2.1.

Improved Spectrometer-Microscope

The improved SM consists of a wide-field fluorescence microscope (IX73, Olympus, Japan) equipped with a metal halide lamp (HGLGPS, Olympus, Japan), a $40\times /1.3$ NA oil objective (UPLFLN40XO, Olympus, Japan), a CCD (ORCA-Flash 4.0, Hamamatsu, Japan), a fiber optic spectrometer (QE65 Pro, Ocean Optics, Florida), and a customized excitation filters wheel. As shown in Fig. 1, five different bandpass excitation filters are installed in the excitation filters wheel. In our study, four different excitations of $405/20\mathrm{nm}$ (Ex405), $436/20\mathrm{nm}$ (Ex436), $470/20\mathrm{nm}$ (Ex470), and $480/20\mathrm{nm}$ (Ex480) (Chroma, United States) are used for ExEm-spFRET and mExEm-spFRET measurement. Excitation of $510/20\mathrm{nm}$ (Ex510) (Chroma, United States) is used for emp-PbFRET measurement. Excitations of both Ex405 and Ex436 share the same filter cube that contains a DM455 (455-nm dichroic mirror, D455) (Olympus, Japan) and an ET460lp (long-pass emission filter of 460 nm, LP460) (Chroma, United States). Similarly, excitations of both Ex470 and Ex480 share another filter cube that contains a DM490 (490-nm dichroic mirror, D490) (Olympus, Japan) and an ET495lp (long-pass emission filter of 495 nm, LP495) (Chroma, United States). The illumination intensity can be attenuated in seven discrete steps (0%, 3%, 6%, 12%, 25%, 50%, and 100%), and another neutral density filters controller with three discrete steps (1%, 3%, and empty) can be used for the same purpose. Donor excitation and donor detection (DD) cube containing a DM460 (460-nm dichroic mirror, D460) and an ET480/30m (bandpass emission filter of $480/30\mathrm{nm}$, BP480/30) (Chroma, United States) and acceptor excitation and acceptor detection (AA) cube containing a DM515 (515-nm dichroic mirror, D515) and an ET550/40m (bandpass emission filter of $550/40\mathrm{nm}$, BP550/40) (Chroma, United States) are used to estimate the coexpression of both donor and acceptor in single living cells.

Fig. 1

Illustration of SM. Excitation filters wheel contains five bandpass excitation filters: Ex405, Ex436, Ex470, Ex480, and Ex510. Cube wheel contains four cubes: D455 cube containing a 455-nm dichroic mirror (D455) and a long-pass emission filter of 460 nm (LP460), D490 cube containing a 490-nm dichroic mirror (D490) and a long-pass emission filter of 495 nm (LP495), DD cube containing a 460-nm dichroic mirror (D460) and a bandpass emission filter of $480/30\mathrm{nm}$ (BP480/30), and AA cube containing a 515-nm dichroic mirror (D515) and a bandpass emission filter of $550/40\mathrm{nm}$ (BP550/40). Two detection channels: CCD imaging channel (a) and spectral detection channel (b).

SM has two independent detection modes: (a) microscopic imaging in CCD channel, offering a guidance for finding cells, and (b) spectral detection in spectrometer channel, recording emission spectra of the guided cells in the middle of CCD channel. Each $\text{count}(\lambda )$ or $E(\lambda )$ at the emission wavelength is related to the photons in about a 0.761-nm wavelength range. Spectral detection range is from 460 to 620 nm in this report.

2.2.

Modified ExEm-spFRET Method

ExEm-spFRET method we recently developed²⁵ is modified using a system correction factor ($f_{\mathrm{sc}}$) as follows (mExEm-spFRET):

A donor–acceptor tandem reference with known $E^{\mathrm{ref}}$ and $R_C^{\mathrm{ref}}$ can be used to predetermine $f_{\mathrm{sc}}$ and $r_k$ as follows:

In reality, $E^{\mathrm{ref}}$ can also be determined using some FRET methods, such as three-cube-based acceptor-sensitized emission method (E-FRET)¹⁰ or partial acceptor photobleaching method (emp-PbFRET).²⁶

2.3.

Partial Acceptor Photobleaching Method

Quantitative emp-PbFRET measurement was performed on SM for predetermining the $E^{\mathrm{ref}}$ of a 1D-nA tandem reference which contains one donor ($D$) and $n$ acceptor ($A$). Ex436 excitation was used to excite donor (Cerulean/CFP), and Ex510 excitation was used to selectively excite acceptor (Venus/YFP). Donor detection channel (${\mathrm{CH}}_D$) from 470 to 490 nm was used to selectively collect donor emission, and acceptor detection channel (${\mathrm{CH}}_A$) from 530 to 550 nm was used to mainly collect acceptor emission. The $E^{\mathrm{ref}}$ value of 1D-nA construct can be measured as follows:^11,26

2.4.

Calibration of SM

Careful calibration of SM was carried out with a halogen tungsten lamp (ISP-REF-CAL, Ocean Optics, Dunedin, Florida) just as described previously.²⁴ We first used a spectrometer (QE65 Pro, Ocean Optics, Florida) precalibrated by a standard light source (LS-1-CAL, Ocean Optics, Florida) to measure the spectrum [$E(\lambda )$] of the halogen tungsten lamp. We next used the spectrometer to measure the spectrum [${\text{count}}_{\text{lamp}}(\lambda )$] at the export of our microscope when the halogen tungsten lamp was placed on the objective of the microscope. The emission spectral response was calculated using $K(\lambda )=E(\lambda )/{\text{count}}_{\text{lamp}}(\lambda )$.

2.5.

Reagent and Plasmids

Plasmids DNA of Cerulean (C), Venus (V), CFP, and YFP were purchased from Addgene Company (Cambridge, Massachusetts). FRET tandem constructs, including C32V, CVC, and VCV, were kindly provided by the Vogel lab (National Institutes of Health, Bethesda, Maryland).^19,27 Plasmids DNA of CFP-Bax and YFP-Bax were kindly provided by Dr. Prehn.¹ Plasmid DNA of 18AA was kindly given by Professor Kaminski.¹¹ STS was purchased from Sigma-Aldrich Co. LLC (Santa Clara).

2.6.

Cell Culture and Transfection

HeLa cells obtained from the Department of Medicine, Jinan University (Guangzhou, China) were cultured just as described previously.²⁸ When the cells reached 70% to 90% confluence in a 35-mm glass dish, plasmids were transfected into cells by Turbofect™ (Fermentas Inc., Glen Burnie, Maryland) for 24 h.

3.1.

Calibration of SM

We first measured the emission spectral responses [$K(\lambda )$] of our SM system as shown in Fig. 7. We used the spectrometer to measure the spectrum [$E(\lambda )$, black solid line] of the halogen tungsten lamp and the spectrum [${\text{count}}_{\text{lamp}}(\lambda )$, black dot line] at the export of our microscope with D455 cube (a) and D490 cube (b), respectively, when the halogen tungsten lamp was placed on the objective of our microscope. $K_1(\lambda )$ for D455 cube and $K_2(\lambda )$ for D490 cube are also shown in Fig. 7 (gray solid line). Throughout the paper, emission count spectra with Ex405 or Ex436 excitation were calibrated with $K_1(\lambda )$, and emission count spectra with Ex470 or Ex480 excitation were calibrated with $K_2(\lambda )$.

We found that the emission spectral responses measured during six months were constant, demonstrating the excellent stability of our SM system. Although this calibration step is not mandatory for a precalibrated SM during at least six months, we actually performed this calibration step for every mExEm-spFRET measurement, which can be used as a criterion to determine whether SM is stable. In fact, this calibration step is very simple and can be performed within a few minutes.

3.2.

Excitation–Emission Spectral Fingerprints

Living HeLa cells exclusively expressed Cerulean (donor) or Venus (acceptor) were used to measure the excitation–emission spectral fingerprints of Cerulean and Venus as well as Cerulean–Venus sensitization on SM. Mean background spectrum with four excitations, respectively, collected from 20 nontransfected (no DNA plasmids and vectors) cells was subtracted from the raw count spectrum from cells expressing fluorescent proteins. Figure 2(a) (left) shows the raw emission count spectra of a representative cell expressing Cerulean-only indicated by red circle (inset) with Ex405, Ex436, Ex470, and Ex480 excitation, respectively. After calibration with $K(\lambda )$, emission spectrum of Cerulean with four excitations, respectively, was divided by the maximum value of emission spectrum with Ex436 excitation to obtain the normalized emission spectra [Fig. 2(a), middle]. The relative fluorescence intensities in emission wavelength range of 510 to 530 nm (CH1) with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, are shown in Fig. 2(a) (right). Similarly, Fig. 2(b) (left) shows the raw emission count spectra of a representative cell expressing Venus-only indicated by red circle (inset) with four excitations. Emission spectrum of Venus with four excitations, respectively, was divided by the maximum value of emission spectrum with Ex480 excitation to obtain the normalized emission spectra [Fig. 2(b), middle]. The relative fluorescence intensities in emission wavelength range of 520 to 540 nm (CH2) with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, are shown in Fig. 2(b) (right). Fluorescence intensities in Fig. 2(a) (middle) with Ex436 excitation and in Fig. 2(b) (middle) with Ex470 excitation are normalized to unit area, respectively, as the emission spectra of Cerulean and Venus.

Fig. 2

Excitation–emission spectral fingerprints of Cerulean and Venus as well as Cerulean–Venus sensitization. (a) Raw count spectra (left), normalized intensity spectra after calibration (middle), and the relative fluorescence intensity (right) in emission channel CH1 (from 510 to 530 nm) of a representative HeLa cell expressing Cerulean-only (inset image) with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation. Scale bar: $10\mu \mathrm{m}$; (b) raw count spectra (left), normalized intensity spectra after calibration (middle), and the relative fluorescence intensity (right) in emission channel CH2 (from 520 to 540 nm) of a representative HeLa cell expressing Venus-only (inset image) with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation. Scale bar: $10\mu \mathrm{m}$; (c) unit-area-normalized emission spectra of Cerulean (SEM C) and Venus (SEM V) measured from at least 20 living HeLa cells; (d) unit-area-normalized excitation spectra of Cerulean (SEX C) and Venus (SEX V) measured from at least 20 living HeLa cells; (e) three excitation–emission spectral fingerprints of Cerulean ($S_C$), Venus ($S_V$), and Cerulean–Venus sensitization ($S_S$) obtained by the outer product of the excitation and emission spectra in (c) and (d).

The emission spectra of Cerulean with Ex436 excitation and Venus with Ex470 excitation obtained from at least 20 living HeLa cells expressing Cerulean or Venus were normalized to unit area as the emission spectra of Cerulean (SEM C) and Venus (SEM V) [Fig. 2(c)]. Figure 2(d) shows the unit-area-normalized excitation spectra of Cerulean (SEX C) and Venus (SEX V). The unit-volume-normalized three-dimensional excitation–emission spectral fingerprints of Cerulean ($S_C$), Venus ($S_V$), and Cerulean–Venus sensitization ($S_S$) in Fig. 2(e) were calculated by the outer product of SEX C and SEM C, SEX V and SEM V, and SEX C and SEM V, respectively. In reality, $S_C$ and $S_V$ as well as $S_S$ in Fig. 2(e) were reconstructed by equally dividing the normalized intensity values into 25 grades (pseudocolor), and the equal grades were connected with contours.

The fact that the normalized emission spectra of fluorescent proteins (FPs) (Cerulean/CFP or Venus/YFP) measured from living HeLa or HepG2 cells expressing different levels of FPs are consistent further demonstrates the notion that the absorption and emission spectra of fluorescent proteins are generally very stable.^29,30 Although fluorescence intensity is proportional to the intensity of excitation light, mExEm-spFRET method is independent of the intensity of excitation light. Because three excitation–emission spectral fingerprints ($S_D$, $S_A$, and $S_S$) are normalized to unit volume, different excitation intensity only affects the weight factors ($W_D$, $W_A$, and $W_S$) rather than the ratios of weight factors. In reality, we found that the normalized excitation–emission spectral fingerprints of FPs obtained from living HeLa or HepG2 cells were constant during at least six months, indicating that our SM system is very stable. Therefore, the predetermined $S_C$, $S_V$, and $S_S$ can be directly used for subsequent quantitative mExEm-spFRET measurement without additional measurement. To offset the random fluctuation of count recorded at different emission wavelength, we summated the fluorescence intensity values in an emission wavelength range (CH1 for Cerulean and CH2 for Venus) rather than at single emission wavelength for Cerulean or Venus to obtain their excitation spectra. In reality, the emission wavelength range of 500 to 530 nm should be a better choice for CH1.

3.3.

Predetermination of the Correction Factors ($f_{\mathrm{sc}}$ and $r_k$)

To predetermine the correction factors ($f_{\mathrm{sc}}$ and $r_k$), living HeLa cells expressing C32V were excited with Ex405, Ex436, Ex470, and Ex480 excitation, respectively. Figure 3(a) shows the normalized emission spectra of a representative cell expressing C32V with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation. Figure 3(b) shows the corresponding excitation–emission spectrum ($S_{DA}$), which was linearly unmixed according to Eq. (3) to obtain $W_D=0.37$, $W_A=0.39$, and $W_S=0.27$. Next, we used emp-PbFRET method to measure the $E^{\mathrm{ref}}$ value of C32V for the same cell. Figure 3(c) shows the fluorescent images of the cell with DD cube (left) and AA cube (right), respectively, before (upper panel) and after (lower panel) partial Venus bleaching. Figure 3(d) shows the corresponding normalized count spectrum with Ex436 (solid line) and Ex510 (dot line) excitation, respectively, before (black) and after (gray) partial Venus bleaching with respect to the maximum value of the count spectrum with Ex436 excitation before partial Venus bleaching (black solid line). We summated the normalized count values in ${\mathrm{CH}}_D$ channel and ${\mathrm{CH}}_A$ channel, respectively, to obtain $I_{\mathrm{DD}}=11.94$, $I_{\mathrm{DD}}^{\text{post}}=14.41$, $I_{\mathrm{AA}}=13.45$, and $I_{\mathrm{AA}}^{\text{post}}=6.75$. According to Eq. (6), the corresponding $E^{\mathrm{ref}}$ was 28.8%, and the statistical $E^{\mathrm{ref}}$ value measured from 15 living HeLa cells was $30.1\%\pm 2.9\%$ [Fig. 3(e)]. Substituting $W_D=0.37$, $W_A=0.39$, and $W_S=0.27$ as well as $E^{\mathrm{ref}}=30.1\%$ into Eq. (4) to obtain $f_{\mathrm{sc}}=1.81$, where the quantum yield ratio ($r_Q$) of Venus (0.57) to Cerulean (0.62) is 0.919,^31,32 and into Eq. (5) to obtain $r_k$ was 0.43, where $R_C^{\mathrm{ref}}=1$. Statistical $f_{\mathrm{sc}}$ and $r_k$ values from 20 cells are $1.83\pm 0.12$ and $0.44\pm 0.01$, respectively [Fig. 3(f)].

Fig. 3

Predetermination of correction factors ($f_{\mathrm{sc}}$ and $r_k$) using a Cerulean–Venus tandem construct (C32V). (a) Normalized emission spectra of a representative cell expressing C32V with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation. Scale bar: $10\mu \mathrm{m}$; (b) excitation–emission spectrum corresponding to (a); (c) images of the cell with Ex436 (DD) or Ex510 (AA) excitation before (upper panels) and after (lower panels) partial Venus photobleaching. Scale bar: $10\mu \mathrm{m}$; (d) normalized count spectra of C32V inside the cell indicated by red circles in (c). ${\mathrm{CH}}_{\mathrm{D}}$: 470 to 490 nm and ${\mathrm{CH}}_{\mathrm{A}}$: 530 to 550 nm. (e) Statistical $E^{\mathrm{ref}}$ value from at least 15 living cells using emp-PbFRET method. (f) Statistical $f_{\mathrm{sc}}$ and $r_k$ values from at least 20 living cells.

In many reports, the quantum yield values of donor ($Q_D$) and acceptor ($Q_A$) from literature were directly quoted for quantitative FRET measurement.^{11,14,20,25,28} However, real $Q_D$ and $Q_A$ values are related to not only the optical properties of donor/acceptor but also the emission transmission characteristics of the instrument used. Moreover, it is also inappropriate to consider the $Q_D$ and $Q_A$ values from literature as the real $Q_D$ and $Q_A$ within a bandpass emission wavelength range. We here used the $f_{\mathrm{sc}}$ to correct the ratio of $Q_A$ to $Q_D$ ($Q_A/Q_D$) in our SM system. In fact, the product of the $Q_A/Q_D$ ratio quoted from literature and the $f_{\mathrm{sc}}$ is the real $Q_A/Q_D$ value in our SM system. Therefore, mExEm-spFRET method can measure the real $Q_A/Q_D$ value rather than the referenced $Q_A/Q_D$ value from literature for quantitative FRET measurement.

Generally, $r_k$ is only related to the excitation spectrum of our SM system and the absorption spectra of both donor and acceptor for a given cell line.²⁵ Just as discussed above about the spectral fingerprints, the spectral characteristics of our SM system and donor/acceptor are very stable. Therefore, for a given specific system, the predetermined $f_{\mathrm{sc}}$ and $r_k$ can be directly used for subsequent mExEm-spFRET measurement. In reality, we remeasured $f_{\mathrm{sc}}$ and $r_k$ values for Cerulean–Venus pair inside HeLa cells on our SM system during six months and obtained consistent $f_{\mathrm{sc}}$ and $r_k$ values, further demonstrating the stability of our instrument.

3.4.

Implementation of mExEm-spFRET in Single Living HeLa Cells Expressing C + V, CVC, and VCV

We next performed ExEm-spFRET and mExEm-spFRET method, respectively, on SM to measure the $E$ and $R_c$ values of single living cells expressing unlinked Cerulean plus Venus (C + V), CVC, and VCV, respectively. We measured four emission spectra of the cells excited with Ex405, Ex436, Ex470, and Ex480, respectively. Figures 4(a)–4(c) show the normalized emission spectra (left) of a representative cell expressing C + V (a, with respect to the value of the maximum peak at emission spectrum with Ex480 excitation), CVC (b, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation), and VCV (c, with respect to the value of the maximum peak at emission spectrum with Ex480 excitation), respectively, with different excitations and the corresponding excitation–emission spectra ($S_{\mathrm{DA}}$) (right). The $S_{\mathrm{DA}}$ were linearly unmixed according to Eq. (3) to obtain the weight values of donor, acceptor, and donor–acceptor sensitization (Table 1). Substituting these weight values and $r_k=0.44$ as well as $f_{\mathrm{sc}}=1.83$ into mExEm-spFRET method [Eqs. (1) and (2)] to obtain the corresponding $E$ and $R_c$: 3.1% and 1.01 for C + V, 40.4% and 0.54 for CVC, and 69.2% and 2.59 for VCV. In addition, for the same cells, implementation of ExEm-spFRET method exhibited that the $E$ and $R_C$ values were 5.5% and 1.21 for C + V, 55.3% and 0.56 for CVC, and 80.4% and 2.04 for VCV. Figure 4(d) shows the statistical $E$ and $R_c$ values of C + V, CVC, and VCV from 20 living cells. The $E$ values of CVC and VCV obtained by mExEm-spFRET are consistent with those measured by E-FRET method ($40.0\%\pm 0.7\%$ for CVC and $69.3\%\pm 1.0\%$ for VCV).³³

Fig. 4

Implementation of mExEm-spFRET method on SM system for quantitative FRET measurement in living cells separately expressing C + V, CVC, and VCV. (a–c) Normalized emission spectra (left) and corresponding excitation–emission spectrum (right) of representative cells separately expressing C + V (a, with respect to the value of the maximum peak at emission spectrum with Ex480 excitation), CVC (b, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation), and VCV (c, with respect to the value of the maximum peak at emission spectrum with Ex480 excitation) with Ex405, Ex436, Ex470, and Ex480 excitation, respectively. Scale bar: $10\mu \mathrm{m}$. (d) Statistical $E$ (left) and $R_c$ (right) values of C + V, CVC, and VCV constructs in 20 living HeLa cells obtained by ExEm-spFRET and mExEm-spFRET method, respectively.

Table 1

Weight values for different constructs.

We also used mExEm-spFRET method with two excitations to calculate the $E$ and $R_c$ values of C + V, C32V, CVC, and VCV constructs, respectively, for the same cells (Table 2). mExEm-spFRET method with Ex405 and Ex470, Ex405 and Ex480, Ex 430 and Ex470, or Ex436 and Ex480 excitations showed consistent results, while mExEm-spFRET method with Ex405 and Ex436 excitations obtained an obviously larger $R_c$ value for CVC construct, which may owe to the similarity of fluorescence intensity spectra with Ex405 and Ex436 excitation, respectively. Ex436 and Ex470 or Ex436 and Ex480 excitations should be good choices for quantitative mExEm-spFRET measurement of CFP/Cerulean and YFP/Venus pairs. In reality, we can perform a quantitative mExEm-spFRET measurement with two excitations within 1 s, which is applicable to the monitoring of dynamical events in single living cells.

Table 2

E and RC values of constructs measured by mExEm-spFRET with different excitation wavelengths.

3.5.

mExEm-spFRET Measurement of C32V in the Presence of Free Donor or Free Acceptor

We also used mExEm-spFRET method with four excitation wavelengths to measure the $E$ and $R_c$ values of C32V construct in the presence of free Cerulean or free Venus. Figure 5 shows the $E-R_c$ plot on a cell-by-cell basis for C + V, C32V, C32V + C, and C32V + V, respectively. Unlinked Cerulean plus Venus (C + V) exhibits very low $E$ values independent of $R_c$ (solid squares), whereas C32V exhibits a restricted distribution for $E$ (about 30.9%) and $R_c$ (about 1.02) values (solid triangles). The $E$ values of C32V + C are positively proportional to the corresponding $R_c$ (open triangles), whereas C32V + V has the same $E$ values as C32V (open circles), which is consistent with the previous reports.^15,20,34

Fig. 5

$E$ as a function of $R_c$. Implementation of mExEm-spFRET with four excitation wavelengths on SM for the HeLa cells expressing C32V (solid triangles), C + V (solid squares), C32V + C (open triangles), and C32V + V (open circles), respectively. C32V + C exhibits a positive linear correlation between $E$ and $R_c$, and the slope of dot line is 31.6. In the case of C32V + V, no correlation is observed.

In reality, high concentration of free Cerulean and free Venus may result in the possibility of spurious FRET efficiency by random collision.³⁵ For some bright cells coexpressing Cerulean and Venus (C + V, solid squares in Fig. 5), donor and acceptor may be within the Förster distance and form “spurious donor–acceptor complex,” which leads to a small systematic increase of $E$ as a function of $R_c$.^10,14 Therefore, we should not choose the cells with very high concentration of fluorescent proteins for quantitative measurements.

3.6.

mExEm-spFRET Measurement of STS-Induced Bax Homeoligomerization

Bax is a proapoptotic protein required for the process of mitochondrial outer membrane permeabilization.¹ Some publications, including our previous studies, have demonstrated that STS induces Bax translocation into mitochondria and subsequent homooligomerization.^20,28,36 We here performed mExEm-spFRET method on SM for single living HeLa cells coexpressing CFP-Bax and YFP-Bax. A CFP–YFP tandem reference (18AA) was used to predetermine the $f_{\mathrm{sc}}$ (1.80) and $r_k$ (0.42) values for CFP–YFP pair on our SM system. As shown in Figs. 6(a) and 6(b), emission spectrum with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, was divided by the maximum value of emission spectrum with Ex436 excitation to obtain the normalized emission spectra. Bax distributed evenly in cytosol in the control cell exhibiting 2.85% of $E$ and 0.26 of $R_c$ [Fig. 6(a)], and Bax showed significant clusters in the cell exhibiting 10.23% of $E$ and 0.30 of $R_c$ after $2\text{-}\mu \mathrm{M}$ STS treatment for 6 h [Fig. 6(b)]. Statistical $E$ values are $2.2\%\pm 0.4\%$ for control cells (26 cells) and $14.4\%\pm 5.7\%$ for STS-treated cells (105 cells), indicating that STS induced the formation of mitochondria-associated Bax clusters. Figure 6(c) shows the $E-R_c$ plot on a cell-by-cell basis for control (solid squares) and STS-treated (open squares) cells, respectively. As shown in Fig. 6(c), the fact that the FRET efficiency {apparent FRET efficiency [$E_{\mathrm{app}}=E_{\max }·(DA)/(D_{\text{total}})$]} is very low and independent on the $R_c$ for control cells indicates that Bax does not exist as homooligomer in healthy cells. However, for the STS-treated cells, apparent FRET efficiency ($E_{\mathrm{app}}$) obviously increased in the case of $R_c<1$ [(DA) increases with $R_c$ or ($A_{\text{total}}$)] but kept constant in the case of $R_c>1$ $[(DA)=(D_{\text{total}})]$ with $R_c$ increasing, further demonstrating that all Bax formed homooligomer after STS treatment for 6 h.

Fig. 6

mExEm-spFRET analysis on Bax homooligomerization in single living HeLa cells coexpressing CFP-Bax and YFP-Bax. (a and b) Normalized emission spectra (left) and the corresponding excitation–emission spectrum (right) of a representative cell coexpressing CFP-Bax and YFP-Bax in the absence (a) and presence (b) of STS with Ex405, Ex436, Ex470, and Ex480 excitation, respectively, with respect to the value of the maximum peak at emission spectrum with Ex436 excitation. Scale bar: $10\mu \mathrm{m}$. (c) $E-R_c$ plot from 26 control cells (solid squares) and 105 STS-treated cells (open squares), respectively.