Dynamic regulation of the mitochondrial proton gradient during cytosolic calcium elevations.

Mitochondria extrude protons across their inner membrane to generate the mitochondrial membrane potential (ΔΨ(m)) and pH gradient (ΔpH(m)) that both power ATP synthesis. Mitochondrial uptake and efflux of many ions and metabolites are driven exclusively by ΔpH(m), whose in situ regulation is poorly characterized. Here, we report the first dynamic measurements of ΔpH(m) in living cells, using a mitochondrially targeted, pH-sensitive YFP (SypHer) combined with a cytosolic pH indicator (5-(and 6)-carboxy-SNARF-1). The resting matrix pH (∼7.6) and ΔpH(m) (∼0.45) of HeLa cells at 37 °C were lower than previously reported. Unexpectedly, mitochondrial pH and ΔpH(m) decreased during cytosolic Ca(2+) elevations. The drop in matrix pH was due to cytosolic acid generated by plasma membrane Ca(2+)-ATPases and transmitted to mitochondria by P(i)/H(+) symport and K(+)/H(+) exchange, whereas the decrease in ΔpH(m) reflected the low H(+)-buffering power of mitochondria (∼5 mm, pH 7.8) compared with the cytosol (∼20 mm, pH 7.4). Upon agonist washout and restoration of cytosolic Ca(2+) and pH, mitochondria alkalinized and ΔpH(m) increased. In permeabilized cells, a decrease in bath pH from 7.4 to 7.2 rapidly decreased mitochondrial pH, whereas the addition of 10 μm Ca(2+) caused a delayed and smaller alkalinization. These findings indicate that the mitochondrial matrix pH and ΔpH(m) are regulated by opposing Ca(2+)-dependent processes of stimulated mitochondrial respiration and cytosolic acidification.

Introduction

Mitochondria are mobile intracellular integrators of metabolic and ionic signaling. These multifunctional organelles generate ATP by oxidative phosphorylation, integrate signaling cascades leading to the release of proapoptotic factors (1), and shape cellular Ca2+ signals by taking up and releasing Ca2+ ions (2). By acting as transient Ca2+ buffers, mitochondria alter the propagation of Ca2+ waves, modulate the activity of plasma membrane channels and transporters (3, 4), and facilitate the refilling of intracellular Ca2+ stores (5, 6) (reviewed in Ref. 7). Mitochondrial metabolism and intracellular Ca2+ signaling are closely linked. Increases in the free Ca2+ concentration in the mitochondrial matrix ([Ca2+]mit) activate mitochondrial dehydrogenases to stimulate oxidative phosphorylation (8), and Ca2+ elevations in the intermembrane space stimulate the uptake of substrates of oxidative phosphorylation (9, 10). Moreover, electrophoretic Ca2+ uptake into the matrix is thought to reduce the electrical resistance for pumping protons out of the matrix (11).

Mitochondria are double membrane organelles with an outer membrane permeable to solutes and an inner membrane harboring the respiratory chain complexes. The respiratory chain extrudes protons from the mitochondrial matrix into the intermembrane space to generate a maximal proton-motive force (Δμ̃H) of ∼220 mV across the inner mitochondrial membrane (12). The Δμ̃H is the sum of the electrical potential across the inner membrane (ΔΨ, negative inside) and the pH gradient between the bulk solutions on either side of the membrane (ΔpH, alkaline inside). According to Mitchell's chemiosmotic theory, ΔΨ and ΔpH are thermodynamically equivalent driving forces for the synthesis of ATP by F1F0 ATP-synthase (13) (reviewed in Ref. 14). Recent studies, however, indicate that ATP synthesis by purified F0 complexes reconstituted in liposomes requires a high concentration of protons (pH <6.5) at the source P site of the enzyme (15) (reviewed in Ref. 16). Efficient ATP synthesis thus require not only a high driving force for protons but also a low pH within the cristae of mitochondria. These data imply that a ΔpH of >1 pH unit must be maintained to enable the synthesis of ATP by respiring mitochondria.

The mitochondrial pH gradient (ΔpH) is the sole driving force for the electroneutral transport of many ions and metabolites in and out of the mitochondrial matrix, whereas the activity of several mitochondrial ion exchangers is coupled, directly or indirectly, to the electroneutral movement of protons (17). For instance, the mitochondrial Na+ gradient is clamped close to ΔpH by the electro-neutral mitochondrial 1Na+:1H+ exchanger (mNHE) (reviewed in Ref. 7). The sodium gradient, in turn, drives electrogenic mitochondrial 1Ca+:3Na+ exchange (18) by the recently identified protein NCLX (19) that regulates mitochondrial Ca2+ levels. Electroneutral K+/H+ exchange catalyzed by the protein Letm1 is essential for mitochondrial ionic and volume homeostasis (20, 21). The Letm1 protein was recently proposed to be an electrogenic Ca2+/H+ antiporter with a 1:1 stoichiometry (22), despite earlier studies indicating that Ca2+ enters mitochondria as the fully charged species (23–25) and exits mitochondria with a 3H+:1Ca2+ stoichiometry (26). Regardless of whether Letm1 transports K+ or Ca2+ in exchange for H+, its exchange activity depends on ΔpH. Finally, ΔpH also influences the amplitude of [Ca2+]mit elevations in that the mitochondrial Ca2+-buffering power is determined by the mitochondrial phosphate concentration, which is modulated by ΔpH (27). Mitochondria therefore rely on ΔpH to generate ATP, to move ions and metabolites, and to buffer Ca2+ ions.

The regulation of Δμ̃H has been extensively studied because it is essential for oxidative phosphorylation. Early approaches to estimate Δμ̃H in isolated mitochondria used potassium- and proton-selective electrodes or isotopes to detect changes in the external concentration of these ions in the presence of the K+ ionophore valinomycin to provide a diffusion potential (12, 28). These experiments revealed that at low external K+ concentrations, ΔpH contributes up to 170 mV to the proton-motive force (12). In physiological conditions, however (i.e. high K+ concentrations and no valinomycin), the situation is reversed, and ΔΨ contributes most of the proton-motive force. These findings were validated by subsequent determinations of ΔΨ in isolated mitochondria and intact cells with fluorescent lipophilic cations (reviewed in Ref. 29). Our current knowledge of ΔpH regulation is largely based on early experiments in isolated mitochondria that employed minimal sucrose buffers and H+/K+ ionophores, such as nigericin (30, 31). In intact cells, ΔpH has been estimated around 1.0–1.2 pH units by isotopic measurements of weak acid or bases (32, 33), thus contributing ∼60 mV to Δμ̃H. More recent studies using fluorescent indicators yielded ΔpH values of 0.5–0.9 pH units based on separate static measurements of matrix, cytosolic or inter-membrane space pH in sister cultures of live cells (34). These studies brought us closer to direct measurement of ΔpH in live cells but were unable to resolve the dynamic regulation of ΔpH by factors like Ca2+ uptake, the activity of the mitochondrial permeability transition pore (35), and the presence of uncoupling proteins (36).

In this study, we simultaneously measured cytosolic pH using the well characterized dye 5-(and 6)-carboxy-SNARF-1 (SNARF) and matrix pH using a ratiometric circularly permuted YFP. The concurrent measurements of pH in the cytosol (pHcyto) and mitochondrial matrix (pHmito) provided real-time measurement of the cytosol-matrix pH gradient, and thus of ΔpH, in intact cells. Moreover, this approach enabled us to determine the proton-buffering power of the matrix of intact mitochondria. Unexpectedly, we observed that [Ca2+]cyt elevations were associated with massive cytosolic and matrix pH decreases and consequent decreases in ΔpH.

EXPERIMENTAL PROCEDURES

Reagents

Minimum essential medium with Glutamax (catalog no. 41090), fetal calf serum, penicillin/streptomycin, Lipofectamine 2000, SNARF/AM (C1272), and Fura-2 (F1201) were from Invitrogen. Histamine, 2-APB, orthovanadate, nigericin, monensin, antimycin, oligomycin, and rotenone were from Sigma. Transfast was from Promega. The QuikChange II site-directed mutagenesis kit was from Stratagene. Mutagenesis primers were from Microsynth (Balgach, Switzerland). HyPer plasmids were from Evrogen (Moscow, Russia). YC3.6cyto and 4mitD3-CPV were kindly provided by Drs. Amy Palmer and Roger Tsien.

Cell Culture and Transfection

Culturing of HeLa cells was described previously (37). For epifluorescence microscopy, cells were plated on 25-mm glass coverslips in 35-mm culture dishes. For high throughput experiments, 5000–8000 cells/well were plated on clear bottom, black-walled 96-well plates (catalog no. 655090, Greiner Bio One). Coverslips were transfected with 2 μg of DNA and Lipofectamine 2000 (4 μl) or Transfast (3.5 μl) as per the manufacturer's instructions. Cells on 96-well plates were treated with 1.1 μl of Lipofectamine 2000 and 0.4 μg of DNA in 200 μl of medium/well for 6–24 h and imaged 2 days later.

Cell Permeabilization

MitoSypHer-expressing HeLa cells were permeabilized on the microscope with a 1-min exposure to digitonin (100 μm) in Ca2+-free intracellular buffer: 120 mm KCl, 10 mm NaCl, 1 mm H2KPO4, 20 mm HEPES, 5 mm succinic acid, 1 mm ATP-Mg2+, 0.02 mm ADP-K, 1 mm MgCl2 (1 mm Mg2+-free), 0.5 mm EGTA adjusted to pH 7.4 with KOH. After digitonin washout, pH was stepped between 7.4 and 7.2 by turnover of the bath solution. The addition of Ca2+ or drugs was performed by full bath turnover. Total CaCl2 and MgCl2 was 0.327 and 5.5 mm, respectively, at pH 7.4, and 0.326–5.43 mm at pH 7.2 to give ∼10 μm free Ca2+ buffered with 5 mm HEDTA and 1 mm ATP-Mg2+ as calculated with Max Chelator.

Conversion of HyPer to SypHer

In their description of HyPer, Belousov et al. (38) noted that mutation of either of two H2O2-sensing cysteine residues to a serine caused total loss of HyPer sensitivity to H2O2. We mutated the first cysteine in cytosolic and mitochondria-targeted HyPer to make SypHer. Briefly, the C199S mutation was performed with the QuikChange II site-directed mutagenesis kit as per the manufacturer's instructions using 12 reaction cycles with 5.5 min for elongation in each. Primers were used at 10-fold the suggested concentration, and primers were 5′-agatggtcactctttgcgcgat-3′ (forward) and 5′-atcgcgcaaagagtgaccatct-3′ (reverse), with underlined nucleotides showing the position of the point mutation. PCRs were transformed into XL1-blue bacteria, and minipreps (Gene Elute, Sigma) were prepared from clonal colonies for sequencing to confirm the C199S mutation and the absence of other mutations. Mutated plasmids were transfected into HeLa cells to confirm their sensitivity of pH and lack of response to H2O2 (200 μm).

pH Measurements

Experiments were performed in HEPES buffer (HBSS) solution containing 140 mm NaCl, 5 mm KCl, 1 mm MgCl2, 2 mm CaCl2, 20 mm Hepes, 10 mm glucose, pH set to 7.4 with NaOH at 37 °C. Ca2+-free solution contained 0.2 mm EGTA instead of CaCl2. Cells grown on glass coverslips were inserted into a thermostatic chamber (Harvard Apparatus, Holliston, MA), and solutions were changed by hand. Ratiometric pH and Ca2+ images were acquired with a ×40 objective (1.3 numerical aperture, Zeiss Axiovert s100TV) and a cooled CCD camera (MicroMax, Roper Scientific) (37). For pH imaging, SypHer (cytosolic or mitochondrial) was alternately excited for 200–300 ms at 430 and 480 nm through a 505DCXR dichroic filter and imaged with a 535DF25 band pass filter (Omega Optical). For simultaneous mitochondrial/cytosolic pH measurements, SypHer-expressing cells were loaded with SNARF (5 μm, 0.2% DMSO, 0.01% pluronic F-127, 30 min, ambient temperature) followed by 20 min of de-esterification before 10 min of equilibration on the heated stage. SypHer was imaged as above, and SNARF was excited at 480 nm for 200 ms through the same dichroic filter and imaged with 580DF30 and 640DF35 filters (Omega Optical). Image pairs and quartets were typically acquired every 2–5 s. Fluorescence ratios (F480/430 SypHer, F640/580 SNARF) were calculated in MetaFluor 6.3 (Universal Imaging) and analyzed in Excel (Microsoft) and GraphPad Prism 5.01 (GraphPad). After each experiment, cytosolic and mitochondrial pH was calibrated using nigericin (5 μg/ml) and monensin (5 μm) in 125 mm KCl, 20 mm NaCl, 0.5 mm MgCl2, 0.2 mm EGTA, and 20 mm N-methyl-d-glutamine (pH 9.5–10.0), Tris (pH 8.0, 9.0), HEPES (pH 7.0–7.5), or MES (pH 5.5–6.5). For each cell, a 6-point calibration curve was fitted to a variable slope sigmoid equation with 1/y weighting and constraining the top of the curve to 30 (GraphPad Prism 5.01).

Ca2+ Imaging

Cytosolic Ca2+ was measured with fura-2 (2 μm) or YC3.6 (39), and mitochondrial Ca2+ with 4mitD3-CPV on the epifluorescence system described above. Fura-2 was excited alternately at 340 and 380 nm (200–300 ms/λ) through a 430DCLP dichroic filter and 510WB40 emission filter when recording only Fura-2 or through a 505DCXR dichroic filter and 535DF25 filter when measured simultaneously with SypHer. Cameleons were excited at 430 nm through a 455DRLP dichroic filter and alternately imaged with 475DF15 and 535DF25 emission filters (200–400 ms/λ, Omega Optical).

Cell Lysis, Mitochondrial Isolation, and Western Blotting

Whole cells were lysed in a glass homogenizer. The lysate was centrifuged at 14,000 × g for 20 min, and the protein content of the supernatant was determined using a BCA protein assay (Pierce). The mitochondrial fraction was obtained by differential centrifugation as reported previously (40). 20 μg of total protein (from cytosolic or mitochondrial fraction) was loaded per lane of SDS-PAGE. For immunoblotting, proteins were transferred onto nitrocellulose membrane and probed with the following antibodies: anti-GFP (sc-9996) and anti-Tom20 (sc-11415) (Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) and anti-α-tubulin (T9026) (Sigma). Horseradish peroxidase-conjugated Secondary antibodies (Amersham Biosciences) were used, followed by detection by chemiluminescence (Amersham Biosciences).

Inmunocytochemistry and Confocal Imaging

48 h after transfection cells were fixed using 4% paraformaldhyde for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and incubated in blocking buffer IgG (2 μg/ml) for 1 h. Fixed cells were probed with rabbit polyclonal anti-Tom20 (1:100) for 10 h at 6 °C and labeled with Alexa Fluor 568-conjugated goat anti-rabbit IgG (1:1000) from Sigma for 1 h at room temperature. Confocal images were acquired on a Leica SP5 2 photon microscope using a ×63 oil immersion objective. MitoSypHer was excited at 488 nm, and fluorescence emitted between 510 and 540 nm was collected. Alexa Fluor 568 was excited at 561 nm, and fluorescence emitted between 600 and 630 nm was collected. The confocal pinhole was 154 μm (1 airy disk) for SypHer and 204 μm (1.25 airy disk) for Tom20.

Live Cell Confocal Imaging

Confocal images were acquired with a Zeiss LSM510meta, using a ×63 Achroplan water immersion objective (numerical aperture 0.95) and 12-bit acquisition mode with zoom set to give 130-nm pixels. For mitoSypHer/MitoTracker dual labeling, 2 days after mitoSypHer transfection, cells were incubated with 25 nm MitoTracker Red CMXRos for 45 min at 37 °C in normal culture medium. Medium was replaced with HBSS (ambient temperature) immediately before image acquisition. Images were acquired with simultaneous 488/561 nm excitation with sequential 490- and 565-nm dichroic filters and 505–550 nm bandpass emission (SypHer) and 575 nm longpass (MitoTracker Red CMXRos).

Calculation of H+-buffering Power and Fluxes

Cytosolic and mitochondrial H+-buffering powers (β) were calculated by measuring pH changes in cells exposed to permeant weak acid (butyric acid; pK = 4.82) or weak base (trimethylamine (TMA); pK = 9.80) and estimating the change in internal acid/base concentration as described previously (41). Some investigators assume that weak acids (bases) fully deprotonate (protonate) at physiological pH to simplify calculations. We found that this assumption systematically overestimated the calculated changes in weak acid and base, so we relied on exact calculation by the Henderson-Hasselbach equation. To ensure that we measured intrinsic cytosolic and matrix-buffering power, the Na+/H+ exchanger was inhibited with amiloride (100 μm), and the respiratory chain was inhibited with rotenone (5 μm), antimycin (5 μm), and oligomycin (5 μg/ml). The concentration of TMA in compartment j (cytosol or mitochondria) was estimated by Equation 1, where compartment k (extracellular space or cytosol) is the compartment immediately external to j, and pKTMA is assumed to be equal in all compartments. For butyric acid, the same equation is used, but (pH − pK) replaces (pK − pH). Subsequently, β is calculated by Equation 2, where Δ[TMAH+] = [TMAH+],final − [TMAH+],initial, calculated before and after a change in [TMA]bath, and ΔpH is measured as the midpoint pH between changes in [TMA]bath. Simultaneous measurements of pHcyto and pHmito permitted βmito to be calculated from Equation 1, where the cytosol is compartment k.

High Throughput pH Imaging

Separate halves of a 96-well plate were transfected with SypHer and mitoSypHer. 48 h later, culture medium was replaced with HBSS, and the plate was placed in an ImageXpress Micro plate-reading microscope system (Universal Imaging, Sunnyvale, CA) (humidified, 37 °C). Using laser-assisted autofocusing and automated image acquisition (20× objective, 0.5 numerical aperture), ratio images (F480: ex1 470/45 nm, 500 dichroic1, em1 520/40; F430: ex2 435/25 nm, 465 nm dichroic2, em2 480/35 nm) were acquired over the same grid pattern in each well (4–12 sites/well) with a 12-bit CCD camera. For histamine stimulations, 100 μl of 2× concentrated histamine was added to 100 μl of HBSS in each well with a multichannel pipette, and recording was started within 1 min of the addition. On each plate, SypHer and mitoSypHer ratios were calibrated such that intracellular pH for one well for each probe was clamped at 5.5, 6.0, 6.5, 7.0, 7.5, and 8.0 (supplemental Fig. S3). We created automated algorithms (MetaXpress, Universal Imaging) to define regions around cells in F480 images, to calculate background fluorescence and subtract it from the average F80 and F430 fluorescence intensity in each defined cell region. Single-cell fluorescence values (50–400 per well) were transferred to Excel for sorting and calibration of ratio values. For each well, cells with ratio values greater than two S.D. from the mean value were excluded, and individual ratio values were calibrated against a standard curve generated on each plate.

RESULTS

Probe Generation and Validation

While testing the properties of HyPer, a commercial H2O2-sensing probe based on a circularly permuted YFP (38), we observed that HyPer exhibited ratiometric responses to NH4+-induced (30 mm) matrix alkalinization of similar amplitude to its responses to H2O2 (200 μm) (Fig. 1A, top). We reasoned that HyPer could provide the basis for a ratiometric pH sensor with excellent mitochondrial targeting and brightness. We generated a pH-specific version of HyPer that we dubbed SypHer (synthetic pH sensor) by mutating one of the two H2O2-sensing cysteine residues (C199S) of HyPer, as described by Belousov et al. (38). As expected, cells expressing cytosolic SypHer still exhibited large fluorescence changes upon the addition of NH4+ but failed to respond to H2O2 (Fig 1A, bottom). We then characterized the properties of the SypHer probe in vitro, using protein extracts from HeLa cells expressing cytosolic SypHer. As shown in supplemental Fig. S1, the fluorescence intensity increased markedly with pH at excitation wavelength exceeding 435 nm, with a peak around 490 nm. The 490/420 fluorescence ratio was not altered by the addition of millimolar concentrations of Ca2+, PO4−, and H2O2 (Table 1) but increased by ∼5-fold when the bath pH was increased from 7.0 to 8.0 (supplemental Fig. S1). MitoSypHer was efficiently targeted to the mitochondrial matrix by tandem mitochondrial localization sequences, as demonstrated by its colocalization with the mitochondrial resident protein TOM20 and with MitoTracker (Fig. 1B, Pearson coefficients 0.88 ± 0.03 and 0.91 ± 0.06, n = 4 and 5 cells, respectively) and enriched reactivity for an anti-GFP antibody in mitochondrial fractions (supplemental Fig. S2). Mitochondria-targeted SypHer (mitoSypHer) had spectral properties similar to its parental probe (Fig. 1C) and displayed opposite changes in fluorescence at λex = 420 and λex = 490 when the pH of the organelle was varied with H+ ionophores (supplemental Fig. S2). Using an automated microscope to image cells on 96-well plates, equilibration of the extracellular and organellar pH with H+ ionophores confirmed that the mitoSypHer ratio reported variations in local pH and that resting pHmito in most cells was between 7.5 and 8.0 (supplemental Fig. S3). pH titration curves generated from different wells of the 96-well plate clamped at specific pH values and with cells expressing cytosolic or mitochondria-targeted SypHer closely overlapped (supplemental Fig. S3), demonstrating that the pH responses of SypHer were not altered by the matrix environment and that the calibration protocol equilibrated both compartments. In situ calibration of mitoSypHer ratio against pH showed a pK of 8.71 ± 0.05 and a Hill slope of 0.96 ± 0.05 (Fig. 1D) with an impressive 20-fold increase in ratio when pH increased from 7 to 10. The ratio increased 4-fold in the more physiological pH range of 7–8, (Fig. 1D, inset), permitting accurate measurements of pHmito. Having validated the probe, we next measured resting pHmito on the standard epifluorescence microscope, generating an autonomous calibration curve for each cell imaged. To our surprise, resting pHmito (7.61 ± 0.02, minimum 7.26, maximum 7.91, 79 cells; Fig. 1E) was notably lower than the values of 7.8–8.1 previously reported in intact cells (42–44). To confirm these observations, we measured pHmito and pHcyto with SypHer in a large number of cells, by imaging 96-well plates (supplemental Fig. S3). Mitochondrial pH was 7.64 ± 0.14 (mean ± S.D.) at 37 °C and significantly higher (7.77 ± 0.26, mean ± S.D.) at 25 °C (Fig. 1E), consistent with reports that isolated mitochondria are more polarized at 28 °C versus 37 °C (45). In contrast, pHcyto was slightly higher at 37 °C (7.19 ± 0.14) than at 25 °C (7.12 ± 0.14). Consequently, the pH gradient across the inner mitochondrial membrane, ΔpH, averaged 0.46 units in resting, non-stimulated HeLa cells at 37 °C and increased to 0.65 units when the temperature was decreased to 25 °C. The resting pHmito and ΔpH of quiescent cells at physiological temperatures were thus lower than previously reported.

FIGURE 1.

SypHer characterization. A, changes in HyPer (top) and SypHer ratio (bottom) evoked by an alkaline load (30 mm NH4Cl) and by H2O2 (200 μm). B, confocal images of mitoSypHer (green) in fixed cells stained for TOM20 (red; top) or live cells labeled with MitoTracker Red CMXRos (red; bottom). Merged images (right) show clear mitochondrial targeting of mitoSypHer. Scale bars, 5 μm (top) and 10 μm (bottom). C, excitation spectra of mitoSypHer in situ, at 37 °C using a 505-nm dichroic and 535/25 emission filter. D, in situ calibration of mitoSypHer. Inset, dynamic range at physiological pH. E, average resting pHcyto and pHmito measured by single-cell (left box) and high throughput fluorescence imaging. Boxes show quartiles with circles centered on means; error bars show S.E. for the individual cells and 95% intervals for high throughput data.

TABLE 1

Cell lysates from SypHer-expressing cells were prepared in 100 mm HEPES, 150 mm NaCl, 0.5 mm β-mercaptoethanol, pH 7.00 (n = 3) and 7.45 (n = 3).

Compound	Concentration	Change in ratio (490/440) at pH 7.00	Change in ratio (490/440) at pH 7.45
	mm	-fold	-fold
NaOH (pH 7.00–7.76)	40	2.19 ± 0
NO (SNAP)	0.1	0.99 ± 0.005	1.00 ± 0.002
DTT	1.0	0.99 ± 0.004	1.00 ± 0.004
H2O2	0.1	1.00 ± 0.004	0.99 ± 0.001
Ca2+	1.0	1.01 ± 0.006	1.00 ± 0.006
PO4−	1.0	1.01 ± 0.009	1.00 ± 0.004

Mitochondrial pH Decreases during Cytosolic Ca2+ Elevations

Cytosolic Ca2+ elevations boost mitochondrial metabolism, but whether this effect involves net changes in pHmito is controversial because both alkalinization and acidification of mitochondria have been reported during [Ca2+]cyt elevations (42, 43, 46–48). To determine how changes in [Ca2+]cyt alter ΔpH, HeLa cells expressing mitoSypHer were loaded with the Ca2+-sensitive dye fura-2 to monitor [Ca2+]cyt and pHmito simultaneously. The [Ca2+]cyt elevations evoked by the Ca2+-mobilizing agonist histamine were associated with a significant decrease in pHmito (Fig. 2A), but the [Ca2+]cyt and pHmito responses had different temporal profiles. The [Ca2+]cyt elevation exhibited a transient peak followed by a sustained plateau, whereas the pHmito decay was monophasic (Fig. 2A), even in cells that exhibited Ca2+ oscillations during the sustained phase of the Ca2+ signal (Fig. 2A, insets). Parallel pHcyto recordings with cytosolic SypHer revealed parallel decreases in pHcyto during [Ca2+]cyt elevations but with a smaller amplitude than decreases of pHmito (Fig. 2B). Interestingly, the larger decrease in pHmito than pHcyto was more pronounced during the second histamine stimulation (Fig. 2B). Cell-to-cell variability limited our ability to measure precisely ΔpH in individual cells from independent recordings of pHcyto and pHmito. To overcome this limitation, we simultaneously measured pHmito using mitoSypHer and pHcyto using the pH indicator SNARF, a red-shifted dye whose fluorescence emission does not overlap with the yellow fluorescence of mitoSypHer (Fig. 3B). Calculating the cellwise difference between the pHmito and pHcyto recordings (black and blue traces, respectively), these concurrent measurements enabled us to follow ΔpH (red trace in the bottom panel) in intact cells. As shown in Fig. 3A, each histamine stimulation decreased ΔpH by ∼0.05 (range 0.06 to −0.28; Fig. 3C, iii) because drops in pHmito typically exceeded those in the cytosol. Upon histamine removal, pHcyto recovered to resting levels, whereas a pHmito overshoot was typical following successive stimulations (Fig. 3A). As a result, ΔpH increased from 0.15 ± 0.02 in “resting” cells to 0.44 ± 0.02 following recovery from a third histamine stimulation (Fig. 3C, iv). Using the high throughput imaging, we also observed that histamine-mediated acidification was concentration-dependent and maintained during 25–30 min of stimulation with only modest recovery (supplemental Fig. S3).

FIGURE 2.

Mitochondria acidify during [Ca A, [Ca2+]cyt responses to histamine measured with fura-2 (green; raw ratio) in cells expressing mitoSypHer (black; raw ratio). Traces show the mean response of seven cells in a single field of view with an inset of simultaneous responses from a single cell. B, parallel pHmito (black, mitoSypHer; mean of 11 cells) and pHcyto (blue, cytoSypHer; mean of four cells) acidification following a 30 μm histamine addition. Scale bars, 10 μm.

FIGURE 3.

ΔpH A, simultaneous pHmito (black, mitoSypHer) and pHcyto (blue, SNARF) measurements in cells repeatedly stimulated with histamine (30 μm). For each cell, ΔpH was estimated as pHmito − pHcyto (red). Diamonds and circles, mean ± S.D. from 74 cells. B, typical mitoSypHer (top, λex 480 nm) and SNARF fluorescence images (bottom, λem 580 nm). C, average changes in pHcyto (i), pHmito (ii), and ΔpH (iii) during each histamine addition and in ΔpH (iv) after histamine washout (mean ± S.E. (error bars) of 74 cells). *, p < 0.05 for repeated measures general linear model analysis of variance.

[Ca2+]cyt Dependence of the Decreases in pHmito and pHcyto

A close inspection of Ca2+ and pH responses evoked by histamine revealed that the kinetics of acidification closely matched the instantaneous integral of the Ca2+ elevation, suggesting that acid accumulation was proportional to the total cytosolic Ca2+ load rather than its instantaneous concentration (Fig. 4A). The typical peak-and-plateau [Ca2+]cyt response to histamine is due to initial Ca2+ release from the endoplasmic reticulum through inositol 1,4,5-trisphosphate receptors, followed by sustained Ca2+ influx across store-operated Ca2+ entry channels. To determine if [Ca2+]cyt was necessary for the intracellular acidification, we treated cells with 2-APB, an inhibitor of both inositol 1,4,5-trisphosphate receptor intracellular Ca2+ release channels and store-operated Ca2+ channels. As expected, 2-APB (75 μm) inhibited both the release and influx components of [Ca2+]cyt elevations measured with the FRET-based YC3.6, reducing the integrated Ca2+ response by 86.2 ± 1.7% (supplemental Fig. S4, ). In parallel, 2-APB impaired pHcyto and pHmito decreases and completely prevented the loss of ΔpH evoked by histamine (supplemental Fig. S4, ). To separate the influx from the release component of the [Ca2+]cyt elevation, cells were treated with histamine in the presence of thapsigargin (1 μm) to inhibit the sarco/endoplasmic reticulum Ca2+ ATPase (SERCA) and to deplete endoplasmic reticulum Ca2+ stores. Subsequent restitutions of external Ca2+ caused large [Ca2+]cyt elevations due to Ca2+ entry through store-operated Ca2+ entry channels that were sufficient to induce massive decreases in pHcyto and pHmito (Fig. 4B, i and ii). Rapid Ca2+ extrusion was initiated immediately upon removal of extracellular Ca2+, whereas the onset of pH recovery was delayed until [Ca2+]cyt was almost at resting levels (Fig. 4C, i and ii, insets). Thus, Ca2+-dependent acidification did not require the activity of SERCA, and all conditions that caused [Ca2+]cyt elevations decreased both pHcyto and pHmito.

FIGURE 4.

Ca A, simultaneous [Ca2+]cyto (i; fura) and pHmito (ii; mitoSypHer) recordings illustrate that the change in mitoSypHer ratio (iii; black) closely match the instantaneous integral of the fura-2 ratio (iii; green). Traces are taken from Fig. 2A. B, effect of extracellular Ca2+ removal and restitution on [Ca2+]cyt (YC3.6 cameleon, 4 cells) (i), pHmito (11 cells) (ii), and pHcyto (nine cells) (iii) responses evoked by histamine (30 μm) and thapsigargin (1 μm). Insets show kinetics of SERCA-independent Ca2+cyto clearance (i) and pHmito recovery (ii).

PMCA Underlies the Ca2+-dependent Decreases in pH

Like SERCA, PMCA catalyzes H+/Ca2+ countertransport (49, 50) (reviewed in Refs. 51 and 52). Several lines of evidence indicate that the activity of the PMCA could account for the drop in pH associated with [Ca2+]cyt elevations. First, La3+, at millimolar concentrations known to inhibit the PMCA (53), completely inhibited histamine-induced decreases in pHcyto (not shown) and pHmito while enhancing [Ca2+]cyt elevations (Fig. 5A). Because La3+ is not a specific inhibitor of PMCA, we attempted to isolate the activity of the PMCA by treating cells with thapsigargin and reducing extracellular Na+ to 0–5 mm to inhibit the NCX. Under these conditions, the activity of the PMCA is directly reflected by the rapid decrease in [Ca2+]cyt upon the removal of extracellular Ca2+. As expected, La3+ (5 mm) rapidly and reversibly blocked this [Ca2+]cyt recovery (Fig. 5B, i), with an apparent IC50 of 0.2–0.3 mm (data not shown). Importantly, La3+ not only completely blocked the pHcyto decrease initiated by [Ca2+]cyt elevation (not shown), but it revealed a Ca2+-dependent pHmito alkalinization, particularly under conditions of low extracellular Na+ (Fig. 5B, ii). The PMCA is also inhibited at alkaline extracellular pH (pH) because the ATPase is starved for protons to exchange with cytosolic Ca2+ (54). Increasing pH to 8.2–8.8 prevented the rapid recovery of [Ca2+]cyt upon the removal of histamine, whereas subsequent restoration of physiological pH (7.4) caused an immediate decrease in [Ca2+]cyt, demonstrating the inhibition of the PMCA at alkaline pH (Fig. 5C, i). In parallel experiments, alkaline pH prevented the histamine-induced drop in pHcyto and pHmito (Fig. 5C, ii), consistent with a role for the PMCA in the acid generation. Inhibition of PMCA has also been shown for orthovanadate, with sensitivities ranging from micromolar to millimolar concentrations (53). We observed little to no inhibition of histamine-induced acidification at micromolar concentrations of orthovanadate, but 1 and 10 mm orthovanadate inhibited both pHcyto and pHmito acidifications by ∼30% (data not shown). Thus, regardless of the approach used to inhibit PMCA, decreased Ca2+ extrusion was consistently associated with decreased cellular acidification, strongly suggesting that PMCA was the main source of the acid generated during [Ca2+]cyt elevations.

FIGURE 5.

PMCA mediates the Ca A, La3+ (5 mm) blocks clearance of a [Ca2+]cyt elevation (i) and decrease in pHmito (ii) evoked by histamine (30 μm) when SERCA is blocked with thapsigargin (Tg; 1 μm). B, effect of La3+ (5 mm) on the clearance of [Ca2+]cyt elevations (i) and on decreases in pHmito (ii) evoked by Ca2+ readdition to cells treated with histamine and thapsigargin in low extracellular Na+ to block Ca2+ clearance by the Na+/Ca2+-exchanger. C, effect of alkaline pH (8.8) on [Ca2+]cyt (YC3.6 cameleon, four cells) recovery (i) after histamine removal and the decrease in pHcyto (SNARF) and pHmito (mitoSypHer) (ii) evoked by histamine. La3+ and alkaline pH both prevent Ca2+ clearance and prevent the decreases in cytosolic and mitochondrial pH.

Influence of Mitochondrial H+-buffering Power on ΔpHm

Mitochondria consistently showed larger decreases in pH than concomitant changes in the cytosol during histamine stimulations. The larger pHmito decrease caused a net loss of ΔpH during each agonist application (Fig. 3), to the extent that the loss of ΔpH showed a linear correlation with the absolute value of pHmito immediately prior to the histamine addition (not shown). The net loss of ΔpH during histamine-mediated pHcyto acidification could reflect the facilitated import of protons across the inner mitochondrial membrane. Alternatively, the different amplitudes of decreases in pHmito and pHcyto might reflect different proton buffering capacities (β) of the cytosol and mitochondria. High concentrations of cytosolic pH buffers mitigate intracellular pH changes during surges of intracellular acid production. A lesser pH buffering capacity of mitochondria compared with the cytosol might therefore cause a greater decrease in pHmito than pHcyto for an identical acid load. The pH buffering capacity of the cytosol (βcyto) is reported to be 20–30 mm (55), but the intrinsic buffering power of mitochondria (βmito) has not been determined to date. To obtain this parameter, we simultaneously measured the pHmito and pHcyto changes evoked by the addition of varying concentrations of permeant weak acids (butyric acid) or bases (trimethylamine), the former illustrated in Fig. 6A, i. This well established procedure (41) enabled us to determine the buffering power of the cytosol and of mitochondria, over the physiological range of pH 6.8–8.4. As shown in Fig. 6A, ii, the pH dependence of βcyto and βmito was similar (pKcyto 7.58 ± 0.01, pKmito 7.65 ± 0.03), suggesting that the nature of the pH buffers is similar in both compartments. Reassuringly, measurements of βcyto (20.7 ± 1.3 mm) at physiological pHcyto (6.9–7.4) were consistent with previous reports (55). Over the same pH range, βmito was measurably lower (17.8 ± 1.6 mm, p = 0.012, general linear model analysis of variance). More importantly, both βcyto and βmito rapidly fell at pH values above ∼7.4, such that βmito (∼10 mm) was roughly half of βcyto at the typical resting pHmito of ∼7.6. These experiments provide the first estimates of the pH dependence of βmito and indicate that the buffering capacity of mitochondria rapidly decreases as the organelle alkalinizes. This strongly suggests that the reduced H+ buffering capacity of mitochondria at alkaline pH underlies the loss of ΔpH during histamine stimulation.

FIGURE 6.

Low mitochondrial buffering power at alkaline pH A, i, protocol used to calculate H+-buffering power (β) by the stepwise addition/removal of permeant weak acid. Black traces, mitoSypHer; blue traces, SNARF. Amiloride (100 μm), rotenone (5 μm), antimycin (5 μm), and oligomycin (5 μg/ml) were added to prevent membrane H+ transport. ii, pH dependence of intrinsic H+-buffering power (βcyto and βmito). β values were binned every 0.1 pH units (mean ± S.E.). Dotted lines, β at mean resting pHcyto and pHmito. Solid lines, fitted sigmoid functions. B, calculation of JH for a given histamine-mediated pH change, where β is a function of pH. C, i, changes in pHmito as a function of the changes in pHcyto in the same cell during histamine responses. Regression of δpHmito against δpHcyto has a slope of >1, reflecting the loss of ΔpH during stimulation. ii, mitochondrial versus cytosolic proton fluxes. Regression of the calculated JH versus JHcyto has a slope of <1, indicating that mitochondria resist cytosolic H+ fluxes. Regressions were constrained to pass through the origin.

To quantify the effect of the differential buffering capacity of the two compartments, we used the measured βcyto and βmit values to calculate the net fluxes of protons (JH) underlying the typical pHcyto and pHmito responses in a random subset of cells (Fig. 6B). As expected, correlation of the amplitude of the drop in pHmito (δpHmito) with the drop in pHcyto (δpHcyto) produced a regression slope greater than 1, consistent with the loss of ΔpH (Fig. 6C, i). When changes in pH were converted to net H+ fluxes (JH) with the formula, JH = δpH × β, the positive correlation was retained but with a slope significantly less than unity (Fig. 6C, ii). This meant that the flux of protons reaching the matrix was smaller than the flux of protons entering the cytosol, as expected given that the inner mitochondrial membrane is not freely permeable to protons. Thus, the loss of ΔpH was probably due to the difference between βmito and βcyto at physiological pH values in each compartment. Combined, these findings show that the absolute value of pHmito prior to a stimulus like histamine and the pH dependence of βmito are important, previously unrecognized determinants of the changes in pHmito and ΔpH occurring in living cells.

Ca2+ and H+ Have Opposite Effects on Mitochondrial Matrix pH

The drop in pHmito observed during [Ca2+]cyt elevations is at odds with the known effect of mitochondrial Ca2+ uptake to cause matrix alkalinization (17). To study whether the decrease in mitochondrial pH was due to the concomitant cytosolic acidification that occurred during the [Ca2+]cyt elevation, we measured the changes in pHmito in permeabilized cells exposed to Ca2+ or pH changes. As shown in Fig. 7, a transient decrease in bath pH from 7.4 to 7.2 rapidly and reversibly decreased the matrix pH by ∼0.15 pH units. The subsequent addition of 10 μm Ca2+ at pH 7.4 evoked a matrix alkalinization that developed more slowly and that was of smaller amplitude, increasing matrix pH by ∼0.08 pH unit. Parallel experiments in cells expressing 4mitD3-CPV revealed that the addition of Ca2+ to permeabilized cells caused robust [Ca2+]mit elevations that were largely prevented by inhibition of the Ca2+ uniporter with Ru360 (supplemental Fig. S5). We then added 10 μm Ca2+ during the switch to pH 7.2 to mimic the Ca2+ and pH changes occurring in intact cells stimulated with histamine. This protocol caused a rapid decrease in pHmito followed by a slow recovery to a steady-state value ∼0.09 pH units more acidic than the initial pHmito (Fig. 7). Thus, mitochondria showed the expected pHmito increase in response to Ca2+ elevations, but this alkalinization was masked by a larger acidification when pHcyto was decreased concomitantly.

FIGURE 7.

Opposite effects of Ca A, pHmito responses evoked by the addition of H+ and Ca2+ to permeabilized cells. Bath pH was transiently decreased from 7.4 to 7.2, and then 10 μm Ca2+ was added at pH 7.4 or during the pH switch to mimic the changes occurring in intact cells. B, averaged pHmito responses evoked in permeabilized cells by a decrease in environmental pH to 7.2, by the addition of 10 μm Ca2+ at pH 7.4, or by the addition of 10 μm Ca2+ during a switch to pH 7.2.

Contributions of Mitochondrial Transporters to Mitochondrial H+ Fluxes

The large flux of protons reaching the mitochondrial matrix when the cytosol acidifies indicates that pHcyto and pHmito are in dynamic equilibrium. This is not surprising, given the multiplicity of H+-coupled transporters present in the inner mitochondrial membrane, and we then attempted to identify pharmacologically the transporters involved. Cells were stimulated repeatedly with histamine as in Fig. 3, and inhibitors were added 5 min before the third stimulation. We first inhibited the electron transport chain (ETC) with rotenone and antimycin to block complexes I and III, respectively. As expected, inhibition of the ETC caused an immediate decrease in pHmito and collapsed ΔpH to 30% of resting levels. In these conditions, the decrease in pHmito evoked by histamine closely matched that of pHcyto, and the histamine-induced loss of ΔpH was precluded (Fig. 8A). To quantify the effect of ETC inhibition on mitochondrial H+ fluxes, the ratio of net JH to net JHcyto evoked by histamine was compared in single cells before and after the addition of inhibitors (or vehicle) to account for cell to cell variability in JHcyto. As shown in Fig. 8C, ETC inhibition reduced JH/JHcyto by ∼28 ± 18%. We then used oligomycin to inhibit the F1F0-ATP synthase. Contrary to our expectations, the addition of oligomycin did not increase pHmito over a 5-min period in a majority of cells, possibly because mitochondria were already maximally alkaline after two successive stimulations with histamine. Oligomycin had no apparent effect on the subsequent decrease in pHcyto, pHmito, and ΔpH evoked by a third addition of histamine, and calculation of the protons fluxes revealed that oligomycin did not significantly reduced the average JH/JHcyto (Fig. 8C). Bongkrekic acid, an inhibitor of the adenine nucleotide transloctor, rapidly and selectively acidified mitochondria but did not significantly alter the JH/JHcyto (Fig. 8C). Thus, only inhibition of the ETC consistently decreased proton fluxes (by 20–25%), but this inhibition was associated with a decreased proton-motive force. Next, we investigated whether the mitochondrial Na+/Ca2+ exchanger (NCLX), the major mitochondrial Ca2+ extruder of HeLa cells (56), participates in the rapid flux of protons into the matrix during cytosolic Ca2+ elevations. NCLX activity requires a mitochondrial Na+ gradient generated and maintained by the mNHE, thus effectively coupling the extrusion of one Ca2+ ion with the entry of three protons into the mitochondrial matrix (7). To examine the contribution of the mNCX/mNHE axis to net mitochondrial H+ flux, we used CGP-37157 to inhibit the NCLX (Fig. 8A). CGP-37157 reduced and delayed the onset of pHcyto and pHmito acidification, consistent with Ca2+ sequestration in mitochondria and reduced PMCA activity. Importantly, CGP-37157 reduced JH/JHcyto by ∼45% (Fig. 8C), suggesting that the mNCX/mNHE axis is implicated in the rapid transmission of protons from the cytosol to the mitochondrial matrix.

FIGURE 8.

Contributions of mitochondrial transporters to mitochondrial proton fluxes. A, effects of inhibitors of the ETC (5 μm rotenone, 5 μm antimycin with or without 5 μg/ml oligomycin) and of the mitochondrial NCX (10 μm CGP-37157) on histamine-mediated pHmito (black, mitoSypHer) and pHcyto (blue, SNARF) responses in intact cells. Bar graphs show mean ± S.E. drug effects for pHcyto (blue) and pHmito (gray) acidification and loss of ΔpH (pink). n is displayed in the bars and p values are shown for t tests. B, effects of inhibitors of the ETC and of the mitochondrial NCX (as above) on the pHmito responses evoked by the addition of H+ to permeabilized cells. Bath pH was transiently decreased from 7.4 to 7.2 before and after the addition of the inhibitors. Bar graphs show mean ± S.E. decrease in pHmito evoked by the pH switch from 7.4 to 7.2. C, average effect of the inhibitors on the JH/JHcyto flux ratio measured in intact cells during the third histamine response, using the second histamine response as internal control. Bars, mean ± S.E. (error bars) from n cells (number in bars) from ≥6 coverslips. *1, significantly different from control by Dunnett's test with groupwise α = 0.05. D, average effect of the inhibitors on the JH flux measured in permeabilized cells during a pH switch from 7.4 to 7.2. *2, significantly different (p < 0.05) by one-sample t test from internal control in the absence of inhibitor.

Because our pharmacological analysis of intact cells did not reveal a single clear mechanism of rapid proton flux across the mitochondrial membrane, we turned to permeabilized cells to clarify the mechanisms contributing to the high permeability of mitochondria for protons. As shown in Fig. 8B, the combination of oligomycin, antimycin, and rotenone rapidly decreased pHmito, consistent with inhibition of the ETC. Under these conditions, the amplitude of the drop in pHmito evoked by a rapid drop in bath pH was reduced by ∼38%, and the calculated proton flux was reduced by 30% (Fig. 8D). In contrast, resting pHmito as well as acid-induced proton fluxes were not affected by Ru360 and CGP 37157 (Fig. 8, B and D). Consistent with inhibition of mitochondrial Ca2+ uptake and extrusion, Ru360 reduced the pHmito elevations evoked by the addition of Ca2+ to permeabilized cells (supplemental Fig. S5), whereas CGP 37157 potentiated the Ca2+-induced matrix alkalinization (supplemental Fig. S5). Finally, we tested the effects of the Pi/H+ symport inhibitor mersalyl. Mersalyl (100 μm) induced a rapid increase in matrix pH, followed by a subsequent decrease that was associated with a 70% increase in proton flux (supplemental Fig. S6). The initial alkalinization upon mersalyl addition is consistent with inhibition of coupled Pi/H+ entry, but the subsequent drop in matrix pH and increased proton flux suggest that the inner membrane has been damaged by mersalyl. We then turned to ion substitution experiments to identify the H+ transporters involved in the rapid flux of protons across the inner mitochondrial membrane. As shown in supplemental Fig. S6, phosphate removal increased pHmito, whereas phosphate addition decreased pHmito, consistent with coupled Pi/H+ transport. Conversely, potassium removal (120 mm KCl substituted with 235 mm sucrose) decreased pHmito, whereas the subsequent addition of 50 mm KCl increased pHmito (supplemental Fig. S6). Both effects are consistent with K+/H+ exchange because in our short term recordings, mitochondria are not depleted of K+. A drop in pHmito was also observed in sucrose medium when the bath pH was switched from 7.4 to 7.2 (data not shown). The drop in pHmito could reflect increased K+/H+ exchange activity (i.e. more K+ leaving the matrix in exchange for additional external protons). These recordings show that changes in the cytosolic concentration of Pi and K+ alter the matrix pH in the direction predicted by Pi/H+ symport and K+/H+ exchange.

Together, the pharmacological and ion substitution experiments indicate that the Ca2+ uniporter and the Na+/Ca2+ exchanger NCLX do not substantially facilitate the flux of protons into mitochondria during changes in environmental pH. Instead, Pi/H+ symport and K+/H+ exchange activity probably contribute to the rapid adaptation of the mitochondrial matrix pH to changes in cytosolic pH.

DISCUSSION

In this study, we report the first time-resolved quantitative measurements of the changes in ΔpH and of the trans-mitochondrial proton fluxes occurring in single living cells during physiological stimulation. The determination of ΔpH was permitted by concurrent recordings of mitochondrial and cytosolic pH by combining our new mitochondria-targeted, ratiometric indicator SypHer with the cytosolic dye SNARF. These measurements revealed several novel and unexpected aspects of mitochondrial ionic homeostasis.

First, the resting pH of mitochondria in HeLa cells was unexpectedly low, averaging 7.8 at 25 °C and only 7.6 at 37 °C, a value that increased to 7.8 upon repetitive stimulation with histamine. Previous studies using pH-sensitive GFP mutants reported higher resting pHmito values of ∼8–8.2 in HEK-293, Jurkat, and HeLa cells (42, 44, 57). Others, however, reported lower pHmito values, similar to those we report here. A resting pHmito of 7.8 was reported in ECV304 endothelial cells using a mitochondrial EYFP (34) and a pHmito of 7.7 in Madin-Darby canine kidney cells with the ratiometric dye SNARF (58). Recently, we reported a resting pHmito of 7.2 in rat pancreatic β cells that increase to 7.7 upon nutrient stimulation (46). pHmito is an important parameter that determines the ability of mitochondria to fulfill their metabolic and homeostatic functions. pHmito increases early during apoptosis (44), and a low pHmito modulates the opening of the permeability transition pore (59, 60) (reviewed in Ref. 61). In pancreatic β cells, an alkaline pHmito is required for the delayed [Ca2+]cyto that sustains the secretion of insulin in response to glucose (46). Our current measurements with SypHer indicate that the commonly assumed resting pHmito of ∼8 probably overestimates typical resting pHmito in HeLa cells and several other cell types. Despite this lower than expected value for pHmito, our ΔpH values of 0.5–0.7 pH units derived from static high throughput measurements and 0.2–0.4 pH units derived from dynamic single-cell imaging agree well with the values of ∼0.3–0.8 pH units reported in earlier studies with GFP-based probes (34, 43). Because the intermembrane space is 0.7 pH units more acidic than the bulk cytosol (34), the actual ΔpH is therefore >1 pH unit. Our data thus indicate that the resting pHmito is quite low in quiescent cells and increases in the wake of agonist-induced Ca2+ signaling.

Second, we observed that changes in cytosolic pH were paralleled by changes in mitochondrial pH. Bursts of intracellular proton accumulation, due to the activity of the PMCA (see below), were readily transmitted to the mitochondrial matrix. A similar rapid equilibration of pHmito and pHcyto was reported in MDCK cells during metabolic inhibition and attributed to the proton antiporters of mitochondria (58). Our pHmito recordings from permeabilized cells indicate that the matrix pH is affected both by Pi/H+ symport and K+/H+ exchange activity (supplemental Fig. S6). Because K+ and Pi are present at high concentration in the matrix and in the cytosol, these two transporters probably contribute to the rapid equilibration of the mitochondrial matrix pH during cytosolic pH excursions. Unfortunately, we could not identify, pharmacologically, a single transporter that could account for the proton fluxes, and only inhibition of the mitochondrial respiratory chain consistently decreased the proton fluxes (by 20–30%) both in intact and permeabilized cells (Fig. 8). However, this inhibition probably reflects the decreased proton-motive force rather than decreased proton permeability. On the one hand, this high permeability of mitochondria to protons appears incompatible with the need for mitochondria to maintain a proton-motive force across their inner membrane. The cytosol to matrix pH gradient is essential for the import and export of a large number of metabolites. It should be kept in mind, however, that respiratory chain complexes cluster in cristae, folds of the inner mitochondrial membrane with restricted diffusional access to the rest of the intermembrane space. It is thus conceivable that H+ transporters like the Na+/H+ exchanger and phosphate/H+ symporter, whose substrates require easy access to the cytosol, might preferentially localize outside the cristae. Such an arrangement would allow generation of a large pH gradient across of the cristae membrane to power the F1F0-ATP synthase while accounting for the relatively high permeability of mitochondria to protons in our experiments. Our quantitative pH data also indicate that, despite the rapid adaptation of pHmito to changes in pHcyto, the matrix remained at all times more alkaline than the cytosol, even during large pH excursions (Fig. 3C, ii). These data show that mitochondria can maintain a positive cytosol to matrix pH gradient despite large variations in the absolute pH of their matrix and of their surrounding environment. This hitherto unappreciated property might be important to preserve the bioenergetic function of mitochondria in cells exposed to acid or alkaline loads.

Third, we observed unexpected biphasic changes in mitochondrial pH during stimulation of cells with a Ca2+-mobilizing agonist. [Ca2+]cyt elevations were associated with a massive acidification, whereas removal of the agonist increased pHmito above prestimulatory levels. As a result of the overshoot, ΔpH strongly depended on prior exposure of cells to exogenous stimuli and increased by ∼0.2 pH units after three successive agonist stimulations. This effect was most likely due to the Ca2+-dependent activation of matrix dehydrogenases because matrix alkalinization was also observed in permeabilized cells exposed to 10 μm Ca2+. The Ca2+-induced alkalinization, however, was of smaller amplitude than the acidification evoked by a concomitant decrease in bath pH from 7.4 to 7.2, a condition that mimics the changes occurring in intact cells stimulated with Ca2+-mobilizing agonists. These data indicate that mitochondria slowly alkalinize in response to [Ca2+]cyt elevations but that the alkalinization is masked by a larger and faster acidification as the cytosolic pH drops concomitantly during stimulations with Ca2+-mobilizing agonists.

A prominent mitochondrial acidification was reported in cultured cortical neurons stimulated with glutamate (43) and in endothelial cells stimulated with histamine (62) but not in an earlier pHmito study of HeLa cells (42). We now provide a mechanistic explanation for this [Ca2+]cyt-dependent mitochondrial acidification. Three lines of evidence indicate that the PMCA is the source of the acid produced during [Ca2+]cyt elevations: 1) the acidification was proportional to the total calcium load, a kinetic property expected for a Ca2+/H+ transporter that couples the extrusion of Ca2+ ions to proportional entry of protons into cells; 2) the acidification did not require Ca2+ uptake or Ca2+ release from intracellular stores but was related to plasma membrane Ca2+ fluxes; 3) the acidification was prevented by La3+ or by alkaline pH, two treatments that inhibit the PMCA. The PMCA mediates cytosolic acidification upon toxic stimulation of neurons with domoate (63) and has been postulated to mediate glutamate-induced acidification (43). Our data indicate that the acid generated by the PMCA is transmitted to the mitochondrial matrix. This mechanism might protect cells from excessive Ca2+ elevations because an acidic pHmito inhibits the Ca2+-induced opening of the permeability transition pore (64, 65) and reduces mitochondrial production of potentially toxic radical oxygen species (31). In energized mitochondria, however, an acidic pH has the opposite effect and promotes the opening of the permeability transition pore (60). Thus, whether the transmission of the acid generated by the PMCA to mitochondria has protective or deleterious effects remains to be determined.

Fourth, we report here the first estimates of the pH buffering capacity of mitochondria in intact cells. This parameter was derived from the pH changes evoked by the addition of known concentrations of membrane-permeable weak acid or bases. The determination of βmito enabled us to calculate the fluxes of protons within mitochondria. Our βmito values of 10–18 mm are notably lower than the ∼110 mm measured by phosphate-NMR in perfused liver (66) but consistent with values from isolated mitochondria of ∼22 nmol/mg protein (67) that translate into a βmito of 11–13 mm, assuming a matrix water volume of 1.6–2.0 μl/mg of protein (68). Our values reflect the intrinsic pH buffering capacity of mitochondria, since our solutions were devoid of bicarbonate. Overall, the pH dependence of the βmito and βcyto had a similar apparent affinity for protons, suggesting that similar molecules buffer protons within the two compartments. This is not too surprising, considering that intrinsic pH buffers (i.e. non-bicarbonate) are mainly provided by phosphate groups and side chains of amino acids (69). Importantly, the mitochondrial buffering power dropped rapidly when pHmito increased from 7.6 to 8.0. This suggests that the titrable groups that can bind protons within the matrix of mitochondria have a nearly neutral pK. The sharp decrease in βmito at alkaline pH facilitates the generation of a proton gradient as mitochondria become more alkaline, because fewer protons need to be pumped to generate an equivalent change in pHmito. The low buffering capacity of mitochondria thus facilitates the H+-coupled entry or extrusion of mitochondrial substrates and metabolites during cell activation. On the other hand, the low βmito renders mitochondria vulnerable to surges of cytosolic acid production as a relatively small influx of protons markedly decreases pHmito in alkaline mitochondria. This poor protection was experimentally verified to underlie a larger decrease in pHmito than pHcyto and a consequent decrease of ΔpH during Ca2+-induced acidification.

In summary, we show that the mitochondrial pH is not as stable or as alkaline as generally assumed but is dynamically regulated and can vary widely during physiological stimulations. During [Ca2+]cyt elevations, the acid generated by the plasma membrane Ca2+ ATPase is transmitted to the mitochondrial matrix and decreases ΔpH. The drop in ΔpH is due to the reduced buffering power of mitochondria in the alkaline pH range, which by amplifying the changes in pHmito facilitates the generation but also the dissipation of the proton gradient.