Cortical spreading depression (SD) is a propagating wave of neuronal and glial depolarization that manifests in several brain disorders. However, the relative contribution of neurons and astrocytes to SD genesis has remained controversial. This is in part due to a lack of utilizing sophisticated experimental methodologies simultaneously to quantify multiple cellular parameters. To address this, we used simultaneous two-photon imaging, intrinsic optical imaging, and electrophysiological recordings to ascertain the changes in cellular processes that are fundamental to both cell types including cell volume, pH, and metabolism during SD propagation. We found that SD was correlated in neurons with robust yet transient increased volume, intracellular acidification, and mitochondrial depolarization. Our data indicated that a propagating large conductance during SD generated neuronal depolarization, which led to both calcium influx triggering metabolic changes and H+ entry. Notably, astrocytes did not exhibit changes in cell volume, pH, or mitochondrial membrane potentials associated with SD, but they did show alterations induced by changing external [K+]. This suggests that astrocytes are not the primary contributor to SD propagation but are instead activated passively by extracellular potassium accumulation. These data support the hypothesis that neurons are the crucial cell type contributing to the pathophysiological responses of SD.
Spreading depression (SD) in cerebral cortex occurs in several neurological disorders including migraine (Pietrobon and Striessnig 2003; Sanchez-del-Rio and Reuter 2004), brain trauma (Fabricius et al. 2006), and stroke (Strong and Dardis 2005). SD is a slowly propagating wave of neuronal and glial depolarization that spreads out at 30–50 μm/s. The depolarization occurs at the wave front of SD, which causes a large field potential shift and profound ion redistributions (Leao 1944; Somjen 2001).
It has been hotly debated whether neurons or astrocytes are the primary contributor to SD generation and propagation (for review, see Somjen 2001). Some reports indicate that neuronal depolarization is best timed with the SD wave front (Muller and Somjen 2000), whereas others support the contrary view, that astrocyte depolarization shows better temporal coincidence (Sugaya et al. 1975). Imaging calcium in both cell types during SD has helped this deliberation but has not resolved it. Previous work found that the propagation rate of SD was similar to astrocyte calcium waves (Nedergaard et al. 1995), suggesting that this astrocyte signal might drive SD. However, neuronal calcium waves evoked during SD were found to precede the change in astrocytic calcium, and diminishing astrocyte calcium waves did not block SD (Basarsky et al. 1998; Chuquet et al. 2007). These latter studies suggest the astrocyte may be a mere passive responder to the SD wave and lend credence to the idea that neurons are the principal culprit. However, calcium signals alone cannot adequately place this model on solid footing, especially because calcium entry into cells seems to have little role in the mechanisms underlying SD (Basarsky et al. 1998). When combined with the contradictory electrophysiological evidence of the time course of SD in each cell type, it is clear that other cellular parameters must be examined to see if this discrepancy between neurons and astrocytes still holds.
Changes in cell volume, pH homeostasis, and oxidative metabolism occur rapidly, and cumulating evidence suggests that these cellular changes are important to the pathophysiology of SD by initiating damage processes. First, the interstitial space shrinks dramatically following SD (Jing et al. 1994), suggesting the occurrence of cytotoxic cerebral edema caused by cell swelling, which is a possible cause for brain injury in trauma and ischemic stroke (Kimelberg 1995). Second, cortical extracellular pH (pHo) shows a rapid alkaline transient at SD onset, followed by prolonged acidosis (Menna et al. 2000). Disturbance of pH homeostasis has been postulated to contribute to cell death and genesis of infarction following cerebral ischemia (Nedergaard et al. 1991; Lipton 1999). Third, SD induces mitochondrial membrane depolarization (Bahar et al. 2000), which plays an important role in apoptotic pathways (Orrenius et al. 2003). However, it is still unknown what the relative contributions are of neurons and astrocytes to these pathophysiological changes during SD.
In the present study, we use simultaneous electrophysiological and two-photon imaging approaches to examine cell swelling, pH changes, and mitochondrial membrane potentials in individual neurons and astrocytes during the onset and progression of SD. Changes in these pathophysiological responses 1) provide a tool for examining the temporal profile of cellular activation of both neurons and astrocytes, 2) help establish which cell type is the main contributor to SD propagation as well as SD pathophysiology, and 3) offer needed insight into the cellular mechanisms underlying SD-related neural disorders. We demonstrate that SD evokes a sequence of cellular responses unique to neurons in cortical slices that are not reflected in astrocytes. SD induces a remarkably rapid and reversible cell volume increase, a transient intracellular pH (pHi) decrease, and mitochondrial membrane potential changes exclusively in neurons. In contrast, astrocytes did not exhibit any of these responses during SD, suggesting that they play a passive role by only responding to neuronal activity.
Materials and Methods
Neocortical Slice Preparation and Induction of SD
Slices from parietal cortex were prepared from 20- to 30-day-old thy1–yellow fluorescent protein–positive (YFP+) mice (Feng et al. 2000; C57BL/J and CBA F1 hybrids) and from 14- to 21-day-old Sprague Dawley rats, according to standard procedures. Our experiments were approved by the Canadian Council for Animal Care and the University of British Columbia Animal Care Committee. All experiments were conducted in strict accordance with National Institutes of Health Guide for the Care and Use of Laboratory Animals. Briefly, the animals were anesthetized deeply with halothane and decapitated. The brain was removed quickly, and 400-μm coronal cortical slices (from Bregma 0 to Bregma −3 mm of rat brain) were cut with a vibratome (VT100; Leica, Willowdale, Ontario, Canada) in chilled (0–4 °C) slicing solution containing the following (in mM): 230 sucrose, 26 NaHCO3, 10 D-glucose, 2.5 KCl, 1.25 NaH2PO4, 0.5 CaCl2, and 10 MgSO4, pH 7.3. Then the slices were transferred to a storage chamber with fresh artificial cerebrospinal fluid (ACSF) containing the following (in mM): 126 NaCl, 2.5 KCl, 2.0 MgCl2, 2.0 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 D-glucose, pH 7.3, and were incubated at room temperature for >1 h before recording. All solutions were saturated with 95% O2/5% CO2. Brain slices from YFP+ mice were used in experiments measuring neuronal volume. In all the other experiments, brain slices from rats were utilized.
Individual slices were transferred to a recording chamber with slice supports (Warner Instruments, Hamden, CT) that permit solution flow both above and below the slice and perfused rapidly with oxygenated ACSF (3 ml/min) at 30–32 °C. SD was induced by perfusing ACSF containing 40 mM KCl (high K+) for 80–90 s. The high K+ solution was made by an equimolar replacement of 40 mM KCl for NaCl. The high K+ solution reached 50% of maximal concentration in ∼24 s and reached 80% in ∼55 s in the bath. The high K+ solution contains the following (in mM): 88.5 NaCl, 40 KCl, 2.0 MgCl2, 2.0 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 D-glucose, pH 7.3. SD could be reliably induced in 2 min from the perfusion of high K+ ACSF. There was a low probability of inducing SD if the perfusion speed was less than 0.6 ml/min, even when applied for 2–3 min. At this speed, the [K+] in the bath reached 50% of maximum in ∼50 s and 80% in ∼120 s. The slower perfusion rate probably inactivates certain ion channels due to the slower rate of depolarization as external K+ slowly increases. Tetrodotoxin (TTX; 1 μM) was bath applied to block sodium currents in all the experiments. This allowed us to more selectively examine the cellular changes induced by SD itself and not from alterations in action potential firing that are triggered secondary to SD.
Two-Photon Imaging and Analysis
We performed imaging with a two-photon laser scanning microscope (Zeiss LSM510-Axioskop-2 fitted with a 40X-W/0.80 numerical aperture objective lens) directly coupled to a Mira Ti:sapphire laser (∼100-fs pulses, 76 MHz, pumped by a 5 W Verdi laser; Coherent, Santa Clara, CA). We imaged both neurons and astrocytes at depths >50 μm from the slice surface. Fluorescent neurons were selected from Layers IV and V of the upper somatosensory cortex (S1HL and S1FL). In this region, pyramidal neurons were identified by their shape and the presence of a single, large apical dendrite extending vertically toward the pial surface and clearly distinct from the small circular interneurons. Astrocytes were loaded with sulforhodamine 101 (SR101; Molecular Probes, Eugene, OR) by bath application (25 μM for 20 min) followed by washing for at least 20 min. To load astrocytes with 2',7'-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein, acetoxymethyl ester (BCECF/AM; Molecular Probes), slices were incubated with 20 μM of BCECF/AM (in 0.3% dimethyl sulfoxide) for 1 h. To load astrocytes with Rhodamine 123 (Rh123; Molecular Probes), slices were incubated with 10 μM Rh123 for 5 min. Neurons were labeled with YFP or by injection of BCECF (free acid; Molecular Probes), Rh123, or Fluor® 594 hydrazide (Molecular Probes) from the recording electrode. YFP was excited at 890 nm and was detected with external detectors with a 535-nm (30 nm bandpass) filter. The BCECF and Rh123 fluorophores were excited at 840 nm and were detected with detectors with a 535-nm (30 nm bandpass) filter (Supplementary Fig. 2A). The SR101 fluorophore was excited at 840 or 890 nm and was detected with external detectors with a 630-nm (60 nm bandpass) filter. The Alexa 594 fluorophore was excited at 840 nm and was detected with detectors with a 630-nm (60 nm bandpass) filter. Changes of intrinsic optical signals (IOSs) were imaged by acquiring the two-photon near infrared laser light (840–890 nm) in the transmitted optical path with an external photomultiplier (PMT) simultaneously with acquisition of epifluorescence signals via a separate set of PMTs.
Fluorescence signals were defined as ΔF/F = [(F1 − B1) − (F0 − B0)]/(F0 − B0), where F1 and F0 represent fluorescence within the cell cytoplasm at any given time point and at the beginning of the experiment, respectively, and B1 and B0 are the background fluorescence at the same time point and at the beginning of the experiment, respectively. Background values were taken from an adjacent area of the imaged cell. To control for changes in BCECF or Rh123 fluorescence induced by changes in cell volume, we loaded the cells with another inert dye with different emission spectra (Alexa 594 or SR101) to ratio and normalize BCECF or Rh123 fluorescence intensity. The fluorescent intensity of the point of interest was selected from cell soma not including the nucleus region. Normalized fluorescence signals were defined as ΔF/R = (ΔF/F)/(ΔFinert/Finert) where ΔF/F is fluorescence from BCECF or Rh123 fluorophore, and ΔFinert/Finert is fluorescence from the inert dye Alexa 594 or SR101. A representative fluorescence change of normalized BCECF intensity is shown in Supplementary Figure 2B.
Normalized IOSs were defined as ΔT = (T1 − T0)/T0, where T1 and T0 are transmitted light intensity of a small region in the field of view at a certain time point and at the beginning, respectively. The rate of change for IOS was determined by the first derivative of ΔT (dΔT/dt, where ΔT is the change in IOS and t is time). We developed criteria using the maximal rate of rise of the IOS change to quantitatively differentiate between the fast component of the IOS, which we show to be associated with SD, and the slow components of the IOS response that are caused by the elevated [K+]o (Fig. 1B). In 10 randomly selected experiments, the maximal rate of rise of the slow component was <2.0% ΔT/s (mean 0.95 ± 0.10% ΔT/s, n = 10). Therefore, if the maximum rate of rise exceeded 2.0% ΔT/s, it was regarded as the fast component and thus indicated the occurrence of SD (arrows in Fig. 2B). The time when the rate obviously exceeded the baseline was taken as SD onset. The validity of the threshold criterion was confirmed as the peak rate of rise after SD onset (7.47 ± 0.18% ΔT/s, n = 9) was significantly larger than the peak rate of rise when SD was not induced (0.86 ± 0.09% ΔT/s, n = 10, P < 0.001, Fig. 1C).
For imaging cell volume changes, three-dimensional time-lapse images were taken during the perfusion of high K+ ACSF. At each time point, a two-dimensional projection image was made of the three-dimensional image that was composed of a Z-stack of 5 or 6 frames obtained at 2-μm increments. All the imaging experiments were recorded with Zeiss LSM 3.2, and images were exported into a series of time-lapse images. The image sequences were recentered, and the volume changes were calculated as the changes of cross-section area with the “analysis particles” function of ImageJ 1.32. When a specific threshold value was set to the whole image sequence, the analysis particles function automatically, outlined the cell shape, and calculate the cross-section area at all the time points.
Whole-cell voltage-clamp recordings from cortical neurons were obtained at 30–32 °C. Patch electrodes (5–7 MΩ) were pulled from 1.5 mm outer diameter thin-walled glass capillaries (150F-4; World Precision Instruments, Sarasota, FL) in three stages on a Flaming-Brown micropipette puller (model P-97; Sutter Instruments, Novato, CA) and were filled with intracellular solution containing the following: 135 mM K-gluconate, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 1.1 mM ethyleneglycol-bis(2-aminoethylether)-N,N,N',N'-tetra acetic acid, 0.1 mM CaCl2, 4 mM Mg-ATP, 0.5 mM Na-GTP, pH 7.2; 20 μM of Ru360 (Merck, Darmstadt, Germany) or 10 μM of cyclosporin A (CsA; Sigma, St Louis, MO) was included in the intracellular solution for the corresponding experiment. For imaging experiments, 100 μM of BCECF (free acid), 10 μM of Rh123, or 25 μM of Alexa 594 was also included in the intracellular solution. Normally, the intracellular fluorescence became stable within 10 min of establishing whole-cell recordings. Membrane potentials were voltage clamped at –70 mV in neurons and at –80 mV in astrocytes. Membrane potential measurements were obtained with respect to the bath ground not the extracellular voltage in the brains slice so that there may be a small voltage measurement error during the 3–5 mV external DC shifts of SD. DC-coupled field potentials were recorded with glass micropipettes filled with ACSF (resistance, 1–3 MΩ). Membrane currents and field potentials were monitored with MultiClamp 700B amplifier (Molecular Devices, Union City, CA), acquired via a Digidata 1320 series analog-to-digital interface onto a Pentium computer with Clampex 9.0 software (Molecular Devices). Data were sampled at 10 kHz, and most were low-pass filtered (four-pole Bessel) at 1 kHz.
In all cases, Student's t tests were used for statistical comparisons, with P < 0.01 considered significant. Values are reported as the mean ± standard error of the mean.
Imaging Cortical SD with IOS
IOSs, which are generated by changes in tissue light scattering or transmittance (MacVicar and Hochman 1991; Andrew and MacVicar 1994), have been used to monitor SD propagation in brain tissue (Basarsky et al. 1998; Anderson and Andrew 2002). We observed that a wave of SD in cortical slices was reliably induced by perfusing ACSF containing 40 mM of KCl with a high speed (3 ml/min) (Anderson and Andrew 2002) and caused dramatic alterations to the IOS (Fig. 1A and Supplementary Movie 1). High perfusion rates alone, without elevated K+, did not induce SD. The IOS changes that occurred when high K+ was perfused and successfully induced SD consisted of two components: an initial transient peak in light transmittance that appeared on top of a much slower IOS increase (Fig. 1B). Even though these two IOS components were easily discernable, plotting the rate of the IOS change made their difference even more dramatic. The rate of change in the initial component displayed a rapid increase and a sharp peak, followed by a rapid decrease and overshoot past baseline (peak rate of rise: 7.47 ± 0.18% ΔT/s, n = 9, Fig. 1B). This method of analysis also revealed that this IOS component propagated through the tissue with the SD wave at 26.0 ± 4.0 μm/s (n = 9) (Fig. 1D–F). This speed is approximately 50% of the SD propagation rate reported without TTX and is most likely due to a TTX-mediated reduction in the total SD-inward current (Muller and Somjen 1998, 2000). In contrast, the rate of change in the second slower IOS component was much slower (rate of rise: 0.35 ± 0.06% ΔT/s, n = 10, P < 0.001, Fig. 1B) showing no distinct peak and no apparent propagation through the tissue. This indicated that the slow secondary IOS change was not directly associated with SD and was instead due to a general slice swelling as a result of the elevated [K+]o per se (Anderson and Andrew 2002). To confirm this idea, we perfused cortical slices with high K+ at a lower perfusion speed (<0.6 ml/min), so that SD was not evoked. As expected, instead of the biphasic response, the IOS showed a gradual, nonpropagating increase (rate of rise: 0.27 ± 0.04% ΔT/s, n = 10, Fig. 1C), similar to the slow, secondary IOS component observed during SD (P = 0.24). The timing differences between the initial SD-relevant and the secondary SD-irrelevant IOS allowed us to quantitatively differentiate between them (see Materials and Methods) and use the initial rapid IOS component as an indicator of the occurrence of SD for subsequent optical experiments.
The IOS at Single-Cell Resolution Was Temporally Correlated with Electrophysiological Signals during SD
We then examined how the IOS components correlated with the electrophysiological signals generated by SD (Czeh et al. 1993) to determine how the IOS could be used as a marker of the time course of SD. In order to ascertain which electrophysiological phase of SD was correlated with the rapid IOS component, we simultaneously imaged IOS signals and recorded either extracellular field potentials or whole-cell membrane currents (Fig. 2). In response to an SD wave, a field potential shift, measured with an extracellular electrode placed adjacent to a neuron soma (<10 μm), occurred temporally coincident with propagation of the rapid IOS component (center of Fig. 2A; Fig. 2B). Whole-cell voltage-clamp recordings from individual neurons showed that typically a slowly developing inward current was first induced reflecting the membrane depolarization due to high K+ perfusion (amplitude 14.5 ± 3.4 pA/pF, n = 8) before an SD wave was triggered. However, at the initiation of SD, a rapid onset, large amplitude inward current (28.3 ± 5.5 pA/pF, n = 8) was observed that peaked at the same time as the rapid IOS increase (Fig. 2C). This current, which we termed ISD, caused the depolarization observed in SD and has been shown previously to be coincident with the extracellular potential shift (Czeh et al. 1993). ISD was not observed in the absence of SD (as indicated by the IOS). Only the initial inward current from the high K+ perfusion itself continued to slowly develop (amplitude 17.8 ± 3.4 pA/pF, n = 6, Fig. 2D). Whole-cell recording from astrocytes showed a much smaller ISD with an amplitude of 9.2 ± 3.9 pA/pF (n = 4). The occurrence of ISD in astrocytes was also coincident with the IOS change (Supplementary Fig. 1). These data demonstrated that the rapid IOS change reliably denotes the onset of SD as measured by either the extracellular potential shift or the depolarizing inward current from individual cells.
Neuronal and Astrocytic Volume Changes Evoked by SD
We next imaged neuronal and astrocytic morphological changes to determine which type of cell contributes to the interstitial space shrinkage during SD. Measurements of cell volume were acquired simultaneously with either the IOS or electrophysiological indicators to verify the occurrence of SD and to determine how volume changes were temporally related to SD onset. We first examined whether neurons underwent swelling by imaging single YFP-positive cortical neurons in mouse slices and measuring their maximal cross-sectional area from a Z-stack of images as described in the Materials and Methods. When high K+ was perfused at a lower speed so that SD was not induced, there was no obvious neuronal volume change (0.8 ± 0.5%, n = 10, Fig. 3B, top), consistent with a previous report showing that neuron volume is resistant to osmolarity changes (Andrew et al. 2007). Interestingly, at SD onset, a rapid increase in neuronal volume was initiated, reaching a maximum (11.0 ± 0.9%, P < 0.001) in 72 ± 3 s (n = 10), followed by gradual recovery to baseline in 5–7 min (Fig. 3A,B bottom and Supplementary Movie 2). To determine whether astrocytes exhibit similar volume changes, we imaged astrocytes stained with SR101 (Nimmerjahn et al. 2004). In contrast to neurons that were resistant to slow increases in [K+]o, astrocytes gradually swelled (35.6 ± 3.4%, n = 10) (Kimelberg 2000) when high [K+]o was perfused and gradually recovered their volume during washout (Fig. 3C, top). The change in astrocyte volume coincided with the slow phase of the IOS change, indicating a standard swelling response from the elevated [K+]o. When SD was induced, astrocytes did not show any further swelling (40.2 ± 1.7%, n = 10, P = 0.62 compared with slow perfusion, Fig. 3C, bottom, and Fig. 3E). These results indicate that astrocytes only display passive swelling in response to high K+, whereas neurons exhibit a concerted and SD-specific volume increase that is distinct from standard osmotic swelling. These data also suggest that the pathophysiological reduction of interstitial volume observed at SD onset (Jing et al. 1994) is of neuronal, rather than astrocytic origin.
Neuronal versus Astrocytic Intracellular pH Changes during SD
To gain insight into the dysregulation of pH homeostasis during SD (Menna et al. 2000), we used the pH indicator BCECF to examine the pHi dynamics in neurons and astrocytes. Neurons were loaded with cell-impermeable BCECF through the patch pipette. A cell labeled with BCECF showed quite uniform staining in the cytosol and the nucleus. To control for changes in BCECF fluorescence induced by changes in cell volume, another inert dye, Alexa 594, was also injected intracellularly and was used to ratio and thus normalize BCECF fluorescence intensity (Fig. 4A). When SD was not evoked, high K+ induced an increase in BCECF fluorescence in neurons (7.8 ± 1.5% ΔF/R, n = 7, Fig. 4B), suggesting a slight alkaline shift. In contrast, when SD was induced, the BCECF fluorescence showed a transient decrease that was temporally coincident with the rapid IOS change and ISD (−21.7 ± 1.9% ΔF/R, n = 7, Fig. 4C), indicating a pHi decrease at the onset of SD. To examine pHi changes in astrocytes, these cells were loaded with the cell-permeable BCECF/AM and SR101. The latter provided a control for volume-induced changes by normalizing BCECF fluorescence to that of SR101 (Fig. 4D). Either BCECF injected through the patch pipette or bulk loading of BCECF/AM resulted in an even staining in the cell soma without showing specific incorporation in the nucleus. Without the induction of SD, high K+ per se caused a gradual increase in astrocytic BCECF fluorescence (19.7 ± 3.6% ΔF/R, n = 5, Fig. 4E), indicating an intracellular alkalization that was similar but relatively greater than that observed in neurons. However, in direct contrast to neurons, astrocytes did not show a transient decrease of BCECF fluorescence at the onset of SD but instead showed an increase similar to that observed in response to high K+ itself (27.1 ± 3.8% ΔF/R, n = 8, Fig. 4F). To test the possibility that the astrocytic pH response to SD was concealed by the pHi increase to high K+ alone, we locally injected a small volume of KCl (3 M) in a distant and discrete region of the cortex to evoke an SD wave that would then propagate into the imaged region (Supplementary Fig. 3). Under these conditions, astrocytes displayed an immediate increase in BCECF fluorescence, indicating a rapid alkalization of pHi at SD onset. This effect is likely due to the rapid increase in the [K+]o caused by SD (Brinley et al. 1960) which is known to raise pHi in astrocytes (Chesler and Kraig 1989). These data demonstrate that neurons, but not astrocytes, display a distinct acidification during SD. This transient acidosis experienced by neurons may be associated with the interstitial pH changes at SD onset reported by others (Menna et al. 2000).
Mitochondrial Membrane Potential Changes during SD
As SD induces mitochondrial membrane depolarization (Bahar et al. 2000), which is thought to play an important role in apoptosis (Orrenius et al. 2003), we used the mitochondrial membrane potential (Δψm) indicator Rh123 to examine Δψm dynamics in brain slices (Kovacs et al. 2005) during SD. To observe the Δψm in neurons, Rh123 was included in the whole-cell patch pipette. Because Rh123 is membrane permeable, Rh123 also stained cells in the surrounding neuropil, resulting in some puncta background staining. To differentiate the boundary of the recorded neuron and the Rh123 signal from adjacent cells, Alexa 594 was also included in the intracellular solution (Fig. 5A). Before SD onset, high K+ induced a relatively small and gradual increase in Rh123 fluorescence (Fig. 5C). Immediately following the onset of SD, the cytosol Rh123 fluorescence increased (21.2 ± 2.0% ΔF/R, n = 6) with a rapid time course, reaching a maximum approximately 2 s later than the rapid IOS change and ISD (Fig. 5D,F). This indicated that a profound but transient mitochondrial membrane depolarization occurs in neurons just after SD onset (Fig. 5D,F). Because calcium homeostasis is a crucial factor in mitochondrial function (Duchen 2004), we tested the role for calcium entry during SD by perfusing slices with Ca2+-free extracellular solutions to determine whether the mitochondrial depolarization was dependent upon increased cytosolic [Ca2+]. Consistent with previous studies (Basarsky et al. 1998), SD induction and propagation (15 ± 3 μm/s, n = 5) was not blocked by removal of extracellular calcium. However, the maximal rate of rise in the IOS was slower in Ca2+-free solutions (2.3 ± 0.4% ΔT/s, n = 5, P < 0.001) compared with control (7.5 ± 0.5% ΔT/s, n = 9). In the absence of extracellular calcium, the SD-induced mitochondrial depolarization was completely abolished compared with control (0.5 ± 0.6%, P < 0.001, n = 5, Fig. 5G,H). The Δψm depolarization may result from either opening of the mitochondrial permeability transition pore (MPTP), which is induced by uptake of excessive calcium into mitochondria (Skulachev 2006; Nicholls 2008), or uptake of cytosolic calcium through the mitochondrial uniporter (Kirichok et al. 2004). We tested the potential role of the MPTP by including the inhibitor CsA in the intracellular solution (Kovacs et al. 2005) but found no significant change in the Δψm depolarization induced by SD compared with control (18.5 ± 1.5% ΔF/R, P = 0.29, Fig. 5G). However, in the presence of Ru360, a specific inhibitor for the calcium uniporter, the Δψm depolarization was markedly reduced during SD (10.7 ± 0.5% ΔF/R, n = 5, P < 0.01). Collectively, these data suggest that SD-induced mitochondrial depolarization depends upon calcium influx into the cytosol of the neuron and subsequent uptake into mitochondria through the calcium uniporter. This effect, however, is not required for the induction or propagation of SD but instead reflects a consequence of the SD event.
We next examined Rh123 changes in astrocytes and compared the responses with that observed in neurons. Astrocytes were loaded with Rh123 using bath application that led to bright labeling of these cells, which was confirmed by co-labeling with SR101 (Fig. 6). When high K+ was perfused but did not evoke SD, astrocytes showed a gradual increase in Rh123 fluorescence (10.6 ± 2.5% ΔF/R, n = 12), similar to that of neurons. However, the degree of the fluorescence change was variable among astrocytes within the same field of view (Fig. 6E). When SD occurred, most astrocytes did not show an additional increase in Rh123 intensity (Fig. 6F) and the average maximal increase of Rh123 fluorescence was not significantly different from high K+ alone (12.4 ± 2.9% ΔF/R, n = 17, P = 0.65, Fig. 6G). Therefore, these results indicate that neurons are the major locus for the mitochondrial membrane potential changes induced by SD propagation.
We have used a combination of two-photon laser scanning microscopy (TPLSM) and electrophysiological techniques to show changes in cell volume, pHi, and mitochondrial membrane potential in neurons versus astrocytes during SD in cerebral cortex in vitro. A transient increase in light transmittance was observed to be coincident with electrophysiological signals of SD, indicating that imaging IOS changes using the infrared ultrafast laser can be used as a reliable indicator of the SD onset. Imaging individual YFP-labeled neurons showed that SD evoked a rapid but transient volume increase in neurons, but not in astrocytes. At the onset of SD, neurons also showed a transient decrease of pHi and a transient mitochondrial membrane depolarization. The Δψm was dependent upon cytosolic calcium changes and was caused by calcium uptake into mitochondria through calcium uniporters. Astrocytes did not exhibit the same changes in pHi and Δψm during the onset of SD.
IOS and KCl Induction of SD
SD-related IOS changes are complex, involving multiple underlying mechanisms such as cell swelling, dendritic beading, neuronal secretion, and/or mitochondrial volume changes (Salzberg et al. 1985; Aitken et al. 1999). We observed that the SD-related IOS showed a transient change that consisted of an initial increase followed by a rapid decrease in transmitted light, a pattern consistent with several previous reports (Anderson and Andrew 2002; Fayuk et al. 2002). Contrary to these results, some studies have shown an increase in light scattering, which decreases light transmittance during the initial phase of SD (Snow et al. 1983; Aitken et al. 1998; Vilagi et al. 2001; Buchheim et al. 2002). These differences may be due to the arrangement of the optical detectors or the wavelengths used in the experiment. We examined this transmitted image at infrared wavelengths, in which IOSs are principally generated by light scattering (Ba et al. 2002), resulting in an increase in light transmittance as cells swell. Despite the complexity of the IOS though, our data clearly show that by taking the first derivative of this signal, the rapid rise and peak is a robust and faithful indicator of the SD wave front and an indispensable tool for temporally aligning SD with other pathophysiological events within the cortical slice.
Our results provide evidence for the existence of differential signals generated by neurons as compared with astrocytes during SD and high K+. The rapid IOS component that was correlated with SD onset occurred simultaneously with several events only observed in neurons. The most dramatic change observed was a remarkably large yet quickly reversible swelling, which showed maximal volume increase 60 s after the peak IOS. This latency suggests that this effect may be caused by the large conductance associated with the rapid depolarization of SD as massive and prolonged ion influx would dramatically enable sustained water movement into the cell. The shorter time course of the IOS transient compared with the duration of neuronal swelling suggests that the transient increase in light transmittance was not solely due to neuronal volume changes in the soma and dendritic shafts but perhaps additionally reflects yet to be described cellular processes. Finally, the slow nonpropagating IOS increase induced by KCl was coincident with the astrocytic volume increase. This was similar in all slices whether or not SD was evoked. Therefore, the slower component of IOS was likely due to the change in light scattering caused by astrocytic swelling.
Contribution of Neurons versus Astrocytes to the Interstitial Volume Changes
Based on measurements of the concentration of extracellular cell-impermeable indicators like tetramethylammonium and tetraethylammonium, it has been concluded that the interstitial space shrinks rapidly at SD onset (Hansen and Olsen 1980; Jing et al. 1994). However, in these studies, it was unclear which cell type was responsible. A recent study has shown that neuronal swelling was likely to be the main cause for the tissue swelling during SD in vivo (Takano et al. 2007). Our observations demonstrate that astrocytes swell only as a result of the elevated [K+]o and not because of SD, while neurons were essentially opposite, showing resistance to elevated [K+]o and only swelled in response to SD. In astrocytes, the high K+ would lead to potassium uptake followed by water entry through aquaporins to compensate for the subsequent osmolarity shift (Kimelberg 2000). This is consistent with data showing that astrocytes express high levels of aquaporin 4 (Nielsen et al. 1997). In contrast, neurons lack aquaporin channels (Rash et al. 1998; Badaut et al. 2004), explaining their inability to osmotically respond to K+ elevation. These ideas have been supported in a recent imaging study, in which neurons fail to change cellular volume during acute osmotic stress whereas astrocytes rapidly swell (Andrew et al. 2007). Our results suggest that SD-induced neuronal swelling is probably due to the ionic fluxes during the large conductance increase during SD. However, it is currently unclear which ion channels are opening to allow the movement of water.
Intracellular pH Changes
Another novel observation in our study is that pHi rapidly acidifies in neurons during SD. Previous studies of pH measurements mainly focused on pHo. For instance, Kraig et al. (1983) have used a double-barreled liquid membrane pH ion-selective micropipette to obtain rapid and repeatable pHo measurements during SD. They reported that SD was accompanied by an initial small alkaline shift in pHo and was followed by a long-lasting acid phase (Kraig et al. 1983; Tong and Chesler 1999). Although this study showed that the pHo changes resulted from the membrane transport of H+ and HCO3−, these extracellular measurements could not differentiate between the contributions of pH transients in neurons versus astrocytes. The development of fluorescent pH indicators provided the ability to obtain continuous real-time monitoring of pHi in individual cells. BCECF is the most widely used pH indicator because it is well retained in cells (i.e., low leakage rate), its pKa (6.98) is close to natural pHi, and it can be used for quantitive pH measurements with ratiometric imaging (Paradiso et al. 1984) or for imaging qualitative pH changes using TPLSM. We used BCECF imaging with TPLSM to provide spatiotemporal determination of pHi changes in neurons versus astrocytes. During SD we observed remarkably different pHi transients in neurons versus astrocytes. The rapid pHi decrease in neurons suggests a transient proton influx or HCO3− efflux, which accompanies the large ionic flux at the onset of SD. In contrast to the neurons, astrocytes increased pHi gradually when [K+]o was elevated, but they did not exhibit any unique changes to SD. The gradual pHi change in high K+ was consistent with previous studies which showed that raising [K+]o or electrical stimulation caused intracellular alkalization in astrocytes (Chesler and Kraig 1987, 1989). This alkalization is due to the astrocytic Na+-HCO3− cotransporter, which transports 2 or 3 HCO3− for each Na+. Astrocytic membrane depolarization induced by elevated [K+]o shifts the electrochemical gradient toward increased influx of HCO3− thereby raising pHi (Chesler 2003). This suggests that neuronal activities likely account for the pHo transients at the onset of SD.
Δψm Changes during SD
There are several mechanisms underlying mitochondrial depolarization, including mitochondrial Ca2+ cycling that is dependent on the calcium uniporter and Na+/Ca2+ exchanger, Ca2+-dependent mitochondrial membrane transition, or a disruption of the ADP/ATP and the oxidized/reduced nicotinamide adenine dinucleotide ratio (Duchen 2004). Under normal physiological conditions, the electrochemical potential gradient across the mitochondrial membrane drives the uptake of calcium from cytosol into mitochondria. The calcium is taken up through the uniporter on the inner membrane and it therefore depolarizes Δψm (Gunter et al. 1994). It has been reported that mitochondrial calcium uptake follows the cytosolic calcium elevation with a time lag (Rakhit et al. 2001). This is in line with our observations that the mitochondrial signal occurs later than the IOS and electric signals. The measured increase in cytosolic Rh123 fluorescence during SD suggests depolarized mitochondrial membrane potential (Baracca et al. 2003), which would reduce oxygen consumption by mitochondria thereby possibly transiently increasing pO2 (Kadenbach 2003).
In conclusion, our results suggest that astrocytes are not active contributors to the generation of SD. Astrocytes appear to merely follow the intense depolarization of neurons and respond to the increased [K+]o which is known to occur during SD. The neuronal depolarization we observed during SD is unusual in that it still propagated in TTX and calcium-free external solutions, a combination that blocked both Na+- and Ca2+-dependent spikes and synaptic transmission. Our imaging experiments suggest that it is most likely a nonselective cation conductance that allowed the substantial changes to pH. In addition, the depolarization during SD triggered calcium influx into neurons that led to mitochondrial depolarization. Our observations suggest that neurons are the key locus for the pathogenesis induced by ischemia or other disorders that involved SD. It will be important to extend our observations to in vivo models of SD (Takano et al. 2007) as an intact blood supply may differentially affect SD mitochondrial responses. However, our results are consistent with the conclusion from Somjen's overview of the literature that “in SD, neurons lead and glial cells follow” (Somjen 2001). These data help clarify the role of neurons and astrocytes in SD and provide an important framework toward a mechanistic comprehension of the neuropathology.
Canadian Institutes of Health Research; Fondation Leducq. B.A.M. is a Canada Research Chair in Neuroscience and Michael Smith Distinguished Scholar; Heart and Stroke Foundation of Canada (HSFC) and Michael Smith Foundation for Health Research (MSFHR) to N.Z.; Alberta Heritage Foundation for Medical Research, HSFC, MSFHR, and the Natural Sciences and Engineering Council of Canada (to G.R.J.G.).
Conflict of Interest: None declared.