A time-resolved fluorescence anisotropy imaging method for studying nanoscale clustering of proteins or lipids was developed and evaluated. It is based on FRET between the identical fluorophores (homo-FRET), which results in a rapid depolarization of the fluorescence. The method employs the time-resolved fluorescence anisotropy decays recorded in a confocal microscope equipped with pulsed excitation and time-gated detection. From the decay the limiting anisotropy r inf was derived, which is a direct measure for the number of fluorophores per cluster. The method was evaluated by imaging GPI-GFP, a lipid raft marker. Small clusters were observed in the plasma membrane while the cytoplasm and the Golgi contained predominantly monomers.
© 2007 Optical Society of America
1.1 Hetero and homo-FRET
Advanced fluorescence microscopy based methods are indispensable tools in molecular cell biology1, 2. For example, molecular scale colocalisation of proteins or lipids can be imaged by means of Förster resonance energy transfer (FRET) [1–4]. Here, the excited-state energy of a fluorescent probe (donor) is transferred to another probe (acceptor), provided that (1) the donor-acceptor distance is less than ~10 nm, (2) there is spectral overlap between the donor’s emission and the acceptors absorption band and (3) the donor’s emission and the acceptors absorption transition dipole moments are not perpendicularly oriented. FRET results in quenching of the donor emission and sensitized emission of the acceptor. The FRET efficiency can be quantified using the donor/acceptor intensity ratio, or using spectral imaging. Furthermore, donor quenching results in a reduction of its fluorescence lifetime. Fluorescence lifetime imaging (FLIM) is a straightforward way to quantify the FRET efficiency in a reliable way [5, 6]. It is often regarded as the method of choice for FRET imaging.
The conditions for FRET to occur can be also met for a pair of identical fluorophores. Significant spectral overlap between the emission and absorption spectra of a fluorophore is often observed [2, 4, 7, 8]. As a result, energy transfer may take place between two or more identical fluorophores (homo-FRET). Hetero-FRET can also be used for studying homoclustering. However, it is much less ‘sensitive’ than homo-FRET. In the case of hetero-FRET a dimer can consist of D+A, A+D, D+D and A+A, only half of the combinations result in hetero-FRET. Furthermore, hetero-FRET requires approximately equal amounts of donor and acceptor labeled molecules and its interpretation can be complicated when larger clusters are being formed. As a result of the complications related to hetero-FRET based imaging of clusters, homo-FRET imaging methods have recently gained interest [4, 7–11].
Homo-FRET can not be imaged using standard hetero-FRET methodologies . Methods based on the difference in spectral properties between donor and acceptor (i.e., intensity ratio imaging or spectral imaging) are obviously not suitable. Moreover, homo-FRET does not change the donor’s lifetime; the reduction in the lifetime of the directly excited donor is fully compensated by the additional emission of the indirectly excited ‘donor’. Therefore, also FLIM can not be used. Since homo-FRET predominantly takes place between differently oriented fluorophores, it results in depolarization of the emission. This depolarization can be conveniently quantified using the fluorescence anisotropy (r) and homo-FRET can be imaged using fluorescence anisotropy imaging methods. Several approaches for imaging homo-FRET have been described in the literature [4, 7–11]. In these studies, steady-state anisotropy images and time-resolved decays were recorded separately to derive homo-FRET efficiencies and inter fluorophore distances [7, 8].
1.2 Cluster size imaging
An important feature of homo-FRET is that it can provide information about the number of fluorophores (N) per cluster. Clustering of proteins is often observed in cells and cluster size quantification is of great biological interest. For example, in cell signaling the binding of a ligand is often accompanied by receptor dimerization . Also the clustering of certain lipids in membrane domains or lipid rafts has been suggested to play an important role in signaling [7, 13]. To study cellular processes that involve clustering, cluster size imaging methods are essential. So far, however, methods to directly image molecular scale clustering (in vivo) are not available.
In the present work it is shown that the average cluster size per pixel can be imaged using time-resolved fluorescence anisotropy imaging. The limiting value of the anisotropy r inf in the time-resolved fluorescence anisotropy decay can be directly related to the cluster size N. Interestingly only a limited number of time points and modest time resolution are required to determine the limiting anisotropy. In this work a confocal time-gated fluorescence anisotropy imaging microscope is used . The setup utilizes two detection modules that were originally developed for time-gating based FLIM , one for each polarization direction. The detection modules (LiMo) collect the emission in four 2 nanosecond wide consecutive time gates. This affords the recording of course time-resolved anisotropy decay images.
Several other time-resolved fluorescence anisotropy imaging microscopes have been described in literature. These microscopes are all based on wide-field frequency-domain [9, 16] and wide-field time-domain [17, 18] imaging. In the current work a confocal time-resolved fluorescence anisotropy imaging system is used. The optical sectioning yields higher contrast and more details in the anisotropy and cluster size images. Furthermore, it allows the recording of 3D cluster size images.
The potential of imaging protein cluster sizes using confocal time-gated anisotropy imaging is demonstrated using cells expressing GFP linked to glycosylphosphatidylinositol (GPI-GFP). GPI-anchored proteins include different receptors for Thy-1, folate and the cellular prion protein. Clustering or crosslinking of these receptors trigger transmembrane signal transduction and internalization. GPI-anchored proteins are found in the detergent resistant membrane fraction and are frequently used as so-called lipid raft markers . Chemical crosslinking using short crosslinkers and FRET experiments suggest that the GPI-anchored proteins exist as small clusters in the plasma membrane . In the present work, it is shown that GPI-GFP expressing cells exhibit subcellular heterogeneities in GPI-GFP cluster sizes. Furthermore, controlled photobleaching yielded more details about the cluster size distribution of N-mers. GPI-GFP appears to be present in small clusters in the plasma membrane while cytoplasmic GPI-GFP appears monomeric.
2. Theoretical background
2.1 Fluorescence anisotropy and homo-FRET
with I par and I per the emission parallel and perpendicular respectively to the excitation polarization direction. Since the orientations of two identical dye molecules will in general be different, energy transfer between the fluorophores will decrease the anisotropy of the emission . The anisotropy can, however, also decrease due to other effects than homo-FRET; the most common source of depolarization is rotation of the fluorophore. Directly after exciting the fluorescent molecules, the maximum initial anisotropy r0 is 0.4. The exact value of r0 depends on the relative orientation of the absorption and emission transition dipole moments of the fluorophore; a value of r0 =0.4 is found when the two dipole moments are parallel. Depending on the size of the molecule and the viscosity of the surrounding medium, rotation of molecules can occur on a sub-nanosecond to nanosecond timescale. This time scale may very well be similar to the time scale at which homo-FRET takes place. To discriminate between the rotation and homo-FRET, time-resolved fluorescence anisotropy decay curves can be recorded. For a free (spherical) rotor in solution in the absence of homo-FRET, this decay is described by:
with Φ the rotation correlation time of the rotor. In the presence of homo-FRET both rotational depolarization and homo-FRET depolarization take place and the anisotropy decay has a multi-exponential character. Reliable analysis of the decay may be complicated. To measure homo-FRET it is advantageous to employ slowly rotating dye molecules such as GFPs (rotational correlation time Φ ≫ τ). Now, rotations can be ignored and the analysis of the anisotropy decay is comparatively simple. The time-resolved anisotropy decay can be written as :
Here, ω denotes the homo energy transfer rate and r inf the limiting anisotropy at time infinity. ω and the homo-FRET efficiency E are defined as:
where R0 is the Förster distance, R the inter fluorophore distance and τ the fluorescence lifetime.
2.2 Fluorescence anisotropy and cluster size
The number of fluorophores per cluster is directly related to the anisotropy . For a monomer, in the absence of rotation the average anisotropy amounts to r mono = r 0 = 0.4. For a dimer, however, the initial anisotropy of the excited donor equals r = 0.4 whereas the anisotropy of the donor that is indirectly excited after homo-FRET amounts to r et = 0.016 [see Fig. 1(top)] [21, 22]. When energy transfer is very fast (E≈1) and reversible, both donors have equal probabilities of emitting the photon and the average anisotropy of the dimer amounts to r dimer = 0.2. For clusters of size N the observed average anisotropy equals:
When the energy transfer is less efficient, i.e. E < 1, most emission will come from the directly excited donor and there will be less depolarization by homo-FRET. The dependency of the steady-state anisotropy on the transfer rate in oligomers of size N is described by :
Figure 1(bottom) shows graphs of the relation between the cluster size N and the steady-state anisotropy r for various values of the transfer efficiency. Dimers (N = 2) have a steady-state anisotropy r = 0.2 when E≈1, whereas this same anisotropy value corresponds to trimers (N = 3) when E= 0.5.
Different cluster sizes can be discriminated by employing Eq. (6). This requires knowledge of the homo-FRET efficiency. Unfortunately, experimental determination of ω is difficult in time-resolved imaging experiments because high signal levels are required to obtain sufficient photon statistics [9, 23]. It should be realized, however, that in time-resolved anisotropy decays [as described by Eq. (3)] only the increase in homo-FRET transfer rate ω is responsible for increased values of the steady-state anisotropy for e.g. dimers (see Fig. 2). When r(t) = r inf, the probability of emitting a photon is equal for all fluorescent molecules in the cluster. This condition is already met after ~2 ns for homo-FRET efficiencies higher than 0.5 [see Fig. 2(b)]. The value of the limiting anisotropy r inf in the time-resolved anisotropy decay therefore equals the steady state anisotropy in the case of E = 1, and:
Images of r inf can be directly converted to images of the protein cluster size, without any prior knowledge of the homo-FRET transfer rate.
3.1 NIH 3T3 fibroblasts expressing GPI-GFP
NIH 3T3 cells were cultured in Dulbecco’s modified Eagle’s medium (Invitrogen) supplemented with 7.5% foetal bovine serum (v/v) and 2 mM L-glutamine at 37°C in the presence of 5% CO2 under a humidified atmosphere. For the generation of a cell-line stably expressing GPI-GFP, NIH 3T3 cells were transfected with plasmid DNA encoding this construct (kindly provided by dr. P. Keller, Dresden) using Lipofectamin 2000 (Invitrogen; Breda, The Netherlands). Stable expression was achieved by prolonged culturing in medium containing selective amounts of Zeocin (Invitrogen). For imaging, cells were seeded on 18 mm coverslips and grown for 2 days until 30% confluence. Cells were fixed with freshly prepared 4% formaldehyde at 37°C for 20 minutes, and quenched by 100 mM glycine. Coverslips were mounted with Mowiol and stored at -20°C until further use. These preparations are not expected to influence the observed cluster sizes24.
3.2 Time-gated fluorescence anisotropy imaging set-up
A modified confocal scanning laser microscope (CSLM, Nikon PCM 2000) was employed. The excitation polarization direction was defined by a linear polarizer (Meadowlark, Frederick, CO, USA) and a broadband polarizing beam splitter cube (PBS, OptoSigma, Santa Ana, CA, USA) was used to split the emission in a parallel and a perpendicular channel with respect to the excitation light. The excitation light was provided by a frequency doubled picosecond pulsed Ti:Sa laser (Tsunami, Spectra Physics) tuned to 460 nm; a pulse picker was used to reduce the repetition rate to 8.2 MHz. The excitation pulses were transferred to the confocal microscope using a single-mode optical fiber. For imaging, a NA=1.20 / 40x water immersion objective (Plan Apo, Nikon) was used. Such high NA objectives yield lower value of r 0 due to the effect of the high acceptance angle on the anisotropy measurement [7–9, 11, 25]. Here, a value of r 0 = 0.33 was found instead of 0.40. Each of the two emission channels (“parallel” and “perpendicular”) was equipped with its own detection system consisting of a high quantum efficiency fiber-coupled PMT (Hamamatsu H7422P-40), a pre-amplifier and a fluorescence lifetime imaging module (LiMo, Nikon Instruments BV, Badhoevedorp, The Netherlands) . The LiMo captures four images using four consecutive time gates of approximately 2 ns width. By employing two such modules, one for each polarization direction, a four channel time-resolved anisotropy decay can be acquired for each pixel.
Determination of the time-resolved anisotropy using the two separate detection channels requires careful synchronization in time and correction for their difference in sensitivity. The former was achieved by using a aqueous solution of Rose Bengal (τ F = 70 ps) as a reference. The time offset of the first gate with respect to the excitation pulse was adjusted in such a way that the first gate opened after 90% of the Rose Bengal fluorescence was emitted. The correction factor for the difference in transmission (c tr) between both channels was determined by recording an anisotropy image of an aqueous solution of fluorescein. In this case, fast rotation of the fluorophore will result in emission that is completely depolarized before the second gate opens. Consequently, the anisotropy in the last three gates will be zero, and I par = c tr.I per. Finally, a correction was applied to correct for small difference in gate width between the parallel and perpendicular channel. These differences were determined using a sample containing 10 μM GFP monomers in 50/50 glycerol/buffer. In this solution, rotation of the fluorophore is much slower than the fluorescence lifetime (Φ ≫ τ). Consequently, the anisotropy will remain constant at the r 0 level. For every gate, a correction factor was determined that ensures that the anisotropy in this gate is identical to the steady state anisotropy. The absence of concentration dependent homo-FRET in the GFP solution was checked by lowering the concentration; no increase in anisotropy was observed.
To record high time resolution fluorescence anisotropy decays, the two LiMo’s were replaced by two time-correlated single photon counting boards (TimeHarp, Picoquant, Germany). Again, an aqueous solution of fluorescein was used to calculate the difference in transmission of the two polarization channels, and a solution of Rose Bengal was used for temporal synchronization of the channels.
3.3 Data handling
Cluster size images were obtained by extracting r inf-values from the time-gated fluorescence anisotropy images. The anisotropy values in the last gates are, in good approximation, equal to r inf, provided that E is high enough. Based on Fig. 2(a) this condition is met for the three last gates provided that E > 0.5. Four-gate anisotropy decays were created by binning the intensities I par and I per per gate in regions of interest. In the anisotropy imaging experiments a threshold of I par,inf + 2. I per,inf > 300 counts was applied to all images. In theory, this number of counts corresponds to a standard deviation in the anisotropy of 0.05 .
3.4 Simulation of photo-bleaching
The effect of photo-bleaching on the anisotropy of N-mers was simulated using MS Excel. First order chemical kinetics, equal photo-bleaching probabilities for every GFP and inter fluorophore distances smaller than the Förster distance R 0 were assumed7. Three model distributions of N-mers were generated. For each initial distribution of N-mers, the distribution after photo-bleaching a certain fraction of molecules was calculated. This can be conceived as follows. Bleaching of a population of N-mers will result in a distribution of N-mers, (N-1)-mers, (N-2)-mers, etc. For example, bleaching a fraction x of the GFP trimers will result in the following distribution: (1-x)3 trimers, 3.x.(1-x)2 dimers, 3.x 2.(1-x) monomers and x 3 fully bleached GFP. The final distributions of N-mers from each initial N-mer were summed per final N-mer. Next, the fraction of the total signal for each final N-mer was calculated. This was repeated for a range of bleaching percentages, thus creating a simulated bleaching curve.
The three models of N-mer distributions were adjusted to match the initial experimental cluster size. Model 1 assumes only monomer and dimers; the fraction of the latter was tuned. Model 2 is based on a Gaussian distribution of N-mers that is centered at N = 0; tuning was achieved by varying the width of this distribution. Finally, model 3 involves monomers and a varying fraction of large clusters. Here, a Gaussian distribution with a mean cluster size of 5 was assumed. The amplitude of the distribution was adjusted to match the initial cluster size. These models are depicted in Fig. 7(c).
4. Results and discussion
4.1 Evaluation of the set-up
The rotational mobility of fluorescein in solutions of varying glycerol content was used to evaluate the time-resolved fluorescence anisotropy imaging setup. Anisotropy decays were recorded using both time-gated imaging and time-correlated single photon counting (TCSPC). The resulting decays are shown in Fig. 3. The black lines correspond to the anisotropy decays recorded using TCSPC (in non-imaging mode) and the bars correspond to the average time-gating based anisotropy decays over the 160*160 pixels of the image. The anisotropy decay slows down in a more viscous environment. Importantly, the decays recorded using the time-gating match the ones recorded using TCSPC. For fluorescein in 75/25 glycerol/water, fitting with a simple mono-exponential decay yielded Φ = 6.4 ± 0.15 ns (TCSPC) and Φ = 6.6 ± 0.6 ns (time-gating).
4.2 Cluster size imaging in cells
To evaluate the potential of the approach in cellular imaging, clustering of GPI-anchored GFP was analysed. In Fig. 4, an image of a NIH 3T3 fibroblast stably expressing GPI-GFP is shown. In the intensity image of this cell, clear labeling of the plasma membrane can be observed and plasma membrane ruffles are visible [Fig. 4(a)]. Also in the cytoplasm some structures can be distinguished; the relatively high intensity region is the Golgi apparatus. The anisotropy image reveals clear heterogeneities in anisotropy [Fig. 4(b)]. Particularly large differences in anisotropy are observed between the Golgi [blue bordered region of interest (RoI)] and the ruffles (e.g. green bordered RoI). In the absence of homo-FRET the anisotropy in a monomeric solution of GFP (in 50/50 glycerol/buffer) was found to be 0.33. This serves as a reference value for the anisotropy level of monomers, i.e. in absence of homo-FRET. The lower anisotropy in the plasma membrane and ruffles can be attributed to homo-FRET in GPI-GFP clusters.
The average time-gated anisotropy decays of ruffles and Golgi are shown in green and blue respectively [Fig. 4(d)]. The decay of the monomeric reference solution of GFP (in 50/50 glycerol/buffer) is shown in yellow. The constant level of the anisotropy of the GFP monomers gives the initial anisotropy in the decays of GPI-GFP in the cell, i.e., r mono = r 0 = 0.33. In these decays, the occurrence of homo-FRET can be readily observed. The anisotropy decays much faster than can be expected from GFP rotation alone. This indicates homo-FRET takes place in clusters of GPI-GFP. The anisotropy in the last gates is to a good approximation equal to r inf (a direct measure of the cluster size, see Eq. (7). Here, r inf was estimated from the average anisotropy in the last three gates. A slow decay can still be observed in these gates, probably due to inefficient homo-FRET or minor rotation of GFP. However, the determination of the cluster size is hardly affected by such a slow decay and the averaging strongly improves the signal-to-noise ratio of the rinf determination. The dashed lines in Fig. 4(d) indicate the r inf levels for various cluster sizes. This shows that the effect of the slow decay on the cluster size determination is much less than 0.1. In the ruffle RoI r inf = 0.207, whereas in the Golgi RoI r inf = 0.280. The corresponding cluster sizes are 1.6 and 1.2, respectively. The values of r inf in Fig. 4(b) are also calculated from the intensities in the last three gates. The corresponding GPI-GFP cluster size image was calculated using these r inf values [Fig. 4(c)].
To exclude the possibility of concentration induced homo-FRET, scatter plots of intensity versus anisotropy were made  [Fig. 4(e)]. The blue dots in this figure correspond to the Golgi RoI, while the green dots correspond to the ruffle RoI in the plasma membrane. In case of concentration induced homo-FRET a correlation between intensity and anisotropy is expected. Figure 4(e) shows that in ruffles there is no significant correlation between anisotropy and intensity. This indicates that in ruffles small nanoscale clusters are present and that intensity is only a measure of the number of clusters, not their size. In the Golgi RoI below an intensity of 400 counts a concentration-dependence might be present. It should, however, be noted that in the Golgi GPI-GFP is still “under construction” (the attachment of the GPI anchor to the GFP takes place in the Golgi), a situation not representative for the properties of GPI-GFP in membranes.
To explore cell dependent variations in GPI-GFP clustering intensity and cluster size images were recorded of four different cells (Fig. 5). Clear local differences in both the intensity and cluster sizes can be observed. In general, the cluster size in the Golgi is low and in the membrane regions significant clustering can be observed. Strong ruffling of the plasma membrane seems to be accompanied by high GPI-GFP clustering [Fig. 5(b)]. In the absence of ruffling, some cells exhibit less clustering in the plasma membrane [Figs. 5(c) and 5(d)]. To further explore the local heterogeneities in N, a 3D-stack of images of a GPI-GFP expressing cell was recorded (Fig. 6). It is clear that in the plasma membrane the cluster size is larger than in the interior of the cell.
An important advantage of the confocal approach to image cluster sizes is the improved contrast and suppression of out-of-focus signals. This is particularly important for cluster size imaging in plasma membrane systems. Discriminating out-of-focus contributions from the signal improves the accuracy of the cluster size determination.
4.3 Effects of photo-bleaching on the cluster size
Besides quantification of the average cluster size per pixel, also information on the distribution of N-mers can be obtained from a controlled photo-bleaching series of cluster size images. The effect of photo-bleaching on homo-FRET was initially exploited by Sharma et al 7. Their modeled data was compared to the experimentally observed decrease in anisotropy after photo-bleaching. The analysis, however, involved an assumption of the values of r N (N = 2, 3, 4). In the current methodology these values are readily available. Moreover, it is not necessary to incorporate the anisotropy in the modeling; the effect of photo-bleaching on the cluster size can be directly compared to a bleaching series of cluster size images.
To obtain more detailed information on cluster size distributions photo-bleaching curves of a NIH 3T3 cell expressing GPI-GFP were recorded. The decrease in cluster size after photo-bleaching is plotted [red dots in Fig. 7(b)] for two RoI in the cell [see Fig. 7(a)]. These results are compared to three models of N-mer distributions, (1) monomers and a variable fraction of dimers, (2) Gaussian distributions of variable width centered at N = 0, and (3) monomers and a Gaussian distribution of variable width centered at N av = 5 [see Fig. 7(c)]. For these three models, the theoretical photo-bleaching curves are plotted as well [dark blue, light blue and green lines for model 1, 2 and 3, respectively in Fig. 7(b)]. The experimental data is reasonably well described by model 2 for both RoI. This suggests that both in the ‘normal’ plasma membrane and in ruffles GPI-GFP is present in small clusters. Clusters of size larger than two must be present to account for the measurements. However, the measurements can not be explained by a mix of monomers and large clusters. When assuming that model 2 is correct, the fraction of GFP in monomers, dimers and oligomers (N≥3) amounts to 0.35, 0.37 and 0.28 respectively in ruffles and 0.40, 0.38 and 0.21 respectively in the membrane.
The results presented here demonstrate the potential of confocal time-gated fluorescence anisotropy imaging for the study of homo-FRET in cells. The time-resolved anisotropy decays reveal the occurrence of homo-FRET. Importantly, the r inf derived from the anisotropy images can be readily transferred to cluster size images without information about the energy transfer efficiency. The method is expected to be more accurate than approaches based on steady state anisotropy. The high sensitivity of the method afforded for the first time to record cluster size maps of single cells. Moreover, the implementation in a confocal microscope afforded the recording of 3-D cluster size images. The cluster size images showed that the interior of cells can have significantly different cluster sizes than the plasma membrane. Finally, it was demonstrated that photo-bleaching can be exploited to obtain additional information about the distribution of N-mers in clusters.
This work is financially supported by the NWO/ALW MtC program.
References and links
1. A. Periasamy, Methods in Cellular Imaging, 1st ed. (Oxford University Press, Oxford, 2001).
2. B. Valeur, Molecular Fluorescence. Principles and Applications (Wiley-VCH, Weinheim, 2002).
3. V. E. Centonze, M. Sun, A. Masuda, H. Gerritsen, and B. Herman, “Fluorescence resonance energy transfer imaging microscopy,” Method. Enzymol. 360, pp. 542–560 (2003). [CrossRef]
5. A. Esposito, H. C. Gerritsen, and F. S. Wouters, “Fluorescence lifetime heterogeneity resolution in the frequency domain by lifetime moments analysis,” Biophys. J. 89, 4286–4299 (2005). [CrossRef] [PubMed]
6. J. Sytsma, J. M. Vroom, C. J. De Grauw, and H. C. Gerritsen, “Time-gated fluorescence lifetime imaging and microvolume spectroscopy using two-photon excitation,” J. Microsc. 191, 39–51 (1998). [CrossRef]
7. P. Sharma, R. Varma, R. C. Sarasij, Ira, K. Gousset, G. Krishnamoorthy, M. Rao, and S. Mayor, “Nanoscale organization of multiple GPI-anchored proteins in living cell membranes,” Cell 116, 577–589 (2004). [CrossRef] [PubMed]
8. I. Gautier, M. Tramier, C. Durieux, J. Coppey, R. B. Pansu, J. C. Nicolas, K. Kemnitz, and M. Coppey-Moisan, “Homo-FRET microscopy in living cells to measure monomer-dimer transition of GFP-tagged proteins,” Biophys. J. 80, 3000–3008 (2001). [CrossRef] [PubMed]
9. D. S. Lidke, P. Nagy, B. G. Barisas, R. Heintzmann, J. N. Post, K. A. Lidke, A. H. A. Clayton, D. J. Arndt-Jovin, and T. M. Jovin, “Imaging molecular interactions in cells by dynamic and static fluorescence anisotropy (rFLIM and emFRET),” Biochem. Soc. T. 31, 1020–1027 (2003). [CrossRef]
11. A. Squire, P. J. Verveer, O. Rocks, and P. I. H. Bastiaens, “Red-edge anisotropy microscopy enables dynamic imaging of homo-FRET between green fluorescent proteins in cells,” J. Struct. Biol. 147, 62–69 (2004). [CrossRef] [PubMed]
12. G. Krauss, Biochemistry of Signal Transduction and Regulation, 2nd ed. (Wiley-VCH, Weinheim, 2001). [CrossRef]
14. A. N. Bader, E. G. Hofman, P. van Bergen en Henegouwen, and H. C. Gerritsen, “Confocal time-resolved fluorescence anisotropy imaging,” Proc. SPIE 6441, art. no. 64410C (2007). [CrossRef]
15. C. J. de Grauw and H. C. Gerritsen, “Multiple time-gate module for fluorescence lifetime imaging,” Appl. Spectrosc. 55, 670–678 (2001). [CrossRef]
16. A. H. A. Clayton, Q. S. Hanley, D. J. Arndt-Jovin, V. Subramaniam, and T. M. Jovin, “Dynamic fluorescence anisotropy imaging microscopy in the frequency domain (rFLIM),” Biophys. J. 83, 1631–1649 (2002). [CrossRef] [PubMed]
17. K. Suhling, J. Siegel, P. M. P. Lanigan, S. Leveque-Fort, S. E. D. Webb, D. Phillips, D. M. Davis, and P. M. W. French, “Time-resolved fluorescence anisotropy imaging applied to live cells,” Opt. Lett. 29, 584–586 (2004). [CrossRef] [PubMed]
18. J. Siegel, K. Suhling, S. Leveque-Fort, S. E. D. Webb, D. M. Davis, D. Phillips, Y. Sabharwal, and P. M. W. French, “Wide-field time-resolved fluorescence anisotropy imaging (TR-FAIM): Imaging the rotational mobility of a fluorophore,” Rev. Sci. Instrum. 74, 182–192 (2003). [CrossRef]
19. D. A. Brown and E. London, “Functions of lipid rafts in biological membranes,” Annu. Rev. Cell Dev. Biol. 14, 111–136 (1998). [CrossRef]
20. F. Tanaka and N. Mataga, “Theory of time-dependent photo-selection in interacting fixed systems,” Photochem. and Photobiol. 29, 1091–1097 (1979). [CrossRef]
22. V. M. Arganovich and M. D. Galanin, Electronic excitation energy transfer in condensed matter (North-Holland Publishing, Amsterdam, 1982).
23. K. A. Lidke, B. Rieger, D. S. Lidke, and T. M. Jovin, “The Role of Photon Statistics in Fluorescence Anisotropy Imaging,” IEEE T. Image Process. 14, 1237 (2005). [CrossRef]
24. J. D. Bancroft and A. Stevens, Theory and practice of histological techniques (Churchill Livingstone, 1982).
25. D. Axelrod, “Fluorescence Polarization Microscopy,” Method. Cell Biol. 30, 333–352 (1989). [CrossRef]