Simultaneous Monitoring of Ionophore- and Inhibitor-mediated Plasma and Mitochondrial Membrane Potential Changes in Cultured Neurons*

Although natural and synthetic ionophores are widely exploited in cell studies, for example, to influence cytoplasmic free calcium concentrations and to depolarize in situ mitochondria, their inherent lack of membrane selectivity means that they affect the ion permeability of both plasma and mitochondrial membranes. A similar ambiguity affects the interpretation of signals from fluorescent membrane-permeant cations (usually termed “mitochondrial membrane potential indicators”), because the accumulation of these probes is influenced by both plasma and mitochondrial membrane potentials. To resolve some of these problems a technique is developed to allow simultaneous monitoring of plasma and mitochondrial membrane potentials at single-cell resolution using a cationic and anionic fluorescent probe. A computer program is described that transforms the fluorescence changes into dynamic estimates of changes in plasma and mitochondrial potentials. Exploiting this technique, primary cultures of rat cerebellar granule neurons display a concentration-dependent response to ionomycin: low concentrations mimic nigericin by hyperpolarizing the mitochondria while slowly depolarizing the plasma membrane and maintaining a stable elevated cytoplasmic calcium. Higher ionomycin concentrations induce a stochastic failure of calcium homeostasis that precedes both mitochondrial depolarization and an enhanced rate of plasma membrane depolarization. In addition, the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone only selectively depolarizes mitochondria at submicromolar concentrations. ATP synthase reversal following respiratory chain inhibition depolarizes the mitochondria by 26 mV.

Although natural and synthetic ionophores are widely exploited in cell studies, for example, to influence cytoplasmic free calcium concentrations and to depolarize in situ mitochondria, their inherent lack of membrane selectivity means that they affect the ion permeability of both plasma and mitochondrial membranes. A similar ambiguity affects the interpretation of signals from fluorescent membrane-permeant cations (usually termed "mitochondrial membrane potential indicators"), because the accumulation of these probes is influenced by both plasma and mitochondrial membrane potentials. To resolve some of these problems a technique is developed to allow simultaneous monitoring of plasma and mitochondrial membrane potentials at singlecell resolution using a cationic and anionic fluorescent probe. A computer program is described that transforms the fluorescence changes into dynamic estimates of changes in plasma and mitochondrial potentials. Exploiting this technique, primary cultures of rat cerebellar granule neurons display a concentration-dependent response to ionomycin: low concentrations mimic nigericin by hyperpolarizing the mitochondria while slowly depolarizing the plasma membrane and maintaining a stable elevated cytoplasmic calcium. Higher ionomycin concentrations induce a stochastic failure of calcium homeostasis that precedes both mitochondrial depolarization and an enhanced rate of plasma membrane depolarization. In addition, the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone only selectively depolarizes mitochondria at submicromolar concentrations. ATP synthase reversal following respiratory chain inhibition depolarizes the mitochondria by 26 mV.
Whereas synthetic and natural ionophores are powerful tools for manipulating cellular physiology, their inherent lack of specificity means that they influence the ion permeability of several membranes in the cell. The two classes of ionophores that are most commonly used are the electroneutral Ca 2ϩ /2H ϩ exchangers typified by ionomycin, and protonophores, such as carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). 2 Ionomycin is most commonly used with the assump-tion that it will generate a stable, moderately elevated, cytoplasmic free Ca 2ϩ concentration, [Ca 2ϩ ] c with maintained cell viability. However, ionomycin also intercalates into the inner mitochondrial membrane where it provides an additional pathway for Ca 2ϩ efflux from the matrix in parallel with the native Ca 2ϩ /Na ϩ antiporter, setting up a protondissipating, i.e. uncoupling, Ca 2ϩ cycle that is controlled by the activity of the mitochondrial Ca 2ϩ uniporter and hence by [Ca 2ϩ ] c (1). The bioenergetic consequences of this are usually ignored, although they could have profound effects on cellular function. Conversely, protonophores such as FCCP that are widely employed to depolarize mitochondria in intact cells can affect plasma membrane potentials at higher concentrations (2).
An equally important ambiguity surrounds the widespread use of cationic, membrane permeant, fluorescent probes as mitochondrial membrane potential indicators (3)(4)(5) because their uptake and equilibrium accumulation within the mitochondrial matrix of the cell is responsive equally to the plasma membrane potential, ⌬ p , and the mitochondrial membrane potential, ⌬ m . With the explosion of interest in the multiple roles played by in situ mitochondria in cell physiology and pathology (reviewed in Refs. 6 and 7) has come the importance of accurately monitoring changes in ⌬ m in intact cells. The complicating role of the plasma membrane was recognized at an early stage (4) and is exacerbated in studies in which both ⌬ m and the plasma membrane potential (⌬ p ) change during the experiment. We have previously presented a semiquantitative technique for the interpretation of whole cell cationic indicator traces under such conditions (8) based on curve fitting and estimates of likely changes in ⌬ p . The approach has proven useful for deciphering experiments in "quench mode" (when the probe concentration in the matrix is sufficient for reversible aggregation) as well as for interpreting traces obtained with probes with differing permeability rate constants (8 -11). Quench mode is only applicable to experiments in which rapid step changes in ⌬ m occur while the cell is being imaged (8). Conversely, low probe loadings that avoid matrix quenching must be employed to follow slow changes in potential as well as to estimate pre-existing values of ⌬ m in cell populations (reviewed in Ref. 12). Under these latter conditions there is serious ambiguity as to whether the observed change in fluorescence is due to a difference in ⌬ p , ⌬ m , or both.
It is evidently important to monitor both potentials. Anionic membrane-permeant probes are excluded from polarized cells due to the negative plasma membrane potential but partition increasingly into the cell upon plasma membrane depolarization. Because of their negative charge these probes are not accumulated by mitochondria. Anionic oxonol dyes have been used for several years to monitor changes in ⌬ p (13)(14)(15)(16), but their usefulness is compromised by a relatively slow equilibration across the plasma membrane. Recently a proprietary plasma membrane potential assay kit (Molecular Devices, Sunnyvale, CA) has become available in which the problem of background fluorescence from the high extracellular probe concentration is suppressed by a hydrophilic quencher (17).
In this paper a technique was developed for combining the use of the Molecular Devices fluorescent anion (which is termed PMPI, for "plasma membrane potential indicator") with the established cationic indicator tetramethylrhodamine methyl ester (TMRM ϩ ). At the same time we have devised a curve-fitting spreadsheet to interpret the traces. As well as providing a more quantitative means of compensating the TMRM ϩ signal for changes in ⌬ p , this combined technique allows for the first time simultaneous and continuous monitoring of ⌬ p and ⌬ m in cultured neurons. Whereas the technique is validated for cerebellar granule neurons exposed to a variety of ionophores and inhibitors on a confocal microscope, with suitable calibration the methodology is equally applicable to any attached cell preparation and can be used with non-confocal imaging. Finally the curve-fitting spreadsheet may also be used for single-probe studies of ⌬ m , where ⌬ p is invariant or its changes can be estimated.

MATERIALS AND METHODS
Reagents-TMRM ϩ fluo-4 and fluo-5F were from Molecular Probes (Eugene, OR). PMPI is a proprietary component of the Membrane Potential Assay Kit (R-8042) from Molecular Devices Corp., Sunnyvale, CA. All other reagents were from Sigma.
Preparation of Cerebellar Granule Neurons-Cerebellar granule neurons were prepared from 7-day-old Wistar rats as previously described (18) with modifications. Briefly, cells were plated into coverslip-based 8-well chambers (LabTek, Naperville, IL) previously coated with 33 g/ml polyethyleneimine, at a density of 380,000 cells per 0.8-cm 2 well. Cultures were maintained in minimal essential medium supplemented with 10% fetal bovine serum, 30 mM glucose, 20 mM KCl, 2 mM glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. 24 h after plating, 10 M cytosine arabinoside was added to inhibit growth of non-neuronal cells. Cell cultures were maintained at 37°C in an incubator with a humidified atmosphere of 5% CO 2 , 95% air and used for experiments at 12-14 days in culture.
Plasma Membrane Potential Indicator-An individual vial from a Molecular Devices "membrane potential assay kit, explorer format" (R-8042) containing a proprietary plasma membrane potential indicator was reconstituted in 1 ml of distilled water, dispensed into 50-l aliquots, and frozen (PMPI stock). For spectral analysis 20 l of PMPI stock was diluted with 250 l of water and extracted with 250 l of octanoyl alcohol to separate the anionic indicator from the aqueous quencher present in the PMPI stock. Excitation and emission spectra ( Fig. 1) were determined with a PerkinElmer LS50 scanning spectrophotometer for PMPI in octanoyl alcohol and compared with tetramethylrhodamine in water.
Simultaneous Monitoring of PMPI and TMRM ϩ Fluorescence-Cerebellar granule neurons (CGN) were washed and incubated (37°C, pH 7.4) for 45 min prior to imaging with a medium (low K-medium) containing 3.5 mM KCl, 120 mM NaCl, 1.3 mM CaCl 2 , 0.4 mM KH 2 PO 4 , 5 mM NaHCO 3 , 1.2 mM Na 2 SO 4 , 15 mM D-glucose, 20 mM Na-TES, 1 M tetraphenylboron, 5 nM TMRM ϩ , and 0.5 l/ml PMPI stock. An identical medium in which 120 mM NaCl was substituted by 120 mM KCl was prepared (high-K medium). 5 nM TMRM ϩ is below the limit for probe aggregation and quenching within the matrix. In some experiments the TMRM ϩ concentration was varied. The PMPI concentration was chosen to obtain a signal comparable with that of TMRM ϩ . The presence of tetraphenylboron (TPB Ϫ ) facilitates the equilibration of TMRM ϩ and other lipophilic cations (4,8,19) across the plasma membrane. No effect of 1 M TPB Ϫ on the equilibrium distribution of either PMPI or TMRM ϩ was detected. No extracellular PMPI fluorescence could be detected and addition of 0.5% Triton X-100 abolished both the TMRM ϩ and PMPI signals associated with the cells.
Cells were imaged on a Zeiss Pascal confocal Axiovert 100M microscope equipped with a computer-driven stage. The technique is equally applicable to non-confocal imaging, and unless otherwise stated, the pinhole diameter was increased to give an optical slice of 10 m to allow collection of the defocused signal from individual somata. Fluorescence was excited in single track mode with the 514-nm band of an argon laser. Because the emission peaks of the two indicators are separated by only 20 nm (Fig. 1) the spectral overlap must be corrected. To accomplish this the emitted epifluorescence was split with a 570-nm dichroic mirror and detected in channel 1 through a Chroma 595-650-nm filter and channel 2 with a Chroma 525-575-nm filter. Crossover between the two channels was quantified with cell preparations loaded with a single probe. The values obtained were dependent upon the amplifier gains for the two channels and so these were kept constant throughout. Laser power was adjusted to obtain an optimal image and did not exceed 5% when used with the ϫ20 air objective. Whereas the Zeiss software allows for pixel by pixel correction of crossover between the two emission spectra using its subtractive function, it is equally valid to correct single-cell time courses in a spreadsheet. Scan averaging was performed for the high resolution studies shown in Fig. 2 using the ϫ63 oil immersion objective with line mode summing of four repetitive scans at 1% laser power to minimize phototoxicity.
Curve-fitting Computer Simulation-The simulation to convert the TMRM ϩ and PMPI fluorescence traces into a time course of changes in ⌬ p and ⌬ m is described in supplementary data where it may be accessed as an Excel spreadsheet. The mathematical background to the simulation is derived below. When applied to data (Figs. 3, 6 -8, and 10) the experimental   Fig. 2 depicts equatorial confocal slices through the somata of 12 DIV CGNs equilibrated with PMPI and 5 nM TMRM ϩ . In the polarized cells, PMPI fluorescence is faint and associated with the plasma membrane with slight cytoplasmic fluorescence. Depolarization with 25 mM K ϩ results in a increase in cell-associated PMPI fluorescence following entry of the probe into the cytoplasm although the probe remains excluded from the nucleus. The decrease in TMRM ϩ fluorescence is due to the decreased probe accumulation across the plasma membrane in the partially depolarized cells (see Equation 5).

RESULTS
Computer Simulation-The computer simulation can be accessed in supplementary data. The theoretical basis of the simulation is developed below.
PMPI Calibration-It is first necessary to calibrate the enhancement in the PMPI fluorescence as a function of plasma membrane depolarization. This does not follow an ideal Nernstian relationship, but instead is determined empirically by quantifying the fluorescent enhancement obtained when ⌬ p is depolarized by increasing KCl concentrations. The relationship between ⌬ p and the K ϩ concentration of the medium (Fig. 3A) was calculated from electrophysiological data in Laritzen et al. (20) applying the Goldman-Hodgkin-Katz equation. An initial value for the ⌬ p of 8 -12 DIV rat CGNs in 3.9 mM KCl media of Ϫ83 mV at 37°C was calculated. A series of experiments were then performed with PMPI-equilibrated CGNs in which step increases in K ϩ concentrations were made by removing defined volumes of low-K medium and replacing this with an equal volume of high-K medium to give final K ϩ concentrations from 4 to 80 mM. The mean fluorescent enhancement for 10 cell bodies was determined for each K ϩ concentration and plotted as a function of ⌬ p (Fig. 4A). The best fit with the empirical fluorescent enhancement was obtained with a second-order regression curve, where E is the fluorescent enhancement relative to cells in 3.9 mM KCl medium and ⌬ KCl the calculated plasma membrane potential at a given KCl concentration. It should be noted that the fluorescent enhancement is considerably less than that predicted for the change in free cytoplasmic PMPI concentration from the Nernst equation (dashed line in Fig. 3A), presumably due to complicating factors of probe binding to membranes and proteins and changes in fluorescence yield. Using this empirical relationship, an observed change in whole cell PMPI fluorescence can be calibrated as a function of ⌬ p .
Single-channel Monitoring of ⌬ p -The PMPI probe may be used in the absence of TMRM ϩ and tetraphenylboron to monitor changes in ⌬ p using the empirical curve fit in Equation 1. In this case the only additional parameter that is required to produce a dynamic read-out of potential is the rate constant for the redistribution of PMPI across the plasma membrane. This is estimated from the kinetics of the PMPI fluorescence enhancement resulting from a step change in ⌬ p caused by an increase in KCl concentration. The assumption is made that the rate of change in fluorescence is a first-order function of the disequilibrium between the instantaneous and final probe distribution and is proportional to the rate constant for the PMPI re-equilibration across the plasma membrane, where ␦f/␦t is the rate of increase in signal, f (t) is the fluorescence at time t, f (f) is the final fluorescence, and k PMPI is the rate constant (s Ϫ1 ). Using this approach, a good curve-fit for the KCl jump is obtained by adopting a rate constant of 0.04 s Ϫ1 for PMPI equilibration across the plasma membrane (data not shown). It must be noted that this value refers to the granule cell soma and will differ for other cells. CGNs were equilibrated with PMPI and 5 nM TMRM ϩ in low-(3.9 mM) K medium as described under "Materials and Methods." Images were corrected for crossover between the fluorophores (Equations 1 and 2). After imaging cells in low-K medium, media was exchanged to increase [K ϩ ] to 25 mM and the field was imaged again when re-equilibration was complete. The lower panels show the intensity profiles through the cell, scale in micrometers. Note the reciprocal changes in probe intensity, the lack of co-localization between the PMPI and TMRM ϩ fluorescence, and the exclusion of PMPI from the nucleus.
To validate the responsiveness of PMPI fluorescence to changes in ⌬ p , 12 DIV CGNs were depolarized with NMDA, kainate, or ouabain (data points in Fig. 3, B-D). The computer simulation (detailed in supplementary data) was used to curve-fit the experimental data. The solid lines in Fig. 3, B-D, show the close fit obtained with the experimental data points when the ⌬ p time courses shown in Fig. 3, E-G, are input into the simulation. Comparison with the final KCl depolarization, calculated to decrease ⌬ p to Ϫ40 mV, demonstrates how extensive the plasma membrane depolarization is in the cell population and also that in both cases there is a partial recovery of membrane potential, perhaps due to receptor desensitization, during the exposure. It is notable that the repolarization of the plasma membrane following addition of NBQX to terminate ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid/kainate receptor activation is slower than that following NMDA receptor inhibition by MK801. The extent of depolarization occurring during ␣-amino-3-hydroxy-5-methyl-4isoxazolepropionic acid/kainate receptor activation correlates closely with electrophysiological data reported by Kiedrowski and Mienville (21). CGNs were equilibrated with PMPI in 3.9 mM K medium as described under "Materials and Methods." A, varying volumes of 120 mM K ϩ medium were substituted to give final K ϩ concentrations from 4 to 80 mM with a proportionate decrease in Na ϩ . Plasma membrane potentials were calculated from the Goldman equation assuming conductances for K ϩ , Na ϩ , and Cl Ϫ of 1:100:1. (see text). The fluorescent enhancement relative to 4 mM K ϩ was determined for the mean of 10 cell bodies with no contaminating membrane fragments at each K ϩ concentration (closed squares, K ϩ concentration in parentheses). The trend line (solid) is given by a second-order regression curve (see Equation 1). The  Single-channel Monitoring of ⌬ m -The simulation described in the supplementary data is valid for the interpretation of experiments in which just the cationic indicator is present, as long as the plasma membrane potential is constant during the experiment, or changes in ⌬ p can be estimated. The validity extends to traces obtained under nonquench and quench conditions, and with cationic probes that are rapidly permeant (e.g. TMRM ϩ in the presence of tetraphenylboron) or slowly permeant (e.g. rhodamine 123).
Mathematic Basis of the Simulation-The basic simplifying assumptions are similar to those previously reported (8).
(i) The equilibrium distribution of TMRM ϩ (and under non-quench conditions its fluorescence) is a simple consequence of the Nernstian distribution of the probe across both the plasma and mitochondrial membranes; thus no corrections are made for binding, non-ideal behavior or spectral changes resulting from probe redistribution (see Ref. 22).
(ii) The matrix volume is assumed to remain constant during an experiment. This is an approximation, because in practice in some conditions such as extensive Ca 2ϩ loading, alteration in mitochondrial morphology consistent with matrix swelling can be observed (23).
(iii) As a consequence of the enormous surface/volume ratio of the mitochondrial inner membrane/matrix compared with the plasma membrane/ cytoplasm, probes equilibrate much more rapidly across the mitochondrial membrane than across the plasma membrane (8). For these studies it is valid to assume that re-equilibration of TMRM ϩ between matrix and cytoplasm occurs within the sampling interval of the experiment.
Interpretation of the ⌬ m profile underlying changes in the whole cell TMRM ϩ fluorescence is somewhat complex and depends on the initial values of ⌬ p and ⌬ m , the fraction of the cell volume occupied by the mitochondrial matrices, the quench limit (the concentration at which the probe forms non-fluorescent aggregates within the mitochondrial matrix) and, for dynamic determinations, the rate constant for the equilibration of the probe across the plasma membrane. Each of these parameters will now be derived for the cerebellar granule neuron preparations employed in the current study.
Initial Values of ⌬ p and ⌬ m -The derivation of an initial value of Ϫ83 mV for ⌬ p for CGNs in 3.9 mM K medium at 37°C has been discussed above. The consensus value for the initial ⌬ m of in situ mitochondria respiring between States 3 and 4 is close to 150 mV, based on semiquantitative data obtained for the mitochondria within isolated nerve terminals (24) and isolated hepatocytes (25) and is consistent with data obtained with isolated mitochondria (see for example, Ref. 26). This initial value is adopted for the simulation. The equilibrium concentration of TMRM ϩ in the cytoplasm (c) and mitochondrial matrix (m) relative to the external medium (e) is given at 37°C by the following equations. The Fraction of the Cell Volume Occupied by the Mitochondrial Matrices-The relative contributions of the cytoplasmic and matrix pools of a cationic probe to whole cell fluorescence will depend on the relative volumes of the two compartments. One way to estimate the volume fraction x of the soma occupied by the mitochondrial matrix is to determine the residual cytoplasmic fluorescence after mitochondrial depolarization by calculating the ratio of the whole cell TMRM ϩ fluorescence (in non-quench mode) for a cell with depolarized mitochondria (e.g. in the presence of myxothiazol to inhibit the respiratory chain and oligomycin to block the ATP synthase) relative to the same cell with polarized mitochondria prior to the addition of inhibitors (Fig. 4A). If the fraction of the cell occupied by the mitochondrial matrices is x then the ratio of the whole cell fluorescence of depolarized versus polarized mitochondria, i.e. (⌺TMRM ϩ ) depol /(⌺TMRM ϩ ) pol , will be given as Equation 6.
The ratio determined experimentally in 10 random cells was 0.121 Ϯ 0.01, and substituting this value into Equation 6 gives a value for the volume fraction x of 2.3 Ϯ 0.02% when ⌬ m is 150 mV (Fig. 4A). It must be emphasized that this assumes that there is no potential-independent binding of TMRM ϩ to components of the cell.
For these studies it is valid to assume that re-equilibration of TMRM ϩ between matrix and cytoplasm in response to a step change in ⌬ m occurs within the sampling interval of the experiment. When calculating fluxes across the plasma membrane the cytoplasm plus matrix can thus be considered to be a single compartment. The apparent "volume" of this combined compartment is equivalent to a cytoplasm whose volume is increased by the factor 1 ϩ x ϫ 10 ⌬m/61 from Equation 6. This "expansion factor" is critical for the understanding of the dynamic TMRM ϩ fluorescence response to a change in ⌬ p or ⌬ m . Qualitatively, a step decrease in ⌬ m will decrease the apparent volume and thus increase the concentration in the cytoplasm, leading to redistribution across the plasma membrane to restore the Nernst equilibrium.
Establishment of the Quench Limit for TMRM ϩ in the Mitochondrial Matrix-For the simulation to be valid under both non-quench and quench conditions, it is important to establish the concentration of TMRM ϩ in the matrix that initiates aggregation. Cells were equilibrated with TMRM ϩ concentrations from 50 to 10 nM (Fig. 4B). A mixture of rotenone, oligomycin, and FCCP was added to cause a rapid mitochondrial depolarization that is accompanied by a transient "spike" in whole cell fluorescence if sufficient probe was present in the matrix to cause aggregation and quenching. Optimal simulation of these traces using the values determined above for the cell parameters was obtained with a value of 140 M for the quench limit for TMRM ϩ in the mitochondrial matrix (Fig. 4C).
Estimation of the Rate Constant for TMRM ϩ Re-equilibration Across the Plasma Membrane-A minimum estimate for the rate constant in the presence of 1 M TPB Ϫ was obtained by assuming that the collapse of ⌬ m on addition of a 10-fold excess of FCCP (Fig. 8B, cf. A) is fast compared with the sampling interval. By curve fitting the experimental points in Fig. 8B with this proviso, a good fit is obtained with a rate constant not less than 0.015 s Ϫ1 . It must be emphasized that these values pertain to the somata of rat cerebellar granule cells. Other cells will differ depending on their size, and the probe will redistribute across thin processes more rapidly than into the cell body (8).
Simultaneous Monitoring of Changes in ⌬ p and ⌬ m by Dual Labeling with PMPI and TMRM ϩ -With appropriate choices of excitation and emission wavelengths it is possible to combine the determinations of ⌬ p and ⌬ m in a single experiment with both probes (see "Materials and Methods"). In the present study this was achieved by exciting at 514 nm and collecting dual emissions at 525-570 and 595-650 nm. Singletrack excitation has the advantage that crossover between the channels can be corrected pixel by pixel by the Zeiss software, but crossover could be reduced by exciting PMPI at about 530 nm and TMRM ϩ at 570 nm with cell by cell crossover corrected for in a spreadsheet. It must be emphasized that confocal microscopy is not obligatory for these studies.
The technique for combining the ⌬ p and ⌬ m simulations is detailed in the supplementary data. Briefly, the empirical constants determined above are first entered. The simulated PMPI fluorescence time course is then curve-fitted to the experimental PMPI fluorescence. Finally the TMRM ϩ simulation is curve-fitted to the experimental TMRM ϩ data points.
Monitoring ⌬ p Removes Inherent Ambiguity from Mitochondrial Membrane Potential Determinations- Fig. 5 shows simulated TMRM ϩ and PMPI fluorescence traces generated by the program using the constants derived above for CGNs. The simulations show the predicted responses to sudden or slow partial depolarization of the plasma and mitochondrial membranes for cells equilibrated with low (non-quenching) and high (quenching) concentrations of TMRM ϩ . Quench conditions are only applicable to one specialized condition, namely a the sudden change in ⌬ m (e.g. Fig. 5C) producing a transient spike in cell fluorescence as the quenched probe is diluted into the cytoplasm. This has been exploited to examine the relationship between the initial Ca 2ϩ -mediated mitochondrial depolarization seen in CGNs upon NMDA receptor activation and the subsequent survival of the cells (8) and in several studies investigating the final catastrophic collapse of ⌬ m in this context (27,28). However, partial depolarization often produces no change in steady-state signal (Fig. 5C) and studies that only determine initial and final fluorescence can form erroneous conclusions. Quench mode is also relatively insensitive to slow mitochondrial depolarization (Fig. 5G), whereas the response to slow mitochondrial hyperpolarization (not shown) is indistinguishable from that resulting from plasma membrane depolarization (Fig. 5, D and H).
Non-quench mode, typically requiring less than 10 nM TMRM ϩ , is equally applicable to rapid (Fig. 5A) and slow (Fig. 5E) changes in ⌬ m , but virtually identical traces are produced by equivalent changes in ⌬ p (Fig. 5, B and F). The parallel monitoring of ⌬ p removes this ambiguity (Fig. 5, A and B and E and F). It should be noted that the PMPI signal faithfully reflects the changes in ⌬ p under all conditions. Uniport and Antiport K ϩ Ionophores Have Opposing Effects on ⌬ p and ⌬ m -The K ϩ -uniport ionophore valinomycin and the K ϩ /H ϩ exchange ionophore nigericin are valuable tools in isolated mitochondrial studies to equilibrate, respectively, ⌬ m (29) or ⌬pH (26) with the transmembrane K ϩ gradient. As with all ionophores, however, their lack of membrane selectivity means that their action in intact cells is complex, particularly because K ϩ conductances at the plasma membrane play the major role in the maintenance of ⌬ p . Valinomycin should slightly hyperpolarize the plasma membrane by bringing ⌬ p closer to the K ϩ diffusion potential while collapsing the membrane potential of the mitochondria in the high K ϩ cytoplasm. Fig. 6A shows that both of these changes in potential can be detected by the present methodology. It must be emphasized that valinomycin causes matrix swelling (30), which was not allowed for in the present simulation.
The K ϩ /H ϩ ionophore nigericin produces a complex TMRM ϩ trace (Fig. 6B) that would be difficult to interpret without the parallel monitoring of ⌬ p . Curve fitting with the simulation indicates a rapid partial depolarization of the plasma membrane, consistent with the efflux of K ϩ from the cytoplasm in exchange for protons. This is followed by a mitochondrial hyperpolarization of about 30 mV as the ionophore intercalates into the inner mitochondrial membrane, collapsing ⌬pH and allowing a compensatory increase in ⌬ m . This result suggests that the initial mitochondrial ⌬pH is at least Ϫ0.5 pH units (equivalent to 30 mV of protonmotive force).
Glycolysis Maintains ⌬ p but Lowers ⌬ m -In contrast to experiments performed with isolated mitochondria, respiratory chain inhibition in intact cells does not lead to a total collapse of ⌬ m , because the ATP synthase operating in reverse can function as an alternative proton pump driven by the hydrolysis of glycolytically generated ATP and maintain a suboptimal ⌬ m . Because the direction and rate of the ATP synthase will be in part governed by thermodynamic disequilibrium, it follows that the ⌬ m maintained by glycolysis will be lower than the ⌬ m generated by respiring mitochondria. Fig. 7A quantifies this difference for the present preparation. The experimental and simulated traces can be superimposed assuming a 26-mV drop in ⌬ m on addition of myxothiaxol to inhibit Complex III. Note that ⌬ p is maintained, indicating that glycolysis is sufficiently active to supply ATP for the plasma membrane Na ϩ /K ϩ -ATPase, the major utilizer of ATP in the neuron, indeed a slight ⌬ p hyperpolarization can be detected. In this and most subsequent figures the average responses of 10 cells are shown. The histogram in Fig. 7A shows the typical cell to cell variability in the membrane potential changes. In contrast, when ATP synthase reversal is prevented by oligomycin (Fig. 7B), myxothiaxol initiates a progressive collapse of ⌬ m .

Low Micromolar Concentrations of FCCP Depolarize the Plasma
Membrane-Protonophores such as FCCP are universally employed to depolarize mitochondria in intact cells. It is established that very high concentrations of protonophores also depolarize the plasma membrane (2), however, it is apparent from Fig. 8 that although 0.25 M FCCP is selective for the mitochondrial membrane potential, as little as 2.5 M FCCP initiates a biphasic plasma membrane depolarization that almost totally collapses ⌬ p over a period of 8 min. Whereas the mechanism of this plasma membrane depolarization was not further investigated, it is not due to a failure of mitochondrial ATP production because oligomycin was present in both experiments. It is thus apparent that great care must be taken when titrating in protonophore to intact cell preparations to avoid the metabolic and ionic complications inherent in an unrecognized plasma membrane depolarization.

Concentration-dependent Effects of Ionomycin on [Ca 2ϩ ] c , ⌬ m , and
⌬ p -The Ca 2ϩ /2H ϩ exchange ionophores ionomycin, A23187, and 4-bromo-A23187 have been used extensively to raise [Ca 2ϩ ] c in a great variety of studies. However, with relatively few exceptions (e.g. Refs. 1 and 31) the possible effects on in situ mitochondrial bioenergetics have tended to be ignored. The sequence of events is difficult to predict: an increase in [Ca 2ϩ ] c as a result of ionophore action at the plasma membrane will increase the activity of the mitochondrial Ca 2ϩ uniporter leading to increased uptake into the matrix. However, at the same time the ionophore action at the inner membrane will introduce an additional pathway for Ca 2ϩ efflux from the matrix, in parallel with the endogenous Ca 2ϩ /2Na ϩ exchanger. This enhanced Ca 2ϩ cycling is driven by proton re-entry into the matrix and thus begins to "uncouple" the proton circuit in the same way as a conventional protonophore. Indeed in an earlier study we investigated the relationship between [Ca 2ϩ ] c and the respiration of isolated nerve terminals in the presence of increasing ionophore (1). On the other hand, although this Ca 2ϩ cycling would be predicted to lower ⌬ m , the ionophore additionally collapses the mitochondrial pH gradient (32) that could have a nigericin-like effect (Fig. 6B) of allowing a compensatory increase in ⌬ m . Finally matrix acidification will destabilize any matrix Ca 3 PO 4 complex (40). The dynamic range of individual Ca 2ϩ indicators is limited, in the present context this makes it difficult to distinguish between a modest controlled rise in [Ca 2ϩ ] c and an uncontrolled catastrophic Ca 2ϩ deregulation as Ca 2ϩ entry across the plasma membrane overwhelms the Ca 2ϩ efflux and sequestration mechanisms of the cell. In Fig. 9 parallel experiments are shown where low ionomycin concentrations are added to CGNs loaded with either the high affinity fluo-4 (K d ϭ 345 nM) or the low affinity fluo-5F (K d ϭ 2300 nM). Ca 2ϩ homeostasis was only maintained for most cells during the period of the experiment with the lowest concentration of ionomycin (0.5 M). With 1 M ionomycin and fluo-5F (Fig. 9E) a sudden secondary stochastic rise in [Ca 2ϩ ] c is seen and with 3 M ionophore this Ca 2ϩ deregulation is seen in almost all cells. Thus in the entire fields 9% of cells deregulated with 0.5 M ionomycin, 33% with 1 M, and 92% with 3 M ionomycin within 13 min of ionophore addition.
When cells were loaded with both fluo-5F and TMRM ϩ (Fig. 10A) cells imaged during stochastic Ca 2ϩ deregulation showed a low [Ca 2ϩ ] c and retained mitochondrial TMRM ϩ labeling (e.g. cell 1), a high [Ca 2ϩ ] c and no TMRM ϩ labeling (e.g. cell 3), or a high fluo-5F fluorescence with a retained TMRM ϩ signal (e.g. cell 2). When the time courses of individual cells was followed it was found that the initiation of the secondary rise in [Ca 2ϩ ] c (a in Fig. 10, B and C) always preceded the initiation of fall in TMRM ϩ fluorescence (b in Fig. 10, B and C). Fig. 10B shows the time interval between these two events for individual cells within a given field.
As discussed earlier, the decrease in TMRM ϩ fluorescence cannot be interpreted without parallel measurement of ⌬ p , because it could be a consequence of depolarization of the mitochondrion, the plasma membrane, or both. It is additionally important to relate changes in plasma FIGURE 6. The K ؉ -uniport ionophore valinomycin and the K ؉ /H ؉ antiport ionophore nigericin have opposing effects on CGN ⌬ p and ⌬ m . CGNs were equilibrated in low-K medium with PMPI and TMRM ϩ . A, where indicated by the arrow, 0.5 M valinomycin plus 5 g/ml oligomycin was added. Note the transient plasma membrane hyperpolarization and the collapse of ⌬ m as the ionophore clamps the mitochondrial membrane potential to the now negligible K ϩ gradient across the inner mitochondrial membrane. B, addition of 0.5 M nigericin to the cells causes a partial depolarization of the plasma membrane followed by a mitochondrial hyperpolarization of 30 mV, consistent with a collapse of a pH gradient of at least 0.5 units across the inner membrane and the compensatory increase in ⌬ m . Unless otherwise stated, this and subsequent experimental traces are the mean of 10 randomly chosen somata with backgrounds subtracted. In this and subsequent figures, the experimental data points are represented as square data points and the fitted computer simulations by the continuous lines. ⌬ p (sim) and ⌬ m (sim), membrane potential time courses that generate the fitted computer simulations in A and B. and mitochondrial potentials to the biphasic increase in [Ca 2ϩ ] c ; however, because the spectra of PMPI and fluo-5F superimpose, it is necessary to perform two parallel experiments, one with PMPI and TMRM ϩ and a second with fluo-5F and TMRM ϩ (Fig. 10C). To establish the temporal relationships between the parameters, a representative cell was selected from each experiment that showed a similar delayed collapse in TMRM ϩ fluorescence. The time axes of the two experiments were adjusted to exactly synchronize the rapid collapse in the TMRM ϩ signal and establish the temporal sequence of events.
The initial effect of the ionophore addition is to raise [Ca 2ϩ ] c to a plateau. The simulation (Fig. 10D) indicates that addition of 1 M ionomycin is accompanied by a mitochondrial hyperpolarization of about 10 mV that persists through the first stage of controlled [Ca 2ϩ ] c elevation. This is similar to the effect of nigericin in collapsing ⌬pH and allowing ⌬ m to increase in compensation (Fig. 6B) and is consistent with the matrix acidification that will result from Ca 2ϩ /2H ϩ exchange across the inner membrane (32). At the same time ⌬ p starts very slowly to depolarize. For this representative cell, a rapid uncontrolled rise in [Ca 2ϩ ] c is initiated 4 min after ionomycin addition. At this point the decline in ⌬ p accelerates, but is still relatively slow. Importantly ⌬ m is maintained until [Ca 2ϩ ] c has risen to several micromolar (based on the K d of the indicator) when it suddenly collapses. This indicates importantly that it is the failure of cytoplasmic Ca 2ϩ homeostasis in the presence of ionomycin that initiates mitochondrial depolarization, rather than a collapse of mitochondrial bioenergetics causing the Ca 2ϩ deregulation.
This analysis demonstrates the two distinct modes of ionomycin action: either a modest elevation in [Ca 2ϩ ] c with maintained mitochondrial integrity, or a stochastic failure of Ca 2ϩ homeostasis and resulting mitochondrial depolarization. It is clearly essential to establish which mode is operative when interpreting experiments in which the Ca 2ϩ ionophores are added with no independent monitor of mitochondrial integrity.

DISCUSSION
Parallel Monitoring of Plasma and Mitochondrial Membrane Potentials-All techniques for monitoring changes in in situ mitochondrial membrane potentials rely on the interpretation of experiments in which membrane-permeant cations are initially allowed to accumulate across the plasma and mitochondrial membranes. The two modes that are employed for fluorescent cationic indicators, quench mode and non-quench mode, have to be carefully distinguished, because they can give opposing responses to changes in ⌬ m (Fig. 5). In quench mode (typically achieved with TMRM ϩ by equilibrating neurons with 50 -100 nM probe) aggregation and quenching occurs in the matrix with the result that mitochondrial depolarization results in an increase in whole cell fluorescence as the probe is released into the cytoplasm. Such an increase is transient, however, because the "excess" cytoplasmic probe redistributes to restore the Nernst equilibrium across the plasma membrane (Fig. 5). For this reason "quench mode" is best studied with  Note that the higher concentration of protonophore causes a biphasic depolarization of the plasma membrane. The kinetics of TMRM ϩ re-equilibration following the high FCCP addition were assumed to be limited by probe redistribution and mitochondrial depolarization was assumed not to be ratelimiting. The resulting rate constant was used for the experiment in Fig. 7A. Each experimental trace is the mean of 10 randomly chosen somata with backgrounds subtracted. The experimental data points are represented as square data points and the fitted computer simulations by the continuous lines. ⌬ p (sim) and ⌬ m (sim), membrane potential time courses that generate the fitted computer simulations in A and B.
probes such as rhodamine 123 that are relatively slowly permeant across the plasma membrane (re-equilibration between matrix and cytoplasm is always very rapid due to the enormous surface to volume ratio of the matrix). Rhodamine 123 is, however, difficult to quantify because its low membrane permeability means that cells are usually loaded by brief exposure to very high (i.e. micromolar) concentrations of probe (33). One advantage of quench mode is that cell and mitochondrial depolarizations produce opposite responses that can frequently be distinguished kinetically (8). However, quench mode cannot accurately monitor slow changes in ⌬ m when release from the matrix occurs at a comparable rate to re-equilibration across the plasma membrane (Fig. 5).
Non-quench mode (achieved, for example, with neurons equilibrated with 1-10 nM TMRM ϩ ) can follow slow changes in ⌬ m and distinguish between populations of cells with differing steady-state potentials, but cannot distinguish at whole cell resolution between a change in ⌬ m and ⌬ p (Fig. 5). The present study reveals that significant changes in ⌬ p occur with commonly used ionophores at low micromolar concentrations, such as FCCP (Fig. 8) and ionomycin (Fig. 10), whereas the fluorescence response of cationic indicators to large plasma membrane depolarizations (Figs. 2 and 3) can readily be misinterpreted as reflecting changes in ⌬ m (34).
Anionic oxonol dyes such as di-BAC 4 (3) have been used extensively to monitor changes in ⌬ p in cultured cells (e.g. Refs. 35 and 36) although until recently their utility has been limited by sensitivity and rate of re-equilibration. The availability of the proprietary FLIPR (Plasma) Membrane Potential Indicator Kit (37) incorporating an extracellular quencher has largely overcome these problems, with its large dynamic range and rapid response. The concentration of probe can be drastically reduced below that recommended to balance the emission intensity with the 5 nM TMRM ϩ used in this study with no loss of response. Calibration of the response with varied potassium media and application of the Goldman-Hodgkin-Katz equation is straightforward and the resulting traces can be used both to obtain information on ⌬ p per se and also to correct the TMRM ϩ response for changes in plasma membrane potential.
Uses and Limitations of the Simulation-In 2000 we (8) published a detailed study in which the mechanistic basis underlying the whole cell fluorescence of probes such as TMRM ϩ was investigated. It was possible to devise a simple Excel program based on three principles, probe distribution toward a Nernst equilibrium across both plasma and mitochondrial membranes, much faster equilibration across the inner mitochondrial membrane than across the plasma membrane, and probe aggregation at a critical matrix concentration. Whereas the original simulation has proven valuable for predicting and interpreting the whole cell fluorescence response of cells equilibrated with cationic probes in both quench and non-quench mode (8) the lack of information about changes in ⌬ p is a significant limitation. The present simu- lation removes many of the indeterminate parameters from the calculation. Whereas the parameters were determined for the cell bodies of CGNs, the approach is readily applicable to other cell types once values for the initial membrane potentials, mitochondrial volume fraction, and rate constants for the two probes across the plasma membrane are determined.
A number of studies have monitored changes in ⌬ m in intact cells by following the ratio of fluorescent intensity between mitochondria-rich and mitochondria-poor regions of single cells (38,39). This approach must be used with low non-quenching concentrations of probe, and although in theory this removes the dependence on the plasma membrane potential, there are considerable errors in estimating the tiny difference in fluorescent intensity between the mitochondria-poor cytoplasm and the background (see Fig. 2).
Whereas the combined use of the two indicators provides considerably more precise information than has previously been available, there are a number of limitations. First, as with all such studies, changes in matrix volume are not allowed for, although in the present study the mitochondrial swelling accompanying, for example, valinomycin addition (Fig. 6A), will not influence the interpretation because the depolarized mitochondria are depleted of TMRM ϩ . Second, high quenchmode concentrations of TMRM ϩ (ϳ50 nM) appear to interfere with the PMPI response (data not shown). Third, the large correction to the TMRM ϩ signal required with the extensive plasma membrane depolarizations caused by excitatory ionotropic receptor activation (Fig. 3) introduces a significant uncertainty in the calculated ⌬ m . Quenchmode experiments with slowly equilibrating probes remain the best way to detect step changes in ⌬ m under these special circumstances (8).
Distinguishing Plasma and Mitochondrial Membrane Potential Changes in Response to Ionophore Addition-The contrasting effects of the K ϩuniport ionophore valinomycin and the K ϩ /H ϩ exchanger nigericin on ⌬ p and ⌬ m (Fig. 6) serve to validate the methodology. In particular the transient plasma membrane hyperpolarization with valinomycin ( Fig. 6A) is consistent with an increased K ϩ conductance of the membrane, whereas the nigericin-induced mitochondrial hyperpolarization mirrors studies with isolated mitochondria (26) and is consistent with a conversion of the transmembrane pH gradient into an enhanced membrane potential. This suggests that low nigericin concentrations may be of use in investigating the role of ⌬ m in superoxide generation in intact cells as well as the effect of matrix acidification on the stability of the matrix Ca 3 PO 4 complex (40).
The present study reveals a number of novel details concerning the bioenergetic consequences of ionophore addition to intact cells. Protonophores have been extensively used to depolarize mitochondria in cells as a means of inhibiting oxidative phosphorylation or mitochondrial Ca 2ϩ transport with the implicit assumption that ⌬ p is unaffected, although it is known that extremely high protonophore concentrations can depolarize the plasma membrane (2). However, the present study reveals that even in the presence of oligomycin, when ATP production is purely glycolytic and independent of mitochondrial bioenergetic status, the low micromolar concentrations commonly employed in cell studies can also lead to extensive plasma membrane depolarization (Fig. 8B). This of course introduces considerable ambiguity into the interpretation of such studies, particularly in neurons with their plethora of voltage-gated ion channels.
Ca 2ϩ /2H ϩ antiport ionophores such as ionomycin are widely employed either to elevate [Ca 2ϩ ] c to a sustainable plateau or deliberately to induce cell death. High ionomycin concentrations induce a stochastic loss of FIGURE 10. Sequential effects of ionomycin on [Ca 2؉ ] c , ⌬ m , and ⌬ p . A, CGNs were preincubated in low-K medium in the presence of 0.5 M fluo-5F-AM (green) and 5 nM TMRM ϩ (red). After washing the cells and re-adding low-K medium with TMRM ϩ , 2 M ionomycin was added. The confocal image was taken 7 min after addition of the ionophore. Note the presence of cells that maintain a low [Ca 2ϩ ] c and retain mitochondrial TMRM ϩ (e.g. cell 1), cells that display a high [Ca 2ϩ ] c but retain TMRM ϩ (e.g. cell 2), and cells that display a high [Ca 2ϩ ] c but have lost TMRM ϩ fluorescence (e. g. cell 3). B, time interval between initiation of the rapid rise in [Ca 2ϩ ] c (a in C) and the initiation of the decrease in TMRM ϩ fluorescence (b in C) for single neuronal cell bodies from the experiment depicted in A. C, CGNs were incubated in low-K medium in the presence of PMPI and 5 nM TMRM ϩ . Where indicated 1 M ionomycin was added and the PMPI and TMRM ϩ fluorescence time courses were recorded. The experimental data points for a single representative cell body are shown. In parallel, cells loaded with 0.5 M fluo-5F-AM and 5 nM TMRM ϩ were exposed to 1 M ionomycin. A cell soma was selected that showed the same kinetics of TMRM ϩ loss as the first cell and the trace of fluo-5F fluorescence was adjusted on the x axis to synchronize the mitochondrial depolarizations. Note that Ca 2ϩ deregulation precedes the loss of TMRM ϩ . D, computer simulation of changes in ⌬ m and ⌬ p that produce the best fit (solid lines in C) with the experimental traces. Note the biphasic responses of both potentials to ionomycin addition. mitochondrial membrane potential in cultured neurons (31), but the mechanism of the depolarization and its relationship to elevated [Ca 2ϩ ] c is unclear. In addition to its action at the plasma membrane, intercalation of the ionophore into the inner mitochondrial membrane will acidify the matrix (32), destabilize any matrix Ca 3 PO 4 complex (40) and release Ca 2ϩ into the cytoplasm, induce dissipative Ca 2ϩ cycling that can uncouple the mitochondria (1), and ultimately induce a mitochondrial permeability transition (41,42). The present study clarifies the causal relationship between the elevated [Ca 2ϩ ] c and mitochondrial during this stochastic Ca 2ϩ deregulation. Prior to deregulation it is possible to detect a mitochondrial hyperpolarization consistent with the decrease in transmembrane ⌬pH induced by the ionophore (32). Importantly, the rapid increase in [Ca 2ϩ ] c occurs before the plasma membrane has depolarized sufficiently to trigger the activation of voltage-activated Ca 2ϩ channels (Fig. 10, C and D) and before the mitochondria start to depolarize (Fig. 10). Thus mitochondrial bioenergetic failure is a consequence, rather than a cause, of the cytoplasmic Ca 2ϩ deregulation.
Conclusions-The present study is an attempt to improve the precision with which changes in mitochondrial membrane potential can be monitored in intact neurons. We have recently described a "cell respirometer" (43) that allows the respiratory rates of coverslip-attached cells to be continuously monitored and allows calculation of ATP turnover, proton leak, and reserve respiratory capacity. Used in combination these novel techniques may allow in situ mitochondrial bioenergetics to be quantified with a precision approaching that for isolated mitochondria while avoiding the pitfalls in mitochondrial isolation and incubation in a non-physiological environment.