Imaging of fluorescence anisotropy during photoswitching provides a simple readout for protein self-association

Monitoring of protein oligomerization has benefited greatly from Förster Resonance Energy Transfer (FRET) measurements. Although donors and acceptors are typically fluorescent molecules with different spectra, homo-FRET can occur between fluorescent molecules of the same type if the emission spectrum overlaps with the absorption spectrum. Here, we describe homo-FRET measurements by monitoring anisotropy changes in photoswitchable fluorescent proteins while photoswitching to the off state. These offer the capability to estimate anisotropy in the same specimen during homo-FRET as well as non-FRET conditions. We demonstrate photoswitching anisotropy FRET (psAFRET) with a number of test chimeras and example oligomeric complexes inside living cells. We also present an equation derived from FRET and anisotropy equations which converts anisotropy changes into a factor we call delta r FRET (drFRET). This is analogous to an energy transfer efficiency and allows experiments performed on a given homo-FRET pair to be more easily compared across different optical configurations.


Supplementary Note 1
Here, we consider homo---FRET between fluorescent proteins, which are known to have a high intrinsic anisotropy due to their relatively slow rotational correlation times. Energy transfer between fluorescent proteins leads to polarization scrambling and decreased anisotropy. We approached this problem by describing homo---FRET as a sensitized emission experiment with the parallel and perpendicular emission channels being analogous to the donor and FRET emission channels, respectively. We include all steps in our approach to this problem which will allow readers to follow our reasoning and check our work. To begin, consider the coordinates associated with an anisotropy experiment.
In a typical spectrometer based experiment, excitation light polarized in the z direction is transmitted along the x axis and excites a sample. The emission is detected in one or both directions along the y axis. Emission is collected through polarization filters which transmit light polarized parallel with the z axis (blue arrow in the figure) or polarized parallel with the x axis (perpendicular to z, red arrow in the figure). The intensity difference between Iz and Ix polarized light indicates the polarization of the sample. Polarization can provide a readout of the fluorophore's rotational motion which may report on its microenvironment or molecular binding events. Polarization (P) is defined by = % − ' % + ' (1) where Iz and Ix represent light polarized parallel with the z axis (parallel with the excitation light polarization) or polarized parallel with the x axis (perpendicular to the excitation light polarization).
In keeping with our sensitized emission analogy, Iz represents the donor channel and Ix represents the FRET channel. Similar to sensitized emission, Ix will contain the energy transfer signal (ET) and crosstalk from non---FRETing donor molecules as well as direct excitation and emission for this orientation. ' = ET + crosstalk (2) x z Light source Detector y polarizer For polarization measurements on a population of molecules in the absence of homo---FRET, any given P and Iz values will produce an expected Ix. Here, we consider this signal to be the non---FRETing crosstalk component. Therefore, measurements on these control samples (designated et0) will provide the necessary control (Ix et0/Iz et0) ratio to determine the non---FRETing crosstalk ratio. In a population of molecules undergoing homo---FRET, the ratio Ix et1/Iz et1 (et1 is used to designate an experimental sample undergoing energy transfer) will differ due to increased perpendicular polarized light and decreased parallel polarized light. However, for molecules in this population not undergoing homo---FRET, the ratio between the measured Iz et1 value and a corresponding crosstalk component (Ix ct) will be the same.
Here, the subscript ct is used to designate the crosstalk signal. Therefore, for the non---FRETing population, ' 78 Before proceeding with the derivation based on spectrometer configurations, we instead consider both excitation and detection of light along the x---axis which is the configuration for an epifluorescence microscope. Here, we would collect z---polarized and y--polarized light and similar equations could be derived if we simply substitute Iy for Ix. = % − F % + F (7) ET F 7 = F 98: − % 98: F 98; % 98; (8) Thus, to account for the decrease in Iz fluorescence due to homo---FRET and find the total sensitized emission signal (ETc), we must add the values of ETx c and ETy c.
ET 7 = ET ' 7 + ET F 7 (9) However, in most polarized imaging or spectroscopy experiments, we do not collect data along both x and y axes and only have one perpendicular channel. This simplifies by noting that if we assume the population of molecules is randomly oriented, then x---polarized light can be considered equal to that of y---polarized light. Therefore, Ix = Iy, and the total isotropic fluorescence ( (15) To complete our analogy with a sensitized emission experiment, we need to combine the anisotropy terms with a FRET efficiency equation from 1 , where ETc represents the corrected energy transfer signal, Id represents the donor fluorescence in the donor channel, and G is a factor relating the sensitized emission signal to the loss in donor fluorescence. G must be determined for each donor---acceptor pairing on each microscope to provide an accurate FRET measurement. This factor accounts for light transmission differences in the emission pathways as well as differences in quantum efficiencies for the donor and acceptor.
= 7 ⁄ Q + 7 ⁄ (16) Continuing with our sensitized emission analogy, we must also relate the signal observed in the FRET channel (^) to the signal in the donor channel ( || ). The || and ^ emission pathways for an instrument may be equivalent with the exception of the polarizer orientation, but optical components can introduce a polarization bias which makes anisotropy measurements inaccurate. To compensate, a factor, called g, describes any polarization bias in the instrument and is generally determined for any anisotropy experiment. It is most easily determined using small, freely diffusing and rotating fluorescent molecules which should display no polarization bias and the determined g is used to scale the intensity ^ to that of || . Thus, the fully corrected anisotropy equations also have the g factor, which we insert into Supplementary Equations 14 and 15. = || − ^ || + 2^ (17) ET 7 = 2^ 98: − 2 ||98: ^ 98; ||98; (18) Since the g factor determined for anisotropy measurements adequately compensates for instrument related signal differences in I^ and I || and the donor and acceptor are the same type of molecule with no differences in extinction coefficient or quantum yield, the G factor normally associated with FRET experiments in Supplementary Equation 14  (20) Here, we must point out where our analogy with a hetero---FRET sensitized emission experiment markedly diverges and we introduce a term to replace E for the energy transfer efficiency we are reporting. Treating anisotropy measurements in this manner cannot detect all homo---FRET energy transfer between molecules since the molecules with parallel dipole orientations will undergo efficient energy transfer but will not produce a detectable difference anisotropy. Moreover, any homo---FRET signal in the parallel channel is treated as non---FRET signal. Thus, the energy transfer indicated in Supplementary Equation 20 is actually a conversion of the perpendicular channel signal into a value representing the percent signal change. Although we consider delta r FRET (drFRET) to be akin to hetero---FRET efficiencies, it is denoted differently here since we report only on the increased perpendicular channel signal and we wish to avoid confusion when comparing FRET efficiencies for the same protein---protein interactions using homo---FRET versus hetero---FRET approaches. Therefore, substitution into Supplementary Equation 20 provides the following. (40) The final Supplementary Equation 40 comes with at least one caveat. The anisotropy for the control should represent the anisotropy of the molecule of interest in the absence of FRET. While this seems obvious, it may not be practical to determine this value for some molecules. For instance, a fluorescently labeled protein may have different anisotropies as a monomer versus a dimer even in the absence of homoFRET since the rotational dynamics of the dimer might be different compared to the monomer. In such an example, the dimer in the absence of FRET would be expected to have a higher anisotropy. The presence of homo---FRET would reduce the anisotropy to ret1, but using the ret0 value determined from a monomer would underestimate the energy transfer. Such a small difference may have little consequence if one wishes only to compare FRET efficiencies measured across laboratories under numerous conditions, but we feel obligated to point out this potential pitfall.
On the other hand, photobleaching or photoswitching off of the fluorescent tags can provide a readout of the anisotropy of the molecule of interest in the absence of reduced energy transfer. When photoswitched on, anisotropy measurements of photoswitchable fluorescent proteins provide ret1 at the initial fluorescence level. As the molecules photoswitch off over time, they will no longer be able to FRET with neighboring molecules. Thus, the anisotropy will increase to conditions closely resembling a true ret0, where the measured anisotropy is dictated by rotational motion of the protein complex and is independent of FRET. or V5 (open circles) were imaged and photobleached while collecting parallel and perpendicular polarized fluorescence emission. The anisotropy was determined and displayed as a function of the fluorophore photobleached. The data points were fitted to a linear equation to determine the change in anisotropy. The steady---state anisotropy values for the Venus oligomers were determined from single images of the oligomers expressed in COS---7 cells combined with the data points collected at the onset of the photobleaching experiment (inset; white columns). Data represent mean±sem (n=37, 36, 24, 32, and 29 for V1, V2, V3, V4, and V5, respectively). ANOVA tests indicated significant differences in conventional steady---state anisotropy for all oligomers (p---value <0.05; Cohen's d values ranged from 0.77 -6.87). (e) Linear fits of the data in c were used to determine the a" b" d" c" e" f" anisotropy of Cerulean and Cerulean---Cerulean before photobleaching (white columns), after photobleaching (black columns) and the difference (hatched columns). Data represent mean±sem (n=15 and 11 for Cerulean and Cerulean---Cerulean, respectively). ANOVA tests indicated significant differences between the pre---and post---photobleaching Cerulean---Cerulean anisotropy values (p---value <0.05; Cohen's d = 1.49). Two---tailed t---tests indicated a significant difference in the delta r between Cerulean and Cerulean---Cerulean (p---value <0.05; Cohen's d = 2.27) (f) Linear fits were used to determine the change in anisotropy (delta r) of the Venus oligomers during the photobleaching experiment (hatched columns). Data represent mean±sem (n=6, 6, 5, 6, and 6 for V1, V2, V3, V4, and V5, respectively).
ANOVA tests indicate significant differences in delta r for all oligomers ( Based on the studies of Runnels and Scarlata 5 , the expected changes in anisotropy as a function of cluster size are heavily dependent on the ratio of the separation distance to R0. Thus, it is noteworthy that the upward curvatures of the photoswitching or photobleaching profiles depend on the separation distance between the molecules being ≤ 0.8 R0 6 . (a) The oligomers were simulated using a binomial distribution described in Yeow and Clayton 6 . Dronpa has an R0 5.3nm and an assumption for this approach is that the average distance between the fluorophores is 0.8*R0. (b) The backbone beta---barrel structure of Dronpa molecule is shown to be approximately 4 nm in length and 2 -2.5 nm in diameter 7 . However, inclusion of the side chains and space filling models suggest that likely the closest distance between two fluorophores will be 3nm. (c) For this example, two Dronpa molecules are linked with a 5---amino acid linker and moved as far apart as possible. The chromophore to chromophore distance could reach 6 nm. As noted previously 8,9 , assuming an average k 2 =2/3 for fluorescent proteins is not valid but would represent the most liberal estimate and result in the highest possible R0. Assuming this best case scenario and using the overlap integrals between the absorption and emission spectra for Dronpa and Venus, we calculated the Dronpa---Dronpa and Venus---Venus R0 values to be 5.3nm. Therefore, the chromophores for our oligomer chimeras would need to be within 4.25 nm a" b" e" c" 6"nm" 6"nm" 11"nm" 3"nm"~2 "-"2.5"nm" 4"nm" f" d" "0.8"R 0! to meet ≤0.8 R0 criteria. (d) Simulations were performed in which 100,000 Dronpa molecules are randomly assembled into dimers with a limitation of 3nm minimum distance and 6nm maximum distance between the chromophores. The arrow indicates the approximate 0.8*R0 for Dronpa homo---FRET. (e) For this example, five Dronpa molecules are linked with 5---amino acid linkers and are shown in one potential configuration. This suggests that the distances between each of the chromophores can be variable for a Dronpa5 chimera. (f) Simulations were performed in which 100,000 Dronpa molecules are randomly assembled into pentamers with a limitation of 3nm minimum distance and 6nm maximum distance between each chromophore in the linear sequence. These show an even broader distribution with the separation distances up to 20 nm for each molecule in the chain. Dronpa images were produced using the Swiss PDB Viewer using the Dronpa structure pdb file 2IE2 7 .