red (glutamate) bars. Time scale is indicated by black bars, bckg denotes background signal measured 400 μm above the neuron layer. Quantification of O₂ consumption after fccp and glutamate treatment (g,h,i). The maximum increase in O₂ consumption was measured as sustained average and single peak value. Statistic analysis revealed no significant difference in the O₂ consumption increase induced by FCCP or glutamate in neuron clusters (left 2 graphs) or single neurons (right 2 graphs).

Stepwise increases in glutamate concentrations consistently caused incremental increases in O₂ consumption (Fig. 4c,d). Concentrations causing a maximal response ranged from 10 μM to 32 μM for neuron clusters and 7 μM to 32 μM in single neurons. Sustained O₂ consumption lasted up to 10 min in clusters and 1-5 min in single neurons. O₂ consumption typically declined faster in single neurons than in neuron clusters, suggesting that measurements from neuronal populations overestimate the duration of O₂ consumption responses to glutamate of individual neurons within the population. Further increases in glutamate concentration caused no further increase in O₂ consumption. On average, glutamate increased O₂ consumption in neuron clusters to 213 ± 35% (n=4) of control (Fig. 4g). Thus, the magnitude of the increase of O₂ consumption induced with glutamate was similar to that induced with FCCP.

In order to establish that glutamate can indeed induce maximum O₂ consumption in neurons, FCCP was added to neurons shortly after glutamate. No further increase in O₂ consumption was induced by subsequent additions of FCCP (Fig. 4e,f). In this series of experiments glutamate alone increased O₂ consumption to 250 ± 22% over baseline (n=3) and subsequent addition of FCCP had no additional effect (247 ± 47% of the basal level; n=3; Fig. 4h). If compared with results from cerebellar granule neurons published previously (Jekabsons and Nicholls, 2004; Johnson-Cadwell et al., 2007; Yadava and Nicholls, 2007), our data indicate that cortical neurons require a lower dose of glutamate to reach a maximal respiration level (33 μM in cortical neurons versus 100 μM in cerebellar neurons), consistent with the higher sensitivity to glutamate toxicity of cortical neurons. The maximal respiration of cerebellar granule neurons has been published to be 250 to 275% higher than baseline when incubated in artificial cerebrospinal fluid (Johnson-Cadwell et al., 2007; Yadava and Nicholls, 2007), which is slightly higher than data shown here in cortical neurons, and may contribute to the higher resistance to excitotoxicity of cerebellar granule neurons.

Simultaneous recordings of O₂ consumption, [Ca²⁺]_i and mΔψ in single neurons exposed to glutamate

In order to record O₂ consumption together with [Ca]_i and mΔψ, the self-referencing system was mounted on a spinning disk confocal microscope (Fig. 5a). The main advantage of a spinning disk confocal microscope is that the image acquisition time is reduced relative to a conventional laser scanning confocal microscope, therefore exposing neurons to minimal laser excitation. This advantage was of critical importance for our studies, as can be seen in Figure 5a, where laser excitation resulted in noise artifacts on the O₂ flux recording when exciting the mΔψ indicator tetramethylrhodamine ester (TMRE⁺). TMRE⁺ (unlike the [Ca²⁺]_i indicator fluo-4 which was removed from the medium after cell loading) was maintained at a fixed concentration in the culture medium for the duration of the experiment and, due to its positive charge, adhered to the (negatively charged) O₂ electrode. Laser excitation thus resulted in artifacts in the current measured by the electrode. In addition, by binding TMRE⁺, the electrode exhibited increasing fluorescence over time (excitation at 568 nm, Fig. 5b). This resulted in an unexpected advantage in that we were able to visualize the position of the electrode even when no visible light source was present. As a trade-off, self-referencing of the O₂ electrode resulted in additional image artifacts in the form of a saw tooth line in the TMRE⁺ recordings, not seen in the fluo-4 imaging (Fig. 5c). However, in the system we used, these artifacts were modest and the higher signal-to-noise ratio in subsequent O₂ recordings allowed reliable simultaneous recordings of O₂ consumption, [Ca²⁺]_i and mΔψ in single neurons.

Figure 5. Simultaneous recording of O₂ consumption, [Ca]_i and mΔψ in single neurons.

O₂ recording artefacts caused by laser light (a). TMRE⁺ accumulation on the negatively charged platinum electrode (b). No accumulation of Fluo-4 on the platinum electrode (c). O₂ consumption after addition of 100 nM FCCP (d). O₂ consumption after exposure to glutamate (e,f) with slightly different courses of events and a comparatively more resistant neuron in (e) and a more sensitive neurons in (f). Black arrows denote the start of glutamate addition with concentration indicated on top (in uM). O₂ consumption scale is on the left, fluorescence scale on the right, timescale is indicated by black bars. Red traces represent TMRE⁺ signal, green fluo-4 and blue O₂ consumption traces. Glutamate addition would occasionally cause artefacts in TMRE⁺ (18, 24, 32 uM) and O₂ traces (24 μM) (d).

A typical time course of events after addition of FCCP is shown in Figure 5d. FCCP initially caused an increase in O₂ consumption; interestingly, mΔψ remained stable initially, most likely because of reverse activity of the mitochondrial ATP synthase. Once O₂ consumption reached maximal levels, it rapidly declined together with mΔψ. The increase in [Ca²⁺]_i commenced shortly after the onset of the increase in O₂ consumption, but before maximal respiration was reached. With the subsequent decrease in O₂ consumption, [Ca²⁺]_i continued to rise progressively.

A typical sequence of events after glutamate addition is seen in Figures 5e and f where a subtoxic concentration of glutamate produced a small reversible [Ca²⁺]_i peak (typically two-fold over baseline), followed by a small reversible increase in O₂ consumption which returned to near baseline levels within 5 min (Fig. 5e). A higher toxic concentration of glutamate produced a large increase of [Ca²⁺]_i (approximately 5-fold over baseline) that was immediately followed by a large increase in O₂ consumption which remained at this maximal level for 1-5 min and then gradually declined towards zero (Fig. 5e,f). The time it took for neurons to undergo complete loss of respiration varied, but all neurons reached this point within 30 min. During the time period of maximal O₂ consumption, [Ca²⁺]_i occasionally remained constant, but generally increased at a reduced rate. With excitotoxic concentrations of glutamate, during the decline of O₂ consumption [Ca²⁺]_i increased progressively during the 30 min recording period (Fig. 5d). Toxic doses varied among neurons, some neurons reached maximal O₂ consumption with glutamate concentrations as low as 7 μM, while others required higher concentrations. A concentration of 33 μM was toxic for all neurons tested. TMRE⁺ fluorescence typically increased transiently after the addition of glutamate and then shortly before maximal O₂ consumption TMRE⁺ fluorescence peaked and then declined (Fig. 5e,f). Subsequently, O₂ consumption declined. At the end of a typical 30 min recording session a neuron would have zero O₂ consumption, with a concomitant seven-fold increase in fluo-4 fluoresence and a five-fold reduction in TMRE⁺ fluorescence, indicating a metabolically inactive neuron. Importantly, these experiments show that a neuron can have a declining TMRE⁺ fluorescence while sustaining maximal respiration. These representative examples also demonstrate that, using these novel, integrated technologies we can dissect subtle differences in the time courses of changes of O₂ consumption, [Ca²⁺]_i and mΔψ in response to glutamate receptor activation versus mitochondrial uncoupling in single neurons. Glutamate receptor activation results in a rise in [Ca²⁺]_i, with a subsequent rise in O₂ consumption, while FCCP causes a reverse trend with O₂ consumption increased first, followed by an increase in [Ca²⁺]_i.

Simultaneous recording of pΔψ and mΔψ

Since TMRE⁺ fluorescence is influenced by pΔψ and mΔψ (Nicholls, 2006; Ward et al., 2007), we recorded both potentials simultaneous, to check for possible confounding effects. pΔψ Was monitored using the plasma membrane sensitive dye DiBAC₄ and mΔψ was monitored with TMRE⁺ (Suppl. Fig. 2). After addition of glutamate an intial ~10% increase in both DiBAC₄ and TMRE⁺ fluorescence was observed, indicating a slight increase in mΔψ and a slight decrease in pΔψ. This phase was followed by a simultaneous decline in both pΔψ and mΔψ, indicating a depolarizing plasma membrane and mitochondrial membrane. A similar sequence of events was observed by Nicholls (2006) in cerebellar granule neurons exposed to ionomycin.

Delayed NMDA receptor inhibition modifies glutamate-induced changes of O₂ consumption, [Ca²⁺]_i and mΔψ in single neurons

Because NMDA receptors are mainly responsible for the excitotoxic actions of glutamate in the cortical neuron cultures of the present study, we determined the effects of delayed treatment of neurons with the selective NMDA receptor antagonist MK801 (10 μM) on glutamate-induced changes of [Ca²⁺]_i, O₂ consumption and mΔψ in individual neurons. Rather than adding MK801 at fixed time points after glutamate addition, we added MK801 at metabolically-relevant time points, namely, either after O₂ consumption and [Ca²⁺]_i had started to increase but were not at their peak, or shortly after the initial increase in [Ca²⁺]_i but before any increase in O₂ consumption. Representative examples are shown in Figure 6a and b. When MK801 was added shortly before maximal O₂ consumption had been attained, it had little or no effect on the sequence of events described before (Fig. 6a). However, when MK801 was added before any increase in O₂ consumption had occurred, MK801 was able to attenuate the increases in O₂ consumption and [Ca²⁺]_i, resulting in only a small transient increase of [Ca²⁺]_i and little or no increase in O₂ consumption (Fig. 6b).

Figure 6. Effect of MK801 on O₂ consumption after application of glutamate.

Examples of two recordings. Scales and color codes are identical to Figure 5. Simultaneous recordings of O₂ consumption, [Ca]_i and mΔψ in single neurons after glutamate exposure and subsequent addition of MK 801, either after [Ca]_i and O₂ consumption had risen (a) or before any increase in [Ca]_i and O₂ consumption was detected.

Discussion

The present technology for simultaneous recording of O₂ consumption, [Ca²⁺]_i and mΔψ allowed us to elucidate relationships between mitochondrial metabolic activity and Ca²⁺ dynamics in individual live neurons. Our data indicate that ATP depletion is critical for the underlying pathophysiology of glutamate toxicity in cortical neurons. Thus after glutamate receptor activation [Ca²⁺]_i levels increase first followed very shortly afterwards by O₂ consumption. O₂ consumption in neurons reaches maximal levels within 1 to 5 minutes. This is an indicator of ATP depletion in itself because O₂ consumption will only reach maximal levels in ATP depletion settings. In addition, within 1 min after glutamate addition, ATP levels start to drop and level off to a minimum within 15 min, despite mitochondria respiring at maximal capacity during this time period. Although ATP molecules are still in the cell, the ATP/ADP ratio dropped to a level that will render the neuron inoperable. So far no critical threshold for a neuronal ATP/ADP ratio is known and only by measuring O₂ consumption, mΔψ and ATP/ADP levels a conclusion on the neuronal energy status can be made. Interestingly, at the onset of glutamate receptor activation, TMRE⁺ fluorescence is actually increasing together with increased O₂ consumption, suggesting that initially increased respiration leads to increased proton transport across the inner mitochondrial membrane and increased mΔψ. Notably, during the phase of increased TMRE⁺ fluorescence pΔψ, as indicated by DiBAC₄ fluorescence, declines only very moderately. We believe that this indicates that in the initial phase after glutamate application increased O₂ consumption and ATP production help the neuron to stabilize pΔψ by pumping Ca²⁺ and Na⁺ out of the cytosol. A similar initial increase in mΔψ and simultaneous small decrease in pΔψ was observed by Nicholls (2006) after exposure of cerebellar granule neurons to 1 μM ionomycin. However, TMRE⁺ fluorescence starts to decline shortly afterwards, despite maximal respiration. Together with the decline of TMRE⁺ fluorescence we observed a simultaneous rapid decline in pΔψ, as indicated by DiBAC₄ fluorescence, which is again in line with the data of Nicholls (2006). One possible explanation is that neurons are running out of ATP and therefore neither pΔψ nor mΔψ can be maintained and decline simultaneously. Indeed, Computer simulations used by Nicholls (2006) cand Ward et al. (2007) that calculate pΔψ and mΔψ in order to best fit the curves of TMRE⁺ and DiBAC4 fluorescence in an acute excitotoxic setting both indicated a simultaneous decline of pΔψ and mΔψ. However it must be clearly stated that this is not the only possible explanation. Alternatively it is possible that pΔψ starts to decline first and the very first decrease in TMRE⁺ fluorescence is an artifact of the declining pΔψ, while mΔψ will start to decline later at a time point that can not be exactly determined, due to the fact that TMRE⁺ fluorescence depends both on pΔψ and mΔψ (Nicholls 2006; Ward et al., 2007). Therefore more work is needed to conclusively establish the exact timing of mΔψ decline. However, it is still of interest to show for the first time that TRME⁺ fluorescence starts to decline while O₂ consumption is maximal, as the reverse case (i.e. TMRE+ fluorescence declines after O₂ consumption declines) would indicate that mitochondrial dysfunction is the initial step that leads to mΔψ decline. This important aspect of glutamate toxicity that only be detected by a simultaneous single neuron examination of O₂ consumption and mΔψ, demonstrating an advantage of the presented method over other approaches, such as studying isolated mitochondria or using a cell respirometer (Jekabsons and Nicholls, 2004; Johnson-Cadwell et al., 2007) in which only a small subset of neurons (~100) is imaged out of a much larger population from which O₂ consumption is measured.

The focus of this paper is on neuronal O₂ consumption, ATP concentration and ATP/ADP ratio during glutamate toxicity. Nevertheless, we assume that AMP levels will increase after exposure to glutamate. This may lead to an activation of AMP activated kinase (AMPK), an important enzyme for the regulation of cell metabolism (Misra 2008). But at the same time AMPK will be in short supply of one of its substrates, i.e. ATP and at the moment it is not clear what impact AMPK activation has in acute glutamate toxicity.

Our data reveal previously unknown heterogeneity among neurons in both the magnitudes and time courses of changes of O₂ consumption, [Ca²⁺]_i and mΔψ, even though we examined only neurons with a similar pyramidal-like morphology. Possible factors important for differential responses to glutamate among cortical neurons may be the differential expression of glutamate receptors (Mattson et al., 1991; Pickard et al., 2000), Ca²⁺-binding proteins (Fiumellie et al., 2000), and proteins that modify mitochondrial energy metabolism (Boero et al, 2003). The baseline O₂ consumption of a single cortical neuron was calculated to be 12 (± 2.1) fmol O₂*min^-1. Values recorded from cerebellar granule neurons grown on coverslips range between 0.8 and 1.2 fmol O₂*min^-1 (Jekabsons and Nicholls, 2004). Cortical neurons are expected to have a higher consumption since they are larger. However, values recorded from large populations on coverslips may underestimate the actual O₂ consumption rates (Jekabsons and Nicholls, 2004).

Nicholls et al. who have done the most extensive work on O₂ consumption and glutamate toxicity found different responses after glutamate, depending on preparation batch of neurons and experimental protocol (Jekabsons and Nicholls, 2004; Budd and Nicholls, 1996; Yadava and Nicholls, 2004). In their most extensive study of glutamate toxicity and O₂ consumption Jekabsons and Nicholls (2004) found a submaximal increase in O₂ consumption in response to glutamate and concluded that ATP depletion did not take place, and that mitochondrial function was preserved for up to 3 hr in cerebellar granule neurons after exposure to 250 μM glutamate plus glycine in Mg²⁺ free ACSF with 25 mM KCl. It therefore appeared that mitochondrial failure was not an early event in the cell death cascade. Later on Yadava and Nicholls¹¹ reported that neurons did, in fact, maximally increase O₂ consumption after glutamate exposure (100 μM glutamate plus glycine in Mg²⁺ free ACSF with 5 mM KCl) and ATP depletion was thus concluded to be of importance. They explained their discrepant findings by the fact that different concentrations of KCl in the ACSF were used in the two studies. This explanation is counterintuitive, as Jekabsons and Nicholls (2006) reported that 25 mM KCl actually increases O₂ consumption for up to 3 hr. Also this explanation would imply that after glutamate exposure mitochondrial function could be preserved by raising the KCl concentration. Moreover, using a low (3.5 mM) KCl concentration in cerebellar granule neurons will start potassium deprivation-induced apoptosis, making it more difficult to interpret results. Finally, it should be mentioned that the same group reported that cerebellar granule neurons benefit from depolarizing mitochondria shortly before glutamate treatment by adding rotenone and oligomycin, concluding that mitochondrial accumulation of Ca²⁺ and precipitation of CaPO₄ crystals are key events in glutamate toxicity (Budd and Nicholls, 1996). These results were explained by using neuron preparations with high glycolytic potential. In our hands cortical neurons always increased O₂ consumption maximally upon glutamate exposure, independent of neuron batch or culture conditions. We used ACSF with 3.5 mM KCl as this is the buffer of choice for cortical neurons. In earlier experiments we saw no reduction of cytosolic Ca²⁺ concentration when 25 mM KCl was used (M.G., unpublished data).

We have also demonstrated that the period in which glutamate toxicity can be reversed using MK801 is very short. In order to prevent mitochondrial failure, MK801 must be added before O₂ consumption increases; adding MK801 shortly before O₂ consumption peaks will fail to blunt the toxic effects of glutamate. A likely explanation for this is that the initial plasma membrane depolarization and Na⁺ and Ca²⁺ influx results in excessive ATP demand for the plasma membrane Na⁺/K⁺- and Ca²⁺-ATPases, such that the amount of ATP required for restoration of the ion gradients exceeds the amount that can be generated with mitochondrial respiration. In such a depolarized neuron with an ATP deficit, blockade of NMDA receptors will fail to prevent the continued increase of [Ca²⁺]_i because influx of Ca²⁺ through voltage-gated channels exceeds the amount of Ca²⁺ that is extruded.

The ability to measure O₂ consumption, [Ca²⁺]_i and mΔψ simultaneously in single cells within a population of interacting cells provides the opportunity to understand the relationships of mitochondrial function and Ca²⁺ at a level of resolution not previously possible. We have shown how this technique provides valuable insight to the mechanism of glutamate toxicity in neurons, pointing to the pivotal role of cellular ATP depletion. Apart from unraveling glutamate toxicity, we believe this technique will be valuable for establishing how various intercellular signaling mechanisms and subcellular signaling pathways affect mitochondrial respiration, not only in neurons, but also in other cells in which energy metabolism and Ca²⁺ signaling are of particular importance to cell physiology and disease, including cardiac myocytes and pancreatic β-cells (Civelek etal., 1997; Cortassa et al., 2003). Our method will also allow the measurement of O₂ consumption in cases where cell numbers are limited, such as preparations of neurons from adult animal brain or human biopsies (Brewer 1997).

Supplementary Material

1

Supplemental Figure 1. O₂ consumption after single dose application of glutamate. O₂ consumption was recorded in single neurons (a,b) and neuron clusters after a single dose application glutamate that was high enough to elicit a maximal response. Blue traces represent O₂ consumption (expressed as flux J), black bar indicates time scale, red bar indicates duration of treatment, the dose of glutamate is indicated above the red bar. Bckg denotes background signal, recorded 400 μm above the neuron layer.

Supplemental Figure 2. Simultaneous recording of pΔψ and mΔψ after glutamate exposure. pΔψ was recorded with the membrane potential sensitive fluorescent dye (DiBAC₄ (3)) and mΔψ was recorded with TMRE⁺. Addition of 20 μM glutamate is indicated by black arrows, time scale indicated by black bars. The green trace represents DiBAC₄ fluorescence, indicative of pΔψ (increased fluorescence represents decreased membrane potential) and the red trace indicates TMRE⁺ fluorescence, indicative of mΔψ. Each graph represents one cortical neuron soma.

Abbreviations

ACSF

artificial cerebrospinal fluid

ATP

adenosine triphosphate

[Ca²⁺]_i

intracellular calcium concentration

ETC

electron transport chain

FCCP

carbonyl cyanide p-trifluoromethoxy-phenylhydrazone

mΔψ

mitochondrial membrane potential

NMDA

N-methyl D-aspartate

TMRE+

tetramethylrhodamine ester