- Email: [email protected]
Detection of the homo- and hetero-interaction of proteins using fluorescence resonance energy transfer spectrum method
ARTICLE IN PRESS
Optik
Optics
Optik 121 (2010) 57–62 www.elsevier.de/ijleo
Detection of the homo- and hetero-interaction of proteins using fluorescence resonance energy transfer spectrum method Chen Wanga, Feng Zhangb, Guiying Wanga, Ya Chenga, Zhizhan Xua, a
State Key Laboratory of High Field Laser Physics, Shanghai Institute of Optics and Fine Mechanics, CAS, Shanghai 201800, China b Department of Neurobiology, Institute of Neuroscience, Second Military Medical University, Shanghai 200433, China Received 12 December 2007; accepted 15 May 2008
Abstract Fluorescence spectrometry based on fluorescence resonance energy transfer (FRET) principle is a simple but effective tool for investigating protein–protein interactions. In this paper, we report a spectrometry to quantify FRET efficiency based on our home-designed spectral probe system and spectral data-processing procedure. In our method, the fluorescence spectrum from each specimen is recorded at two wavelengths 454 and 502 nm. Least-squares linear fitting algorithm is applied directly to decompose the spectra of donor and acceptor under these two wavelengths to obtain FRET efficiency, which takes both spectral intensity and spectral profile into account compared with traditional three-step analysis. This system and the data-processing procedure enabled us to detect the homo-interaction and hetero-interaction of proteins in living cell. r 2008 Published by Elsevier GmbH. Keywords: Fluorescence spectroscopy; Spectral unmixing; FRET; CFP; YFP
1. Introduction Since the last decade of the twentieth century, fluorescence resonance energy transfer (FRET) technique has found widespread applications in various fields owing to its advantage of providing direct and noninvasive detection of protein interaction in vivo [1,2]. Up to now, most FRET detection is achieved in fluorescence microscope through two- or multi-channel imaging [3,4]. In those FRET imaging systems, emission filters are specially designed to collect emissions only in the short wavelength range of the donor or in the long Corresponding author. Tel.: +86 21 69918201
E-mail addresses: [email protected] (C. Wang), [email protected] (Z. Xu). 0030-4026/$ - see front matter r 2008 Published by Elsevier GmbH. doi:10.1016/j.ijleo.2008.05.012
wavelength range of the acceptor in order to avoid cross-talk among the image channels. The problem is that for FRET to work, the donor emission and acceptor excitation spectra must overlap (high overlap is good), but for good signal-to-noise ratio imaging one must avoid collecting the ‘‘wrong’’ photons [5]. A potential alternative collection method is fluorescence spectrometry that would collect all the photons from both the donor and acceptor simultaneously without the requirement of any filter. Therefore, fluorescence spectrometry could produce straightforward and instrument-independent FRET result for recording full-spectrum information instead of only images’ intensity. Accompanied with the increasing application of spectral FRET, data-processing methods to extract FRET information from fluorescent spectrum
ARTICLE IN PRESS 58
C. Wang et al. / Optik 121 (2010) 57–62
have also drawn great attention. Overton and Kota have established three-step analysis procedure to quantify FRET efficiency [6,7], which determine emission induced by FRET by subtracting the cell autofluorescence, then emission of donor, and the last the acceptor emission due to direct excitation. Although the threestep analysis procedure is a robust method, it can be a tough exercise, requiring the precise control of cells number in each sample and the precise calculation of normalized factor. And the three-step analysis procedure is costly in terms of time, requiring performing all the subtractions step by step. Furthermore, the nature behind this approach is to quantify FRET only from spectral intensity without considering spectral profile. Recent publications have demonstrated the feasibility of spectrally unmixing procedures [8–10] in quantitative FRET studies. Gu and his colleagues succeeded to resolve the spectral bleed-through by combining the spectral unmixing technique with FRET by acceptor photobleaching, and retrieved the FRET efficiency by spectrally unmixing the donor emissions before and after photobleaching [11]. Raicu et al. [12] demonstrated collection of 16 images of fluorescence emissions from each sample over 5 nm-wide windows separated by 1 nm gaps to form emission fluorescence spectra using a wavelength-tunable confocal laser scanning microscope, and then separated the spectra generated by donor and acceptor through spectral unmixing. However, those unmixing methods applied for FRET microscopy, either combined with acceptor photobleaching or confocal microscopy, may all complicate the experimental operation or the data analysis. In this paper, we report a spectrometry based on a simple spectral linear unmixing algorithm and a selfdeveloped compact spectral probe system for studying the interaction of proteins inside living cells. In our approach, spectral unmixing is applied directly to spectral decomposition of fluorescence spectra obtained in our probe system. No spectrum construction or other supplementary technology is needed. The fluorescence spectrum from each specimen (co-transfected with donor and acceptor) is recorded at the two wavelengths, which are 454 and 502 nm (for the cyan fluorescent protein (CFP)–yellow fluorescent protein (YFP) FRET pair in our case). Least-squares linear fitting procedure gives out the best-fitting curves to experimental composite spectra based on the knowledge of member spectra at the two wavelengths, respectively, and the fraction of each component can be worked out. Through comparing the relative fraction factors of acceptor emission components under two wavelengths, the FRET efficiency is obtained. Our approach takes both spectral intensity and spectral profile into account compared with three-step analysis, and is more simple and straightforward to quantify FRET efficiency compared with other spectral unmixing microscopy. We applied
our system to detect FRET between proteins of interest fused with CFP and YFP in living Cos7 cell. FRET information from specimen is analyzed by spectral unmixing and comparing with three-step analysis procedure. Two approaches gave the similar results, suggesting homo- and hetero-interaction of target proteins in vivo, which implies that our system is sufficient for investigating the interaction of proteins in living cell conveniently.
2. Material and method 2.1. Spectral probe system The experimental setup is shown in Fig. 1. An Ar ion, nine lines, single-mode laser (MellesGriot, CA, USA) was used as the excitation light source. The cooling fan was mounted away from optical table to prevent fluctuation of laser power .The laser beam with a diameter of 2 mm was focused with an achromatic lens (f ¼ 100 mm) into a 10 10 40 mm3 quartz sample cell. A shutter is placed in front of the laser source to avoid excess illumination of sample while no signal collection is undertaken. The fluorescence signal that transmitted at the right angle was focused into the slit of a monochromator (Triax180, Jobin-Yvon, Horiba, France) through a focusing lens (f ¼ 50 mm), and the front entrance slit of monochromator was set as 20 mm to ensure a high spectral resolution as well as a good collection efficiency of the fluorescence photons. A 16bit CCD (Jobin-Yvon, Horiba, France) was mounted at the output of the monochromator to record the fluorescence spectra.
2.2. Cell culture and transfection Cos7 cell line were seeded in 12-well overnight with Dulbecco’s modified Eagle’s medium (DMEM), which contains 10% fetal bovine serum and glutamine. Cells
shutter L1
Sample cell
Power meter
L2 Monochromator
Computer
CCD
Fig. 1. Schematic drawing of FRET spectral probe system.
ARTICLE IN PRESS C. Wang et al. / Optik 121 (2010) 57–62
were transiently transfected with 4 mg of plasmid DNA for single transfection or 2 mg of each plasmid DNA for cotransfections. DNA plasmid coding for CFP, YFP, CFP–GASTVPRDPPVAT–YFP, target proteins fused with CFP and YFP were used in our experiment (gifts from Dr. Feng Zhang and Professor Cheng He in Second Military Medical University in China). Twenty four hours after transfection, the cells were washed with PBS twice, then treated with trysin/EDTA, next resuspended with PBS, and lastly added into quartz cell for detection.
composite spectra by minimizing the sum of the squares of the offset, we can estimate the fractions of YFP emission of each specimen at those two wavelengths. It is known that FRET results in a decreased fluorescence intensity of CFP and an increased fluorescence intensity of YFP. Here we used FRET ratio (FR) to quantify the enhanced YFP emission (i.e. FRET signal), which was calculated as a proportion between total YFP emission (denoted as TYFP) and direct excitation of YFP (denoted as DYFP) as follows: FR ¼ TYFP454 =DYFP454 ¼ A2 =A4 .
2.3. FRET measurement In our study, CFP and YFP were used for FRET pair. For CFP excitation, wavelength was adjusted to 454 nm. For YFP excitation, wavelength of 502 nm other than 514 nm was chosen to greatly reduce cross-talk of excitation photons into fluorescence signal since YFP emission ranges from 500 to 650 nm. Given the optical characteristic of CFP and YFP, grid (300 lines/mm) with blaze wavelength at 560 nm is selected and CCD scans were performed from 400 to 700 nm. The output laser powers of different wavelengths were both set at 18 mW. The spectra were recorded by a single shot at the 454 and 502 nm with the identical detection conditions for each sample, such as exciting power, placed position of samples, collection angle, and integration times (set to 0.5 s).
2.4. Spectrum unmixing analysis For quantitative FRET spectral measurements, spectrum unmixing analysis was performed. We first recorded spectra of CFP, YFP, target protein I-CFP, target protein I-YFP, target protein II-CFP, target protein II-YFP at two different wavelengths (454 and 502 nm) from cells expressing only a single fluorescence entity to establish endmember spectral libraries for the two wavelengths. Full-spectrum fluorescence from specimen was then collected under the two excitation wavelengths, respectively, which was considered as a linear combination of endmember spectra. That can be denoted as S 454 ¼ A1 CFP454 þ A2 YFP454 þ A3 Background454 , S 502 ¼ A4 YFP502 þ A5 Background502
(1)
S was detected spectrum from samples; CFP and YFP stood for spectra of donor and acceptor, background was also considered as one component spectrum here; A1–A5 were scaling factors. The subscripts indicated the wavelengths used for excitation. By linear spectral fitting to composite spectra on the basis of knowledge of endmember spectra in which leastsquares technology was used to find best-fitting curve to
59
(2)
FR is a unitless index equal to the fractional increase in YFP emission due to FRET, it would be unity when there is no FRET. As the amount of FRET increases, FR will raise above unity.
2.5. Three-step analysis procedure We compare our spectrum unmixing procedure with the three-step analysis procedure mentioned above in this paper. Briefly, we first irradiated cells at 454 nm and recorded emission spectra, next subtracted the components from the background, CFP emission, and YFP emission due to the direct excitation, resulting in the YFP emission spectrum only contributed by FRET as follows: FR ¼ ½FRET CFP454 =ðQ YFP502 Þ
(3)
to qualify FRET efficiency, where FRET, CFP454 and YFP502 are the background-subtracted spectrum, CFP emission spectrum excited by 454 nm and YFP emission spectrum excited by 502 nm of cells. Q is the ratio between relative fluorescence efficiencies of YFP taken at 454 and 502 nm excitations.
2.6. Statistics All results were repeated three times and data were presented as the mean7SE values.
3. Result 3.1. Spectral characteristics of CFP, YFP and their fused proteins To perform spectral unmixing, spectral library was firstly established. Cells were transfected with plasmids encoding CFP, YFP, target protein I-CFP, target protein I-YFP, target protein II-CFP, target protein II-YFP so that each sample would express only a single fluorescent entity. We then excited the various fluorescent entities with both 454 and 502 nm and collected
ARTICLE IN PRESS C. Wang et al. / Optik 121 (2010) 57–62
Intensity (AU)
12000
CFP target protein I-CFP target protein II-CFP
9000 6000
influence on CFP optical characteristic. However, the fluorescence peaks of both spectra remained around 476 and 502 nm, and the fluorescence bandwidth was also unchanged, exhibiting the same required spectral properties for FRET experiment. In the case of YFP and YFP tagged proteins, almost no difference was observed except a minor decrease in peak intensity. For three-step analysis, we measured quantum ratios (Q) of YFP (Fig. 2c), target protein I-YFP, target protein II-YFP under excitation wavelengths of 454 and 502 nm together with the above spectral information. The quantum ratios were calculated as the ratio of integrated area under YFP spectrum curve, and we obtained the results of 0.2770.07, 0.2270.05, 0.2270.09, respectively.
20000
positive control
10000 Intensity (AU)
spectral profiles. We also irradiated control cells in PBS that did not express CFP or YFP at these two wavelengths to record the background signal from cell autofluorescence together with other scattering light. Thus pure CFP, YFP, target protein I-CFP, target protein I-YFP, target protein II-CFP, target protein IIYFP emission spectrum will be obtained after background-subtraction (Fig. 2a and b). It is important to note that CFP and its fusion protein irradiated by 502 nm yields no signal after background removal, which demonstrated that CFP cannot be excited at the wavelength of 502 nm. Comparing CFP with CFP tagged proteins, their spectral profiles were somehow different, showing that fusion with a target protein had
Intensity (AU)
60
15000 10000 5000 0 400
3000
450 500 550 600 Wavelength (nm)
6000 4000 2000 0 400
650
15000
15000 10000 5000
5000
20000
500 550 600 650 Wavelength (nm)
700
YFP under 502nm YFP under 454nm
15000 10000 5000 0 400
500 600 Wavelength (nm)
700
Fig. 2. Spectra profiles of CFP, YFP and their fusion proteins in COS7 cells: (A) comparison of spectrum among CFP, target protein I-CFP, target protein II-CFP, (B) comparison of spectrum among YFP, target protein I-YFP, target protein IIYFP, and (C) quantitative comparison of YFP fluorescence intensity at the excitation wavelength 454 and 502 nm.
650
Total YFP emission Direct YFP emission
5000 0
400
9000
500 600 Wavelength (nm)
Negative CFP
6000 3000
400
500 600 Wavelength (nm)
500 600 Wavelength (nm)
700
0
450
450 500 550 600 Wavelength (nm)
10000
0
12000
10000
0 400
Intensity (AU)
YFP target protein I-YFP target protein II-YFP
Intensity (AU)
Intensity (AU)
20000
700
Intensity (AU)
500 600 Wavelength (nm)
Positive CFP
20000
700
1500 Intensity (AU)
400
Intensity (AU)
15000
0
negative control
8000
700
Total YFP emission direct YFP emission
1000 500 0 400
500 600 Wavelength (nm)
700
Fig. 3. FRET information from positive and negative control are analyzed by spectral unmixing and compared with threestep analysis procedure: (A) sketch map of the original fluorescence emission from positive control cells (Cos7 cells were transfected with YFP–CFP fusion plasmid) at 454 nm (red line) and 502 nm (green line). The fit results by spectral unmixing at the two wavelengths were represented as black dash lines. (B) Same analysis as in (A) for negative control cells (Cos7 cells were co-transfected with YFP and CFP). (C) Emission due to FRET was determined by three-step procedure for positive control. Black line means original fluorescence emission at 454 nm, and red line means the bleedthrough of CFP which was normalized to the CFP fluorescence peak at 476 nm (hereafter same). (D) Following analysis of (C), green line means total YFP emission which comprised two components, YFP emission due to direct emission and increased YFP emission due to FRET. The former was represented by blue line. (E) and (F) Same analysis as in (C) and (D) for negative control cells.
ARTICLE IN PRESS C. Wang et al. / Optik 121 (2010) 57–62
YFP fluorescence emission due to FRET. Meanwhile, data from cells co-transfected with CFP and YFP gave negative result FR ¼ 1.170.22. Negative control result further confirms that FRET of positive control did not come from over-expression of CFP and YFP together. In addition, we also perform data analysis by three-step procedure, which was shown in Fig. 3b and c. The emission spectra comprise of three components: the bleedthrough of CFP (red curve), the total emission of YFP include FRET (green blue) and the direct emission of YFP (blue curve). The real FRET signal is quantified as the area under green line subtracting that under blue line. The results are FR ¼ 4.370.55 and 0.970.13. All the results showed good arrangement of our FRET spectrum system.
3.2. Positive and negative control Positive control and negative control were used in our FRET studies to confirm that the system works well. For positive control, COS7 cells were transfected with CFP–YFP, which is linked by 15 amino acids GASTVPRARDPPVAT. For negative control, COS7 cells were co-transfect with CFP and YFP. We performed the FRET analysis by spectrum unmixing as shown in Fig. 3a. The red line represented raw spectrum data under the excitation wavelength of 454 nm, which included CFP fluorescence, YFP fluorescence and background noise obtained with this excitation wavelength. The green line represented raw spectral data under the excitation wavelength of 502 nm, which included only YFP fluorescence and background noise. Black dash lines represented fitting results for these two wavelengths using computer program. Data from COS7 cells transfected with CFP–YFP gave the result FR ¼ 4.370.55, demonstrating significant increase in
(A)
Intensity (AU)
2000
4000 3000 2000
0 450
(C)
500 550 600 Wavelength (nm)
650
400
target protein I-CFP + target protein II-YFP
6000 4000 2000 0 400
450
(D)
3000 Intensity (AU)
Intensity (AU)
target protein II-CFP + target protein II-YFP
(B)
1000
0 400
Intensity (AU)
5000
4000
6000
To visualize homo-interaction of proteins, CFP and YFP were fused to the C terminal of target proteins,
target protein I-CFP + target protein I-YFP
6000
8000
3.3. Homo- and hetero-interaction of target proteins
500 550 600 Wavelength (nm)
650
target protein I-CFP + target protein I-YFP target protein I-CFP FRET target protein I-YFP direct
2000
1000
0 450
(E)
500 550 600 Wavelength (nm)
650
target protein II-CFP + target protein II-YFP target protein II-YFP FRET target protein-YFP direct
5000 4000 3000 2000
400
8000 Intensity (AU)
Intensity (AU)
8000
61
500 600 Wavelength (nm)
(F)
target protein I-CFP + target protein II-YFP target protein I-CFP FRET target protein II-YFP
6000 4000 2000
1000 0 400
450
500 550 600 Wavelength (nm)
650
0 400
450
500 550 600 Wavelength (nm)
650
Fig. 4. Homo- and hetero-interaction of target proteins in living cells were studied by FRET. CFP/YFP fusion proteins were expressed separately and together in COS7 cells for FRET experiments. Cells expressing these fusion proteins were excited at 454 and 502 nm, and their emission spectra were recorded. Comparison of the fitting results by spectral unmixing with the raw data of (A) target protein I-CFP/YFP, (B) target protein II-CFP/YFP, and (C) target protein I-CFP/target protein II-YFP. The red and green line represented raw spectrum data of experiment at 454 and 502 nm, respectively, black dash line represented fitting results. Emission due to FRET was determined by three-step procedure as described under ‘‘Materials and methods’’. (D) target protein I-CFP/YFP, (E) target protein II-CFP/YFP, and (F) target protein I-CFP/target protein II-YFP.
ARTICLE IN PRESS 62
C. Wang et al. / Optik 121 (2010) 57–62
respectively. Fluorescence from cells co-expressing target protein I-CFP and target protein I-YFP or target protein II-CFP and target protein II-YFP were collected as described in ‘‘Materials and methods’’. Quantitative FRET analyses were carried out (Fig. 4). The FRs were 1.670.43 for target protein I-CFP/YFP and 2.270.36 for target protein II-CFP/YFP, indicating self-oligomerization of target proteins in the living cells. Furthermore, hetero-interaction between target proteins was also observed and the FR value was 2.070.50. Applying three-step procedure analysis to spectral data yields 1.4870.56, 2.6370.15, 1.870.30 for target protein I-CFP/YFP, target protein II-CFP/YFP and target protein I-CFP/target protein II-YFP, respectively.
4. Conclusion In this work, we have setup a compact spectral probe system to perform FRET detection. Unlike the collection mode used in intensity-based microscope, we detected fluorescence signal in the direction perpendicular to the incident laser, avoiding the use of optical filters in this arrangement. This change facilitates the application to various fluorophore pairs. We also developed a data analysis procedure based on linear spectral unmixing. The greatest advantages of our method are simpleness and easy to use. Once the spectral library set up, FRET efficiency can be determined quickly since it is only needed to obtain spectral information under two wavelengths for one specimen, which can be achieved by two shots. And the following analysis by spectral linear mixing is easy to perform using computer program and can be done within a few seconds. All of these advantages make it a suitable method to perform fast FRET detection.
Acknowledgments The work was supported by grants from National Basic Research Program of China (Grant no. 2006CB806000) and (Grant no. 2002CB713808). The authors thank
Professor Cheng He at the Second Military Medical University for his great contribution to this work.
References [1] C. Berney, G. Danuser, FRET or no FRET: a quantitative comparison, Biophys. J. 84 (2003) 3992–4010. [2] T. Haraguchi, Live cell imaging: approaches for studying protein dynamics in living cells, Cell Struct. Funct. 27 (2002) 33–334. [3] G.W. Gordon, G. Berry, Xiao Huan Liang, B. Levine, B. Herman, Quantitative fluorescence resonance energy transfer measurements using fluorescence microscopy, Biophys. J. 74 (1998) 2702–2713. [4] R.B. Sekar, A. Periasamy, Fluorescence resonance energy transfer (FRET) microscopy imaging of live cell protein localizations, J. Cell Biol. 160 (2003) 629–633. [5] George McNamara’s Multi-probe Microscopy document, /http://home.earthlink.net/mpmicroS. [6] M.C. Overton, K.J. Blumer, Use of fluorescence resonance energy transfer to analyze oligomerization of G-protein-coupled receptors expressed in yeast, Methods 27 (2002) 324–332. [7] P. Kota, P.J. Reeves, U.L. Rajbhandary, H.G. Khorana, Opsin is present as dimmer in COS1 cells: identification of amino acids at the dimmer interface, PNAS 103 (2005) 3054–3059. [8] R. Lansford, G. Bearman, S.E. Fraser, Resolution of multiple green fluorescent protein color variants and dyes using two-photon microscopy and imaging spectroscopy, J. Biomed. Opt. 6 (2001) 311–318. [9] T. Haraguchi, T. Shimi, T. Koujin, Spectral imaging fluorescence microcopy, Genes Cells 7 (2002) 881–887. [10] J.W. Boardman, Inversion of imaging spectrometry data using singular value decomposition, in: Proceedings of the IGARSS ’89, 1989, pp. 2069–2072. [11] Y. Gu, W.L. Di, D.P. Kelselld, D. Zicha, Quantitative fluorescence resonance energy transfer (FRET) measurement with acceptor photobleaching and spectral unmixing, J. Microsc. 215 (2004) 162–173. [12] V. Raicu, D.B. Jansma, R.J. Dwayne Miller, Protein interaction quantified in vivo by spectrally resolved fluorescence resonance energy transfer, Biochem. J. 385 (2005) 265–277.
Optik
Optics
Optik 121 (2010) 57–62 www.elsevier.de/ijleo
Detection of the homo- and hetero-interaction of proteins using fluorescence resonance energy transfer spectrum method Chen Wanga, Feng Zhangb, Guiying Wanga, Ya Chenga, Zhizhan Xua, a
State Key Laboratory of High Field Laser Physics, Shanghai Institute of Optics and Fine Mechanics, CAS, Shanghai 201800, China b Department of Neurobiology, Institute of Neuroscience, Second Military Medical University, Shanghai 200433, China Received 12 December 2007; accepted 15 May 2008
Abstract Fluorescence spectrometry based on fluorescence resonance energy transfer (FRET) principle is a simple but effective tool for investigating protein–protein interactions. In this paper, we report a spectrometry to quantify FRET efficiency based on our home-designed spectral probe system and spectral data-processing procedure. In our method, the fluorescence spectrum from each specimen is recorded at two wavelengths 454 and 502 nm. Least-squares linear fitting algorithm is applied directly to decompose the spectra of donor and acceptor under these two wavelengths to obtain FRET efficiency, which takes both spectral intensity and spectral profile into account compared with traditional three-step analysis. This system and the data-processing procedure enabled us to detect the homo-interaction and hetero-interaction of proteins in living cell. r 2008 Published by Elsevier GmbH. Keywords: Fluorescence spectroscopy; Spectral unmixing; FRET; CFP; YFP
1. Introduction Since the last decade of the twentieth century, fluorescence resonance energy transfer (FRET) technique has found widespread applications in various fields owing to its advantage of providing direct and noninvasive detection of protein interaction in vivo [1,2]. Up to now, most FRET detection is achieved in fluorescence microscope through two- or multi-channel imaging [3,4]. In those FRET imaging systems, emission filters are specially designed to collect emissions only in the short wavelength range of the donor or in the long Corresponding author. Tel.: +86 21 69918201
E-mail addresses: [email protected] (C. Wang), [email protected] (Z. Xu). 0030-4026/$ - see front matter r 2008 Published by Elsevier GmbH. doi:10.1016/j.ijleo.2008.05.012
wavelength range of the acceptor in order to avoid cross-talk among the image channels. The problem is that for FRET to work, the donor emission and acceptor excitation spectra must overlap (high overlap is good), but for good signal-to-noise ratio imaging one must avoid collecting the ‘‘wrong’’ photons [5]. A potential alternative collection method is fluorescence spectrometry that would collect all the photons from both the donor and acceptor simultaneously without the requirement of any filter. Therefore, fluorescence spectrometry could produce straightforward and instrument-independent FRET result for recording full-spectrum information instead of only images’ intensity. Accompanied with the increasing application of spectral FRET, data-processing methods to extract FRET information from fluorescent spectrum
ARTICLE IN PRESS 58
C. Wang et al. / Optik 121 (2010) 57–62
have also drawn great attention. Overton and Kota have established three-step analysis procedure to quantify FRET efficiency [6,7], which determine emission induced by FRET by subtracting the cell autofluorescence, then emission of donor, and the last the acceptor emission due to direct excitation. Although the threestep analysis procedure is a robust method, it can be a tough exercise, requiring the precise control of cells number in each sample and the precise calculation of normalized factor. And the three-step analysis procedure is costly in terms of time, requiring performing all the subtractions step by step. Furthermore, the nature behind this approach is to quantify FRET only from spectral intensity without considering spectral profile. Recent publications have demonstrated the feasibility of spectrally unmixing procedures [8–10] in quantitative FRET studies. Gu and his colleagues succeeded to resolve the spectral bleed-through by combining the spectral unmixing technique with FRET by acceptor photobleaching, and retrieved the FRET efficiency by spectrally unmixing the donor emissions before and after photobleaching [11]. Raicu et al. [12] demonstrated collection of 16 images of fluorescence emissions from each sample over 5 nm-wide windows separated by 1 nm gaps to form emission fluorescence spectra using a wavelength-tunable confocal laser scanning microscope, and then separated the spectra generated by donor and acceptor through spectral unmixing. However, those unmixing methods applied for FRET microscopy, either combined with acceptor photobleaching or confocal microscopy, may all complicate the experimental operation or the data analysis. In this paper, we report a spectrometry based on a simple spectral linear unmixing algorithm and a selfdeveloped compact spectral probe system for studying the interaction of proteins inside living cells. In our approach, spectral unmixing is applied directly to spectral decomposition of fluorescence spectra obtained in our probe system. No spectrum construction or other supplementary technology is needed. The fluorescence spectrum from each specimen (co-transfected with donor and acceptor) is recorded at the two wavelengths, which are 454 and 502 nm (for the cyan fluorescent protein (CFP)–yellow fluorescent protein (YFP) FRET pair in our case). Least-squares linear fitting procedure gives out the best-fitting curves to experimental composite spectra based on the knowledge of member spectra at the two wavelengths, respectively, and the fraction of each component can be worked out. Through comparing the relative fraction factors of acceptor emission components under two wavelengths, the FRET efficiency is obtained. Our approach takes both spectral intensity and spectral profile into account compared with three-step analysis, and is more simple and straightforward to quantify FRET efficiency compared with other spectral unmixing microscopy. We applied
our system to detect FRET between proteins of interest fused with CFP and YFP in living Cos7 cell. FRET information from specimen is analyzed by spectral unmixing and comparing with three-step analysis procedure. Two approaches gave the similar results, suggesting homo- and hetero-interaction of target proteins in vivo, which implies that our system is sufficient for investigating the interaction of proteins in living cell conveniently.
2. Material and method 2.1. Spectral probe system The experimental setup is shown in Fig. 1. An Ar ion, nine lines, single-mode laser (MellesGriot, CA, USA) was used as the excitation light source. The cooling fan was mounted away from optical table to prevent fluctuation of laser power .The laser beam with a diameter of 2 mm was focused with an achromatic lens (f ¼ 100 mm) into a 10 10 40 mm3 quartz sample cell. A shutter is placed in front of the laser source to avoid excess illumination of sample while no signal collection is undertaken. The fluorescence signal that transmitted at the right angle was focused into the slit of a monochromator (Triax180, Jobin-Yvon, Horiba, France) through a focusing lens (f ¼ 50 mm), and the front entrance slit of monochromator was set as 20 mm to ensure a high spectral resolution as well as a good collection efficiency of the fluorescence photons. A 16bit CCD (Jobin-Yvon, Horiba, France) was mounted at the output of the monochromator to record the fluorescence spectra.
2.2. Cell culture and transfection Cos7 cell line were seeded in 12-well overnight with Dulbecco’s modified Eagle’s medium (DMEM), which contains 10% fetal bovine serum and glutamine. Cells
shutter L1
Sample cell
Power meter
L2 Monochromator
Computer
CCD
Fig. 1. Schematic drawing of FRET spectral probe system.
ARTICLE IN PRESS C. Wang et al. / Optik 121 (2010) 57–62
were transiently transfected with 4 mg of plasmid DNA for single transfection or 2 mg of each plasmid DNA for cotransfections. DNA plasmid coding for CFP, YFP, CFP–GASTVPRDPPVAT–YFP, target proteins fused with CFP and YFP were used in our experiment (gifts from Dr. Feng Zhang and Professor Cheng He in Second Military Medical University in China). Twenty four hours after transfection, the cells were washed with PBS twice, then treated with trysin/EDTA, next resuspended with PBS, and lastly added into quartz cell for detection.
composite spectra by minimizing the sum of the squares of the offset, we can estimate the fractions of YFP emission of each specimen at those two wavelengths. It is known that FRET results in a decreased fluorescence intensity of CFP and an increased fluorescence intensity of YFP. Here we used FRET ratio (FR) to quantify the enhanced YFP emission (i.e. FRET signal), which was calculated as a proportion between total YFP emission (denoted as TYFP) and direct excitation of YFP (denoted as DYFP) as follows: FR ¼ TYFP454 =DYFP454 ¼ A2 =A4 .
2.3. FRET measurement In our study, CFP and YFP were used for FRET pair. For CFP excitation, wavelength was adjusted to 454 nm. For YFP excitation, wavelength of 502 nm other than 514 nm was chosen to greatly reduce cross-talk of excitation photons into fluorescence signal since YFP emission ranges from 500 to 650 nm. Given the optical characteristic of CFP and YFP, grid (300 lines/mm) with blaze wavelength at 560 nm is selected and CCD scans were performed from 400 to 700 nm. The output laser powers of different wavelengths were both set at 18 mW. The spectra were recorded by a single shot at the 454 and 502 nm with the identical detection conditions for each sample, such as exciting power, placed position of samples, collection angle, and integration times (set to 0.5 s).
2.4. Spectrum unmixing analysis For quantitative FRET spectral measurements, spectrum unmixing analysis was performed. We first recorded spectra of CFP, YFP, target protein I-CFP, target protein I-YFP, target protein II-CFP, target protein II-YFP at two different wavelengths (454 and 502 nm) from cells expressing only a single fluorescence entity to establish endmember spectral libraries for the two wavelengths. Full-spectrum fluorescence from specimen was then collected under the two excitation wavelengths, respectively, which was considered as a linear combination of endmember spectra. That can be denoted as S 454 ¼ A1 CFP454 þ A2 YFP454 þ A3 Background454 , S 502 ¼ A4 YFP502 þ A5 Background502
(1)
S was detected spectrum from samples; CFP and YFP stood for spectra of donor and acceptor, background was also considered as one component spectrum here; A1–A5 were scaling factors. The subscripts indicated the wavelengths used for excitation. By linear spectral fitting to composite spectra on the basis of knowledge of endmember spectra in which leastsquares technology was used to find best-fitting curve to
59
(2)
FR is a unitless index equal to the fractional increase in YFP emission due to FRET, it would be unity when there is no FRET. As the amount of FRET increases, FR will raise above unity.
2.5. Three-step analysis procedure We compare our spectrum unmixing procedure with the three-step analysis procedure mentioned above in this paper. Briefly, we first irradiated cells at 454 nm and recorded emission spectra, next subtracted the components from the background, CFP emission, and YFP emission due to the direct excitation, resulting in the YFP emission spectrum only contributed by FRET as follows: FR ¼ ½FRET CFP454 =ðQ YFP502 Þ
(3)
to qualify FRET efficiency, where FRET, CFP454 and YFP502 are the background-subtracted spectrum, CFP emission spectrum excited by 454 nm and YFP emission spectrum excited by 502 nm of cells. Q is the ratio between relative fluorescence efficiencies of YFP taken at 454 and 502 nm excitations.
2.6. Statistics All results were repeated three times and data were presented as the mean7SE values.
3. Result 3.1. Spectral characteristics of CFP, YFP and their fused proteins To perform spectral unmixing, spectral library was firstly established. Cells were transfected with plasmids encoding CFP, YFP, target protein I-CFP, target protein I-YFP, target protein II-CFP, target protein II-YFP so that each sample would express only a single fluorescent entity. We then excited the various fluorescent entities with both 454 and 502 nm and collected
ARTICLE IN PRESS C. Wang et al. / Optik 121 (2010) 57–62
Intensity (AU)
12000
CFP target protein I-CFP target protein II-CFP
9000 6000
influence on CFP optical characteristic. However, the fluorescence peaks of both spectra remained around 476 and 502 nm, and the fluorescence bandwidth was also unchanged, exhibiting the same required spectral properties for FRET experiment. In the case of YFP and YFP tagged proteins, almost no difference was observed except a minor decrease in peak intensity. For three-step analysis, we measured quantum ratios (Q) of YFP (Fig. 2c), target protein I-YFP, target protein II-YFP under excitation wavelengths of 454 and 502 nm together with the above spectral information. The quantum ratios were calculated as the ratio of integrated area under YFP spectrum curve, and we obtained the results of 0.2770.07, 0.2270.05, 0.2270.09, respectively.
20000
positive control
10000 Intensity (AU)
spectral profiles. We also irradiated control cells in PBS that did not express CFP or YFP at these two wavelengths to record the background signal from cell autofluorescence together with other scattering light. Thus pure CFP, YFP, target protein I-CFP, target protein I-YFP, target protein II-CFP, target protein IIYFP emission spectrum will be obtained after background-subtraction (Fig. 2a and b). It is important to note that CFP and its fusion protein irradiated by 502 nm yields no signal after background removal, which demonstrated that CFP cannot be excited at the wavelength of 502 nm. Comparing CFP with CFP tagged proteins, their spectral profiles were somehow different, showing that fusion with a target protein had
Intensity (AU)
60
15000 10000 5000 0 400
3000
450 500 550 600 Wavelength (nm)
6000 4000 2000 0 400
650
15000
15000 10000 5000
5000
20000
500 550 600 650 Wavelength (nm)
700
YFP under 502nm YFP under 454nm
15000 10000 5000 0 400
500 600 Wavelength (nm)
700
Fig. 2. Spectra profiles of CFP, YFP and their fusion proteins in COS7 cells: (A) comparison of spectrum among CFP, target protein I-CFP, target protein II-CFP, (B) comparison of spectrum among YFP, target protein I-YFP, target protein IIYFP, and (C) quantitative comparison of YFP fluorescence intensity at the excitation wavelength 454 and 502 nm.
650
Total YFP emission Direct YFP emission
5000 0
400
9000
500 600 Wavelength (nm)
Negative CFP
6000 3000
400
500 600 Wavelength (nm)
500 600 Wavelength (nm)
700
0
450
450 500 550 600 Wavelength (nm)
10000
0
12000
10000
0 400
Intensity (AU)
YFP target protein I-YFP target protein II-YFP
Intensity (AU)
Intensity (AU)
20000
700
Intensity (AU)
500 600 Wavelength (nm)
Positive CFP
20000
700
1500 Intensity (AU)
400
Intensity (AU)
15000
0
negative control
8000
700
Total YFP emission direct YFP emission
1000 500 0 400
500 600 Wavelength (nm)
700
Fig. 3. FRET information from positive and negative control are analyzed by spectral unmixing and compared with threestep analysis procedure: (A) sketch map of the original fluorescence emission from positive control cells (Cos7 cells were transfected with YFP–CFP fusion plasmid) at 454 nm (red line) and 502 nm (green line). The fit results by spectral unmixing at the two wavelengths were represented as black dash lines. (B) Same analysis as in (A) for negative control cells (Cos7 cells were co-transfected with YFP and CFP). (C) Emission due to FRET was determined by three-step procedure for positive control. Black line means original fluorescence emission at 454 nm, and red line means the bleedthrough of CFP which was normalized to the CFP fluorescence peak at 476 nm (hereafter same). (D) Following analysis of (C), green line means total YFP emission which comprised two components, YFP emission due to direct emission and increased YFP emission due to FRET. The former was represented by blue line. (E) and (F) Same analysis as in (C) and (D) for negative control cells.
ARTICLE IN PRESS C. Wang et al. / Optik 121 (2010) 57–62
YFP fluorescence emission due to FRET. Meanwhile, data from cells co-transfected with CFP and YFP gave negative result FR ¼ 1.170.22. Negative control result further confirms that FRET of positive control did not come from over-expression of CFP and YFP together. In addition, we also perform data analysis by three-step procedure, which was shown in Fig. 3b and c. The emission spectra comprise of three components: the bleedthrough of CFP (red curve), the total emission of YFP include FRET (green blue) and the direct emission of YFP (blue curve). The real FRET signal is quantified as the area under green line subtracting that under blue line. The results are FR ¼ 4.370.55 and 0.970.13. All the results showed good arrangement of our FRET spectrum system.
3.2. Positive and negative control Positive control and negative control were used in our FRET studies to confirm that the system works well. For positive control, COS7 cells were transfected with CFP–YFP, which is linked by 15 amino acids GASTVPRARDPPVAT. For negative control, COS7 cells were co-transfect with CFP and YFP. We performed the FRET analysis by spectrum unmixing as shown in Fig. 3a. The red line represented raw spectrum data under the excitation wavelength of 454 nm, which included CFP fluorescence, YFP fluorescence and background noise obtained with this excitation wavelength. The green line represented raw spectral data under the excitation wavelength of 502 nm, which included only YFP fluorescence and background noise. Black dash lines represented fitting results for these two wavelengths using computer program. Data from COS7 cells transfected with CFP–YFP gave the result FR ¼ 4.370.55, demonstrating significant increase in
(A)
Intensity (AU)
2000
4000 3000 2000
0 450
(C)
500 550 600 Wavelength (nm)
650
400
target protein I-CFP + target protein II-YFP
6000 4000 2000 0 400
450
(D)
3000 Intensity (AU)
Intensity (AU)
target protein II-CFP + target protein II-YFP
(B)
1000
0 400
Intensity (AU)
5000
4000
6000
To visualize homo-interaction of proteins, CFP and YFP were fused to the C terminal of target proteins,
target protein I-CFP + target protein I-YFP
6000
8000
3.3. Homo- and hetero-interaction of target proteins
500 550 600 Wavelength (nm)
650
target protein I-CFP + target protein I-YFP target protein I-CFP FRET target protein I-YFP direct
2000
1000
0 450
(E)
500 550 600 Wavelength (nm)
650
target protein II-CFP + target protein II-YFP target protein II-YFP FRET target protein-YFP direct
5000 4000 3000 2000
400
8000 Intensity (AU)
Intensity (AU)
8000
61
500 600 Wavelength (nm)
(F)
target protein I-CFP + target protein II-YFP target protein I-CFP FRET target protein II-YFP
6000 4000 2000
1000 0 400
450
500 550 600 Wavelength (nm)
650
0 400
450
500 550 600 Wavelength (nm)
650
Fig. 4. Homo- and hetero-interaction of target proteins in living cells were studied by FRET. CFP/YFP fusion proteins were expressed separately and together in COS7 cells for FRET experiments. Cells expressing these fusion proteins were excited at 454 and 502 nm, and their emission spectra were recorded. Comparison of the fitting results by spectral unmixing with the raw data of (A) target protein I-CFP/YFP, (B) target protein II-CFP/YFP, and (C) target protein I-CFP/target protein II-YFP. The red and green line represented raw spectrum data of experiment at 454 and 502 nm, respectively, black dash line represented fitting results. Emission due to FRET was determined by three-step procedure as described under ‘‘Materials and methods’’. (D) target protein I-CFP/YFP, (E) target protein II-CFP/YFP, and (F) target protein I-CFP/target protein II-YFP.
ARTICLE IN PRESS 62
C. Wang et al. / Optik 121 (2010) 57–62
respectively. Fluorescence from cells co-expressing target protein I-CFP and target protein I-YFP or target protein II-CFP and target protein II-YFP were collected as described in ‘‘Materials and methods’’. Quantitative FRET analyses were carried out (Fig. 4). The FRs were 1.670.43 for target protein I-CFP/YFP and 2.270.36 for target protein II-CFP/YFP, indicating self-oligomerization of target proteins in the living cells. Furthermore, hetero-interaction between target proteins was also observed and the FR value was 2.070.50. Applying three-step procedure analysis to spectral data yields 1.4870.56, 2.6370.15, 1.870.30 for target protein I-CFP/YFP, target protein II-CFP/YFP and target protein I-CFP/target protein II-YFP, respectively.
4. Conclusion In this work, we have setup a compact spectral probe system to perform FRET detection. Unlike the collection mode used in intensity-based microscope, we detected fluorescence signal in the direction perpendicular to the incident laser, avoiding the use of optical filters in this arrangement. This change facilitates the application to various fluorophore pairs. We also developed a data analysis procedure based on linear spectral unmixing. The greatest advantages of our method are simpleness and easy to use. Once the spectral library set up, FRET efficiency can be determined quickly since it is only needed to obtain spectral information under two wavelengths for one specimen, which can be achieved by two shots. And the following analysis by spectral linear mixing is easy to perform using computer program and can be done within a few seconds. All of these advantages make it a suitable method to perform fast FRET detection.
Acknowledgments The work was supported by grants from National Basic Research Program of China (Grant no. 2006CB806000) and (Grant no. 2002CB713808). The authors thank
Professor Cheng He at the Second Military Medical University for his great contribution to this work.
References [1] C. Berney, G. Danuser, FRET or no FRET: a quantitative comparison, Biophys. J. 84 (2003) 3992–4010. [2] T. Haraguchi, Live cell imaging: approaches for studying protein dynamics in living cells, Cell Struct. Funct. 27 (2002) 33–334. [3] G.W. Gordon, G. Berry, Xiao Huan Liang, B. Levine, B. Herman, Quantitative fluorescence resonance energy transfer measurements using fluorescence microscopy, Biophys. J. 74 (1998) 2702–2713. [4] R.B. Sekar, A. Periasamy, Fluorescence resonance energy transfer (FRET) microscopy imaging of live cell protein localizations, J. Cell Biol. 160 (2003) 629–633. [5] George McNamara’s Multi-probe Microscopy document, /http://home.earthlink.net/mpmicroS. [6] M.C. Overton, K.J. Blumer, Use of fluorescence resonance energy transfer to analyze oligomerization of G-protein-coupled receptors expressed in yeast, Methods 27 (2002) 324–332. [7] P. Kota, P.J. Reeves, U.L. Rajbhandary, H.G. Khorana, Opsin is present as dimmer in COS1 cells: identification of amino acids at the dimmer interface, PNAS 103 (2005) 3054–3059. [8] R. Lansford, G. Bearman, S.E. Fraser, Resolution of multiple green fluorescent protein color variants and dyes using two-photon microscopy and imaging spectroscopy, J. Biomed. Opt. 6 (2001) 311–318. [9] T. Haraguchi, T. Shimi, T. Koujin, Spectral imaging fluorescence microcopy, Genes Cells 7 (2002) 881–887. [10] J.W. Boardman, Inversion of imaging spectrometry data using singular value decomposition, in: Proceedings of the IGARSS ’89, 1989, pp. 2069–2072. [11] Y. Gu, W.L. Di, D.P. Kelselld, D. Zicha, Quantitative fluorescence resonance energy transfer (FRET) measurement with acceptor photobleaching and spectral unmixing, J. Microsc. 215 (2004) 162–173. [12] V. Raicu, D.B. Jansma, R.J. Dwayne Miller, Protein interaction quantified in vivo by spectrally resolved fluorescence resonance energy transfer, Biochem. J. 385 (2005) 265–277.