In the previous section, the problem was posed that the differing spatial resolution of donor and acceptor channel leads to image artifacts when using ratiometric methods to calculate the FRET efficiencyEFRET:
EFRET= IFRETA ID+IFRETA
Instead of comparing the sensitized acceptor emission IFRETA to the quenched donor emissionID, one could compare it to the directly excited acceptor emission IdirA :
RFRET=IFRETA IdirA
Fortunately, the spatial resolutions of acceptor and FRET channels can be rendered al-most equal with optimized STED parameters, and a donor with a short lifetime (section 11.6). The fluorescence intensity in the acceptor channel is of course proportional to the number of acceptor molecules and if the excitation intensity is constant, the signal can be used for normalization: more sensitized acceptor emission would correspond to a higher intensity ratio.
The intensity ratios EFRET and RFRET of fluorescence emitted by the previously in-troduced single molecule FRET constructs were calculated for confocal and STED measurements. The resulting ratios for measurements versus STED intensity and pulse delay are plotted in Fig. 20.2. Unfortunately, the s.d. is rather large, probably due to the low photon counts. Hence, the same intensity ratios were calculated from simulated data and are shown in Fig. 20.3.
112 Chapter 20 Finding a measure for the FRET efficiency compatible with STED
The absolute values ofEFRETdetermined from experimental data and simulations do not fully agree with each other and with the theoretically expected values stated in the legend of Fig. 20.3. However, the proportions are reproduced as expected, i.e. DNA 79 has the highest FRET efficiency, etc. The same is observed forRFRET: the absolute values derived from simulation and experiment differ, in this case because the direct excitation of the acceptor was stronger in the experiment. Yet, the curves match reasonably well within the error margins. Most importantly, it is evident that both measures depend both on the STED intensity and delay, i.e. the same FRET construct appears to have a different FRET efficiency under different illumination conditions. Thus, both intensity ratios are not a reliable measure for the FRET intensity in STED images.
The reason that RFRET also correlates with the STED parameters is that the directly excited acceptor intensityIdirA simply decays either by stimulated or spontanous emission, whereas the amount of sensitized acceptor emissionIFRETA during and after the STED pulse depends on the STED intensity and delay. This can be seen in the lifetime histograms in chapter 18, when comparing the curves of DNA 83 (no FRET) with the curves of the other FRET constructs. The detected intensity corresponds to the area under the curve.
20.2.1 Combining intensity ratio and photon arrival time
It was discussed in previous sections thatEFRETis not suitable for calculating the FRET efficiency in STED images, because the donor channel, being depleted much less, has a significantly lower resolution than the FRET and the acceptor channel. Instead,RFRET was proposed as a measure, because FRET and acceptor channel can have almost the same resolution if the experimental parameters are adjusted. However, the intensity emitted in both channels depends on the STED settings each in a different way. Hence RFRET also correlates with them, which can be seen in Fig. 20.2 and Fig. 20.3. Most importantly, the curves ofRFRETcross each other, meaning that under certain conditions FRET pairs with different transfer rates could not be distinguished.
To overcome this uncertainty, it was proposed to combineRFRETwith lifetime informa-tion to better identify different FRET pairs. The first moment of the acceptor lifetime histograms, the mean photon arrival timeEtA, was used. In case of high FRET, more photons arrive earlier, whereas in low FRET, photons arrive later, as can be estimated from the histograms shown in Fig. 28.3.
In order to test if the different FRET constructs could be reliably distinguished by a combination ofRFRETandEtA, both values were calculated for each measured FRET pair.
Then, scatter plots were constructed withRFRETas x-coordinate andEtA as y-coordinate of each measured data point, which are shown in Fig. 20.4 and Fig. 28.5. The ellipses
20.2 Normalizing to the acceptor emission 113
Fig. 20.2: Fluorescence intensity ratiosEFRETandRFRETwere calculated from images of single molecule FRET constructs detailed in table 15.1, which were acquired with different STED powers and pulse delays. Error bars correspond to the s.d. between individual pairs.
Fig. 20.3: Fluorescence intensity ratiosEFRETandRFRETwere calculated from simulations of single molecule FRET pairs with the same efficiencies as detailed in table 15.1.
114 Chapter 20 Finding a measure for the FRET efficiency compatible with STED
have a diameter of two s.d., meaning they enclose 95% of the data points belonging to each FRET construct.
In the first graph in Fig. 20.4, one can see that in confocal measurements, the FRET constructs can be discriminated rather well, the s.d. ellipses overlap only a little.
Variations of the STED settings change the shape and location of each distribution in the plot. Consequently, the overlap between distributions varies, e.g. at 25mW STED, DNA 80 and DNA 81 almost completely overlap, whereas at 50mW STED, they are distinct, but DNA 81 and DNA 82 completely overlap. In conclusion, the combination of intensity ratioRFRETand mean photon arrival timeEtAalso does not allow a reliable discrimination of the different FRET pairs in STED images. Analogous scatter plots for different STED pulse delays are shown in supplementary Fig. 28.5.
One could try to use not only the first moment (mean), but also the second moment (variance) and maybe the third moment (skewness) to describe the histogram shape and thus distinguish the FRET pairs by their lifetime histograms in addition to their spectral information.