Patch-clamp recording is considered to be the reference method to quantitatively measure the plasma membrane potential Vp in human cells. A disadvantage in respect of routine use is that this method is time-consuming and laborious thus usually resulting in measuring values for only a small number of cells (1, 2). Therefore, voltage-sensitive anionic and cationic dyes, which partition between the extracellular staining buffer and the cytoplasm under the influence of Vp according to a Nernstian distribution, have been applied as an alternative (3–12).
Due to their strong accumulation in mitochondria (13, 14), cationic dyes report Vp alone only when the mitochondrial membrane potential Vm remains constant during the measurement. To ensure this condition, Vm was specifically abolished by ionophores in some studies (15–17). With increasing concentration, the monomer form of cationic dyes, among them carbocyanines as the most popular dye class, is usually more and more displaced by dimers or higher aggregates. This process takes place both with dye dissolved in aqueous solution or bound to cell compartments, and is accompanied by distinct spectral shifts in absorbance and fluorescence maxima leading to quenching of the fluorescence of the monomers (18–21). For example, DiSC3(5) monomers show red fluorescence with a maximum at about 660 nm, while dimers are nonfluorescent, and the fluorescence of higher aggregates (J-type) peaked at about 770 nm. Interestingly, based on the different spectral ranges of the fluorescence of monomers and aggregates, a ratiometric measurement by flow cytometry could be successfully performed on bacteria using the carbocyanine DiOC2(3) (22). Unfortunately, the high dye concentration of 30 μM required for this approach seems to be relatively toxic to cells (22). With other cyanines, severe cell damages were observed even at concentrations exceeding only 5 nM (5). In any case, a uniform fluorescence response over the whole voltage range can only be ensured when the quenching limit of the cationic dye used is known. Otherwise, misinterpretations of results cannot be excluded.
In contrast, bis-barbituric acid oxonols (often referred to as bis-oxonols), the preferred anionic compound class, are considered largely nontoxic (8, 23–25). Formation of aggregates presumably does not seem to occur (26), and dye uptake as a response to depolarizing treatments was found to be accompanied by an increase in cell fluorescence even when the bis-oxonol concentration was 25 μM (27). The nonspecific staining of mitochondria can additionally be avoided from the start (6, 7, 28, 29). Consequently, these dyes are particularly suitable if a steady state or slow changes in Vp are to be determined (30, 31). Though bis-oxonols have been in common use for more than two decades, however, there is still not even approximately a comparable methodological approach in respect of the staining procedure. Unlike with cationic dyes, concentration itself obviously has been so far not considered as a possible source of unfavorable effects on Vp measurements, as oxonol concentrations used ranged from 15.5 nM (32) to 25 μM (27) with flow cytometry, from 5 nM (33, 34) to 5 μM (2, 35) with microtiter plate reading and microscopic imaging. Striking differences can also be found for the indicated time of cell staining to ensure dye equilibration: with the same dye concentration of 5 μM only 10 min (35) or 60 min (2).
As a prerequisite for the determination of absolute values, a calibration curve is indispensable to establish a definite relation between fluorescence and Vp. Some authors simultaneously performed measurement with microscopic imaging and microelectrodes or patch clamping (1, 4, 36, 37). Approaches which use a series of different iso-osmotical compositions of the external buffer and addition of an appropriate ionophore (usually valinomycin or gramicidin, according to whether a cationic or an anionic dye is used) to definitely alter Vp on the basis of the Nernst potential for the extra and intracellular cations sodium and potassium are less laborious and also well founded (35, 38–44). To simplify the procedure, however, the intracellular cation concentrations were usually only estimated (35, 39, 41–43, 45, 46) rather than actually measured. In this respect, the method provided by Krasznai et al. (8) is very convenient, as intracellular cations do not need to be considered for calibration. Cells totally depolarized by paraformaldehyde (PFA) (8, 47) are stained with a series of extracellular DiBAC4(3) concentrations. After reaching equilibrium, intracellular free dye concentration then equals the respective extracellular dye concentration. As significant dye depletion in the buffer as a result of cellular dye uptake could not be observed over a wide range of cell densities and dye concentrations, a calibration curve can be established with the total cell fluorescence as a function of the known initial extracellular dye concentration (8). The intracellular free dye concentration and thus also Vp can then be calculated for any polarized cell population of the same cell type. To achieve totally depolarized cells, the calibration procedure has been modified in other studies using gramicidin (12) or methanol (48).
There are obviously several accepted approaches in respect of cell staining (dye concentration, time to reach equilibrium) and calibration (different depolarization treatments). However, influences of the respective approach itself on results have not been sufficiently investigated so far. Therefore, this study focuses on methodological aspects of Vp measurement with special interest in the reliable determination of absolute values by flow cytometry using IGR1 melanoma cells as a model and applying the anionic oxonol dye DiBAC4(3) over a wide range of concentrations. In addition, some experiments were performed on adherently growing Chinese hamster ovary CHO-K1 cells and human B lymphoblastoid LCL-HO cells growing in suspension.
MATERIALS AND METHODS
Chemicals and Incubation Buffers
Bis(1,3-dibutylbarbituric acid)trimethine oxonol, DiBAC4(3), was obtained from Molecular Probes (Invitrogen detection technologies, Eugene, OR). Gramicidin A (Fluka 50845) and methotrexate (MTX) (Sigma A-6770) were from Sigma-Aldrich (Taufkirchen, Germany), propidium iodide (Calbiochem 537059) from Merck Biosciences (Schwalbach, Germany). DiBAC4(3) was dissolved in DMSO to obtain a stock solution of 25 mM. Then, staining solutions with different concentrations were made by further dilution with DMSO (12.5 μM to 12.5 mM). Gramicidin was also dissolved in DMSO at a concentration of 2 mg/mL. Cells were usually suspended in a HEPES buffer containing 1 mM CaCl2, 1 mM MgCl2, 10 mM HEPES, 140 mM NaCl, 5.4 mM KCl, 10 mM glucose (Na-buffer) with the exception of the series with different [K+]o concentrations (5.4, 10, 20, 33, 66, 140 mM), where sodium and potassium were iso-osmotically adjusted (the sum was 145.4 mM each). In the following, the special buffer with 140 mM K+ is referred to as K-buffer. For dead cell discrimination, propidium iodide (Calbiochem 537059, Merck Biosciences, Schwalbach, Germany) was dissolved in Na-buffer at concentrations of 100, 200, 400, and 800 μg/mL.
IGR1 melanoma cells were maintained at 37°C in a humidified atmosphere containing 9% CO2 and cultured in phenol red-free Dulbecco's modified Eagle's medium (DMEM) (Gibco Life Technologies, Eggenstein, Germany) supplemented with 10% fetal calf serum. For patch clamping, cover slips 5 mm in diameter were placed in 35-mm culture dishes, and cells were seeded at a number of 4 × 104 per dish (measurement on day 3 of culture). As the flow cytometric experiments require a very high cell number, cells were usually seeded at a density of 1.2 × 106 in 75-cm2 culture flasks and prepared for measurement on day 3 of culture. To compare the resting potential between flow cytometry and patch clamping, however, cell density was reduced to 4 × 105 at plating in this special case. In this manner, a comparable confluence between cells intended for patch clamp or flow cytometry could be achieved. In addition, one experiment was done in parallel by flow cytometry and patch clamping using the same trypsinized cells for determination of the resting potential (0.4 × 106 cells plated in 75-cm2 culture flasks, grown for 3 days). In order to arrest the cells at the G1-S transition, the medium was exchanged on day 2 of culture for medium containing 1 μM MTX for at least 20 h. MTX was then removed by washing two times with PBS followed by cultivation in MTX-free medium for 8 h. As a result of changes in the distribution of cell cycle phases, MTX-synchronized cell populations possess an enhanced protein content due to the accumulation at G2/M (control vs. MTX-synchronization (proportions in %): G0/1 phase 57.1 ± 2.1 vs. 17.7 ± 3.2, S phase 30.4% ± 3.1% vs. 25.3% ± 2.3%, G2/M phase 12.1 ± 1 vs. 57 ± 5.4; significant difference for G0/1 and G2/M between control and MTX-synchronized cells, Student's t-test, three independent experiments).
Chinese hamster ovary cells CHO-K1 were cultured in phenol red-free DMEM/Ham's F12 medium (PromoCell GmbH; Heidelberg, Germany) supplemented with 10% fetal calf serum. Cells were plated in 75-cm2 culture flasks at a density of 0.75 × 106 and maintained at 37°C in a humidified atmosphere containing 5% CO2. On day 2 of culture (confluence 70–80%), cells were trypsinized and prepared for the measurements.
The human B lymphoblastoid cell line LCL-HO were cultured in phenol red-free RPMI 1640 medium (PromoCell GmbH) completed with 10% fetal calf serum and maintained as described for CHO-K1 cells. Cells were seeded in 75-cm2 culture flasks and grown as a single cell suspension. Measurements were performed at different cell densities as indicated in Results.
All cell lines were purchased from the German Collection of Microorganisms and Cell Cultures GmbH (Braunschweig, Germany).
Electrophysiological Measurement of Membrane Potential
The measurement was performed in the whole-cell configuration in current clamp mode using the amplifier Axopatch 1D (Axon Instruments, Foster City, CA). Microelectrodes were pulled from borosilicate glass and filled with the intracellular (pipette) solution containing (in mM) 145 KAsp, 10 glucose, 10 HEPES, 3 EGTA, 1 CaCl2, 5.5 MgCl2, 5 NaATP, 3 BAPTA, adjusted to pH 7.2 with KOH. The pipette resistance was in the range of 5 to 8 MΩ.
IGR1 cells on cover slips were placed in a flow chamber and perfused with the Na-buffer. For the series where sodium was iso-osmotically replaced by different potassium concentrations, a fast and reproducible exchange of solutions at the extracellular site of the cell under examination was achieved using the U-tube technique (49). Additional measurements without the U-tube were performed for cells perfused with Na- (resting potential) or K-buffer (depolarization).
For the experiment with trypsinized IGR1 cells, cells were attached onto a poly-L-lysine covered bottom of a 0.2-mL measuring chamber and continuously superfused with Na-buffer.
Laser Scanning Microscopy
LSM was carried out on a DM IRE2 (Leica, Bensheim, Germany) equipped with a 488 nm argon laser and an HCX PL APO 100x/1.40-0.7 OIL CS objective. The filter combinations used were set to the optical conditions of the FL1 channel of the flow cytometer FACScan (see below). IGR 1 cells on 42-mm cover slips were placed in a measuring chamber. The respective staining solution was then injected into the buffer followed by incubation for at least 10 min at 37°C. For the estimation of the fluorescence of free dye in the cytoplasm (see Data Analysis), cells were covered by K-buffer. Then, images for three dye concentrations per cover slip could be acquired (negligible change in cell volume for the corresponding time in K-buffer). Fluorescence values were compensated for offset. The latter was determined by the measurement of K-buffer without dye. Saturation of the analog-to-digital converter was avoided by appropriate settings of the photomultiplier voltage.
Spectrofluorimetry was performed with an F-2500 fluorescence spectrophotometer (Hitachi, Tokyo, Japan). First, a volume of 3 mL of cell-free Na-buffer or the respective cell suspension (0.5 and 1 × 106 cells/mL) was transferred into standard quartz cuvettes. DiBAC4(3) was then added by injection of an appropriate volume of staining solution followed by incubation for at least 10 min. An even distribution of dye and cells was ensured by magnetic stirring. DiBAC4(3) was excited with 488 nm (slit width 2.5 nm) to match to the laser of the FACScan flow cytometer, and fluorescence emission was scanned in the range of 500 to 700 nm (slit width 5 nm). In addition, fluorescence at 530 and 630 nm as a measure for the green and red peak was recorded by scanning the excitation wavelength [slit width 2.5 nm (excitation) or 5 nm (emission)] between 480 and 520 nm (530 nm) or 480 and 620 nm (630 nm). In the case of cell suspensions, bulk cell fluorescence was calculated by subtraction of background fluorescence from the initial measuring values. For this, an extra cuvette containing cell-free buffer with the same oxonol concentration as used for the respective cell suspension was simultaneously prepared and measured. Despite the low quantum yield of the dye in aqueous solution in principle, the green peak fluorescence of buffer ranged from 9.9% (100 nM) to 49.3% (3,250 nM) of the total fluorescence intensity (0.5 × 106 cells/mL), obviously a result of the path length of the cuvette (1 cm). Cell autofluorescence was also determined. However, it was negligible for all emission scans and the excitation scan of the fluorescence at 530 nm. So, all emission scans and the 530-nm excitation scans were only compensated for buffer fluorescence. In contrast, the initial fluorescence at 630 nm was relatively low at low dye concentrations. Therefore, these scans were compensated for buffer and cell autofluorescence.
Depending on the cell density, emission scans showed some differences. Green peak fluorescence of lower dye concentrations was higher in relation to the maximum (for either cell density at 3,200 nM), and the ratio of the maximum red to the maximum green peak height was increased at the lower cell density (1,600 nM: 5.4 vs. 3.3%, 25,000 nM: 38.4 vs. 29.9%). In particular based on the latter, data are only shown for 0.5 × 106 cells/mL.
IGR1 and CHO-K1 cells were detached from the culture dishes by short trypsinization for 2 min and then transferred into 30-mL glass test tubes. Addition of medium containing 10% FCS blocked trypsin activity. Then, when necessary after division into appropriate subpopulations, cells were usually washed two times in the respective HEPES buffer. Cells intended for Vp dissipation by fixation were washed two times in phosphate-buffered saline followed by treatment with 2% PFA (8) or 70% methanol (48) in PBS at 4°C for 1 h. After fixation, cells were washed two times in Na-buffer. Final cell density was adjusted to 4 × 105 cells/mL for all preparations. Then, a volume of 1 mL cell suspension each was transferred into FALCON tubes. The respective gramicidin and oxonol solutions were added at a volume of 1 μL. Hence, the concentration of DMSO in the cell suspensions never exceeded 0.2%. Final dye concentrations are given in Results and figure legends. Independent of the cell type used, an incubation time of at least 10 min was sufficient to reach dye equilibration in resting cells (Fig. 1A). Gramicidin treatment was performed for at least 30 min (50) with 2 μg/mL (corresponds to about 1 μM). Proved by patch clamping, this concentration was effective in depolarizing cells to zero mV (40). According to our protocol, gramicidin addition was immediately followed by oxonol injection. Simultaneous incubation with gramicidin and oxonol between 30 and 120 min led to the same calculated membrane potential (Fig. 1B). Unless indicated otherwise, measurements were therefore generally performed 40 min after exposure to gramicidin and dye. During incubation, cells were kept at 37°C (tubes covered by aluminum foil). For experiments in which data were acquired for different times of treatment, extra samples were prepared for each time so that the measurements could be performed one immediately after the other. To recognize viability in nonfixed cells, propidium iodide (PI) was added to the suspensions 5 min before analysis at a volume of 5 μL. To avoid overlapping with the oxonol fluorescence in the FL3 channel, the final concentration of PI was adapted to the respective oxonol concentration (final PI concentration 0.75 to 6 μg/mL). For the calculation of cell cycle phases, cells were processed for DNA staining with PI at a final concentration of 50 μg/mL. All independent experiments were done in triplicate for each preparation. Except trypsinization, the same procedures were applied to LCL-HO cells.
Flow cytometry was performed with a FACScan from Becton Dickinson (San Diego, CA) equipped with a 488-nm air-cooled argon laser. Oxonol fluorescence was measured in the FL1 (bandpass 530/30 nm) and FL3 channel (longpass 650 nm) (fluorescence height FL1-H or FL3-H), and PI for cell cycle analysis in the FL2 channel (bandpass 585/42 nm) (fluorescence area FL2-A). Time-resolved data were acquired in intervals of 1 s. In the case of a series of different preparations, samples were run for at least 20 s one after the other. In order to largely exclude dead cells and debris from measurement, an appropriate FSC/SSC gate was already applied during data acquisition. Under certain conditions, oxonol attachment to the tube system of the flow cytometer has to be considered (51). Measurement of cells stained with 25 nM DiBAC4(3) preceded by rinsing with 25,000 nM (the highest concentration used for cell staining in this study) led to a threefold enhancement in fluorescence for the first and still 2.4-fold for the second sample due to dye release from the tube system of the flow cytometer. Therefore, excess dye from the tubes has to be removed by careful rinsing before measurements of samples with distinctly lower oxonol concentrations can be done after samples with correspondingly higher concentrations. However, the attachment of dye onto the silicone tubes was not a problem in experiments where dye concentration was systematically increased every three samples to obtain measuring values for calibration. After replacing a sample by one with a higher dye concentration, a stable fluorescence was achieved after running for a short time.
The wide range of oxonol concentrations applied required different settings of the photomultiplier voltage to prevent the 10-bit output of the analog-to-digital converter from saturation. Therefore, we determined the respective difference in channel number with samples of the same fluorescence at definite voltage ramps. When required, measured fluorescence channel data were then recalculated after export as ASCII files (see data analysis) to values as if measured at 400 V thus allowing quantitative comparison.
Cell size was measured with a Coulter Particle Analyzer Z2 (Beckman Coulter, Fullerton, CA) as a corresponding sphere diameter followed by calculation of the cell volume.
Flow cytometric FCS data were exported as ASCII files using the freeware WinMDI 2.8 (written by Joseph Trotter). To discriminate between dead and viable cells by propidium iodide staining, appropriate gates were defined in the dotplot FL3/time additional to that based on FSC/SSC. Transformations and calculations were performed with Excel 2000 (Microsoft Corporation, USA). Curve fitting was done in Origin 7 (Origin Lab Corporation, Northampton, MA).
The following explanations are related exclusively to the green fluorescence FL1-H of DiBAC4(3) as the parameter which is used as a measure of Vp. Calibration was based on the method provided by Krasznai et al. (8) assuming that the intracellular free dye concentration Di of totally depolarized cells equals the known initial extracellular dye concentration De. When not otherwise indicated, we applied gramicidin rather than PFA-fixation to obtain Vp dissipation, a modification previously described by Newell and Schlichter (12). The graphs which represent calibration curves were created in accordance with Krasznai et al. (8) as a logarithmic function of De. By interchanging the variables, however, the respective inverse was mathematically fitted so that any Di could be calculated from the total cell fluorescence F as follows:
where xo, yo, A1, A2, b1, and b2 are constants obtained from curve fitting. A series of calibration curves was established section by section based on Eq. (1) with dye concentrations of 0/12.5/25/50/100 nM (12.5, 25, and 50 nM), 0/25/50/100/400 nM (100 nM), 0/25/50/100/400/1,600/3,200 nM (400 and 1,600 nM), and, as the fit with Eq. (1) was not satisfactory, on Eq. (2) with a series of 0/25/50/100/400/1,600/3,200/4,800/6,250 nM (3,200, 4,800, and 6,250 nM) (in parentheses the concentrations used for the ultimate measurements of the absolute membrane potential). The latter calibration curve could also be applied to Vp calculations for 400 and 1,600 nM yielding results comparable to those with Eq. (1). Vp was then calculated according to the Nernst potential for the anionic oxonol dye:
Under the specific condition that F is linearly related to the respective intracellular free dye concentration and the resulting straight line intersects with zero, the Nernst equation can be reformulated by:
where Fdep is the fluorescence of totally depolarized cells (Di = De).
Fluorescence response R (in %) was defined as
where F0 is the fluorescence of untreated cells in the resting state. The ratio of R and the corresponding change in Vp were reported as sensitivity (in %/mV).
This study includes an estimation of the contribution of the cell volume-dependent free intracellular dye to the total cell fluorescence. Calculation based on flow cytometric data will give the increase in total cell fluorescence in percent as a result of an increase only in the cytoplasmic space which doubles the cell volume:
where fK and vK are the normalized fluorescence or volume of gramicidin-treated cells after exposure to K-buffer for 90 min in percent (fluorescence and cell volume of depolarized cells in Na-buffer were set to 100%). Using confocal laser microscopic imaging, the proportion of the fluorescence of free dye ffree was calculated as follows:
where Fin is the mean fluorescence of a region of interest (ROI) defined within the range of the cytoplasm of totally depolarized cells (Di = De), Fout is the mean fluorescence of the same ROI shifted to the cell-free buffer. As images of confocal microscopy represent thin sections, it can be assumed that ROIs inside and outside the cell have the same volume.
At a concentration of 25 nM DiBAC4(3), the green fluorescence FL1-H of resting cells with a Vp of −40.5 ± 1.9 mV was about 90-fold (87.8 ± 7.9) and of gramicidin-depolarized cells about 400-fold (397.3 ± 42.9) the autofluorescence (means of six independent experiments). For CHO-K1 and LCL-HO cells, the following values for the ratio of FL1-H and autofluorescence were calculated based on −40.5 mV and 25 nM as well (six single values of two independent experiments each): CHO-K1 resting cells 59.5 ± 2.4, depolarized cells 271.4 ± 28.8; LCL-HO resting cells 38.4 ± 3.3, depolarized cells 174.3 ± 10.3. So, it was not required to compensate the measuring values for autofluorescence even at the lowest dyeconcentration used (IGR1 12.5 nM, CHO-K1 and LCL-HO 25 nM).
Statistical analysis was performed using SigmaStat software (Jandel Scientific, Erkrath, Germany). A P < 0.05 was considered to be statistically significant.
Spectrofluorimetry of DiBAC4(3) in Cell-Free Buffer and Bound to IGR1 Cells
Fluorescence of DiBAC4(3) dissolved in cell-free Na-buffer showed a unimodal distribution up to 50 μM with concentration-dependent quenching above 6.25 μM (Figs. 2A and 2B). Additional measurements with concentrations from 5 to 10 μM by steps of 1 μM revealed that fluorescence intensity peaked at 9 μM (data not shown). Oxonol-stained cells emitted a fluorescence spectrum with a green peak at around 520 nm and, visible only at 1.6 μM and above, a red peak at about 620 nm (Figs. 2C and 2D). In contrast to flow cytometry, where cell fluorescence in resting cells continuously increased up to 25,000 nM, the green fluorescence of DiBAC4(3) showed a maximum at 3,200 nM (Fig. 2C). This suggests that spectrofluorimetric measurement implies an inner filter effect. In addition to intensity changes, concentration-dependent spectral shifts of the peak maxima could be observed. The overall shift was 8 nm for dye in buffer (516 to 524 nm) or 10 nm for cell suspensions (517 to 527 nm) (100 to 25,000 nM).
Excitation scans were performed for the fluorescence at 530 (data not shown) and 630 nm (Fig. 3). In the case of 530 nm, excitation maxima were found between 497 and 511 nm for dye in buffer or 500 to 511 nm for cell suspensions (100 to 25,000 nM). The fluorescence at 630 nm was optimally excited at 590 nm (dye in buffer) or 605 nm (DiBAC4(3)-stained cells) (Figs. 3A and 3B). In contrast to dye in buffer, a second effective excitation maximum of the red peak could be observed in the green wavelength range in cell suspensions. A distinct shift of 13 nm occurred only for this green excitation in cells (500 to 513 nm, 400 to 25,000 nM) (Fig. 3B).
Influence of Cell Volume and Extracellular Potassium Concentration on Calculated Membrane Potential
Exclusive of the incubation with gramicidin and for a short time with K-buffer, depolarizing treatments led to significant changes in cell volume of IGR1 cells (Table 1). The cell volume of PFA-fixed LCL-HO cells was increased by 25.1% ± 8.5% (volume of control: 0.62 ± 0.03 pL). This results in differences in the amount of free dye in the cytoplasm even if cells have the same membrane potential. However, IGR1 cells stained with 25, 1,600, 4,800, and 25,000 nM DiBAC4(3) and depolarized by gramicidin or K-buffer supplemented with gramicidin (incubation in K-buffer for 90 min) did not show significant differences in fluorescence (data not shown). Despite the considerable increase in cell volume, the calculated membrane potential of cells in K-buffer was significantly higher than that of the zero-mV-control in Na-buffer only at the high oxonol concentration of 4,800 nM (Fig. 4A). K-buffer alone led time-dependently to slightly repolarized Vp values both for 25 and 1,600 nM (Fig. 4B). The data presented in Table 2 represent an estimate of the increase in total cell fluorescence upon an increase in cell volume of 100% (FCM data) or the proportion of the fluorescence of free dye as percentages of the total cell fluorescence (confocal laser microscopy).
Table 1. Change in cell volume caused by different depolarizing treatments of IGR1 cells on day 3 of culture (calculated from Coulter particle size data)
Number of single measurements (in parentheses: number of independent experiments, n = 3 each).
One-way ANOVA using Bonferroni's t-test as a post hoc analysis (cell volume of untreated cells as a control remained unchanged between 30 and 120 min after preparation for flow cytometry, mean cell volume: 2.76 ± 0.13 pL).
Gramicidin (30 min)
−1.96 ± 1.6
Gramicidin (60 min)
−1.48 ± 1.7
Gramicidin (90 min)
−2.34 ± 2.2
K-buffer (30 min)
3.1 ± 10.5
K-buffer (60 min)
26.55 ± 4.2
P < 0.05
K-buffer (90 min)
42.35 ± 5.5
P < 0.05
K-buffer (120 min)
71.84 ± 16.4
P < 0.05
K-buffer (90 min)/gramicidin
77.1 ± 6.6
P < 0.05
30.41 ± 9.9
P < 0.05
−42.35 ± 1.3
P < 0.05
Table 2. Estimates of the contribution of the fluorescence of free dye dissolved in the cytoplasm to the total cell fluorescence in IGR1 cells
Data reflect the increase in total cell fluorescence in percent as a result of an increase only in the cytoplasmic space which doubles the cell volume (means of six single measurements of two independent experiments).
Data represent the fluorescence proportion of free dye as percentages of total cell fluorescence (n = 3 cells each).
0.36 ± 0.79
0.12 ± 0.04
0.32 ± 0.14
1.26 ± 1.98
1.32 ± 0.31
1.72 ± 0.28
3.72 ± 1.62
1.84 ± 0.25
1.59 ± 0.24
2.72 ± 1.86
Figure 5A shows the membrane potential of IGR1 cells on day 3 of culture as a function of the external potassium concentration. Calibration for flow cytometry was based on gramicidin-depolarized cells. Independent of the respective method, a logarithmic relationship between the potassium concentration of the buffer and Vp could be observed. At 140 mM potassium (considering the time-dependent slight repolarization of cells in K-buffer described earlier, cells were suspended for a maximum of 40 min in K-buffer), values close to zero were determined both with flow cytometry and patch clamping [25 nM: 2.8 ± 1.4, 400 nM: 1.9 ± 1.4 mV, 1,600 nM: 2.4 ± 0.7 mV; patch clamping: 1.49 ± 1.47 mV (n = 10 cells) or −2.46 ± 2.58 mV (n = 11 cells)]. The differences between resting Vp values obtained from patch clamping (−52.99 ± 13.03 mV, n = 27 cells) and flow cytometry (25 nM: −40.2 ± 0.7 mV, 400 nM: −40.1 ± 0.6 mV, 1,600 nM: −41.3 ± 1.2 mV) were obviously a result of different culture conditions (Table 3). The high number of cells required for the measurement of potassium series with flow cytometry could only be ensured with acceptable time and effort by using a higher cell density already at plating (1.2 × 106 per 75 cm2 flask) compared to that of cells intended for patch clamping (4 × 104 per 35-mm dish). Using the same trypsinized IGR1 cell population (plated at a density of 0.4 × 106 cells, trypsinized on day 3 of culture), however, comparable Vp values of −49.4 ± 1.3 mV (flow cytometry) or −46.5 ± 15 mV (patch clamping, n = 6) were measured. Additional experiments confined to the determination of the resting potential showed that Vp values measured with the two methods at the same cell confluence (30–40%) were also comparable when cells were independently grown [three independent experiments by flow cytometry: −47.5 ± 0.96 mV; two additional measurements by patch clamping: −47.67 ± 13.6 mV (n = 15 cells) and −43.9 ± 9.1 mV (n = 12 cells)]. Cell cycle analysis of these cells intended for patch clamping or flow cytometry verified a significant difference only for the G2/M phase (14.6% ± 2.4% vs. 9.86% ± 2%, t-test).
Table 3. Membrane potential of IGR1 cells at different culture confluence and cell cycle state calculated from flow cytometric results (oxonol concentration 25 nM)
CELL NUMBER AT PLATING
DAY OF CULTURE AT MEASUREMENT
ESTIMATED CELL CONFLUENCE BEFORE CELL DETACHMENT (%)
Membrane potential values and cell cycle phase proportions represent means ± SD of one experiment (n = 3).
The number of independent experiments with similar results (n = 3 each) are given in parentheses. Significance could only be found between the lowest and highest value (−38.2 ± 1.15 mV and −43.24 ± 1.33 mV) of the series with 70–80% confluence (comparison of results with the same confluence using one-way ANOVA, all pairwise tested, Student-Newman-Keuls as a post hoc test).
At low dye concentrations (25 nM), cell fluorescence depends exponentially on Vp. However, this function turned into a straight line with increasing external dye concentration (Fig. 5B).
DiBAC4(3) Staining Possesses a Quenching Limit of Green Fluorescence
The calibration curve obtained from gramicidin-depolarized IGR1 cells (Figs. 6A and 6B) revealed a green fluorescence maximum at 6,250 nM DiBAC4(3) (three independent experiments) or 4,800 nM (one independent experiment, t-test: not significantly different from the fluorescence with 6,250 nM). In contrast, the fluorescence of untreated control cells increased continuously up to 25,000 nM (three independent experiments). Except fluorescence intensity, the same results were obtained using LCL-HO at a density of 0.5 × 106 cells/mL (Figs. 6E and 6F) and, only performed for gramicidin-depolarized cells, CHO-K1 cells (data not shown). Unexpectedly, there obviously exists a quenching limit for the DiBAC4(3) fluorescence of depolarized cells. So the calculation of membrane potential (Figs. 6C and 6G) and sensitivity (Figs. 6D and 6H) was only performed for values obtained from cells with a maximum dye concentration of 6,250 nM.
Membrane potential calculation based on appropriate exponential fits of the calibration curve (referred to as exponential) yielded the same value independent of dye concentration in IGR1 cells (Fig. 6C) (mean of all concentrations: −38.3 ± 1.2 mV). In contrast, this applied to LCL-HO cells only up to 1,600 nM (mean Vp of 25 to 1,600 nM −45.1 ± 1.6 mV). At higher concentrations, a significant hyperpolarization could be observed leading to Vp values of −53 ± 2.2 mV (3,200 nM) or −59.7 ± 1.8 mV (6,250 nM) (Fig. 6G). Because of the virtually linear relation between fluorescence and dye concentration for both depolarized and resting cells (Figs. 6A and 6B), a reliable Vp calculation using Eq. (4) was possible up to 100 nM (marked as linear). This result was based on at least two independent experiments for all cell lines (IGR-1: 12.5, 25, 50, and 100 nM; CHO-K1 and LCL-HO: 25, 50, and 100 nM). Using IGR1 cells, a third independent experiment led to a significant difference of 4.5 mV at 100 nM to the calculated membrane potential based on exponential fitting and Eq. (3) (−42.6 ± 1.3 mV). The resting potential of CHO-K1 cells was −37.6 ± 0.8 mV and −35.1 ± 0.7 mV.
Calculated Membrane Potential as a Result of Gramicidin- and Fixation-Based Calibration
Similar to gramicidin-depolarized cells with a higher protein content (MTX-synchronized), fixed cells emit a higher green fluorescence FL1 (Figs. 7A and 7C) than gramicidin-depolarized cells in spite of having the same protein content. Interestingly, the red fluorescence FL3 (Figs. 7B and 7D) is correspondingly reduced. Figure 8 shows that fixation-based calibration may lead to incorrect membrane potential calculations for unfixed cells. The extent of this deviation from the actual membrane potential is dependent on the fixative, the external dye concentration and the actual membrane potential. For this experiment, LCL-HO cells were used at a density of 0.8 × 106/mL (mean Vp of 50 to 1,600 nM: −50.1 ± 1.8 mV). The resting potential of IGR1 in Figure 8A was −40.7 ± 0.8 mV (mean of 25 to 1,600 nM).
The cellular bis-oxonol distribution consists of free dye dissolved in the aqueous phase of the cytoplasm and of bound dye forming complexes with cell components such as proteins. Therefore, it has been generally assumed that there is a nonpotential associated fluorescence, which varies with changes in cell volume (8, 52–54), and thus the amount of intracellular free dye, and changes in protein content or the number of dye binding sites (30). Although it is known from microscopic observation that the quantum yield of DiBAC4(3) in an aqueous solution is relatively low compared to that of dye bound to hydrophobic cellular binding sites (4), a quantitative assessment of this nonspecific fluorescence proportion in whole single cells is still lacking. So far, only spectrofluorimetric data showing an increase in fluorescence of dye bound to cell compartments have been provided. For example, Rink et al. (55) found an increase in quantum efficiency of up to 20-fold when DiSBAC2(3) binds to membranes or proteins. Another study (31) found a two to threefold enhancement of fluorescence upon DiBAC4(3) binding to BSA as a model for proteins with hydrophobic binding sites. Using flow cytometry, this study demonstrates for the first time that differences in cell volume do not have relevant effects on total cell fluorescence (Table 2). Compared to gramicidin-depolarized cells in Na-buffer, an increase in cell volume of gramicidin-treated cells in K-buffer by 77% (Table 1) led to a maximum difference in the calculated membrane potential of only 2.87 ± 1.3 mV (4,800 nM, Fig. 4A). In contrast to cell volume, changes in protein content cannot be considered negligible in respect of nonspecific variations in total cell fluorescence. As expected, enhancement of protein content in MTX-synchronized, gramicidin-depolarized IGR1 cells was accompanied by an increase in fluorescence compared to the gramicidin-treated control (Fig. 7A).
It is known that the cell-associated green fluorescence of DiBAC4(3) tends toward saturation with increasing dye uptake (8, 31). The results of this study demonstrate, to our knowledge for the first time, that even quenching occurs at very high concentrations in depolarized cells (Figs. 6A and 6E). By applying a dye concentration of 25 μM (27), changes in membrane potential over the whole voltage range may therefore easily be misinterpreted. It can be ruled out that an interaction between dye and gramicidin leading to impaired gramicidin channel conductance caused this quenching, as there was no significant difference between the fluorescence of gramicidin- and K-buffer-depolarized cells at 25 μM (data not shown). However, an additional red fluorescent peak was found with increasing dye concentration and thus obviously associated with saturation and quenching (Figs. 2C and 2D). By analogy to the cationic cyanine DiOC2(3) (22), it seems reasonable to assume that aggregate formation of dye bound to cell components occurred at correspondingly high dye concentrations and gave rise to this peak. An increase in the number of aggregates associated with a resultant decrease in that of monomers would then be accompanied by quenching of the green peak. The effect of at least two processes, saturation and highly probable aggregation, may also be an explanation for the finding that the calibration curves could not be satisfactory fitted by Eq. (1) but required an exponential function of second order [Eq. (2)] at dye concentrations above 3,200 nM. Recordings of the fluorescence intensity of IGR1 cells at 630 nm as a function of the excitation wavelength and dye concentration revealed that the red peak presented two excitation maxima at 500–520 and 605 nm (Fig. 3B). So, if aggregates are localized at a correspondingly close distance to monomers, additional quenching would be caused by FRET from monomers to aggregates; a process which has also been suggested to occur in relation to the excitation of J-type aggregates of the cationic cyanine dye DiSC3(5) bound to DNA (20). In contrast to cell-associated fluorescence, the red peak of dye in buffer exhibits only one excitation maximum at 590 nm (Fig. 3A). This explains the lack of the red peak in the emission scans of dye in buffer excited with 488 nm. In addition, the results of spectrofluorimetry reveal considerable shifts in fluorescence and excitation maxima in the green wavelength range. In flow cytometry with fixed optical channels and laser excitation at 488 nm, slight wavelength shifts, even without change in intensity, may result in distinct changes in the fluorescence values obtained by this method. Therefore, measuring values of oxonol fluorescence are a result of complex processes and depend on binding, spectral, and quenching behavior. Because of the high differences in DiBAC4(3) concentration higher than 3,200 nM (Fig. 6), the exact quenching limit could not be specified. Still, the overall concordance between the results obtained from three different cell lines for both linearity at low dye concentrations and quenching was unexpected. A possible explanation could be that the cell lines used exhibit similar protein or binding site concentrations.
Serious objections have been raised to the use of fixed cells for the construction of calibration curves (22, 24). The results of this study support this view. A flow cytometric experiment suggests that, unlike gramicidin, PFA-fixation leads not only to the depolarization of the cell membrane but also of cell organelles such as mitochondria. As compared to resting IGR1 cells, the fluorescence of the cationic styryl dye LDS-751 (25 nM) that accumulates in mitochondria was decreased by 30.4% ± 4.7% for gramicidin-treated cells and still by 73.8% ± 0.8% for PFA-fixed cells (data not shown). It can be assumed that this also applies to methanol. This additional depolarization may facilitate the uptake of dye in cell organelles and should thus contribute to an increase in nonspecific fluorescence. In fact, both methanol and paraformaldehyde made more dye binding sites accessible or at least considerably increased quantum efficiency of the green fluorescence compared to gramicidin-depolarized cells (Fig. 7). In the case of paraformaldehyde, this could be additionally based on its ability to greatly alter the charge of lipid membranes and, although to a lesser degree, of proteins (56). Methanol causes obviously even more pronounced changes in proteins and lipids than PFA. This can also be concluded from measurements of Raman vibrations (57), which support the observation of this study. Consequently, significant deviations of the calculated from the actual membrane potential of unfixed cells can occur, the extent of which proved to be dependent on the fixative, the staining conditions, and the actual membrane potential (Fig. 8). In contrast, gramicidin treatment not only dissipates the membrane potential to zero mV (40, 58) but obviously leaves the dye binding characteristics of cells unchanged. This can be concluded from the results of this study for cells in K-buffer using gramicidin-based calibration. Membrane potential obtained from patch clamping and flow cytometry was virtually identical (Fig. 5A) and consistent with that of other studies using patch clamping (59, 60). Even the slight repolarization after prolonged incubation in K-buffer (Fig. 4B) is supported by previous results (7, 38). In the last analysis, the calibration method provided by Krasznai et al. (8) remains well-founded in principle and is presumably generally valid when based on unfixed cells with known membrane potential, for example gramicidin-depolarized cells (12). In addition to PFA-fixed cells, another study used viable cells exposed to a buffer containing 20 mM potassium and specified the resulting curve drawn from fluorescence and extracellular dye concentration as a calibration curve (61). According to our logic, however, this term should be reserved to curves which make the calculation of absolute Vp values for a corresponding sample of the same cell type with any Vp possible. So, it seems to be most convenient to establish the calibration curve with totally depolarized cells, as their membrane potential is definitely known without the need of any other measurement. This is exactly the advantage of the method suggested by Krasznai et al. (8).
The mean membrane potential of a whole cell population is not a constant value but dependent on the distribution of the cell cycle phases (62, 63) (Table 3). It is known that there are cell-type specific differences in polarization of cells escaped from cycling by entering the G0 phase. Vp can then be more negative or else more positive compared to normally growing cells (64). For example, the pronounced increase in G0/G1 of IGR1 cells grown to confluence was associated with a more depolarized Vp (Table 3, day 5), while a nonadherent B cell line showed a hyperpolarized Vp at high cell density and G1 arrest (16). Confirming earlier results obtained from patch clamping (63), however, a comparable depolarization of IGR1 cells also occurred at very low confluence associated with high proliferation (Table 3, day 2, 20% confluence). This could be attributed to the increase in S phase leading to a drop in Vp (63). In contrast, the hyperpolarized state of cells at low confluence on day 3 of culture (Table 2) could be a result of an increase predominantly in the number of cells progressing through G1 (62). In respect of Vp measurement using charged dyes, it must therefore be taken into account that changes in the cell cycle can often be accompanied by changes in the mean protein content. Consequently, calibration should be repeated for every new experiment, if identical cycling states cannot be ensured for following measurements. Thus, an approach which simplifies the calibration procedure would be extremely useful. Two microscopic imaging studies (36, 65) assumed rather than proved linearity between dye concentration and fluorescence and replaced the concentrations Di and De in the Nernst equation by the fluorescence of inside (Fin) and outside (Fout) the cell. The term Fout, however, does not correspond to the fluorescence actually measured and should correctly be Fdep (the fluorescence inside the cell at a free dye concentration in the cytosol which equals the concentration of the external buffer). Because of the low number of cells (n = 5) and only three different K+ concentrations (36) or the measurement of totally depolarized cells with only one dye concentration for calibration (65), the validity of this approach still needs to be confirmed by a method with a high statistical certainty. This study could prove by flow cytometry that there actually exists a concentration range [upper limit about 100 nM DiBAC4(3)], where the relation between total cell fluorescence and intracellular free dye concentration is virtually linear and the resulting straight line intersects with zero (Figs. 6A and 6E). So calibration can be confined to the determination of the fluorescence of only a totally depolarized cell population stained with the same dye concentration as intended for the actual measurement, and Vp calculation can be reliably performed according to Eq. (4). In contrast, the linear relation between total cell fluorescence and membrane potential at considerably higher concentrations (4, 31, 37, 66) (Fig. 5B) is not a substitute for calibration. If the resting potential is unknown and a semiquantitative calibration was performed by the addition of KCl aliquots (31), only changes in Vp can be calculated on a millivolt scale (30, 31). Based on linearity, however, calibration could also be simplified as two definite K+ concentrations, which already satisfy the requirements of this approach. The dependence of the two different linear relationships on dye concentration is just what may be expected from fundamental principles of the fluorescence of Nernstian dyes. There is strong evidence that the exponential relation between dye concentration and fluorescence can be considered linear for sufficiently small concentrations far from binding saturation (67). At dye concentrations where this relation can actually only be expressed by an exponential function of first order [Di is proportional to exp(a × F)], membrane potential must be directly proportional to the fluorescence of a Nernstian dye [Di is also proportional to exp(b × Vp)]. The sign of the slope depends on the charge of the dye and the mode of action (quench or nonquench mode). It is clear that the concentration range which fulfills the respective condition depends on the specific dye used and the object under study. Cells stained with the cationic dyes TMRM (14) and DiSC3(3) (67) showed a linear relation of fluorescence and dye concentration between 20 and 100 nM (TMRM) or below 50 nM [DiSC3(3)]. Using rhodamine 6G at 600 nM (10) and DiSC3(5) at 1 μM (68), Vp and F have been found to be directly proportional. This study verified DiBAC4(3) ranges between 25 and 100 nM (F linearly related to D) and, predicted from measurement and exponential curve fitting, between about 1,600 nM and 3,200 nM (F linearly related to Vp; actually measured at 400 and 1,600 nM, Fig. 5B).
For reliable measurements, the time of incubation with DiBAC4(3) must primarily ensure dye equilibration. Resting cells used in this study incorporated most dye within a few minutes, and incubation for at least 10 min was sufficient to achieve a long-lasting stable fluorescence (Fig. 1). In other studies, incubation times between 2 min (69) and 60 min (47) have been indicated. Dye concentrations also differ considerably. DiBAC4(3) have often been applied between 1 and 5 μM (31, 66, 70, 71) or even at 10 μM and above (27, 72, 73) without giving any reasons for using such high concentrations. In contrast, 100 nM DiBAC4(3) (6, 54) or below (32–34, 39, 41) also seems to be advantageous for reasons other than calibration. For example, some oxonols including DiBAC4(3) have been proven to influence anion transport at concentrations above 1 μM (74–76). In addition, BK channels widely expressed in human and animal cells (77–83) can even be activated at DiBAC4(3) concentrations lower than 100 nM (84). Therefore, dye-mediated side effects cannot be precluded from the start. Consequently, staining with adequate dye concentrations may be indispensable in affected cell systems. The results of this study suggest that this does not apply to IGR1 cells but to LCL-HO cells. While DiBAC4(3) fluorescence reliably reported the membrane potential in IGR1 cells as long as the dye concentration was below the quenching limit (Fig. 6C), LCL-HO cells responded with a hyperpolarization of up to 15 mV to DiBAC4(3) concentrations higher than 1,600 nM (Fig. 6G). This seems to be consistent with the activation of an ion channel leading to K+ efflux which is not expressed on the membrane of IGR1 cells. Nevertheless, another aspect of using low dye concentrations is of importance in IGR1 cells too. Sensitivity is clearly a function of external dye concentration, although 1 %/mV, as determined by Bräuner et al. (4) with 2 μM, has been cited as an apparently typical characteristic of DiBAC4(3) (see “The Handbook,” Molecular Probes). Based on the fluorescence response from resting potential to zero, for IGR1 cells the same value of about 8%/mV was determined between 12.5 and 100 nM (a result of the linear relation between dye concentration and fluorescence in this range) followed by a continuous negative slope resulting in 2.9%/mV at 1,600 nM and 1.3%/mV at 6,250 nM (Fig. 6D). Other flow cytometric studies found about 10%/mV with 100 nM (6) or 3 %/mV with 1.6 μM (61). Thus, a comparison of this parameter between different dyes including DiBAC4(3) at the high concentration of 5 μM (31) cannot characterize the behavior of the bis-oxonol in principle.
In summary, fixation-based calibration procedures proved to be applicable over the whole Vp range only at low dye concentrations. In contrast, gramicidin-based calibration led to reliable Vp determinations largely independent of Vp and dye concentration. As a response to total depolarization, DiBAC4(3) concentrations higher than 6.25 μM cause fluorescence quenching in IGR1, CHO-K1, and LCL-HO cells. Independent of dye concentration, nonpotential related fluorescence variations are only based on changes in the amount of cellular bound dye, whereas the contribution of the free dye dissolved in the aqueous phase of the cytoplasm to total cell fluorescence is negligible. DiBAC4(3) concentrations between 50 and 100 nM seem to be optimal in order to simplify calibration, to ensure high sensitivity, and to avoid unwanted side effects.
The authors thank K. Hölsken, U. Schramm, and C. Bruhne for their excellent technical assistance.