G-TRACE: rapid Gal4-basedcell lineage analysisin DrosophilaCory J Evans1,6, John M Olson1,6, Kathy T Ngo1,Eunha Kim1, Noemi E Lee1, Edward Kuoy1,Alexander N Patananan1, Daniel Sitz1,PhuongThao Tran1, Minh-Tu Do1, Kevin Yackle1,Albert Cespedes1, Volker Hartenstein1–3,Gerald B Call1,5 & Utpal Banerjee1–4

We combined Gal4-UAS and the FLP recombinase–FRT and

fluorescent reporters to generate cell clones that provide spatial,

temporal and genetic information about the origins of individual

cells in Drosophila melanogaster. We named this combination the

Gal4 technique for real-time and clonal expression (G-TRACE). The

approach should allow for screening and the identification of real-

time and lineage-traced expression patterns on a genomic scale.

The yeast transcriptional activator Gal4 has been extensively used asa genetic reporter in Drosophila1. To date, thousands of transgeniclines have been generated that exhibit distinct Gal4 expressionpatterns during development and are available for analysis2,3. Herewe describe a Drosophila screening tool, which we name the Gal4technique for real-time and clonal expression (G-TRACE) system,developed by the University of California Los Angeles (UCLA)Undergraduate Research Consortium in Functional Genomics4,5.This system reveals real-time Gal4 expression, similar to traditionalGal4 analysis, but also marks cell lineages derived from Gal4-expressing cells (Fig. 1a and Supplementary Fig. 1). The keyfeature is that the initiation of lineage reporter expression is Gal4-dependent whereas maintenance of lineage reporter expression isnot. G-TRACE makes use of fluorescent protein reporters for bothreal-time analysis (RFP) and lineage-based analysis (enhancedGFP (EGFP)), thereby increasing screening throughput as there isno requirement for antibody staining (Fig. 1a).

To initiate lineage tracing, Gal4 mediates the expression ofFLP recombinase, which in turn removes an FRT-flanked tran-scriptional termination cassette inserted between a Ubiquitin-p63E(Ubi-p63E) promoter fragment and the EGFP open reading frame.Thereafter, the Ubi-p63E promoter maintains EGFP expressionperpetually in all subsequent daughter cells, independent of Gal4

activity. These ‘FLP-out’ constructs have previously been used to assessbiological function including lineage information6–10; G-TRACE is toour knowledge the first FLP-out system to use the Ubi-p63E promoterand a direct EGFP reporter.

To demonstrate the usefulness of the G-TRACE system, we eval-uated Gal4 lines with well-defined real-time expression patterns at thethird larval instar but for which expression patterns at earlier stageswere less clear (Fig. 1b–g and Supplementary Fig. 2). In the wingimaginal disc, hedgehog-gal4 (hh-gal4) mediated RFP expressionthroughout the posterior compartment (Fig. 1b). Examination ofthe cognate lineage-traced EGFP pattern revealed uniform expressionlimited to the posterior compartment (Fig. 1c), confirming priorgenetic evidence. Both patched-gal4 (ptc-gal4) and decapentaplegic-gal4 (dpp-gal4), which respond to Hedgehog signaling, expressedRFP in a stripe of anterior compartment cells along the anterior-posterior compartment boundary (Fig. 1d,f). However, despitesimilar real-time expression at this developmental stage, EGFPexpression in ptc-gal4 marked the entire anterior compartment(Fig. 1e) but only a portion of the anterior compartment in dpp-gal4 (Fig. 1g), indicating differential regulation of these genesduring wing development.

The G-TRACE system worked well across all tissues, includingthe brain and lymph gland (Supplementary Figs. 3 and 4); weanalyzed real-time expression and lineage tracing for the Gal4enhancer trap (Gal4-ET) line NP0114 in the brain (Fig. 1h,i) andCollagen-gal4 (Cg-gal4) in the lymph gland (Fig. 1j,k). The methodwas also useful for identifying Gal4 lines with expression patternsrestricted to early development (that is, showing EGFP expression,but no RFP expression at later developmental stages), such as forHemese-gal4 (He-gal4), a putative blood-specific marker11, in thelate third instar brain (Fig. 1l,m) and for the Gal4-ET line NP0829in the lymph gland at the same stage (Fig. 1n,o).

We found many lines with expression patterns in distinct brainlineages (Fig. 2a). For example, in Gal4-ET lines NP0114 (Fig. 2b–d)and NP0189 (Fig. 2e,f), current (RFP) expression (third instar) wasrestricted to the central brain, whereas a substantial number oflineage-traced populations were found in the optic lobes and surfaceglia. In contrast to these lines, the Gal4-ET line NP0829 exhibitedwidespread current RFP and lineage expression in the optic lobe(Fig. 2g–i). A summary of Gal4 expression patterns in four differenttissues and representative examples for each Gal4 line screenedduring the course of this work is available in SupplementaryTables 1–4 and Supplementary Data as well as at http://www.mcdb.ucla.edu/research/banerjee/urcfg/.

To better understand how late stage patterns arise, we selected asubset of Gal4 lines for analysis at earlier stages (Fig. 3 and

RECEIVED 26 MARCH; ACCEPTED 8 JUNE; PUBLISHED ONLINE 26 JULY 2009; DOI:10.1038/NMETH.1356

1Department of Molecular, Cell and Developmental Biology, 2Department of Biological Chemistry, 3Molecular Biology Institute, 4Eli and Edythe Broad Center ofRegenerative Medicine and Stem Cell Research at University of California, Los Angeles, University of California, Los Angeles, Los Angeles, California, USA. 5Presentaddress: Department of Pharmacology, Midwestern University, Glendale, Arizona, USA. 6These authors contributed equally to this work. Correspondence should beaddressed to U.B. (banerjee@mbi.ucla.edu).

NATURE METHODS | VOL.6 NO.8 | AUGUST 2009 | 603

BRIEF COMMUNICATIONS

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

Supplementary Fig. 5); in each case, we identified the earliest timepoint at which RFP was expressed in the tissue or area of interest. Inthe late third instar, real-time expression of OK107-gal4 was

restricted to the mushroom body of thecentral brain, but a large fraction of theoptic lobe expressed EGFP (Fig. 3a,b).Consistent with this pattern, OK107-gal4was expressed in many cells in the region ofthe developing optic lobe by the late firstinstar (Fig. 3c,d). Unpaired-gal4 (upd-gal4)drove RFP expression in the optic lamina atthe late third instar12 (Fig. 3e), whereasEGFP was expressed throughout the opticlobe (Fig. 3f). The earliest expression ofupd-gal4 in the developing optic lobe wasduring the early stages of the second instar(Fig. 3g,h). The hh-gal4 line lacked real-time expression in the late third-instarbrain but generated a large cluster ofEGFP-expressing cells in the posterior ven-

tral portion of the optic lobe (Fig. 3i,j). We determined that hh-gal4was expressed in the early first instar in a small cluster of cellslocated ventrally within the posterior region of the optic lobe

Enhancer gal4A AP

A

A

MZ

NP0114

NP0829

MZ

CZ CZ

P

P A P

A P

P

hh-gal4

ptc-gal4

dpp-gal4

He-gal4

Cg-gal4

RFP

RFPEGFPDNA

UAS

Stop

‘Cell color’

Real-timeexpression

Lineageexpression

Real-time and lineageexpression

Ubi p63

Ubi p63

UAS

FRT

FRT

FRT nEGFP

nEGFP

RFP

FLP

a b c

d e

f g

RFPRFPEGFPDNA

h i j k

l m n o

Figure 1 | The G-TRACE analysis system.

(a) Schematic of the molecular mechanisms of the

G-TRACE system. Gal4 activates the expression of

RFP and FLP recombinase. Cells expressing FLP

recombinase then excise the FRT-flanked stop

cassette separating the Ubi-p63E promoter and

nuclear EGFP (nEGFP) open reading frame. This

initiates expression of EGFP, which is heritably

maintained in all daughter cells. The ’cell color’

gradient represents the apparent color of cells

after initiation of Gal4 expression. (b–o) Fluor-

escence images showing G-TRACE analysis in

hh-gal4 (b,c), ptc-gal4 (d,e), dpp-gal4 (f,g),

NP0114 (h,i), Cg-gal4 (j,k), He-gal4 (l,m) and

NP0829 (n,o) reporter lines, showing current RFP

expression alone (b,d,f,h,j,l,n) and merged with

lineage-traced EGFP expression (c,e,g,i,k,m,o).

Images show the wing imaginal disc (b–g), the

brain (h,i,l,m) or the lymph gland (j,k,n,o), all at

the late third instar. A, anterior wing compart-

ment; P, posterior wing compartment. CZ, cortical

zone; MZ, medullary zone. Scale bars, 50 mm.

Figure 2 | G-TRACE analysis reveals new cell

lineage markers. (a) Schematic of lineages

from larval central brain neuroblasts.

(b–f) Fluorescence images showing G-TRACE

analysis in the Gal4-enhancer trap lines

NP0114 (b–d) and NP0189 (e,f). Images in dand f show magnification of the boxed regions

in c and e, respectively. MB, mushroom body;

OL, optic lobe. (g) Schematic depicting optic

lobe development. Cross-section of a brain

hemisphere at early and late stages showing

the outer anlagen (OA; red) and inner anlagen

(IA; gray) of the optic lobe. Neuroblasts (NB;

green) at the lateral edge form neurons of the

medullary primordium (MP), and medial edge

neuroblasts form neurons of the laminar

primordium (LAP). The inner optic anlage (IA)

forms the lobula primordium (LP). (h) Fluorescence image showing G-TRACE analysis in the Gal4-enhancer trap line NP0829. OOA, outer optic anlage.

(i) Magnification of the boxed region in h. Scale bars, 50 mm.

Neuroblast

Neuroblast

Olderneurons

Youngneurons

Early larva

OL

MB

NP0114

NP0189 Surface glia

MB

MB

OL

OA

MP

IA IA

MP MPMPNeurons

LPLP

LAP

OA OA OANB

IA IA

OL

Late larva

Recently bornneurons

Oldneurons

Early larva Late larva

Neuroblasts andprecursors

Neuroblast

Opticfurrow

Opticfurrow

Olderneurons

OOA

NP0829

Youngneurons

a

b c d

e

h i

g

f

604 | VOL.6 NO.8 | AUGUST 2009 | NATURE METHODS

BRIEF COMMUNICATIONS

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

(Fig. 3k,l). In the lymph gland, hand-gal4 expression was absent atthe late third instar, but the vast majority of lymph gland cellsexpressed EGFP at this stage (Fig. 3m,n). Using G-TRACE, wefound hand-gal4 to be expressed in the lymph gland in late-stageembryos (Fig. 3o,p), consistent with reports of hand gene13

expression in the embryonic cardiogenic mesoderm.The G-TRACE system places Gal4 expression patterns in a

developmental context by combining real-time expression analysiswith a cell-autonomous FLP-out marking system, which is parti-cularly valuable in providing information about gene expression,such as transient expression patterns, that may be limited orinaccessible using other reporter systems (for example, standardGal4 expression analysis or analysis of spatially random heat shock-induced clones). Temporal analysis (Fig. 3) can be used to distin-guish between broad EGFP expression owing to small numbers ofprecursors that proliferate extensively (that is, actual cell clones)and Gal4 expression among many post-mitotic or slowly prolifer-ating cells at later stages. For any line, a subset of Gal4-expressingcells may not initiate EGFP expression owing to a ‘threshold effect’.We found that the efficiency of generating EGFP-marked cell clonescorrelated primarily with driver strength, that is, the amount andduration of Gal4 expression. Strong drivers, such as hh-gal4 andubiquitin-gal4 (Supplementary Fig. 6), initiated lineage tracing in

nearly all Gal4-expressing cells, whereas weaker drivers, such aseyeless-gal4, caused unmarked clones to appear at high frequency(data not shown). Drivers that expressed Gal4 in a gradient, suchas hand-gal4, reproducibly generated both marked and unmarkedlineages (Supplementary Fig. 6). Thus, reliable Gal4-based lineagepatterns as defined by the G-TRACE system were those that werehighly reproducible among multiple samples.

In summary, the G-TRACE system is very useful for theidentification and characterization of Gal4 lines with early and/ortransient expression patterns and in describing the development oftissues with complex lineages. Furthermore, G-TRACE has thepotential to be valuable in studying stem cells and how theycontribute to the growth and maintenance of tissues, as well as instudying phenomena such as cell death, senescence and dediffer-entiation. Future analyses may benefit from modified G-TRACEsystems with different reporters (for example, membrane-associated GFP) or in which Gal4 activity can be modulated duringdevelopment, such as with Gal80ts, a temperature-sensitive inhi-bitor of Gal4 function.

METHODSMethods and any associated references are available in the onlineversion of the paper at http://www.nature.com/naturemethods/.

Note: Supplementary information is available on the Nature Methods website.

ACKNOWLEDGMENTSK.T.N., Eu.K., N.E.L., Ed.K., A.N.P., D.S., P.T., M.-T.D., K.Y. and A.C. participated asundergraduate research students in the UCLA Undergraduate Research Consortiumin Functional Genomics, which is supported by a Howard Hughes Medical InstituteProfessor’s Award to U.B. C.J.E. was supported by a US National Institutes ofHealth-UCLA Vascular Biology training grant T32HL69766. J.M.O. and G.B.C.were supported by the Howard Hughes Medical Institute Professor’s award and wereinstructors in the UCLA Undergraduate Research Consortium in FunctionalGenomics. This work was supported by National Institutes of Health grantsR01EY008152 and R01HL067395 to U.B. We thank G. Struhl (ColumbiaUniversity) and P. O’Farrell (University of California, San Francisco) for providingDNA reagents.

AUTHOR CONTRIBUTIONSC.J.E. and U.B. devised the G-TRACE system. C.J.E., Eu.K., K.Y. and A.C. generatedthe G-TRACE system. C.J.E., K.T.N., Eu.K., N.E.L., Ed.K., A.N.P., D.S., P.T., M.-T.D.,K.Y. and A.C. generated data. C.J.E., J.M.O., G.B.C. and U.B. provided instructionand supervision. V.H. provided the brain schematics and helped interpret brainlineages. J.M.O. and K.T.N. generated the online database. C.J.E. and U.B. wrotethe manuscript.

Published online at http://www.nature.com/naturemethods/.Reprints and permissions information is available online athttp://npg.nature.com/reprintsandpermissions/.

1. Elliott, D.A. & Brand, A.H. Methods Mol. Biol. 420, 79–95 (2008).2. Pfeiffer, B.D. et al. Proc. Natl. Acad. Sci. USA 105, 9715–9720 (2008).3. Hayashi, S. et al. Genesis 34, 58–61 (2002).4. Chen, J. et al. PLoS Biol. 3, e59 (2005).5. Call, G.B. et al. Genetics 177, 689–697 (2007).6. Pignoni, F. & Zipursky, S.L. Development 124, 271–278 (1997).7. Ito, K. et al. Development 124, 761–771 (1997).8. Struhl, G. & Basler, K. Cell 72, 527–540 (1993).9. Weigmann, K. & Cohen, S.M. Development 126, 3823–3830 (1999).10. Jung, S.H. et al. Development 132, 2521–2533 (2005).11. Kurucz, E. et al. Proc. Natl. Acad. Sci. USA 100, 2622–2627 (2003).12. Yasugi, T. et al. Development 135, 1471–1480 (2008).13. Han, Z. & Olson, E.N. Development 132, 3525–3536 (2005).

Late third instar

OK107-gal4

a b c d

e f g h

i j k l

n o p

OK107-gal4

upd-gal4upd-gal4

hh-gal4

hand-gal4 hand-gal4

hh-gal4

Late 1IL

Late 1IL

Embryo

LG

DVm

2IL

RFP fluorescence,current

RFP fluorescence,current

RFP fluorescence, current

EGFP fluorescence, lineage

RFP fluorescence,current

EGFP fluorescence,lineage

Earlier developmental stages

Figure 3 | Stage dependence of G-TRACE patterns. (a–p) Fluorescence images

showing G-TRACE analysis with OK107-gal4 (a–d), upd-gal4 (e–h), hh-gal4

(i–l) and hand-gal4 (m–p) at late and early developmental stages, as

indicated. DV, dorsal vessel; LG, lymph gland; line surrounds cardioblasts of

the embryonic DV in o,p. 1IL, first instar larvae; 2IL, second instar larvae.

DNA is stained with ToPro-3 iodide (blue). Scale bars, 50 mm (a,b,e,f,i,j,m,n),

25 mm (c,d,g,h,k,l), and 10 mm (o,p).

NATURE METHODS | VOL.6 NO.8 | AUGUST 2009 | 605

BRIEF COMMUNICATIONS

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

ONLINE METHODSDrosophila lines. The uncharacterized Gal4 enhancer trap linesscreened in this study were obtained from the Gal4 Enhancer TrapInsertion Database, a subdivision of the Drosophila GeneticResource Center, Kyoto Institute of Technology. The followinglines were obtained from the Bloomington Drosophila StockCenter: hh-lacZ, ptc-gal4, ptc-lacZ, dpp-gal4, dpp-lacZ andOK107-gal4. Other drivers used in this study were obtained fromthe indicated individuals: Antp-gal4 (S. Cohen, Temasek LifeSciences Laboratory, Singapore), Cg-gal4 (C. Dearlolf, Massachu-setts General Hospital), dome-gal4 (S. Noselli, Centre National dela Recherche Scientifique, Nice), Dot-gal4 (D. Kimbrell, Universityof California, Davis), gcm-gal4 (L. Zipursky, University of Cali-fornia, Los Angeles), hand-gal4 (Z. Han, University of Michigan),hh-gal4 (T. Kornberg, University of California, San Francisco),upd-gal4 (Y.H. Sun, Academia Sinica, Taiwan).

Molecular cloning and the generation of the G-TRACE system.Approximately 2 kb of DNA immediately upstream of the Ubi-p63E (CG11624) open reading frame was PCR-amplified as anEcoRI-KpnI fragment from a previously cloned promoter region(provided by P. O’Farrell, University of California, San Francisco)and cloned into the pCRII vector (Invitrogen). Primer sequencesare available in Supplementary Table 5. This EcoRI-KpnI fragmentwas then subcloned into the Drosophila transformation vectorpStinger (Drosophila Genomics Resource Center), which contains

a multiple cloning site upstream of the nuclear EGFP readingframe. A ‘stop cassette’, containing a transcriptional terminationsequence flanked by an inverted pair of FRT recombination sites,was isolated as a KpnI restriction fragment from the constructD237 (kindly provided by G. Struhl, Columbia University) andsubcloned into the KpnI site between the Ubi-p63E promoter andthe EGFP reading frame. Transgenic flies were generated using thepUbi-63E-STOP-Stinger vector and a commercial transformationservice (BestGene, Inc.). A triple-recombinant chromosomewas generated using UAS-RedStinger14 (w1118; P{UAS-RedStinger}4/CyO, Bloomington Drosophila Stock Center), UAS-FLP (y1 w*;P{UAS-FLP1.D}JD1, Bloomington Drosophila Stock Center) andpUbi-63E-STOP-Stinger insert 9F6 and was balanced over CyO.

Analysis of Gal4 expression patterns. The Gal4-expressing lineswere individually crossed on standard fly food at 25 1C. Wander-ing third instar larvae were dissected in 1� PBS (pH 7.4) andtissues were fixed in 4% formaldehyde in 1� PBS for 30 min,followed by three washes (10 min each) with 1� PBS and 0.4%Triton-X 100 (Fisher) detergent (PBST). During the second 1�PBST wash, To-Pro3 iodide (Invitrogen) was added to stain nuclei.Tissues were mounted on slides in Vectashield (Vector Labora-tories) and imaged using a Zeiss Apotome or a BioRad Radiance2000 confocal microscope with standard techniques.

14. Barolo, S., Castro, B. & Posakony, J.W. Biotechniques 36, 436–440, 442 (2004).

NATURE METHODS doi:10.1038/nmeth.1356

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.