3.1.

Reagent and Plasmids

Cerulean, Venus-kras, and YFP-Bax plasmids were purchased from Addgene Company (Cambridge, Massachusetts). The FRET-standard constructs, including Cerulean-32-Venus (C32V, Addgene plasmid 29396), Cerulean-5-Venus-5-Cerulean (CVC, Addgene plasmid 27788), Venus-5-Cerulean-5-Venus (VCV, Addgene plasmid 27788), and Venus-5-Cerulean-5-Venus-6-Venus (VCVV, Addgene plasmid 27789), were kindly provided by the Vogel lab (National Institutes of Health, Bethesda, Maryland).

3.2.

Cell Culture and Transfection

Huh-7 cells, a human hepatocellular carcinoma cell line, were obtained from the Department of Medicine, Jinan University, Guangzhou, China. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Grand Island, New York) containing 10% fetal calf serum (FCS; Sijiqing, Hangzhou, China) at 37°C under 5% ${\mathrm{CO}}_2$ in a humidified incubator. For transfection, cells were cultured in DMEM containing 10% FCS in a 35-mm glass dish with a density of $4\times {10}^4\mathrm{cells}/\mathrm{ml}$ at 37°C under 5% ${\mathrm{CO}}_2$ in a humidified incubator. After 24 h, when the cells reached about 70% to 90% confluence, plasmid was transfected into the Huh-7 cells for 24-48 h. Turbofect™ (Fermentas Inc., Glen Burnie, Maryland) was used as a transfection reagent.

3.3.

Statistical Analysis

Results are expressed as $\mathrm{mean}\pm \mathrm{SD}$. Data were analyzed by repeated measures ANOVA with parametric methods using the statistical software SPSS 18.0 (SPSS, Inc., Chicago, Illinois). Throughout the work, $P$ values less than 0.05 were considered to be statistically significant.

4.1.

mlux-FRET: Modified lux-FRET Theory

With two excitation wavelengths, lux-FRET can simultaneously measure the absolute $E$ and $R_{\mathrm{C}}$ of the FRET samples including $D\sim A$ FRET constructs and free donors and acceptors.^11,17 In addition, for the time-invariant and uniform acceptor/donor ratio such as the tandem construct sensors, after an initial spectrally resolved dual-excitation calibration, the user can perform repetitive single-excitation wavelength measurements to quantify $E$ at high temporal resolution.^11,17 Here, we developed a modified lux-FRET method, mlux-FRET, to measure the absolute $E$ and $R_{\mathrm{C}}$ of the FRET samples including $mD\sim nA$ FRET constructs, free donors, and acceptors. Considering an FRET sample containing $mD\sim nA$ FRET constructs, free donors, and free acceptors, the emission spectra should be a linear combination of five contributions from the free donor, the donor in $mD\sim nA$ FRET constructs, the free acceptor and the acceptor in $mD\sim nA$ FRET constructs including direct excitation and sensitized emission:

Sorting these terms in Eq. (1) according to those with emission characteristics of the donor and acceptor, respectively, we obtain

We define two apparent relative donor and acceptor concentrations, respectively, as follows:

We can obtain the corresponding excitation ratio ($r^{\mathrm{ex},i}$) at ${\lambda }^i$ excitation from the reference measurements [Eqs. (3) and (4)]

Considering two different excitation wavelengths, Eqs. (5) and (6) represent three independent equations [since Eq. (5) does not rely on the extinction coefficient ratio ${\gamma }^i={\epsilon }_D^i/{\epsilon }_A^i$]. From Eqs. (5), (6), and (8), we can obtain the apparent efficiency ($E_{\mathrm{app}}$) and $R_C$:

In fact, the lux-FRET method proposed by Wlodarczyk et al. has the same $E{f}_D$ and $R_{\mathrm{C}}$ equations as Eqs. (9) and (11) except that the $E{f}_A$ equation is different from Eq. (10) and is given as^11,17

4.2.

spE-FRET: Spectra-Based 2-Channel E-FRET Method

Based on the measured ${\mathrm{count}}^i(\lambda )$ ($i=1$, 2) spectra at two different excitations, we developed a spectra-based E-FRET method, spE-FRET, to calculate the $E$ and $R_{\mathrm{C}}$ values. $E$ was calculated as follows:

We can also measure $R_{\mathrm{C}}$ as follows:

5.1.

Spectral Fingerprints of Venus and Cerulean in Living Huh-7 Cells

In order to use our platform to measure the spectral fingerprints of Cerulean [$e_D(\lambda )$] and Venus [$e_A(\lambda )$] in living Huh-7 cells, we first used a standard lamp to calibrate the instrument. Calibration was performed with a halogen tungsten lamp calibrated by a spectrometer (QE65 Pro, Ocean Optics, Florida) that was calibrated by a standard light source (LS-1-CAL, Ocean Optics, Florida). Figure 2(a) showed the reference spectrum [$E(\lambda )$] of the halogen lamp directly measured by the calibrated QE65 Pro spectrometer, the spectrum [${\mathrm{count}}_{\mathrm{lamp}}(\lambda )$] of the halogen lamp obtained by our spectroscopic widefield fluorescence microscope, and the corresponding spectral sensitivity calibration curve: $K(\lambda )=E(\lambda )/{\mathrm{count}}_{\mathrm{lamp}}(\lambda )$. The same $K(\lambda )$ curves were obtained at an interval of 3 months [Fig. 2(b)], demonstrating the stability of the spectrometer-microscope platform.

Fig. 2

Spectral fingerprints of Cerulean [$e_D(\lambda )$] and Venus [$e_A(\lambda )$] inside living Huh-7 cells. (a) Spectral sensitivity calibration curve $K(\lambda )$ of the spectrometer-microscope platform. (b) Spectral sensitivity calibration curves $K_1(\lambda )$ and $K_2(\lambda )$ measured at intervals of 3 months. (c) Spectra of a representative cell expressing Cerulean and the corresponding background from a cell without expressing fluorescent proteins. Scale bar: $10\mu \mathrm{m}$. (d) Normalized emission spectra of Cerulean inside living cells from at least 15 cells. (e) Same as (c) except for Venus. Scale bar: $10\mu \mathrm{m}$. (f) Same as (d) except for Venus from at least 15 living cells.

Figure 2(c) showed the spectra of a representative living Huh-7 cell expressing Cerulean (upper two lines) and the corresponding background from cells without expressing FPs (lower two lines), showing that the autofluorescence and background were very low. The mean spectra from cells without expressing FPs under the same measurement conditions were subtracted from the measured fluorescence spectrum. Although 436-nm excitation excited Cerulean more effectively [Fig. 2(c)], the normalized spectrum [Fig. 2(d), black line] was the same as the normalized spectrum with a 405-nm excitation [Fig. 2(d), dashed line], and this spectrum was used as the spectral fingerprint [$e_D(\lambda )$] of Cerulean. Similarly, the normalized spectrum of Venus in living Huh-7 cells with 436-nm excitation was used as the spectral fingerprint [$e_A(\lambda )$] of Venus [Fig. 2(e)]. In reality, $e_D(\lambda )$ and $e_A(\lambda )$ were obtained from at least 15 living cells.

Although the measured $\mathrm{count}(\lambda )$ of Cerulean decreased with the decreasing exposure time, normalization showed the same $\mathrm{count}(\lambda )$ spectra (data not shown), indicating that the exposure time of spectrometer did not influence the spectral fingerprints. The shortest exposure time of spectrometer is 8 ms, and in this report, the exposure time is 50 ms.

The stability of the platform is very important for FRET quantification in living cells. Some FRET quantification methods need to calibrate the instrument setup.^9,10,15 For example, three external references are required to determine the instrument setup calibration factor ($G$) for the E-FRET method^9,15 and the sp-RET method needs to rigorously calibrate the spectral response of the complete optical setup.¹⁰ Although some FRET quantification methods do not need to calibrate the optical setup, the correction coefficients for removing spectral crosstalks also depend on the stability of the instrument.^11,16,17 All SE-FRET methods need complicated crosstalk correction,^{9–11,15–19} and these correction coefficients are intrinsic to the given fluorophores and optical setup as well as the illumination.^{1,9,10,17,21,27} In reality, the fluorophores and illumination light are very stable.^9,21,27 Therefore, once these correction coefficients and calibration factors are predetermined for a given choice of fluorophores and instrument, it is not necessary to measure them for every experiment with a steady-state FRET instrument.^{1,9,10,19,21,27}

Although our spectrometer-microscope platform is very stable over at least 3 months [Fig. 2(b)], periodic calibration consistently improves the fit quality and the reliability of the measurements. In reality, calibration of the optical setup is very easy, and the normative calibration can be completed in a few minutes. Since the EC is independent of the emission controller, we, here, detected the two spectra at 405- or 436-nm excitation with the same emission channels. Therefore, we used the same calibration curve to correct the measured spectra with different excitations, which is contrary to the sp-RET method proposed by Levy et al.¹⁰

In fact, the calibration step [Fig. 2(a)] is not mandatory. Spectral fingerprints of the donor [$e_D(\lambda )$] and acceptor [$e_A(\lambda )$] are independent of instrument and illumination, thus they can also be obtained by using other instruments such as a spectrofluorometer.¹⁷ Moreover, once $e_D(\lambda )$ and $e_A(\lambda )$ were determined, we did not need to measure them for every experiment in the specific cell line. In order to insure the accurate measurement of total donor and acceptor emissions, the acquisition range should overlap well with their emission spectra. The spectral ranges chosen in this study fulfill this requirement, covering $\sim 96.3\%$ and $\sim 96.8\%$ of the Cerulean and Venus emission spectra, respectively [Figs. 2(d) and 2(f)].

5.2.

Implementation of mlux-FRET for Measuring the $E$ of C32V in Living Cells

To use mlux-FRET to measure the $E$ of C32V in living cells, we first measured four reference emission spectra of the Cerulean-only sample and the Venus-only sample, respectively, with two different excitations, and then measured two emission spectra of the C32V sample at the two different excitations under the same measurement conditions. Figure 3(a) showed the images of cells separately expressing Cerulean and Venus with 405-nm (${\lambda }^1$) and 436-nm (${\lambda }^2$) excitations, respectively, and the reference spectra inside the living cell indicated by the red circles ${\mathrm{count}}_D^1(\lambda )(C_D^1)$, ${\mathrm{count}}_A^1(\lambda )(C_A^1)$ and ${\mathrm{count}}_D^2(\lambda )(C_D^2)$, ${\mathrm{count}}_A^2(\lambda )(C_A^2)$ were shown in Figure 3(b). Excitation ratios were determined from the above spectra as: $r^{\mathrm{ex},1}=2.29$ [Fig. 3(c)] and $r^{\mathrm{ex},2}=1.21$ [Fig. 3(d)], using quantum yield values of donor and acceptor, $Q_D=0.62$ and $Q_A=0.57$.²⁸ Figure 3(e) showed the images of representative cells expressing C32V under the same measurement conditions as Fig. 3(a), and the spectra and mlux-FRET unmixing (black line) for the live cell indicated by red circles were shown in Figs. 3(f) and 3(g). The estimated parameters were ${\alpha }^1=8.69$, ${\alpha }^2=6.20$, and ${\delta }^1=5.35$. Substitute $m/n=1$, $r^{\mathrm{ex},i}$, ${\alpha }^i$, and ${\delta }^i$ values into Eqs. (9) to (11) to obtain $E=29.77\%$ and $R_{\mathrm{TC}}=0.445$. The statistical results from at least 13 cells expressing C32V were $E(E=E{f}_D=E{f}_A)=28.67\pm 0.95\%$ and $R_{\mathrm{TC}}=0.442\pm 0.008$.

Fig. 3

Implementation of mlux-FRET on the spectrometer-microscope platform for measuring the $E$ of C32V construct inside single living Huh-7 cells. (a) Images of cells separately expressing Cerulean and Venus with 405- or 436-nm excitation. Scale bar: $10\mu \mathrm{m}$. (b) Emission count spectra of Cerulean and Venus inside the living cell indicated by the red circles in (a). (c) and (d) Calculation of excitation ratios $r^{\mathrm{ex},i}$. (e) Image of cells expressing C32V with 405- or 436-nm excitation under the same measurement conditions as the Cerulean-only or Venus-only sample in (a). Scale bar: $10\mu \mathrm{m}$. (f) and (g) Spectral unmixing of C32V with 405-nm (e) or 436-nm (d) excitation.

In reality, we must be very careful when obtaining the reference spectra of Venus with 405-nm excitation due to the very low extinction coefficients of Venus at 405 nm [Fig. 3(a)]. Although we can use strong 405-nm excitation to excite Venus, this strong illumination significantly bleached the Venus in C32V (data not shown). Therefore, we must choose cells expressing a high level of Venus as a reference. Although Wlodarczyk et al. used 458- and 488-nm excitations to perform lux-FRET for ECFP-EYFP pair,¹⁷ it is very hard to use a 488 nm laser line to excite Cerulean or ECFP,¹⁰ which was confirmed by the fact that the $r^{\mathrm{ex},2}$ (488-nm excitation) was 0.02, which is much less than the $r^{\mathrm{ex},1}=2.29$ (458-nm excitation).¹⁷ For 458- and 488-nm excitations, an alternative choice is to use GFP as a donor and YFP or Venus as an acceptor for the implementation of mlux-FRET on a confocal microscope.

The acceptor cross-excitation fraction is intrinsic to the extinction coefficient ratio (${\gamma }^i$) of the donor to acceptor and their relative concentrations. Different ${\gamma }^i$ at two different excitations is necessary for the mlux-FRET method according to Eq. (4). Although the two ratios ${\gamma }^i$ ($i=1$, 2) are related to the cell lines to some extent, they depend mainly on the excitation wavelengths intrinsic to the excitation light source and the transmission properties of the instrument for a given donor-acceptor pair. 405-nm excitation can effectively excite Cerulean/CFP but can hardly excite Venus/YFP, 458-nm excitation can effectively excite both Cerulean/CFP and Venus/YFP, and 488-nm excitation can effectively excite Venus/YFP but hardly excite Cerulean/CFP. Although the ${\gamma }^1/{\gamma }^2$ ratio with 458- and 488-nm excitations is 114.5,⁹ it is inapplicable to the CFP (Cerulean)–YFP (Venus) pair due to the very low extinction coefficient of CFP (Cerulean) at the 488-nm wavelength. For our spectrometer-microscope platform, 405- and 436-nm excitations (${\gamma }^1/{\gamma }^2=1.89$) induced significantly different spectra for C32V, CVC, VCV, and VCVV constructs [Figs. 5(b)–5(e)]. Another better choice, in this view, is the 430- and 445-nm excitations for the CFP (Cerulean)–YFP (Venus) pair.

5.3.

Implementation of spE-FRET for Measuring the E of C32V in Living Cells

For spE-FRET, we first calculated the spectral crosstalk coefficients according to Eq. (15). We choose 470.107 to 470.883 nm emission data as $\mathrm{count}({\lambda }_1)$ and 530.452 to 531.223 nm emission data as $\mathrm{count}({\lambda }_2)$. From the measured spectra in Fig. 3(b), the crosstalk correction coefficients for Cerulean and Venus on our spectrometer-microscope platform were $a=0.142$, $b=0.001$, $c=2.650$, and $d=0.665$. In order to determine the parameter $G$, we next used a strong 500-nm excitation to selectively bleach partial acceptors inside the same cell as that in Fig. 3(e). Figure 4(a) showed the images of cells in Fig. 3(e) before and after partially photobleaching the acceptor, and the corresponding count spectra of the cell indicated by the red circle in Fig. 4(a) were shown in Fig. 4(b). The degree of PbFRET ($x$) was 35.46% [inset figure in Fig. 4(b)]. Substitute ${\mathrm{count}}^1({\lambda }_1)=70.478$, ${\mathrm{count}}^{1,\mathrm{post}}({\lambda }_1)=78.677$, ${\mathrm{count}}^1({\lambda }_2)=170.020$, ${\mathrm{count}}^{1,\mathrm{post}}({\lambda }_2)=143.801$, ${\mathrm{count}}^2({\lambda }_2)=773.200$, and ${\mathrm{count}}^{2,\mathrm{post}}({\lambda }_2)=660.001$ [Fig. 4(b)] into Eq. (16) to obtain $G=1.49$, and the statistical $G=1.46\pm 0.05$ from at least 13 living cells. Substitute $G=1.46$ and these count values into Eqs. (13) and (17) to obtain $E=27.75\%$ and $R=0.128$ for the cells indicated by the red circle in Fig. 4(a), and the statistical results from 13 living cells were $E=28.22\pm 3.21\%$ and $R=0.129\pm 0.007$.

Fig. 4

Implementation of spE-FRET method on the spectrometer-microscope platform for measuring the $E$ of C32V construct inside single living Huh-7 cells. (a) Images of cells expressing C32V with 405 nm (left panels) or 436 nm (right panels) excitation before (upper panels) and after (lower panels) partially photobleaching Venus. Scale bar: $10\mu \mathrm{m}$. (b) Count spectra of C32V inside the living cell indicated by red circles in (a). Inset figure: emission spectra of C32V inside the living cells indicated by red circle in (a) before (upper line) and after (lower line) partially photobleaching Venus. ${\lambda }_1$: 470.107 to 470.883 nm and ${\lambda }_2$: 530.452 to 531.223 nm.

With spE-FRET, the sensitized fluorescence ${\mathrm{count}}^{i,\mathrm{SE}}({\lambda }_2)$ is calculated by subtraction of the major crosstalk components from ${\mathrm{count}}^1({\lambda }_2)$, and the minor crosstalk components from ${\mathrm{count}}^1({\lambda }_1)$ and ${\mathrm{count}}^2({\lambda }_2)$ according to Eq. (14), which is applicable only when the donor channel selectively or mainly collects donor emission.⁹ For ${\lambda }_1=500$ to 510 nm and ${\lambda }_2=560.454$ to 561.221 nm, $a=0.142$, $b=0.613$, $c=0.056$, and $d=0.014$, where $b$ is not ignorable, and the corresponding $G=0.054\pm 0.002$ from at least 12 living cells and $E=23.10\%$, which is much less than 27.75%, further demonstrating our recent study that spE-FRET is also very sensitive to the acceptor emission bleedthrough to the donor channel.²⁴ It is generally difficult to choose an appropriate optical filter as a donor emission channel to selectively collect donor emission on a conventional widefield fluorescence microscope.^9,22 However, it is very easy to choose a wideband ${\lambda }_2$ (525 to 535 nm) and a narrowband ${\lambda }_1$ (476 to 476.69 nm) to make the crosstalk coefficient $b=0.0001$ negligible for the Cerulean-Venus pair on our spectrometer-microscope platform. Moreover, choosing 485.624 to 486.399 nm emission data as $\mathrm{count}({\lambda }_1)$, and 520.426 to 550.466 nm emission data as $\mathrm{count}({\lambda }_2)$ for the GFP–Venus pair, the corresponding $b=0.002$ is also negligible.

5.4.

Quantifying Tandem Constructs with Different Acceptor/Donor Stoichiometries in Single Living Huh-7 Cells Using mlux-FRET and spE-FRET Methods

We next used mlux-FRET and spE-FRET methods to measure the $E$ values of C32V, CVC, VCV, and VCVV constructs inside single living Huh-7 cells. For a live cell expressing the FRET tandem construct, we first measured the spectra at two different excitations. Figure 5(a) showed the images of representative cells expressing C32V, CVC, VCV, and VCVV, respectively, under the same measurement conditions as the Cerulean-only and Venus-only samples in Fig. 3(a). The corresponding normalized spectra of the cell indicated by the red circle in Fig. 5(a) with ${\lambda }^i$ excitation and the corresponding values calculated by mlux-FRET and spE-FRET, respectively, were shown in Figs. 5(b)–5(e). The statistical $E$ values from at least 16 live Huh-7 cells in four independent experiments calculated by the two methods showed consistent results for C32V, CVC, VCV, and VCVV constructs, respectively [Fig. 5(f)], and the corresponding $R_{\mathrm{C}}$ values calculated by spE-FRET and mlux-FRET methods were shown in Fig. 5(g).

Fig. 5

Implementation of both mlux-FRET and spE-FRET methods on the spectrometer-microscope platform for measuring the $E$ and $R_{\mathrm{C}}$ values of tandem constructs with different acceptor/donor stoichiometries inside single living Huh-7 cells. (a) Images of cells separately expressing C32V, Cerulean-5-Venus-5-Cerulean (CVC), Venus-5-Cerulean-5-Venus (VCV), and Venus-5-Cerulean-5-Venus-6-Venus (VCVV) tandem construct with 405- (upper panels) and 436-nm excitations (lower panels). Scale bar: $10\mu \mathrm{m}$. (b)–(e) Count spectra normalized to unit area of C32V (b), CVC (c), VCV (d), and VCVV (e) constructs inside the cell indicated by red circles in (a) with 405- and 436-nm excitation, respectively. (f) Statistical $E$ values of C32V, CVC, VCV, and VCVV tandem constructs in living Huh-7 cells obtained by mlux-FRET and spE-FRET methods, respectively. (g) Statistical $R_{\mathrm{C}}$ values obtained by spE-FRET and mlux-FRET from Huh-7 cells exclusively expressing CVC, VCV, and VCVV tandem construct. (h) $E{f}_D$ and $E{f}_A$ of C32V, CVC, VCV, and VCVV tandem constructs with different donor/acceptor stoichiometries in living Huh-7 cells.

The $E{f}_A$ formalism [Eq. (10)] of mlux-FRET is different from that [Eq. (12)] of lux-FRET. As shown in Fig. 5(h), the $E{f}_A$ values calculated by Eq. (10) were consistent with the $E{f}_D$ values, while the $E{f}_A$ values calculated by Eq. (12) were inconsistent with the $E{f}_D$ values except for the C32V construct, verifying the ability of mlux-FRET methods to determine the $E{f}_A$ for $mD-nA$ system.

In fact, both the mlux-FRET and spE-FRET methods can be used to measure $E{f}_D$ and $R_{\mathrm{C}}$ values of the $mD\sim nA$ construct where the “$m$” and “$n$” are unknown. However, a mlux-FRET method can be used to measure the $E{f}_A$ value of $mD\sim nA$ construct only when the $m/n$ ratio is known. In reality, the relative concentration of each molecule contained in a biological complex can be predetermined by biotechnology such as flow cytometry, mass spectrometry, or western blotting.¹⁴