`
`
`TFS1018
`
` Tse
`ecembertSaco
`x
`“a
`
`
`a pet 22 ee! aa
`
`ublished by the American Physiologica?Society
`
`
`1
`
`1
`
`TFS1018
`
`
`
`Physiological Genomics
`
`Editor-in-Chief: V. DZAU
`
`Deputy Editor: M. J. FINLEY AUSTIN
`
`Senior Editors:
`R. AEBERSOLD
`P.O. BROWN
`
`Associate Editors:
`W.AIRD
`S.ALPER
`P.ANDREWS
`J . DERISI
`
`Editorial Board:
`H. ABURATANI
`M.ALTHERR
`F. BAAS
`\V. BERRETTINI
`\V. BORON
`C. A. R. BOYD
`D. BROOK
`H. BUSSEY
`C. T. CASKEY
`K R. CHIEN
`G. CHURCH
`P. CORVOL
`A. DOIV!INICZAK
`S. FIELDS
`lVI. FISHl\'lAN
`
`A. W. COWLEY, JR.
`D. HOUSMAN
`
`R. !VIULLIGAN
`R. ROSENBERG
`
`M. EISEN
`S.GULLANS
`H. JACOB
`A. J. LINK
`
`J. PELLETIER
`R. PRATT
`L.STAUDT
`L.ZON
`
`J. FRIEDl\·lAN
`A. FUKAM IZU
`D.GANTEN
`W. GUGGINO
`J. HALL
`S. HARRAP
`J. R. HAYWOOD
`I. HUMPHERY-SMITH
`S. JZUMO
`J. KRIEGER
`J.KUNES
`T.KURTZ
`E.S.LAN DER
`B. LEFKOWITZ
`H. LEHRACH
`
`H. LESTER
`H. LEWIN
`M.LOVETT
`H. MCDEVITT
`R. JVJCJNNES
`J. MULLENS
`M.PRINTZ
`J. RAPP
`R.ROMAN
`A. RUT
`L. SAMUELSON
`J. SANES
`C.SAPER
`J.SASSARD
`C. SIGl\'IUND
`
`S. SNYDER
`M. BENTO SOARES
`G. STAJ\JATOYANNOPOULOS
`M. STOLL
`I<. STRANGE
`J. TAKAI-IASI-11
`R. TANZ I
`S.TONEGAWA
`L.-C. TSUI
`T. UNGER
`H. WE ILER
`M.J. WELSH
`M. WHITE
`B. ZABEL
`
`Publications Committee
`D. J. BENOS , Chair
`D.H.ALPERS
`R. D. l\1.ANNING, JR.
`J. A. SCHAFER
`S. H. WRIGHT
`
`Ex officio
`M.REICH
`Publication s Manager
`and Executive Editor
`W. F. BORON, President
`M. FRANK, Executive Director
`
`A. G. TRUDGETT
`Editorial Manager
`Z. ERIM
`M. GOODWIN
`Senior Copy Edt.tors
`
`Physiologica l Genomics (ISSN-
`1094-8341) is published by the
`American Physiological Society,
`9650 Rockville Pike, Bethesda, MD
`20814-3991. Subscription Prices:
`Institutional: $200.00, US; $215.00,
`Canada a nd Mexico; $230.00, else(cid:173)
`where. Non-member individua l:
`$133.00, US; $150.00, Canada and
`Mexico; $165.00, elsewhere. APS
`member: $65.00. S ubscriptions
`from outside the United States are
`payable in US currency or full
`equivalent . J apanese yen price is
`available from om agent: USACO
`
`Corporation, 13-12 Shimbashi 1-
`Chome, Minato-Ku, Tokyo 105, J a(cid:173)
`pan. Subscriptions are accepted on
`a calendar-year basis only. S ingle
`Copies and Bach Issues: When
`availa ble, $50.00 each postpaid.
`
`Copyri ght © 1999 by the American
`Physiological Society. Printed in
`the United States of America . The
`code at the bottom of the first page
`of an a rticle indi cates the copy(cid:173)
`right owner's consent that copies
`of that article may be made be(cid:173)
`yond that permitted by sections
`
`107 and 108 of the US Copyright
`Law-unless the copies a re for gen(cid:173)
`eral distribution, for advertising,
`for creating new works, or for re(cid:173)
`sale-provided the per-copy fee is
`paid through the Copyright Clear(cid:173)
`a nce Center, Inc., 222 Rosewood
`Dr., Danvers, M.A. 01923 . 1094-
`8341/99 $5 .00.
`
`Editorial and Business Correspon(cid:173)
`de nce: Physiological Genomics,
`9650 Rockville Pike, Bethesda, MD
`20814-3991. E-mail : www.physiol(cid:173)
`genomics.org
`
`2
`
`
`
`Physiological Genomics
`MANDATORY SUBMISSION FORM
`
`Title:
`
`oResearch Artic le o BriefCommunication o Perspecti ve o News and Reports o Ed itorial o Lette r to Edi tor o Review o Too lbox
`
`A Signature Below Certifies Complian ce W ith th e Following Statem ents
`Copyright T ra ns fer. In consideration of th e acceptance of th e above wo rk for publication, I do hereby ass ign and transfer to The
`American Physio log ical Society all rights, title, and interest in and to th e copyright in the above-titled work. If any changes in
`authorsh ip (order, deletion s, or addition s) occur after the manuscript is submitted, agreement by all authors for such changes must be
`on file with the Soc iety . An author's name may only be removed at his/her own request. (Note: material prepared by empl oyees of the
`u.s. govern me nt in th e course of their offici al duties cannot be copyrighted.)
`*For com mercial companies authorized agent signatures are allowed for copyright transfer but authors must sign fo r authorshi p
`responsib ilities.
`Aut horship Res ponsibilities to be signed by each author
`Each autho r is required to attest that the manuscript is not currentl y under consideration elsewhere and the research reported will not be submi tted for
`publication elsewhere until a fi nal decision has been made as to its acceptability by the journal;
`the manuscript is truthful original wo rk without fa brication, fra ud, or plagiarism; each author has made an important scientific contri bution to the
`study and is tho roughly fa mili ar with the primary data; and each author has read the complete manuscript and takes responsibility fo r the content and
`completeness of the manuscript and understands that if the paper or part of the paper is foun d to be faul ty or fraudulent, he or she shares the
`responsibi lity. If the paper depends critically on another paper that is as yet unpubli shed, three copies of that paper should be included fo r reviewers.
`In cases in which obtaining a signature from each author would delay rev iew, the corresponding author's signature is suffic ient prov ided that the
`correspond ing author understands that he or she signs on behalf of all of the authors who have not signed the fo rm . It should also be un derstood by
`all authors that the corresponding author has signed th e manuscript submission fo rm as their proxy.
`Conflict of Inter est D isclosu re. A ll funding sources supporting the work and all in sti tutional or corporate affi liations of min e are
`acknowledged. Except as di sclosed on a separate attachment, I certi fy th at I have no commercia l assoc iati ons (e.g. , consultancies, stock
`ownership, equi ty interests, patent-li censing arrangements) th at might pose a conflict of interest in connection w ith the submitted
`article (letter attached).
`
`0 Please check if this article was written as part of the officia l duties of an employee of the U. S. G ove rnm ent
`
`Sign ature from EACH author required in th e sam e order a s on the manuscript ti t le page (Fax Signatures, Multi ple Forms
`Acceptab le). For more than 7 authors, use an extra sheet.
`
`Signature
`
`Signature
`
`Print Name
`
`Print Name
`
`Date
`
`Date
`
`Signatu re
`Date
`Print Name
`------------------------------------------------------------------------------------------------------
`Signature
`Print Name
`Date
`-----------------------------------------------------------------------------------------------------
`Signature
`Print Name
`Date
`
`Date
`Signature
`Print Na me
`---------------------------------------------------------------------------------------
`Signature
`Print Name
`Date
`
`'Commercial company agent's signature (if applicable)----------------------------------------------------------------------
`
`Co rresponding Author
`
`Phone --------------------- I' ax
`
`Mailing Address -----------------------------------------------------------------------------------------------
`
`--------------------------------------------------------------- E-mail ------------------------------------
`
`Please fax to 301 -571 -1814
`
`3
`
`
`
`.if
`
`~-
`
`WhY FRET over genomics?
`
`review
`
`DINO A. DE ANGELIS
`Cellular B iochem istry and Biophysics Program, Memoria l Sloan-Kettering
`Cancer Cente1; N ew York, N ew Yorll 10021
`
`DeAngelis, Dino A. Why FRET over genomics? Physiol. Genomics 1:
`93-99, 1999.-Gen etic information is being uncover ed quickly and in
`vast amounts through the largely a utomated sequencing of genomes from
`all kinds of organisms. As this information becomes available, enormous
`ch allenges a re emerging on three levels: first, functions will h ave to be
`assigned to individual gen e products; second, factors that influen ce the
`expression level of these gene products will h ave to be identified; and
`third, allelic variants th at act alone or in combination to give rise to
`complex traits will have to be ch ar acterized. Because of th e sh eer size of
`genomes, methods that can streamline or a utomate th ese processes are
`highly desirable. Fluorescence is an attractive readout for such high(cid:173)
`throughput tasks because of th e availability of equipment designed to
`detect light-emitting compounds with great speed and high capacity. The
`following is a n overview of th e achievemen ts and potential of fluorescence
`resonance energy tra nsfer (FRET) as applied in three areas of genomics:
`t he identification of single-nucleotide poly.morphisms, th e detection of
`protein-protein interactions, a nd th e genomewide analysis of regulatory
`sequences.
`green fluorescent protein; proximity imaging; high-throughput screen(cid:173)
`ing; single-nucleotide polymor phism; fluorescence-activated cell sor ting;
`digital imaging spectr oscopy
`
`FLUORESCENCE RESONANCE ENERGY TRANSFER
`
`Fluorescence r esonance en ergy transfer (FRET) is a
`quantum m echanical phenomenon that occurs between
`a fluorescence donor (D ) and a fluorescence acceptor (A)
`in close pr oximity (usually < 100 A of separ ation) if th e
`emission spectrum of D overlaps with th e excitation
`spectrum of A (Fig. 1A) (15). Under optimal FRET
`conditions, illumination at the excitation wavelength of
`D results in tra nsfer of ener gy, not photons, to A. If A is
`nonfluor escent, it can be r eferred to as a quencher (Q );
`in this case, FRET between D and Q r esults in a n et
`decrease in photon emission from D (Fig. 1C) . If A is
`also fluorescent, the decrease in emission from D is
`concomitant with an increase in fluorescence at the
`emission wavelen gth of A (Fig. 1B). The efficien cy of
`energy transfer (E ) decreases very rapidly with increas(cid:173)
`ing distance (R ) between the donor and acceptor, accord(cid:173)
`ing to th e r elationship E rx [1 + (RIR0 )6]- l, wh er e R 0 is
`the distance at which E is 50%. To illustr a te this point,
`simply making R = 2 X R 0 decreases FRET effici ency
`from 50% to ~ 1%.
`
`Received 3 June 1999; accepted in fi nal form 29 J~1ly 1999.
`Ar ticle published online before print . See web site fo r date of
`Publication (http://physiolgenomics.physiplogy.org).
`
`FRET-induced ch a n ges in fluorescence intensity
`(wh en A is a nonfluorescent quen ch er, Q) or in th e r atio
`of emission inten sities of th e donor and acceptor (wh en
`A is fluorescen t) h ave been widely exploited in biology
`to monitor proximity relationships between appropri(cid:173)
`ately labeled m acromolecules (3 1, 33). Absolute dis(cid:173)
`t ances between donor and acceptor pair s can be mea(cid:173)
`sured with FRET; it h as been coined a "spectroscopic
`r uler" (33). However, absolute distance measurements
`ar e not u su ally a ttempted becau se F RET efficiency is
`also influenced by th e angle between the dipole mo(cid:173)
`ments of D a nd A, a par a meter th a t is often difficult t o
`qu an tify. Th e most common application of th e tech (cid:173)
`nique is to dyn a mically monitor th e r elative distance
`between D and A aft er chan ges induced in the system
`under study.
`
`ANALYSIS OF SINGLE-NUCLEOTIDE POLYMORPillSMS
`WITH FRET
`
`One of th e greatest ch allenges facing geneticists is
`th e analysis of complex traits and diseases. Complexity
`arises because ther e is no simple correspondence be(cid:173)
`tween genotype and phenotype, for instance, when
`muta tions in a ny one of several genes r esult in the
`sam e phenotype or wh en mutations in several genes
`ar e r equired to give rise to a particular trait (20) . One
`way of zeroing in on candidate gen es is through gen etic
`
`1094-8341199· $5.00 Copyright © 1999 the Am erican Physiological Society
`
`93
`
`4
`
`
`
`94
`
`REVIEW
`
`A
`
`Fig. 1. Fluorescence resonance en ergy tra nsfer (FRET).
`A: excitation (dotted lines) and emission (solid lines)
`spectra of a n ideali zed fluorescence donor (D, white
`lines) and fluorescence acceptor (A, black lines) pair.
`Parti a l overlap between emission spectrum of D and
`excitation spectrum A is shown in shaded ar ea. Wave(cid:173)
`length a nd inten sity va lues ar e in arbitrary units . B:
`schematic representation of a fluorescence donor (D)
`and acceptor (A) pair. B, left: D and A are not wit hin
`FRET distance; excitation ofD res ults in emission from
`D. B, rig ht: D and A are within FRET dista nce; excita (cid:173)
`tion of D results in tran sfer of energy to A (white
`lightning), ca using net increase in a mount of light
`emitted fTom A concomitant with decrease in amount of
`light emitted from D. C: same as in B, but acceptor is a
`nonfluorescent quencher (Q); energy transfer (white
`lightning) res ults in net decrease in fluorescence inten(cid:173)
`sity of D.
`
`B
`
`wavelength
`
`c
`
`association studies that compare the prevalence of a
`marker (or a set of markers) between affected and
`nonaffected individuals. Single-nucleotide polymor(cid:173)
`phisms (SNPs) are genetic markers that are densely
`distributed throughout the genome, at an average
`frequency of 1 per 1,000 base pairs (8). Organizations
`such as the National Human Genome Research Insti(cid:173)
`tute (http://www.nhgri.nih.gov/) and the recently
`created SNP Consortium (23) are creating public data(cid:173)
`bases cataloguing several hundreds of thousands of
`these markers. These databases, coupled with efficient
`methodologies to detect SNPs, will facilitate the identi(cid:173)
`fication of genes responsible for complex disorders such
`as diabetes, hypertension, or schizophrenia.
`Although SNPs can be successfully analyzed by
`traditional methods based on gel separation (Sanger or
`Maxam-Gilbert sequencing, restriction fragment length
`polymorphism), several more efficient/automatable al(cid:173)
`ternatives have been devised (19) . Most methods are
`based on differences in hybridization to a given oligo(cid:173)
`nucleotide probe between a perfectly matched and a
`mismatched sequence. For instance, DNA chips consist
`of thousands of oligonucleotide probes arrayed at high
`density on a solid support in a predetermined order
`(37). DNA samples are amplified by PCR using fluores(cid:173)
`cent nucleotide analogs; these labeled PCR products
`can only hybridize to perfectly matching oligonucleo(cid:173)
`tide sequences immobilized on tli.e .chip. Images of the
`hybridized fluorescent probes are obtained by confocal
`microscopy of the chip, allowing direct comparison
`
`between patient and control hybridization patterm
`The main advantage ofthis approach is that it enabl
`the simultaneous detection of thousands of SNPs fro
`a given genomic DNA sample; however, the fact tha
`the probes are fixed to a chip that is difficult
`manufacture hinders the detection of newly cataloguec
`SNPs.
`More flexibility is provided by simple homogeneoll!
`assays that allow the detection ofSNPs in solution; fow
`of these are based on FRET and have been reviewed iD
`detail elsewhere (19). The TaqMan assay (21) is
`protocol that utilizes short oligonucleot ides probes tha
`undergo FRET in their intact state because they an
`labeled at each end with a fluorescence donor a
`acceptor pair (Fig. 2A ). Even in probes longer than 3(
`nucleotides, significant FRET can occur because
`hydrophobic interactions between D and A (5). Tbl
`probe is designed to hybridize to a sequence be·
`amplified in a PCR reaction using Taq polymeraJ
`(downstream of one of the primers). If there is a perfl
`match between the probe and the target, the 5
`exonuclease activity of Taq polymerase can digest tbl
`hybridized TaqMan probe during the elongation cycle
`separating D from A and resulting in a decrease ·
`FRET.
`The "molecular beacons" of 'I'yagi and Kramer (36
`represent a second approach. These oligonucleotidi
`probes are chemically modified with a fluore scen
`donor (EDANS) at their 5' end and a nonfluorescenl
`quenching acceptor (DABCYL) at their 3' end. In
`
`http://physiolgenomics.physiology.org
`
`5
`
`
`
`MISMATCH
`
`A
`
`B 0
`
`X
`
`c
`
`D
`
`REVIEW
`
`MATCH
`
`95
`
`Fig. 2. Four FRET-based approaches for
`analysis of single-nucleotide polymor(cid:173)
`phisms (SNPs) in solution. A: Taql\•Ian
`assay. A , left : TaqMa n probe does not
`match ta rget DNA (gray horizontal li ne;
`mismatch represented by ''X") a nd under(cid:173)
`goes FRET in the intact state upon illumi(cid:173)
`nation of D; 7hq polym erase (Taq) ca n
`extend a PCR primer (arrow ) loca ted up(cid:173)
`stream of mismatched target sequence. A,
`.right: hybridized TaqMa n probe is di(cid:173)
`gested by 5'-exonuclease activity of Taq
`polymerase duri ng exte nsion of PCR
`primer, resul ting in separation of D from
`A a nd disruption of FRET signa l. B: mo(cid:173)
`lecular beacons. B , left : in presence of a
`mismatched target , the probe adopts a
`s tem-and-loop s tructure; because of prox(cid:173)
`imity of 5' a nd 3' ends of probe, donor (D)
`fl uorescence is di rectly coupled to nonfluo(cid:173)
`rescent quenching acceptor (Q ), resulting
`in very low levels of fl uor escence emission
`from D. B, rig ht : in presence of correct
`target , molecular beacon h ybridizes per(cid:173)
`fectly to the ta rget ; D and Q are sepa(cid:173)
`rated, giving rise to a disruption of FRET
`a nd a n increase in em ission from D. C:
`dye-labeled oligonucleotide ligation (DOL).
`C, left : a mismatch at junction between 2
`contiguous oligonucleotides [one labeled
`with a fluorescence donor (D) and t he
`other with an acceptor (A)) prevents DNA
`ligase (Lig) from joini ng the 2 fragments
`together, preventing energy transfer be(cid:173)
`tween D and A. C, rig ht : a perfect match
`allows ligase to li nk the oligonucleoticles,
`resulting in FRET. D: template-directed
`dye-terminator incorpora tion (TDI). DNA
`polymer ase (Pol) a nd an a cceptor-labeled
`n ucleotide ter minator analog correspond(cid:173)
`ing to SNP of interest a re added to a
`fl uorescence donor-labeled oligonucleo(cid:173)
`tide hybridized to a DNA ta rget . D, left: in
`a misma tch situation, the analog is not
`incorpor ated at the 3' end of the primer
`and FRET does not occur. D, rig ht : with a
`perfect ma tch, the ana log is incor pora ted
`a nd FRET resul ts.
`
`absence of a perfectly m at ch ed target , they assume a
`stem-and-loop structure in solution: the loop is a DNA
`strand complem entary to th e tar get, and th e st em is
`formed by intr amolecular base-pair ing of short comple(cid:173)
`ment ary sequences at each end of th e loop (Fig. 2B ).
`This hairpin conformation positions D and Q in ex(cid:173)
`tremely close pr oximity, much closer th a n a pair of
`fluorophores at opposite ends of a r andomly coiled
`oligonucleotide; this effectively quen ch es donor fluores(cid:173)
`cence . In the presen ce of a perfectly m atch ed sequence,
`the oligonucleotide under goes a confor mational ch ange
`that allows the h airpin loop to hybridize to the t ar get ,
`separating D from Q a n d r esulting in a fluor escen ce
`increase (up to 900-fold; Fig. 2B). Th e ability of molecu (cid:173)
`lar beacons to form h air pin structures sig11ificantly
`enhances their specificity compar ed with st andard
`oligonucleotide pr obes of th e same size, allowing th em
`to readily distinguish between a perfect m at ch and a
`single base mism atch (3) . Recently, molecular beacons
`With seven additional donors r anging from blue to r ed
`fluor escence emission have been developed , permitting
`
`th e simulta n eous detection of several target s within
`the sam e tube (35).
`Dye-labeled oligonucleotide ligation (DOL) (7) is a
`third FRET-based technique th at relies on th e extreme
`sensitivity of DNA ligase to mism atch es close t o th e
`ligation site. Th e m ethod ut ilizes two oligonucleotides,
`one labeled with a fluorescen ce donor, th e oth er with an
`acceptor (Fig. 2C) . The probes are designed to hybridize
`adjacen tly on a cont iguous DNA target sequence; if
`both oligonucleotides m atch perfectly, DNA ligase can
`join th e fragm ents togeth er, resulting in FRET. A
`mism at ch locat ed on the 3' end of the ligation site (first
`base of th e oligonucleotide) will r esult in failure of th e
`ligase t o link th e labeled oligonucleotides, thus preven t(cid:173)
`ing FRET.
`Fin ally, in templa t e-direct ed dye-t erminator incorpo(cid:173)
`r ation (TDI) (6), a donor-labeled oligonucleotide primer
`designed with its 3' end immedia tely adj acent to a
`polymor phic nucleotide is allowed to hybridize to a
`given tar get sequ ence (Fig. 2D ). In a subsequent reac(cid:173)
`tion simila r to dideoxy chain t ermination sequencing,
`
`http://physiolgenomics.physiology.org
`
`6
`
`
`
`96
`
`REVIEW
`
`DNA polymerase and an acceptor-labeled nucleotide
`terminator analog corresponding to the SNP of interest
`are added; FRET results only if the analog is incorpo(cid:173)
`rated at the 3' end of the primer, indicating a perfect
`match. Although TDI is extremely easy to optimize, it
`necessitates addition of specific reagents twice after the
`initial reaction setup; the other three FRET-based
`methods are "walk-away" assays that only require
`proper thermal cycling ofthe initial reaction mixture.
`Although these four methods do not approach the
`degree of parallelism or multiplexing offered by high(cid:173)
`density DNA arrays, where simultaneous analysis of
`thousands of SNPs per reaction is customary, they are
`much more flexible: new SNPs can be detected without
`having to remanufacture a custom DNA chip because
`the specific probes and reagents required are easy to
`obtain or synthesize. Moreover, these FRET-based as(cid:173)
`says are fast and simple: with the exception ofTDI (see
`above), all necessary reagents are initially combined
`and changes in FRET can be monitored in real time
`during the course of the assays without having to
`process the samples further and without the help of
`expensive imaging equipment.
`
`ANALYSIS OF CODING SEQUENCES: PROTEIN-PROTEIN
`INTERACTIONS ANALYZED WITH FRET AND PRIM
`
`Because most cellular processes are carried via pro(cid:173)
`tein assemblies, important clues to the function of a
`novel gene product can be obtained by identifYing the
`set of proteins with which it interacts. On a genome(cid:173)
`wide basis, such information would constitute what has
`been termed "protein linkage maps" (13) . Methods able
`to streamline the identification of protein-protein inter(cid:173)
`actions are necessary to cope with the large amount of
`genomic data being generated (4). In the yeast two(cid:173)
`hybrid system (14), the DNA binding domain of a
`transcription factor is fused to a sequence of interest in
`a vector, and a DNA library is fused to the transcrip(cid:173)
`tional activation domain in a second vector. Protein(cid:173)
`protein interactions are detected in cotransformed yeast
`colonies on the basis of reconstituted transcriptional
`activity, which is coupled to a functional effect (usually
`survival on selective media). Inserting a library in each
`of the vectors allows the mapping of protein-protein
`interactions on a genomewide basis, as was demon(cid:173)
`strated with the bacteriophage T7 genome (2); larger
`genomes can be handled with multiple-round two(cid:173)
`hybrid screens (16). In phage-display technologies,
`eDNA libraries are expressed on the surface of phage
`particles directly through fusion to a coat protein (18)
`or indirectly via strong noncovalent interaction with a
`modified coat protein (10). Typically, the phage libraries
`are presented to a target immobilized on a solid sup(cid:173)
`port; nonbinding phages are washed away and specific
`binding phages are enriched and analyzed (27). A
`combination of phage and bacterial display has re(cid:173)
`cently been developed that can potentially detect inter(cid:173)
`actions not only between a target protein and a library
`of coding sequences but also between two libraries (4, 22).
`Since the initial cloning of green fluorescent protein
`from the jellyfish Aequorea vic~oria and its expression
`
`in differ ent cells and organisms . (5, 30), the codin
`sequence h as been mutagenized extensively t o improv
`or alter some of the spectral and biophysical propertie
`ofthe original. The various mutants have recently bee
`grouped under seven classes (34), including varian ·
`emitting light of different wavelengths: green, hlu~
`cyan, and yellow fluorescent proteins (GFP, BFP, CFI:
`and YFP, respectively). The advent of these spectra
`variants has opened up another platform for the detet.
`tion of protein-protein interactions: appropriately se.
`lected mutants can serve as D-A pairs in a FRET
`scheme. In proof-of-concept experiments (17, 25), BFp
`(D) and GFP (A) were linked together via spacers of2o
`or 25 amino acids containing a protease cleavage site
`As a result afforced proximity, substantial intramolecu.
`lar energy transfer occurred from BFP to GFP in th
`uncleaved species; addition of the specific protease l
`to the disappearance of the FRET signal. GFP donors
`and acceptors have since been used in numerous FRET
`based applications (29). Currently, the optimal D-A pair
`consists of optimized CFP and YFP variants, respec
`tively (26). The drawbacks ofusing a 28-kDa protein a:
`a fluorescent tag (large size, possible disruption o
`protein function) are outweighed by the advantages o
`specifically and homogeneously labeling a given protein
`and targeting it to any subcellular location for which
`there is a known consensus sequence.
`The ease with which given coding sequences can be
`fused to GFP suggests that high-throughput methods
`may be devised to detect protein-protein interaction;
`by FRET on a genomewide basis (1) (Fig. 3A) . In such
`scheme, an appropriate host (yeast or bacteria) waul
`be transformed with a vector containing a given co.ding
`sequence (or a library) fused with CFP and a second
`vector with a library fused to YFP and then grown on
`selective media. Only a small fraction of the surviving
`clones are expected to have an altered fluorescena
`emission ratio, indicative of a protein-protein interac
`tion; the major task would then be to identify an
`isolate these positive clones for further analysis . Fortu(cid:173)
`nately, existing technologies can be adapted for such
`task: isolation of positive clones from liquid cultures
`can be performed by iterative cycles of fluorescence(cid:173)
`activated cell sorting (FACS), as was previously don
`for the identification of strongly fluorescent variants~
`GFP (9). Solid-phase screening directly from agar plate!
`containing the transformants is also possible: digi
`imaging spectroscopy (DIS) can record spectral informs
`tion from any two-dimensional surface in a massiveJ
`parallel fashion (39). Thus DIS could be configured
`report the CFP-to-YFP emission ratio of every pixel in
`given field, allowing simultaneous imaging of up to 21
`agar plates in certain setups. Using either FACS or
`DIS, positive clones could be expanded, and the protein
`protein interactions giving rise to changes in emissiot
`ratio could be characterized.
`Whereas FRET is ideally suited to detect heterotyp"
`protein interactions, proximity imaging (PRIM) is
`novel technique that can optically report homotyp
`interactions of GFP-labeled proteins (12). PRIM e~
`ploys the original wild-type GFP with two amino a
`
`h ttp://physiolgenomics. physiology.org
`
`7
`
`
`
`A
`
`B
`
`X
`
`X
`
`X
`
`REVIEW
`
`97
`
`Fig. 3. Analysis of protein-protein interactions with
`green fluorescent protein (GFP)-based probes . A: detec(cid:173)
`tion of heterotypic protein interaction with F RET be(cid:173)
`tween cyan fluorescent protein (CFP) fusion to protein X
`and yellow fluorescent protein (YFP) to protein Y. Under
`optimal proXimity and angular conditions, interaction
`between X andY (rig ht ) causes a decrease in intensity of
`CFP fluorescence concomitant with an increase in inten(cid:173)
`sity of YFP fluorescence. B: detection of homotypic
`protein interaction of protein X using proximity imaging
`(PRIM) with thermotolerant GFP (tt GFP). Under opti(cid:173)
`mal proximity and angular conditions, self-associa tion
`of X will produce a change in exci tation ratio of ttGFP
`(depicted by change in shape of shading within ttGFP
`rectangle).
`
`tt
`GFP
`
`I I
`
`X
`
`tt
`GFP
`~ I
`X
`
`substitutions that help the protein fold at 37°C (32)
`[henceforth referred to as ttGFP (thermotolerant GFP)].
`Unlike the GFP variants with simplified excitation
`spectra used in FRET, ttGFP possesses two excitation
`peaks (at 395 and 475 nm) and one emission peak (at
`510 nm). The ratio of intensities ofthe excitation peaks,
`i.e. , the excitation ratio, is the same for any ttGFP(cid:173)
`labeled protein that is monomeric. However, on dimer(cid:173)
`ization, deviations from this excitation ratio occur (Fig.
`3B), detectable in vitro and in vivo (12, 24) . Unlike
`FRET, which is a quantum mechanical interaction
`between any D-A pair, PRIM is GFP specific and results
`from orientation-dependent structural interactions be(cid:173)
`tween two ttGFP modules that affect their excitation
`ratio. High-throughput identification of proteins or
`protein domains capable of self-association could be
`achieved by adapting FACS or DIS for excitation ratio
`imaging. Exhaustive screening for homotypic interac(cid:173)
`tions by PRIM would be unidimensional, requiring the
`expression of a single library of coding sequences fused
`to ttGFP. By comparison, exhaustive screening for
`pairwise heterotypic interactions (with any of the meth(cid:173)
`ods mentioned above) is a two-dimensional combinato(cid:173)
`rial process in which each coding sequence has to be
`tested against every other coding sequence, necessitat(cid:173)
`ing the expression of two libraries. A genome of 50,000
`coding sequences thus would minimally require the
`screening of 50,000 clones for homotypic interactions
`and 0. 5 X (50,000)2 or 1,250,000,000 clones for hetero(cid:173)
`tYPic interactions.
`Each of the above-mentioned techniques suffers from
`shortcomings in certain respects . Phage display tech(cid:173)
`niques are limited to the study of relatively small- to
`medium-size proteins lacking eukaryotic posttransla(cid:173)
`tional modifications (4). The yeast two-hybrid system is
`known to generate a high number of false positives clue
`to nonspecifically interacting proteins and transcrip(cid:173)
`tional autoactivation by the DNA-binding fusion pro-
`
`tein (2). Moreover, interactions have to take place
`within the yeast nucleus. In contrast , GFP-based FRET
`or PRIM measurements can be performed in bacteria,
`yeast, or other cell types; moreover, interactions can
`theoretically be detected within any subcellular organ(cid:173)
`elle. However, GFP-based measurements are ham(cid:173)
`pered by several problems. 1) The GFP mutants used as
`FRET donors typically photobleach at much faster
`rates than acceptors, complicating the analysis some(cid:173)
`what (26); with PRIM care must be taken not to
`illuminate samples too intensely in the UV range
`because this can cause excitation ratio changes through
`a photoisomerization effect (11). 2) A large number of
`labeled molecules have to be expressed to be detectable,
`preferably at similar levels in the case of FRET with
`CFP- and YFP-tagged proteins; the detection threshold
`
`A
`
`8
`
`Fig. 4. Ana lysis of regulatory regions using [3-lactamase cleavage of
`CCF2 . A: CCF2 consists of blue-emitting coumarin (donor D) linked
`to green-emitting fluorescein (acceptor A) via a [3-lactam bond. In
`cells lacking [3-lactamase, excitation of D results in fluorescence
`emission predominantly from A due to FRET in the intact probe. B: in
`presence of [3-lactamase (scissors), CCF2 is cleaved, disrupting
`FRET; excitation ofD res ults in emission predominantly from D.
`
`http://physiolgenomics.physiology.org
`
`8
`
`
`
`98
`
`REVIEW
`
`is 200 nM or - 2,000 copies/cell for yeast and 200
`copies/cell for E. coli (28). This can pose certain prob(cid:173)
`lems, for instance in cases where a protein is toxic to
`the host above a certain threshold concentration. 3)
`Stringent distance and angular requirements for FRET
`and PRIM signals might contribute to a large number
`of false negatives, i.e., proteins that actually interact
`but do not produce a change. in the emission ratio
`(FRET) or excitation ratio (PRIM) because of unfavor(cid:173)
`able geometry of the GFP modules within the protein
`complex. Despite these shortcomings, screening protein(cid:173)
`protein interactions using GFP-based techniques can
`be a valid alternative, especially when screening for
`homotypic protein interactions or when eukaryotic
`posttranslational modifications or specific subcellular
`localization of the target proteins is required.
`
`ANALYSIS OF REGULATORY SEQUENCES WITH FRET
`
`Functional genomics would be incomplete without an
`understanding of the way in which regulatory se(cid:173)
`quences dynamically con