ExEm-FRET two-hybrid assay: FRET two-hybrid assay based on linear unmixing of excitation-emission spectra

Chenshuang Zhang; Yangpei Liu; Wenfeng Qu; Wenhua Su; Mengyan Du; Fangfang Yang; Tongsheng Chen

doi:10.1364/OE.27.018282

1. Introduction

Fluorescence resonance energy transfer (FRET) two-hybrid assay is an invaluable method for resolving the stoichiometric information of macromolecular complexes within living cells [1–3]. In FRET two-hybrid assay, a single cell is considered to be a “biological cuvette” for titration experiments between molecules. Both E_D and E_A (donor- and acceptor-centric FRET efficiency) of the pairs of donor fluorophore-tagged and acceptor fluorophore-tagged molecules are systematically measured from a lot of single cells with varying concentrations of free donors (D_free) and acceptors (A_free) to obtain the FRET binding curves of both E_D-A_free and E_A-D_free, yielding E_A,max and E_D,max (the maximal E_A and E_D, assuming each acceptor/donor molecule binds to donor/acceptor) as well as the stoichiometric ratio (E_A,max/E_D,max) of the fluorophore-tagged molecules [1].

FRET two-hybrid assay is currently carried out by implementing two existing fluorescence intensity-based FRET approaches: 3³-FRET and E-FRET [1,2]. Based on the measured E_D and E_A images by using E-FRET and 3³-FRET methods, we recently obtained E_A,max and E_D,max by using E_D, E_A and the ratio of donor to acceptor concentration across pixel-to-pixel to determine the stoichiometric ratio of intracellular complexes [2]. Although E-FRET method is a popular approach for quantitative biochemistry of dynamic signal transduction in living cells [1–4], the requirement that all measurements must be implemented under the same conditions partly impedes the application of live-cell quantitative FRET measurements.

Simultaneous unmixing of excitation-emission spectra (ExEm-unmixing) has become a potential technology to resolve FRET signal (ExEm-FRET) due to the inherent ability of overcoming spectral crosstalk (a key issue of quantitative FRET measurement) [5–11]. Recently, a spectral wide-field microscope was developed for microscopic live-cell spectral imaging, and we implemented ExEm-FRET measurement on the microscope in single living cells [6,7,10,11]. The instrument settings for ExEm-FRET measurement can be adjusted for measuring cells under optimal imaging conditions, thus ExEm-FRET measurement greatly reduces the failure rate of quantitative FRET measurements in living cells [10,11]. Moreover, ExEm-FRET measurement has superior stability and robustness and can obtain relatively accurate results even for cells with low signal-to-noise ratio [6,11].

This report aims to develop an ExEm-unmixing-based FRET two-hybrid assay method, named as ExEm-FRET two-hybrid assay. Linear unmixing of the excitation-emission spectra (S_DA) of cells obtains the weight factors of donor (W_D), acceptor (W_A) and donor-acceptor sensitization (W_S), yielding E_D and E_A images. ExEm-FRET two-hybrid assay employs pixel-to-pixel titration curves (the titration curves are obtained by analyzing each pixel of images) of E_D/E_A versus the free acceptor (C_a)/donor(C_d) concentration deduced from the three weight factors for obtaining E_A,max and E_D,max, thus yielding the stoichiometric ratio (E_A,max/E_D,max) of donor-tagged protein binding to acceptor-tagged protein. ExEm-FRET two-hybrid assay requires only a few cells, which greatly reduces the workload of quantitative FRET measurements. Furthermore, due to the superior robustness and sensitivity of ExEm-unmixing, ExEm-FRET two-hybrid assay is especially applicable for cells with low fluorescence signal, thus provides a powerful tool for exploring the molecular mechanism of intracellular signal transduction in living cells. We performed the ExEm-FRET two-hybrid assay on a self-developed automated FRET microscope for evaluating the binding of Bcl-XL to Bad. ExEm-FRET two-hybrid assay for living Hela cells co-expressing Bad-YFP and CFP-Bcl-XL showed 1.345 (in healthy cells) and 1.716 (in apoptotic cells) of the stoichiometric ratio of Bcl-XL binding to Bad, indicating coexistence of both Bad-Bcl-XL dimer and Bad-Bcl-XL₂ trimer in living cells.

2. Theory

2.1 ExEm-FRET two-hybrid assay

The net excitation-emission spectrum (S_DA) of a FRET sample can be linearly resolved into three excitation-emission components as follow [7,10],

\begin{matrix} S_{D A} = (C_{d} K_{D} Q_{D} + C_{d a} K_{D} (1 - E) Q_{D}) S_{D} + C_{d a} K_{D} E Q_{A} S_{S} + C_{A}^{t} K_{A} Q_{A} S_{A} \\ = W_{D} S_{D} + W_{S} S_{S} + W_{A} S_{A}, \end{matrix}

where, E is FRET efficiency of the paired donor-acceptor; C_d, C_da and C^t_A are the concentration of free donor, the paired donor and total acceptor, respectively; K_X is total absorption of X (D for donor, A for acceptor) in all excitation wavelengths used [7,10], and Q_X is quantum yield of X; S_D, S_A and S_S are the unit-area-normalized excitation-emission spectral signatures of donor, acceptor and donor-acceptor sensitization, respectively; W_D, W_A and W_S are the weights of S_D, S_A and S_S, respectively [7]. According to Eq. (1) and the definitions of W_D, W_A and W_S, we also obtain

C_{A}^{t} = \frac{1}{K_{A} Q_{A}} W_{A},

\begin{matrix} C_{D}^{t} = C_{d} + C_{d a} \\ = \frac{1}{K_{D} Q_{D}} (W_{D} + \frac{Q_{D}}{Q_{A}} W_{S}), \end{matrix}

where C^t_D is the concentration of total donor. E_D and E_A are obtained by following equations [10],

E_{A} = E \frac{C_{d a}}{C_{A}^{t}} = \frac{K_{A}}{K_{D}} \frac{W_{S}}{W_{A}},

E_{D} = E \frac{C_{d a}}{C_{D}^{t}} = \frac{W_{S}}{W_{S} + \frac{Q_{A}}{Q_{D}} W_{D}} .

For a D_nDA_nA complex with n_D:n_A of stoichiometric ratio [1,2]

n_{D} D + n_{A} A \underset{K_{d}}{\Leftrightarrow} D_{n_{D}} A_{n_{A}},

K_d is the dissociation constant, and [1,2,12]

K_{d} = \frac{C_{d} C_{a}}{C_{D A}},

where C_a and C_DA are the concentration of free acceptor and donor-acceptor complex, respectively. For a binding interaction with hypothetical n_A:n_D = 1:n_D,

C_{A}^{t} = C_{a} + C_{D A},

C_{D}^{t} = C_{d} + n_{D} C_{D A} .

Combing Eqs. (5) and (6), we obtain following equation,

C_{d}^{2} + (K_{d} + n_{D} C_{A}^{t} - C_{D}^{t}) C_{d} - K_{d} C_{D}^{t} = 0.

We thus obtain C_d as follow,

C_{d} = \frac{b + \sqrt{b^{2} + 4 c}}{2},

where,

b = \frac{1}{K_{D} Q_{D}} (W_{D} + \frac{Q_{D}}{Q_{A}} W_{S}) - n_{D} \frac{1}{K_{A} Q_{A}} W_{A} - K_{d},

c = K_{d} \frac{1}{K_{D} Q_{D}} (W_{D} + \frac{Q_{D}}{Q_{A}} W_{S}) .

According to Eq. (6), C_a is determined as follow,

\begin{matrix} C_{a} = C_{A}^{t} - \frac{C_{D}^{t} - C_{d}}{n_{D}} \\ = \frac{1}{K_{A} Q_{A}} W_{A} - \frac{1}{n_{D}} [\frac{1}{K_{D} Q_{D}} (W_{D} + \frac{Q_{D}}{Q_{A}} W_{S}) - C_{d}] . \end{matrix}

Using dynamic titration curves of E_D versus C_a and E_A versus C_d, E_D,max and E_A,max are obtained. The stoichiometric ratio (n_D/n_A) of the FRET sample is [1,2]

\frac{n_{D}}{n_{A}} = \frac{E_{A, m a x}}{E_{D, m a x}} .

2.2 Determination of collisional FRET

As the spurious E_A and E_D due to the collision between donor and acceptor molecules are linearly proportional to the concentration of donor and acceptor, respectively [1], the magnitude of spurious FRET efficiency is related to C^t_D and C^t_A by the following,

E_{A, s p u r i o u s} = S l o p e_{s p u r i o u s F R E T} \cdot C_{D}^{t} = S l o p e_{s p u r i o u s F R E T} \cdot \frac{1}{K_{D} Q_{D}} (W_{D} + \frac{Q_{D}}{Q_{A}} W_{S}),

E_{D, s p u r i o u s} = S l o p e_{s p u r i o u s F R E T} \cdot C_{A}^{t} = S l o p e_{s p u r i o u s F R E T} \cdot \frac{1}{K_{A} Q_{A}} W_{A} .

and the measured E_A and E_D are corrected by using the flowing equations:

E_{A, c o r r e c t e d} = E_{A} - S l o p e_{s p u r i o u s F R E T} \cdot C_{D}^{t},

E_{D, c o r r e c t e d} = E_{D} - S l o p e_{s p u r i o u s F R E T} \cdot C_{A}^{t} .

2.3 Determination of FRET binding curves

The experimentally measured E_A and E_D must relate to the C_d and C_a, respectively:

E_{A, p r e d} = E_{A, m a x} \cdot \frac{C_{D A}}{C_{A}^{t}} = E_{A, m a x} \cdot \frac{C_{d}}{C_{d} + K_{d}},

E_{D, p r e d} = E_{D, m a x} \cdot \frac{n_{D} \cdot C_{D A}}{C_{D}^{t}} = E_{D, m a x} \cdot \frac{C_{a}}{C_{a} + K_{d} \cdot \frac{1}{n_{D}}} .

Thus, the values of K_d, E_A,max and E_D,max can be estimated with fitting iteratively through least-squares minimization [1], where the condition for finding the optimal solution is that the difference between E_A/E_D and E_A,pred/E_D/pred (Eqs. (3) and 12) is the smallest, and the FRET binding curves of both E_D-C_a and E_A-C_d are obtained as shown in Fig. 1 [1].

Fig. 1 Illustration of live-cell FRET binding curves. (a) Acceptor-centric FRET efficiency (E_A) values measured from single pixels were plotted versus the concentration of free donor (C_d) to form E_A-C_d binding curve. The E_A values of the pixels with excess donor molecules are extremely close to the value of E_A,max (the maximal E_A, assuming each acceptor molecule binds to donor). (b) Donor-centric FRET efficiency (E_D) values measured from single pixels were plotted versus the concentration of free acceptor (C_a) to form E_D-C_a binding curve. The E_D values of the pixels with excess acceptor molecules are extremely close to the value of E_D,max (the maximal E_D, assuming each donor molecule binds to acceptor).

Download Full Size | PDF

2.4 Five steps of implementing ExEm-FRET two-hybrid assay

Step 1: Predetermination of excitation–emission spectral signatures.

Excitation–emission spectral signatures (S_D, S_A and S_S) are determined by using living cells separately expressing donor and acceptor fluorophores, respectively [7,10]. The excitation spectrum of donor/acceptor is obtained by fixing one emission channel and switching different excitation wavelengths. Fixing one excitation wavelength and collecting the fluorescence of different emission channels, the emission spectrum of donor/acceptor is obtained. S_D/S_A is the outer product of the excitation spectrum and the emission spectrum measured from donor-only/acceptor-only sample. S_S is obtained by the outer product of donor excitation spectrum and acceptor emission spectrum.

Step 2: Determination of the system calibration factors (Q_A/Q_D and K_A/K_D).

Both Q_A/Q_D and K_A/K_D values are simultaneously measured by linearly unmixing the excitation–emission spectra of more than one FRET pairs (reference samples) just as described previously [10]. S_DA of the reference samples are linearly unmixed into W_D, W_A and W_S according to Eq. (1). Plot of W_S/W_D versus W_A/W_D for the reference samples indicates a linear relationship with the absolute intercept (Q_A/Q_D) and the reciprocal of slope (K_A/K_D).

Step 3: FRET efficiency calculation of FRET sample.

After obtaining the three contributions (weights: W_D, W_S and W_A) from the S_DA of FRET sample, both E_A and E_D values are calculated according to Eq. (3). The E_A and E_D values still need to be further corrected (as shown in Step 4 below).

Step 4: Correction of spurious FRET.

When fluorophores exist at high concentration, the distance between a donor and an acceptor may be less than the Förster distance by random chance, and the donor and acceptor molecules would undergo FRET even there is no real potential binding interaction [1]. Therefore, to obtain reliable live-cell titration curves of E_D/E_A versus C_a/C_d, the contribution of spurious FRET due to the collision between donor and acceptor molecules must be subtracted (Eq. (11). As shown in Eq. (2), the measured C^t_A and C^t_D are related to 1/K_AQ_A and 1/K_DQ_D, respectively. K_AQ_A value can be assumed to be 1, and the 1/(K_DQ_D) value can be determined according to the values of measured K_A/K_D and Q_A/Q_D.

Step 5: Determination of FRET binding curves.

Pixel-to-pixel titration curves of E_D/E_A versus C_a/C_d deduced from the three weight factors (W_D, W_S, and W_A) are employed for obtaining E_A,max and E_D,max (Fig. 1), thus yielding the stoichiometric ratio (E_A,max/E_D,max) of donor-tagged protein binding to acceptor-tagged protein.

3. Materials and methods

3.1 ExEm-FRET experiments

ExEm-FRET measurements were performed on a self-developed automated FRET microscope. Figure 2(a) depicts the diagram of the automated FRET microscope. A wide-field microscope (IX73, Olympus, Hamburg, Japan) with an objective (60 × /1.42 NA oil, Olympus), a mercury lamp (HGLGPS, Olympus) and a CMOS camera (Flash 4.0, Hamamatsu, Japan) was refitted by utilizing a motorized cube wheel (IX3-RFACA, Olympus) with transition time as few as 500 ms and a motorized emission filter wheel (FW103, Throlabs, Newton, NJ, USA) with transition time as few as 55 ms. Excitation of two wavelengths was changed by switching two band-pass excitation filters (Chroma, USA): 435/20 nm (Ex 1) and 470/20 nm (Ex 2). Five band-pass emission filters (Chroma, USA) were placed in a motorized emission filter wheel for emission spectra measurement: 470/20 nm (Em 1), 490/20 nm (Em 2), 510/20 nm (Em 3), 530/20 nm (Em 4) and 550/20 nm (Em 5). In this report, the exposure time of each image was 500 ms, and a complete ExEm-FRET measurement took about 5.5 s.

Fig. 2 Illustration of ExEm-FRET experiments. (a) Diagram of the wide-field automated FRET microscope. A 435/20 nm band-pass excitation filter and a 455 nm dichroic mirror are placed in Cube 435; A 470/20 nm band-pass excitation filter and a 490 nm dichroic mirror are placed in Cube 470. The emission filter wheel contains 470/20 nm, 490/20 nm, 510/20 nm, 530/20 nm, and 550/20 nm band-pass emission filters. (b) An S_DA contains five images with 435 nm excitation and three images with 470 nm excitation).

Download Full Size | PDF

Image-stack (S_DA) was used to denote the excitation-emission spectral images of FRET sample [7,10]. An S_DA contains eight images, including five emission spectral images with 435 nm excitation and three emission spectral images with 470 nm excitation, respectively (Fig. 2b).

3.2 Cell culture, transfection and plasmids

Hela cells, cell culture, transfection, plasmids (CFP (C), YFP (Y), C80Y, C40Y, C10Y and C4Y) and Staurosporine (STS) have been described previously in our publications [13,14]. CFP-Bcl-XL was kindly supplied by Dr. Gilmore [15]. To generate a plasmid encoding YFP fused to Bad, the coding region of Bad was prepared by PCR (the Bad cDNA from mCherry-Bad), and was ligated into pcDNA3-YFP. Plasmid of mCherry-Bad was kindly provided by Dr. Andrews [16].

Hela cells were washed three times in phosphate buffer solution (PBS) after 24 h transfection and Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum was added before ExEm-FRET imaging in healthy cells. For the samples needed for ExEm-FRET imaging in apoptotic cells, 1 μM STS was added for 7 h after 24 h transfection. Finally, the cells were washed three times in PBS and DMEM containing 10% fetal calf serum was added before ExEm-FRET imaging in apoptotic cells.

4. Result and discussion

4.1 Predetermination of excitation–emission spectral signatures and system calibration factors

Firstly, living Hela cells separately expressing untagged CFP (C) and untagged YFP (Y) were used to predetermine excitation–emission spectral signatures (S_D and S_A as well as S_S) [7,10]. Figure 3(a) shows the emission-spectral images of CFP and YFP with 435 nm and 470 nm excitation, respectively (Left). BG (the intensity of non-fluorescent area) was subtracted from the intensities of the pixels of cells. The measured fluorescence intensities were calibrated by using the emission-spectral responses of the imaging system, and then were normalized to unit area as the emission spectrum of CFP and YFP, respectively [10] (Fig. 3a, Right). Figure 3(b) shows the fluorescence images at 510 nm of cells expressing CFP and fluorescence images at 530 nm of cells expressing YFP with 435 nm excitation and 470 nm excitation, respectively (Top). After BG subtraction and normalization of measured fluorescence intensities, the average excitation spectra of the cells separately expressing CFP and YFP were obtained (Fig. 3b, Bottom). S_D/S_A was the outer product of the excitation spectrum and the emission spectrum measured from donor-only/acceptor-only sample, and S_S was obtained by the outer product of donor excitation spectrum and acceptor emission spectrum (Fig. 3c) [7].

Fig. 3 Excitation-emission spectral signatures of CFP (S_D), YFP (S_A) and CFP-YFP sensitisation (S_S). (a) Left: emission-spectral images of CFP and YFP with 435 nm excitation and 470 nm excitation, respectively; right: average emission spectra of the pixels of the cells separately expressing CFP and YFP, respectively. (b) Top: excitation spectral images of cells expressing CFP and YFP with 435 nm excitation and 470 nm excitation, respectively; bottom: average excitation spectra of the pixels of the cells separately expressing CFP and YFP, respectively. (c) Pseudo-colour images of S_D, S_A and S_S.

Download Full Size | PDF

Four C-Y FRET pairs (C4Y, C10Y, C40Y and C80Y) were used to simultaneously measure K_A/K_D and Q_A/Q_D values as described previously [1,10]. In this report, K_AQ_A value was assumed to be 1, and the measured K_A/K_D and Q_A/Q_D values were 0.237 and 1.667, respectively. We obtained 0.394 of 1/(K_DQ_D) value. The excitation–emission spectral signatures (S_D, S_A and S_S) and K_A/K_D and Q_A/Q_D values for ExEm-FRET imaging just need to be measured periodically due to the stability of our system [10,13].

In order to reduce the effect of pixel mismatch on measurement results, two ways were adopted for reducing the effect of pixel mismatch on our results. Firstly, the pixel mismatches among five detection channels were calibrated by imaging standard ruler sample and corrected in post-processing. Secondly, the cell edges with mismatch were filtered by using a binary mask just as described in our previous report [13]. According to the emission spectra of donor (CFP) and acceptor (YFP), we chose 470/20 and 530/20 nm channels to fabricate a binary mask for filtering the mismatched pixels from FRET images as following steps: (1) Set 1.5*I_BG (I_BG, the intensity of background) as the thresholds of 470/20 and 530/20 nm images, respectively; (2) Generate binary masks of 470/20 and 530/20 nm images by assigning pixels whose gray intensity are less than 1.5*I_BG with 0 and pixels whose gray intensity are not less than 1.5*I_BG with 1; (3) Get the final binary mask through a logical AND of 470/20 and 530/20 nm binary masks.

4.2 Correction of spurious FRET

To obtain reliable titration curves of E_A-C_d and E_D-C_a, the contribution of spurious FRET due to collisions must be subtracted [1,17]. After BG subtraction and calibration by using the system responses, the S_DA of cells co-expressing CFP and YFP (Fig. 4a) was linearly unmixed for obtaining W_A, W_S, and W_D factors according to Eq. (1) (Fig. 4b, Top). According to Eqs. (2) and (3), we obtained the images of C^t_D, E_A, C^t_A, and E_D (Fig. 4b, Bottom) and the corresponding pixel-to-pixel E_A- C^t_D plot (Fig. 4c) and E_D- C^t_A (Fig. 4d). The spurious FRET efficiency due to collisions is linearly proportional to the concentration of fluorescent proteins with 3.4 × 10⁻⁶ of slope in the E_A- C^t_D plot, which can also be obtained from the E_D- C^t_A plot. In our experimental system, we found that the spurious FRET efficiency caused by the collision between untagged CFP and untagged YFP was very small, which can be ignored in our system. However, the spurious FRET efficiency may be prohibitively high when molecules are placed restrictions on membrane, which can be estimated and corrected by localizing donor and acceptor fluorescent proteins on the membrane.

Fig. 4 Measuring collisional FRET in living Hela cells. (a) A representative excitation-emission spectral image (S_DA) of cells co-expressing untagged CFP and untagged YFP. Scale bar: 40 μm. (b) Top: pixel-to-pixel images of W_A, W_S and W_D corresponding to (a); Bottom: the pixel-to-pixel images of C^t_D, E_A, C^t_A and E_D. (c) Pixel-to-pixel E_A- C^t_D plot with a slope of 3.4 × 10⁻⁶. (d) Pixel-to-pixel E_D- C^t_A plot with a slope of 3.4 × 10⁻⁶.

Download Full Size | PDF

4.3 Evaluation of the binding of Bad to Bcl-XL

We next used ExEm-FRET two-hybrid assay to evaluate the interaction of Bad, an apoptotic BH-3 only protein, with Bcl-XL, an anti-apoptotic protein, in healthy cells [16,18]. Figure 5(a) shows an S_DA of living cells co-expressing Bad-YFP and CFP-Bcl-XL (Top). Linear unmixing of the S_DA according to Eq. (1) obtained the corresponding W_A, W_S and W_D images (Fig. 5a, Bottom). The C^t_D and C^t_A mages were obtained by using Eq. (2). The E_A and E_D images were obtained by using Eq. (3) and corrected by Eq. (11) (Fig. 5b). According to Eq. (8a), E_A was plotted against estimated C_d (Fig. 5c, Left), where each black symbol corresponds to the data from a single pixel, and the E_A,max value was about 0.364. Similarly, according to Eq. (8d), E_D was plotted as a function of estimated C_a, and the E_D,max value was about 0.274 (Fig. 5c, Right). 1.282 (0.364/0.274) of the stoichiometric ratio of Bcl-XL binding to Bad was obtained. We analyzed about 50 cells from 10 frames and obtained 1.345 (0.394/0.293) of the stoichiometric ratio of Bcl-XL binding to Bad, indicating that the complexes exist mainly in Bad-Bcl-XL dimer and there may be a small amount of Bad-Bcl-XL₂ trimer in healthy cells.

Fig. 5 Stoichiometric ratio of Bad binding to Bcl-XL in healthy cells. (a) An S_DA of living cells co-expressing Bad-YFP and CFP-Bcl-XL and the corresponding W_A, W_D and W_S images. Scale bar: 40 μm. (b) The corresponding pixel-to-pixel images of E_A, C^t_D ,C^t_A and E_D. (c) Left: pixel-to-pixel E_A-C_d plot with an E_A,max value of 0.364; Right: pixel-to-pixel E_D-C_a plot with a E_D,max value of 0.274.

Download Full Size | PDF

ExEm-FRET two-hybrid assay was also used to evaluate the binding of Bad with Bcl-XL in staurosporine(STS)-induced apoptotic cells [19]. Treatment with 1 μM STS for 7 h induced cell shrinkage (Fig. 6) and apoptosis (data not shown). Figure 6(a) shows an S_DA of apoptotic cells co-expressing Bad-YFP and CFP-Bcl-XL (Top). The corresponding W_A, W_S and W_D images were obtained from the S_DA (Fig. 6a, Bottom), and the corresponding E_A, C^t_D, C^t_A and E_D images were shown in Fig. 6(b). The E_A,max value was about 0.377 from the corresponding E_A-C_d plot, and the E_D,max value was about 0.202 from the corresponding E_D-C_a plot. Statistical result from 50 cells in 10 frames was 1.716 (0.398/0.232) stoichiometric ratio of Bcl-XL binding to Bad, indicating that the complexes exist mainly in Bad-Bcl-XL₂ trimers in apoptotic cells.

Fig. 6 Stoichiometric ratio of Bad binding to Bcl-XL in apoptotic cells treated with 1 μM STS for 7 h. (a) A S_DA of apoptotic cells co-expressing Bad-YFP and CFP-Bcl-XL and the corresponding W_A, W_D and W_S images. Scale bar: 40 μm. (b) The corresponding pixel-to-pixel images of E_A, C^t_D, C^t_A and E_D. (c) Left: pixel-to-pixel E_A-C_d plot with an E_A,max value of 0.377; Right: pixel-to-pixel E_D-C_a plot with a E_D,max value of 0.202.

Download Full Size | PDF

Interactions between pro-apoptotic and anti-apoptotic proteins play an important role in the regulation of apoptosis [20–23]. Bad has been demonstrated to not only abrogate the protective capacity of Bcl-XL but also accelerate the apoptotic response of cells to a death signal [20,23]. It was reported that Bad inhibits the survival functions of Bcl-XL by direct interaction with Bcl-XL, forming hetero-dimer [22,23].

However, our results that the stoichiometric ratio of Bcl-XL binding to Bad is always between 1 and 2 in both healthy and STS-induced apoptotic cells (Figs. 5 and 6) further suggest the existence of Bad-Bcl-XL₂ trimer [2]. In vitro studies have confirmed the existence of Bad-Bcl-XL dimer, but have not observed Bad-Bcl-XL₂ trimer [21–23]. We speculate that the affinity of Bad-Bcl-XL₂ trimer may be not as strong as that of Bad-Bcl-XL dimer, thus Bad-Bcl-XL₂ trimer has not been detected by in vitro studies.

In order to exclude the possibility that dimer and trimer exist separately in different cells, we also analyzed one of the cells in Fig. 5 and obtained 1.240 (0.381/0.307) of the stoichiometric ratio of Bcl-XL binding to Bad (Fig. 7). The results shown in Fig. 7 further indicate the coexistence of both Bad-Bcl-XL dimer and Bad-Bcl-XL₂ trimer in living cells. However, in practical experiments, it is difficult for one cell to satisfy the conditions that both C_d and C_a have a wide concentration range. Therefore, in reality, simultaneous analysis of a few cells was adopted for obtaining more desirable titration curves of E_D-C_a and E_A-C_d.

Fig. 7 Stoichiometric ratio of Bad binding to Bcl-XL in the cell indicated by a red box. (a) An S_DA of the cell indicated by the red box. (b) Left: pixel-to-pixel E_A-C_d plot of the cell indicated by red box in (a) with an E_A,max value of 0.381; Right: pixel-to-pixel E_D-C_a plot with a E_D,max value of 0.307.

Download Full Size | PDF

The 1.345 and 1.716 stoichiometric ratios of Bcl-XL-Bad complexes also indicate that Bad has two binding sites with Bcl-XL. Bcl-XL may bind preferentially to one of the sites and then bind to the other one when Bcl-XL concentration is relative higher than Bad concentration. Our studies and other reports have demonstrated that Bcl-XL has strong anti-apoptotic ability and can prevent STS-induced apoptosis [2,23]. After STS treatment for 7 h, only the cells with relative high expression of Bcl-XL may be used for analysis, which increased the proportion of Bad-Bcl-XL₂ trimer. Therefore, the stoichiometric ratio of Bcl-XL to Bad from STS-induced cells was higher than that from healthy cells.

We for the first time employed ExEm-FRET method to measure the stoichiometric information of macromolecular complexes within living cells. Live-cell ExEm-FRET two-hybrid assay for evaluating the binding of Bad to Bcl-XL in Hela cells yielded similar results to the previous report that revealed the coexistence of both Bad-Bcl-XL dimer and Bad-Bcl-XL₂ trimer in living cells [2]. Recently, our studies have confirmed that ExEm-FRET measurement has superior stability and robustness and can obtain relatively accurate results even for cells with low signal-to-noise ratio [6,10,11]. Due to the superior robustness and sensitivity of ExEm-FRET [6,11], ExEm-FRET two-hybrid assay is very applicable to evaluating stoichiometry of macromolecular complexes in living cells. On our self-developed automated FRET microscope, one switch of the motorized cube wheel and six-time switch of the motorized emission filter wheel are required for a complete ExEm-FRET measurement. Although ExEm-FRET measurement requires about eight spectral images, a complete measurement takes about only 3 s on our self-developed automated FRET microscope when the expression level of fluorescent protein is relatively high.

5. Conclusion

We here develop an ExEm-FRET two-hybrid assay to determine the stoichiometric information of intracellular complexes by linearly unmixing the excitation-emission spectra of cells. Because of using pixel-to-pixel titration curves of donor-centric FRET efficiency (E_D) versus the free acceptor (C_a) and acceptor-centric FRET efficiency (E_A) versus the free donor (C_d) to obtain the E_D,max and E_A,max (the maximal E_A and E_D) values, respectively, ExEm-FRET two-hybrid assay requires only a few cells, which greatly reduces the workload of quantitative FRET measurements. Furthermore, due to the superior robustness and sensitivity of ExEm-unmixing, ExEm-FRET two-hybrid assay is especially applicable for cells with low fluorescence signal, thus provides a powerful tool for exploring the molecular mechanism of intracellular signal transduction in living cells.

Funding

National Natural Science Foundation of China (NSFC) (Grant No. 61875056 and 61527825).

References

1. E. S. Butz, M. Ben-Johny, M. Shen, P. S. Yang, L. Sang, M. Biel, D. T. Yue, and C. Wahl-Schott, “Quantifying macromolecular interactions in living cells using FRET two-hybrid assays,” Nat. Protoc. 11(12), 2470–2498 (2016). [CrossRef] [PubMed]

2. M. Y. Du, F. F. Yang, Z. H. Mai, W. F. Qu, F. R. Lin, L. C. Wei, and T. S. Chen, “FRET two-hybrid assay linearly fitting FRET efficiency to concentration ratio between acceptor and donor,” Appl. Phys. Lett. 112(15), 153702 (2018). [CrossRef]

3. M. G. Erickson, B. A. Alseikhan, B. Z. Peterson, and D. T. Yue, “Preassociation of calmodulin with voltage-gated Ca(²⁺) channels revealed by FRET in single living cells,” Neuron 31(6), 973–985 (2001). [CrossRef] [PubMed]

4. T. Zal and N. R. Gascoigne, “Photobleaching-corrected FRET efficiency imaging of live cells,” Biophys. J. 86(6), 3923–3939 (2004). [CrossRef] [PubMed]

5. A. D. Hoppe, B. L. Scott, T. P. Welliver, S. W. Straight, and J. A. Swanson, “N-way FRET microscopy of multiple protein-protein interactions in live cells,” PLoS One 8(6), e64760 (2013). [CrossRef] [PubMed]

6. F. Lin, C. Zhang, M. Du, L. Wang, Z. Mai, and T. Chen, “Superior robustness of ExEm-spFRET to IIem-spFRET method in live-cell FRET measurement,” J. Microsc. 272(2), 145–150 (2018). [CrossRef] [PubMed]

7. M. Du, L. Zhang, S. Xie, and T. Chen, “Wide-field microscopic FRET imaging using simultaneous spectral unmixing of excitation and emission spectra,” Opt. Express 24(14), 16037–16051 (2016). [CrossRef] [PubMed]

8. Y. Sun, H. Wallrabe, C. F. Booker, R. N. Day, and A. Periasamy, “Three-color spectral FRET microscopy localizes three interacting proteins in living cells,” Biophys. J. 99(4), 1274–1283 (2010). [CrossRef] [PubMed]

9. J. Wlodarczyk, A. Woehler, F. Kobe, E. Ponimaskin, A. Zeug, and E. Neher, “Analysis of FRET signals in the presence of free donors and acceptors,” Biophys. J. 94(3), 986–1000 (2008). [CrossRef] [PubMed]

10. C. Zhang, F. Lin, M. Du, W. Qu, Z. Mai, J. Qu, and T. Chen, “Simultaneous measurement of quantum yield ratio and absorption ratio between acceptor and donor by linearly unmixing excitation-emission spectra,” J. Microsc. 270(3), 335–342 (2018). [CrossRef] [PubMed]

11. W. Su, M. Du, F. Lin, C. Zhang, and T. Chen, “Quantitative FRET measurement based on spectral unmixing of donor, acceptor and spontaneous excitation-emission spectra,” J. Biophotonics 12(4), e201800314 (2019). [CrossRef] [PubMed]

12. G. Heras-Martinez, J. Andrieu, B. Larijani, and J. Requejo-Isidro, “Quantifying intracellular equilibrium dissociation constants using single-channel time-resolved FRET,” J. Biophotonics 11(1), e201600272 (2018). [CrossRef]

13. C. Zhang, Y. Liu, H. Sun, F. Lin, Y. Ma, W. Qu, and T. Chen, “Automated E-FRET microscope for dynamical live-cell FRET imaging,” J. Microsc. 274(1), 45–54 (2019). [CrossRef] [PubMed]

14. F. R. Lin, M. Y. Du, F. F. Yang, L. C. Wei, and T. S. Chen, “Improved spectrometer-microscope for quantitative FRET measurement based on simultaneous spectral unmixing of excitation and emission spectra,” J. Biomed. Opt. 23(1), 1–10 (2018). [PubMed]

15. A. J. Valentijn, A. D. Metcalfe, J. Kott, C. H. Streuli, and A. P. Gilmore, “Spatial and temporal changes in Bax subcellular localization during anoikis,” J. Cell Biol. 162(4), 599–612 (2003). [PubMed]

16. A. Aranovich, Q. Liu, T. Collins, F. Geng, S. Dixit, B. Leber, and D. W. Andrews, “Differences in the mechanisms of proapoptotic BH3 proteins binding to Bcl-XL and Bcl-2 quantified in live MCF-7 cells,” Mol. Cell 45(6), 754–763 (2012). [CrossRef] [PubMed]

17. C. King, S. Sarabipour, P. Byrne, D. J. Leahy, and K. Hristova, “The FRET signatures of noninteracting proteins in membranes: simulations and experiments,” Biophys. J. 106(6), 1309–1317 (2014). [CrossRef] [PubMed]

18. F. Edlich, S. Banerjee, M. Suzuki, M. M. Cleland, D. Arnoult, C. Wang, A. Neutzner, N. Tjandra, and R. J. Youle, “Bcl-x(L) retrotranslocates Bax from the mitochondria into the cytosol,” Cell 145(1), 104–116 (2011). [CrossRef] [PubMed]

19. H. Düssmann, M. Rehm, C. G. Concannon, S. Anguissola, M. Würstle, S. Kacmar, P. Völler, H. J. Huber, and J. H. Prehn, “Single-cell quantification of Bax activation and mathematical modelling suggest pore formation on minimal mitochondrial Bax accumulation,” Cell Death Differ. 17(2), 278–290 (2010). [CrossRef] [PubMed]

20. A. Kelekar, B. S. Chang, J. E. Harlan, S. W. Fesik, and C. B. Thompson, “Bad is a BH3 domain-containing protein that forms an inactivating dimer with Bcl-XL,” Mol. Cell. Biol. 17(12), 7040–7046 (1997). [CrossRef] [PubMed]

21. J. Zha, H. Harada, K. Osipov, J. Jockel, G. Waksman, and S. J. Korsmeyer, “BH3 domain of BAD is required for heterodimerization with BCL-XL and pro-apoptotic activity,” J. Biol. Chem. 272(39), 24101–24104 (1997). [CrossRef] [PubMed]

22. E. Yang, J. Zha, J. Jockel, L. H. Boise, C. B. Thompson, and S. J. Korsmeyer, “Bad, a heterodimeric partner for Bcl-XL and Bcl-2, displaces Bax and promotes cell death,” Cell 80(2), 285–291 (1995). [CrossRef] [PubMed]

23. A. M. Petros, D. G. Nettesheim, Y. Wang, E. T. Olejniczak, R. P. Meadows, J. Mack, K. Swift, E. D. Matayoshi, H. Zhang, C. B. Thompson, and S. W. Fesik, “Rationale for Bcl-xL/Bad peptide complex formation from structure, mutagenesis, and biophysical studies,” Protein Sci. 9(12), 2528–2534 (2000). [CrossRef] [PubMed]

ExEm-FRET two-hybrid assay: FRET two-hybrid assay based on linear unmixing of excitation-emission spectra

Abstract

1. Introduction