PixFRET, an ImageJ plug-in for FRET calculation ... - ResearchGate

More documents

Recommendations

Info

56 J.N. FEIGE ET AL.Fig. 5. pixFRET plug-in interfaces. A: SBT determination interface. B: FRET interface. [Color figurecan be viewed in the online issue, which is available at www.interscience.wiley.com.]tion of the generated FRET image. The backgroundpixels are hence automatically colorized in blue usinga look-up table designed for this purpose.Test of the PixFRET Plug-inTo validate the tools described above, we performed apixel-by-pixel analysis of FRET by computing an imageof two cells expressing different amounts of PPAR-ECFP and RXR-EYFP (Fig. 6) and compared theimpact of using constant or fitted SBT ratios on FRETdisplay. Since these cells express the same ECFP andEYFP fusion proteins, both of them should exhibit similarnormalized FRET values. When constant SBTratios are used, FRET efficiency appears higher in cell2 than cell 1 (Figs. 6B and 6C, left panels), whereas theuse of the expNFRET formula generated an imagewhere the two cells display similar intensities (Figs. 6Band 6C, right panels). This visual appreciation of theadvantage of using expNFRET was confirmed whenthe distributions of pixel intensities within each cellwere plotted as histograms and average FRET valuescalculated (Fig. 6D). When constant SBT ratios wereused, the average normalized FRET value in cell 1 was4.9, versus 13.8 in cell 2, whereas the use of expNFRETgenerated closer mean values (9.6 for cell 1 versus 7.4for cell 2). Indeed, when using an average SBT ratio fornormalized FRET calculation (see formula (1)), mostdonor SBT values are underestimated in cell 2, whereCFP levels are high, and overestimated in cell 1 whereCFP levels are low. This is why overall and consistentwith what was observed at the cell population level inFigure 4, the use of the exponential fit allows a betternormalization of the data by reducing the variability ofFRET signals generated from different ranges of fluorophoreintensity.DISCUSSIONFRET is a technique whose use is rapidly expandingamong cell biologists as it provides very valuableinformation about physical interactions between moleculeswithin cells. Numerous procedures exist andhave been used to determine FRET in living cells,each of them having advantages and drawbacks (Berneyand Danuser, 2003), and we are still in a phasewhere the robustness of the methods has to beimproved (van Rheenen et al., 2004). One importantissue that has received only limited attention so faris the variation of SBT ratios as a function of fluorophoreintensity when performing sensitized emissionFRET. Using the most widely used fluorophore pairin cell biology, i.e. ECFP and EYFP, we show thatunder some common technical circumstances, theseratios are not constant and can vary with fluorophoreconcentration. In cases where the variations of fluorophoreconcentrations are small, assuming constantSBT ratios may be sufficient to get data of satisfactoryaccuracy. However, when these SBT ratios varyimportantly and thereby clearly bias FRET calculation,and when only small variations in FRET efficiencyare expected between cells or samples, solutionsto tackle this problem are required.Variations in SBT ratios as a function of fluorophoreintensity could be assigned here to a differentialresponse of the PMTs used for detection in theDonor and FRET channels. Although using a uniquePMT both for the detection of the donor and of the
PixFRET: AN IMAGEJ PLUG-IN FOR PIXEL-BY-PIXEL ANALYSIS OF FRETFig. 6. expNFRET reduces intercellular variability in pixel-bypixelanalyses. Cos-7 cells were transfected with expression vectorsfor PPARa-ECFP and RXRa-EYFP. A: Images of two cells expressingboth PPAR-ECFP and RXR-EYFP in the CFP and YFP setting. AGaussian blur of 1 was applied to the original image. B: NFRET andexpNFRET images generated by the PixFRET plug-in. C: Same as (B)but the cells are pseudocolorized with a different look-up table to bettervisualize intensity differences. D: Distribution of pixel intensitieswithin the two cells when the NFRET or expNFRET methods areused. The red bar and the number indicate the mean intensity overthe entire cell.acceptor could circumvent this problem, the requiredmechanical movement of the PMT (or of the filtercube wheel in other set-ups) implies that the imagesin the FRET, the donor, and the acceptor channelsare acquired sequentially. While this approach iswell applicable to fixed samples or to slowly diffusingor immobile proteins, it is more problematic forthe study of interactions of highly mobile proteinssuch as PPARs (Feige et al., 2005). Scanning eachchannel between lines, which can only be performedwith two separate PMTs, therefore, limits the diffusionof the interacting complexes during the time ofscanning. It is noteworthy that on a different microscope,using an array of PMTs, the same problemwas encountered. Hence, our results are of generalinterest to scientists performing FRET on a confocalmicroscope. Interestingly, Chen et al. (in press) havealso observed variations in SBT ratios when performingFRET with a 2-photon excitation set-up, andconsistent with our study, they did not observe anyvariation of the SBT ratios on a wide-field microscope,using a CCD camera detection.In cases where SBT ratios vary as a function of fluorescenceintensity, we showed that FRET calculationcan be improved by modeling these variations. This isparticularly important when fluorophore intensitiesvary greatly between cells. We demonstrated here thatthe expNFRET method reduces the variability of theFRET values calculated for the PPAR/RXR interaction.Although in our case, the exponential modeling generatedthe best results, other fits might be tested in situationswhere SBT ratios vary differently.Cells are highly organized and protein distributionas well as interactions are often limited to specificcompartments. It is hence of importance to be able tomap FRET precisely within cells to better characterizethe mode of action of interacting proteins. Toachieve this goal, we developed a plug-in called Pix-FRET for the ImageJ software that generates animage where NFRET is calculated for each pixel. Theplug-in allows one to directly determine the SBTparameters from images acquired when only thedonor or the acceptor are present. As various normalizationmethods have been proposed (Gordon et al.,1998; Xia and Liu, 2001), PixFRET allows the user tochoose the type of normalization desired. To ourknowledge, this is the first freely available programthat offers such possibilities and the source files willbe available upon request, allowing users to programspecific SBT ratio modeling methods according totheir needs.In conclusion, we uncovered the bias that the useof several PMTs may introduce when performingFRET experiments. We show that in cases where aset-up with two PMTs or an array of PMTs is preferredor required, this bias can be circumvented by asimple modeling of SBT ratios. We also developed auser-friendly and free interface, called PixFRET, thatallows a simple and rapid determination of SBTparameters and to display normalized FRET images.Altogether, these results and the PixFRET plug-inwill help cell biologists to increase the precision ofFRET analyses conducted on confocal microscopes aswell as the visualization and localization of interactionsin living cells.ACKNOWLEDGMENTSWe thank Dr. Claude Berney for valuable discussionsand for the reading of the manuscript, Dr. NathalieGarin for access to the ISREC imaging platform, Drs.Peterson and DeGrazia for the gift of the GI-FITCligand, and Dr. Tavaré for the gift of the EYFP–ECFPconstruct.REFERENCESBerney C, Danuser G. 2003. FRET or no FRET: a quantitative comparison.Biophys J 84:3992–4010.Chen Y, Elangovan M, Periasamy A. 2005. FRET data analysis: thealgorithm. In: Periasamy A, Day RN, editors. Molecular imaging:FRET microscopy and spectroscopy. Oxford: Oxford UniversityPress (in press).DeGrazia MJ, Thompson J, Heuvel JP, Peterson BR. 2003. Synthesisof a high-affinity fluorescent PPARgamma ligand for highthroughputfluorescence polarization assays. Bioorg Med Chem 11:4325–4332.Elangovan M, Wallrabe H, Chen Y, Day RN, Barroso M,Periasamy A. 2003. Characterization of one- and two-photon excitationfluorescence resonance energy transfer microscopy. Methods29: 58–73.57
Page 1 and 2: MICROSCOPY RESEARCH AND TECHNIQUE 6
Page 3 and 4: PixFRET: AN IMAGEJ PLUG-IN FOR PIXE
Page 5: PixFRET: AN IMAGEJ PLUG-IN FOR PIXE

PixFRET, an ImageJ plug-in for FRET calculation ... - ResearchGate

Create successful ePaper yourself

Delete template?

Save as template?