Fluorescence resonance energy transfer of GFP and YFP by spectral imaging and quantitative acceptor photobleaching


Dr. A.B. Houtsmuller. Tel: +31 10 7044456; e-mail: a.houtsmuller@erasmusmc.nl


To study protein–protein interactions by fluorescence energy transfer (FRET), the proteins of interest are tagged with either a donor or an acceptor fluorophore. For efficient FRET, fluorophores need to have a reasonable overlap of donor emission and acceptor excitation spectra. However, given the relatively small Stokes shift of conventional fluorescent proteins, donor and acceptor pairs with high FRET efficiencies have emission spectra that are difficult to separate. GFP and YFP are widely used in fluorescence microscopy studies. The spectral qualities of GFP and YFP make them one of the most efficient FRET donor–acceptor couples available. However, the emission peaks of GFP (510 nm) and YFP (527 nm) are spectrally too close for separation by conventional fluorescence microscopy. Difficulties in simultaneous detection of GFP and YFP with a fluorescence microscope are eliminated when spectral imaging and subsequent linear unmixing are applied. This allows FRET microscopy using these tags to study protein–protein interactions. We adapted the linear unmixing procedure from commercially available software (Zeiss) for use with acceptor photobleaching FRET using GFP and YFP as FRET pair. FRET efficiencies up to 52% for a GFP-YFP fusion protein were measured. To investigate the applicability of the procedure, we used two constituents of the nucleotide excision repair system, which removes UV-induced single-strand DNA damage. ERCC1 and XPF form a heterodimeric 5′ endonuclease in nucleotide excision repair. FRET between ERCC1-GFP and XPF-YFP occurs with an efficiency of 30%.


Protein–protein interactions can be detected in living cells in real time with fluorescence resonance energy transfer (FRET) microscopy (Jares-Erijman & Jovin, 2003). FRET is the non-radiative energy transfer from an excited donor fluorophore to an acceptor, which then emits photons at its emission wavelength. The efficiency of this energy transfer depends on several factors: the overlap between donor emission and acceptor excitation spectra, relative orientation of the transition dipoles, the quantum yield of the donor, the extinction coefficient of the acceptor and the (sixth power of the) distance between the fluorophores (Lakowicz, 1999; Piston & Kremers, 2007).

FRET induces a number of characteristic changes in the donor and acceptor fluorescence, which can be detected in various ways: (1) Fluorescence lifetime imaging (FLIM) can be used to determine FRET since the fluorescence lifetime of the donor is reduced when energy transfer to an appropriate acceptor occurs (Hink et al., 2002; Clegg et al., 2003). (2) FRET can also be quantitatively assessed by measuring the indirect excitation of acceptor fluorophores resulting from energy transfer by a directly excited donor at the appropriate wavelength (Honda et al., 2001; Ponsioen et al., 2004). This method is generally referred to as sensitized emission. A number of correction protocols have been described for the accurate microscope calibration and data collection that is required with this technique (Gordon et al., 1998; Hoppe et al., 2002; Jares-Erijman & Jovin, 2003). (3) Acceptor photobleaching (abFRET) is a technique that measures FRET by irreversibly photobleaching the acceptor through application of high-intensity light at the acceptor excitation wavelength, and thereby abrogating the energy transfer. When the acceptor is made non-fluorescent by intense irradiation of the excitation wavelength (acceptor photobleaching), the donor will regain its original fluorescence. This resulting increase of donor fluorescence after abFRET can therefore be used to quantitatively determine FRET efficiency (Bastiaens et al., 1996; Karpova et al., 2003; Gu et al., 2004; Van Munster et al., 2005).

At present, there is a large availability of possible donor–acceptor pairs that can be used to study FRET in living cells, but most live-cell FRET studies have employed the cyan and yellow variants of GFP, CFP and YFP, as donor and acceptor fluorophores, respectively (Karpova et al., 2003; Gu et al., 2004; Ponsioen et al., 2004; Van Munster et al., 2005). A potential problem of the CFP-YFP FRET couple is that the fluorescence intensity or brightness (quantum yield × molar extinction coefficient) of CFP is approximately four times lower than that of YFP (Rizzo et al., 2004). This makes it difficult to detect FRET signals in cells expressing small amounts of tagged proteins. In contrast to CFP, GFP has a quantum yield comparable to YFP and its emission spectrum overlaps more closely with the excitation spectrum of YFP than CFP. This higher overlap results in an increase in FRET efficiency (Patterson et al., 2000). Moreover, many research groups have generated numerous GFP fusions of different proteins of interest and extension of research to interaction studies only requires generation of YFP-tagged potential interaction partners if GFP-YFP could be used as a FRET couple. However, simultaneous imaging of GFP and YFP is not possible using conventional microscopy because of the large overlap of the two emission spectra. Here, we describe a quantitative acceptor photobleaching method, based on spectral imaging, for live-cell interaction studies using GFP and YFP as a FRET couple.


Experimental set-up

To study protein–protein interactions in living-cell nuclei, we have set up a system to measure fluorescence energy transfer from GFP to YFP. First, we optimized a linear unmixing procedure to separate GFP and YFP signals using spectral confocal microscopy. Subsequently, we investigated the possibility to use this FRET pair in a quantitative live-cell abFRET assay by determining FRET values in cells expressing GFP-YFP fusion protein and unfused GFP and YFP at different ratios. Finally, we validated the applicability of the method in living cells expressing two interacting DNA repair proteins (ERCC1 and XPF) tagged with GFP and YFP.

Linear unmixing

The spectral overlap between GFP and YFP is too large to allow separate detection using conventional microscopy (Fig. 1(A)). To use GFP and YFP in protein–protein interaction studies by means of FRET, we investigated the use of spectral imaging and subsequent linear unmixing to separate these two fluorophores. In linear unmixing, reference spectra are used to determine the contribution of multiple fluorophores to a measured spectrum. We first obtained reference spectra of GFP and YFP by separately recording cells that express either GFP or YFP (Fig. 1(B)). Using spectral imaging settings of a Zeiss LSM 510 META (Zeiss, Jena, Germany), GFP and YFP spectra were recorded in eight channels (between 486 nm and 572 nm). Excitation of GFP and YFP by 488 nm or 514 nm (for YFP) resulted in detection of laser-light reflection in the spectral images (data not shown). To avoid this problem, both fluorophores were excited by 458 nm (Fig. 1(A)). Subsequently, spectra of cells expressing both GFP and YFP were recorded and linear unmixing was performed (Fig. 1(C)). The least-squares method was used to fit normalized reference spectra of GFP and YFP to the spectrum measured in cells expressing both GFP and YFP (Dickinson et al., 2001; Hiraoka et al., 2002; Gu et al., 2004). The contribution of GFP and YFP to a recorded spectrum is described by


where Itot(λ) is the total fluorescence at wavelength λ, g and y are the relative abundance factors of GFP and YFP, respectively, and Ig(λ) and Iy(λ) are the fluorescence intensities at wavelength λ for the GFP and YFP reference spectra, respectively. To fit normalized GFP and YFP reference spectra to measured fluorescence spectra in the nuclei of living cells, values for g and y were determined that minimize the sum:


where i indicates a spectral channel (eight channels in total, see above).

Figure 1.

GFP and YFP absorption and reference spectra and unmixing. (A) Absorption (continuous line) and emission (dotted line) spectra of GFP (green) and YFP (yellow). The 458-nm line (purple vertical line) was used for excitation and the 514-nm line (green vertical line) for acceptor photobleaching. The 488-nm line (light blue) was not used for excitation because it interfered with spectral imaging. The GFP and YFP emission spectra (dotted lines) show a large overlap, which makes these two fluorophores inseparable with conventional microscopy using band-pass filters. Reference for GFP and YFP spectra: Asahi Spectra USA Inc, http://www.asahi-spectra.com/. (B) Background corrected normalized reference spectra of GFP and YFP (Ig (λ) and Iy (λ), respectively). Boxes outlined in white indicate spectral channels. The light blue vertical line represents the 488-nm laser and indicates that this laser line was detected in the first spectral channel. Actual GFP and YFP emission spectra are shown as dashed lines. (C) Unmixing example before and after acceptor photobleaching. Black continuous lines represent measured spectra (I (λ)) containing GFP and YFP signals. Green lines are fits of the GFP reference spectrum (Ig (λ)) to Itot (λ), and yellow lines are fits of the YFP reference spectrum (Iy (λ)) to Itot (λ). Dotted lines are the sum of the GFP and YFP fits. The similarity of the black continuous and the dotted lines show the accuracy of the fits. Images on the right show the fluorescence signal in each of the eight spectral channels before and after photobleaching. Yellow circles represent the region of acceptor photobleaching and Itot (λ) measurement (the nucleus). Values for g and y are given.

For accurate unmixing of fluorophores with large variations in intensities (i.e. before and after acceptor photobleaching) background correction of reference spectra is essential. Because background values differ in every spectral channel, we recorded spectra of the sample background signals and subtracted these from the fluorescence spectra recorded in cells in the same image before applying linear unmixing.

Acceptor photobleaching

To determine FRET efficiencies, we used the acceptor photobleaching method. In this method, images are taken before and after photobleaching of the majority of acceptor molecules of the investigated FRET pair, in our case YFP. If any interactions leading to energy transfer were present in the cell, photobleaching of the acceptor will lead to an increase of donor fluorescence, as it is no longer quenched by the acceptor. Acceptor photobleaching was performed with a high-intensity laser pulse at a wavelength of 514 nm. To calculate FRET efficiency F in acceptor bleaching experiments, we calculated the fraction of donor molecules that was quenched by the acceptor before photobleaching as:


where Da is the donor intensity after and Db the donor intensity before abFRET. Before substituting Da and Db with the values for GFP after and before abFRET, Ga and Gb, respectively, two corrections were performed. Since the GFP and YFP excitation spectra overlap to some extent (Fig. 1(A)), a fraction of the GFPs will also be bleached. To determine the fraction of GFP bleached at 514 nm, relative to the fraction of bleached YFP, we subjected cells expressing either YFP or GFP to 514 nm laser pulses of increasing intensity. In the range of 514 nm laser intensities that were typically applied to deplete approximately 70–85% of YFP, the fraction (k) of GFP that was photobleached was 0.32 ± 0.03 of the YFP bleached fraction. This k value can be used to calculate the corrected GFP value after photobleaching YFP. inline image, with Yb the YFP value before photobleaching and Ya the YFP value after photobleaching, then defines the fraction of GFP that was bleached and Ga divided by 1 minus this fraction gives Ga:


Equation 4 can be rewritten as:


To correct for incomplete acceptor photobleaching, the difference between GFP before and after bleaching (GaGb) was divided by the fraction of YFP that was bleached inline image and then added to the GFP intensity before photobleaching (Gb):


where Ga is the corrected GFP intensity. Equation 6 can be rewritten as:


Equations 5 and 7 were substituted into the general abFRET equation (see Eq. 3):




Obviously, to get FRET percentages, Eq. 9 was simply multiplied by 100.

Quantitative acceptor photobleaching

To validate the method, we generated fusion constructs of GFP and YFP and of CFP and YFP, both with flexible linkers, as references (van Royen et al., 2007). To determine the efficiency of the GFP-YFP FRET couple, CHO cells were transfected with either a YFP-IRES-GFP (YIG) construct or with the GFP-YFP fusion (GYFP) construct. The internal ribosomal entry site (IRES) between GFP and YFP in YIG ensures that both GFP and YFP are separately translated from the same mRNA molecule. Three images were taken before and three images after acceptor photobleaching. No significant photobleaching was detected as a result of making three consecutive images at the laser intensity used (data not shown). Transfection with YIG results in the separate expression of free GFP and YFP, which should not show any FRET. The physical linkage between donor and acceptor by means of flexible linker of the GYFP construct should result in a high FRET signal. Indeed, performing abFRET on cells expressing YIG resulted in a small decrease of GFP fluorescence, 32% of the decrease in YFP fluorescence (see value 0.32 for k above and in Fig. 2(A)), whereas performing abFRET on cells expressing GYFP resulted in a clear increase of the GFP signal (Fig. 2(B)).

Figure 2.

GFP-YFP FRET quantification. (A) Cells were transfected with the YFP-IRES-GFP (YIG) construct and acceptor photobleaching (abFRET) was performed. Spectral images were unmixed with the Zeiss unmixing procedure. Red images represent YFP signals and green images GFP. No increase of GFP fluorescence was detected after abFRET. The graph in the right panel shows the measured spectra before and after abFRET of a typical cell transfected with YIG. (B) Cells were transfected with the GFP-YFP fusion construct (GYFP) and abFRET was performed. Spectral images were unmixed with the Zeiss unmixing procedure. Red images represent YFP signals and green images GFP. Strong increase of GFP fluorescence was detected after abFRET. The graph in the right panel shows the measured spectra before and after abFRET of a typical cell transfected with GYFP. (C) Cells were transfected with both YIG and GYFP in five different relative concentrations (closed diamonds): (0) Only YIG, (1/3) two-thirds YIG and one-third GYFP, (1/2) half of each, (2/3) one-third YIG and two-thirds GYFP and (1) only GYFP. A direct dependence was found between the relative concentration of GYFP (the interacting molecules) and FRET. Maximum FRET efficiency (GYFP only) was 52%. Error bars indicate two times the SEM.

Next, the dependence of measured FRET intensities on the fraction of interacting molecules was studied. Cells were co-transfected with GYFP and YIG in five different concentration ratios (0:1, 1:2, 1:1, 2:1 and 1:0, respectively, Fig. 2(C)). As expected, an increasing relative amount of the GYFP fusion-construct resulted in an increase in FRET (Fig. 2(C), closed diamonds). Maximum FRET was measured in cells transfected with only the GY construct (52 ± 1.2%). As a comparison, the same experiment (with slightly different relative concentrations 0:1, 1:1, 2:1, 4:1 and 1:0) was performed with the CFP-YFP FRET couple (Fig. 2(C), open diamonds). The same dependence of FRET intensities on relative concentrations of transfected fusion construct was observed, but FRET values were lower than for the GFP-YFP FRET couple (32 ± 1.4% for CYFP). These results indicate that abFRET combined with linear unmixing can be used to quantitatively assess molecular interactions. Furthermore, the high maximum efficiency of the GFP-YFP FRET couple makes the dynamic range of detectable changes in FRET larger than for CFP-YFP and probably most other fluorophore couples.

Biological example: ERCC1 and XPF

To study the applicability of abFRET and linear unmixing to biologically relevant proteins expressed at physiological levels, we used components of the nucleotide excision repair (NER) system, ERCC1 and XPF. ERCC1 and XPF form a heterodimer, which is responsible for the 5′ incision of the damaged strand during NER (Park et al., 1995; Houtsmuller et al., 1999). Formation of this heterodimer has been shown to be required for the stabilization of both proteins (Sijbers et al., 1996; Yagi et al., 1997; de Laat et al., 1998a). To investigate whether the interaction between ERCC1 and XPF could be detected by FRET, we generated two constructs: ERCC1-GFP and XPF-YFP. Both proteins were tagged at the C-termini, which also harbour the mutual interaction domains (de Laat et al., 1998b). Expression of ERCC1-GFP in 43-3B (ERCC1 –/–) and XPF-YFP in UV47 (XPF –/–) Chinese hamster ovary (CHO) cell lines restores the UV-C sensitivity to wild-type levels [(Houtsmuller et al., 1999) and data not shown], indicating the functionality of the tagged proteins. ERCC1-GFP was stably transfected into the ERCC1 –/– CHO cell line 43-3B, and this cell line was then transfected with XPF-YFP. Both ERCC1-GFP and XPF-YFP are mainly present in the nucleus but ERCC1-GFP can also be found in the cytoplasm (Fig. 3(A)). After abFRET, the GFP signal increased indicating FRET had taken place (Figs 3(A) & (B)). Quantification of the FRET between ERCC1-GFP and XPF-YFP showed a relatively high FRET percentage of 30 ± 2.4% (Fig. 3(C)).

Figure 3.

ERCC1 and XPF. (A) ERCC1 mutant CHO cells (43–3B) stably expressing ERCC1-GFP were transiently transfected with XPF-YFP and acceptor photobleaching was performed. Spectral images were unmixed with the Zeiss unmixing procedure. Shown are merged images with ERCC1-GFP in green and XPF-YFP in red. Both ERCC1-GFP and XPF-YFP are mainly located in the nucleus. ERCC1-GFP can also be found in the cytoplasm. After abFRET, the YFP signal is reduced and the GFP slightly increases. Circles indicate the region that is plotted in (B). (B) Graph of the abFRET procedure of cell shown in (A). Three images were taken before and three after abFRET. The black vertical bar indicates the moment of photobleaching. Upon photobleaching with 514 nm YFP (red) signal is reduced and GFP (green) increases. (C) Average FRET of ten ERCC1-GFP and XPF-YFP expressing cells is plotted in green. FRET intensities of YIG and GYFP are plotted alongside for comparison. (Average FRET between ERCC1-GFP and XPF-YFP is 30 ± 2.4%). Error bars indicate standard deviations.

Two factors that have to be taken into account when interpreting FRET intensities of protein–protein interactions are (1) the average FRET efficiency of the interacting molecules and (2) the fraction of donor-tagged molecules interacting with acceptor-tagged molecules. Both ERCC1 and XPF were tagged with their fluorophores near the respective interaction domains (Tripsianes et al., 2005), so the average FRET efficiency of the interacting ERCC1-GFP and XPF-YFP was likely close to that of the GFP-YFP fusion construct. Therefore, if every ERCC1-GFP molecule would interact with an XPF-YFP molecule, the expected FRET efficiency would be close to 52%, as it is for GFP-YFP. ERCC1 and XPF are known to interact with a stoichiometry of 1 to 1 (Park et al., 1995). However, in this experiment ERCC1-GFP was interacting both with endogenous XPF and with XPF-YFP. The ERCC1-GFP molecules that were interacting with endogenous XPF did not transfer energy through FRET to a YFP, but they were partly photobleached by the 514-nm laser pulse. Consequently, the presence of this non-FRETting ERCC1-GFP pool results in an underestimation of FRET values. Therefore, the surprisingly high FRET efficiency of ERCC1-GFP and XPF-YFP of 30% indicates that a very large fraction of ERCC1-GFP and XPF-YFP molecules were interacting. Most likely no free ERCC1-GFP exists in these cells.


We have used linear unmixing combined with acceptor photobleaching FRET to study interaction between the spectrally very close GFP and YFP-tagged molecules in living cells.

Spectral imaging

One of the requirements for efficient FRET is the significant overlap between the emission spectrum of the donor with the excitation spectrum of the acceptor. With conventional two-channel microscopy, separating two highly overlapping fluorophores requires many corrections and is always accompanied by loss of signal, due to use of only two channels instead of eight. Spectral imaging reduces the amount of corrective measures and allows collection of light from the whole spectrum of each fluorophore thus resulting in a better signal-to-noise ratio. Quantitative FRET measurements between CFP and YFP with the aid of spectral imaging have been performed previously (Gu et al., 2004; see also Neher & Neher, 2004). It is concluded that even though FRET studies with these fluorophores are possible on conventional microscopes, spectral unmixing has the advantage of eliminating spectral cross talk as well as providing relative concentrations of the fluorophores.

A recent development that eliminates spectral cross talk of GFP and YFP is the dark yellow fluorescent protein called REACh. REACh has retained absorption properties of YFP while having lost its fluorescence (Ganesan et al., 2006). The dynamic range of donor-based FRET detection methods such as donor quenching and FLIM is enhanced with REACh as FRET acceptor since photons from the entire GFP spectrum can be used. An advantage of using YFP over REACh is the direct detection of the acceptor-tagged proteins.

FRET between GFP and YFP has also been measured by FLIM. Positive FRET results in longer average YFP lifetimes, which can be detected by FLIM microscopy without the need of spectral separation (Harpur et al., 2001). However, this method is only quantitative in studies of intramolecular FRET (proteins tagged with GFP on one side and YFP on the other) because without spectral separation, the presence of unpaired fluorophores will obscure lifetime measurements.

Live-cell abFRET

Acceptor photobleaching implies the destruction of fluorescence from the acceptor fluorophore and is therefore less suitable for longer time-series imaging than other FRET detection methods. However, if a reversibly photobleachable acceptor fluorophore is used, such as the recently developed dronpa (Ando et al., 2004; Habuchi et al., 2005), this technique can possibly be adapted for continuous measurements (Jares-Erijman & Jovin, 2003). When such a fluorophore is used as an acceptor of FRET, it can be returned to its fluorescent state by photoactivation after each abFRET pulse in order to obtain a new FRET measurement by abFRET at the next time point. However, in a system as used here, repeated bleaching will lead to loss of the donor so for longer time-series measurements of FRET, a different donor would be required.


In conclusion, we present a method that takes full advantage of the optimal energy transfer characteristics of GFP and YFP and straightforward FRET detection by acceptor photobleaching.

FRET detection between other closely overlapping fluorophores, such as GFP2 and YFP (Zimmermann et al., 2002) or YFP and mOrange, will also benefit from this method.

Material and methods


Microscopy experiments were performed on a Zeiss LSM 510 META confocal microscope. All images were recorded with the same filter settings, objective, detector offset and laser intensities and similar detector gain to ensure proper unmixing. Fluorophores were excited by the 458-nm line of a 30-mW argon laser, a 458/514 dichroic mirror was present in the excitation path and imaging occurred through a 63 × 1.4 NA objective. For spectral imaging, fluorescence intensities were detected in eight channels corresponding to the wavelength range between 486 and 572 nm.

Cell culture

Chinese hamster cells (AA8 and 43-3B) were cultured under standard conditions in DMEM/Ham's F10 complemented with 10% FCS and antibiotics at 37°C and 5% CO2.


FuGENE6 (Roche Applied Science, Almere, The Netherlands) was used as a transfection reagent. Transfections were performed according to manufacturer's protocol. Total DNA concentrations for double transfections was 1 μg per 3 μL FuGENE6.

Data analysis

Data were analyzed with the LSM software of Zeiss (AIM version 3.2) and Microsoft Excel.


This work was supported by the Dutch Organization for Scientific Research (NWO): ZonMW 912-03-012 (CD), 917-46-371 (ABH), 917-46-364 (WV) and by ESF 855-01-072 (MvR).