2.1.

Cell Cultures and Experiments

Human osteosarcoma 143B and HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Invitrogen Corp., Carlsbad, California) containing $100\text{units}∕\mathrm{ml}$ penicillin G, $100\mu \mathrm{g}∕\mathrm{ml}$ streptomycin sulfate, $0.5\mu \mathrm{g}∕\mathrm{ml}$ amphotericin B, and 5% fetal bovine serum (FBS) (Biological Industries, Kibbutz Beit Haemek, Israel) at $37°\mathrm{C}$ in a humidified atmosphere with 5% $\mathrm{C}{\mathrm{O}}_2$ . At $24\mathrm{h}$ before fluorescence lifetime imaging and drug treatments, cells at a density of $2\times {10}^4\text{cells}∕{\mathrm{cm}}^2$ were seeded onto $24\mathrm{mm}$ diameter round glass coverslips (Paul Marienfeld GmbH & Co., Lauda-Konigshofen, Germany) that had been coated with FBS. These coverslips were then kept in dishes and cultured in DMEM inside an incubator for $24\mathrm{h}$ . After $24\mathrm{h}$ , cells were completely attached onto the coverslip, ready to grow (in the early log phase of cell proliferation curve), and ready for fluorescence imaging. Immediately before fluorescence imaging, cells were washed twice using phosphate-buffered saline solution and then transferred to a cell chamber designed for viewing live cell specimens. A 1 mL aliquot of 5 mM HEPES ( $N$ -2-hydroxyethylpiperazine- $N^{\prime }$ -2-ethanesulfonic acid) buffer ( $5\mathrm{mM}$ KCl, $140\mathrm{mM}$ NaCl, $2\mathrm{mM}$ $\mathrm{Ca}{\mathrm{Cl}}_2$ , $1\mathrm{mM}$ $\mathrm{Mg}{\mathrm{Cl}}_2$ , $10\mathrm{mM}$ glucose, pH 7.4) was added into the cell chamber to nourish cells and to avoid light absorption of the red color of the culture medium. Fluorescence lifetime images of grouped cells were acquired at 1 to 3 sites per coverslip before treatment. The field of view (FOV) of each image was $100\times 100\mu \mathrm{m}$ . The FLIM system was equipped with a microscope cage incubator (Oko-lab, Naples, Italy) to maintain the optimal temperature during experiments. To optimize the NADH fluorescence intensity and to acquire high photon counts for statistical processing of the data, the imaging was conducted at room temperature $(20°\mathrm{C}\text{to}23°\mathrm{C})$ .³¹ Additional measurements were performed at the physiological temperature $(35°\mathrm{C}\text{to}37°\mathrm{C})$ in both controls and STS-treated cells to ascertain that the results of NADH lifetime measurements were not affected by the room temperature conditions.

Our previous experience with STS-induced apoptosis showed that the activation of caspase 3 occurred at $8\mathrm{h}$ or later after $25\mathrm{nM}$ STS treatment and the apoptotic bodies appeared at $24\mathrm{h}$ after treatment.³² Maeno ²⁹ showed that the release of cytochrome $c$ from mitochondria and the activation of caspase 3 occurred at $\sim 2\text{to}4\mathrm{h}$ after $4\mu \mathrm{M}$ STS treatment of HeLa cells. Thus we randomly chose a medium dose $\left(1\mu \mathrm{M}\right)$ of STS (Sigma-Aldrich, St. Louis, Missouri). Hydrogen peroxide is known to induce both apoptosis and necrosis depending on the concentration (e.g., $10\text{to}100\mu \mathrm{M}$ for apoptosis, and $1\text{to}10\mathrm{mM}$ for necrosis) used.³⁰ We chose to use $1\mathrm{mM}$ ${\mathrm{H}}_2{\mathrm{O}}_2$ (Sigma-Aldrich) for necrosis induction. Time-lapsed fluorescence lifetime images were obtained at the same site (same FOV) before, immediately after $\left(0\text{to}15\min \right)$ , and up to $10\mathrm{h}$ after treatment with STS or ${\mathrm{H}}_2{\mathrm{O}}_2$ . Controls are fluorescence lifetime images of cells without and before treatment.

2.2.

Caspase 3 Activity Assay

Cells were disintegrated in $100\mu \mathrm{l}$ lysis buffer [ $12.5\mathrm{mM}$ Tris-HCl, $1\mathrm{mM}$ dithiothreitol, $0.125\mathrm{mM}$ ethylenediaminetetraacetic acid (EDTA), 5% glycerol, and an aliquot of complete protease inhibitor mixture (Roche Applied Sciences, Mannheim, Germany), pH 7.0] on ice for $30\min$ and centrifuged at $9000g$ for $10\min$ at $4°\mathrm{C}$ . A $50\mu \mathrm{g}$ aliquot of protein was incubated with $20\mu \mathrm{M}$ Ac-DEVD-AFC (Calbiochem, San Diego, California), a fluorescent substrate of caspase 3, in 500 μl of assay buffer [50 mM Tris-HCl, 1 mM EDTA, and 10 mM EGTA (ethyleneglycol-bis-(β-aminoethylether)- $N$ , $N$ , $N^{\prime }$ , $N^{\prime }$ -tetraacetic acid), pH 7.0] at $37°\mathrm{C}$ for $30\min$ in the dark. The fluorescence intensity was determined by spectrofluorometry (Hitachi F-3000, Tokyo, Japan) at an excitation wavelength of $380\mathrm{nm}$ and an emission wavelength of $508\mathrm{nm}$ as described previously.^{32, 33}

2.3.

NADH FLIM

Time-domain FLIM was performed with a $60\times 1.45$ numerical apenture PlanApochromat oil objective lens (Olympus Corp., Tokyo, Japan) on a modified two-photon laser scanning microscope (FV300 with the IX71 inverted microscope, Olympus Corp.) as described previously.³⁴ In this study, samples were excited at $750\mathrm{nm}$ (two-photon) by a mode-locked Ti:Sapphire Mira F-900 laser, pumped by a solid-state continuous wave $532\mathrm{nm}$ Verdi laser (both from Coherent Inc., Santa Clara, California). The scanning speed of the FV300 was controlled externally by a function generator (AFG310, Tektronix Inc., Beaverton, Oregon).

Fluorescence photons were detected in a non-descanned mode by a photon-counting photomultiplier (H7422P-40; Hamamatsu Photonics K.K., Hamamatsu, Japan). Time-resolved detection was conducted by the single-photon-counting SPC-830 printed circuit board (Becker & Hickl GmbH, Berlin, Germany). A bandpass filter of $450\pm 40\mathrm{nm}$ (Edmund Optics Inc., Barrington, New Jersey) was inserted in the emission path for the detection of NADH that emits with the maximum at $450\mathrm{nm}$ .³⁵ Additional short pass and IR cutoff filters were used to reject reflected or scattered excitation light at $750\mathrm{nm}$ . The average laser power measured at the focal plane of the objective was $\sim 3\text{to}5\mathrm{mW}$ , which was lower than the reported laser power of two-photon damage³⁶ and was found to be optimal for the prevention of photobleaching. All the images were taken at $256\times 256\text{pixels}$ resolution with the acquisition time in the range of $900\text{to}1800\mathrm{s}$ for enough photon count statistics at the given laser power for further data analysis. To compare the emitted fluorescence intensity between controls and treated cells, we adapted the photon count from the peak time channel (and with fixed binning parameters) of the recorded decay curve at the brightest point of a FLIM image.

2.4.

FLIM Data Analysis

Data were analyzed with the commercially available SPCImage version 2.8 software package (Becker & Hickl GmbH) via a mathematical convolution of a model function and the instrument response function (IRF), and fitting to the experiment data. To calculate the lifetime from the composite decays of NADH, we convolved an IRF, $I_{\mathit{instr}}$ , with a double-exponential model function, defined in Eq. 1, with offset correction for the ambient light and/or dark noise $I_0$ to obtain calculated lifetime decay function $I_c\left(t\right)$ in Eq. 2

The model parameters (i.e., $a_i$ and ${\tau }_i$ ) were derived by SPCImage software by fitting the calculated decay $I_c\left(t_k\right)$ , defined in Eq. 2, to the actual data $I_a\left(t_k\right)$ through minimizing the goodness-of-fit ${\chi }_R^2$ function defined in Eq. 3 using the Levenberg-Marquardt search algorithm

3.1.

NADH Fluorescence Intensity and Lifetime Images of Intact Cells

Figure 1 shows representative images, fitting curves, and lifetime distribution from control 143B cells. It includes a fluorescence intensity image [Fig. 1a], fluorescence lifetime image [Fig. 1b] using two-photon FLIM, the corresponding fluorescence decay curves (measured: blue; fit: red), the fitting residuals [Fig. 1c], and normalized lifetime histograms [Fig. 1d] of the $256\times 256$ lifetime image as shown in Fig. 1b. This entire image was acquired over a period of $1300\mathrm{s}$ . In control cells, the average photon count collected over the given period of time was approximately 600 in the peak time channel.

Fig. 1

Measurement of the intensity and average lifetime of NADH fluorescence in 143B cells. The intensity and average lifetime images of NADH fluorescence in 143B cells are shown in gray (a) and color scale (b), respectively, using two-photon FLIM with the excitation wavelength at $750\mathrm{nm}$ . The FOV of each image is $100\times 100\mu \mathrm{m}$ . The scale bar in (a) is $20\mu \mathrm{m}$ . A representative fluorescence lifetime decay curve (blue curve in the top panel), the fit (red curve in the top panel), and the residuals (bottom panel) are shown (c). The normalized histograms over the 256 x 256 pixel lifetime image shown in (b) were plotted for ${\tau }_1$ , ${\tau }_2$ , and ${\tau }_{\mathrm{ave}}$ (d).

Figure 1a shows that the fluorescence signals in the cytoplasm displayed punctuate perinuclear patterns, which were attributed to mitochondria-associated NADH.^{19, 38} No fluorescence was seen in the nuclei or on the nuclear membrane. In Fig. 1b, each pixel represents the average lifetime $\left({\tau }_{\mathrm{ave}}\right)$ of the short $\left({\tau }_1\right)$ and the long $\left({\tau }_2\right)$ lifetime components weighted by their relative contributions $a_1$ and $a_2$ , respectively, such that ${\tau }_{\mathrm{ave}}=(a_1{\tau }_1+a_2{\tau }_2)∕(a_1+a_2)$ . The scale of the color mapping was chosen for later comparison with FLIM images of treated cells so that the blue color represents the longer lifetime (maximum $4000\mathrm{ps}$ ) and the red color presents the shorter lifetime (minimum $200\mathrm{ps}$ ). Overall, the NADH fluorescence in the control 143B cells exhibited an average lifetime $\left({\tau }_{\mathrm{ave}}\right)$ of $\sim 1300\mathrm{ps}$ , which corresponds to the yellowish color in the selected color scale. The lifetime across the entire FOV or even within a single cell was not homogeneous: some pixels tend to be greenish and some tend to show orange color. Figure 1c demonstrates the fitting curve obtained from Fig. 1b, with the goodness-of-fit residuals shown in the bottom panel of Fig. 1c. Finally, the distribution of the averaged lifetime $\left({\tau }_{\mathrm{ave}}\right)$ from all pixels shown in Fig. 1b is presented in Fig. 1d, which depicts the normalized histogram $\left(H_{\mathrm{norm}}\right)$ of ${\tau }_{\mathrm{ave}}$ . The histogram shows a peak $\left({\tau }_{\text{peak}}\right)$ at $\sim 1300\mathrm{ps}$ with a full width at half maximum $\sim 500\mathrm{ps}$ . This broad distribution illustrates the inhomogeneous color distribution observed in Fig. 1b. To compare with previously published results regarding the values of the NADH short $\left({\tau }_1\right)$ and long $\left({\tau }_2\right)$ lifetime components, we obtained ${\tau }_1$ and ${\tau }_2$ from the entire image pixels by SPCImage software and plotted their normalized histogram in the same figure [Fig. 1d]. The histograms of ${\tau }_1$ and ${\tau }_2$ showed peaks at $\sim 600$ and $\sim 3000\mathrm{ps}$ , respectively. These results are close to the previously published lifetime values of free $(\sim 400\text{to}500\mathrm{ps})$ and bound NADH $(\sim 2000\text{to}3000\mathrm{ps})$ .^{22, 23} The observation that ${\tau }_2$ had a broader distribution than ${\tau }_1$ may be attributed to the wide distribution of lifetimes of NADH molecules bound to different proteins.^{39, 40}

We acquired NADH lifetime images from a total of 35 sites of control HeLa cells (from 15 coverslips) and 11 sites of control 143B osteosarcoma cells (from 8 coverslips). The values of ${\tau }_{\text{peak}}$ from all of the lifetime images were analyzed, recorded, and averaged. Table 1 summarizes the total number of sites $\left(N_{\text{site}}\right)$ and the corresponding $\text{mean}\pm \text{standard}$ error (SE) of the ${\tau }_{\mathit{\text{peak}}}$ values from control HeLa cells and control 143B cells. The results showed that HeLa and 143B cells had the average ${\tau }_{\text{peak}}$ values equal to $1360\pm 24$ $(N=35)$ and $1260\pm 49\mathrm{ps}$ $(N=11)$ , respectively. The two mean values are not significantly different as judged by two-tailed Student’s $t$ test $(p\text{value}=0.09)$ . The mean value of all control ${\tau }_{\text{peak}}$ values is $1336\pm 22\mathrm{ps}$ , which will be used later in Figs. 4 and 6 for comparison.

Table 1

Summary of the number of sites (Nsite) of NADH lifetime images acquired from the number of coverslips (Ncoverslip) in control, STS-treated, and H2O2 -treated cells. The number of sites is the same as the number of coverslips in treated cells because of time sequence measurements at the same site. The corresponding means (±SE) of the peaks (τpeak) of the average lifetime (τave) distributions were calculated and listed with the unit of picoseconds (ps). τpeak was defined as the peak position (or the maximal value) of the τave histogram from each 256×256 pixel lifetime image. For STS-treated cells, only the lifetime at the first time point (0to15min) was listed. For H2O2 -treated cells, the lifetimes at both 0 to 15 and 0to25min after treatments ( Nsite at each time point is 2) were averaged and listed to account for a larger number of the samples for averaging.

3.2.

Cell Morphological Change and Caspase 3 Activation Confirmed STS-Induced Apoptosis and $H_{\mathit{2}}{O}_{\mathit{2}}$ -Induced Necrosis

Although STS-induced apoptosis and high dose ${\mathrm{H}}_2{\mathrm{O}}_2$ -induced necrosis have been well documented, we observed the cell morphology and measured caspase 3 activity to confirm the cell death pathway at the drug dose we chose. Figure 2a shows the light microscopic images of HeLa cells $15\min$ and $5\mathrm{h}$ after either ${\mathrm{H}}_2{\mathrm{O}}_2$ or STS treatment. All the images were taken at the same magnification. Cell swelling or eruption, a typical feature when cells are under necrotic pathway, was observed in both $15\min$ and $5\mathrm{h}$ after ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment. In contrast, the typical feature of apoptotic cells, cell shrinkage, was observed in STS-treated cells. Figure 2b shows the normalized caspase 3 activities of HeLa cells before (controls), $15\min$ after, and $2\mathrm{h}$ after either ${\mathrm{H}}_2{\mathrm{O}}_2$ or STS treatment. Caspase 3 activity was significantly higher at $2\mathrm{h}$ after STS treatment compared with controls, indicating that STS-treated cells were undergoing apoptosis. However, no significant increase of caspase 3 activity was observed in cells at $15\min$ or $2\mathrm{h}$ after ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment.

Fig. 2

The morphology (a) and normalized caspase 3 activity (b) of HeLa cells after treatment with ${\mathrm{H}}_2{\mathrm{O}}_2$ and STS, respectively. The FOV of each image is $180\times 180\mu \mathrm{m}$ . The scale bar in (a) is $30\mu \mathrm{m}$ .

3.3.

NADH Fluorescence Lifetime Increased Immediately after STS-Induced Apoptosis

Figure 3 shows the representative FLIM images (in color mapping) before, 0 to 15, 30 to 45, and $60\text{to}75\min$ after $1\mu \mathrm{M}$ STS treatment at the same FOV of HeLa cells. All the figures are displayed using the same scale of color bar as in Fig. 1b that the minimal and maximal lifetime is $200\mathrm{ps}$ (red) and $4000\mathrm{ps}$ (blue), respectively. Similar to Fig. 1b, control HeLa cells [Fig. 3a] had $a\sim 1300\mathrm{ps}$ lifetime (yellowish image in the selected color scale). The acquisition time of this control image was $900\mathrm{s}$ or $15\min$ . Immediately after STS treatment, we continuously acquired FLIM images with the acquisition time of $15\min$ per image for up to $90\min$ (the images acquired at 15 to 30, 45 to 50, and $75\text{to}90\min$ after treatment are not shown) at the same site to trace the response of the specimen. Apparently, the color-coded lifetime image exhibited blueshift (increased lifetime) at $0\text{to}15\min$ [Fig. 3b] after treatment, and then exhibited redshift at $60\text{to}75\min$ [Fig. 3d]. The nuclear areas tended to decrease in size with the increment of treatment time. A ringlike structure surrounding the nucleus [arrow head in Fig. 3c] was observed. In addition, the overall intensity of NADH fluorescence significantly increased immediately after STS treatment so that the laser power was decreased to escape the photomultiplier tube saturation. The recorded average photon count for the STS-treated cells was more than 1000, approximately twice higher than that in controls (600 counts), in the peak time channel of the recorded decays.

Fig. 3

Effect of staurosporine treatment on the FLIM images of HeLa cells. The FLIM images of HeLa cells from the same sites before (a), $0\text{to}15\min$ (b), $30\text{to}45\min$ (c), and $60\text{to}75\min$ (d) after treatment with 1 $\mu \mathrm{M}$ STS. The FOV of each image is $100\times 100\mu \mathrm{m}$ . The scale bar in (a) is $20\mu \mathrm{m}$ .

The distributions of the average lifetime $\left({\tau }_{\text{ave}}\right)$ [Figs. 3a, 3b, 3c, 3d] were plotted on the top panel of Fig. 4 . The curve labels (a) to (d) correspond to the results of Figs. 3a, 3b, 3c, 3d, respectively. An immediate lifetime increase after STS treatment was observed that the histogram shifted from the lower lifetime [curve (a)] to the higher lifetime [curve (b)] values. Then, the lifetime decreased and the histogram shifted toward the controls and became broader. The peak (i.e., maximal) value of ${\tau }_{\text{ave}}$ histograms, ${\tau }_{\text{peak}}$ , increased from $\sim 1300\mathrm{ps}$ [controls, curve (a)] to $\sim 3500\mathrm{ps}$ at $0\text{to}15\min$ after STS treatment [curve (b)], and then decreased to $\sim 2200\mathrm{ps}$ at $60\text{to}75\min$ .

Fig. 4

Effect of staurosporine treatment on the distributions of the average lifetime $\left({\tau }_{\mathrm{ave}}\right)$ of the FLIM images of HeLa and 143B cells. The top panel shows the plot of the distributions of the average lifetime $\left({\tau }_{\mathrm{ave}}\right)$ corresponding to the FLIM images in Figs. 3a, 3b, 3c, 3d. The bottom panel shows the plot of the mean and standard error (SE) of the ${\tau }_{\mathrm{peak}}$ as the function of time in all STS-treated cells. The number of sites acquired at each time point was labeled below each data point. The mean ${\tau }_{\mathrm{peak}}$ value $(\pm \mathrm{SE})$ of controls (i.e., $1336\pm 22\mathrm{ps}$ ) was plotted (star) in the same figure.

Similar results were repeatedly observed in a total of 6 sites (6 coverslips) of HeLa cells and 5 sites (5 coverslips) of 143B cells and are summarized in the bottom plot of Fig. 4. Here the number of sites is the same as the number of coverslips because time sequence measurements were performed at the same site of cells. The bottom panel of Fig. 4 plotted the mean and standard error (SE) of the ${\tau }_{\text{peak}}$ at all time points over all STS-treated HeLa and 143B cells. The number of sites acquired was decreased as a function of time and was labeled below each data point in the figure. The time point was assigned as the mean value of the acquisition period of time. For example, $7.5\min$ represents the data point acquired at $0\text{to}15\min$ after treatment, except that the latest point was labeled as $600\text{to}675\min$ . The mean ${\tau }_{\text{peak}}$ value $(\pm \mathrm{SE})$ of all controls (i.e., $1336\pm 22\mathrm{ps}$ ) was plotted (star) in the same figure for comparison. The mean ${\tau }_{\text{peak}}$ values $(\pm \mathrm{SE})$ for HeLa and 143B cells at $0\text{to}15\min$ after STS treatment are listed in Table 1. The mean ${\tau }_{\mathit{\text{peak}}}$ values at $0\text{to}15\min$ after STS treatment for HeLa or 143B cells are significantly different from those values of the controls ( $p\text{value}=1.4\times {10}^{-16}$ ). The two mean values of HeLa and 143B cells (i.e., $3570\pm 56$ versus $3448\pm 76$ ) are not significantly different $(p\text{value}=0.23)$ .

3.4.

No NADH Fluorescence Lifetime Change in High Dose $H_{\mathit{2}}{O}_{\mathit{2}}$ -Induced Necrosis

Figure 5 shows the representative FLIM images (in the same color mapping as previous FLIM images of controls and STS-treated cells) before, 0 to 15, 30 to 50, and $50\text{to}70\min$ after treatment with $1\mathrm{mM}$ ${\mathrm{H}}_2{\mathrm{O}}_2$ , taken from the same FOV of HeLa cells. In contrast to the FLIM images of STS-treated cells shown in Fig. 3, all images in Fig. 5 tend to be yellowish in the same selected color scale, which indicates that there was no significant lifetime change after ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment. The fluorescence intensity or photon counts tended to decrease after ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment [as seen in the changes of signal-to-background noise from Figs. 5a, 5b, 5c, 5d], which in part is due to the NADH oxidation¹⁶ by ${\mathrm{H}}_2{\mathrm{O}}_2$ . The recorded average photon count for the ${\mathrm{H}}_2{\mathrm{O}}_2$ -treated cells was $\sim 100$ to 200 (i.e., 3 to 6 times lower than that of controls) in the peak time channel.

Fig. 5

Effect of ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment on the FLIM images of HeLa cells. The FLIM images of HeLa cells were recorded from the same sites before (a), $0\text{to}15\min$ (b), $30\text{to}45\min$ (c), and $45\text{to}60\min$ (d) after treatment of the cells with $1\mathrm{mM}{\mathrm{H}}_2{\mathrm{O}}_2$ . The FOV of each image is $100\times 100\mu \mathrm{m}$ . The scale bar in (a) is $20\mu \mathrm{m}$ .

The top panel of Fig. 6 plotted the normalized histogram of ${\tau }_{\text{ave}}$ of NADH lifetime images shown in Figs. 5a, 5b, 5c, 5d. The curves (b) to (d) overlapped with one another. There is slight difference between curve (a) and curves (b) to (d). Similar to the bottom panel of Fig. 4, the bottom panel of Fig. 6 plotted the mean error and standard error of the ${\tau }_{\text{peak}}$ at all time points over all ${\mathrm{H}}_2{\mathrm{O}}_2$ -treated HeLa cells. The time point was assigned the same way as in Fig. 4 except the latest time point labeled as $135\text{to}185\min$ . The mean ${\tau }_{\text{peak}}$ value $(\pm \mathrm{SE})$ of controls (i.e., $1336\pm 22$ ) was plotted (star) in the same figure for comparison. The number of sites acquired at each time point was labeled below each data point in the figure. We observed no significant difference in ${\tau }_{\text{peak}}$ at all time points. To get more data points for averaging and comparing with the results of controls and apoptotic cells in Table 1, the mean ${\tau }_{\text{peak}}$ values $(\pm \mathrm{SE})$ for both time points at $0\text{to}15\min$ $(N=2)$ and $0\text{to}25\min$ $(N=2)$ after treatment were calculated and listed in Table 1. The mean ${\tau }_{\text{peak}}$ values at $0\text{to}25\min$ after ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment in HeLa cells are marginally different from the values of control HeLa cells $(p\text{value}=0.05)$ .

Fig. 6

Effect of ${\mathrm{H}}_2{\mathrm{O}}_2$ treatment on the distributions of the average lifetime $\left({\tau }_{\mathit{ave}}\right)$ of the FLIM images of HeLa cells. The top panel plotted the distributions of the average lifetime $\left({\tau }_{\mathit{ave}}\right)$ corresponding to the FLIM images in Figs. 5a, 5b, 5c, 5d. The bottom panel plotted the mean and standard error (SE) of the ${\tau }_{\mathit{\text{peak}}}$ as a function of time in all ${\mathrm{H}}_2{\mathrm{O}}_2$ -treated cells. The number of sites acquired at each time point was labeled below each data point. The mean ${\tau }_{\mathit{\text{peak}}}$ value $(\pm \mathrm{SE})$ of the controls (i.e., $1336\pm 22\mathrm{ps}$ ) was plotted (star) in the same figure.