seeing the wood through the trees: a review of techniques ......granules almost invariably have...

Analytical Biochemistry 291, 175–197 (2001)doi:10.1006/abio.2000.5006, available online at http://www.idealibrary.com on

REVIEW

Seeing the Wood through theTrees: A Review of Techniques forDistinguishing Green FluorescentProtein from EndogenousAutofluorescence

Nicholas Billinton* and Andrew W. Knight†,1

*Department of Biomolecular Sciences and †Department ofInstrumentation and Analytical Science, University ofManchester Institute of Science and Technology, P.O. Box88, Manchester M60 1QD, United Kingdom

Nicholas Billinton Andrew W. Knight

Published online March 16, 2001

The green fluorescent protein (GFP)2 from the jelly-fish Aequorea victoria has recently leapt to prominencewithin numerous biological fields. Interest in the use ofGFP has grown enormously over the past 5 years. Thenumber of papers concerning GFP held in the Web ofScience database, which abstracts the scientific litera-ture (http://wos.mimas.ac.uk), rose from 13 in the pe-riod 1981–1994, to over 3400 in the period 1995 toOctober 2000. This is due to the ability to clone andheterologously express GFP genes in a diverse range ofcells and organisms, from bacteria and yeast, to plantsand mammals, coupled with favorable properties suchas high stability, minimal toxicity, noninvasive detec-tion, and the ability to generate the highly visible flu-orophore in vivo in the absence of external cofactors.Thus, GFP has become a truly versatile marker forvisualizing physiological processes, monitoring subcel-

1 To whom correspondence and reprint requests should be ad-dressed. Fax: 44-161-200-4911. E-mail: [email protected].

2 Abbreviations used: GFP, green fluorescent protein; FMN, flavinmononucleotide; FAD, flavin adenine dinucleotide; AGEs, advancedglycation end-products; FFI, 2-(2-furoyl)-4(5)-(furanyl)-1H-imida-zole; BFPs, blue fluorescent proteins; FITC, fluorescein isothiocya-nate; LED, light emitting diode; CCD, charge coupled devices; PMT,photomultiplier tubes; RGB, red–green–blue; CLSM, confocal laserscanning microscopy; TPLSM, two-photon laser scanning micros-copy; FLIM, fluorescence lifetime imaging microscopy; CE, capillaryelectrophoresis; LB, Luria-Bertani; FRET, fluorescence resonance

energy transfer; CFP, cyan fluorescent protein; YFP, yellow fluores-cent protein; yEGFP, yeast-enhanced GFP.

0003-2697/01 $35.00Copyright © 2001 by Academic PressAll rights of reproduction in any form reserved.

lular protein localization, distinguishing successfultransfection, or reporting on gene expression. As such apowerful tool, GFP now impinges on almost every areaof biological research.

However, unless GFP is very highly expressed ordensely localized, the fluorescence signal from GFPwill invariably be contaminated with endogenous cel-lular or media fluorescence. Researchers often thinkthey have failed in effectively using GFP as a marker,when actually they have succeeded. Their real problemwas simply the detection of GFP in the sea of otherfluorescent species within biological material, whichfluoresce at a similar wavelength. Cells in most organ-isms exhibit a natural fluorescence, commonly called“autofluorescence,” from a range of species includingmetabolites and structural components. This has beencharacterized in such diverse examples as human epi-dermal cells (1), rat hepatocytes (2), Drosophila mela-nogaster (3), numerous fungi (4–6), and plants such asmaize (7) and alfalfa (8). Many cellular metabolitesexhibit autofluorescence, and since cellular extractsare often crucial components of culture media, suchmedia can also be intensely autofluorescent, com-pounding the problem.

Autofluorescence spectra are generally broad andencompass most of the visible spectral range, overlap-ping the emission wavelengths of green fluorescentprotein and its many derivatives. When measuring

GFP, the presence of autofluorescence often leads to

175

176 BILLINTON AND KNIGHT

low signal-to-noise ratios, restriction of detection sen-sitivity, and in some cases even failure to detect orvisualize GFP at all. Autofluorescence also reduces thecontrast and clarity of GFP labeled structures in fluo-rescence microscope images.

This paper will first review the current opinion onthe source, chemical nature, and spectroscopic charac-teristics of autofluorescent molecules across a widerange of species. Then, for the first time, the variousmethods open to the researcher to overcome the prob-lems of GFP determination in the presence of autofluo-rescence will be systematically and critically reviewed.The methods discussed all contribute to sensitive de-tection, accurate localization, or sharp visualization ofGFP.

The discussion will include instrumentation ap-proaches that exploit differences between GFP andautofluorescent species in terms of their fluorescenceemission wavelength, bandwidth, anisotropy, and life-time. Alternatively, chromatographic and other sepa-ration techniques are included which rely on differ-ences in chemical and physical properties. Also coveredare microscopy methods used to specifically target lo-calized GFP while discriminating against autofluores-cent structures and organelles. In addition, softwaremethods which can be used to remove autofluorescenceand enhance GFP visualization in fluorescence micro-scope images are considered.

Chemical methods for reducing autofluorescence in-clude the careful selection of growth media and pH,and the use of colored dyes which specifically quenchautofluorescence. Photochemical methods include pho-tobleaching, photoconversion of GFP to produce alter-native fluorescence wavelengths, and the use of fluo-rescence resonance energy transfer. Finally,biochemical methods for GFP signal enhancementcover the use of more active promoters, and the choiceof GFP mutants to give brighter fluorescence or alter-native emission wavelengths.

It is hoped that by gaining greater insight into thesources of autofluorescence, and making use of one ormore of the variety of autofluorescence discriminationmethods discussed, the researcher will be able to “seethe wood through the trees.”

POTENTIAL BIOCHEMICAL SOURCES OFAUTOFLUORESCENCE

Flavins

Flavins are ubiquitous coenzymatic redox carriers inthe metabolism of most organisms and as photorecep-tors in plants and fungi. Flavin mononucleotide (FMN)and flavin adenine dinucleotide (FAD) are the mostcommon flavins and are derivatives of riboflavin (vita-min B ). FMN and riboflavin are the most intensely
2fluorescent flavins, with maximal emission at 530 nm,
while FAD is about 10 times less fluorescent. As longago as the 1940s, these properties were being exploitedin the quantification of flavins. In a study of vitamincontent in the malpighian tubes of Periplaneta ameri-cana (American roach), the presence of riboflavin wasascertained and quantified by its “yellow–green” fluo-rescence (9).

Two early studies to identify the sources of intracel-lular autofluorescence provided further important evi-dence that flavoproteins are a major cause. In the first,excitation of various cells with blue light (488 nm)resulted in emission spectra with a fluorescence peakbetween 540 and 560 nm (10). This corresponds to thefluorescence characteristics of flavoproteins which aretypically 10–30 nm red-shifted with respect to the freeflavin. The second study analyzed the fluorescence ofviable mammalian cells and revealed broad excitationmaxima centered around 380 and 460 nm with maxi-mal emission at 520 nm. This was suggested to arisefrom intracellular riboflavin, flavin coenzymes, and fla-voproteins in the mitochondria (11).

Cumulative evidence confirms that flavins excitedwith blue light (450–490 nm) fluoresce with a green–yellow color (500–560 nm), spectroscopic characteris-tics not dissimilar to those of GFP. Table 1 lists exam-ples of autofluorescent cellular components. This listdemonstrates the broad range of cell types exhibitingsuch fluorescence, along with their spectral properties.

NAD(P)H

Nicotinamide-adenine dinucleotide (NAD1/NADH)and nicotinamide-adenine dinucleotide phosphate(NADP1/NADPH), like FAD and FMN, are fluorescentcoenzymes crucial to many reactions in most cell types.Since both NAD1 and NADP1 have equal standardreduction potentials, they shall be referred to asNAD(P)H. The report of van Duijn on examples ofprimary fluorescence for users of fluorescence micros-copy (12) quotes nicotinamide as exhibiting “bluish-white” fluorescence in the free state, but green fluores-cence when observed in the reticuloendothelial systemof phagocytic cells, responsible for the degradation offoreign bodies and material. The blue fluorescence isdependent on the redox state and environment of thecoenzyme. It must be in the reduced state to exhibitfluorescence. Binding to protein cofactors results inenhancement of fluorescence intensity and/or altersthe wavelength of maximal emission, resulting in ablue-shift (13). Maximal excitation of free NAD(P)Hoccurs at ;360 nm with emission at ;460 nm, andthere is a ;20 nm blue-shift for the protein-boundcoenzyme (440 nm). These properties were shown tohold true for baker’s yeast (Saccharomyces cerevisiae),rat liver cells, and Chinese hamster ovary (CHO) cells

(11, 13).

ldeawerbfla

asatot

L

to

F

N

L

177DISTINGUISHING GREEN FLUORESCENT PROTEIN FROM ENDOGENOUS AUTOFLUORESCENCE

Two independent studies of cellular autofluorescencein viable mammalian cells (11) and S. cerevisiae (14)ed to the conclusion that cellular fluorescence could beivided into two parts: (i) excitation at 365 nm withmission at 445 nm, and (ii) excitation maxima at 380nd 440 nm with emission at 520 nm. After comparisonith contemporary literature on the fluorescence prop-rties of known metabolites, these were seen to beepresentative of NAD(P)H and flavins, principally ri-oflavin, respectively. It is also worth noting that theuorescence intensity due to NAD(P)H was between 50nd 100 times that due to flavins (11).Thus far, autofluorescent sources affecting green-

nd blue-emitting fluorescent proteins have been con-idered. However, the broad spectral diversity nowvailable from the ever-expanding range of GFP mu-ants means that sources of background fluorescencether than green-emitters must be considered. A selec-ion of such potential sources is briefly discussed below.

ipofuscins

The term “lipofuscin” refers to a dark lipopigmenthat emits orange–yellow fluorescence. It is something

TAB

Common Biochemical Sources of Autofluorescencwith Their Respective Emission and Excitatio

Autofluorescentsource Organism/tissue

lavins CHO cellsRat hepatocytesBovine and rat neural cellsGoldfish inner earPeriplaneta americana

AD(P)H Rat cardiomyocytesS. cerevisiae, rat hepatocytesCHO cells

ipofuscins Rhesus monkey, rat, human medullaRat heartMuscle, myocardium, hepatocytesHuman brainRat liverRat retina

AGEs Chinese hamster and human lensesIn vitroHuman corneaDiabetic human skin

Collagen and elastin Human aorta and coronary arteryHuman skin

Protoporphyrin IX Human psoriatic skinLignin Phyllostachys pubescens

Pinus radiataChlorophyll Green algae thylakoid

f an enigmatic substance, found to positively stain for

lipid, carbohydrate, and protein, though inconsistentlybetween different cell types and in the hands of differ-ent researchers. Lipofuscin is also called “age pigment”since its strong autofluorescence is seen more fre-quently with increasing age, in the cytoplasm of post-mitotic cells (reviewed in 15). It has been found in cellsfrom a wide range of species and appears as irregularyellow–brown granules of 1–20 mm diameter. Thesegranules almost invariably have single membrane en-velopes.

Lipofuscinogenesis (lipofuscin formation) is age-re-lated and may result from the gradual failure of thecell’s recycling systems (reviewed in 16). Material dam-aged by continual exposure to oxidative and free radi-cal attack accumulates, reacting and cross-linking toform large, heterogeneous fluorescent complexes. Thefluorescence properties of lipofuscin appear to be sub-ject to differences of opinion but most studies agreethat maximal excitation of lipofuscin fluorescence oc-curs at near-visible ultraviolet wavelengths (330–390nm). Autofluorescent emission can vary from blue (420nm) (17) to an orange–yellow in color (540–560 nm)(18), or encompass a wide spectral range (450–650

1

n a Wide Variety of Cell Types and Organisms,axima as Reported in the References Given

Emissionwavelength/nm

Excitationwavelength/nm Reference

520 380, 460 (11)525 468 (2)540–560 488 (10)540 450 (61)“Yellow–green” UV (9);509 ;395 (114)440–470 366 (13, 14)440–450 360 (11)“green” 460–490 (20)550 450–490 (19);540–560 360 (18)481–673 435 (115)430 345 (116).510 390–490 (117)434 365 (24)440–450 370 (22)385 320 (23)440 365 (25).515 476 (26)470–520 442 (27)635 442 (27)UV 470–510 (29)488 ;530 (30)488 685 (740) (31)

LE

e in M

nm). In a study of the relationship between lipofusci-


nogenesis and glutathione concentration, cultured ratcardiomyocytes fluoresced at 550 nm when excited at450–490 nm (19). These wavelengths are sufficientlyclose to those of GFP to cause spectral overlap andhence interference. Furthermore, fluorescence micros-copy of human, monkey, and rat neural tissues dem-onstrated green fluorescence from lipofuscin-likeautofluorescent pigments, using optical filters for fluo-rescein (20), which has almost identical spectroscopicproperties to the commonly used S65T GFP and otherenhanced derivatives.

AGEs

AGEs are advanced glycation end-products, whereglycation refers to nonenzymatic glycosylation, alsoknown as the Maillard reaction. These reactions con-sist of the addition of an aldehyde sugar to an aminogroup, resulting in condensation products known asAmadori products. Amadori products can undergo sub-sequent rearrangements and reactions to form struc-turally diverse complexes that are called AGEs. Theseare better characterized than lipofuscins, the termAGE representing a large family of mostly nitrogenous,cyclic, and conjugated complexes or polymers that maybe free or bound to proteins or other macromolecules.

Several AGEs have been chemically identified. Anexample is 2-(2-furoyl)-4(5)-(furanyl)-1H-imidazole(FFI) (21), which is autofluorescent and has a fluores-cence spectrum [excitation at 370 nm and emission at440–450 nm (22)] similar to that of aged eye lensproteins and collagen. Because of the involvement ofsugars (primarily glucose) in advanced Maillard reac-tions, AGEs are associated with disease states such asdiabetes mellitus. Changes in the cornea of the eye areassociated with diabetes, and these include increasedcorneal autofluorescence. Modification of collagen byAGEs is implicated in this increased fluorescence (320nm excitation and 385 nm emission) (23). A more re-cent study, utilizing more AGE-appropriate wave-lengths and filter sets, also demonstrated that diabeticlenses (hamster and human) are more fluorescent thannondiseased lenses, with a peak excitation of 365 nmand emission at 434 nm (24). The diabetes link wasfurther shown when pentosidine, an AGE, was foundat elevated levels in the skin collagen of long-termdiabetes patients, coinciding with increased fluores-cence (370 nm excitation and 440 nm emission) (25).

Autofluorescence from AGEs is most significant inthe measurement of blue-shifted derivatives of GFP,the so-called blue fluorescent proteins (BFPs) andClontech’s GFPuv (see Table 3). The BFPs exhibitmuch lower quantum yields than their green coun-terparts, resulting in relatively poor signal-to-noise

ratios.

Elastin and Collagen

Autofluorescence of structural components in ani-mals has already been mentioned in relation to the roleof collagen in AGE formation in disease and aging.However, structural proteins like collagen and elastinhave further autofluorescence properties. A study oflaser-excited autofluorescence in the aorta and the cor-onary artery of humans revealed green–yellow fluores-cence from collagen and elastin fibers with excitationat 476 nm (26).

Protoporphyrin IX

Another study on human skin revealed autofluores-cence at wavelengths similar to those for GFP (27). Theskin of psoriatic patients was examined spectrofluoro-metrically, revealing that normal skin showed a broadpeak of fluorescence emission between 470 and 520 nm(excitation at 442 nm). This clearly corresponds withthe peak emission from such GFPs as the S65T deriv-ative and the EGFP, and is presumably a result of theelastin and collagen autofluorescence discussed above.However, while psoriatic skin from the same patientsshowed the same blue–green broad emission, therewas a further distinct peak of emission at 635 nm. Theextra peak in the disease state was shown to be causedby increased levels of protoporphyrin IX, which is anintermediate in the biosynthetic pathway for heme.While 635 nm is a significantly longer wavelength thanthat at which the traditional GFPs emit (;510–520nm), a red fluorescent protein was recently discovered(28). The red fluorescent protein, now marketed byClontech as DsRed, has the most red-shifted spectralproperties of the known fluorescent proteins, with maxi-mal excitation at 558 nm and emission at 583 nm (seeFig. 8). Red autofluorescent sources such as protoporphy-rin IX are likely to be important background componentsin the measurement of DsRed fluorescence.

Lignin

Plants have other autofluorescence problems thatare disruptive to GFP detection. Of course, the obviousproblem is the reflection of visible green light by plants(see chlorophyll section below), ensuring that detectionof GFP fluorescence in weakly expressed systems isvery difficult. However, another problem is lignin, thechief component of woody tissue, which has an associ-ated blue autofluorescence. Attempted detection ofGFP fluorescence in transgenic maize plants revealedblue background autofluorescence that appeared to beless intense in younger leaves (7). This was suggestedto be due to the lower lignin content of the cell walls inyounger samples. However, the blue autofluorescencecan appear green depending upon the environment of

the sample. For example, in work with bamboo, UV

oiathiu

C

PANVVR


excitation resulted in blue fluorescence until sampleswere treated with ammonia, when fluorescence inten-sity increased and changed color to green (the peakshifting from 470 nm to 510 nm) (29). Another study,on Pinus radiata mild compression wood, directly dem-nstrated green fluorescence from lignified tissue afterllumination at 488 nm (argon-ion laser) and with im-ging at 530 nm (30), although the authors suggesthat there may have been significant contribution fromolocellulose at these wavelengths. These two studies

ndicate the spectral changes that can occur by manip-lating the sample environment.

hlorophyll

Chlorophyll (in particular types a and b) is the mostabundant of the photosynthetic chromophores and isrelated to protoporphyrin IX of hemoglobin and myo-globin. Both chlorophylls a and b absorb strongly in thedeep blue and red regions of the visible spectrum andthus green light that is reflected back from the planttissue can mask GFP fluorescence. However, chloro-phyll is also autofluorescent with maximal excitationat 488 nm and maximal emission at 685 nm with abroad shoulder at 740 nm (31). In an early study to useGFP as a vital marker in plants, the Aequorea GFPwas utilized as a marker for gene expression and forFACS analysis in Arabidopsis (32). Bright green fluo-rescence from GFP was detectable in ;50% of trans-fected cells, even in the presence of blue autofluores-cence from the vacuole and red autofluorescence fromthe chloroplasts. However, overlap of GFP fluorescenceand red chlorophyll autofluorescence resulted inpatches of bright yellow fluorescence. Stewart makesthe point in a discussion on monitoring transgenicplants using GFP that GFP expression must be highenough to mask the red fluorescence from chlorophyll(33).

Drosophila

Drosophila warrants a special mention since Plautzand Kay (3, 34) cite autofluorescence as the greatestdifficulty encountered using GFP in the fly. When ex-cited at GFP wavelengths, the cuticle of the fly exhibitsa yellow fluorescence with a peak separated by 50 nmfrom that of GFP. Fluorescence from yolk in vitello-genic and mature oocytes, and young embryos, can alsomake GFP fluorescence difficult to detect (35, 36).

Media Components and Related Molecules

Only aromatic amino acids exhibit significantautofluorescence. Tryptophan fluoresces with thegreatest quantum yield, with maximal excitation at287 nm and emission centered at 348 nm. Tyrosine and

phenylalanine are less fluorescent with excitation at

275 and 260 nm and emission centered at 303 and 282nm, respectively. The tyrosine derivatives, adrenalineand noradrenaline, exhibit spectral properties signifi-cantly red-shifted from tyrosine. Adrenaline is excitedat 330 and 420 nm and exhibits fluorescence at 525 nm,while noradrenaline is excited at 325 and 395 nm withmaximal emission at 515 nm. These are potentiallyimportant for GFP detection in human tissues. In ad-dition, melatonin is a derivative of tryptophan and isexcited at 300 nm with emission at 540 nm, which mayinterfere with the detection of red-shifted GFP mu-tants. Histidine is not naturally fluorescent althoughits imidazole ring is reactive and can form fluorescentproducts. For example, L-histidine can react with N-bromosuccinimide to produce a species that is maxi-mally excited at 400 nm and exhibits fluorescence at520 nm (37).

The adenine biosynthesis mutants in budding yeast,ade1 and ade2, appear pink under normal room light-ing, due to the accumulation of a metabolic intermedi-ate (phosphoribosylaminoimidazole). This pigment in-terferes with GFP detection under FITC filterconditions, since the pink cells appear bright yellow orgreen (38). Addition of supplementary adenine to thegrowth medium reduces the autofluorescent interfer-ence, although ade strains should be avoided wherepossible.

Finally a number of vitamins are autofluorescentdue to their delocalized ring structures. Table 2 illus-trates selected examples of autofluorescent vitaminsand their spectral properties. Environmental condi-tions can significantly alter the spectral properties de-tailed here. However, these effects are beyond thescope of this review and have been discussed elsewhere(37).

INSTRUMENTATION METHODS FORAUTOFLUORESCENCE DISCRIMINATION

Optimized Filter Sets

In the majority of spectroscopic techniques that have

TABLE 2

Selected Autofluorescent Vitamins and Their SpectralCharacteristics (37)

VitaminEmission

wavelength/nmExcitation

wavelength/nm

yridoxine 350–400 300–340scorbic acid 430 350icotinamide 460 360itamin D 470 390itamin A 490 340iboflavin 500–560 450–490

been used to visualize or quantify GFP expressed in

bbJtrcrE

ocic(tFacstwabGetc

ppatrb


cells and organisms, carefully selected optical filtershave been employed to discriminate between GFP andnaturally occurring autofluorescence. Ideally such fil-ters pick out areas of the spectrum such that GFP canbe excited, and its fluorescence transmitted, withgreater efficiency than the autofluorescent species.Due to the remarkably similar spectral characteristicsof fluorescein and common GFP variants, some work-ers have been able to simply use the widely availablefilter sets for the fluorescent probe fluorescein isothio-cyanate (FITC), fitted to most fluorescence micro-scopes. The FITC filter set is suitable for viewing thecommonly used blue excitation enhanced GFPs, suchas S65T and its derivatives, while UV filter sets can beused for viewing wild-type and UV-shifted GFPs suchas the BFPs. However, in general, this approach is onlysuccessful in cases where the GFP to autofluorescenceratio is high, for example by the use of a highly ex-pressed, densely localized or especially bright GFP mu-tant. Since the excitation and emission peaks ofautofluorescent components invariably overlap or lievery close to those of GFP, standard FITC filter setsare often not specific enough to adequately discrimi-nate GFP from autofluorescence.

A typical fluorescence filter set will consist of anexcitation filter, a dichroic beamsplitter, and an emis-sion filter. In a fluorescence microscope these threecomponents are generally combined in a “filter cube” or“block.” Other instrumentation uses similar filters invarious orientations; however, the selection criteriawill be broadly similar. Filter sets dedicated to GFPapplications have recently been made commerciallyavailable. These can be tailored in size to suit mostmicroscopes and filter holders; however, they are rela-tively expensive, costing several hundred pounds perset. Suppliers such as Chroma Technology Corp. andOmega Optical Inc. (both Brattleboro, VT) market over30 different filter sets for GFP, each dedicated to aparticular GFP variant or to overcome a specificautofluorescence problem. This diversity highlights thereasons why most researchers have chosen to optimizetheir own filter sets, often from a combination of dif-ferent suppliers (39): first, because the autofluores-cence arises from a disparate, and often unknown,range of molecules, which vary enormously betweendifferent cells and species and hence is specific to agiven application; and second, because new geneticallyengineered GFP mutants with diverse spectroscopicproperties are constantly being developed.

The choice of excitation filter is dependent primarilyon the light source used and the excitation character-istics of the selected GFP, as well as more specificautofluorescence considerations. Several common lightsources are employed for GFP excitation. The mercuryarc lamp produces a series of useful lines between 400

and 500 nm, whereas the xenon arc lamp and halogen m

lamp produce a relatively broad continuous spectrumin the visible range. For all of these sources a fairlybroad bandpass excitation filter would be the one ofchoice to allow sufficient light transmission for efficientGFP excitation, such as a 475/40 with transmission inthe range 455 to 495 nm. However, for many specimensthe use of a broad bandpass filter may lead to a reduc-tion in overall signal-to-noise ratio, since the increasein autofluorescence will outweigh the increase in theGFP fluorescence signal. In this case, narrower bandexcitation filters may be more usefully employed suchas a 475/20. Excitation of GFP with the intense bluelaser line of the argon ion laser (488 nm), allows theuse of very narrow excitation filters, such as 488/10 or488/3 interference filters. This combination limits theautofluorescence to that which arises only from thosespecies whose excitation spectra significantly overlapthe excitation peak of GFP.

Randers-Eichhorn et al. reported a filter set used forexcitation of GFP in Escherichia coli using a brightlue light-emitting diode (LED) (40). The brightestlue LEDs manufactured by the Nichia Corporation ofapan, have a luminous intensity of 2000 m cd, cen-ered at 470 nm and extending from 430 to 510 nm. Theelatively wide bandwidth of the LED emission canause interference problems and is best used with aelatively narrow bandpass excitation filter, Randers-ichhorn et al. employing a 458/10 interference filter.The choice of emission filter is dependent primarily

n the type of detector employed and the emissionharacteristics of the GFP used. Common detectorsnclude the eye, charge coupled devices (CCD), videoameras, photographic film and photomultiplier tubesPMT). Depending on the application in question, ei-her a bandpass or long-pass emission filter is used.or more qualitative applications where GFP is visu-lized using the eye or detectors capable of recordingolor images, long-pass filters may be used because theharp bright green of GFP may be distinguished fromhe broad spectrum autofluorescence (3). In caseshere only intensity and no color data is recorded, suchs with a PMT or monochrome camera, then narrowandpass filters are employed to distinguish betweenFP and autofluorescence, such as a 510/20 interfer-nce filter. This is especially the case when quantita-ive data is required, such as in fluorimetric or flowytometric methods.Further optimization of standard filter sets may sim-

ly be made by the addition of extra long-pass or band-ass filters. However, as a significant proportion of thevailable fluorescence is lost with each additional fil-er, typically 10 to 30%, sensitivity will be furthereduced. In addition, further fluorescence signal maye lost in the process of collimating the light, since

ost interference and bandpass filters effectively dis-

fihm


criminate wavelength for only light striking the filterat 90° to their surface.

Finally the choice of dichroic beamsplitter used inmicroscope filter cubes depends entirely on spectralproperties of the chosen excitation and emission filters.The dichroic filter is placed at a 45° angle to the incom-ing excitation light, and should ideally reflect greaterthan 90% of the excitation light toward the sample,while transmitting more than 90% of the fluorescenceback to the microscope eyepiece or photodetector. Thecutoff wavelength of the dichroic filter should fall be-tween the transmission windows of the excitation andemission filters thus eliminating cross-talk interfer-ence, and is typically in the range 490–505 nm forGFP.

A diagram of the spectral characteristics of a com-mercial filter set designed especially for researchershaving problems with autofluorescence swampingtheir GFP signal is shown in Fig. 1. The use of nar-rower band filters, in this example to restrict excitationand emission wavelengths, leads however to a muchdimmer image than the standard GFP filter set, and istherefore only recommended where autofluorescence isa particular problem, such as in plant research. Adiscussion of useful filter sets for GFP and a selectionof protocols for the visualization of GFP using variousinstruments and in specific cells and organisms hasbeen documented by Endow and Piston (41).

While it is imperative to optimize the filter set forGFP visualization, the quality of the other elements of

FIG. 1. Transmission spectra of Omega Optical Incorporated GFPlter set XF116, specifically designed for researchers experiencingigh autofluorescence. Adapted from company literature with per-ission of Omega Optical Inc. (Brattleboro, VT).

the microscope should not be neglected. High-quality

objective lenses, internal prisms, light transducers,and other components should be used where possible,so as not to introduce light which may be perceived asan erroneous fluorescence signal, or which may obscurethe fluorescence signal being measured.

Dual-Wavelength Differential Fluorescence Correction

Dual-wavelength differential fluorescence correctionmethods exploit the fact that autofluorescence excita-tion and emission spectra are generally broad with fewfeatures, extending over several hundred nanometers,whereas the excitation and emission spectra of thetarget fluorophore, such as GFP, are relatively narrowand well-defined stretching over typically 50 to 100nm. There are two main approaches in this technique:(i) the use of two excitation wavelengths with fluores-cence measurement at a single wavelength, and (ii) theuse of a single excitation wavelength with fluorescencemeasurement at two wavelengths.

Steinkamp and Stewart reported a method for reduc-ing background autofluorescence in flow cytometry us-ing different lasers to provide two excitation wave-lengths; however, the principles could be applied tomany other experimental methods and instruments(42). Although based on FITC, the method should alsobe suitable for GFP. An argon-ion laser line (488 nm)was used to optimally excite the fluorophore plus theautofluorescent components of the cell, while a kryptonlaser line (413 nm) was used to principally excite onlythe autofluorescence. The fluorescence readings thatresult from excitation by each laser respectively aremeasured in the same wavelength region, and sub-tracted on a cell by cell basis, using a differentialamplifier to zero the autofluorescence and amplify theremaining signal from the fluorophore.

In the flow-cytometric method the two laser beamswere physically spaced 250 mm apart, such that the cellpassed first through one laser beam and then throughthe second, giving two distinct fluorescence pulse sig-nals. Alternatively, a fixed sample could be illuminatedfirst with one laser light source and then the other toobtain the two separate fluorescence measurements.The method assumes that autofluorescence excitationis nearly uniform between the two laser wavelengths.This is often the case since several species with over-lapping fluorescence spectra contribute toward theautofluorescence signal. If however this is not the case,then the input signal amplitudes can be adjusted tobalance the differential amplifier output to zero usingautofluorescent unlabeled control cells. Figure 2 illus-trates the principle of the technique.

One advantage of this technique is that as long asthe two lasers are exciting the same autofluorescentchromophores in the cell, the correction method will

automatically compensate for factors affecting fluores-

42,


cence in general, such as fluctuations in temperatureor pH. In addition, in flow cytometry the correction canbe made on a cell by cell basis, so a mixture of high andlow autofluorescent cells can be analyzed together. Themain disadvantage is the need for an additional nar-row band light source focused to excite the same area ofthe sample, which increases the cost and complexity ofexisting instrumentation. Additional fast processingelectronics are also required for flow-cytometric appli-cations.

These disadvantages are overcome in the conversedual wavelength correction method whereby a single

FIG. 2. Conceptual diagram for illustrating (A) fluorescence excautofluorescence and the 488- and 413-nm laser excitation wavelengexcitation, (C) fluorescence emission of autofluorescence for 413-nbetween the fluorescence emission spectra from 488 and 413 nm, i.efluorescence measurement region, 515–545 nm. Adapted from Ref.Wiley & Sons, Inc.

excitation light source is used, and two different emis-

sion or other wavelength parameters are measured.Roederer and Murphy explored this approach, againfor FITC analysis in flow cytometry, but in such a wayas to be applicable for GFP and in a wide range ofinstrumental formats (43). Excitation of both the flu-orophore and autofluorescence was via a single argon-ion laser line (488 nm). The fluorescence of the fluoro-phore, plus that of the interfering autofluorescentcomponents of the cell, was measured at the optimum530 nm. This measurement was then mathematicallycorrected using software by subtracting the autofluo-rescence component of the signal, which was calculated

ion spectra of fluorescein isothiocyanate (FITC) and clone 9 cell, (B) fluorescence emission of FITC plus autofluorescence for 488 nmaser excitation, and (D) the mathematically calculated differencehe emission spectra for FITC corrected for autofluorescence. Greenreprinted with permission of Wiley–Liss Inc., a subsidiary of John

itatthsm l., t

by correlation to the emission intensity at 625 nm.


Again this method assumes a broad autofluorescenceemission against a narrow emission from the targetfluorophore, which shows very low intensity at thesecond wavelength. The method also assumes that thelight detector responds with equal sensitivity to thetwo fluorescence wavelengths selected. Yet if this is notthe case, software parameters can be adjusted to com-pensate using data from unlabeled control cells. Inter-estingly, the interfering autofluorescent component ofthe analytical signal, measured at 530 nm, can beestimated from a variety of alternative parameters,provided that they are highly correlated with theautofluorescence of unlabeled control cells, and are notsignificantly affected by the fluorescence of the probeitself. Roederer and Murphy found a reasonable corre-lation between side-scattered excitation light at 90°and the autofluorescence intensity. This is assumablydue to a relatively constant cytoplasmic concentrationof autofluorescent cell constituents, and thus autofluo-rescence is directly proportional to cell volume andthus size.

Fluorescence Polarization

A method of discriminating GFP from autofluores-cence based on the observation that GFP producessignificantly polarized fluorescence when excited withplane polarized light has been recently reported by theauthors (44, 45). When a sample containing small flu-orescent molecules is illuminated with plane polarizedlight, those molecules with their electronic transitionmoment aligned parallel to the electric vector of theexcitation light are excited. The subsequent fluores-cence emission will be largely unpolarized, however,since the molecules are free to rotate during the timetaken for the electronic transitions of fluorescence tooccur. That is to say their rotational correlation time ismuch shorter than their fluorescence lifetime, andtheir molecular orientation effectively becomes ran-domized before emission occurs. However, since GFP isa relatively large molecule (27 kDa) and the actualfluorophore element is rigidly encapsulated within thecylindrical structure, it rotates in solution at a slowrate compared to its fluorescence lifetime. Swami-nathan et al. determined the fluorescence lifetime ofGFP to be 2.9 and 2.6 ns in aqueous solution and cellcytoplasm, respectively, an order of magnitude fasterthan the corresponding rotational correlation times of20 and 36 ns (46). By contrast, fluorescein had a rota-tional correlation time of 120 ps, 40 times shorter thanits fluorescence lifetime.

It was found that when GFP, either in free solutionor within whole yeast cells, was excited with planepolarized light from an argon-ion laser, the fluores-cence intensity perpendicularly polarized with respect

to the laser light source (I') was only approximately

50% of that aligned parallel, with respect to the laserlight source (I||). In contrast, for fluorescein, the fluo-rescence intensities measured parallel and perpendic-ular with respect to the polarization of the laser lightsource were approximately equal. Thus the degree ofpolarization (P), defined as (I|| 2 I')/(I|| 1 I'), wasmeasured as 0.398 for GFP and 0.004 for fluorescein(45).

It was speculated in this work that autofluorescencein whole cells and conditioned media may contain asignificant contribution from smaller fluorescent spe-cies, such as flavins and NAD(P)H, which would ex-hibit low fluorescence polarization. To discriminate be-tween GFP and autofluorescence, the measuredfluorescence signal is the difference between I|| and I',which will be large for GFP, much lower for smallerautofluorescent species, and approximately zero forother small green fluorophores such as fluorescein. Themethod was successfully applied to improving thequantification of GFP expressed in naturally autofluo-rescent transgenic yeast strains (45).

Although dedicated fluorescence polarization spec-trometers are commercially available, it is relativelyeasy to adapt existing instrumentation to obtain fluo-rescence polarization measurements. There are severalinstrument configurations that can be adopted. Using asingle fluorescence detector placed behind a polarizingfilter, the plane of polarization of the light source canbe switched between two orthogonal planes to enableI|| and I' measurements to be made. Adjustment of theplane of polarization may be made by use of a rotatingpolarizing filter or a polarizing beamsplitter. Similarly,the polarization of the light source can be fixed andpolarization of light transmitted to the fluorescencedetector may be selected to record I|| and I'. Thus aconventional fluorescence spectrometer can be readilyadapted simply by slotting inexpensive polarizing fil-ters either side of the cuvette or sample holder androtating one through 90° between measurements. Amore robust arrangement which involves no movingparts is shown in Fig. 3 incorporating a fluorescenceflow cell. Two detectors are used, both at right angles tothe path of the laser light source. Using fixed polariz-ing filters, detectors 1 and 2 simultaneously measureI|| and I', respectively. Once assembled the sensitivityof the two detectors is equalized (electronically or usingsoftware) such that I|| 5 I' and P 5 0, using thefluorescence of 8-hydroxypyrene-1,3,6-trisulfonic acidtrisodium salt, which has fluorescence characteristicsvery similar to S65T GFP, while demonstrating ex-tremely low fluorescence polarization (45). The disad-vantage with this approach is the need for an addi-tional fluorescence detector. Proof of principle isdemonstrated is Fig. 4, which shows the response ofboth detectors as fluorescein was added to a culture of

yeast cells expressing GFP. While the magnitude of the

atio

osm


fluorescence signal on each channel increased linearlywith fluorescein concentration, the difference signal(I|| 2 I') due to the GFP present remained constant.Hence, the signal due to fluorescein, used to simulateautofluorescence, was effectively cancelled out.

Provided the interfering species produces fluores-cence that is less polarized than GFP, i.e., has a lowerpolarization ratio, then by taking a difference measure-ment (I|| 2 I'), a greater percentage of the autofluo-rescent signal will be removed from the final measure-ment than that of the signal due to GFP. Hence, anincrease in signal-to-noise ratio will result. The beauty

FIG. 3. Schematic diagram of the fluorescence polariz

FIG. 4. Variation in fluorescence signals as fluorescein is added toa culture of genetically modified yeast cells expressing GFP. (F)Fluorescence signal from the parallel filter detector 1 (I||). (E) Flu-rescence signal from the perpendicular filter detector 2 (I'). (Œ) Theignal for GFP, taken as the difference between these respectiveeasurements (I|| 2 I'). The lines shown are the linear regressions

of the measured fluorescence signal versus fluorescein concentration.

Reproduced from Ref. 45, with permission of the Royal Society ofChemistry.

of this method is that because it relies on a property ofthe light other than wavelength, it can be used todiscriminate GFP from autofluorescence with substan-tially the same spectral properties.

Postmeasurement Image Correction (Use of ImageAnalysis Software)

Fluorescence microscope images captured using dig-ital cameras, or CCD detectors can be manipulatedusing image analysis software to improve the discrim-ination of GFP-laden cells, or labeled cell components,over background autofluorescence. Most commonlyavailable image manipulation software packages willallow simple adjustments of brightness, contrast, andsharpness, for example, that may improve any type ofimage. In addition, the feature of automatic edge de-tection may allow GFP-labeled cell structures to bespecifically highlighted. With a color image, given suf-ficient wavelength separation between GFP andautofluorescence, adjustments such as color balance,hue, and color transformations may also be applied.For example, this may be useful where the pure greenof GFP fluorescence is seen against yellow–greenautofluorescence. The simplest and crudest method toselect green fluorescence from a color image is to con-vert it to red–green–blue (RGB) format, and displayonly the green channel signal. All of these transforma-tions are however somewhat subjective and involvecorruption of the original data. Such transformationsare usually carried out to enable qualitative observa-tions or for aesthetic enhancements.

The authors have used a method of analyzing pixelbrightness to remove autofluorescence contaminatingfluorescence microscope images of GFP-laden yeastcells, which is demonstrated in Figs. 5 and 6. Someimage analysis software packages, such as Corel

n experiment incorporating a flowthrough optical cell.

Photo-Paint, are able to analyze the brightness of in-

bc


dividual pixels in an image. A histogram is producedshowing the number of pixels of each brightness on ascale from 0 to 255, where 0 is black and 255 is purewhite on a monochrome image, or pure bright green onthe green channel of an RGB image. In our application,yeast cells were genetically engineered to express GFPin response to DNA damage repair (47). Fluorescencemicroscope images showed distinct backgroundautofluorescence originating from the growth media,which to some extent obscured the yeast cells. It wasfound that analysis of pixel brightness in areas of theimage containing only autofluorescence showed a par-ticularly narrow distribution in brightness values. Inother words, the intensity of the background autofluo-rescence was quite uniform across the microscope im-age. When an area of the image containing a yeast cellwas analyzed, the same narrow pixel brightness distri-bution of the autofluorescence was still visible, but it

FIG. 5. Two intensity histograms showing pixel brightness distri-utions for images of background autofluorescence and a GM yeastell in the presence of autofluorescence.

was alongside a wider distribution of brighter pixelsthat comprised the image of the yeast cell. Using the

analysis of the pure autofluorescence, a thresholdvalue of brightness could be selected above which theautofluorescence was not visible. In the example shownin Fig. 5, this value was chosen as 115. The softwarecan then remove from the image all pixels which fallbelow the threshold value, and thus the autofluores-cence is stripped from the image leaving the yeast cellsclearly visible, as shown in Fig. 6.

Furthermore some quantitative data can then bederived. The total GFP fluorescence intensity, propor-tional to GFP concentration, can be obtained by inte-grating pixel intensity above the threshold value bypixel count (48). However, if the source of the autofluo-rescence physically overlaps that of the GFP under themicroscope, it will contribute toward the apparent GFPbrightness and will need to be estimated and sub-tracted. Second, there may also be a need to calibratethe brightness measurement with the fluorophore con-centration, rather than assume that a pixel brightnessvalue of 100, for example, represents twice the concen-tration of GFP with a brightness of 50. This could bereadily carried out with fluorescein. Third, the thick-ness of the sample beneath the microscope slide wouldneed to be uniform and reproducible. Dual wavelengthfluorescence correction techniques, if relevant, can alsobe applied to images since the same software will usu-ally allow the subtraction of two images.

Confocal Laser Scanning Microscopy

Confocal laser scanning microscopy (CLSM) is anestablished technique for obtaining high-resolution flu-orescence images and three-dimensional reconstruc-tions of a variety of biological specimens. In CLSM, anexpanded laser beam is reflected by a dichroic mirrorand focused to a small spot within the sample by meansof a microscope objective lens. The fluorescence pro-duced is collected by the same objective and passesthrough the dichroic mirror toward a photomultiplier.A confocal aperture (a pinhole, typically a few hundredmicrometers in diameter) is placed in front of the pho-todetector. This is positioned such that fluorescentlight from points within the specimen that are outsidethe focal plane of the laser, “the out of focus light,” isobscured from the detector. In this way it is possible tocarry out “optical sectioning,” as the laser is swept backand forth across the sample. The analogue light signalfrom the photomultiplier is converted to a digital signaland recorded as a function of laser position in a pixel-based image constructed by a computer. By building upa series of images it is possible to perform depth pro-filing and produce a three-dimensional image of thesample (49, 50).

However, if the autofluorescence occurs at a wave-length similar to that of GFP and if the autofluorescent

species is evenly dispersed throughout the area of in-

cu


terest within the sample, then CLSM is susceptible tothe same problems with autofluorescence as other con-ventional microscopy techniques. The main advantageof CLSM stems from restricting the excitation light tothat material confined within the confocal spot, incases where either the GFP or autofluorescence ishighly localized. Thus, discrete subcellular compart-ments can by specifically analyzed, either by pickingout just the areas of the sample where GFP has beentargeted or by avoiding areas where autofluorescenceis more prevalent. The focused laser could also be usedto photobleach specific autofluorescent areas of thesample. The main drawback of this technique is theinefficient use of excitation light. Only fluorescence

FIG. 6. (A) Fluorescence microscope image of yeast cells expressingmicroscope image after software correction to remove autofluorescenMiyan, Department of Biomolecular Sciences, UMIST.FIG. 8. Excitation and emission spectra for enhanced blue, cyan,demonstrating the diversity of excitation and emission wavelengthswith permission from Clontech Laboratories, Inc. (Palo Alto, CA).

photons which pass back unhindered through the sam-

ple matrix to the detector will be detected. Scatteredphotons, which are often in the majority, are excludedby the confocal aperture. The increase in the intensityof the excitation light needed to compensate for thisloss results in increased photobleaching of the sample,both at the focal spot and along the light path passingthrough the sample matrix.

Two-Photon Laser Scanning Microscopy

The main disadvantage of CLSM, discussed above,can be largely avoided by the use of two-photon laserscanning microscopy (TPLSM). In TPLSM, shortpulses of near-infrared laser light (typically 100 fs at

FP in medium showing autofluorescence. (B) The same fluorescence(see text for details). Photograph courtesy of M. G. Barker and J. A.

een, yellow, and the newly discovered DsRed fluorescent proteins,rrently commercially available in fluorescent proteins. Reproduced

Gce

gr

100 MHz) are focused onto the sample in the same way,


and using essentially the same instrumentation, as forCLSM. Two-photon excitation occurs upon simulta-neous absorption of two separate photons, each havinghalf the energy required to cause transition to theexcited state. For example, 778-nm light is used toproduce excitation at 389 nm, which has been used toimage GFP in mammalian cells (49). The very steepdependence of the absorption rate on light intensity inTPLSM means that excitation and subsequent fluores-cence only occur within the diffraction limited focalspot, where light intensity is high. Virtually no fluo-rescence excitation occurs above or below the focalspot; hence, a confocal pinhole is not required for spa-tial resolution. Most fluorescence photons, whetherscattered on leaving the sample or not, can be collectedand contribute toward the measured signal. Thus thesensitivity of the technique is enhanced and lower in-tensity excitation light can be employed which in turnreduces photobleaching (51). Scattered and reflectedexcitation light is easily removed from the fluorescenceemission due to the large wavelength separation. Thelonger wavelength also penetrates sample tissue morereadily. In addition, any scattered excitation light willbe too weak to promote two-photon excitation of otherfluorescent material outside the area of interest.

This approach does however have some disadvan-tages. The first is the need for a more complex andexpensive pulsed laser light source. The second is thatcommon wavelengths from infrared lasers produce ex-citation at wavelengths significantly lower than theoptimum 490 nm. At such wavelengths, autofluores-cence can increase due to more efficient excitation ofNADH, and photobleaching of GFP has been reportedto be more rapid (49).

Time-Resolved Fluorescence Microscopy

Time-resolved fluorescence spectroscopy is oftenused to quantify fluorescently labeled species in thepresence of nonspecific background autofluorescence.In its simplest form a pulsed light source is used, andthe fluorescence of the labeled species is measuredafter a time interval in which the short-lived autofluo-rescence has decayed. The method is best applied tofluorescent species with especially long fluorescent life-times (.15 ns), such as some transition metal or lan-thanide chelates. Since GFP has a particularly shortfluorescent lifetime (2.6 ns for S65T GFP), this ap-proach may be less suitable; however, its success willdepend on the fluorescence lifetime characteristics ofthe autofluorescence specific to a given application.

The fluorescence lifetime of different GFP mutantsvaries considerably, from 1.3 ns for cyan fluorescentprotein to 3.7 ns for yellow fluorescent protein. Thisvariation has enabled different GFPs to be distin-

guished in multiple labeling applications using time

resolved techniques. Pepperkok et al. (52) used fluores-cence lifetime imaging microscopy (FLIM) to resolvethree coexpressed GFP mutants. In FLIM, the excita-tion light source is a pulsed laser and images are cap-tured using conventional optics with a time-gated CCDor similar detector. A series of time-gated images arecollected between each excitation pulse, and thereforethe decay of fluorescence is recorded simultaneouslyfor each pixel in the field of view. Software algorithmscombine the data from each successive image to deter-mine the fluorescence lifetime for each pixel. Data col-lection and subsequent processing typically take a fewminutes to complete and result in a series of fluores-cent lifetime maps, each highlighting a different rangeof fluorescent lifetimes and thus different species. Pro-vided the average fluorescence lifetime of the GFPmutant used and the autofluorescent species are sig-nificantly different, this approach should enable thetwo to be distinguished. Since the parameter measuredis a spectroscopic property other than wavelength, onlya single dichroic and long-pass filter is required. Thisenables a large proportion of the light emitted to berecorded, enhancing sensitivity, and thus lowering theintensity of the excitation light needed. This in turnreduces photochemical degradation of GFP. The resolv-ing power of the technique can also be increased bycombining the fluorescence lifetime and wavelengthinformation.

The main disadvantage of this technique is the needfor complex and expensive instrumentation, includinga pulsed laser and fast image acquisition and dataprocessing hardware. In addition, the fluorescence life-time of GFP mutants is affected by temperature, fusionto other proteins, and to a lesser extent variations inpH. For example, Pepperkok et al. observed a 400-psdecrease between room temperature (22°C) and 37°C,and a 400- to 500-ps decrease upon fusion to anotherprotein.

Capillary Electrophoresis

Capillary electrophoresis (CE) is a powerful elec-troseparation technique, which is widely used in bio-chemical analysis due to the small sample volumesrequired, rapid analysis time, high resolving power,and the ability to directly inject samples. Molecules tobe separated migrate through a buffer solution underthe influence of an electric field, with mobilities thatdepend on their molecular size and charge. Separationtakes place in a capillary tube, across which a voltageof typically tens of kilovolts is applied. Capillary elec-trophoresis has been used as a method to physicallyseparate GFP from other autofluorescent species inmedia and cellular extracts, and to allow its directquantification. Using CE, GFP is best detected using

laser-induced fluorescence. Fluorescence is collected at

d


90° to the capillary using a microscope objective, cou-pled with a spatial mask and photomultiplier tube.Low limits of detection for both wild-type (53) andS65T GFP (54) have been reported, which are in theorder of 10212 M. With the low injection volume of CE,typically around 20 nl, this limit of detection corre-sponds to approximately 10220 mol, or just a few thou-sand molecules.

For CE, cellular extracts containing GFP must becarefully prepared by laborious procedures includingrepeated washing and lysis of the cells, followed bycentrifugation, filtration, and resuspension steps.However, once prepared, GFP can easily be separatedfrom other autofluorescent species and quantified. Forexample, Fig. 7A is an electropherogram showing theseparation of S65T GFP from autofluorescent mediaimpurities, in the analysis of an extract of humanembryonic kidney (HEK293) cells, from the work ofMalek and Khaledi (54). These workers have also re-ported the quantification of GFP in whole HEK293cells using CE, without the need for cell lysis, as shownin Fig. 7B. In this way, cells can be effectively sepa-rated from autofluorescent media components andthen GFP can be quantified on a cell by cell basis. Theresolving power of CE will also allow separation ofdifferent forms or mutants of GFP, even if their wave-length characteristics are identical, since separationdepends on structural properties. For example, Korf etal. separated the two naturally occurring isoforms ofwild-type GFP in under 5 min using a simple boratebuffer at pH 8.5 (55). Similarly, free and immune-complexed GFP can be resolved.

One disadvantage of the technique is the need tooptimize the buffer conditions, such as ionic strengthand pH, to suit both the electroseparation and the GFPfluorescence, and in the analysis of whole cells, alsomaintain the cell integrity. Interestingly, adverse con-ditions that severely impair the fluorescence of freeGFP have been found not to diminish fluorescence de-tection of immune-complexed GFP (55). However, CEis inherently a disruptive and invasive technique, interms of either the need for cell lysis, or in removingcells from growth media for analysis in a separationbuffer, and it requires expensive instrumentation in-cluding a high voltage power supply and componentsfor laser-induced fluorescence detection.

CHEMICAL/PHOTOCHEMICAL METHODS OFREDUCING AUTOFLUORESCENCE

Choice of Growth Medium

Careful choice of growth medium might seem to bean obvious and simple way of reducing autofluores-cence, but it can be crucial in enabling the detection ofin vivo fluorescence. This was particularly relevant in

the early stages of our own work requiring detection of

g

fluorescence from GFP, expressed in Saccharomycescerevisiae (56). The combination of a weak promoter(from the RAD54 gene) driving expression of an S65T

erivative of GFP and an intensely autofluorescentrowth medium, made in situ GFP detection extremely

FIG. 7. Electropherograms of (A) the separation of S65T GFP fromautofluorescent components in HEK293 cell extract. Buffer: 10 mMTris–HCl, pH 7.5. Applied voltage, 40 kV. (B) The separation of asingle HEK293 cell containing GFP from autofluorescent media com-ponents. Buffer: 10 mM Tris–HCl, pH 8.0. Applied voltage 40 kV.Adapted from Ref. 54.

difficult.


The first step in improving the fluorescence detectionsystem was to remove as much contaminating mediumfluorescence as possible. To achieve good yields of cells,a complex rich medium (YPD, yeast extract, peptone,and dextrose) (57) was initially used. However, thisparticular medium emits intense broad band fluores-cence centered at 511 nm, the wavelength used todetect S65T GFP fluorescence. The combination of poorexpression and weak emission meant that the GFPsignal was completely masked by autofluorescence inthis medium.

A complex minimal medium (SD, synthetic dextrosemedium) (57) was utilized as a lower fluorescent alter-native. SD was found to produce 10-fold less autofluo-rescence at the GFP emission wavelength than the richcounterpart, but still resulted in a poor signal-to-noiseratio. A defined minimal medium, normally employedin fermentation experiments [adapted from (58)], wassubsequently tried. This medium does not contain thehighly fluorescent component of SD, yeast nitrogenbase. The investigator has complete control over theconstituents of the fermentation medium (called F1),allowing it to be refined to specific requirements. Thebasic composition of F1 is a mixture of inorganic salts,trace elements, and vitamins, supplemented with glu-cose and additional nutrients according to auxotrophicrequirement. Thus, the levels of autofluorescent com-ponents can be minimized, resulting in negligibleautofluorescence (600-fold less than the rich medium).

Medium fluorescence is a common problem, oftenmentioned in passing under Materials and Methodssections of publications. For example, one group at-tempting to use GFP as a quantitative reporter ofpromoter activity in E. coli note that LB (Luria-Ber-tani) broth is unsuitable for the liquid culture of GFP-expressing cells due to its “high and variable back-ground fluorescence” (59). One of the constituents ofLB is intensely fluorescent yeast extract. The problemwas overcome by employing M9 minimal medium, adefined mixture of salts supplemented with the appro-priate carbon source and amino acids if required. Afurther method for overcoming this problem is tochange the medium prior to fluorescence measure-ment. Phosphate-buffered saline can be used as a pre-detection washing agent to remove contaminating flu-orescent species and it also behaves as a lowbackground fluorescence detection medium (10, 60).However, it is noteworthy that removal of autofluores-cent growth medium from cells by washing can reducethe longevity of fragile cells on the microscope slide.Hence, observations must be made rapidly before celllysis occurs.

A more involved method was utilized by Aubin (11).Various mammalian cell lines were grown in specificmedia, supplemented with calf serum. Aubin extracted

fluorescent components from the serum using dextran–

charcoal treatment. After mixing and incubating atroom temperature, the charcoal was removed by cen-trifugation and the serum was further purified by mi-crofiltration. This process removed 95% of detectablefluorescent components.

Finally, many growth media contain phenol red as apH indicator to ensure that the growth medium is notcontaminated. This indicator can interfere with GFPfluorescence detection and therefore the use of a phenolred-free medium is suggested.

Selection of low autofluorescent culture mediumfrom the outset or prior to fluorescence detection maynot always be practicable, but where possible, it iscertainly one of the simplest approaches for the reduc-tion of background autofluorescence.

Chemical Treatment

The abrogation of media autofluorescence can stillleave significant disruptive fluorescence from intracel-lular sources or residual species after extraction. Asidefrom the instrumentation methods discussed, there arespecific chemical treatments that can be employed toaid autofluorescence discrimination.

Of particular importance, given their culpability incellular autofluorescence, is the redox state of pyridinedinucleotides and flavins. The pyridine dinucleotidesare primarily fluorescent in the reduced state(NAD(P)H) while the flavins are more fluorescent inthe oxidized state. Flavins can be chemically reducedto lessen their fluorescent contribution. Metcalf re-duced riboflavin by injection of a weak solution of so-dium dithionite, but also noted that the nonfluorescentstate was transient, fluorescence returning slowly withreoxidization by air or rapidly with dilute hydrogenperoxide (9). Dithionite reduction is still used as anassay for the presence and concentration of flavins bymeasuring the change in fluorescence emission at;520 nm before and after reduction (61, 62). However,the application of this method to the reduction ofautofluorescence must come with the caveat of un-known consequences for pyridine dinucleotide fluores-cence. As Schneckenburger et al. point out (14), alter-ing the redox state of one of these two coenzymes willconcomitantly enhance the autofluorescence of theother. Another way of reducing oxidized flavins lies intheir light sensitivity. Irradiation with light in thepresence of electron donors such as EDTA readily re-duces oxidized flavins (63). Further treatments for theinhibition of flavin fluorescence include cyanide andhypoxia (59), although cyanide also reduces the nico-tinamide nucleotides to the fluorescent state (13). Bes-sey and co-workers (64) showed that FAD fluorescencecould be quenched by the presence of purines such as,

adenosine, adenine, and adenylic acid.

(sTtn

bc

t


Lipofuscin is thought to accumulate mainly in post-mitotic cells, so utilization of such cells for fluorescencestudies should be avoided where possible. Schnell et al.20) studied the effect of chemical treatments on tissueections from monkey, rat, and human neural tissue.hey revealed that lipofuscin-like autofluorescence inhese tissues could be significantly reduced or elimi-ated by treatment with 1–10 mM CuSO4 in a 50 mM

ammonium acetate buffer (pH 5) or 1% Sudan black Bin 70% ethanol. The same group investigated the ef-fects of these treatments on immunofluorescent label-ing and found that pretreatment with CuSO4 or Sudanlack B also reduced the intensity of the immunocyto-hemical labeling (20). While CuSO4 reduced immuno-

fluorescence in a dose-dependent manner, its effects onlipofuscin-like autofluorescence were more pro-nounced.

Choice of Fixative

An attractive property of GFP for microscopists is itstolerance to fixatives such as formaldehyde and glutar-aldehyde (65). Retention of GFP fluorescence in suchfixatives allows viewing and localization of GFP inpreserved tissue. However, formaldehyde enhancesautofluorescence from flavins (60) in all cell types andglutaraldehyde has a tendency to yellow with age anddevelops a bluish fluorescence which may interferewith the GFP signal (66). The use of formaldehyde as afixative can diminish the fluorescence of some GFPfusion proteins (67) while others precipitate out of so-lution, forming misleading bright spots throughout thecell (68). The use of either of these fixatives prior toviewing GFP fluorescence means that the cells or tis-sue should be thoroughly washed to remove all tracesof the chemical. GFP is very sensitive to some nailpolishes used to seal coverslips onto microscope slides(65) and so the use of molten agarose or rubber cementis recommended as a sealant. Complete loss of GFPfluorescence is observed in the presence of absoluteethanol (66), making it unsuitable as a fixative andalso implying that GFP fluorescence is unlikely towithstand the complete dehydration required for par-affin or plastic embedding. Mounting and embeddingmedia, also used for sectioning work, can be an addi-tional source of autofluorescence. Those of natural or-igin, such as Canada balsam or glycerin–albumin areamong the worst culprits, and hence the use of one ofmany commercially available, synthetic, low fluores-cence mounting media is recommended. Note that wa-ter-soluble resins are preferable if the mount is to beremoved before viewing with a fluorescence micro-scope, as the use of alcohol will diminish GFP fluores-

cence.

sm

Selection of pH

In studies on rat liver cell membranes (62), it wasshown that while dithionite reduces flavins to a verylow fluorescent state, treatment with hydrogen perox-ide to reoxidize the flavins did not restore 100% fluo-rescence. Dithionite reacts with hydrogen peroxide toproduce sulfuric acid, thus lowering the pH. It was thechange in pH that was responsible for preventing fullrestoration of flavin fluorescence (64, 69). Hence, flavinfluorescence can be pH controlled. This is further cor-roborated by research into eosinophil autofluorescence,in which FAD fluorescence is quantified by the differ-ence between the fluorescence intensity at pH 7.7 and2.6 (70).

However, care should be taken in adjusting the pH tocontrol autofluorescence, as the intensity of GFP fluo-rescence is also pH-dependent. While the wild-typeGFP shows a relatively constant brightness from pH 5to 10, the commonly used S65T mutants are twofoldbrighter at pH 7 than at pH 6 (71).

Purification

If GFP localization or fluorescence at the individualcell level is not a requirement, it is possible to extractGFP for fluorescence quantitation. The high stabilityand fluorescence durability of GFP make it suitable forextraction. GFP fluorescence is only eliminated byharsh conditions including extreme pHs, 6 M guani-dinium chloride at 90°C, or 1% SDS at 65°C (72).

A crude extraction previously employed in this lab-oratory was adapted from the method used by Johnsonand Kolodner (73) and is detailed elsewhere (56). Yeastcells expressing GFP were repeatedly washed in waterfollowed by an “extraction” buffer to remove all tracesof medium fluorescence before resuspension in a“crushing” buffer. Cell lysis was performed by vortex-ing with 0.45-mm diameter glass beads and cell debriswas removed by centrifugation after removal of lysatefrom the beads. The resulting crude extracts were ad-justed to pH 11 with 1 M TRIS-base and then assayedfor GFP fluorescence. After measurement, extractswere treated with 1% SDS at 65°C for 15 min to elim-inate GFP fluorescence. Subsequent fluorescence mea-surement then allowed quantitation of the proportionof autofluorescence in the GFP signal. This crude ex-traction procedure permitted detection of GFP fluores-cence in cells that was otherwise inhibited by autofluo-rescence and absorption effects of the cellular matrixand cell wall.

Traditional protein purification techniques can beused to extract GFP after in vivo expression. De-schamps et al. employed chromatographic techniqueso purify recombinant GFP from an E. coli expressionystem (74). After creating extracts from E. coli, am-

onium sulfate at 40% saturation was added, leaving

iSs

tpsstpsiraeq

A

rfsab

ttsaawGa0wton2hc6(ce

fftaftoodsrtcpm

D

aitpatl

P

Gt


GFP in the supernatant, and then increased to 70%saturation to precipitate GFP. The GFP was solubi-lized in ammonium sulfate for purification on a size-exclusion column, with elution by reduction of the saltconcentration. The GFP concentrate was then passedthrough an ion-exchange column and GFP fractionswere identified by spectroscopic characteristics. Fur-ther traces of contaminants were removed by a finalHPLC size-exclusion step in low-salt buffer.

Purification of polyhistidine-tagged GFP has beendemonstrated using a new purification matrix of ce-ramic hydroxyapatite (Ca5(PO4)3OH) loaded with dif-ferent metal ions (75). Zn(II) was found to be the bestmetal ion for this purpose. The tagged GFP bound tothe column was eluted using either phosphate buffersor EDTA, that competed for phosphate groups in thehydroxyapatite or complexed with the metal ions. Buff-ered extraction procedures involving repeated centrif-ugation were still required to prepare GFP-containingcytoplasmic extracts.

A recent paper demonstrated that organic solventpurification of GFP from E. coli is possible (76). Thesoluble protein fraction was collected after sonicationof cells, followed by removal of nonfluorescent, insolu-ble material. Ammonium sulfate was added to thegreen fluorescent supernatant to a final saturationlevel of 70% before ethanol extraction. GFP was thenextracted back into an aqueous phase from the ethanolextract, using less polar solvents. The GFP content ofthis extract was found to be more than 90% of the totalproteins with a yield of ;60% of the total GFP presentnitially. The final step in the purification was phenyl-epharose chromatography and the resulting GFPhowed the appropriate spectroscopic characteristics.Numerous fusions of cellular proteins or signal pep-

ides with GFP have been made to follow real-timerotein translocations (77–81) and kinetics of proteinecretion (82–84). Secretion of GFP from cells into theurrounding media could be used to enhance its detec-ion since cellular autofluorescence and scatteringroblems are removed. Laukkanen et al. (85) demon-trated that GFP could be secreted from insect cellsnto the culture medium, giving easily detectable fluo-escence. Combined with the utilization of lowutofluorescence culture medium or phosphate-buff-red saline, this could be a sensitive method of GFPuantitation.

utofluorescence Quenching

Fluorescence quenching is a simple in situ method ofeducing nonspecific autofluorescence without the needor additional instrumentation or optics. Microscopeections or whole cells for cytometry are stained withn appropriate dye which, for example, quenches the

road band autofluorescence, while allowing visualiza-

tf

ion of the narrower band GFP fluorescence. For effec-ive quenching the dye used should have an absorbancepectrum that significantly overlaps that of theutofluorescence emission. For example, van der Geestnd Petolino stained transgenic plant tissue sectionsith toluidine blue to allow the detection of wild-typeFP, which was otherwise obscured by cellularutofluorescence (7). Sections were stained for 30 s in.5% toluidine blue, which significantly quenched cellall autofluorescence. Nontransformed, control sec-

ions stained with the dye showed no appreciable flu-rescence. Toluidine blue has a peak absorbance at 626m, while absorbance at 400 and 500 nm is low, at onlyand 12% of the peak respectively. Other dyes which

ave been successfully used for quenching autofluores-ence include methylene blue (peak absorbance 668/09 nm) and trypan blue (peak absorbance 607 nm)86). Fluorescence emission from these compounds oc-urs in the far red region of the spectrum and is thusasily blocked with appropriate filters.Provided a dye with suitable spectroscopic properties

or a given application is available, there are severalurther criteria to fulfill. The dye chosen should be ableo permeate cell membranes, otherwise a cell perme-bilization process is required. A low-molecular-weightacilitates this and also ensures rapid diffusionhroughout the cell. However, since autofluorescence isften concentrated in particular structures or regionsf the cell, the dye also needs to be able to evenlyiffuse into these specific intracellular areas. The dyehould also exhibit low cytotoxicity if it is to be used ineal time in vivo studies. Finally the concentration ofhe dye used must be carefully optimized, as too high aoncentration will significantly quench the GFP fluoro-hore. Finding a dye to satisfy all these criteria is theain disadvantage of this technique.

rosophila

In preparing live eggs for imaging, the highlyutofluorescent chorion can be easily removed by pull-ng on the dorsal appendages of oocytes with fine-ipped forceps or by rolling the embryos gently on aiece of double-sided sticky tape (87). Since theutofluorescence of Drosophila is very much localizedo specific structures, confocal microscopy is particu-arly applicable (see earlier section).

hotochemical Methods (Photobleaching,Photoconversion, and FRET)

Several workers have observed that when viewingFP-labeled cells or organelles under the microscope,

he broad spectrum yellow–green autofluorescence ofhe surrounding media or cellular material tends to

ade with prolonged illumination, and the GFP signal


is then more readily resolved. This is presumably dueto photobleaching of the autofluorescent species.

GFP is relatively resistant to photobleaching, itsphotochemical stability arising from the fact that thefluorophore element of the protein is encapsulatedwithin the barrel-like structure, well shielded fromchemical reactants such as O2. The rate of photobleach-ing of wild-type and S65T GFPs is reported to be lessthan half that of fluorescein (46, 88). This difference inphotobleaching rates between GFP and other fluores-cent species may be exploited to enhance the detectionof GFP in the presence of autofluorescence. Improve-ments may be observed when using a fluorescence mi-croscope, for example, if the image is captured severalseconds after the sample illumination has beenswitched on. However, care should be taken if thisapproach is to be adopted for quantitative work, sinceGFP will also be photobleached to some extent duringthis period, and the rate of photobleaching may varydepending on the presence of modulating species in thesurrounding medium. The use of antifade agents suchas Vectashield (Vector Laboratories, Inc., Burlingame,CA) as mounting media for fluorescence microscopycan reduce the rate of photobleaching and hence in-crease the measurement and integration time avail-able for imaging. This can result in clearer images andimproved GFP signal discrimination.

Zylka and Schnapp have used a photobleachingmethod to reduce interference from media autofluores-cence, thought to originate primarily from riboflavin,when imaging live mammalian cells (39). In order notto unduly illuminate the cells expressing GFP, a smallvolume of growth medium alone was exposed to short-wavelength UV light for 2 min, in a light box contain-ing four 15-W, 300-nm bulbs. Shortly before viewingthe cells, their growth medium was replaced by theUV-treated medium. In addition to photobleachingautofluorescent compounds, this method may, how-ever, destroy media components which are critical forsustaining the cells, or form photochemical derivativesharmful to the cell. The process of switching mediaprior to measurement also necessitates an extra labo-rious step and thus may be less suitable for realisticreal-time monitoring of GFP in situ.

Any method that enables excitation of a particularGFP to occur at an alternative wavelength may beuseful in avoiding areas of the spectrum which prefer-entially excite autofluorescent species. Similarly meth-ods which enable the fluorescence emission of GFP tobe shifted to an alternative wavelength may help toavoid areas of the spectrum where autofluorescence isprimarily occurring. A few such methods exist, how-ever they have yet to be applied specifically to improv-ing autofluorescence discrimination.

Wild-type GFP shows a major absorption/excitation

peak at 395 nm with a lesser peak at 475 nm, and

emission centered at 508 nm. It has been known forsome time that prolonged irradiation of wild-type GFPwith UV or blue light causes a progressive decrease inthe 395 absorption peak, and corresponding increase inthe 475-nm absorption peak, due to photo-inducedisomerism of the molecule (88, 89). This occurs espe-cially with UV irradiation in the range 280–395 nm,but is also observed with blue light up to 490 nm (71).The photoconversion process reverses only slowly overa period of days in the dark (89). Thus, by first irradi-ating wild-type GFP with UV light, the major absorp-tion peak can be effectively moved from the UV to theblue region of the spectrum. The use of 475 nm forexcitation thus reduces autofluorescence from speciessuch as NAD(P)H which primarily absorb in the UV.However, during illumination, photobleaching pro-cesses will also occur causing both 395 and 475 nmabsorption peaks to decline by approximately the sameextent, with a corresponding loss in sensitivity. Blueexcitation-enhanced GFPs, such as S65T and its deriv-atives, do not show the UV absorption peak or thisform of photoisomerism, only slow photobleachingupon illumination with UV or blue light (88).

Elowitz et al. discovered that a variety of GFPs,including wild-type, S65T, and EGFP, can undergo aphotoconversion to a red fluorescent species upon irra-diation with blue light, from an argon-ion laser ormicroscope lamp, under anaerobic conditions (90). Thered fluorescent species thus formed absorbs green lightat 525 nm and fluoresces with a maximum at 600 nm.Thus GFP photoactivated in this way can be viewedusing a rhodamine filter set. Anaerobic conditions canbe brought about by the use of oxygen scavengers suchas a combination of glucose oxidase, catalase, and glu-cose, or occur in microorganisms housed in a sealedenvironment that have exhausted their available oxy-gen. This low oxygen concentration requirement is incontrast to the need for oxygen in the latter stages ofchromophore development after GFP expression. Thered-emitting form of GFP remains stable for at least24 h, but its chemical nature is currently unknown.After exposure to a flash of blue light the red emissiondevelops relatively slowly with a time constant of 0.7 s,suggesting that photoactivation is a two-stage process:irradiation stimulating fast transition to an excitedintermediate, which subsequently decays slowly to thered emitting state, the latter process occurring eventhe dark. Thus by photoactivation in this way both theexcitation and emission characteristics of GFP can beradically altered.

Fluorescence resonance energy transfer (FRET) is aphenomenon whereby a fluorescent molecule, the do-nor, transfers excitation energy in a nonradiative way,usually by dipole–dipole interaction, to a neighboringfluorophore, the acceptor. Energy transfer requires

that the emission spectrum of the donor fluorophore,

tel(g4flwp

dvttfeemGssteaqtog5ameeTgttrn


overlaps that of the absorption spectrum of the accep-tor fluorophore, and that the two fluorophores are invery close proximity (,100 Å apart) with their respec-ive transition moments at an angle other than 90° toach other. Chromophore-mutated GFPs show excel-ent spectral overlap and hence make good FRET pairs91). For example, BFP emits at 466 nm making it aood donor to enhanced GFP (S65T), which absorbs at88 nm (92). A superior example is the pairing of cyanuorescent protein (CFP), with emission at 505 nm,hich is an excellent donor to the yellow fluorescentrotein (YFP), with absorbance at 514 nm (91).The efficiency of the FRET process is highly depen-

ent on the proximity of the fluorophores, being in-ersely related with the sixth power of separation dis-ance. Therefore it has mainly found applications inhe study of protein–protein interactions, protein con-ormation changes, and proteolytic processes. How-ver, FRET could be used to provide an alternativemission or excitation wavelength for a particular GFPutant. The technique would require an additionalFP, i.e., the corresponding FRET partner. Their close

patial proximity can be achieved either by the expres-ion of a fusion protein, with the two GFP mutantsethered by a short spacer of ,20 residues (93), or byxpressing the two GFP mutants separately, each withn appropriate protein tag attached that will subse-uently strongly interact and thus bring the two GFPsogether (92). The excitation and emission wavelengthsf the GFP fusion protein are thus separated to a muchreater extent than in the single GFP, and are 382 and09 nm, respectively, for the BFP–EGFP pair, and 452nd 527 nm, respectively for the CFP–YFP pair. Thisakes the selection and optimization of optical filters

asier and greatly reduces cross-talk interference, i.e.,xcitation light detected by the fluorescence detector.he disadvantages are obviously the need for greaterenetic modification to allow expression of the addi-ional GFP mutant, and that any biochemical effecthat changes the distance between fluorophores or theelative orientation of their transition dipoles will sig-ificantly change the efficiency of the FRET process.

MOLECULAR BIOLOGICAL APPROACHES TO THEENHANCEMENT OF GFP DETECTION

The Use of Spectrally Different GFP Mutants

The cloning of the wild-type GFP-encoding gene (94)from Aequorea victoria provided the means for creatingnew variant fluorescent proteins with altered fluores-cence properties. In 1994, Chalfie et al. (65) showedthat GFP was a genetically encoded reporter with norequirement for an enzymatic substrate or cofactorand, as such, could be expressed in both prokaryoticand eukaryotic cells, thus realizing its potential as a

marker and reporter. Wild-type Aequorea GFP was

originally used as the reporter molecule; however, itsexcitation at 395 nm results in increased backgroundautofluorescence and photoisomerization, both proper-ties that are undesirable in a reporter molecule. Thus,knowledge of the primary structure of GFP was used tocreate mutants with improved characteristics.

Table 3 lists some of the more common commerciallyavailable derivatives of GFP with the most pronouncedspectral alterations. Several specific examples of mu-tations will be discussed below but others are discussedin detail elsewhere (see references in Table 3). Muta-tions in the 20 amino acids around and including thechromophore (residues 55–74) tend to alter the spec-tral characteristics and stability of the chromophore,while other mutations tend to affect protein foldingand solubility. Table 3 and Fig. 8 illustrate the widespectral diversity available from GFP mutants. Thisdiversity allows the investigator to choose the mostappropriate fluorescent protein with excitation andemission wavelengths significantly removed fromthose of the interfering autofluorescence.

Early GFP mutants with altered spectral propertieswere isolated by screening bacterial colonies after per-forming random mutagenesis. Heim et al. (95) isolatedseveral mutants with altered spectral characteristics,of which two of particular interest are discussed here.One mutant was found to have a tyrosine to histidinesubstitution at position 66 (Y66H) that resulted in asignificantly blue-shifted GFP derivative (BFP). Thismutant was maximally excited at 382 nm with peakemission at 448 nm, although its fluorescence intensitywas only 57%, relative to the wild-type GFP. The sec-ond mutant of interest was found to have a tyrosine totryptophan substitution at position 66 (Y66W), result-ing in another blue-shifted derivative, although not assignificantly shifted as the BFP. The Y66W mutantexhibited maximal excitation at 458 nm and emissionat 480 nm and is known as a CFP. Ormo et al. (96) used

TABLE 3

A Selection of GFP Derivatives and Alternative Fluores-cent Proteins with Diverse Spectral Characteristics through-out the Visible Spectrum

Fluorescentprotein

Emissionwavelength (nm)

Excitationwavelength (nm) Reference

BFP/EBFP 440 380 (95)CFP/ECFP 480 458 (95)Wild-type GFP 508 395 (475) (118)EGFP 508 489 (101)GFPuv 509 395 (119)S65T GFP 510 488 (97)yEGFP 515 490 (105)YFP/EYFP 527 513 (96, 104)DsRed 583 558 (28)

a site-directed approach to mutate threonine 203


(T203) which lies adjacent to the chromophore in themature protein. A T203 to tyrosine (T203Y) mutant,which also incorporated two other chromophore muta-tions, had excitation and emission peaks of 513 and527 nm, respectively. These red-shifted spectral char-acteristics were significantly different from other mu-tants and were distinguishable using the appropriatefilter sets on a fluorescence microscope. This new mu-tant was termed a YFP.

Use of Specifically Enhanced GFPs

A key way of increasing the GFP fluorescence toautofluorescence ratio is to increase the fluorescenceintensity of the GFP chromophore itself. Thus, themutation approach has also been applied to enhancingGFP fluorescence intensity. The first breakthrough influorescence intensity-enhancing mutations was madeby Heim et al. (97) after employing site-

seeing the wood through the trees: a review of techniques ......granules almost invariably have...

Documents