computer program for analyzing donor photobleaching fret image series

10
Computer Program for Analyzing Donor Photobleaching FRET Image Series Gergely Szentesi, 1,2 Gyorgy Vereb, 1 G abor Horv ath, 1 Andrea Bodn ar, 2 Akos F abi an, 1 J anos Matk o, 3 Rezs } oG asp ar, 1 S andor Damjanovich, 1,2 L aszl oM atyus, 1 * and Attila Jenei 1 1 Department of Biophysics and Cell Biology, Research Center for Molecular Medicine, Medical and Health Science Center, University of Debrecen, Debrecen, Hungary 2 Cell Biophysics Research Group of the Hungarian Academy of Sciences, Hungary 3 Department of Immunology, Eotvos Lor and University, Budapest, Hungary Received 21 April 2005; Revision received 6 June 2005; Accepted 7 June 2005 Background: The photobleaching fluorescence reso- nance energy transfer (pbFRET) technique is a spectro- scopic method to measure proximity relations between fluorescently labeled macromolecules using digital ima- ging microscopy. To calculate the energy transfer values one has to determine the bleaching time constants in pixel-by-pixel fashion from the image series recorded on the donor-only and donor and acceptor double-labeled samples. Because of the large number of pixels and the time-consuming calculations, this procedure should be assisted by powerful image data processing software. There is no commercially available software that is able to fulfill these requirements. Methods: New evaluation software was developed to analyze pbFRET data for Windows platform in National Instrument LabVIEW 6.1. This development environment contains a mathematical virtual instrument package, in which the Levenberg-Marquardt routine is also included. As a reference experiment, FRET efficiency between the two chains (b2-microglobulin and heavy chain) of major histocompatibility complex (MHC) class I glycoproteins and FRET between MHC I and MHC II molecules were determined in the plasma membrane of JY, human B lym- phoma cells. Results: The bleaching time constants calculated on pixel-by-pixel basis can be displayed as a color-coded map or as a histogram from raw image format. Conclusion: In this report we introduce a new version of pbFRET analysis and data processing software that is able to generate a full analysis pattern of donor photo- bleaching image series under various conditions. q 2005 International Society for Analytical Cytology Key terms: photobleaching; photobleaching fluorescence resonance energy transfer; LabVIEW; computer program Investigation of protein-protein interactions is impor- tant in understanding the structure-function relations in living cells. Fluorescence resonance energy transfer (FRET) techniques are excellent tools for determining association patterns of transmembrane proteins at the cell surface. With the help of FRET, molecular dimensions and molecular proximity can be measured and determined in functioning, live cells and provide information that would be hindered by the implicit aggregating nature of other classic approaches. The theory of FRET was first developed by Theodor Forster (1). The FRET process is a dipole-dipole interac- tion in which an excited donor fluorophore transfers its energy to an acceptor molecule in close vicinity (1–10 nm) in a nonradiative way. The main application of FRET as a spectroscopic ruler (2) is based on the fact that the rate of energy transfer depends on the inverse sixth power of the distance between the two interacting molecules. The common way to measure FRET processes is by the decrease in donor fluorescence in the presence of an acceptor, which can be accompanied by an increase of fluorescence of the acceptor (if the acceptor is a fluores- cent molecule). In recent years numerous FRET-based techniques have been developed for flow cytometry (3–5) and microscopy (6–10). Several biological structures have been successfully investigated by these methods: recep- tors involved in immune response (11–13), growth factor receptors on tumor cells (14–16), determination of lipid Contract grant sponsor: Hungarian Academy of Sciences (OTKA); Contract grant numbers: F034487, F046497, T043509, T042618, T037831; Contract grant sponsor: Ministry of Health and Welfare (ETT); Contract grant numbers: 013/2001, 603/2003, 602/2003, 532/2003; Contract grant sponsor: RET; Contract grant number: 06/2004. Contract grant sponsor: B ek esy Fellowship, Hungarian Ministry of Education. *Correspondence to: L aszl oM atyus, Department of Biophysics and Cell Biology, Medical and Health Science Center, University of Debrecen, P.O. Box 39, H-4012 Debrecen, Hungary. E-mail: [email protected] Published online 12 September 2005 in Wiley InterScience (www. interscience.wiley.com). DOI: 10.1002/cyto.a.20175 q 2005 International Society for Analytical Cytology Cytometry Part A 67A:119–128 (2005)

Upload: independent

Post on 02-May-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Computer Program for Analyzing DonorPhotobleaching FRET Image Series

Gergely Szentesi,1,2 Gy€orgy Vereb,1 G�abor Horv�ath,1 Andrea Bodn�ar,2 �Akos F�abi�an,1

J�anos Matk�o,3 Rezs}o G�asp�ar,1 S�andor Damjanovich,1,2 L�aszl�o M�atyus,1* and Attila Jenei11Department of Biophysics and Cell Biology, Research Center for Molecular Medicine, Medical and Health Science Center,

University of Debrecen, Debrecen, Hungary2Cell Biophysics Research Group of the Hungarian Academy of Sciences, Hungary

3Department of Immunology, E€otv€os Lor�and University, Budapest, Hungary

Received 21 April 2005; Revision received 6 June 2005; Accepted 7 June 2005

Background: The photobleaching fluorescence reso-nance energy transfer (pbFRET) technique is a spectro-scopic method to measure proximity relations betweenfluorescently labeled macromolecules using digital ima-ging microscopy. To calculate the energy transfer valuesone has to determine the bleaching time constants inpixel-by-pixel fashion from the image series recorded onthe donor-only and donor and acceptor double-labeledsamples. Because of the large number of pixels and thetime-consuming calculations, this procedure should beassisted by powerful image data processing software.There is no commercially available software that is able tofulfill these requirements.Methods: New evaluation software was developed toanalyze pbFRET data for Windows platform in NationalInstrument LabVIEW 6.1. This development environmentcontains a mathematical virtual instrument package, inwhich the Levenberg-Marquardt routine is also included.

As a reference experiment, FRET efficiency between thetwo chains (b2-microglobulin and heavy chain) of majorhistocompatibility complex (MHC) class I glycoproteinsand FRET between MHC I and MHC II molecules weredetermined in the plasma membrane of JY, human B lym-phoma cells.Results: The bleaching time constants calculated onpixel-by-pixel basis can be displayed as a color-coded mapor as a histogram from raw image format.Conclusion: In this report we introduce a new versionof pbFRET analysis and data processing software that isable to generate a full analysis pattern of donor photo-bleaching image series under various conditions. q 2005

International Society for Analytical Cytology

Key terms: photobleaching; photobleaching fluorescenceresonance energy transfer; LabVIEW; computer program

Investigation of protein-protein interactions is impor-tant in understanding the structure-function relations inliving cells. Fluorescence resonance energy transfer(FRET) techniques are excellent tools for determiningassociation patterns of transmembrane proteins at the cellsurface. With the help of FRET, molecular dimensions andmolecular proximity can be measured and determined infunctioning, live cells and provide information that wouldbe hindered by the implicit aggregating nature of otherclassic approaches.

The theory of FRET was first developed by TheodorF€orster (1). The FRET process is a dipole-dipole interac-tion in which an excited donor fluorophore transfers itsenergy to an acceptor molecule in close vicinity (1–10nm) in a nonradiative way. The main application of FRETas a spectroscopic ruler (2) is based on the fact that therate of energy transfer depends on the inverse sixth powerof the distance between the two interacting molecules.The common way to measure FRET processes is by thedecrease in donor fluorescence in the presence of an

acceptor, which can be accompanied by an increase offluorescence of the acceptor (if the acceptor is a fluores-cent molecule). In recent years numerous FRET-basedtechniques have been developed for flow cytometry (3–5)and microscopy (6–10). Several biological structures havebeen successfully investigated by these methods: recep-tors involved in immune response (11–13), growth factorreceptors on tumor cells (14–16), determination of lipid

Contract grant sponsor: Hungarian Academy of Sciences (OTKA);

Contract grant numbers: F034487, F046497, T043509, T042618, T037831;

Contract grant sponsor: Ministry of Health and Welfare (ETT); Contract grant

numbers: 013/2001, 603/2003, 602/2003, 532/2003; Contract grant

sponsor: RET; Contract grant number: 06/2004. Contract grant sponsor:

B�ek�esy Fellowship, Hungarian Ministry of Education.

*Correspondence to: L�aszl�o M�atyus, Department of Biophysics and Cell

Biology, Medical and Health Science Center, University of Debrecen, P.O.

Box 39, H-4012 Debrecen, Hungary.

E-mail: [email protected] online 12 September 2005 in Wiley InterScience (www.

interscience.wiley.com).

DOI: 10.1002/cyto.a.20175

q 2005 International Society for Analytical Cytology Cytometry Part A 67A:119–128 (2005)

microdomain structures (17,18), and protein conforma-tion (19,20).

Fluorescence-activated cell sorting and analysis or flowcytometry is one solution for high-speed quantitative ana-lysis of cell. With this technique cell populations can bestudied and evaluated relatively fast and the energy trans-fer efficiency values can be determined on a cell-by-cellbasis. Because of the large number of cells, it can providerelatively good statistics (21), but the energy transfervalues are averaged for each cell, so no information isavailable about the diversity of these values on the cell sur-face of single cells.

Different fluorescent microscopic techniques havebecome very popular during the last 15 years for studyingvarious components of the cell. Parallel to the evolutionof these techniques has been the realization of the impor-tance of the lateral distribution and topology of the differ-ent cell surface components. There are several spectro-scopic methods adapted and applied together with micro-scopic techniques. For the imaging techniques theapplication of FRET for inspection of receptor associationsprovides an extra possibility for surpassing the actual dif-fraction limited resolution of the light microscope.

Upon continuous excitation of the donor molecules,their fluorescence intensity decreases due to photobleach-ing, an irreversible, oxygen-dependent photochemicaldegradation process. For a double-labeled sample, i.e.,with donor and acceptor molecules present, FRETbetween the fluorophores opens an additional relaxationpathway for the excited donor molecules. Thus thedecline in the number of donor excited states by photo-bleaching takes longer. This method was designated‘‘donor photobleaching FRET (pbFRET)’’ at the time of itsintroduction (22–24).

Because of the uneven distribution of cell surface com-ponents, there are several regions at the labeled surfacethat demonstrate different fluorescence intensities. Thesespatial heterogeneities can also be very important toanswer relevant biological questions (25–28).

To determine the transfer efficiency (E), two series ofimages should be recorded, one from the sample labeledsolely by donor and one from a sample double-labeled bydonor and acceptor. A convenient means for generatingthe ‘‘acceptor-alone’’ reference state is to photobleach theacceptor in an appropriate region of the sample and con-venient combined protocols for donor and acceptorphotobleaching have been implemented (29).

The photobleaching decay curves can be obtained fromthe image series by fitting exponential function to eachpixel series (22–24). At the time when the initial fluores-cence intensity decreases to 1/e, part of its initial value isdefined as the photobleaching time constant. The energytransfer efficiency can be calculated from the bleachingtime constants of the donor-only labeled and donor-accep-tor double-labeled samples.

Several commercially available programs have optionsto record and display bleaching image series and to gener-ate bleaching curves from the recorded images, but theydo not support curve fitting or only single curve fitting is

possible (30). Because of the large number of pixels (ingeneral 512 3 512 pixel image size) and the time-consum-ing calculations, this procedure should be assisted by apowerful image processing software.We have developed a program to calculate FRET effi-

ciency from raw image file series in Windows 9x and Win-dows XP operating systems. The program automaticallygenerates the bleaching decay curves from the image ser-ies and fits them with optional single, double, or tripleexponential functions. The bleaching time constant valuescan be displayed and saved in color-coded images (‘‘taumap’’) and histograms. The program also calculates statisti-cal values of histograms to determine the mean bleachingtime constant value and its standard deviation. We providea short description of the theoretical background ofpbFRET and the functionality of the software in additionto a representative experiment to demonstrate how theprogram works.

THEORETICAL BACKGROUNDFluorescence Resonance Energy Transfer

In a fluorescence process, light of intensity I0 excitesthe donor fluorophore from the ground state D0 at a ratekex to one of the electronic excited levels D*. The exciteddonor molecule has several ways for relaxation, includingthe radiative fluorescence (kfl) and the nonradiative inter-nal conversion (kic) and intersystem crossing (kisc), withoverall decay rate constant krel. Introducing a suitableacceptor, a new decay path is opened by the nonradiativeresonance energy transfer with the rate ket. F€orster’s the-ory shows that the rate of energy transfer (ket) for a singledonor-acceptor pair is given by the following formula:

ket ¼ krelR0

R

� �6

ð1Þ

where R0 is the F€orster distance that corresponds to anenergy transfer efficiency probability of 0.5 for a givendonor-acceptor pair. It depends on the index of refraction ofthe conveying medium, the fluorescence quantum yield ofthe donor in the absence of energy transfer, the mutualorientation of the transition dipole moment of the donorand acceptor molecules, and the overlap integral of thedonor emission and acceptor excitation spectra. The energytransfer efficiency E is defined by the following equation:

E ¼ ket=ðkrel þ ketÞ ¼ 1� sdasd

ð2Þ

where sda 5 1/(krel 1 ket) and sd 5 1/krel are the fluores-cence lifetimes of the excited state D*, with and withoutenergy transfer, i.e., in the presence or absence of anacceptor, respectively.

Determination of Energy Transfer byPhotobleaching

It is possible to measure the sda and sa in the nano-second time domain and to determine the E value; how-

120 SZENTESI ET AL.

ever, considerable simplification can be achieved if thefluorescence lifetime dependence of the fluorophore onthe photobleaching time constant s1 is used.

Each molecule in the excited state has a given probabil-ity (kb) to be decomposed by photochemical processes.To simplify the equations, the overall rate of the relaxationprocesses is marked with krel, excluding energy transferand photobleaching (Fig. 1). Two coupled differentialequations can be formed (22–24,31–33), one for the exci-tation (rate of populating the excited state) and one forthe relaxation (rate of depletion of the excited state) pro-cess of the donor molecules. By solving the equations, thefollowing two estimated photobleaching time constantscan be obtained:

s1 � ðkrelÞ=ðkexkbÞ ð3Þ

s01 � ðket þ krelÞ=ðkexkbÞ ð4Þ

in the absence and presence of acceptor molecules,respectively. The photobleaching time constant (s01) of thefluorophore is increased by energy transfer because of theadditional pathway for relaxation (Fig. 2). By measuring

the lifetime s1 and s 01 by photobleaching experiments, the

energy transfer efficiency can be determined by the fol-lowing formula:

E ¼ 1� s1=s01 ð5Þ

MATERIALS AND METHODSCells

The Epstein-Barr virus–transformed human B lympho-blastoid cell line, JY, was grown in RPMI 1640 mediumcontaining 10% heat-inactivated fetal calf serum, 2 mMI-glutamine, and 50 lg/ml gentamicin in humidified aircontaining 5% CO2 at 37�C.

Monoclonal Antibodies

W6/32 (immunoglobulin G2aj) specific for the heavychain of major histocompatibility complex (MHC) class I(HLA-A, -B, -C), L368 (immunoglobulin G1j) binding b2-microglobulin (the light chain of MHC class I), and L243(immunoglobulin G2a) specific for MHC class II (HLA-DR)were prepared from hybridoma supernatants (kindly pro-vided by F. Brodsky, University of California at San Fran-cisco, USA) by protein A-affinity chromatography. Aliquotsof purified monoclonal antibodies (mAbs) were conju-gated with 6-(fluorescein-5-carboxamido) hexanoic acidsuccinimidylester (SFX) or 6-(tetramethyl-rhodamine-5-(and-6)-carboxamido) hexanoic acid succinimidyl ester(TAMRAX; Molecular Probes, Eugene, OR, USA) as pre-viously described (34). Labeling ratios were determinedby a spectrophotometer. The fluorescently conjugatedmAbs retained their affinity according to competition withidentical, unlabeled antibodies.

Labeling of Cells With Fluorescent Antibodies

Cells were washed twice and suspended in phosphatebuffered saline (PBS; pH 7.4) at a concentration of 0.5 to1 3 106 cells/50 ll and were incubated with saturatingamounts of SFX- and/or TAMRAX-conjugated mAbs for45 min on ice. Thereafter cells were washed twice in PBSand fixed with 1% formaldehyde in PBS on ice for 30 min.During labeling special care was taken to keep the cells atice-cold temperature to avoid induced aggregation of cellsurface molecules.

FIG. 1. Photophysical scheme of the com-bined process of fluorescence resonance energytransfer and donor photobleaching. The possi-ble relaxation ways of an excited donor mole-cule are displayed in the figure, where D*, D#,D, and A* represent the excited, bleached andground states of the donor molecule, and theexcited state of the acceptor molecule, respec-tively. The overall rate of relaxation is markedwith krel, excluding energy transfer and photo-bleaching, which are denoted by ket and kb,respectively.

FIG. 2. Bleaching curves of single- and double-labeled samples withbleaching time constants ss1 and s01. The bleaching time constant (s10) ofthe transfer sample (labeled with donor and acceptor) is larger than thatof the donor-only labeled sample (s1) because of the additional way forrelaxation. Thus the resulting bleaching curve is flatter and the initialfluorescence intensity is lower.

121PBFRET CALCULATIONS FROM IMAGE SERIES

Donor Photobleaching Procedure andImage Acquisition

Photobleaching data collection was developed for aninverted microscope (Zeiss Axiovert 100). The fluores-cence microscope is configured with exchangeable XBO75-W and HBO 100-W lamps. The optical path is shown inFigure 3. The instrument incorporates an electronic shut-ter that allows time-gated excitation. During one period ofexcitation (1 to 3 s) the intensified video camera collectsseveral frames (full frame is 512 3 480 pixels, but 1/4, 1/16, 1/64 parts can also be obtained), which are digitizedby a Matrox frame grabber card at 8-bit resolution. Themeasurement also can be performed on a confocal laserscanning microscope.

The irradiance level under which the emission processis measured has to be considered (35) because the photo-bleaching process can be influenced by different levels ofexcitation. At low levels of excitation (low irradiance)photobleaching is not significant because there is noappreciable depletion of the ground state of the donormolecule. If the applied level of excitation is too high, thevarious excited states of donor molecules are saturated,and the long-lived intermediate states (i.e., triplet state)are in a steady-state distribution. Under this condition thefluorescence no longer is linearly dependent on the excita-tion intensity and the photobleaching rate constant doesnot provide a direct measure of FRET efficiency. Thus amoderate light level is necessary to perform photobleach-ing measurement because in this case the rate of photo-bleaching and the fluorescence intensity of donor mole-cules linearly depend on the intensity (photon flux) ofexcitation light.

The number of images to be collected depends on thebleaching time constant of the fluorophore and the energytransfer efficiency. The higher the bleaching time con-stant, the longer illumination time required for bleachingthe same amount of donor molecules. The occurrence of

energy transfer also slows down the bleaching processbecause it opens an additional way for relaxation ofexcited state donor molecules, and thus the mean bleach-ing time constant of donor fluorophore increases.

Determination of Photobleaching TimeConstant Values

The pixel-by-pixel distribution of photobleaching timeconstants is best determined by fitting the photobleachingdecay curve in each pixel with an exponential function.In accordance with previous studies (36–38), the follow-ing double exponential equation gave the best fit with thesmallest least-squares errors in our experiments:

IðtÞ ¼ A1e�t=s1 þ A2e

�t=s2 þ I1 ð6Þ

where I(t) is the fluorescence intensity at time t, I1 is thebackground fluorescence, A1 and A2 are the amplitudes ofthe two components, and s1 and s2 are the bleaching timeconstants. Nonetheless, the program described herein isdesigned to be versatile and can optionally fit one or threeexponentials in addition to the above formula.From the fitted parameters the weighted mean bleach-

ing time constant (Æsæ) can be calculated as follows:

hsi ¼ A1s1 þ A2s2A1 þ A2

ð7Þ

To perform the exponential fitting for each pixel, theintensity values of every pixel has to be collected to a lin-ear array from all images of bleaching series (Fig. 4). Thephotobleaching decay curves are generated by plottingthese intensity values as a function of time (in arbitraryunits).

FIG. 3. Schematic drawing of a conven-tional inverted epifluorescence microscopewith mercury arc-lamp excitation used forpbFRET measurements. Dotted lines repre-sent the optical pathways. The electric PC-driven shutter is installed after the lamp togenerate the excitation (ex.) light pulses.The excitation and emission (em.) wave-lengths are selected by a filter cube contain-ing three different filter configurations thatcan be adjusted manually. The emitted fluor-escent light can be recorded with an intensi-fied video camera (CCD).

122 SZENTESI ET AL.

Flatfield Correction

As discussed under Donor Photobleaching Procedureand Image Acquisition, the bleaching time constant of afluorophore is a function of the intensity of the excitationlight. At sufficiently low intensities that are below satura-tion, the equilibrium percentage of excited state fluoro-phores is a linear function of the illumination flux.Further, if reactants around the fluorophore are in greatexcess, photobleaching becomes a quasi-monomolecularreaction, with a rate constant inversely proportional tothe excited state population and, hence, excitation inten-sity. To correct for the inhomogeneity of excitation inten-sity, we have used a uranyl acetate glass fluorescence stan-dard of even fluorophore distribution. By measuring thefluorescence intensity of the standard, the calculatedmean bleaching time constant values can be correctedaccording to the following formula:

hsii;j;corr ¼ hsii;jcfi;j ð8Þ

where Æsæi,j and Æsæi,j,corr are the uncorrected and correctedmean bleaching time constant values of pixel [i, j], and cfi,j isthe correction factor of pixel [i, j] calculated by equation:

cfi;j ¼ Ii;j � Bgi;j

hI � Bgi ð9Þ

where Ii,j is the fluorescence intensity of pixel [i, j] in thestandard image of the uranyl acetate glass, Bgi,j is the darkcurrent value from the same pixel, and ÆI 2 Bgæ is theaverage fluorescence intensity from the same background-corrected image.

Fitting Algorithm

Our software uses the Levenberg-Marquardt nonlinearfitting algorithm to fit bleaching curves, with three dif-ferent formulas, single, double, or triple exponentialfunctions. The fitting procedure is based on the mini-mization of the v2 merit function to find the best fitparameters, with nonlinear dependences. The minimiza-tion must proceed iteratively by giving trial values forthe parameters and the algorithm improves the trialsolution until the v2 stops (or effectively stops)decreasing (39). LabVIEW contains a mathematical vir-tual instrument (VI) package in which the Levenberg-Marquart subVI is also included. The subVI icon of theLevenberg-Marquardt (Lev-Mar subroutine) is shown inFigure 5.The input parameters are the following:

1. Standard deviation: the array of standard deviations,r[i], for data point (x[i],y[i]). The default value is 1.0 ifthe standard deviations are all equal or unknown.2. X and Y: the array of data points representing inde-

pendent variables.3. Initial guess coefficients: initial array of trial values.4. Max iteration: maximum number of iterations.5. Derivative: the formula for the derivative of a fitted

function can be specified in the ‘‘Formula Node’’ (graphi-cal mathematical expression evaluator) on the block dia-gram.

The output parameters of the subVI are the following:

1. Covariance: matrix of covariances C.

FIG. 4. Bleaching series. Arepresentative series of imagesobtained in a photobleachingexperiment. The bleaching decaycurve of a single pixel (1) and itsdouble exponential fitted curve(solid line) are also plotted.

FIG. 5. The scheme of subVI Levenberg-Marquardt of the developing program LabVIEW. The objects, structures, and subroutines are ‘‘hidden’’ inside sub-VIs, represented by single icons on the developer interface. Each one of these can have input and output parameters indicated with different color pinsdepending on the type of parameters. The subVI Levenberg-Marquardt has six input (left side of the icon) and five output (right side) parameters, describedin detail under Materials and Methods.

123PBFRET CALCULATIONS FROM IMAGE SERIES

2. Best-fit coefficients: set of coefficients that minimizev2, defined in the following equation:

:v2 ¼Xn�1

i¼0

yi � f ðxi;a1:::amÞri

� �2

ð10Þ

3. Best fit: fitted data array. Values are computed byusing the best-fit coefficients.

4. MSE: mean squared error.5. Error: returns any error or warning condition from

the subVI.

Detailed information about this subVI can be found inthe Help of LabVIEW (LabVIEW Help, National Instru-ments Corporation, Austin, TX, USA).

RESULTSSoftware

The program was developed in National InstrumentLabVIEW 6.1 (National Instruments Corporation). Wechose this software because of its powerful graphicaldevelopment environment, which provides severalobjects (graph controls, image display control) to aiddesigning the program, and it contains several precom-piled mathematical VIs and routines.

The minimum system requirements are the following:

� Windows 9x or later versions of the Windows operat-ing system

� IBM-compatible PC with Pentium 166 MHz� 128 MB RAM� Desktop area 10243 768 pixels� 32-Bit color palette� Mouse or compatible pointing device

Description of the Program

User interfaces. This software was developed to han-dle the image files generated by the acquisition softwareof our instrument (Attofluor, Zeis, Oberkochen Germany),i.e., 8-bit resolution raw image files with dimensions 60 364, 120 3 128, 240 3 256, and 480 3 512. This imagetype has no header or other additional information aboutthe structure of the file. We chose this image format forour application because most of the image analysis soft-ware have the ability to convert image files to this format.Thus this program can be used to analyze bleaching imageseries measured by any other instrument (e.g., confocallaser scanning microscope).

The main window of the program contains only fourbuttons to run the subprograms, Analyze Pixel, Multifit,Multi Analyze, and Distribution Viewer, in sequence.

Analyze Pixel subprogram. The aim of this subprogramis to help the user decide which type of function providesthe best fit to the photobleaching curves. The interfacecontains a frame that shows the first image of the bleach-

ing series. The pixel of interest can be chosen by movinga small yellow cursor over the first image. Next to theimage display frame, a plot is placed to show the photo-bleaching and the fitted curve together, which can besaved as a tab-delimited ASCII text file. A table is placedunder the image window in which the initial and theresult fitting parameters are displayed. The Movie subVIcan be opened from this window to see the bleaching pro-cess as a movie and to check whether the studied cellsshifted during acquisition.Multifit subprogram. Fitting all the bleaching curves in

all series would be a remarkably time-consuming proce-dure. The Multifit subprogram provides an interfacewhere all bleaching series and their initial fitting para-meters can be adjusted before analysis. The thresholdintensity and the standard images (for flatfield correction)can be adjusted also. It is also possible to include anddelete image series from the analysis procedure beforestarting the overall fitting course. After all relevant seriesare added to the series list and all initial parameters aredefined, the program automatically fits the bleachingcurves of all pixels above the threshold and writes theresults into a tab-delimited ASCII text file. The structure ofthis temporary file (resu.lts) is shown in Figure 6.Multi Analyze subprogram. After the fitting procedure

of the Multifit, the results can be surveyed in this subpro-gram. The first image is displayed in a frame that includestwo cursors with which one rectangular part of the imagecan be selected. Selecting one portion of the imageenables choosing only a subset of the calculated bleachingtimes according to user-adjusted minimum and the maxi-mum bleaching times, fluorescence intensities, and themaximum of the standard error of mean of the fitting. Sli-ders are placed on the interface to change acceptanceranges of these parameters. The selected pixels and theirfitting parameters are appended to an ASCII text result filefor further analysis. A plot is also included to display thebleaching and fitted curves together. The fitting para-meters can be displayed in a frame as a color-coded map.Distribution Viewer subprogram. With this subpro-

gram it is possible to collect the results of the individualmeasurements as one array and display the bleaching timeconstants in a histogram. The histograms can be fittedwith Gaussian functions and the resulting parameters aredisplayed in the histogram frame. Photobleaching timeconstants can be gated on the histograms with three dif-ferent colored gates: red, green, and yellow. The color-coded photobleaching time constants are overlaid on thefirst image of the bleaching series. The color code is gener-ated according to the maximum and minimum values ofthe photobleaching time constant subset. The lowest andhighest values of the subset get the darkest and brightesttones of the gate color. The color-coded image can besaved as a 24-bit bitmap image.Test sample. To demonstrate the functionality of the

program, FRET efficiencies between the two chains (b2m,heavy chain) of MHC class I and between MHC class I(b2m) and MHC class II were determined at the surface ofJY human B lymphoma cells. Proteins were targeted by

124 SZENTESI ET AL.

SFX- or TAMRAX-conjugated mAbs serving as donor andacceptor, respectively.

FRET efficiencies were determined by using the follow-ing samples:

1. JY cells labeled with SFX-L368 (targeting b2m;donor)

2. JY cells simultaneously labeled with SFX-L368 andTAMRAX-W6/32 (targeting b2m and MHC class I heavychain, respectively; donor 1 acceptor)

3. JY cells labeled with SFX-L368 and TAMRAX-L243(targeting b2m and MHC class II, respectively; donor 1acceptor)

4. JY cells labeled with TAMRAX-L243 (binding MHCclass II; acceptor-only labeled)

5. Unlabeled JY cells

Because the photobleaching process is highly sensitiveto oxygen concentration, experiments were performedon the mixture of donor-only and double-labeled sam-ples to ensure the same level of dissolved oxygen. Todistinguish donor-only and double-labeled cells, at theend of the experiment an additional image was recordedin the acceptor channel (kex 5 543 nm, kem > 580 nm).The sample labeled with TAMRAX-L243 alone (acceptor-only labeled sample) was used to check whether theintensity of the acceptor was affected by donor excita-tion. If the acceptor is also bleached upon donor photo-bleaching, it loses its ability to accept energy from theexcited donors and, as a result, the bleaching time con-stant decreases. The fifth sample was used as a back-ground control.

To prove the applicability of our software for any acqui-sition system, the same samples were measured on a ZeissLSM 510 microscope (lasers: Ar ion at 488 for donor exci-tation, HeNe at 543 nm for acceptor excitation), andimage series were converted to raw format and analyzedwith our software. Four image series of two to three cellswere taken from each sample (�10 cells/sample). Wedetermined 600 to 1,500 photobleaching time constants(s) per cell, i.e., the mean time constants were calculatedfrom �7,000 data for each sample. The resulting distribu-tion histograms of ‘‘tau’’ values along with representativeimage triplets (images taken in the donor and acceptorchannels and the color-coded map of photobleaching timeconstants) for the double-labeled samples are shown inFigure 7.The mean values of photobleaching time constants for

donor-only labeled cells (sample 1) were independent ofwhether they were determined in the mixture of samples1 and 2 or that of samples 1 and 3: the calculated valueswere 1.69 6 0.35 and 1.65 6 0.48, respectively. In thecase of the double-labeled cells, the mean constants forsamples 2 and 3 were 2.62 6 0.56 and 1.78 6 0.56,respectively. FRET efficiencies for the SFX-L368/TAMRAX-W6/32 (b2m 2 MHC class I heavy chain) and for the SFX-L368/TAMRAX-L243 (b2m 2 MHC class II) pairs were35.5% and 6.9%, respectively.

DISCUSSION

The photophysical consequences of the FRET phenom-enon offer several ways to measure FRET efficiency. Oneof these possibilities is the acceptor-depletion FRET tech-nique (13,40–42), which is a useful pixel-by-pixel based fluor-

FIG. 6. The structure of the result file gen-erated by the subprogram Multifit. The out-put parameters of the fitting functions arestored pixel by pixel in a tab-delimited ASCIIlist mode file. The columns contain the Xand Y positions of the pixel, the mean photo-bleaching time constant and the meansquared error values, and the separate expo-nential function parameters. The initial inten-sity of the pixel is also stored at the end ofeach row.

125PBFRET CALCULATIONS FROM IMAGE SERIES

escence method. Using conventional fluorescence farfieldimage microscopic technique, a whole energy transfermap can be generated for cell surface receptors. With thismethod, some corrections have to be considered in thecalculation because the results can be affected by back-ground fluorescence and photobleaching of the donormolecules during acceptor bleaching. Despite these pro-blems, the acceptor-depletion method has the obviousadvantage that FRET can be determined from a single sam-ple labeled with donor and acceptor.

Using the donor pbFRET method, the above-mentionedtwo technical drawbacks (background fluorescence andacceptor bleaching upon donor excitation) can be mini-mized. This technique is a powerful microscopic methodto study the proximity and distribution of cell surface pro-teins on a molecular scale. To determine the energy trans-fer efficiency, two series of images should be recorded,one with a cell sample labeled solely with donor fluoro-phores and one with cells labeled with donor and accep-tor. From the image series the distribution of the cell sur-face molecules can be studied. Moreover, from the imageseries the photobleaching time constants can be calcu-

lated by fitting the decay curves with an exponential func-tion. To increase the accuracy of the mean of photobleach-ing time constant, values can be determined on a pixel-by-pixel basis. In general, the method allows for the selectionof any desired region of interest in various regions of sin-gle cells, cell conjugates, or thin tissue slices.Although the method is simple and does not need a

complicated hardware configuration, there is a lack of ana-lysis software because commercially available softwarecan only record and display a bleaching image series andgenerate bleaching curves from the recorded images, butusually do not support curve fitting algorithms or only sin-gle exponential curve fitting is possible.Here we present a LabVIEW application, with which

the raw image series can be processed to determine thephotobleaching time constant values. The resulting valuescan be plotted as a color-coded photobleaching time con-stant map to investigate the spatial distribution of the cellmembrane components. Further, histograms can be gener-ated from the fitting coefficients calculated on a pixel-by-pixel basis. The histograms can be gated by different colorgates and fitted with Gaussian functions. The gated values

FIG. 7. Representative pbFRET experiments between the two chains (b2m and heavy chain) of MHC class I (A–C, G) and between MHC class I (b2m) andMHC class II (D–F, H) glycoproteins on JY human B lymphoma cells. The two chains of MHC I were targeted by SFX-L368 and TAMRAX-W6/32 mAbs,respectively, whereas MHC II was labeled by TAMRAX-L243 mAb. SFX was used as donor and TAMRAX as acceptor. Photobleaching experiments were per-formed on the mixture of donor-only and double-labeled cells as detailed in the Sample Run. A, D: First images of the bleaching series recorded in the donorchannel. B, E: Images recorded in the acceptor channel after finishing donor photobleaching. C, F: Color-coded map of photobleaching time constants(‘‘tau’’ map). G, H: Distribution histograms of photobleaching time constants. The presence of FRET resulted in the shift of the distribution histograms ofdouble-labeled cells (solid lines) toward the higher values in comparison with those of donor-only labeled cells (dashed-dotted lines). This is also clearly visi-ble on the tau maps. In the case of G, the shift is more prominent, i.e., the two chains of MHC I molecules are in closer proximity than are b2m and MHCII; thus the probability of FRET is higher.

126 SZENTESI ET AL.

are displayed on the first image of the captured image ser-ies by the gating color.

The program can be used to analyze bleaching imageseries measured by any instrument because it was devel-oped to process raw image files. The program uses thisimage format because different equipments have severalimage formats to store the measured data, but most imageanalysis programs are able to convert image files to rawformat.

The fitted data are saved to a simple ASCII text file, sothey can be opened by any data processing software. Thesoftware package contains tools to convert the resultantdata map to Tagged Image File Format (TIFF) or ImageCytometry Standard (ICS) image formats for further analy-sis. These tools for formatting results are also installed bythe software. Details about their usage can be found in thesoftware manual.

The application can be used without an existing Lab-VIEW developer environment because the LabVIEW run-time engine is automatically installed by the Install Shieldwizard. The software is freely distributed and can beobtained free of charge from the authors. The softwarepackage includes a complete help and tutorial to help thefirst-time user.

In conclusion, we have introduced a new version ofpbFRET analysis and data processing software that are ableto generate a full analysis pattern of donor photobleachingdecay image series at various conditions, independently ofthe complexity of the decay process and therefore of thenature of applied fluorophore. It allows for analysis ofselected regions of interest on single cells or high through-put screening statistics on a large number of cells. Thedonor pbFRET approach together with the acceptor-deple-tion method used in conventional fluorescence or confo-cal laser scanning microscopes are of a continuously wideinterest nowadays in studying molecular aspects of cellu-lar recognition, communication, and signal processes inliving cells, particularly in cell biology, immunobiology,and neurobiology.

LITERATURE CITED1. F€orster T. Zwischenmolekulare Energiewanderung und Fluoreszenz.

Ann Phys 1948;2:55–75.2. Stryer L, Thomas DD, Carlsen WF. Fluorescence energy transfer mea-

surements of distances in rhodopsin and the purple membrane pro-tein. Methods Enzymol 1982;81:668–678.

3. Matyus L. Fluorescence resonance energy transfer measurements oncell surfaces. A spectroscopic tool for determining protein interac-tions. J Photochem Photobiol B 1992;12:323–337.

4. Szollosi J, Damjanovich S, Matyus L. Application of fluorescence reso-nance energy transfer in the clinical laboratory: routine and research.Cytometry 1998;34:159–179.

5. Vereb G, Matko J, Szollosi J. Cytometry of fluorescence resonanceenergy transfer. Methods Cell Biol 2004;75:105–152.

6. Damjanovich S, Vereb G, Schaper A, Jenei A, Matko J, Starink JP, FoxGQ, rndt-Jovin DJ, Jovin TM. Structural hierarchy in the clustering ofHLA class I molecules in the plasma membrane of human lymphoblas-toid cells. Proc Natl Acad Sci USA 1995;92:1122–1126.

7. Lidke DS, Nagy P, Barisas BG, Heintzmann R, Post JN, Lidke KA, Clay-ton AH, rndt-Jovin DJ, Jovin TM. Imaging molecular interactions incells by dynamic and static fluorescence anisotropy (rFLIM andemFRET). Biochem Soc Trans 2003;31:1020–1027.

8. Nagy P, Bene L, Balazs M, Hyun WC, Lockett SJ, Chiang NY, WaldmanF, Feuerstein BG, Damjanovich S, Szollosi J. EGF-induced redistribu-

tion of erbB2 on breast tumor cells: flow and image cytometric energytransfer measurements. Cytometry 1998;32:120–131.

9. Gu Y, Di WL, Kelsell DP, Zicha D. Quantitative fluorescence resonanceenergy transfer (FRET) measurement with acceptor photobleachingand spectral unmixing. J Microsc 2004;215:162–173.

10. Jares-Erijman EA, Jovin TM. FRET imaging. Nat Biotechnol 2003;21:1387–1395.

11. Damjanovich S, Matko J, Matyus L, Szabo G Jr, Szollosi J, Pieri JC, Far-kas T, Gaspar R Jr. Supramolecular receptor structures in the plasmamembrane of lymphocytes revealed by flow cytometric energy trans-fer, scanning force- and transmission electron-microscopic analyses.Cytometry 1998;33:225–233.

12. Dornan S, Sebestyen Z, Gamble J, Nagy P, Bodnar A, Alldridge L, DoeS, Holmes N, Goff LK, Beverley P, Szollosi J, Alexander DR. Differentialassociation of CD45 isoforms with CD4 and CD8 regulates the actionsof specific pools of p56lck tyrosine kinase in T cell antigen receptorsignal transduction. J Biol Chem 2002;277:1912–1918.

13. Vamosi G, Bodnar A, Vereb G, Jenei A, Goldman CK, Langowski J,Toth K, Matyus L, Szollosi J, Waldmann TA, Damjanovich S. IL-2 andIL-15 receptor alpha-subunits are coexpressed in a supramolecularreceptor cluster in lipid rafts of T cells. Proc Natl Acad Sci USA 2004;101:11082–11087.

14. Diermeier S, Horvath G, Knuechel-Clarke R, Hofstaedter F, Szollosi J,Brockhoff G. Epidermal growth factor receptor coexpression modu-lates susceptibility to Herceptin in HER2/neu overexpressing breastcancer cells via specific erbB-receptor interaction and activation. ExpCell Res 2005;304:604–619.

15. Nagy P, Jenei A, Damjanovich S, Jovin TM, Szolosi J. Complexity of sig-nal transduction mediated by ErbB2: clues to the potential of recep-tor-targeted cancer therapy. Pathol Oncol Res 1999;5:255–271.

16. Nagy P, Vereb G, Sebestyen Z, Horvath G, Lockett SJ, Damjanovich S,Park JW, Jovin TM, Szollosi J. Lipid rafts and the local density of ErbBproteins influence the biological role of homo- and heteroassociationsof ErbB2. J Cell Sci 2002;115:4251–4262.

17. Szollosi J, Nagy P, Sebestyen Z, Damjanovich S, Park JW, Matyus L.Applications of fluorescence resonance energy transfer for mappingbiological membranes. J Biotechnol 2002;82:251–266.

18. Vereb G, Szollosi J, Matko J, Nagy P, Farkas T, Vigh L, Matyus L, Wald-mann TA, Damjanovich S. Dynamic, yet structured: the cell mem-brane three decades after the Singer-Nicolson model. Proc Natl AcadSci USA 2003;100:8053–8058.

19. Bagossi P, Horvath G, Vereb G, Szollosi J, Tozser J. Molecular modelingof nearly full-length ErbB2 receptor. Biophys J 2005;88:1354–1363.

20. Gaspar R Jr, Bagossi P, Bene L, Matko J, Szollosi J, Tozser J, Fesus L,Waldmann TA, Damjanovich S. Clustering of class I HLA oligomerswith CD8 and TCR: three-dimensional models based on fluorescenceresonance energy transfer and crystallographic data. J Immunol 2001;166:5078–5086.

21. Szentesi G, Horvath G, Bori I, Vamosi G, Szollosi J, Gaspar R, Damjano-vich S, Jenei A, Matyus L. Computer program for determining fluores-cence resonance energy transfer efficiency from flow cytometric dataon a cell-by-cell basis. Comput Methods Programs Biomed 2004;75:201–211.

22. Jovin TM, Arndt-Jovin DJ. Digital imaging of fluorescence resonanceenergy transfer. Applications in cell biology. Cell structure and func-tion by microspectrofluorimetry. 1989:99–115.

23. Jovin TM, Arndt-Jovin DJ. FRET microscopy: digital imaging of fluores-cence energy transfer. Application in cell biology. In: Kohen E,PloemJS,Hirschberg JG, editors. Cell structure and function by microspec-trofluorimetry. Orlando: Academic Press; 1989. p 99–117.

24. Young RM, Arnette JK, Roess DA, Barisas BG. Quantitation of fluores-cence energy transfer between cell surface proteins via fluorescencedonor photobleaching kinetics. Biophys J 1994;67:881–888.

25. Bodnar A, Jenei A, Bene L, Damjanovich S, Matko J. Modification ofmembrane cholesterol level affects expression and clustering of classI HLA molecules at the surface of JY human lymphoblasts. ImmunolLett 1996;54:221–226.

26. Bodnar A, Bacso Z, Jenei A, Jovin TM, Edidin M, Damjanovich S,Matko J. Class I HLA oligomerization at the surface of B cells is con-trolled by exogenous beta(2)-microglobulin: implications in activationof cytotoxic T lymphocytes. Int Immunol 2003;15:331–339.

27. Szabo G Jr, Weaver JL, Pine PS, Rao PE, Aszalos A. Cross-linking ofCD4 in a TCR/CD3-juxtaposed inhibitory state: a pFRET study. Bio-phys J 1995;68:1170–1176.

28. Nagy P, Vereb G, Sebesty�en Z, Horv�ath G, Lockett SJ, Damjanovich S,Park JW, Jovin TM, Sz€ollosi J. Lipid rafts and the local density of ErbBproteins influence the biological role of homo- and heteroassociationsof ErbB2. J Cell Sci 2002;115.

29. Bastiaens PIH, Jovin TM. Fluorescence resonance energy transfermicroscopy. Cell biology: a laboratory handbook. Volume 3.New York: Academic Press; 1998. p 136–146.

127PBFRET CALCULATIONS FROM IMAGE SERIES

30. Gadella TWJ, Jovin TM. Fast algorithms for the analysis of single anddouble exponential decay curves with a background term. Applica-tion to time-resolved imaging microscopy. Bioimaging 1997;5:19–39.

31. Nagy P, Vamosi G, Bodnar A, Lockett SJ, Szollosi J. Intensity-basedenergy transfer measurements in digital imaging microscopy. Eur Bio-phys J 1998;27:377–389.

32. Periasamy A. Fluorescence resonance energy transfer microscopy: amini review. J Biomed Opt 2001;6:287–291.

33. Bene L, Szentesi G, Matyus L, Gaspar R, Damjanovich S. Nanoparticleenergy transfer on the cell surface. J Mol Recogn 2005;18:236–253.

34. Sz€ollosi J, Horejsi V, Bene L, Angelisova P, Damjanovich S. Supramole-cular complexes of MHC class I, MHC class II, CD20, and tetraspanmolecules (CD53, CD81, and CD82) at the surface of a B cell line JY.J Immunol 1996;157.

35. Jovin TM, Arndt-Jovin DJ, Marriott G, Clegg RM, bert- Nicoud M, Bes-tyen Z.Distance, wavelength and time: the versatile 3rd dimensions inlight emission microscopy. In: Herman B, Jacobson K, editors. Opticalmicroscopy for biology. New York: Wiley-Liss; 1990. p 575–602.

36. Song L, Hennink EJ, Young IT, Tanke HJ. Photobleaching kinetics offluorescein in quantitative fluorescence microscopy. Biophys J 1995;68:2588–2600.

37. Song L, Varma CA, Verhoeven JW, Tanke HJ. Influence of the tripletexcited state on the photobleaching kinetics of fluorescein in micro-scopy. Biophys J 1996;70:2959–2968.

38. Song L, van Gijlswijk RP, Young IT, Tanke HJ. Influence of fluoro-chrome labeling density on the photobleaching kinetics of fluoresceinin microscopy. Cytometry 1997;27:213–223.

39. William HP, Brian PF, Saul AT, William TV.Numeriacla recipes in C.Cambridge: Cambridge University Press; 2005.

40. Bastiaens PI, Majoul IV, Verveer PJ, Soling HD, Jovin TM. Imaging theintracellular trafficking and state of the AB5 quaternary structure ofcholera toxin. EMBO J 1996;15:4246–4253.

41. Panyi G, Vamosi G, Bacso Z, Bagdany M, Bodnar A, Varga Z, Gaspar R,

Matyus L, Damjanovich S. Kv1.3 potassium channels are localized in

the immunological synapse formed between cytotoxic and target

cells. Proc Natl Acad Sci USA 2004;101:1285–1290.42. Vereb G, Meyer CK, Jovin TM.Novel microscope-based approaches

for the investigation of protein-protein interactions in signal transduc-

tion. In: Heilmeyer LMG Jr, editor. Interacting protein domains, their

role in signal and energy transduction. Volume H102. New York:

Springer-Verlag; 1997. pp. 49–52.

128 SZENTESI ET AL.