Imaging pH Dynamics Simultaneously in Two Cellular Compartments Using a Ratiometric pH-Sensitive Mutant of mCherry.

The regulation of pH is essential for proper organelle function, and organelle-specific changes in pH often reflect the dynamics of physiological signaling and metabolism. For example, mitochondrial energy production depends on the proton gradient maintained between the alkaline mitochondrial matrix and neutral cytosol. However, we still lack a quantitative understanding of how pH dynamics are coupled between compartments and how pH gradients are regulated at organelle boundaries. Genetically encoded pH sensors are well suited to address this problem because they can be targeted to specific subcellular locations and they facilitate live, single-cell analysis. However, most of these pH sensors are derivatives of green and yellow fluorescent proteins that are not spectrally compatible for dual-compartment imaging. Therefore, there is a need for ratiometric red fluorescent protein pH sensors that enable quantitative multicolor imaging of spatially resolved pH dynamics. In this work, we demonstrate that the I158E/Q160A mutant of the red fluorescent protein mCherry is an effective ratiometric pH sensor. It has a pKa of 7.3 and a greater than 3-fold change in ratio signal. To demonstrate its utility in cells, we measured activity and metabolism-dependent pH dynamics in cultured primary neurons and neuroblastoma cells. Furthermore, we were able to image pH changes simultaneously in the cytosol and mitochondria by using the mCherryEA mutant together with the green fluorescent pH sensor, ratiometric-pHluorin. Our results demonstrate the feasibility of studying interorganelle pH dynamics in live cells over time and the broad applicability of these sensors in studying the role of pH regulation in metabolism and signaling.

Introduction

In eukaryotic cells, pH compartmentalization is critical for cellular processes, such as mitochondrial energy production, protein degradation in lysosomes, and post-translational protein modification in the endoplasmic reticulum.[1,2] In the brain, for example, organelle pH gradients are essential for proper neurophysiology. The acidification of synaptic vesicles provides the proton motive force for neurotransmitter loading, and the alkalinization of the mitochondrial matrix provides the proton motive force for adenosine 5′-triphosphate (ATP) synthesis, both of which are required for the energetically expensive process of neurotransmission.[3,4] Because neurotransmission is fundamentally a pH-dependent process, pH is also a useful indicator of activity. For example, transient pH fluctuations in the cytosol of neurons occur due to proton and ion fluxes and mitochondrial pH fluctuates in response to energy consumption during action potential generation.[5−7]

Aberrant changes in pH are also commonly seen in diseases. For example, significant pH changes are seen during neurological disorders such as stroke and ischemia, where pH dynamics during hypoxia plays an important role in cell survival.[1,8] In cancer, altered pH homeostasis can occur[9] and the Warburg effect and metabolic reprogramming can cause intracellular alkalinization and extracellular acidification, both of which may play important roles in cell survival.[10−12] Furthermore, regulation of organelle pH has been linked to oncogenic signaling[13,14] but we still need new pH sensors to accurately study organelle pH changes in the context of the entire cell.

In biological imaging studies, spatially resolved pH dynamics are commonly visualized using pH-sensitive dyes, such as BCECF or SNARF,[15] and genetically encoded green and yellow fluorescent protein-based pH sensors, such as pHluorin and SypHer.[16,17] Genetically encoded sensors are advantageous because they can be targeted to subcellular locations. However, to study the role of proton exchange and buffering between different compartments in pH regulation, we need a toolbox of both green and red fluorescent pH sensors for multiplex imaging. Currently, many of the red fluorescent protein (RFP) sensors, such as pHTomato,[18] pHuji,[19] and mNectarine,[20] report pH changes on the basis of fluorescence intensity alone. These sensors have high dynamic range and are very useful for the detection of events such as synaptic vesicle release.[21] However, ratiometric sensors are better suited for quantifying pH changes because they are insensitive to variations in sensor concentration and photobleaching, which facilitates the comparison of pH dynamics between independent experiments. The ratiometric red fluorescent protein pH sensor pHRed[22] has been used to monitor pH fluctuations in mitochondria[23] and to measure peroxisomal pH.[24] However, pHRed has a pKa of 6.9, which limits its sensitivity in alkaline compartments because it exhibits smaller changes in its signal at pH greater than 8. The availability of spectral variants and pH sensors with a range of pKa values would facilitate the study of pH changes in various cellular compartments.

In this study, we demonstrate that the I158E/Q160A mutant of mCherry, originally reported by Piatkevich et al.,[25] is an effective ratiometric red fluorescent protein pH sensor. We characterize the pH-dependent fluorescence properties of the mCherry mutant protein in solution, and we also characterize its pH sensing performance in live cells. To demonstrate its use in biological studies, we show that it reports pH dynamics caused by changes in neuronal activity and metabolism. Furthermore, we demonstrate that it is spectrally compatible with the green fluorescent pH sensor ratiometric-pHluorin, facilitating the visualization of pH changes in mitochondria and the cytosol simultaneously within the same live cell.

Results

Characterization of mCherryEA as a Ratiometric pH Sensor

We first demonstrated that the mCherry(I158E/Q160A) mutant, which we refer to as “mCherryEA”, is a ratiometric pH sensor. The I158E and Q160A mutations were originally engineered to support excited-state proton transfer (ESPT) in mCherry to generate a long Stokes shift (LSS) variant (Figure A).[25] This mutant has not been used as a long Stokes shift RFP because at neutral pH, it exhibits two excitation peaks corresponding to the protonated and deprotonated chromophore, with a single emission peak (Figure B). However, we hypothesized that existence of the two peaks would instead make the mutant an effective ratiometric pH sensor because it is proposed that the protonation state of Glu158 is sensitive to the pH of the surrounding solution,[25] resulting in pH-dependent protonation of the chromophore.

Figure 1

Characterization of purified mCherryEA in solution. (A) Structure of wild-type mCherry (PDB 2H5Q) showing the location of the E158 mutation relative to the chromophore. (B) The pH-dependent fluorescence excitation and (C) emission spectra (λex = 440 nm) (n = 6). Fluorescence was normalized to total integrated fluorescence. (D) The pH titration curves for the fluorescence intensity with 455 nm (black, dashed) and 585 nm (red, solid) excitation, with emission at 630 nm. (E) The pH titration curve for the F585nm/F455nm ratio (n = 7). The pH titration curves for the (F) extinction coefficient, (G) quantum yield (QY), and (H) brightness (ε·QY) (n = 2). Data were fit to a Boltzmann equation, ratio = minimum + (maximum – minimum)/(1 + exp((pH – pKa)/slope)). Errors are standard deviation (std).

To determine whether the mCherryEA mutant exhibits a pH-dependent ratiometric change in the excitation spectra peaks, we characterized the pH response of purified mCherryEA protein in solution. The mutant showed pH-dependent changes in its excitation spectra with a ratiometric change in excitation peaks at 455 and 585 nm, without a significant spectral shift in the emission peak at 610 nm (Figure B,C). We did not observe residual green emission from immature chromophore, and thus mCherryEA is compatible with blue, cyan, green, and yellow fluorescent proteins that exhibit spectrally distinct emission peaks. As predicted, the proton-transfer network[25] results in an “inverse” pH dependence with an increase from pH 5.5 to 9, causing a 2.80 ± 0.14-fold increase in the 455 nm peak (protonated form) and 2.54 ± 0.20-fold decrease in the 585 nm peak (deprotonated form), with a pKa of 7.8 (Figure D). Fold change was measured by dividing the highest fluorescence ratio by the lowest fluorescence ratio. The intensity ratio (F585nm/F455nm) showed a 6.90 ± 0.83-fold (n = 7, mean ± std) increase with decrease in pH from 9 to 5.5, and the pKa was 7.29 ± 0.03 for the ratio response (Figure E). The pH response was insensitive to variation of salts (NaCl, KCl, MgCl2, and K-gluconate), oxidative stress (H2O2 and dithiothreitol (DTT)), and temperature (25–37 °C) (Figure S-1).

To compare the brightness of the mCherryEA mutant relative to wild-type mCherry, we measured the extinction coefficient (ε) and quantum yield (QY) of each protein in solution. At pH 7.5, the mutant (ε = 11 650 M–1 cm–1, QY = 0.05) was dimmer compared to wild-type mCherry (ε = 72 000 M–1 cm–1, QY = 0.22).[26] Both the excitation peaks showed an increase in QY with increasing pH (Figure F). However, the ε530nm decreased with increasing pH and ε440nm remains relatively unchanged (Figure G). The brightness (ε·QY) for the 455 nm peak showed a larger change compared to that of the 585 nm peak, and it drives the dynamic range of the ratio (Figure H). These results show that the mCherryEA mutant can serve as a ratiometric pH sensor, and therefore it could be an important addition to the toolset of quantitative pH sensors. However, due to the relatively low brightness of the purified protein, we next characterized the sensor expressed in live cells and found it to be well suited for live-cell ratiometric imaging.

Live-Cell Ratiometric pH Imaging

To test mCherryEA as a pH sensor in cellular imaging, we expressed the protein in mammalian cells. The brightness of mCherryEA was comparable to that of pHRed in several mammalian cell lines using exactly the same illumination and imaging conditions (Figures and S-2), and we did not observe any cell toxicity during extended imaging sessions that lasted over 2 h. Differences in sensor characteristics in live cells compared with purified protein measurements have been observed for other fluorescent proteins and sensors, although the exact reasons are unknown.[27,28] Thus, despite the low ε and QY measured for purified protein in solution, in live cells, mCherryEA exhibits sufficient brightness to provide a high fluorescence signal over the background.

Figure 2

mCherryEA reports live-cell pH changes in different cell types. (A) Top: differential interference contrast (DIC) and fluorescence images showing the expression of mCherryEA in the cytosol of Neuro2A cells. Bottom: representative pseudocolored image sequence showing changes in the pixel-by-pixel fluorescence ratio over time in response to a transient NH4Cl pulse. (B) The pH response upon exposure to 10 mM NH4Cl measured as the F575nm/F440nm ratio over time (n = 20). (C) The pH response in primary astrocytes expressing mCherryEA upon exposure to 10 mM NH4Cl (n = 11).

We next tested mCherryEA’s pH response by exposing the cells to ammonium chloride (NH4Cl). It is well-established that exposure to NH4Cl causes a transient alkalinization and reacidification upon washout.[29] The responsiveness of the mutant was demonstrated in Neuro2A cells expressing mCherryEA in the cytosol (Figure A). The cells were exposed to 10 mM NH4Cl for 5 min and then washed with imaging solution. The pH response was determined by measuring the ratio F575nm/F440nm for single cells over time (Figure A). As expected, the addition of NH4Cl caused alkalinization of the cytosol, followed by reacidification upon removal of NH4Cl (Figure B). We found that mCherryEA exhibited a 2.3 ± 0.2-fold change in ratio signal (n = 20, mean ± std) in response to the NH4Cl transient, which was comparable to the 4.2 ± 0.7-fold change observed for pHRed (n = 29, mean ± std) (Figures B and S-2). We saw similar responses in cultured primary astrocytes (Figures C and S-3) and in HEK-293 cells (Figure S-3). Note that the lag in pH response to NH4Cl is caused by the slow perfusion delay and variability in mixing in the live-cell imaging chamber, which also contributes to overall differences in the response. Interestingly, we observed that primary astrocytes regulate cytosolic pH more strongly than Neuro2A cells. That is, the astrocytes exhibited a rebound neutralization during the NH4Cl exposure, which was not observed in Neuro2A cells (Figure B,C). It is not clear if this is an active or passive homeostatic mechanism, but future experiments could address the energy dependence of the response by pairing mCherryEA with one of the currently available green fluorescent ATP sensors.[30−32]

We did observe that long-term expression of both wild-type mCherry and mutant mCherryEA resulted in the formation of red fluorescent puncta in cultured astrocytes after several days (Figure S-3). This may be due to protein accumulation in lysosomes, which has been observed for other red fluorescent proteins.[33−35] To avoid complications caused by puncta formation, all subsequent experiments were carried out 2 days after transfection when neurons and Neuro2A cells did not show any puncta and therefore did not interfere with its use or analysis.

Live-Cell pH Calibration

To calibrate the pH response, we performed an in situ pH titration in Neuro2A cells, expressing cytoplasmic mCherryEA using the ionophore nigericin. Nigericin is a K+/H+ ionophore, which equilibrates the intracellular pH and extracellular pH when high-potassium imaging solution is used.[36] The cells were exposed to nigericin solutions to clamp cytosolic pH from pH 5.5 to 9, and steady-state values were measured over a period of 15–30 min (Figure S-4). mCherryEA in cells has a pKa of 7.31 ± 0.01 (n = 3, mean ± std) (Figure A), consistent with the pKa measured with the purified protein in solution (Figure E). We also carried out an end-of-experiment nigericin pH calibration after exposing cells to an ammonium chloride transient pulse, demonstrating that it is possible to normalize ratio signals into pH values (Figures B and S-5).

Figure 3

In situ pH calibration of mCherryEA in the cytosol and mitochondria live cells. (A) pH titration of Neuro2A cells expressing cytosolic mCherryEA (red line, n = 6, 10 cells each) or ratiometric-pHluorin (green dashed line, n = 3, 10 cells each) using nigericin. (B) pH change upon exposure to a transient 10 mM NH4Cl pulse in Neuro2A cells expressing cytosolic mCherryEA that was calibrated using nigericin at the end of the experiment (n = 4). (C) Example DIC and fluorescence images of a Neuro2A cell showing colocalization of mito-mCherryEA and MitoTracker Deep Red. Cell 1 expressed mito-mCherryEA, but cell 2 was not transfected. Cell 1 shows colocalization (yellow) of mito-mCherryEA (green) and MitoTracker (red). (D) pH titration of Neuro2A cells expressing mito-mCherryEA (red line) and mito-ratiometric-pHluorin (green dashed line) using nigericin plus monensin (n = 3, 4–15 cells each). (E) pH change upon exposure to a transient 10 mM NH4Cl pulse in Neuro2A cells expressing mito-mCherryEA that were calibrated using nigericin plus monensin at the end of the experiment (n = 7). Bars indicate std.

The pKa of mCherryEA also makes it well suited for studying pH changes in the mitochondria matrix, which can fluctuate between neutral and alkaline conditions (pH 7–8).[17,37] We first demonstrated efficient targeting of mito-mCherryEA in Neuro2A cells using the mitochondrial stain, MitoTracker Deep Red, and high-magnification images confirm colocalization of mito-mCherryEA and MitoTracker (Pearson correlation coefficient 0.93 ± 0.06, n = 12 cells) (Figures C and S-6). The in situ pH titration of Neuro2A cells expressing mito-mCherryEA was carried out using nigericin plus monensin to clamp mitochondrial pH from pH 5.5 to 9,[17] and we measured a pKa of 7.18 ± 0.09 (n = 3, mean ± std) similar to that of both purified protein and in situ cytosolic values (Figure D). Again, we carried out an end-of-experiment pH calibration using nigericin plus monensin following an ammonium chloride treatment to demonstrate that ratio signals can be normalized to mitochondrial pH values (Figure E). Interestingly, we observed that mitochondrial pH resided near neutral pH, as has been observed for HeLa cells in other studies.[17]

For comparison to mCherryEA, ratiometric-pHluorin[16] is a green fluorescent pH sensor that is also based on a mutated single fluorescent protein. In the cytoplasm, mCherryEA exhibited a maximum fold change of ratiomax/ratiomin = 3.84 ± 0.91 (mean ± std, n = 6), which is similar to ratiometric-pHluorin expressed in the cytosol of cells (ratiomax/ratiomin = 2.83 ± 0.20, pKa = 6.38 ± 0.77, n = 3) (Figure A). The maximum fold change was similar for mCherryEA in mitochondria (ratiomax/ratiomin = 4.42 ± 1.08, mean ± std). However, mito-ratiometric-pHluorin showed a dampened dynamic range (ratiomax/ratiomin = 1.36 ± 0.16, pKa = 6.83 ± 0.10, mean ± std, n = 3) (Figure D).

Thus, we have demonstrated that mCherryEA can be used for calibrated pH measurements in the cytosol and mitochondria of live cells. Ideally, in situ pH calibrations are carried out on a cell-by-cell basis at the end of each experiment to transform the ratio signal into an absolute pH value.[17] However, in practice, ionophores and permeabilization reagents such as nigericin and monensin can cause cell death in sensitive cell types such as primary neurons or after experimental paradigms that already cause significant cell stress and the low yield of successful calibrations can be prohibitive. Despite this complication, relative pH changes within a compartment can be measured more easily and have proven sufficient for studying many physiological processes, including metabolism and neuronal activity.[38−40] Therefore, we next demonstrated that mCherryEA can be used together with ratiometric-pHluorin to measure relative changes in compartment-specific pH, taking advantage of the fact that the spectral cross talk between mCherryEA and ratiometric-pHluorin is minimal (≤3%). Although the ratio signals that we measure cannot be used to directly compare absolute differences in pH between compartments, they can be used to observe correlations between relative pH changes. We conducted two proof-of-concept experiments. First, we demonstrated that both cytosolic and mitochondrial pH exhibit activity-dependent acidification in neurons. To do this, we expressed mCherryEA in cytosol and ratiometric-pHluorin in mitochondria. Second, we demonstrated that relative changes in cytosolic and mitochondrial pH correlate with metabolic inhibition. To do this, we also demonstrate that mCherryEA and ratiometric-pHluorin can be used in either compartment.

Neuronal Activity-Dependent pH Dynamics

Neuronal activity involves membrane depolarization, and membrane depolarization results in the acidification of neurons.[8,41] Membrane depolarization can be caused by neurotransmitters such as glutamate or a rise in external potassium chloride, and both proton fluxes through channels and metabolic generation of acid equivalents can contribute to the pH dynamics.[42,43] To test the utility of mCherryEA in visualizing this activity-dependent acidification, we expressed mCherryEA in the cytosol of cultured mouse hippocampal neurons (Figure A). Upon transient stimulation with 10 μM glutamate, we were able to visualize acidification in the cytosol, as expected (Figures B,C first arrow and S-7). This was followed by wash-in of 100 μM ATP before a second pulse of glutamate in the presence of ATP (Figure B,C second arrow). ATP was then washed out prior to a third glutamate application (Figure B,C third arrow). The activity-dependent acidification of the cytosol could be observed with repeated glutamate stimulation, and it was not modified by the application of neuromodulators, such as extracellular ATP acting on neuronal purinergic receptors.[44] To validate that mCherryEA was responsive to both acidic and alkaline changes in pH, we exposed neurons to a transient pulse of NH4Cl. Cells exhibited a biphasic response, demonstrating that it is functionally responsive in neurons, similar to other cell types (Figure ). Because energy metabolism contributes significantly to activity-dependent acidification, we also investigated pH dynamics more closely in both the cytosol and mitochondria.

Figure 4

Activity-dependent pH changes in neurons. (A) Representative images showing cortical neurons: DIC, cytosolic mCherryEA (red, λex = 575 nm, λem = 632 nm), mitochondrial ratiometric-pHluorin (green, λex = 475 nm; λem = 525 nm), and merged overlay. (B) Average pH change over time in the cytosol (red) and mitochondria (green, dashed) upon addition of 10 μM glutamate (arrows) in the presence and absence of 100 μM ATP (n = 3). (C) Example of a single-cell response. (D–E) Hippocampal neurons coexpressing cytosolic mCherryEA and mitochondrial ratiometric-pHluorin. (D) Average pH change over time in the cytosol (red) and mitochondria (green, dashed) upon exposure to 15 mM KCl and 10 μM glutamate for 5 min (n = 11). (E) Example of a single-cell response. Cells were treated with 10 mM NH4Cl at the end of the experiment to validate the sensor response. Bars indicate std.

Neuron excitation is a highly energy consuming process, requiring increased glycolysis and mitochondrial respiration for ATP production. Thus, the increased energy demand drives pH changes in both the cytosol and the mitochondrial matrix, and we used a dual pH sensor imaging approach to test how strongly the compartment-specific pH dynamics is coupled to activity. By coexpressing cytosolic mCherryEA and mitochondria-targeted ratiometric-pHluorin[45] in the same neurons (Figures and S-8), we were able to observe synchronous activity-dependent pH changes with similar dynamics in cytosol and mitochondria. Notably, the spectral cross talk between the mCherryEA red fluorescence and ratiometric-pHluorin green fluorescence channels is minimal (≤3%) and does not account for the synchronous changes in cytosolic and mitochondria pH. Thus, the ratio signals report that activity-dependent acidification occurred in each compartment. Although we were unable to carry out pH calibrations due to the toxicity of nigericin and monensin to the cultured neurons, the responsivity of both sensors was validated at the end of the experiment by exposing the cells to NH4Cl. The activity-dependent pH responses observed in both the cytosol and mitochondria could indicate coupling of neuronal activity to mitochondrial ATP synthesis or uncoupling of pH compartmentation due to increased Ca2+ flux. For example, Azarias et al. saw glutamate-induced acidification in mitochondria of astrocytes, which they report was due to loss of cytosol-mitochondrial pH gradient. They also saw pH-dependent loss of metabolism in mitochondria.[46] Further calibration and studies with metabolic sensors would help us better understand the effect of increased neuronal activity on pH and neuron metabolism. Overall, in this proof-of-concept study, we validated that mCherryEA can be used to measure pH dynamics in cultured neurons. We were also able to observe activity-dependent pH changes in different compartments by expressing mCherryEA in the cytosol and ratiometric-pHluorin in the mitochondria of the same cell.

We next tested whether simultaneous imaging of mCherryEA and ratiometric-pHluorin could reveal compartment-specific differences in metabolism-dependent pH dynamics. Since the pKa of mCherryEA also makes it well suited for studying pH changes in the neutral to alkaline range in mitochondria, we set out to test the utility of mCherryEA in measuring mitochondrial pH dynamics. We confirmed the efficient localization of mCherryEA and ratiometric-pHluorin using confocal imaging to demonstrate that mCherryEA could be efficiently targeted to the mitochondrial matrix when co-transfected with cytosolic ratiometric-pHluorin (Figure ).

Figure 5

Confocal microscopy. Dual-compartment imaging with mito-mCherryEA and cyto-ratiometric-pHluorin. Scale bar is 20 μm.

The mCherryEA sensor showed excellent mitochondrial targeting when co-expressed, and therefore we next studied the metabolism-dependent pH changes in a neuroblastoma cell line using well-established inhibitors for glycolytic and oxidative phosphorylation pathways.[47,48]

Metabolism-Dependent pH Dynamics

Metabolism and pH are strongly coupled, with perturbations in the metabolic pathway known to cause pH changes.[1,48] In cancer cells, increased glycolytic activity is known to cause acidification of tumor extracellular environment and alkalinization of the cytosol.[10,12,49] Recent studies have also implicated a link between organelle pH and oncogenic signaling.[50,51] For example, Kondapalli et al. report increased luminal pH due to mutations in Na+/H+ exchanger (NHE9) in patient glioblastomas.[13] However, studies on compartment-specific pH changes in neuroblastoma due to metabolic changes are lacking.

To investigate this with our dual sensor imaging strategy, we studied pH changes in the Neuro2A mouse neuroblastoma cell line coexpressing ratiometric-pHluorin and mitochondria-targeted mCherryEA (Figures A–C and S-9). Cells were grown in the presence of glucose or in the absence of glucose, using galactose-supplemented media.[52] Cells were then imaged during sequential treatment with metabolic inhibitors to reveal pH changes linked to glycolysis versus mitochondrial respiration. To inhibit glycolysis, the cells were first exposed to iodoacetic acid (IAA). IAA inhibits the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase and is frequently used to inhibit glycolysis in a variety of cell types, including Neuro2A cells, as determined by measurement of ATP, reduced nicotinamide adenine dinucleotide, and other metabolic parameters.[30,53−57] To next inhibit mitochondrial electron transport and oxidative phosphorylation, cells were exposed to a combination of the mitochondrial membrane ionophore, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and the ATP synthase inhibitor, oligomycin.[58−60]

Figure 6

Metabolic inhibition causes differential acidification in the cytosol vs mitochondria. (A) Representative overlay image showing Neuro2A cells expressing cytosolic ratiometric-pHluorin and mitochondria mCherryEA. (B) Average pH change over time in the cytosol (green, dashed) and mitochondria (red). Treatment with the glycolytic inhibitor 1 mM iodoacetic acid (IAA) caused minor acidification in the cytosol with significant difference in the ratio before and after addition of IAA in individual cells. Although, no acidification was detected in mitochondria (asterisks) for the population, two of the three cells showed significant decrease. Treatment with the mitochondrial inhibitors 5 μM oligomycin and 1 μM FCCP (O/F) caused acidification in both compartments, and in particular, the cytosol exhibited large acidification. (D–F) The metabolism-dependent pH dynamics were independent of the sensor. (D) Representative overlay image showing Neuro2A cells expressing cytosolic mCherryEA and mitochondrial ratiometric-pHluorin. (E, F) Similar pH dynamics were observed when the localization of mCherryEA and ratiometric-pHluorin was switched. The sensors were validated at the end of the experiment by adding 30 mM NH4Cl and 10 mM acetic acid (n = 3, 30 cells total). Error bars are std. Scale bar is 10 μm. mCherryEA (red, λex = 575 nm, λem = 632 nm). ratiometric-pHluorin (green, λex = 475 nm, λem = 525 nm).

As expected, metabolic inhibition with 1 mM IAA caused a small decrease in pH in the cytosol, which is consistent with glycolytic inhibition. Although the pH change before (ratio = 1.39 ± 0.04, mean ± std, n = 3) and after (ratio = 1.48 ± 0.05) IAA addition is not significant for the population average (p = 0.28, n = 3), the pH change is significant (p < 0.05) in the cytosol of individual cells, highlighting the power of single-cell analysis to reveal differences obscured by ensemble means. The pH change in the mitochondria before (ratio = 0.90 ± 0.04) and after (ratio = 0.90 ± 0.02) IAA addition was not significant (p = 0.80) for the population average (n = 3), but two cells show significant pH decreases (p ≤ 0.015) (Figure B,C). Further calibration would be required to compare the difference in responses between the compartments, but we were clearly able to observe relative pH changes in each compartment. Subsequent treatment with 1 μM FCCP and 5 μM oligomycin (O/F) blocked oxidative phosphorylation and caused further acidification in the cytosol (before O/F ratio = 1.48 ± 0.05; after O/F ratio = 2.08 ± 0.17, p = 0.06, n = 3) and also in mitochondria (before O/F ratio = 0.90 ± 0.02; after O/F ratio = 1.07 ± 0.05, p = 0.02, n = 3). Individual cells showed a significant decrease in pH in both the compartments (p < 0.001). Although blocking the ATP synthase with oligomycin alone would increase mitochondrial pH, the presence of FCCP causes coupling between mitochondrial and cytosolic pH resulting in acidification of the mitochondria. Interestingly, the metabolism-dependent pH dynamics may be dependent on glycogen stores, fatty acid supply, or gluconeogenic amino acid supply because IAA also caused acidification when glucose is replaced with galactose.[52] The IAA-dependent acidification during galactose replacement indicates that extracellular glucose supply is not strictly required for the effect (Figure S-9). However, future metabolic studies will be needed to explore this further.

Importantly, the compartment-specific differences in pH dynamics are consistent with the compartmentation of metabolic processes and we validated that our observations are not an artifact of the sensors themselves. To do this, we switched the localization of the two sensors, now expressing mCherryEA in the cytosol and ratiometric-pHluorin in the mitochondria. We again observed that FCCP and oligomycin treatment caused the greatest change in pH of the cytosol and of the mitochondria compared with IAA treatment, confirming that the compartment-specific pH dynamics are independent of the specific sensor (Figure D–F). We cannot interpret the difference between cytosol and mitochondria dynamics because of the dampened dynamic range in mitochondria, but in the future, additional pH calibration of the sensor in mitochondria and other organelles will enable compartment-specific differences in pH dynamics to be quantified with respect to the physiological process under study.

These studies show the mCherryEA sensor can be used simultaneously with ratiometric-pHluorin to monitor coupled pH changes in two compartments. In addition, the mCherryEA mutant, a ratiometric red pH sensor with a pKa of 7.3, is well suited for quantitative imaging in neutral to alkaline organelles, such as cytosol (pH 7.2), endoplasmic reticulum (pH 7.2), mitochondria (pH 8), and peroxisome (pH 7).[61]

Discussion and Conclusions

We demonstrated here that mCherryEA is a ratiometric red fluorescent pH sensor that adds to the current toolbox of genetically encoded indicators available for multicolor live-cell imaging. An excitation ratiometric sensor exhibits two peaks in its fluorescence excitation spectrum, and for fluorescent protein-based sensors, the peaks are characteristic of the ionization states of the chromophore. The I158E/Q160A mutant of mCherry was originally engineered by Piatkevich et al. to shift the ionization state of the chromophore from the deprotonated to protonated form to generate a long Stokes shift (LSS) mutant.[25] Structural analysis indicated that an aspartate or glutamate at residue 160 could facilitate excited-state proton transfer (ESPT), and therefore the I160E mutation was made to create a proton-relay network between S158, E160, and the hydroxyl group of the chromophore.[25] ESPT mutants have been similarly engineered for LSS mutants of mCherry and other RFPs.[25,62] Piatkevich et al. also showed that the E160 interacts with solvent protons,[25] and we showed here that this interaction makes it a suitable solvent pH sensor. In a similar manner, pHluorin was engineered using the S202H mutation in wild-type green fluorescent protein (GFP). Although wild-type GFP exhibits excitation peaks characteristic of the protonated and deprotonated states, it does not exhibit a ratiometric shift in ionization in response to pH changes. Like the engineered glutamate in mCherryEA, the residue S202, which is involved in the proton-relay network, was mutated to histidine (pKa = 6.0) in pHluorin to increase its sensitivity to solvent pH. Thus, our work characterizing mCherryEA highlights the utility of engineering ESPT mutants of RFPs to generate new pH sensors.

Interestingly, in our characterization of mCherryEA, we found that despite the low quantum yield and extinction coefficient measured from the purified protein, the brightness of mCherryEA expressed in live cells was comparable to that of pHRed under same imaging conditions. Similar photophysical and photochemical differences between solution studies versus live-cell studies have been observed for other fluorescent protein biosensors, but the exact causes are not well understood.[27,28] We then compared the pH sensing characteristics of mCherryEA to that of two other single fluorescent protein-based pH sensors. We found that mCherryEA has a more alkaline pKa of 7.3 compared to that of both ratiometric-pHluorin and pHRed, making mCherryEA useful for monitoring pH changes in neutral to alkaline compartments. We also found that the dynamic range for mCherryEA (4-fold change) is comparable to that for ratiometric-pHluorin and several commonly used fluorescence resonance energy transfer (FRET) sensors, such as AKAR, EKAR, and JNKAR, which show 1.2–2-fold change in cells.[63−65] In the future, further mutagenesis could help improve the dynamic range of mCherryEA to be comparable to that of the multidomain green fluorescent pH sensor SypHer (10-fold change)[17,27] (Figure S-2) and other FRET-based sensors such as the calcium sensor Twitch (4–10-fold change).[66]

A major contribution of mCherryEA is to enable multicolor live-cell imaging because it is red fluorescent and should be compatible with green fluorescent sensors. As a proof-of-concept, we simultaneously imaged mCherryEA and ratiometric-pHluorin in different compartments in live neurons and neuroblastoma cells. We demonstrated that the negligible spectral cross talk between mCherryEA and ratiometric-pHluorin makes it possible to measure pH changes in the cytosol and mitochondria of the same cell. We observed relative changes in compartment-specific pH that correlated with neuronal activity and metabolic inhibition. However, because the sensors show decreased dynamic range in mitochondria, we could not directly compare the differences in pH between compartments. To compare pH changes between organelles, it is necessary to carry out an in situ pH calibration at the end of each experiment on a cell-by-cell basis, as demonstrated by Poburko et al. for example.[17] We demonstrate that in situ calibrations can be done with cytosolic or mitochondrial-targeted mCherryEA using nigericin and monensin. However, we also found that nigericin and monensin can cause extensive cell death after long metabolic manipulations or with primary neuron cultures, precluding our ability to carry out in situ calibrations in some scenarios. In the absence of calibrated measurements, mCherryEA can still be used to observe relative changes in pH within a compartment.

In conclusion, we demonstrated that the mCherryEA mutant is a red fluorescent, ratiometric pH sensor with pKa of 7.3, which is more alkaline than other red fluorescent pH sensors such as pHRed. It can be used to measure compartment-specific pH changes simultaneously with green fluorescent pH sensors such as ratiometric-pHluorin in several different cell types, and in principle, it is spectrally compatible with sensors of other analytes that utilize blue, cyan, green, or yellow fluorescent proteins.

Methods

Materials

Unless otherwise stated, chemicals were purchased from Sigma-Aldrich; molecular biology enzymes were purchased from New England Biolabs (NEB); and cell culture media and supplements were purchased from Invitrogen. Neuro2A and HEK-293 were purchased from ATCC, C57BL/6, and FVB mice were purchased from Charles River Laboratories. MitoTracker Deep Red FM was a kind gift from Dr. Qing Deng at Purdue University.

Molecular Biology

pRSETb-mCherry(wt) and GW1-mCherry(wt) were mutated using the NEB Q5 site-directed mutagenesis kit to generate the mCherry(I158E/Q160A) mutant. Mitochondria-targeted mCherryEA was cloned by fusing a tandem 4xCoxVIII signal sequence[67] to the N-terminus. GW1-ratiometric-pHluorin and GW1-SypHer were generated by subcloning ratiometric-pHluorin from VV064: 1xCox8-ratiometric-pHluorin and SypHer from SypHer-mt into GW1 vector using NEB HiFi reactions. VV064: 1xCox8-ratiometric-pHluorin in the FCK vector was a gift from Adam Cohen (Addgene plasmid # 58502), and SypHer-mt was a gift from Nicolas Demaurex (Addgene plasmid # 48251).

Protein Expression and Purification

pRSETb-mCherryEA was transformed into BL21(DE3) cells and grown in 500 mL of autoinduction media (ForMedium) at 37 °C overnight and transferred to 4 °C for 3 days. Protein was purified using HisTrap Nickel columns (GE Healthcare) according to manufacturer instructions. Purified protein was dialyzed into storage solution (5 mM 3-(N-morpholino)propanesulfonic acid (MOPS), 150 mM NaCl, and 5% glycerol, pH 7.4) and stored at −80 °C.

Steady-State Fluorescence Spectroscopy

pH titration was performed in solutions containing 10 mM each of Bis-Tris, MOPS, and Tris plus 100 mM NaCl adjusted to pH values ranging from 5.5 to 9. Protein samples were diluted to 1–2 μM, and fluorescence was measured in a microplate reader (BioTek Synergy H5). The excitation spectrum was measured using a monochromator set to scan from 350 to 600 nm, with emission set at 630 nm. The fluorescence was normalized by calculating the ratio of excitation peaks at 585 by 450 nm. To test for environmental interference, the mutants were titrated in buffered solutions containing either 100 mM NaCl, KCl, or K-gluconate and the addition of 1 mM MgCl2, 1 mM CaCl2, 1 mM H2O2, and 3 mM DTT. For temperature dependence, the fluorescence measurement was taken with the reader set to 23, 25, 31, and 37 °C.

Cell Culture and Transfections

Neuro2A (ATCC CCL-131) and HEK-293 (ATCC CRL-1573) cells were cultured at 37 °C in a 5% CO2 humidified air incubator in Dulbecco’s modified Eagle’s medium (DMEM) media containing 10% cosmic calf serum (HyClone). Cells were transfected using Effectene (Qiagen) as per manufacturer’s instructions and imaged after 2 days.

Neuron and Astrocyte Cultures

All animal procedures were performed in strict accordance with recommendations provided in the National Institutes of Health Guide for the Care and Use of Laboratory Animals, according to protocols approved by the Purdue Animal Care and Use Committee and the Purdue University Laboratory Animal Program to minimize pain and suffering. Cortical and hippocampal neurons were isolated from P0 mice and maintained in Neurobasal media supplemented with 5–25 mM glucose, 0.2 mM pyruvate, 0.5 mM GlutaMAX, B-27, penicillin, and streptomycin (Pen/Strep). Neurons were transfected after 7–9 days using calcium phosphate method.[68] Cortical astrocytes were isolated from P0 to P4 mice and maintained in DMEM/F12 media supplemented with 10% fetal bovine serum and Pen/Strep. Astrocytes were transfected using Lipofectamine 3000 (Invitrogen), as per manufacturer’s instructions.

Extinction Coefficient (ε) and Quantum Yield (QY)

The concentration of protein containing mature chromophore was quantified by measuring the absorbance of the protein in 1 M NaOH at 450 nm (ε: 44 000 M–1 cm–1), as previously described.[18] Twenty-eight percent of purified mCherryEA corresponded to mature chromophore. The concentration from alkaline denaturation method was used for ε and QY measurements. For ε measurement, absorbance and fluorescence spectra of the protein at dilutions of 5–20 μM were measured. The ε was calculated according to the Beer–Lambert equation. For QY measurement, slopes from absorbance versus fluorescence at different pH were measured with wild-type mCherry as the standard for excitation at 530 nm (QY = 0.22). QY at 440 nm was calculated relative to 530 nm QY.

Sensor Characterization in Live Cells

NH4Cl: pH response of the sensors in live cells was tested by adding 10 mM NH4Cl to the imaging solution (mM: 15 N-(2-hydroxyethyl)piperazine-N′-ethanesulfonic acid (HEPES), 1.25 NaH2PO4, 10 glucose, 120 NaCl, 3 KCl, 2 CaCl2, 1 MgCl2, and 3 NaHCO3, pH 7.3). Cytosol pH calibration: Neuro2A cells transfected with mCherryEA or ratiometric-pHluorin were seeded in a multiwell plate and exposed to different pH ranging from 5.5 to 9 in a high-potassium solution (mM: 15 HEPES, 1.25 KH2PO4, 10 glucose, 123 KCl, 2 CaCl2, 1 MgCl2, pH 7.3) in the presence of 2–5 μM nigericin.[22,27,60] Mitochondria pH calibration: Neuro2A cells transfected with mito-mCherryEA or mito-pHluorin seeded in a multiwell plate were exposed to pH 5.5–9 in a high-potassium solution (mM: 15 HEPES, 10 glucose, 123 KCl, 20 NaCl, 2 CaCl2, 1 MgCl2) in the presence of 5 μg/mL nigericin and 5 μM monensin. For individual cell calibration, Neuro2A cells were treated with 10 mM NH4Cl. After washing the cells with imaging solution, a three-point calibration was performed with high-potassium solutions buffered at pH 6.0, 7.5, and 8.0.

Live-Cell Imaging

Cells were imaged using an Olympus IX83 fluorescence microscope with a 20×/0.75 NA and 60×/1.35 oil objective illuminated by a Lumencor SpectraX light engine and equipped with an Andor Zyla 4.2 sCMOS camera. Lumencor power levels were typically set at 5–10% for each ratio channel for cytosolic sensor and 15–30% for mitochondrial sensor. mCherryEA and pHRed were excited using 575/25 and 438/29 nm band-pass filters, and emission was collected through a 632/60 nm band-pass filter. Ratiometric-pHluorin was excited using 475/34 and 395/25 nm band-pass filters, and emission was collected through a 525/50 nm band-pass filter. SypHer was excited using 475/34 and 438/29 nm band-pass filters, and emission was collected through a 525/50 nm band-pass filter. The exposure time was set between 50 and 200 ms for all experiments, and fluorescence signals were at least 3-fold above the background in all fluorescence channels. For mCherryEA and pHRed comparison, cells expressing each sensor were imaged on the same days under the exact same illumination and imaging conditions (exposure times, light-emitting diode power, etc.). Neurons were perfused at 1 mL/min with artificial cerebrospinal fluid (mM: 15 HEPES, 120 NaCl, 3 KCl, 2 CaCl2, 1 MgCl2, 3 NaHCO3, 1.25 NaH2PO4, 5 glucose, 0.2 pyruvate, and 0.5 glutamax, pH 7.4). To confirm targeting of GW1-4xCox8-mCherryEA, 10 nM MitoTracker Deep Red (Invitrogen) was added to Neuro2A cells expressing mito-mCherryEA and treated as per manufacturer’s instruction. High-magnification images were captured using the 60×/1.35 oil objective. MitoTracker Deep Red was excited using 631/28 nm band-pass filter and emission was collected through a ET706/95 nm band-pass filter.

Data Analysis

Images were acquired with IQ (Andor) and analyzed with ImageJ. The mean and background intensities were measured for the images. The ratio was calculated by dividing the background-subtracted means of images excited at the higher wavelength by the lower wavelength. For pixel-by-pixel measurement, fluorescence images were background-subtracted and a threshold was set to reject background pixels. Ratio images were obtained by dividing the images captured by exciting at the higher and the lower excitation wavelengths.