Single Enzyme Experiments Reveal a Long-Lifetime Proton Leak State in a Heme-Copper Oxidase

Heme-copper oxidases (HCOs) are key enzymes in prokaryotes and eukaryotes for energy production during aerobic respiration. They catalyze the reduction of the terminal electron acceptor, oxygen, and utilize the Gibbs free energy to transport protons across a membrane to generate a proton (ΔpH) and electrochemical gradient termed proton motive force (PMF), which provides the driving force for the adenosine triphosphate (ATP) synthesis. Excessive PMF is known to limit the turnover of HCOs, but the molecular mechanism of this regulatory feedback remains relatively unexplored. Here we present a single-enzyme study that reveals that cytochrome bo3 from Escherichia coli, an HCO closely homologous to Complex IV in human mitochondria, can enter a rare, long-lifetime leak state during which proton flow is reversed. The probability of entering the leak state is increased at higher ΔpH. By rapidly dissipating the PMF, we propose that this leak state may enable cytochrome bo3, and possibly other HCOs, to maintain a suitable ΔpH under extreme redox conditions.


units ("pH jump"). Data is shown for WT cytochrome bo 3 (top right), E286C cytochrome bo 3 (bottom left) and liposomes without protein (bottom right).
For the pH jump experiments, the pH was reduced by adding 0.1 mL of an acidified electrolyte solution (20 mM MOPS, 30 mM Na 2 SO 4 , pH <6.5) to 0.6 mL solution in the cell (20 mM MOPS, 30 mM Na 2 SO 4 , pH 7.4) to give the final pH of ~6.9, which was measured after the experiment. Due to the natural distribution of the intravesicular pH of single lipid vesicles, a distribution in ΔpH is formed in the pH jump experiments. For the pH jump experiments, the proteoliposomes (top right and bottom left) were prepared with higher protein-to-lipid ratio (1% w/w) than the single enzyme experiments to ensure that almost all proteoliposomes contained protein. The data in the top left graph is the same as in Figure 5 (left) of the main text, but limited to the data points with ΔpH below 1 unit. Importantly, the data of the top left plot also contains single-enzyme data where cytochrome bo 3 has "stalled", but not entered a statistically significant leak state (e.g., with leaking rates close to zero). In contrast, Figure 5   The sequence of movies taken from the same electrode Figure S4 Distribution of the lifetime of the proton uptake/release activity of WT cytochrome bo 3 in absence (left) and presence (right) of valinomycin, and for E286C cytochrome bo 3 (middle). Electrochemical potential to reduce the ubiquinone pool was only applied for 100s, limiting the measured proton uptake/release lifetime in this study to 100s. Experimental conditions are described in the Material and Methods section in the main text, as is the data analysis providing the proton translocation rates.     Figure S9 Examples of full-scale images. The HPTS fluorescence image is from the HPTS channel with the excitation peaked at 410nm (top) and ATTO633 (bottom). In this pair of images, only the central part of the images are in sharp focus. Vesicles out of focus appear to be larger and will be rejected (see the Material and Methods). We propose that some vesicles are out of the focal plane while others are in-focus because: 1) the glass-backed gold electrode may have uneven thickness across the field of view (179-by-151 µm) despite being very smooth locally; 2) the electrochemical cell is not machined precisely enough to provide a perfectly flat support for the electrode which is also perfectly aligned to the optics of the microscope; 3) the microscope stage itself may not be perfectly perpendicular to the light path.

HPTS fluorescence
ATTO fluorescence S9 Figure S10 The HPTS-pH calibration curve. The fitting of the pH calibration curve is shown in this figure. Error bars for the data represent the standard error of the mean at each pH point. The data fit to the following formula: The black solid curve is the calibration curve. The green dashed curve is the curve calculated using the parameters at the lower bounds of the 95% confidence interval, and the red dashed curve is the curve of parameters at the higher bounds. The fitted parameters are given in Table  S1. (proteo)liposome ratio, the chance that two (proteo)liposome both containing an enzyme being close together is even lower. Therefore, we propose that the vast majority of the analyzed spots is due to separated single vesicles on the electrode.

Errors and confidence levels, pH calibration procedure
The fitting of the pH calibration curve returns a range for each parameter with the 95% confidence level ( Figure S10 and Table S1). The authors compared the results from using the parameters at the lower/higher bound to those from using the fitted parameter, and found negligible differences. The proton translocation and leaking rates are sensitive to the differences in pH (converted from HPTS fluorescence). Therefore they are good indicators for how sensitive S11 the data analysis is to the pH calibration. When using HPTS-pH parameters at their lower or higher bounds, the differences in proton translocation and leaking rates are statistically insignificant. The null hypothesis is not rejected in all of the 2-sample t-tests between each pairs at the 95% confidence level. The final part of the data analysis procedure assigns different types to each vesicle depending on the fitting results, e.g., "empty vesicle/no protein", "active protein, outward facing orientation, no leak state", "active protein, inward facing orientation, entered leak state", etc. Using the HPTS-pH parameters at the lower or higher bounds does not affect the assigned types. When using parameters at the lower bound, 0.43% of vesicles were assigned differently; when using parameters at the higher bound, 0.86% of vesicles were assigned differently. These levels of differences should be considered negligible. Therefore, the data analysis procedure is robust against uncertainties in the HPTS-pH conversion.

Errors and confidence levels, fitting procedure
The other source of the uncertainty is the noise in the time-traces, which includes noise from the camera, possibly resulting in false proton translocation/leaking events. The data analysis procedure written in MATLAB in this study takes into account the noise level of individual timetraces when fitting the time-traces. As described in Material and Methods, the fitting procedure does not over-fit the time-traces (fitting too many joint segments to the data) because it rejects a fit if the difference between the fit and raw data has a standard deviation lower than 110% of the independently determined standard deviation. The latter is determined for the individual vesicles by analyzing the trace before the electrochemical potential is switch on and it is assumed that the pH inside the vesicle is constant during this time. 110% was chosen instead of 100% to make this control more stringent. After the fitting is accepted based on the standard deviation, a best model is chosen based on the adjusted R 2 values because different models have different degrees of freedom (related to fitted parameters and number of data points) which adjusted R 2 takes into account. Hence, noisier traces are fitted more "conservatively" (i.e., less likely to have a proton transport, leaking, or stalling event assigned to it).
In addition to this, there is a separate statistical test to ensure that proton translocation/leaking events are not caused by random noise by analyzing the linearly fitted segments with a nonparametric Kendall test at the 95% confidence level. A non-parametric test was chosen for their increased robustness because they do not assume a distribution of the input data. Although it is safe to assume that the noise follows a normal distribution, the Kendall test was used to increase the robustness of the analysis (at the expense of sensitivity). If a Kendall test "accepts" a segment (with a rate), the number of protons that are transported in time are higher than could be caused by random noise (at the used confidence level). This is interpreted as a "real" proton S13 translocation/leaking event. In other words, uncertainty in the final conclusion is carefully controlled at each stage of the data analysis procedure and at each stage a 95% confidence level is used with a model that disfavors false positives (at the expense of sensitivity). Because a 95% confidence level was used in the Kendall test and because HPTS-pH calibration within the 95% confidence interval generates the same results, 95% confidence level can be considered an indicator of the statistical stringency of the data analysis procedure. As described in the Material and Methods, other confidence levels were tested for the Kendall test (99% and 90%) and the results are qualitatively identical. In conclusion, for the numerical measurements in this study, e.g., proton translocation rate, a 95% confidence interval represents the uncertainty.

Activity of cytochrome bo 3 and the effect of valinomycin
Wild-type cytochrome bo 3 Figure S4).

Independent verification of experimental setup and data analysis
The experimental setup and data analysis was verified by a set of independent experiments where we changed several of the methods. In this Supporting Information are the methods of these independent experiments, the results of which are shown in Figure S11 and S12 below. The methods differ from those in the main manuscript in the following: reconstitution protocol, pHsensitive dye, microscopic technique, electrode fabrication, the addition of valinomycin and the data analysis procedure. The independent experiments show the same leak state as discussed in the main manuscript.

Protein purification and reconstitution
Wild-type cytochrome bo 3 was purified from GO105/pJRhisA. 3 Proteins were purified similarly as previously reported. 4 Cytochrome bo 3 concentration was determined by its Soret band at 409 nm (ε = 188 mM -1 cm -1 ). Protein was reconstituted following a method published previously, 3 except that the pH sensitive dye, 0.88 mM SNARF-1 (Life Technologies) was used. The lipid mixture used for the reconstitution was E. coli polar lipid extract (Avanti Polar Lipids, Inc., Alabama, U.S.A.) mixed with the following: 1% (w/w) ubiquinone-10 and 0.1% (w/w) ATTO-633 labeled DOPE (ATTO-TEC GmbH, Germany) and 0.1% protein (w/w). The protein-to-lipid weight percentage added for the reconstitution is 0.2%.
The SAM layer was made by incubating the gold electrodes in 1 mM 6-MH solution in isopropanol overnight at 20° C. (Proteo)liposomes were sparsely absorbed to the gold electrode by incubating the gold surface with a diluted suspension (2 µg/mL lipid) for 30 min. at 20° C followed by a thorough washing step to remove unbound (proteo)liposomes.
The samples were examined with a Leica TCS SP5 inverted confocal microscope using an oil SNARF ratio-to-pH calibration. This cross emission is expected to be dependent on the diameter and thus the number of labeled lipids per liposome. To correct for this we formed liposomes without SNARF-1, but otherwise identical conditions to the experiments. We found no cross emission on the APD1 channel and a minimal cross emission on the APD2 channel (Fig. S13).
We used this data to correctly calculate the ratio of the two emission wavelengths of SNARF-1.
An automated algorithm for signal processing of proton count time traces was developed. Noise was removed from the traces by employing low-pass filtering, with a given number of coefficients (≅ number of points in the trace) and cut-off frequency. To find the cut-off frequency for filtering we examined control proton count traces (liposomes without proteins and S18 proteoliposomes without applied voltage). Peaks above a given amplitude threshold value were detected in a second derivative of the filtered proton count traces. Center position of peak is referred to as a rate change point (RCP). The amplitude threshold value for peak detection was identified by examining control proton count traces using various values of amplitude thresholds and finding the one that allowed the algorithm to detect peaks in <10% of control traces. The rate of proton uptake/release was determined by fitting a line between two neighboring RCPs in the filtered proton count trace. Finally, proton count traces reflecting an active proton uptake/release were selected by applying a slope threshold, the value of which is above 90% of slopes detected in the control traces.
In addition to the independent experiments with different methods, a further analysis was done on the data shown in the main text to check that leaking events are unable to form ΔpH, as expected. If the leak event is due to a second enzyme with opposite orientation, the second enzyme should be able to form a ΔpH of opposite sign. For example, for an enzyme orientated such that protons are taken up from the proteoliposome, the pH inside the vesicle increases.
Upon entering the leak state, the protons leak back into the proteoliposome and the pH decreases again. To confirm that this decrease in pH is not due to a second enzyme in the proteoliposomes that releases protons to the inside, it was checked that the final pH is not lower than that measured prior to the start of the experiment. We found that only 0.2% of proteoliposomes have such behavior, compared to the leaking relative frequency of 7.2%. We conclude that oppositely oriented protein complexes cannot account for the observed leaking events.

S19
As shown in Figure S11 and S12 below, the leak state and the effect of valinomycin were reproduced in the independent experiment performed using different materials and methods. In Figure S11, it is also clear that the leak rates are higher than the proton uptake/release rates of cytochrome bo 3 and the proton leakage due to the proton permeability of the (proteo)liposome, further confirming the observed leak state is not due to proteoliposomes with multiple enzymes that are oppositely orientated or the lipid membrane itself. The same conclusion was drawn from  In the pH jump experiment, liposomes were prepared identical to the proteoliposomes but without cytochrome bo 3 . After the immobilization of liposomes on the surface, a ΔpH was formed by increasing the pH of the solution by about 1 pH unit. Leak rates were monitored as for proton translocation by cytochrome bo 3 . All methods are described in the "Independent verification" section.