L E T T E R S

Conventional drug discovery approaches require a prioriselection of an appropriate molecular target, but it is often not obvious which biological pathways must be targeted toreverse a disease phenotype1,2. Phenotype-based screens offerthe potential to identify pathways and potential therapies thatinfluence disease processes. The zebrafish mutation gridlock(grl, affecting the gene hey2) disrupts aortic blood flow in aregion and physiological manner akin to aortic coarctation inhumans3–5. Here we use a whole-organism, phenotype-based,small-molecule screen to discover a class of compounds that suppress the coarctation phenotype and permit survival to adulthood. These compounds function during thespecification and migration of angioblasts. They act toupregulate expression of vascular endothelial growth factor(VEGF), and the activation of the VEGF pathway is sufficient to suppress the gridlock phenotype. Thus, organism-basedscreens allow the discovery of small molecules that amelioratecomplex dysmorphic syndromes even without targeting theaffected gene directly.

Genetic suppressor screens may identify second-site mutations thatmodify the effect of an existing genetic mutation. By screening forphenotypic rescue, this approach identifies components of pathwaysand parallel pathways otherwise difficult to identify by biochemicalmeans6. We reasoned that screens for chemical suppressors of a geneticmutation, carried out in a vertebrate model organism, might provideinsights into vertebrate-specific processes, permit temporal dissectionof their timing and provide small-molecule drug-like leads.

To test the feasibility of this approach, we screened for chemical sup-pressors of the zebrafish gridlock (grl) mutation, which affects thegene hey2. Zebrafish hey2 is an ortholog of the mammalian hey2 gene(also known as HRT2, CHF1, HERP1, HESR2) and encodes a bHLHtranscriptional repressor3. Zebrafish grlm145/m145 mutants, which havehypomorphic mutations in the hey2 gene, suffer from a malformedaorta that prevents circulation to the trunk and tail3,4. The embryosdevelop collateral vessels around the blockage that are sometimes suf-ficient to permit survival. Both the location of the aortic deformities

and the compensatory collaterals are similar to congenital dysplasias ofthe aorta in humans such as coarctation of the aorta5. In the mouse,null mutations of the hey2 gene cause deformities of the major vesselsand also defects of the heart7–9.

We arrayed grlm145/m145 embryos in 96-well plates, exposed them tosmall molecules from a structurally diverse chemical library andexamined their circulation after 48 h of development. Two of 5,000small molecules examined suppressed the gridlock phenotype ingrlm145/m145 embryos (Fig. 1), restoring circulation to the tail (Fig. 1b).Restoration of tail circulation did not involve changes in the numberof collaterals or the timing of their formation. Instead, the morpholog-ical defect in the aorta itself was resolved, and normal flow through theaorta occurred, as confirmed by microangiography (Fig. 1a,b) and his-tology (Fig. 1e,f).

The two suppressors identified are structurally related compounds(Fig. 1c) that rescue the gridlock phenotype in a dose-dependent man-ner (Fig. 1d). The compound GS4012 is a more potent suppressor thanGS3999, and so it was used for all subsequent experiments. Treatmentwith 2.5 µg/ml GS4012 rescues circulation to the tail in approximately55% of grlm145/m145 embryos (167 of 301 embryos tested), whereastreatment with 7.5 µg/ml GS4012 rescues circulation to the tail in allgrlm145/m145 embryos (46 of 46 embryos tested, see Fig. 1d). Other thanthe correction of the aortic defect, no changes in vascular morphologywere apparent when embryos were treated with GS4012 at these con-centrations, although some vascular abnormalities were observed athigher doses.

Angioblasts in zebrafish are first evident around 14 h postfertiliza-tion (h.p.f.), and by 16 h.p.f. they begin to aggregate at the midline10,11.The lumen of the aorta begins to form in the posterior trunk at 17.5 h.p.f., but is not complete until just before the initiation of circu-lation 24 h.p.f. Similarly, the critical period for GS4012 suppressionwas between 12 and 24 h.p.f. (Fig. 2a). Little or no suppression wasobserved when treatment began after 24 h.p.f. When GS4012 wasadded during gastrulation and washed away at later and later timepoints, the effectiveness of suppression was increased, reaching maximal suppression by 30 h.p.f. Suppression was maintained afterremoval of the suppressor, and rescued mutants remained viable

1Developmental Biology Laboratory, Cardiovascular Research Center, Massachusetts General Hospital, Charlestown, Massachusetts 02129, USA and Department ofMedicine, Harvard Medical School, Boston, Massachusetts 02115, USA. 2Howard Hughes Medical Institute, Department of Chemistry and Chemical Biology, HarvardUniversity, Cambridge, Massachusetts 02138, USA. 3Present address: Department of Medicine, Vanderbilt School of Medicine, Nashville, Tennessee 37232, USA.4Present address: Novartis Institutes for Biomedical Research, Cambridge, Massachusetts 02139, USA. Correspondence should be addressed to R.T.P.(peterson@cvrc.mgh.harvard.edu).

Published online 18 April 2004; doi:10.1038/nbt963

Chemical suppression of a genetic mutation in azebrafish model of aortic coarctationRandall T Peterson1, Stanley Y Shaw1,2, Travis A Peterson1, David J Milan1, Tao P Zhong1,3, Stuart L Schreiber2,Calum A MacRae1 & Mark C Fishman1,4

NATURE BIOTECHNOLOGY VOLUME 22 NUMBER 5 MAY 2004 595NATURE BIOTECHNOLOGY VOLUME 22 NUMBER 5 MAY 2004 595

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureb

iote

chno

logy

L E T T E R S

to adulthood. Therefore, suppression is irreversible and largely complete before actual formation of the aorta. Furthermore, the suppressor’s activity is required precisely during the period ofangioblast cell fate determination and migration, suggesting a directrole in these processes.

We used real-time PCR to quantify the expression of several genesbelieved to play roles in angioblast cell fate determination or mig-ration. Expression of one of these genes, VEGF, was upregulated in grlm145/m145 embryos by exposure to GS4012 (Fig. 2b). Treatmentwith GS4012 also increased VEGF mRNA levels in wild-type embryos

596 VOLUME 22 NUMBER 5 MAY 2004 NATURE BIOTECHNOLOGY

N

S

O

CH3

N

S

O2N

b

c d

e f

0

20

40

60

80

100

0 2.5 5 7.5 10

Concentration (µg/ml)

Tail

circ

ulat

ion

(%)

AVA

V

a

Figure 1 Chemical rescue of a genetic cardiovascular defect. (a) Microangiogram of an untreated grlm145/m145 embryo 48 h.p.f. (b) Microangiogram of agrlm145/m145 embryo 48 h.p.f. treated with the small molecule GS4012 beginning 6 h.p.f. (c) Chemical structures of the two small molecules identified asgridlock suppressors by whole-organism chemical screening. Top, GS4012, 4-[2-(4-methoxy-phenylsulfanyl)-ethyl]-pyridine. Bottom, GS3999, 4-[2-(4-nitro-phenylsulfanyl)-ethyl]-pyridine. (d) Dose response curve for suppression of the gridlock mutant phenotype by GS4012. grlm145/m145 embryos were exposed to the indicated concentrations of GS4012 beginning 6 h.p.f. and were transferred to embryo buffer without GS4012 at 30 h.p.f. Embryos were scored for a circulation to the tail 48 h.p.f. Tail circulation percent represents the number of embryos with tail circulation divided by total embryos with circulationanywhere (e.g., cranial circulation). (e–f) Transverse sections comparing the histology of the aorta in grlm145/m145 embryos left untreated (e) or treated with7.5 µg/ml GS4012 (f). A, dorsal aorta; V, cardinal vein.

Figure 2 Chemical suppressors of gridlock act beforeaorta formation and upregulate VEGF expression. (a) Timing of gridlock suppression by GS4012.grlm145/m145 embryos were treated with 2.5 µg/mlGS4012 at the indicated times and allowed to develop to 48 h.p.f. (triangles) or treated with 7.5 µg/ml GS4012at 5 h.p.f. and allowed to develop until the indicatedtimes, at which point they were transferred to embryobuffer without GS4012 and allowed to develop to 48h.p.f. (squares). The rescue percentage for each timepoint represents the difference between the percentagesof wild-type circulation observed in treated and untreatedembryos. (b) The effect of GS4012 on expression ofgenes involved in formation of the aorta. Asterisk, notsignificant. (c) The effect of GS4012 treatment on VEGFmRNA levels as determined by quantitative RT-PCR.Percent change relative to wild-type DMSO-treatedembryos (wt) is shown for grlm145/m145 embryos (grl) and wild-type embryos treated with 5 µg/ml (wt5) or 7.5 µg/ml (wt7.5) GS4012. (d) Expression of hey2(diamonds) and VEGF (squares) mRNA duringdevelopment. Message quantity was normalized toglyceraldehyde 3-phosphate dehydrogenase quantity ateach time point and is expressed in relative units, withthe most abundant time point quantity arbitrarily set at 10. (e) Time course of VEGF induction by GS4012.Fold induction represents the ratio of VEGF mRNA intreated versus untreated embryos. Error bars indicate the standard deviation in three replicates.

–20–10

0102030405060

8

10

6

4

2

0–20–10

0102030405060

2.5

21.5

10.5

0

0

20

40

60

80

100

0 10 20 30 40 50

0 10

wt grl

shh flt4 grl efnB2 vegf

wt5 wt7.5

20 30 40 50

Res

cue

(%)

Per

cent

age

chan

ge

Per

cent

age

chan

geM

essa

ge q

uant

ity(r

elat

ive

units

)

VE

GF

ind

uctio

n (fo

ld)

Treatment length (h)

0 12 24 36 48 60 72

Hours post fertilization

Treatment time (hpf)

P < 0.002

P = 0.003

P < 0.025

****

a b

c

e

d

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureb

iote

chno

logy

L E T T E R S

(Fig. 2c). VEGF plays an important role in formation of the aorta.VEGF122 expression in the Xenopus hypochord acts as a chemoattrac-tant, stimulating angioblast migration to the midline before aorta for-mation12. Targeted disruption of VEGF in mice causes severelydisrupted aortic development13,14, and a VEGF promoter haplotype isassociated with an increased risk for cardiovascular defects (includingcoarctation of the aorta) in humans with deletions in the DiGeorgelocus 22q11 (ref. 15). In zebrafish, hey2 and VEGF transcript levels aretemporally correlated throughout development (Fig. 2d). Both genesare upregulated between 10 and 24 h.p.f., the period of greatest sensi-tivity to gridlock suppression, and are expressed in adjacent tissuesduring this period3,16. VEGF induction by GS4012 treatment isobserved at many stages of development (Fig. 2e).

To test whether the increased VEGF expression induced by GS4012could be responsible for suppression of the gridlock phenotype, weinjected grlm145/m145 embryos with an expression plasmid containingthe murine VEGF cDNA (pmVEGF). Approximately 5% of uninjectedor vector-injected embryos showed wild-type circulation. (The pene-trance of the gridlock phenotype is typically 95% but varies slightlyfrom clutch to clutch, possibly reflecting the existence of unidentifiedmodifier genes). However, about one-third of the mutant embryosinjected with pmVEGF showed wild-type circulation to trunk and tail(Fig. 3). Therefore, VEGF overexpression is sufficient to suppress themutant phenotype in grlm145/m145 embryos. Chemical suppression ofthe gridlock phenotype appears to be more efficient than injection ofpmVEGF. This may be because plasmid-driven VEGF expression ismosaic or because the chemical suppressors have additional mecha-nisms of action beyond VEGF induction.

Because the gridlock suppressors promote vessel formation inzebrafish, we wondered if they would have similar activity in a mam-malian system. We tested the effect of GS4012 on endothelial celltubule formation using cultured human cells. Human umbilical veinendothelial cells (HUVECs) treated with GS4012 formed tubule net-works that appeared more extensive and robust than those formed byuntreated HUVECs (Fig. 4a,b) and were comparable to those formedby HUVECs treated with recombinant VEGF (Fig. 4c,d). Tubulesformed in the presence of GS4012 covered nearly twice as muchmatrix surface area as untreated tubules (Fig. 4d). Therefore, the grid-lock suppressor appears to promote vessel formation in mammalian aswell as zebrafish systems. Given that GS4012 promotes VEGF expres-sion in zebrafish, we reasoned that GS4012 might promote tubule for-mation in human cells through increased VEGF expression. We usedquantitative PCR to test VEGF transcript levels in HUVEC cells. Cellstreated with GS4012 express significantly (P = 0.04) more VEGF thancells treated with dimethyl sulfoxide (DMSO) (Fig. 4e), suggesting thatthe increased VEGF expression in HUVECs treated with GS4012 maycontribute to the increased tubule formation in those cells.

The hey2 gene encodes a transcriptional repressor that drivesangioblast differentiation to an arterial fate rather than a venous fate17.It appears to function downstream of Notch in this bimodal cell fatedecision17–19. VEGF plays several roles in vessel formation, includingstimulating arterial differentiation, expanding the number ofangioblasts and drawing them towards the midline during vessel for-mation12,20,21. Thus, it is possible that rescue of the gridlock mutationby VEGF-inducing GS4012 treatment or by VEGF cDNA injectionreflects stimulation of these compensatory processes. Regardless of theexact nature of the relationship between hey2 and VEGF, it is clear thatthey are related functionally—overexpression of one compensates forhypomorphic mutation of the other. These data are consistent withgenetic studies that place hey2 downstream of VEGF and Notch signal-ing in arterial differentiation pathways20. Although GS4012 induces

expression of VEGF, and VEGF overexpression is sufficient to suppressthe gridlock phenotype, it remains possible that VEGF induction is notthe primary mechanism by which GS4012 suppresses the gridlockphenotype in vivo. For example, in addition to its effects on VEGFexpression, GS4012 may also stimulate compensatory pathways thatare parallel to or downstream of VEGF signaling.

Beyond their utility for developmental biological studies, zebrafishappear to be well suited for modeling genetic diseases. Several muta-tions identified by forward genetic screening cause phenotypes inzebrafish that are similar to human congenital disease phenotypes22,23.In a number of cases where the mutations in human and zebrafishhave been identified, orthologous genes are responsible for the humanand zebrafish phenotypes24,25. Furthermore, the physiological effectsof many drugs have been shown to be comparable in humans andzebrafish26,27. Therefore, insights gained from zebrafish suppressorscreens may be directly applicable to human disease. One shortcomingof conventional approaches to developing therapies for genetic dis-eases has been the difficulty of determining what molecular targetscould compensate for the effects of the mutated genes. This studydemonstrates the feasibility of unbiased screening for such targetsdirectly in a whole, vertebrate organism.

NATURE BIOTECHNOLOGY VOLUME 22 NUMBER 5 MAY 2004 597

b

c

0

5

10

15

20

25

30

35

Circ

ulat

ion

to t

he t

ail (

%)

a

NI pCS2 VEGF

19/418 9/187

74/231

Figure 3 VEGF cDNA injection suppresses the gridlock mutation. (a) Thepercentage of grlm145/m145 embryos showing a circulation to the tail that are not injected (NI), injected with empty plasmid (pCS2) or with plasmidcontaining the murine VEGF cDNA. Circulation was assessed 48 h.p.f., and results are expressed as the number of embryos with a circulation to the tail divided by the number of embryos with a circulation to the head.(b,c) Microangiograms of grlm145/m145 embryos not injected (b) or injectedwith pmVEGF (c).©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

L E T T E R S

Gene therapy has been viewed as a potential means of compensatingfor genetic mutations, but several obstacles have limited the success ofthis approach28. For example, gene transfer to the recipient is ineffi-cient and not uniform, and once the gene is transferred, it remains dif-ficult to control the level of expression of that gene. Concerns have alsobeen raised about the safety of the viral vectors used in gene therapy29.A small molecule–based approach offers a number of advantages.Small molecules can be administered more efficiently and uniformlywithout the use of viral vectors. Most important, dosing can be con-trolled throughout the process, making it possible to maintain aneffect for long periods of time or to terminate exposure as needed.

Recently, it has become possible to isolate zebrafish lines with muta-tions in virtually any gene30. By combining this technology with thesuppressor screening approach described here, it should be possible tomodel many disease processes and identify small molecules capable ofmodifying those processes in new ways.

METHODSZebrafish care. All zebrafish experiments were approved by the MassachusettsGeneral Hospital Subcommittee on Research Animal Care.

Screening for suppressors of the gridlock mutation. Forty pairs of homozy-gous grlm145/m145 zebrafish were mated, and fertilized eggs were arrayed threeembryos per well in 96-well plates containing 0.25 ml E3 embryo buffer (5 mMNaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM MgSO4). At the 50% epibolystage, small molecules from the DiverSet E library (Chembridge) were added toeach well, yielding a final concentration of approximately 2 µg/ml. Embryoswere incubated at 28.5 °C for 48 h. Circulation to the tail was assessed visuallyusing a dissecting microscope.

Quantitative PCR. Total RNA was extracted from groups of embryos using theTRIZOL method and reverse transcribed using random priming and

SuperScript II following the manufacturers’ instructions. Five groups of20 grlm145/m145 embryos were left untreated or treated with 7.5 µg/ml GS4012 at6 h.p.f. (Fig. 2b). Expression was quantified at 24 h.p.f. and is depicted as thepercentage change in transcript levels in GS4012-treated embryos relative totranscript levels in control-treated embryos. The P value represents significancein the pairwise comparison of transcript levels between GS4012-treated andcontrol embryos as determined using the paired t-test. For Figure 2c, fourgroups of 40 embryos were used for each condition. VEGF levels were normal-ized in relation to β-actin expression. P values represent significance as deter-mined using the paired t-test. In Figure 2e, grlm145/m145 embryos were treatedwith 7.5 µg/ml GS4012 at 6 h.p.f. and allowed to develop for the indicatedlength of time. mRNA levels were normalized to β-actin expression. For allexperiments, cDNA was quantified using an Applied Biosystems SequenceDetection System 7000. Taqman probes were used for hey2, VEGF, EfnB2a andglyceraldehyde 3-phosphate dehydrogenase (G3PDH) cDNAs. The SYBR greenmethod was used to quantify sonic hedgehog and Flt4 cDNAs. Primer andprobe sequences used were as follows: hey2, 5′-GTCCGACAGCGACATGGAT-3′, 5′-ACCATTGCTTTGGCCAGAGT-3′, probe 5′-6FAM-AAACCATTGATGTGGGCAGCCAGAATAA-MGBNFQ-3′; zebrafish VEGF, 5′-TGCTCCTGCAAATTCACACAA-3′, 5′-ATCTTGGCTTTTCACATCTGCAA-3′, probe 5′-6FAM-TGCAATGCAAGTCCAGACA-MGBNFQ-3′; human VEGF, 5′-CCAAGGCCAGCACATAGGA-3′, 5′-CTTTCTTTGGTCTGCATTCACATTT-3′, probe 5′-6FAM-ATGAGCTTCCTACAGCAC-MGBNFQ-3′; EfnB2a, 5′-GGAACACCACGAACACCAAGT-3′, 5′-TCCCCAATCTGCGGATACAG-3′, probe 5′-6FAM-TGCCGGGACAGGGTCTGGT-MGBNFQ-3′; zebrafish G3PDH, 5′-GATAACTTTGTCATCGTTGAAGGTCTT-3′, 5′-CGGTCTTCTGTGTTGCTGTGA-3′,probe 5′-VIC-TGAGCACTGTTCATGC-MGBNFQ-3′; human G3PDH, assays-on-demand reagents (Applied Biosystems); sonic hedgehog, 5′-ACAATCCCGACATTATCTTTAAGGA-3′, 5′-TGTCTTTGCATCTCTGTGTCATGA-3′; Flt4,5′-CTGCTCTGGGAGATTTTCTCACTAG-3′, 5′-TCGTTTGCAGAAATCTTCATCAA-3′.

cDNA injections. Embryos generated by mating homozygous grlm145 mutantswere injected at the 1- to 4-cell stage with 1 nl pCS2+ vector or pCS2+-contain-ing murine VEGF cDNA at a concentration of 20 ng/µl in 0.3× Danieau’s buffer(17 mM NaCl, 2 mM KCl, 0.12 mM MgSO4, 1.8 mM Ca(NO3)2 and 1.5 mMHEPES, pH 7.6). Embryos were incubated at 28.5 °C for 48 h. Circulation to thetail was assessed visually using a dissecting microscope.

Fluorescence microangiography. Fluorescence microangiography was done asdescribed4.

Endothelial cell tubule formation assay. Passage 4 HUVECs maintained inEGM-2-MV BulletKit medium (Cambrex) were treated for 11 h with GS4012(5 µg/ml), VEGF (10 ng/ml; R&D Systems) or DMSO dissolved in medium.Cells were trypsinized, resuspended in M199 medium containing 1% FBS(GIBCO-Invitrogen) and either GS4012, VEGF or DMSO at the same concen-trations as before and plated into a 48-well plate precoated with 100 µl ofGrowth Factor Reduced Matrigel (BD Biosciences). After 14 h at 37 °C and 5% CO2, wells were carefully washed once with PBS and photographed under low magnification. The extent of tubule formation was quantified bycounting the number of pixels whose signal intensity exceeded a definedthreshold (MetaMorph software, Universal Imaging). Image processing andanalysis were standardized for all samples and experiments were done in tripli-cate or quadruplicate.

ACKNOWLEDGMENTSThis work was supported by grants from the National Institutes of Health, by the Ned Sahin Research Fund for Restoring Developmental Plasticity and by asponsored research agreement with Novartis Institutes for Biomedical Research.S.Y.S. is a Physician-Postdoctoral Fellow and S.L.S. is an Investigator at the Howard Hughes Medical Institute. S.L.S. is also supported by a grant from theDonald W. Reynolds Cardiovascular Clinical Research Center (University of TexasSouthwestern Medical Center).

COMPETING INTERESTS STATEMENTThe authors declare competing financial interests (see the Nature Biotechnologywebsite for details).

Received 27 October 2003; accepted 13 February 2004

598 VOLUME 22 NUMBER 5 MAY 2004 NATURE BIOTECHNOLOGY

0.20

0.40.60.8

1.0

1.21.41.6

120,000

100,000

80,000

60,000

40,000

20,000

0DMSO GS4012 VEGF NT 2 h 4 h

Tub

ule

form

atio

n(a

rbitr

ary

units

)

Rel

ativ

e V

EG

F ex

pre

ssio

n P = 0.04

a b c

d e

Figure 4 GS4012 promotes endothelial cell tubule formation. (a–c) Net-works of tubules formed by HUVECs treated with DMSO (a), 5 µg/mlGS4012 (b) or 10 ng/ml recombinant VEGF (c). (d) The extent of tubuleformation in HUVECs treated with DMSO, GS4012 or VEGF. Tubuleformation was quantified as described in Methods. Error bars represent the standard deviation. (e) Relative VEGF expression levels in HUVECstreated with DMSO (NT) or with 5 µg/ml GS4012 for 2 h or 4 h. Messagequantity for each sample was normalized to glyceraldehyde 3-phosphatedehydrogenase quantity and is expressed in relative units, with the DMSO-treated value arbitrarily set at 1.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureb

iote

chno

logy

L E T T E R S

Published online at http://www.nature.com/naturebiotechnology/

1. Crews, C.M. & Splittgerber, U. Chemical genetics: exploring and controlling cellularprocesses with chemical probes. Trends Biochem. Sci. 24, 317–320 (1999).

2. Schreiber, S.L. Target-oriented and diversity-oriented organic synthesis in drug dis-covery. Science 287, 1964–1969 (2000).

3. Zhong, T.P., Rosenberg, M., Mohideen, M.A., Weinstein, B. & Fishman, M.C. gridlock,an HLH gene required for assembly of the aorta in zebrafish. Science 287,1820–1824 (2000).

4. Weinstein, B.M., Stemple, D.L., Driever, W. & Fishman, M.C. Gridlock, a localizedheritable vascular patterning defect in the zebrafish. Nat. Med. 1, 1143–1147(1995).

5. Towbin, J.A. & McQuinn, T.C. Gridlock; a model for coarctation of the aorta? Nat.Med. 1, 1141–1142 (1995).

6. St. Johnston, D. The art and design of genetic screens: Drosophila melanogaster. Nat.Rev. Genet. 3, 176–188 (2002).

7. Sakata, Y. et al. Ventricular septal defect and cardiomyopathy in mice lacking thetranscription factor CHF1/Hey2. Proc. Natl. Acad. Sci. USA 99, 16197–16202(2002).

8. Donovan, J., Kordylewska, A., Jan, Y.N. & Utset, M.F. Tetralogy of fallot and other con-genital heart defects in hey2 mutant mice. Curr. Biol. 12, 1605–1610 (2002).

9. Gessler, M. et al. Mouse gridlock: no aortic coarctation or deficiency, but fatal cardiacdefects in hey2–/– mice. Curr. Biol. 183, 37–48 (1997).

10. Eriksson, J. & Lofberg, J. Development of the hypochord and dorsal aorta in thezebrafish embryo (Danio rerio). J. Morphol. 244, 167–176 (2000).

11. Fouquet, B., Weinstein, B.M., Serluca, F.C. & Fishman, M.C. Vessel patterning in theembryo of the zebrafish: guidance by notochord. Dev. Biol. 183, 37–48 (1997).

12. Cleaver, O. & Krieg, P.A. VEGF mediates angioblast migration during development ofthe dorsal aorta in Xenopus. Development 125, 3905–3914 (1998).

13. Carmeliet, P. et al. Abnormal blood vessel development and lethality in embryos lack-ing a single VEGF allele. Nature 380, 435–439 (1996).

14. Ferrara, N. et al. Heterozygous embryonic lethality induced by targeted inactivation ofthe VEGF gene. Nature 380, 439–442 (1996).

15. Stalmans, I. et al. VEGF: a modifier of the del22q11 (DiGeorge) syndrome? Nat. Med.

9, 173–182 (2003).16. Liang, D. et al. The role of vascular endothelial growth factor (VEGF) in vasculogene-

sis, angiogenesis, and hematopoiesis in zebrafish development. Mech. Dev. 108,29–43 (2001).

17. Zhong, T.P., Childs, S., Leu, J.P. & Fishman, M.C. Gridlock signaling pathway fash-ions the first embryonic artery. Nature 414, 216–220 (2001).

18. Iso, T., Chung, G., Hamamori, Y. & Kedes, L. HERP1 is a cell type-specific primarytarget of Notch J. Biol. Chem. 277, 6598–6607 (2002).

19. Nakagawa, O. et al. Members of the HRT family of basic helix-loop-helix proteins actas transcriptional repressors downstream of Notch signaling. Proc. Natl. Acad. Sci.USA 97, 13655–13660 (2000).

20. Lawson, N.D., Vogel, A.M. & Weinstein, B.M. Sonic hedgehog and vascular endothe-lial growth factor act upstream of the Notch pathway during arterial endothelial dif-ferentiation. Dev. Cell 3, 127–136 (2002).

21. Ash, J.D. & Overbeek, P.A. Lens-specific VEGF-A expression induces angioblastmigration and proliferation and stimulates angiogenic remodeling. Dev. Biol. 223,383–398 (2000).

22. Shin, J.T. & Fishman, M.C. From Zebrafish to human: modular medical models.Annu. Rev. Genomics Hum. Genet. 3, 311–340 (2002).

23. Dooley, K. & Zon, L.I. Zebrafish: a model system for the study of human disease. Curr.Opin. Genet. Dev. 10, 252–256 (2000).

24. Xu, X. et al. Cardiomyopathy in zebrafish due to mutation in an alternatively splicedexon of titin. Nat. Genet. 30, 205–209 (2002).

25. Roman, B.L. et al. Disruption of acvrl1 increases endothelial cell number in zebrafishcranial vessels. Development 129, 3009–3019 (2002).

26. Langheinrich, U. Zebrafish: a new model on the pharmaceutical catwalk. Bioessays25, 904–912 (2003).

27. Milan, D.J., Peterson, T.A., Ruskin, J.N., Peterson, R.T. & MacRae, C.A. Drugs thatinduce repolarization abnormalities cause bradycardia in zebrafish. Circulation 107,1355–1358 (2003).

28. Thomas, C.E., Ehrhardt, A. & Kay, M.A. Progress and problems with the use of viralvectors for gene therapy. Nat. Rev. Genet. 4, 346–358 (2003).

29. Check, E. Cancer risk prompts US to curb gene therapy. Nature 422, 7 (2003).30. Wienholds, E., Schulte-Merker, S., Walderich, B. & Plasterk, R.H. Target-selected

inactivation of the zebrafish rag1 gene. Science 297, 99–102 (2002).

NATURE BIOTECHNOLOGY VOLUME 22 NUMBER 5 MAY 2004 599

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureb

iote

chno

logy