Article Information

• PDF (309 kb)
• Full Text with Large Figures
• Supporting Material
• Cited by in Scopus (6)

PubMed

• Articles by Michelle A. Digman
• Articles by Enrico Gratton

Related Articles

• Cerulean, Venus, and VenusY67C FRET Reference Standards

  Cerulean, Venus, and VenusY67C FRET Reference Standards Biophysical Journal, Volume 91, Issue 12, 15 December 2006, Pages L99-L101 Srinagesh V. Koushik, Huanmian Chen, Christopher Thaler, Henry L. Puhl and Steven S. Vogel Abstract Förster's resonance energy transfer (FRET) can be used to study protein-protein interactions in living cells. Numerous methods to measure FRET have been devised and implemented; however, the accuracy of these methods is unknown, which makes interpretation of FRET efficiency values difficult if not impossible. This problem exists due to the lack of standards with known FRET efficiencies that can be used to validate FRET measurements. The advent of spectral variants of green fluorescent protein and easy access to cell transfection technology suggests a simple solution to this problem: the development of genetic constructs with known FRET efficiencies that can be replicated with high fidelity and freely distributed. In this study, fluorescent protein constructs with progressively larger separation distances between donors and acceptors were generated and FRET efficiencies were measured using fluorescence lifetime spectroscopy, sensitized acceptor emission, and spectral imaging. Since the results from each method were in good agreement, the FRET efficiency value of each construct could be determined with high accuracy and precision, thereby justifying their use as standards. Abstract | Full Text | PDF (162 kb)

• Investigating Interactions Mediated by the Presynaptic Prote...

  Investigating Interactions Mediated by the Presynaptic Protein Bassoon in Living Cells by Foerster's Resonance Energy Transfer and Fluorescence Lifetime Imaging Microscopy Biophysical Journal, Volume 94, Issue 4, 15 February 2008, Pages 1483-1496 Mini Jose, Deepak K. Nair, Wilko D. Altrock, Thomas Dresbach, Eckart D. Gundelfinger and Werner Zuschratter Abstract Neuronal synapses are highly specialized structures for communication between nerve cells. Knowledge about their molecular organization and dynamics is still incomplete. The large multidomain protein Bassoon plays a major role in scaffolding and organizing the cytomatrix at the active zone of neurotransmitter release in presynaptic boutons. Utilizing immunofluorescence techniques, we show that Bassoon is essential for corecruitment of its synaptic interaction partners, C-terminal binding protein 1/brefeldin A-dependent ADP-ribosylation substrate and CAZ-associated structural protein, into protein complexes upon heterologous expression in COS-7 cells. A combination of Foerster's resonance energy transfer and fluorescence lifetime imaging microscopy in the time domain was adopted to investigate the potential for the association of these proteins in the same complexes. A direct physical association between Bassoon and CtBP1 could also be observed at synapses of living hippocampal neurons. Simultaneous analysis of fluorescence decays of the donor and the acceptor probes along with their decay-associated spectra allowed a clear discrimination of energy transfer. Abstract | Full Text | PDF (1759 kb)

• Quantitative Multiphoton Spectral Imaging and Its Use for Me...

  Quantitative Multiphoton Spectral Imaging and Its Use for Measuring Resonance Energy Transfer Biophysical Journal, Volume 89, Issue 4, 1 October 2005, Pages 2736-2749 Christopher Thaler, Srinagesh V. Koushik, Paul S. Blank and Steven S. Vogel Abstract Protein labeling with green fluorescent protein derivatives has become an invaluable tool in cell biology. Protein quantification, however, is difficult when cells express constructs with overlapping fluorescent emissions. Under these conditions, signal separation using emission filters is inherently inefficient. Spectral imaging solves this problem by recording emission spectra directly. Unfortunately, linear unmixing, the algorithm used for quantifying individual fluorophores from emission spectra, fails when resonance energy transfer (RET) is present. We therefore sought to develop an unmixing algorithm that incorporates RET. An equation for spectral emission incorporating RET was derived and an assay based on this formalism, spectral RET (sRET), was developed. Standards with defined RET efficiencies and with known Cerulean/Venus ratios were constructed and used to test sRET. We demonstrate that sRET analysis is a comprehensive, photon-efficient method for imaging RET efficiencies and accurately determines donor and acceptor concentrations in living cells. Abstract | Full Text | PDF (536 kb)

• …more

Copyright © 2008 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 94, Issue 2, L14-L16, 15 January 2008

doi:10.1529/biophysj.107.120154

Biophysical Letters

The Phasor Approach to Fluorescence Lifetime Imaging Analysis

Michelle A. Digman^*, Valeria R. Caiolfa^†, ‡, Moreno Zamai^†, ‡ and Enrico Gratton^*, ,  

^* Laboratory for Fluorescence Dynamics, Department of Biomedical Engineering, University of California, Irvine, California
^† Department of Molecular Biology and Functional Genomics, San Raffaele Scientific Institute, Milan, Italy
^‡ IIT Network Research, Unit of Molecular Neuroscience, San Raffaele Scientific Institute, Milan, Italy

Address reprint requests and inquiries to Enrico Gratton.

In fluorescence experiments multiple lifetime components arise from different molecular species or different conformations of the same molecule 1,2. Changes of lifetime values are interpreted in terms of molecular interactions 3,4. The measurement of the fluorescence decay in fluorescence lifetime imaging microscopy (FLIM) can be performed using several methods: the time-correlated single photon counting 5, the frequency-domain, and the time-sampling approach 6,7,8. One difference between FLIM and cuvette measurements is that in FLIM, only a few photons (∼1000) are collected per pixel 9. This is barely enough to distinguish a single from a double exponential decay, which is required to determine if two species are present in the same pixel.

The analysis of FLIM data collected in the time domain proceeds by fitting the decay at each pixel using one or two exponentials and identifying decay times and amplitudes with molecular species and their relative abundances. A problem with this approach is that many of the fluorescent proteins used in microscopy display a complex decay behavior 10. Another problem with exponentials is that there is correlation between amplitude and characteristic exponential times. The analysis of the decay at each pixel (∼10⁵ pixels in an image) in using exponentials is a formidable computational problem that requires expertise to correctly extract the information about the number and abundance of the molecular species 11,12.

The phasor approach presented here has the potential of simplifying the analysis of FLIM images, avoiding some of the problems of the exponential analysis and providing a graphical global view of the processes affecting the fluorescence decay occurring at each pixel 8,13,14. To emphasize the novelty of the approach and to demonstrate that the phasor concept is not unique to the frequency domain, in this letter we use data collected with the time-correlated single photon counting method. We also show that quantitative evaluation of fluorescence resonance energy transfer (FRET) efficiencies in the phasor plot does not require fitting exponentials.

The phasor method transforms the histogram of the time delays at each pixel in a phasor, which is like a vector (Supplementary Eq. 1, Supplementary Material ). The values of the sine-cosine transforms are represented in a polar plot as a two-dimensional histogram (phasor plot). Each pixel of the image gives a point in the phasor plot. The phasor plot is also used in a reciprocal mode in which each (occupied) point of the phasor plot can be mapped to a pixel of the image. Since every molecular species has a specific phasor, we can identify molecules by their position in the phasor plot.

In Fig. 1, the precision of the determination of the phasors of cyan fluorescent protein (CFP), yellow fluorescent protein (YFP), and enhanced green fluorescent protein (EGFP) is sufficient to identify their origin. The resolution of phasors depends on the counts in each pixel (see figure caption). The phasors corresponding to fibronectin background and to cellular autofluorescence are spread over a larger area due to the low intensity (fibronectin) and to pixel heterogeneity (cellular autofluorescence). The phasor of the Raichu-Rac1 construct is also spread along a “trajectory” due to different amounts of FRET.

Display large version of this figure

Figure 1

(A) Phasor plot of the FLIM images shown above. Transfected cells are: (B) CHO-K1 with YFP (306); (C) MEF with paxillin-CFP (442); (D) CHOK1 with EGFP (1537); and (E) CHO-K1 with the RAICHU-Rac1 construct (665). (F) Untransfected cells to measure phasor of cell autofluorescence and fibronectin (573). The numbers in parenthesis indicate counts of the brightest pixel. The red dashed line corresponds to one possible FRET trajectory passing through the clusters of phasors corresponding to cells in the field of view.

The rule of phasor addition (which is the same as the vector addition with normalization; Supplementary Eq. 4, Supplementary Material ) helps in identifying the origin of points in a phasor cluster. If two molecular species are coexisting in the same pixel, all possible weighting of the two species give phasors distributed along a straight line joining the phasors of the two species. In the case of three molecular species, the possible realizations of the system fill a triangle where the vertices correspond to the phasors of the pure species. Just observing the clustering of points in the phasor plot is sufficient to determine that some pixels of the image contain two (or more) molecular species. The relative fluorescence of the species can be obtained in a quantitative way using the “phasor calculator” that graphically implements Supplementary Eq. 4 .

Fig. 2 shows the image of a CHO-K1 cell expressing paxillin-EGFP in a three-dimensional collagen matrix and of a region of the collagen matrix. The collagen fibers display very weak fluorescence with a very short lifetime (different from zero). The pixels that contain a combination of EGFP 1 and collagen emission 2 distribute along a line joining the phasors of the EGFP and collagen. The segment joining the phasor of EGFP and collagen is plotted. As the operator moves the cursor along the segment, the calculator (that solves Supplementary Eq. 4 ) displays the relative fractional contribution of the two components.

Display large version of this figure

Figure 2

CHO-K1 cells transfected with paxillin EGFP in a three-dimensional collagen matrix. (A) Phasor plot of the two images. Region of the sample with a transfected cell (B) and with the collagen matrix (E). (C and F) Pixels in the image selected by the cursor at position 1 (EGFP) are highlighted in pink. (D and G) Pixels selected at position 2 (collagen). Position 3 represents weak background fluorescence. Pixels with multiple contributions lie along the line joining the EGFP and the collagen phasor (green line) or the background autofluorescence and the collagen phasor (black line). The cursor is used interactively to select a given cluster of phasors along the lines. The computer displays the relative contribution of the two species for each position of the cursor along the line.

For interacting species (e.g., FRET pair) in which the presence of the acceptor reduces the lifetime of the donor, the resulting phasor of the interacting species cannot lie in the line joining the phasor of the two noninteracting species, but will be in a different part of the phasor plot corresponding to the quenched species. For FRET the information about the molecular interaction is obtained by localizing the position of clusters of phasor in the phasor plot rather than by resolving the decay of the donor in exponential components. The quantitative information about the interaction (the FRET efficiency) can be obtained using the “FRET calculator” that implements graphically Supplementary Eqs. 4 and 5 .

Fig. 3 shows images of cells transiently transfected with Cerulean (c) and Cerulean-Venus (c-v) constructs. The c-v complex display FRET due to the proximity of the two proteins. The FRET calculator is used to measure the FRET efficiency corresponding to the specific point along the FRET trajectory. The calculator also combines the phasor of the unquenched donor (c) and the phasor of the background (AF), which were determined independently using the c-only cell and a nontransfected cell. The amount of fluorescence of the donor and background to combine changes by the operator until the FRET trajectory passes through the experimental points in the phasor plot. The reciprocal property of the phasor cursor is illustrated in Figure 3CE, which shows the correspondence of phasors that cluster in regions of the c-only phasors and of the c-v phasors (the pink highlighted regions).

Display large version of this figure

Figure 3

MEF cells expressing cerulean and cerulean-venus constructs. (A) Phasor plot showing the clustering of phasors of cell B (c only) and (D) c-v construct. The curved trajectory corresponds to FRET efficiency (0–100%,) and for a percentage of background and donor unquenched. The green line joins the phasor of the cerulean and the autofluorescence phasor, independently determined. The FRET efficiency at the cursor position is 0.208. The counts in the brightest pixels were 457 and 192 for the c-v and c-only cell, respectively. The fraction of background was 9.5% and the fraction of unquenched donors was 0.9%. Image C shows pixels selected at the position of the c-only cell and image E when the position of the cursor is as shown in A.

Figure 1 and Figure 2 and Figure 3 show that specific clustering of points in the phasor plot and along trajectories is sufficient to establish the physical origin of the changes in lifetime at different pixels. This is done without resolving the decay in exponential components, but by inspection of the phasor plot. The identification in the image where these processes occur is done using the reciprocal property of the phasor plot by which every point of the phasor plot can be identified with a pixel in the image.

Although no fits are performed, the exploration of the phasor plot using the cursor and the calculator provides quantitative results for several common situations such as combination of multiple decay components in a pixel (linear trajectories) and FRET (specific curved trajectories). The calculator approach in which the computer draws in the phasor plot the trajectories of selected processes allows full quantitation of the parameters of specific process.

In this letter we intentionally omitted the calculation of the “lifetime” of the phasor. We have shown that its knowledge is unnecessary to identify a specific molecular species (Fig. 1), to determine the fractional contributions of molecular species at one pixel (Fig. 2), or to calculate FRET efficiencies (Fig. 3). If the phasor falls on the universal circle (see Supplementary Material ) it can be unequivocally associated with a lifetime value. If the phasor is not on the universal circle, the corresponding molecular species must have a complex decay. In this case, estimators of lifetime values can be obtained using phasor plots of higher Fourier harmonics. This subject has been discussed in the context frequency-domain fluorescence lifetime determinations and it is beyond the scope of this article. Here we emphasize that the determination of lifetime components is unnecessary when the decay is represented in the phasor plot.