REVIEW

Ye Chen . James D. Mills

Ammasi Periasamy

Protein localization in living cells and tissues using FRET and FLIM

Accepted October 30, 2003

Abstract Interacting proteins assemble into molecularmachines that control cellular homeostasis in livingcells. While the in vitro screening methods have theadvantage of providing direct access to the geneticinformation encoding unknown protein partners, theydo not allow direct access to interactions of theseprotein partners in their natural environment inside theliving cell. Using wide-field, confocal, or two-photon(2p) fluorescence resonance energy transfer (FRET)microscopy, this information can be obtained fromliving cells and tissues with nanometer resolution. Oneof the important conditions for FRET to occur is theoverlap of the emission spectrum of the donor with theabsorption spectrum of the acceptor. As a result ofspectral overlap, the FRET signal is always contami-nated by donor emission into the acceptor channel andby the excitation of acceptor molecules by the donorexcitation wavelength. Mathematical algorithms arerequired to correct the spectral bleed-through signal inwide-field, confocal, and two-photon FRET micro-scopy. In contrast, spectral bleed-through is not an issuein FRET/FLIM imaging because only the donorfluorophore lifetime is measured; also, fluorescencelifetime imaging microscopy (FLIM) measurements

are independent of excitation intensity or fluorophoreconcentration. The combination of FRET and FLIMprovides high spatial (nanometer) and temporal (nano-second) resolution when compared to intensity-basedFRET imaging. In this paper, we describe variousFRET microscopy techniques and its application toprotein-protein interactions.

Key words fluorescence resonance energy transfer(FRET) � confocal FRET (C-FRET) � two-photonFRET (2p-FRET) � fluorescence lifetime imagingmicroscopy (FLIM) � FLIM-FRET � CCAATenhancer binding protein a (C/EBPa) � dimerization �traumatic brain injury � tissue FRET � acceptor-photobleaching FLIM-FRET

Introduction

In living cells, proteins interact to achieve cellularhomeostasis. Protein assemblies are traditionally stu-died using biophysical or biochemical methods such asaffinity chromatography or co-immunoprecipitation.Recently, two-hybrid and phage-display methods havebeen used for detecting protein-protein interactions.These in vitro screening methods have the advantage ofproviding direct access to the genetic informationencoding unknown protein partners (Cunningham,2001). These techniques, however, do not allow directaccess to interactions of these protein partners in theirnatural environment inside the living cell. Using FRETmicroscopy, this information can be obtained withnanometer resolution (Clegg, 1996; Gordon et al., 1998;Cubitt et al., 1999; Miyawaki et al., 1999; Ng et al.,1999; Kenworthy et al., 2000; Kraynov et al., 2000;Periasamy, 2001; Day et al., 2003; Jares-Erijman andJovin, 2003).

New imaging technologies, coupled with the devel-opment of new genetically encoded fluorescent labelsand sensors and the increasing capability of computer

Ye Chen � Ammasi PeriasamyW.M. Keck Center for Cellular ImagingUniversity of VirginiaCharlottesville, VA 22904, USA

James D. MillsDepartment of NeurosurgeryUniversity of Virginia Health Sciences CenterCharlottesville, VA 22908, USA

Ammasi Periasamy ( .*)Departments of Biology and Biomedical EngineeringUniversity of VirginiaCharlottesville, VA 22904, USATel: (434) 243-7602Fax: (434) 982-5210e-mail: ap3t@virginia.edu

U.S. Copyright Clearance Center Code Statement: 0301–4681/2003/7109–528 $ 15.00/0

Differentiation (2003) 71:528–541 r International Society of Differentiation 2003

software for image acquisition and analysis, haveenabled more sophisticated studies of protein functionsand processes ranging from gene expression to second-messenger cascades and intercellular signaling (Roesseland Brand, 2002). FRET microscopy relies on theability to capture fluorescent signals from the interac-tions of labeled molecules in single living or fixed cellsand tissues. If FRET occurs, the donor channel signalwill be quenched and the acceptor channel signal will besensitized or increased (Herman, 1998; Herman et al.,2001). By measuring these effects, FRET microscopicimaging can verify close molecular associations betweencolocalized donor- and acceptor-labeled fusion proteinsthat are far beyond the resolution of traditionalfluorescent microcopy.

FRET is a process involving the radiationlesstransfer of energy from a donor fluorophore to anappropriately positioned acceptor fluorophore (Forster,1965; Stryer, 1978; Van Der Meer et al., 1994; Wu andBrand, 1994; Lakowicz, 1999). FRET can occur whenthe emission spectrum of a donor fluorophore signifi-cantly overlaps (430%) the absorption spectrum of anacceptor (Fig. 1), provided dipoles of the donor andacceptor fluorophores are in favorable mutual orienta-tion. Because the efficiency of energy transfer variesinversely with the sixth power of the distance separatingthe donor and acceptor fluorophores, the distance overwhich FRET can occur is limited to between 1 and10 nm. When the spectral, dipole orientation, anddistance criteria are satisfied, excitation of the donorfluorophore results in sensitized fluorescence emissionfrom the acceptor, indicating that the tagged proteinsare separated by o10 nm.

Several FRET techniques exist based on wide-field,confocal, and 2p microscopy as well as FRET/FLIM,each with its own advantages and disadvantages. They

are used for various biological applications such asstudies of organelle structure, conjugated antibodies,cytochemical identification, and oxidative metabolism(Sekar and Periasamy, 2003). This paper describes andcompares the above microscopy techniques, usingdimerization of a transcription factor expressed asfusion proteins with CFP and YFP to demonstrate theutility of the various approaches. Also, we discuss theimplementation of confocal and two-photon FRET intissue in an in vivo model of neurological disease usingaldehyde-fixed rat brain tissue sections.

Different FRET techniques

Different intensity-based imaging techniques that canutilize the FRET method include wide-field, confocal,and two-photon microscopy (Periasamy et al., 2001).All FRET microscopy systems require neutral densityfilters to control the excitation light intensity, a stableexcitation light source (Hg or Xe or combination arclamp; UV, visible, or infrared lasers), a heated stage or achamber to maintain the cell viability, and appropriatefilter sets (excitation, emission, and dichroic) for theselected fluorophore pair. Very important are goodsensitivity detectors, filters, and objective lenses. High-sensitivity detectors help to reduce data acquisitiontime, and narrow band pass filters for excitation andemission reduce the spectral bleed-through noise. Thesecarefully selected filter combinations also improve thesignal-to-noise (S/N) ratio for the FRET signals.Basically, the FRET signal is created by exciting thedouble-labeled (donor and acceptor) protein(s) ofinterest within the cell with the donor excitationwavelength and collecting both donor emission andacceptor emission. When all the conditions for FRETare satisfied, the donor signal will be quenched due tothe transfer of energy and the acceptor signal will besensitized (increased). The acceptor emission is theFRET signal, which is usually contaminated by spectralbleed-through, which should be corrected as explainedin the FRET data analysis section.

Wide-field FRET (W-FRET) microscopy

Any fluorescence microscope (inverted or upright) canbe converted to W-FRET microscopy. There are anumber of papers in the literature for various proteinstudies using the W-FRET system (Jovin and Arndt-Jovin, 1989; Kam et al., 1995; Day, 1998; Gordon et al.,1998; Varma and Mayor, 1998; Periasamy and Day,1999; Kraynov et al., 2000; Day et al., 2003). For W-FRET, it is advisable to use a single dichroic to acquirethe donor (D) and acceptor (A) images for the donorexcitation wavelength in the double-labeled specimen.This can be achieved by using excitation and emission

Fig. 1 Absorption and emission spectra of CFP-YFP fluorophores.DSBT, donor spectral bleed-through into the acceptor (or FRETchannel); ASBT, acceptor molecule spectral bleed-through signaldue to the acceptor being excited by the donor wavelength. Wide-field donor excitation (ExD) 440/21 nm, emission (EmD) 480/30 nm; acceptor (ExA) 520/15 nm, EmA 535/26 nm. Confocal ExD457, EmD 485/30 nm; ExA 514, EmA 528/50 nm.

529

filter wheels in the microscope system. This option helpsto reduce any spatial shift of donor and acceptorchannel images, because the processed FRET image isobtained through pixel-by-pixel calculation as describedin the FRET data analysis section.

Even though W-FRET microscopy is the simplestand most widely used technique, there is a majorlimitation to W-FRET in that the emission signalsoriginating from above and below the focal planecontribute to out-of-focus signals that reduce thecontrast and seriously degrade the image. Digitaldeconvolution microscopy in the W-FRET systemhelps to localize the proteins at different opticalsections, but this requires an intensive computationalprocess to remove the out-of-focus information fromthe optical sectioned FRET images (Periasamy andDay, 1998, 1999). For protein interactions taking placehomogeneously over a wider area of a cell (e.g.,nucleus), as described in this paper, W-FRET is anentirely suitable technique.

Confocal FRET (C-FRET) microscopy

Laser scanning confocal FRET (C-FRET) microscopyovercomes the limitation of out-of focus informationowing to its capability of rejecting signals from outsidethe focal plane and acquiring the signal in real time(Kenworthy et al., 2000; Pozo et al., 2002; Wallrabe et al.,2003). This capability provides a significant improvementin lateral resolution and allows the use of serial opticalsectioning of the living specimen (Pawley, 1995; Le-masters et al., 2001). By selecting appropriate filtercombinations, one can configure any commerciallyavailable confocal microscopy system for FRET imaging.A disadvantage of this technique is that the wavelengthsavailable for excitation of different fluorophore pairs arelimited to standard laser lines. Standard laser lines doallow C-FRET to be used for a number of fluorophorecombinations including CFP-YFP or ds-RED, GFP-rhodamine or Cy3, FITC or Alexa488-Cy3, Alexa488-Alexa555, and Cy3-Cy5 (Kenworthy et al., 2000;Periasamy, 2001; Day et al., 2003; Elangovan et al.,2003; Mills et al., 2003; Wallrabe et al., 2003).

Also, in one-photon wide-field or confocal microscopy,illumination occurs throughout the excitation beam path,in an hourglass-shaped pattern. This results in absorptionalong the excitation beam path, giving rise to substantialfluorescence emission both below and above the focalplane. Excitation of other focal planes contributes tophotobleaching and photodamage in the specimen planesthat are not being involved in imaging. This can beameliorated by multi-photon/2-photon microscopy.

Two-photon FRET (2p-FRET) microscopy

As mentioned above, the advantage of C-FRET overW-FRET lies in the ability to reject the out-of-focus

signal that originates from outside the focal plane. Asignificant improvement over W-FRET and C-FRET isachieved by eliminating the above/below-focal-planesignal altogether, by limiting excitation to only thefluorophores at the focal plane. This is precisely whattwo-photon excitation microscopy does (Denk et al.,1990; Periasamy et al., 1999). The fluorophores exhibittwo-photon absorption at approximately twice (740 nm)their one-photon absorption wavelengths (370 nm),while the emission for below-focal-plane (BFP) labeledspecimen is the same as that of one-photon (420 nm),allowing the specimen to be imaged in the visiblespectrum. When an infrared laser beam is focused on aspecimen, illumination takes place at a single point andthe fluorescence emission is localized to the vicinity ofthe focal point. The fluorescence intensity then falls offrapidly in the lateral and axial direction. The infraredillumination in two-photon excitation penetrates deeperinto the specimen than visible light excitation due to itshigher energy, making it ideal for many applicationsinvolving depth penetration through thick sections oftissue.

Two-photon absorption was theoretically predictedby Goppert-Mayer (1931) and was experimentallyobserved for the first time in 1961 using a ruby laseras the light source (Kaiser and Garrett, 1961). Denk andothers have experimentally demonstrated two-photonimaging in laser scanning confocal microscopy (Denket al., 1990). Two-photon excitation microscopy hasbeen widely used in the area of biomedical sciencesincluding tissue engineering, protein-protein interac-tions, cell, neuron, molecular, and developmentalbiology (Denk et al., 1995; Periasamy et al., 1999;Svoboda et al., 1999; Periasamy, 2001; Diaspro, 2002;Konig and Riemann, 2003; Samkoe and Cramb, 2003;Soeller et al., 2003). Because pulsed lasers are used as anexcitation source, this configuration is an ideal systemfor fluorescence lifetime imaging (FLIM).

In 2p-FRET microscopy, the excitation process isdifferent compared to 1p-FRET microscopy (W-FRETand C-FRET), while the emission process is the same.The 2p-system is suitable for any FRET fluorophorepair because wavelengths are tunable from 700 to1000 nm. The emission spectrum is well separated fromthe excitation spectrum when compared to 1p-FRETmicroscopy. In the case of CFP-tagged molecules, theemission spectrum separation for 1p is about 40 nm (Ex-440 nm and Em-480 nm), compared to 340 nm (Ex-820 nm and Em-480 nm) in the 2p-FRET systems. Thedisadvantage is that a single wavelength may excite thedonor and acceptor molecule at different rates. Forexample, at 820 nm, the CFP molecule is excitedmaximally but the YFP molecule is excited minimally.It is important to remove the YFP molecule fluores-cence signal excited by the CFP excitation wavelength(820 nm) from the FRET channel, as described in theFRET data analysis section. Two-photon-FRET micro-

530

scopy is also advantageous for time-lapse FRETimaging because the cells can be maintained for alonger duration of time compared to the 1p-FRETmicroscopy systems, as there is less photodamage aboveand below the focal plane.

FRET data analysis

In principle, the spectral bleed-through (SBT) signal isthe same for 1p- or 2p-FRET microscopy. In additionto SBT, the FRET signals in the acceptor channel alsorequire correction for spectral sensitivity variations indonor and acceptor channels, autofluorescence, anddetector and optical noise, which contaminate theFRET signal. The details of the corrections and therelevant biological applications have been reported inthe literature (Elangovan et al., 2003; Mills et al., 2003;Wallrabe et al., 2003; www.circusoft.com, accessedNovember, 2003).

In brief, to remove the SBT or cross-talk for 1p- or2p-FRET, seven images are acquired: double-labeled(three images: donor excitation/donor and acceptorchannel; acceptor excitation/acceptor channel), single-labeled donor (two images: donor excitation/donor andacceptor channel), and single-labeled acceptor (twoimages: donor excitation/acceptor channel; acceptorexcitation/acceptor channel) with appropriate filters forPFRET data analysis as described in the literature(Elangovan et al., 2003). Our approach works on theassumption that the double-labeled cells and single-labeled donor and acceptor cells, imaged under thesame conditions, exhibit the same SBT dynamics. Thehurdle we had to overcome was the fact that we hadthree different cells (D, A, and D1A), where individualpixel locations cannot be compared. What could becompared, however, were pixels with matching fluores-cence levels. Our algorithm follows fluorescence levelspixel-by-pixel to establish the level of SBT in the single-labeled cells, and then applies these values as acorrection factor to the appropriate matching pixels ofthe double-labeled cell.

Then, the PFRET, the contamination-removedFRET signal, is (Elangovan et al., 2003)

PFRET ¼ uFRET�DSBT�ASBT ð1Þ

where uFRET is uncorrected FRET (signal in theFRET/acceptor channel), DSBT is donor spectralbleed-through in the FRET/acceptor channel, andASBT is the acceptor bleed-through signal due toexcitation of the acceptor molecule by the donorwavelength (see Fig. 1).

Conventionally, energy transfer efficiency (E) iscalculated by ratioing the donor image in the presence(IDA) and absence (ID) of acceptor. To execute thiscalculation, either the acceptor in the double-labeled

specimen has to be bleached or the donor fluorescenceaverages of two different cells (single and double label)with most likely different dynamics are used in theefficiency calculation. When using the algorithm asdescribed, we indirectly obtained the ID image by usingthe PFRET image (Elangovan et al., 2003). Thesensitized emission in the acceptor channel is due tothe quenching of the donor or energy transferred signalfrom the donor molecule in the presence of acceptor.Therefore, if we add the PFRET to the intensity of thedonor in the presence of acceptor, we obtain the ID.This ID is from the same cell used to obtain the IDA.Hence, the efficiency equation will be modified to obtainthe new transfer efficiency (En) from the same cell, asshown in Equation 3.

E ¼ 1� ðIDA=IDÞ ð2Þ

En ¼ 1� ½IDA=ðIDAþPFRETÞ� ð3Þ

where

ID¼ IDAþPFRET ð4Þ

It is important to note that there are a number of otherprocesses involved in the excited state during energytransfer. The new efficiency (En) is calculated bygenerating a new ID image by including the detectorspectral sensitivity of donor and acceptor channels andthe donor quantum yield with PFRET signal as shownin Equation 5 (Elangovan et al., 2003).

En ¼ 1� fIDA=½IDA þ PFRET�ðCdd=CaaÞ�Qd�g ð5Þ

where

ðCdd=CaaÞ ¼ ½ðPMT gain of donor channel=

PMT gain of acceptor channelÞ � ðspectralsensitivity of donor channel=spectral sensitivity of

acceptor channelÞ� ð6ÞQd5donor quantum yield

In Equation 7, for estimating the distance betweendonor and acceptor, r has changed to rn. Forster’s distanceR0 value was calculated for various fluorophore pairs

r ¼ R0fð1=EÞ � 1g1=6 ð7Þrn¼ R0fð1=EnÞ � 1g1=6 ð8Þ

The energy transfer efficiency was calculated andcompared for the conventional (two different cells, IDA/ID) and new method (same cell, IDA/[IDA1PFRET�(cdd/caa)� Qd]. The error in efficiency between the samecells versus different cells was about 43%. In the sameway, we compared the distance between the donor andacceptor molecule and the error was about 8%(Elangovan et al., 2003). It is important to note thatwe did not consider the quantum yield of the acceptor

531

molecule in our calculation because the energy transferis from donor to acceptor. Also, the self-interactiondominates over the FRET signal in the case ofoverexpressed cells.

Fluorescence lifetime imaging FRET (FLIM-FRET)microscopy

Each of the fluorescence microscopy techniques de-scribed above uses intensity measurements to revealfluorophore concentration and distribution in the cell.Recent advances in camera sensitivities and resolutionshave improved the capability of these techniques todetect dynamic cellular events. Unfortunately, evenwith these improvements in technology, currentlyavailable fluorescence microscopic techniques do nothave high-speed (os) time resolution to fully character-ize the organization and dynamics of complex cellularstructures. In contrast, time-resolved fluorescence mi-croscopy (or lifetime) allows the measurement ofdynamic events at very high temporal resolution. FLIMwas developed to image and study the environmentalbehavior of the living specimen and to study thedynamic behavior from single cell to single molecule(Gadella et al., 1993; Dowling et al., 1998; Straub andHell, 1998; Hanley et al., 2002). Lifetime measurementis independent of spectral bleed-through, fluorophoreconcentration, or unintended photobleaching.

The fluorescence lifetime (t) is defined as the averagetime that a molecule remains in an excited state prior toreturning to the ground state. In practice, the fluores-cence lifetime is defined as the time in which thefluorescence intensity decays to 1/e of the intensityimmediately following excitation (Lakowicz, 1999).Excited-state lifetime measurements are independentof change in excitation light intensity, probe concentra-tions, and light scattering, but highly dependent on thelocal environment of the fluorophore, such as theoccurrence of FRET, changes in pH and temperature,and the presence/absence of calcium ions. Instrumentalmethods for measuring fluorescence lifetimes aredivided into two major categories, frequency-domain(Gratton et al., 1984, 2003; Lakowicz, 1999) and time-domain (Demas, 1983; O’Connor and Phillips, 1984).In this paper we describe the development of thetime-domain method of data acquisition and proces-sing.

The combination of lifetime and FRET (FLIM-FRET) provides high spatial (nanometer) and temporal(nanosecond) resolution (Bacskai et al., 2003; Elango-van et al., 2002; Krishnan et al., 2003). The presence ofacceptor molecules within the local environment of thedonor that permit energy transfer will influence thefluorescence lifetime of the donor. By measuring thedonor lifetime in the presence and absence of theacceptor, one can accurately calculate the distance

between the donor- and acceptor-labeled proteins.While 1p-FRET produces ‘‘apparent’’ E%, that is,efficiency calculated on the basis of all donors (FRETand non-FRET), the double-label lifetime data in 2p-FLIM-FRET usually exhibit two peaks of donorlifetimes (FRET and non-FRET), allowing a moreprecise estimate of distance based on FRET donorsonly (see Figs. 6,7). The former may be sufficientlyaccurate for many situations, and the latter may be vitalfor establishing comparative distances of several pro-teins from a protein of interest.

The energy transfer efficiency (E), the rate of energytransfer (kT), and the distance between donor andacceptor molecule (r) are calculated using the followingequations (Lakowicz, 1999):

E ¼ 1� ðtDA=tDÞ ð9Þ

kT¼ ð1=tDÞðR0=rÞ6 ð10Þ

r ¼ R0fð1=EÞ � 1g1=6 ð11Þ

R0¼ 0:211fk2n�4QDJðlÞg1=6 ð12Þwhere tD and tDA are the donor excited state lifetime inthe absence and presence of the acceptor; R0 is theForster distance—that is, the distance between thedonor and the acceptor at which half the excitationenergy of the donor is transferred to the acceptor whilethe other half is dissipated by all other processes,including light emission; n is the refractive index; QD isthe quantum yield of the donor; and k2 is a factordescribing the relative dipole orientation (normallyassumed to be 2/3; Lakowicz, 1999).

FRET cellular assay

Transcription factor C/EBPa: FRET microscopywas used to characterize intra-nuclear dimer formationfor the transcription factor C/EBPa in living pituitaryGHFT1-5 cells. Members of the C/EBP family oftranscription factors are critical determinants of celldifferentiation. C/EBPa controls the transcription ofgenes involved in energy, including those encodinganterior pituitary growth hormone (GH) and prolactin(PRL) (Jacob and Stanley, 1999). C/EBPa is a basicregion-leucine zipper (a-zip) transcription factor thatforms dimers through contacts in the leucine zipper andbinds to specific DNA elements via the basic region.Day et al. recently showed that GFP-tagged C/EBPaexpressed in mouse pituitary GHFT1-5 cells waslocalized to sub-nuclear sites associated with pericen-tromeric heterochromatin (Day et al., 2001), and thispattern was identical to that for the endogenous proteinin differentiated mouse adipocytes (Tang and Lane,1999). Studies indicate that the b-zip region of C/EBPa(amino acids 244–358) fused to GFP was sufficient for

532

sub-nuclear targeting of the fusion protein in pituitaryGHFT1-5 cells (Day et al., 2001). Because this regioncontains the dimerization domain, we sought todetermine whether the expressed fusion proteins wereassociated as dimers in these sub-nuclear sites.Tissue FRET pair preparation

Animal surgery: Adult male Sprague-Dawley ratswere subjected to an impact acceleration injury.Specifically, rats weighing between 350 and 400 greceived induction anesthesia, were endotracheallyintubated, and maintained on a modified medicalanesthesia machine. A 3 cm midline incision in thescalp was then made, and periostial membranes wereseparated, exposing bregma and lambda. A metal disc10 mm in diameter and 3 mm thick was attached to theskull with cyanoacrylate and centered between bregmaand lambda. The animal was then placed prone on afoam bed with the metal disk directly under a plexiglasstube. A 450 g brass weight was dropped through thetube from a height of 2 meters striking the disk. Shamsurgery animals underwent an identical procedure withthe exception of the weight striking the metal disk. Sixhours following injury, or 1 hr following sham injury,animals were perfused transcardially, and the brain wasremoved. The brainstem, including the cortico-spinal tracts and the medial lemnisci, was then sagitallycut on a vibratome into 40, 100, or 200 micron thicksections.

All procedures involving live animals were approvedby the Institutional Animal Care and Use Committee ofthe University of Virginia, and were performed accord-ing to the principles of the Guide for the Care and Useof Laboratory Animals, published by the Institute ofLaboratory Resources, National Research Council(NIH publication 85-23-2985).

Immunohistochemistry: The tissue sections were in-cubated in polyclonal antibody raised in rabbit againstBcl-xL amino acids 126–188 (sc-7195; Santa CruzBiotechnology, Santa Cruz, CA) at a dilution of 1:200in 1% normal goat serum (NGS) in PBS overnight at41C. Following incubation in primary antibody, thetissue was washed three times in 1% NGS in PBS, andthen incubated in a secondary anti-rabbit IgG antibodyconjugated with Alexa555 fluorophore (MolecularProbes, Eugene, OR) for 2 hr. The tissue was washedthree times in 1% NGS in PBS, and preincubated in10% NGS for 40 min. The tissue was then incubated inthe second primary antibody, monoclonal anti-BADamino acids 1–168 raised in mouse (sc-8044; Santa CruzBiotechnology), at a dilution of 1:50 in 1% NGS in PBSovernight at 41C. The tissue was again washed threetimes in the 1% NGS solution, and incubated in asecondary anti-mouse IgG antibody conjugated withAlexa488 (Molecular Probes) for 2 hr. The tissueunderwent a final wash in 0.1 M phosphate buffer andwas then mounted using an antifade agent (MolecularProbes) and coverslipped.

Biological processes used for different FRETmicroscopy approaches

There is great interest in studying the dynamics ofprotein molecules under physiological conditions be-cause protein-protein interactions mediate many cellu-lar processes. Identification of the interacting proteinpartners is critical in understanding its function andplace in the biochemical pathway, thereby establishingits role in important disease processes. The microscopytechniques described above allow the study of proteinsin multiple ways, including what proteins are expressed,where they are expressed, and where they move overtime.

FRET microscopy relies on the ability to captureweak and transient fluorescent signals efficiently andrapidly from the interactions of labeled molecules insingle living or fixed cells. FRET microscopy has theadvantage that the spatial distribution of FRETefficiency can be visualized throughout the image,rather than registering only an average over the entirecell or population. Because energy transfer occurs overdistances of 1–10 nm, a FRET signal corresponding to aparticular location within a microscope image providesan additional magnification surpassing the opticalresolution ( � 0.2 mm) of the light microscope. Thus,within a voxel of microscopic resolution, FRETresolves average donor-acceptor distances beyond themicroscopic limit down to the molecular scale (0.001–0.01 mm). This is one of the principal and uniquebenefits of FRET microscopic imaging: not only co-localization of the donor- and acceptor-labeled probescan be seen, but intimate interactions of moleculeslabeled with donor and acceptor can be demonstrated.

Positive and negative control for FRET: Two im-portant aspects have to be observed in order to confirmthe occurrence of FRET in protein-protein interactionsin any biological process: (1) removal of the optical,electrical noises and the spectral bleed-through, and (2)demonstration of positive and negative controls. In thedata presented here, for a positive FRET control,images of cells expressing CFP coupled directly to YFPthrough a 15 amino acid linker (CFP-15aa-YFP) wereused. Using three coverslips (donor [D] and acceptor [A]alone expressed cells; and D1A, expressed cells), sevenimages were acquired using the W-FRET microscopysystem and processed as described in the literature(Day, 1998; Periasamy and Day, 1999; Elangovan et al.,2003). The filter configuration for various FRET pairshas been listed in the literature (Periasamy, 2001; Sekarand Periasamy, 2003; www.kcci.virginia.edu; www.chroma.com; www.omegafilters.com; each accessedNovember, 2003). As shown in Fig. 2, the contamina-tion was removed and the PFRET signal was used tocalculate the efficiency and the distance between thedonor and acceptor molecule. The estimated averagedistance separating the linked CFP and YFP deter-

533

mined using the PFRET image was 62.36 � 2.33 A. Theshortest distance from the fluorophore to the outside ofthe fluorescent protein (FP) b-barrel is 11 A (Rekaset al., 2002), which limits the minimum distance thatcan separate the donor and acceptor fluorophores to� 22 A. The 15 amino acid linker separating the FPswould have a maximum extended chain length of � 50A, but flexibility in its conformation should allow a

range of distances separating the fluorophores at anygiven time. When expressed in the living cell, it isdifficult to predict the conformation that this fusionprotein would adopt, but the FRET results provide anestimate of the spatial relationship between the linkedFPs. For negative control, seven images were collectedas described above. In this case, in cells co-expressingthe unlinked CFP- and YFP-tagged 4-hep C/EBPa,

Fig. 2 Positive FRET control.Images of cells expressing CFPcoupled directly toYFP througha 15 amino acid linker (CFP-15aa-YFP) were acquired andanalyzed using wide-field FRET(W-FRET) microscopy. Excita-tion wavelengths as per Fig. 1.Motorized Olympus IX-70(www.olympusamerica.com, ac-cessed November, 2003) epi-fluorescent microscope equip-ped with Hamamatsu Orca-2CCD camera (www.hamamat-suphotonics.com, accessed No-vember, 2003), and excitationand emission filter wheels (Peri-asamy and Day, 1999). All thehardware was driven by Iseeimaging system software. Objec-tive lens 60� with NA 1.4. (A,B) Single-label donor (CFP),donor excitation (Ex) in thedonor and acceptor channel,respectively. (C, D) Single-labelacceptor (YFP), donor Ex andacceptor Ex, respectively, in theacceptor channel. (E, F) Dou-ble-label (CFP/YFP), donor Ex/donor channel and acceptorchannel (unprocessed FRET),respectively. (G) Double-label,acceptor Ex/acceptor channels.(H, I) Gray level pixel distribu-tion of unprocessed FRET andprocessed FRET (PFRET), re-spectively. PFRET image afterprocessing clearly demonstratesremoval of SBT when comparedto (F). Efficiency: false-coloredrendering of energy transferefficiency.

534

which were co-localized but did not interact, no FRETsignal (or negligible) was observed, as demonstrated inthe histogram (see Fig. 3). As seen elsewhere in dynamicbiological processes, one observes a heterogeneousdistribution of efficiency and distances, also demon-strated in data presented in this paper.

The W-FRET system is simple and widely used, butimages acquired using this system contain out-of-focusinformation, which can be removed by digital deconvo-lution (Periasamy and Day, 1999). In contrast, C-FRETand 2p-FRET provide very sharp images of proteinlocalization because these configurations reject out-of-focus information. In addition, two-photon microscopyoffers advantages over one-photon by causing lessautofluorescence and photobleaching, and being ableto excite any fluorophore using infrared excitationwavelengths from 700 to 1000 nm. Moreover, 2p-FRETmicroscopy is a very effective system for tissue FRETimaging (Mills et al., 2003).

Tissue FRET in traumatic axonal injury: Based onexperiments in other model systems, we hypothesizedthat following traumatic axonal injury, increased levelsof intra-cellular Ca21 should activate calcineurin,resulting in the dephosphorylation and translocationof BAD to the mitochondria, where it binds to andsuppresses the anti-apoptotic protein Bcl-xL, leading tocytochrome c-dependent caspase activation and apop-

tosis (Wang et al., 1999). According to this hypothesis,prior to injury, BAD (donor) and Bcl-xL (acceptor)should remain separate, the distance between themshould be greater than 100 A, and no FRET signalshould be detected. Conversely, following injury, ifthere exists heterodimerization between BAD and Bcl-xL, the distance between them should be less than 100A, resulting in FRET signal detection.

For tissue FRET, we performed a series of pre-liminary immunohistochemical dilution and controlexperiments to determine the appropriate workingantibody concentrations and to verify the absence ofnonspecific or cross-reactive interactions (Mills et al.,2003). Using serial sections of tissue, we simultaneouslylabeled injured tissue with donor only (BAD/Alexa488),acceptor only (Bcl-xL/Alexa555), and with both donorand acceptor. The specimens were then examined foraxons demonstrating vacuolization or formation ofretraction bulbs, morphological characteristics of ax-onal injury. Implementing the C-FRET and 2p-FRETmethodologies as described above in the FRET dataanalysis section, seven images were acquired andprocessed (Elangovan et al., 2003; Mills et al., 2003).In tissue processed from animals receiving a shaminjury, we observed axons with normal morphologiesand ubiquitous labeling with both BAD and Bcl-xL,with no FRET (not shown). Six hours post-injury,

Fig. 3 Negative FRET control.Cells co-expressing unlinked CFPand YFP, which were co-localizedbut did not interact, served as anegative FRET control. Sevenimages were acquired as describedin Fig. 2 and processed. Panelscorrespond to those in Fig. 2. (E,F) Double-label, donor Ex/donorchannel and acceptor channel (un-processed FRET), respectively.Having removed the spectral bleed-through from (F), the PFRETimage clearly demonstrates thatno FRET took place. (H, I) Graylevel pixel distributions and effi-ciency: false-colored image con-firms this fact. The his-togram clearly shows a negligiblesignal in the PFRET image.

535

swollen injured axons continued to be ubiquitouslylabeled for both BAD and Bcl-xL; however, these axonsdemonstrated energy transfer efficiencies approaching30%. Comparing the PFRET images based on C-FRETand 2p-FRET shows that the energy transfer efficiencywas considerably higher in the latter, demonstrating theadvantage of the more sensitive two-photon techniquefor this particular application of localizing proteinheterodimerization in tissue up to 200 microns thick(see Fig. 4). Additionally, the 2p-FRET image hasless background compared to the C-FRET image (seeFig. 4).

Lifetime measurements of green fluorescent proteins(GFPs): FLIM measurements are sensitive to compet-ing environmental and competing physical processes,such as resonance energy transfer and quenching, whichcan alter the fluorescence lifetime; thus, measurementsof fluorescence lifetimes provide a very accuratereflection of the probe’s local environment (Ng et al.,1999; Elangovan et al., 2002; Chen and Periasamy,2004). To measure the lifetime of a fluorophoremolecule, we integrated the Becker & Hickl (Berlin,Germany; www.becker-hickl.de, accessed November,2003) FLIM system with the Bio-Rad Radiance2100confocal/multiphoton microscopy (www.cellscience.bio-rad.com, accessed November, 2003) system (Chenand Periasamy, 2004). A Coherent pulsed Ti:sapphirelaser was used as an excitation source (Santa Clara, CA;www.coherent.com, accessed November, 2003). Thelifetime values were determined at each pixel of theintensity image by the nonlinear least square curve-fitting method (Eliceiri et al., 2003; Chen and Periasa-

my, 2004). The lifetime values are calculated based on asingle exponential decay analysis (for single labeleddonor images) and double exponential decay analysis(for both donor- and acceptor-labeled cells). We usedthe system to measure the lifetime for various mutantforms of green fluorescent proteins (blue, cyan, yellow,and green) fused to C/EBPa protein and expressed inthe GHFT1-5 cell nucleus. As shown in the Fig. 5, thelifetime was obtained using single exponential decayand is found to be different for different color variantsas well as the lifetime distribution. The lifetime distri-bution is a heterogeneous phenomenon in a biologicalsystem due to various environmental variations. Thelifetime is insensitive to photobleaching but is sensitiveto environmental changes such as temperature, pH,calcium signaling, protein-protein interactions, and soon (Sanders et al., 1995; Periasamy et al., 1996;Periasamy, 2001; Chen and Periasamy, 2004), which isvery useful to observe the effect of different experi-mental conditions.

Comparison of C-FRET, 2p-FRET, and FLIM-FRET: We describe here the basic methodology ofextracting and interpreting data using different FRETmicroscopy techniques. As explained in the FRET dataanalysis section, seven images were acquired for C-FRET and 2p-FRET and processed to remove SBT.The resultant data are shown in Fig. 6A–6D. In the caseof FLIM-FRET, the donor lifetimes were measured inthe absence (Fig. 6E) and presence of acceptor mole-cules; if FRET occurs, the lifetime of the donor isexpected to decrease in the presence of acceptors (Fig.6F). Lifetime changes occur in the double-labeled CFP/

Fig. 4 Tissue FRET in traumatic axonal injury. Six hours post-injury; tissue labeled with BAD/Alexa488 (donor) and Bcl-xL/Alexa555 (acceptor) demonstrates energy transfer consistent withBAD-Bcl-xL heterodimerization. The same tissue was used for C-FRET and 2p-FRET imaging using Bio-Rad Radiance2100confocal/multiphoton microscopy. A confocal PFRET image wasobtained at a 40 micron depth and we were unable to achievedecent signal beyond that depth. On the other hand, we obtained

two-photon (or multiphoton) PFRET images up to 200 microns indepth. As shown in the figure, the signal appears to be less when wemoved from 40 to 200 microns. This can be attributed to theconcentration of the fluorophore deep inside the tissue and also tothe visible FRET signal lost in the tissue before reaching thedetector. The scale bar represents 10 mm; 20� MIMM NA 0.75;confocal-ExD 488, EmD 528/30, ExA 543, EmA 590/70; two-photon-ExD 790, ExA 730.

536

YFP-tagged C/EBPa as the labeled molecule dimerizesand FRET takes place. The lifetime data of the donor inthe presence and absence of acceptor are crucial for theefficiency and distance calculation, based on Equations9 and 11, respectively.

As shown in the table in Fig. 6, the efficiency (E) anddistance (r) between the donor and acceptor moleculesare calculated using the respective protein complex inC-FRET and 2p-FRET (Fig. 6C), and mean values

were estimated for about eight protein complexes. Wewould expect the same r values in both cases, but the2p-FRET values were higher than those of C-FRET,which we attribute to the difference in signal acquiredfrom the 2p system. The lifetime distribution is differentfor different cellular environments and in this case mostof the molecules are quenched (Fig. 6F), as shown in asingle peak (compared to Fig. 7), but less than thedonor lifetime in the absence of acceptor. The FLIM-FRET methodology provides an accurate measurementof distance between interacting protein molecules. Thequenched lifetime in Fig. 6F is in the range of 0.7–2.2 ns.The efficiency (E) distribution is in the range of 15%–73%, and the distance between donor and acceptormolecules is in the range of 44.58–70.55 A. A singledonor and acceptor molecule can exist at differentproximities, which can be detected or calculated usingFLIM-FRET techniques. Moreover, it helps to identifymultiple molecular interactions because we are follow-ing the fingerprint of each molecular native fluorescentlifetime change. Multiple protein interactions or dis-tance distribution cannot be measured with C-FRETand 2p-FRET microscopy.

Acceptor-photobleaching FLIM-FRET microsco-py: As described earlier, FLIM-FRET microscopywas used to characterize intra-nuclear dimer formationfor the transcription factor C/EBPa in living pituitaryGHFT1-5 cells. As shown in Fig. 7, the dimerizationreduces donor lifetime at the occurrence of FRET whenenergy transfer quenches the donor, resulting in dif-ferent lifetime distributions compared with non-FRET/unquenched donors. Non-FRET/unquenched donors inthe nucleus are those that did not dimerize as shown bythe distribution in Fig. 7D. The first peak is thequenched donor molecule (t15 1.7 ns), and the secondpeak represents the donor molecule (t25 2.4 ns), whichdid not participate in the energy transfer process. Asmentioned, in the intensity-based method, we speak ofapparent energy transfer efficiency as that calculation isbased on all donor molecules, including those that donot participate in FRET. We have also demonstratedthat when we bleach the acceptor molecule (using514 nm), the quenched molecular peak disappears andall that is left is the lifetime distribution of theunquenched molecule alone (Fig. 7E; t52.5 ns). Thisclearly demonstrates the occurrence of FRET. InFLIM, we can separate the FRET and non-FRETdonors on the basis of lifetime distributions, and webelieve that this is a more realistic measurement. Wehave experienced experimental conditions when onlyone peak appears in the double-labeled specimen,suggesting that all or the majority of donors haveactually participated in energy transfer (Fig. 6F).

Estimation of number of molecules involved in theenergy transfer process: As shown in Figs 6E, 6F, and7, at the occurrence of FRET, energy transfer takesplace and results in extreme quenching of donor

Fig. 5 FLIM images of GFPs. Lifetime images of blue, green,cyan, and yellow fluorescent proteins fused to C/EBPa proteinsexpressed in GHFT1-5 cell nuclei. The lifetime distribution inthe graph clearly demonstrates the heterogeneous nature ofthe cellular environment. Respective mean lifetime of the GFPsis shown. Nikon objective lens 60� oil IR NA1.4 (www.nikonu-sa.com); eBFP-Ex 740 and Em 450/80; eGFP-Ex 880 and Em 515/30; eCFP-Ex 820 and Em 485/30; eYFP-Ex 920 and Em 528/30.Mean lifetime (in ns)� eBFP=1.69; eGFP=2.51; eCFP=2.62;eYFP=2.82.

537

fluorescence and decrease of fluorescence lifetime. Thus,with the measured values of donor lifetime in thepresence and absence of acceptor, the concentration ofthe acceptor species can be determined. Moreover, thefluorescence decay function contains the fluorescence ofquenched and unquenched donor molecules at a pixel,and is therefore double exponential (Fig. 8; graph isfrom the Becker-Hickl processing software). Forexample, if the lifetime of the donor molecule at asingle pixel in the absence and presence of acceptor ist25 t25 2.359 and t15 t15 0.715 ns, respectively, theestimated decay component values are 32.1% (a1) and67.9% (a2), as shown in Fig. 8. It is clear that 32.1% ofthe total donor molecules are quenched at thatparticular pixel. The ratio of quenched to unquenchedmolecules is5 a1/a25 32.1/67.95 0.472. This confirmsthat quenching has occurred. In the unquenched

condition, the ratio would be equal to 1.0. By thismethod, the distribution of quenched and unquenchedmolecules under any given experimental conditionsbecomes available. Thus, with the lifetime measure-ments, complete quantitative characterization of themolecular interactions can be made.

Conclusion

We have demonstrated the feasibility of various FRETmicroscopy techniques to characterize the dimerizationof C/EBPa proteins in the GHFT1-5 cell nucleus anddetecting FRET signals deep inside tissue. Thesemethodologies can be used in any biological systemwith various FRET-pair combinations. A particulartechnique should be chosen depending on the experi-

Fig. 6 Comparison of C-FRET,2p-FRET, and FLIM-FRET.C-FRET and 2p-FRET imagesof the quenched donor (A, C)and PFRET images (B, D) areshown. The respective efficiency(E) and distance (r) are shownin the table below the figure.The distance between donor andacceptor molecules appears tobe higher for 2p-FRET com-pared to C-FRET. This may bedue to the difference in metho-dology of acquisition of pho-tons. Both C-FRET and 2p-FRET signals were collected inthe same cell and optics usingthe Bio-Rad Radiance2100 con-focal/multiphoton microscopysystem. For the same cell, thedonor lifetime images were ac-quired in the absence (E) andpresence (F) of acceptor. Asstated in the text, the naturallifetime of the donor (2.62 ns)was reduced to 1.9 ns (meanvalue) due to FRET. Lifetimemeasurements are the accuratevalues of the distance distribu-tion of the dimerization of C/EBP244 protein molecules inthe mouse pituitary GHFT1-5cell nucleus. The distributionvalues are provided in the tablebelow the figure.

538

mental conditions. It is important to note thatirrespective of the selection of techniques, the FRETsignal is always contaminated, which should becorrected with appropriate software. The alternative isFLIM-FRET microscopy. Even though this techniqueis somewhat more complex, it provides an unprece-dented level of information about the protein molecularassociations under physiological conditions at a very

high temporal and spatial resolution. Moreover, thecomplimentary techniques of fluorescence recoveryafter photobleaching (FRAP) and fluorescence correla-tion spectroscopy (FCS; time and image, not describedhere) provide information about the mobility of theprotein molecules and confirm the occurrence of FRET(Wiseman and Petersen, 1999; Lippincott-Schwartzet al., 2001).

Fig. 7 Acceptor-photobleaching FLIM-FRET microscopy. Accep-tor-photobleaching method to demonstrate the occurrence ofFRET (dimerization of CFP-YFP-C/EBP244 protein moleculesin the mouse pituitary GHFT1-5 cell nucleus). Donor lifetimeimage in the presence (A, D) and absence (B, E) of acceptor, and

(D) and (E) are the respective lifetime distributions. (C) Lifetimedecays at the same pixel before (t15 1.7 ns) and after (t5 2.5 ns)bleaching at 514 nm (100%; 7 min). Nikon objective lens 60 � IRNA1.4 was used, laser average power at Ex 820 nm was 5mW, andthe acquisition time was 35 s. w25 1.4.

Fig. 8 Estimation of energy transferred between molecules. Gra-phical demonstration of the existence of two components of decay(t15 t1 and t25 t2) for the donor molecule in the presence ofacceptor at a pixel. The a1 and a2 represent the quenching (a1)

during FRET and non-quenching (a2) molecules. These lifetimemeasurements allow determination of the interacting and non-interacting protein molecules in a single living cell.

539

Acknowledgments We wish to thank Dr. Richard Day forproviding cells and for his valuable discussion. We would like tothank Mr. Horst Wallrabe for the help and support provided inpreparation of this manuscript. We appreciate Drs. Ty Voss, JamesStone, David Okonkwo, and Gregory Helm for discussion. Weacknowledge the funds supported by the University of Virginia andVirginia Commonwealth Neurotrauma Initiative.

References

Bacskai, B.J., Skoch, J., Hickey, G.A., Allen, R. and Hyman, B.T.(2003) Fluorescence resonance energy transfer determinationsusing multiphoton fluorescence lifetime imaging microscopy tocharacterize amyloid-b plaques. J Biomed Opt 8:368–375.

Chen, Y. and Periasamy, A. (2004) Characterization of two-photonexcitation fluorescence lifetime imaging microscopy for proteinlocalization. Microsc Res Tech, in press.

Cunningham, B.A. (2001) Finding a mate. Scientist 15:26–17.Clegg, R.M. (1996) Fluorescence resonance energy transfer. In:

Wang, X.F. and Herman, B. (eds.) Fluorescence imagingspectroscopy and microscopy. Chemical Analysis Series, JohnWiley & Sons, New York, Vol. 137, pp. 179–251.

Cubitt, A.B., Heim, R., Adams, S.R., Boyd, A.E., Gross, L.A. andTsien, R.Y. (1999) Understanding, improving and using greenfluorescent proteins. Trends Biochem Sci 20:448–455.

Day, R.N. (1998) Visualization of Pit-1 transcription factorinteractions in the living cell nucleus by fluorescence resonanceenergy transfer microscopy. Mol Endocrinol 12:1410–1419.

Day, R.N., Periasamy, A. and Schaufele, F. (2001) Fluorescenceresonance energy transfer microscopy of localized proteininteractions in the living cell nucleus. Methods 25:4–18.

Day, R.N., Voss, T.C., Enwright, J.F. III, Booker, C.F.,Periasamy, A. and Schaufels, F. (2003) Imaging the localizedprotein interactions between Pit-1 and the CCAAT/enhancerbinding protein a (C/EBPa) in the living pituitary cell nucleus.Mol Endocrinol 17:333–345.

Demas, J.N. (1983) Excited state lifetime measurements. AcademicPress, New York.

Denk, W., Strickler, J.H. and Webb, W.W. (1990) Two-photonlaser scanning fluorescence microscopy. Science 248:73–75.

Denk, W., Piston, D.W. and Webb, W.W. (1995) Two-photonmolecular excitation in laser-scanning microscopy. In: Pawley,J.B. (ed.) Handbook of biological confocal microscopy. PlenumPress, New York, pp. 445–458.

Diaspro, A. (2002) Confocal and two-photon microscopy: founda-tions, applications and advances. Wiley-Liss Inc., New York.

Dowling, K., Dayel, M.J., Lever, M.J., French, P.M.W., Hares,J.D. and Dymoke-Bradshaw, A.K.L. (1998) Fluorescence life-time imaging with picosecond resolution for biomedical applica-tions. Opt Lett 23:810–812.

Elangovan, M., Day, R.N. and Periasamy, A. (2002) Nanosecondfluorescence resonance energy transfer-fluorescence lifetimeimaging microscopy to localize the protein interactions in asingle living cell. J Microsc 205:3–14.

Elangovan, M., Wallrabe, H., Chen, Y., Day, R.N., Barroso, M.and Periasamy, A. (2003) Characterization of one- and two-photon excitation resonance energy transfer microscopy. Meth-ods 29:58–73.

Eliceiri, K.W., Fan, C.-H., Lyons, G.E. and White, J.G. (2003)Analysis of histology specimens using lifetime multiphotonmicroscopy. J Biomed Opt 8:376–380.

Forster, T. (1965) Delocalized excitation and excitation transfer. In:Sinanoglu, O. (ed.) Modern quantum chemistry. Academic Press,New York, Vol. 3, pp. 93–137.

Gadella, T.W.J. Jr., Jovin, T.M. and Clegg, R.M. (1993)Fluorescence lifetime imaging microscopy (FLIM)—spatialresolution of microstructures on the nanosecond time-scale.Biophys Chem 48:221–239.

Goppert-Mayer, M. (1931) Ueber elementarakte mit quanten-spreungen. Ann Phys 9:273–295.

Gordon, G.W., Berry, G., Liang, X.H., Levine, B. and Herman, B.(1998) Quantitative fluorescence resonance energy transfer mea-surements using fluorescence microscopy. Biophys J 74:2702–2713.

Gratton, E., Limkeman, M., Lakowicz, J.R., Maliwal, B., Cherek,H. and Laczko, G. (1984) Resolution of mixtures of fluorophoresusing variable-frequency phase and modulation data. Biophys J46:479–486.

Gratton, E., Breusegem, S., Sutin, J., Ruan, Q. and Barry, N.(2003) Fluorescence lifetime imaging for the two-photon micro-scope: time-domain and frequency-domain methods. J BiomedOpt 8:381–390.

Hanley, Q.S., Arndt-Jovin, D.J. and Jovin, T.M. (2002) Spectrallyresolved fluorescence lifetime imaging microscopy. Appl Spec-trosc 56:155–166.

Herman, B. (1998) Fluorescence microscopy. 2nd ed. Springer-Verlag, New York.

Herman, B., Gordon, G., Mahajan, N. and Centonze, V.E. (2001)Measurement of fluorescence resonance energy transfer in theoptical microscope. In: Periasamy, A. (ed.) Methods in cellularimaging. Oxford University Press, New York, pp. 257–272.

Jacob, K.K. and Stanley, F.M. (1999) CCAAT/enhancer-bindingprotein a is a physiological regulator of prolactin geneexpression. Endocrinology 140:4542–4550.

Jares-Erijman, E.A. and Jovin, T.M. (2003) FRET imaging. NatBiotechnol 21:1387–1395.

Jovin, T.M. and Arndt-Jovin, D.J. (1989) Luminescence digitalimaging microscopy. Annu Rev Biophys Chem 18:271–308.

Kaiser, W. and Garrett, C.G.B. (1961) Two-photon excitation inCaF2: Eu

21. Phys Rev Lett 7:229–231.Kam, Z., Volberg, T. and Geiger, B. (1995) Mapping of adherensjunction components using microscopic resonance energy trans-fer imaging. J Cell Sci 108:1051–1062.

Kenworthy, A.K., Petranova, N. and Edidin, M. (2000) High-resolution FRET microscopy of cholera toxin B-subunit andGPI-anchored proteins in cell plasma membranes. Mol Biol Cell11:1645–1655.

Konig, K. and Riemann, I. (2003) High-resolution multiphotontomography of human skin with subcellular spatial resolutionand picosecond time resolution. J Biomed Opt 8:432–439.

Kraynov, V.S., Chamberlain, C., Bokoch, G.M., Schwartz, M.A.,Slabaugh, S. and Hahn, K.M. (2000) Localized Rac activationdynamics visualized in living cells. Science 290:333–337.

Krishnan, R.V., Masuda, A., Centonze, V.E. and Herman, B.(2003) Quantitative imaging of protein-protein interactions bymultiphoton fluorescence lifetime imaging microscopy using astreak camera. J Biomed Opt 8:362–367.

Lakowicz, J.R. (1999) Principles of fluorescence spectroscopy. 2nded. Plenum Press, New York.

Lemasters, J.J., Qian, T., Trollinger, D.R., Muller-Borer, B.J.,Elmore, S.P. and Cascio, W.E. (2001) Laser scanning confocalmicroscopy applied to living cells and tissues. In: Periasamy, A.(ed.) Methods in cellular imaging. Oxford University Press, NewYork, pp. 66–87.

Lippincott-Schwartz, J., Snapp, E. and Kenworthy, A. (2001)Studying protein dynamics in living cells. Nat Rev Mol Cell Biol2:444–456.

Mills, J.D., Stone, J.R., Rubin, J.D., Melon, D.E., Okonkwo, D.O.,Periasamy, A. and Helm, G.A. (2003) Illuminating proteininteractions in tissue using confocal and two-photon excitationfluorescent resonance energy transfer (FRET) microscopy. JBiomed Opt 8:347–356.

Miyawaki, A., Griesbeck, O., Heim, R. and Tsien, R.Y. (1999)Dynamic and quantitative Ca21 measurements using improvedcameleons. Proc Natl Acad Sci USA 96:2135–2140.

Ng, T., Squire, A., Hansra, G., Bornancin, F., Prevostel, C.,Hanby, A., Harris, W., Barnes, D., Schmidt, S., Mellor, H.,Bastiaens, P.I. and Parker, P.J. (1999) Imaging protein kinase Caactivation in cells. Science 283:2085–2089.

540

O’Connor, D.V. and Phillips, D. (1984) Time-correlated singlephoton counting. Academic Press, London.

Pawley, J. (1995) Handbook of biological confocal microscopy. 2nded. Plenum Press, New York.

Periasamy, A. (2001) Methods in cellular imaging. OxfordUniversity Press, New York.

Periasamy, A. and Day, R.N. (1998) FRET imaging of Pit-1protein interactions in living cells. J Biomed Opt 3:154–160.

Periasamy, A. and Day, R.N. (1999) Visualizing protein interac-tions in living cells using digitized GFP imaging and FRETmicroscopy. Methods Cell Biol 58:293–314.

Periasamy, A., Wodnicki, P., Wang, X.F., Kwon, S., Gordon,G.W. and Herman, B. (1996) Time resolved fluorescence lifetimeimaging microscopy using picosecond pulsed tunable dye lasersystem. Rev Sci Instrum 67:3722–3731.

Periasamy, A., Skoglund, P., Noakes, C. and Keller, R. (1999) Anevaluation of two-photon excitation versus confocal and digitaldeconvolution fluorescence microscopy imaging in Xenopusmorphogenesis. Micros Res Tech 47:172–181.

Periasamy, A., Elangovan, M., Wallrabe, H., Demas, J.N.,Barroso, M., Brautigan, D.L. and Day, R.N. (2001) Wide-field,confocal, two-photon and lifetime resonance energy transferimaging microscopy. In: Periasamy, A. (ed.) Methods in cellularimaging. Oxford University Press, New York, pp. 295–308.

Pozo, M.A.D., Kiosses, W.B., Alderson, N.B., Meller, N., Hahn,K.M. and Schwartz, M.A. (2002) Integrins regulate GTP-Raclocalized effector interactions through dissociation of Rho-GDI.Nat Cell Biol 4:232–239.

Rekas, A., Alattia, J.R., Nagai, T., Miyawaki, A. and Ikura, M.(2002) Crystal structure of venus, a yellow fluorescent proteinwith improved maturation and reduced environmental sensitiv-ity. J Biol Chem 277:50573–50578.

Roessel, P.V. and Brand, A.H. (2002) Imaging into the future:visualizing gene expression and protein interactions withfluorescent proteins. Nat Cell Biol 4:E15–E20.

Samkoe, K.S. and Cramb, D.T. (2003) Application of an exovochicken chorioallantoic membrane model for two-photonexcitation photodynamic therapy of age-related macular degen-eration. J Biomed Opt 8:410–417.

Sanders, R., Draaijer, A., Gerritsen, H.C., Houpt, P. and Levine,Y.K. (1995) Quantitative pH imaging in cells using confocal

fluorescence lifetime imaging microscopy. Anal Biochem227:302–308.

Sekar, R.B. and Periasamy, A. (2003) Fluorescence resonanceenergy transfer (FRET) microscopy imaging of live cell proteinlocalization. J Cell Biol 160:629–633.

Soeller, C., Jacobs, M.D., Donaldson, P.J., Cannell, M.B., Jones,K.T. and Ellis-Davies, G.C.R. (2003) Application of two-photonflash photolysis to reveal intercellular communication andintracellular Ca21 movements. J Biomed Opt 8:418–427.

Straub, M. and Hell, S.W. (1998) Fluorescence lifetime three-di-mensional microscopy with picosecond precision using a multi-focal multiphoton microscope. Appl Phys Lett 73:1769–1771.

Stryer, L. (1978) Fluorescence energy transfer as a spectroscopicruler. Annu Rev Biochem 47:819–846.

Svoboda, K., Helmchen, F., Denk, W. and Tank, D.W. (1999)Spread of dendritic excitation in layer 2/3 pyramidal neurons inrat barrel cortex in vivo. Nat Neurosci 2:65–73.

Tang, Q.Q. and Lane, M.D. (1999) Activation and centromericlocalization of CCAAT/enhancer-binding proteins during themitotic clonal expansion of adipocyte differentiation. Genes Dev13:2231–2241.

Van Der Meer, W.B., Coker, G. III and Chen, S.-Y.S. (1994)Resonance energy transfer: theory and data. VCH Publishers,New York.

Varma, R. and Mayor, S. (1998) GPI-anchored proteins areorganized in submicron domains at the cell surface. Nature394:798–801.

Wallrabe, H., Elangovan, M., Burchard, A., Periasamy, A. andBarroso, M. (2003) Confocal FRET microscopy to measureclustering of receptor-ligand complexes in endocytic membranes.Biophys J 85:559–571.

Wang, H.G., Pathan, N., Ethell, I.M., Krajewski, S., Yamaguchi,Y., Shibasaki, F., McKeon, F., Bobo, T., Franke, T.F. and Reed,J.C. (1999) Ca21-induced apoptosis through calcineurin depho-sphorylation of BAD. Science 284:339–343.

Wiseman, P.W. and Petersen, N.O. (1999) Image correlationspectroscopy. II. Optimization for ultrasensitive detection ofpreexisting platelet-derived growth factor-b receptor oligomerson intact cells. Biophys J 76:963–977.

Wu, P. and Brand, L. (1994) Review-resonance energy transfer:methods and applications. Anal Biochem 218:1–13.

541