LETTERS

Chemical rescue of cleft palate and midline defects inconditional GSK-3b miceKaren J. Liu1,2*{, Joseph R. Arron1*{, Kryn Stankunas3, Gerald R. Crabtree1,3,4 & Michael T. Longaker2,5

Glycogen synthase kinase-3b (GSK-3b) has integral roles in a vari-ety of biological processes, including development, diabetes, andthe progression of Alzheimer’s disease1–4. As such, a thoroughunderstanding of GSK-3b function will have a broad impact onhuman biology and therapeutics. Because GSK-3b interacts withmany different pathways, its specific developmental roles remainunclear5. We have discovered a genetic requirement for GSK-3b inmidline development. Homozygous null mice display cleft palate,incomplete fusion of the ribs at the midline and bifid sternum aswell as delayed sternal ossification. Using a chemically regulatedallele of GSK-3b (ref. 6), we have defined requirements for GSK-3bactivity during discrete temporal windows in palatogenesis andskeletogenesis. The rapamycin-dependent allele of GSK-3b pro-duces GSK-3b fused to a tag, FRB* (FKBP/rapamycin binding),resulting in a rapidly destabilized chimaeric protein. In theabsence of drug, GSK-3bFRB*/FRB* mutants appear phenotypicallyidentical to GSK-3b2/2 mutants. In the presence of drug, GSK-3bFRB* is rapidly stabilized, restoring protein levels and activity6.Using this system, mutant phenotypes were rescued by restoringendogenous GSK-3b activity during two distinct periods in gesta-tion. This technology provides a powerful tool for defining win-dows of protein function during development.

Common congenital birth defects such as cleft palate, which affectsroughly one in 2,000 births and is associated with approximately 400known human syndromes, have multifactorial genetic and environ-mental causes7. Mouse genetic models provide insight into thecomplex biological interactions that go awry during developmentto produce human disease. However, existing genetic methods,which depend primarily on deletion of target genes, provide limitedtemporal resolution. To improve studies of embryonic development,we have devised a strategy combining pharmacologic and geneticmanipulation of developmentally relevant genes. The advantage ofpharmacologic intervention is that, by targeting at the protein level,gene function can be rapidly and reversibly controlled. Recent appli-cations of small molecules to developmental studies include the use ofcyclopamine to define the roles of Hedgehog signalling8 in develop-ment and cancer, and cyclosporine to characterize calcineurin func-tion during angiogenesis and neurogenesis9,10. Chemical geneticapproaches have also identified new biologically active small mole-cules11–13. However, isolating a specific inhibitor of a target of choicecan be prohibitively laborious and expensive. To circumvent thisdifficulty, we have developed a generalized technology for makinghighly specific drug-dependent alleles using a destabilizing tag(FRB*) and we have applied it to GSK-3b (ref. 6). The 89 amino acidFRB* domain is thermally unstable. Fusion to FRB* transfers thisinstability and reduces the melting temperature of the target protein.

Thus, FRB*-tagged proteins are also unstable and rapidly degraded.Instability of the chimaeric protein is reversed by rapamycin bindingto the FRB* domain6 (Supplementary Fig. 1a–c).

GSK-3b has regulatory roles in many developmentally importantmolecular pathways, including Wnt, NFAT, Hedgehog and insulinsignalling2. GSK-3b homozygous null mice had previously beenreported to die during mid-gestation owing to liver degeneration5.We found, to our surprise, that GSK-3b2/2 mice survived gestationand died perinatally. All of these mice had a complete cleft of thesecondary palate (Fig. 1c); the observed phenotypic range is shown(n 5 17, Fig. 1e). In mice, palatal development occurs between stagese12.5–e15 and is a process in which the bilateral maxillary processesdescend vertically from the maxilla. Subsequently, the palatal shelvesrotate horizontally, meet at the midline, and fuse by the time of e15(schematized in coronal cross-sections in Fig. 1a). Clefts of the sec-ondary palate result from an inability of the palatal shelves to differ-entiate, grow, rotate or fuse14. Mice homozygous for the FRB*-tagged

*These authors contributed equally to this work.

1The Department of Pathology, 2The Stanford Institute for Stem Cell Biology and Regenerative Medicine, 3Department of Developmental Biology, 4Howard Hughes Medical Institute,and the 5Division of Plastic and Reconstructive Surgery, Department of Surgery, Stanford University School of Medicine, Palo Alto, California 94305, USA. {Present addresses:Department of Craniofacial Development, King’s College London, London SE1 9RT, UK (K.J.L.); Department of Immunology Diagnostics, Genentech, 1 DNA Way, South San Francisco,California 94080, USA (J.R.A.).

WT

GSK-3β–/–

GSK-3β–/–

β F*/F* β–/F*

e13.5 e14.5 e15.5e12.5

TP P

Descending Rotating Meeting Fusing

SP

PP

WT

T T

P P

ab c

d e f gGSK-3 GSK-3

Figure 1 | GSK-3b mutants have cleft palates. a, Schematic of palate fusionin coronal cross-section; palate closure occurs between e12.5 and e15.5. Ate12.5, palatal shelves (P, blue) grow downward from the maxillary processes,lateral to the tongue (T). Palatal shelves have rotated and elevated above thetongue by stage e13.5, extended towards the midline at e14.5, and fused bye15.5. b–e, Palate closure. b, Palatal shelves (black arrow) are fully fused ine16.5 wild-type (WT) mice (coronal cross-sections, stained withhaematoxylin and eosin). c, In GSK-3b2/2 mice, palatal shelves (blackarrow) have rotated but do not meet at the midline. d, Palates are fully fusedin e18.5 wild-type mice. The roof of the mouth is pictured (anterior at top ofimage). PP, primary palate; SP, secondary palate; midline marked with blackarrowhead. e–g, GSK-3b2/2 (e), GSK-3bFRB*/FRB* (f) and GSK-3b2/FRB*

(g) mice have cleft palates; e shows representatives of the range of cleftsobserved. (F* is FRB*.)

Vol 446 | 1 March 2007 | doi:10.1038/nature05557

79Nature ©2007 Publishing Group

allele, GSK-3bFRB*/FRB* (n 5 58), all had identical palate defects tonull mice (Fig. 1f), as did 22 out of 23 trans-heterozygotes (GSK-3b2/FRB*, n 5 23, Fig. 1g). At birth, all mutants were present atroughly mendelian ratios (GSK-3b2/2, 22% or 45/203; GSK-3bFRB*/FRB*, 28% or 82/288; GSK-3b2/FRB*, 26% or 23/90). We didnot observe any cleft palates in wild type or heterozygous litter-mates (GSK-3b1/1, n 5 167; GSK-3b1/FRB*, n 5 149; GSK-3b1/2,n 5 115).

Because many signalling pathways regulated by GSK-3 affectosteogenesis15,16, we stained cartilage with alcian blue and ossifiedbone with alizarin red. The three classes of mutants (GSK-3b2/2,GSK-3bFRB*/FRB* and GSK-3b2/FRB*) displayed comparable arraysof defects in skeletal development and ossification. We observed anumber of anomalies including delayed ossification of the skull, ear

bones and cranial base (K.J.L. and M.T.L., unpublished observa-tions). In the present study, we focus on sternal fusion and ossifi-cation. Sternal development occurs later in gestation than palatefusion. The sternal body develops from two mesenchymal bars thatmigrate ventrally during development, meet and fuse at the midline(e13.5 to e16.5). Subsequently, ossification centres (marked in red tomimic alizarin red staining of bone) arise at the points of contactbetween the ribs and the sternum17 (schematic in Fig. 2a; devel-opmental progression of ossification shown in wild-type animals inFig. 2b). In contrast to wild-type animals (Fig. 2b), in GSK-3b2/2

mice, the sternal bars were frequently bifurcated and the appearanceof ossification centres was markedly delayed (Fig. 2d). Often therewere holes in the xiphoid cartilage (asterisk, Fig. 2d). GSK-3bFRB*/FRB* and GSK-3b2/FRB* were identical to GSK-3b2/2 withregard to the sternal phenotype (Fig. 2d, e and data not shown).The delay in ossification was most obvious at early stages; by e18.5,mutants sometimes showed levels of ossification comparable to wild-type, although often in aberrant locations (compare Fig. 2d, e toFig. 2b, c). We examined cell proliferation and death, which areknown to be important in palatal fusion18. There was a mild butstatistically insignificant increase in cell proliferation in mutant ani-mals (Supplementary Fig. 2a). Neither the palatal nor sternal defectsseemed to result from aberrant cell death: both wild-type and mutantanimals showed minimal cell death (Supplementary Fig. 2f, g anddata not shown).

Having confirmed that in the absence of drug treatment the FRB*-tagged allele of GSK-3b mimics the phenotypes of a conventionalknockout, we set out to rescue GSK-3b activity during the criticalperiods of palatogenesis and skeletogenesis. In the presence of rapa-mycin or rapamycin analogues, FRB*-tagged proteins dimerize withendogenous FK506 binding proteins (FKBPs), thus stabilizing thefusion protein and restoring protein levels and activity (Supplemen-tary Fig. 1a–c)6. Because the FRB*-tag was inserted as a knock-in tothe GSK-3b locus, any protein activity restored by drug treatmentresults from expression encoded by the endogenous locus and reflectsendogenous expression and activity, both temporally and spatially. Inthe absence of drug treatment, there is minimal protein expressionand activity6.

Teratogenicity had previously been described in mouse embryostreated with rapamycin during early gestation stages (up to e10.5)19.Forty-eight hour drug treatments from e13.5 onward bypassed theperiod of rapamycin sensitivity and wild-type embryos showed noovert phenotypic abnormalities (Supplementary Fig. 1g) or skeletaldefects. As a further control, all phenotypic analysis was performedby comparison with littermate controls as well as conventionalknockout animals (GSK-3b2/2) treated with rapamycin, and treatedanimals showed no phenotypic abnormalities beyond those observedin knockout animals. However, it is important to note the possibilitythat rapamycin has effects on other tissues not examined in thisstudy.

We wanted to ensure that any rapamycin-dependent rescue ofGSK-3b mutant phenotypes correlated with accessibility of the drugin utero, with the perdurance of rapamycin after the final treatment,and with stabilization of GSK-3b. From each litter treated, we har-vested brain, liver and placental samples for pharmacokinetic andprotein analysis. In brain samples harvested seven hours after thelast dose administered, rapamycin was detectable (data not shown)and GSK-3bFRB* was clearly stabilized (Supplementary Fig. 1d).Although residual rapamycin was detected in embryos 15–27 h afterthe final dose (data not shown), we were unable to detect stabilizationof GSK-3bFRB* 15 h after the last dose (Supplementary Fig. 1e)indicating that GSK-3bFRB* stabilization was reversed after drugtreatment was stopped. Thus, we can achieve a high degree of tem-poral precision of protein stabilization during specific stages ofembryonic development.

Our results demonstrate that GSK-3bFRB* was stabilized bymaternal rapamycin treatment (Supplementary Fig. 1d, f). We

e17.5 e18.5 p0 e18.5

e17.5 e18.5 p0 e18.5

Xiphoid cartilage

Manubrium

WT

*

e13.5 e14.5 e15.5 e16.5 e17.5 e18.5a

b

d e

c

e12.5

WT

GSK-3βF*/F*GSK-3β–/–

Figure 2 | GSK-3b mutants show bifid sternum and delayed sternalossification. a, Schematic of sternal fusion: sternal bars form as twocartilaginous processes medial to the ribs. Sternal bars migrate towards themidline (from e13.5 on), and meet and fuse to form the sternum (by e16.5).Ossification (marked in red to depict alizarin red staining) occurscraniocaudally and continues after birth. b, In wild-type animals thesternum is fused by e17.5; ossification centres appear in a stereotypicalfashion and progress after birth. Representative animals are shown frome17.5 through birth. Animals in b are littermates of those shown incorresponding columns in d. Alizarin red stains ossified bone; alcian bluestains cartilage. c, Wild-type (e18.5); littermate control for e. d, GSK-3b2/2

mutants display aberrant fusion of the sternebrae resulting in nearlycompletely bifid sterna. Xiphoid cartilage (black arrowheads) often has holes(asterisk). e, GSK-3bFRB*/FRB* mutants phenocopy GSK-3b2/2 mutants.

LETTERS NATURE | Vol 446 | 1 March 2007

80Nature ©2007 Publishing Group

did not observe developmental defects in untreated heterozygotes,indicating that 50% of wild type protein was sufficient for properembryogenesis. This led us to predict that stabilizing a portion of thepool of GSK-3b (Supplementary Fig. 1d) might be enough to rescuethe mutant phenotype and would provide a temporal mapping ofGSK-3b requirements during embryogenesis.

In GSK-3b mutants, the palatal shelves clearly descended androtated but did not meet and fuse at the midline (Fig. 1c). There-fore, we predicted that rescue of cleft palate would require stabi-lization of GSK-3bFRB* at or around e14.5 (Fig. 1a). We dosedpregnant dams with rapamycin (5 mg kg21 every 12 h for a total of4 doses) in two-day windows beginning at e13.5–e15.0 and up toe16.5–e18.0. Indeed, administration of rapamycin at the earliest timepoints tested (e13.5–e15.0) rescued palatogenesis completely in56% (5/9) and partially in 11% (1/9) of treated GSK-3bFRB*/FRB*

embryos (Fig. 3b, c). Wild-type animals treated with rapamycin closedtheir palates normally (Fig. 3a). GSK-3b2/2 embryos were not rescuedby parallel drug treatment (Fig. 3c; 0/2 rescued). Consistent with ourhypothesis that GSK-3b is required during a critical early step in palatemorphogenesis, we saw no restoration of palate closure when treatingduring two-day windows beginning e14.5, e15.5 or e16.5 (Fig. 3c; 0/23rescued). This identifies the essential developmental window for GSK-3b in palatogenesis between e13.5 to e15.0.

We next examined the sternal phenotype in rapamycin-treatedanimals. To ensure that rapamycin had no effect on skeletal develop-ment, we first examined treated wild-type animals by alcian blue andalizarin red staining and determined that normal sternal develop-ment was unaffected (Fig. 4a, c, e and g). As shown in Fig. 2b–e,mutant animals had bifurcated sternebrae and delayed sternalossification. We predicted that GSK-3b might be required for bothfusion and ossification, or it might be involved only in fusion,implying that the delay in ossification was secondary to a fusiondefect. Using drug treatments, we defined the critical period forGSK-3b in sternal development as e15.5 to e17.0 (Fig. 4a, b). Thisis consistent with the onset of ossification occurring at e17 andcontinuing after birth17. GSK-3bFRB*/FRB* animals treated duringthis period of gestation displayed an onset of ossification (Fig. 4b,i; 7/7 rescued) comparable with littermate controls. Many also dis-played fully fused sterna (4/7 rescued) and those that did not hadonly minor caudal clefts (3/7 partially rescued). Treatment at the

earliest time point (e13.5–e15.0) failed to restore either sternalfusion or ossification (Fig. 4f), whereas embryos treated frome14.5 to e16.0 showed some improvement in sternal ossification,but not to the level observed with e15.5 to e17.0 (Fig. 4h, i). Ourdata suggest that although there was a correlation between thedegree of sternal fusion and the onset of ossification, full ossificationdid not require complete fusion of the sternum (compare e14.5–e16treatment with e15.5–e17 treatment, Fig. 4i). Again, in comparison,GSK-3b2/2 mice treated for the same period with rapamycin hadthe full loss-of-function phenotype (Fig. 4d, 0/2 rescued). This resultmaps GSK-3b function to a temporal window consistent with a rolefor GSK-3b function in sternal fusion and onset of ossificationbetween e15.5 and e17.0.

The range of rescue seen is likely to be due to variation in thestabilization of GSK-3bFRB*. These differences may be due to in-consistency of drug penetrance resulting from differences in trans-placental blood flow related to embryo position in the uterinehorns20. There is also inherent variability in the timing of organogen-esis from embryo to embryo and from litter to litter, as seen in therange of palatal (Fig. 1) and sternal (Fig. 2) phenotypes observed inuntreated mutant embryos.

Although GSK-3b functions had previously been studied usingconventional gene targeting strategies, an observation of early leth-ality precluded a more careful analysis of GSK-3b’s roles in laterdevelopment. We did not observe the reported early lethality pheno-type, and comparison of the drug-dependent conditional allele withthe existing knockout allele has allowed us to identify novel develop-mental roles for GSK-3b and confirmed the specificity of the con-ditional allele. In the future, combining traditional knockout alleleswith promoter-regulated FRB*-tagged transgenics or conditionalknock-ins of FRB*-tags will allow additional spatial and temporalrefinement of our studies. Drug-dependent rescue of developmentaldefects will provide a powerful tool to study GSK-3b requirementspostnatally and allow us to learn more about GSK-3’s roles in develop-ment and disease. Finally, our data showing the feasibility of small-molecule-based chemical genetics strategies have prospective clinicalimplications. New approaches to rescuing selected developmentaldefects require detailed knowledge of timing and levels of proteinexpression; our studies provide an improved method for definingthese experimental conditions in vivo.

Timelineof drug addition

WT + rapamycin

Embryos analysed at

5 out of 9 1 out of 9 3 out of 9

Later treatments do not rescue (n = 23)

or GSK-3β–/–

a cb

e12.5

Untreated

Phenotypes of rapamycin-treated animals

Vehicle-treated animals

(only includes e17.5 to p0)

CleftGSK-3β–/– 17 0 0

58 0 022 0 1

Partial Closed

e13.5–e15

e14.5–e16

e15.5–e17

e16.5–e18

3

4

16

3

006

00e13.5–e15 2

0

0

0

1

0

0

0

5

e13.5 e14.5 e15.5 e16.5 e18.5 e17.5

GSK-3βF*/F* + rapamycin

GSK-3βF*/F*

GSK-3βF*/F*

GSK-3βF*/F*

GSK-3β–/F*

Figure 3 | Drug-dependent rescue of cleft palate GSK-3bFRB*/FRB* mice inutero. Embryos were treated with four doses of rapamycin at 5 mg kg21 inutero at the indicated time points. Animals were euthanized at e18.5 andscored for palate closure. a, Palates are fully fused in wild-type littermatestreated with rapamycin. Midlines are marked with yellow arrowheads.b, Rapamycin treatment in utero from e13.5 to e15 rescues palatogenesis inGSK-3bFRB*/FRB* embryos. Full rescue observed in 5/9, 1/9 showed a smallcleft in the posterior part of the palate (middle panel, red arrowhead) and 3/9were not rescued. Solid bar in timeline below indicates period of rapamycin

treatment; triangle indicates gestational age at harvest. Red bar correspondswith pictured experiments. Treatment at later time points did not rescuecleft palate (indicated by black bars, n 5 23). c, Distribution of cleft palate inmutant animals. GSK-3b2/2, GSK-3bFRB*/FRB* and GSK-3b2/FRB* mutantshad complete clefts of the secondary palate (97/98 animals). In uterotreatment of GSK-3bFRB*/FRB* embryos (e13.5–e15.0) rescued cleft palate,whereas later treatments did not. Neither GSK-3b2/2 animals nor animalstreated with vehicle (300ml of injection solution—10% polyethylene glycol400, 17% Tween-80 in H2O) were rescued.

NATURE | Vol 446 | 1 March 2007 LETTERS

81Nature ©2007 Publishing Group

METHODSMice. Generation of drug-dependent GSK-3bFRB* mice was described prev-

iously6. These mice will be made available at the Jackson Laboratory (JAX stock

number JR6824, strain GSK3btm1Grc/J). Conventional mutant alleles of GSK-3

(GSK-3b2/1) were a gift from J. Woodgett and have been described previously5.

All of the experiments shown were performed in outbred CD-1 mice; however,

we found the same cleft palate phenotype in the original GSK-3b2/2 and GSK-

3bFRB*/FRB* mice and observed neither hepatic defects nor mid-gestational leth-

ality. Alcian blue and alizarin red staining of bones and cartilage were performed

according to established protocols.

Drug treatments. Rapamycin (sirolimus) was resuspended in a stock solution at20 mg ml21 in N,N-dimethylacetamide (DMA) and stored at –20 uC until use.

Pregnant dams were treated with subcutaneous injections every 12 h during the

time periods indicated. Each injection consisted of 5 mg kg rapamycin21 diluted

in 200ml of injection vehicle (10% polyethylene glycol 400, 17% Tween-80). For

vehicle controls, animals were injected with 200ml of injection vehicle (every 12 h

either for two days or continuously from e13.5 to e18.5) and displayed no

phenotypic abnormalities.

Received 4 August; accepted 27 December 2006.Published online 11 February 2007.

1. Cohen, P. & Frame, S. The renaissance of GSK3. Nature Rev. Mol. Cell Biol. 2,769–776 (2001).

2. Frame, S. & Cohen, P. GSK3 takes centre stage more than 20 years after itsdiscovery. Biochem. J. 359, 1–16 (2001).

3. Doble, B. W. & Woodgett, J. R. GSK-3: tricks of the trade for a multi-tasking kinase.J. Cell Sci. 116, 1175–1186 (2003).

4. Jope, R. S. & Johnson, G. V. The glamour and gloom of glycogen synthase kinase-3.Trends Biochem. Sci. 29, 95–102 (2004).

5. Hoeflich, K. P. et al. Requirement for glycogen synthase kinase-3b in cell survivaland NF-kB activation. Nature 406, 86–90 (2000).

6. Stankunas, K. et al. Conditional protein alleles using knockin mice and a chemicalinducer of dimerization. Mol. Cell 12, 1615–1624 (2003).

7. Murray, J. C. & Schutte, B. C. Cleft palate: players, pathways, and pursuits. J. Clin.Invest. 113, 1676–1678 (2004).

8. Chen, J. K., Taipale, J., Cooper, M. K. & Beachy, P. A. Inhibition of Hedgehogsignaling by direct binding of cyclopamine to Smoothened. Genes Dev. 16,2743–2748 (2002).

9. Graef, I. A., Chen, F., Chen, L., Kuo, A. & Crabtree, G. R. Signals transduced byCa21/calcineurin and NFATc3/c4 pattern the developing vasculature. Cell 105,863–875 (2001).

10. Graef, I. A. et al. Neurotrophins and netrins require calcineurin/NFAT signaling tostimulate outgrowth of embryonic axons. Cell 113, 657–670 (2003).

11. Peterson, R. T. et al. Chemical suppression of a genetic mutation in a zebrafishmodel of aortic coarctation. Nature Biotechnol. 22, 595–599 (2004).

12. Peterson, R. T., Link, B. A., Dowling, J. E. & Schreiber, S. L. Small moleculedevelopmental screens reveal the logic and timing of vertebrate development.Proc. Natl Acad. Sci. USA 97, 12965–12969 (2000).

13. Liu, J. et al. A small-molecule agonist of the Wnt signaling pathway. Angew. Chem.Int. Ed. 44, 1987–1990 (2005).

14. Sharpe, P. M. & Ferguson, M. W. Mesenchymal influences on epithelialdifferentiation in developing systems. J. Cell Sci. 10 (suppl.), 195–230 (1988).

15. Karsenty, G. & Wagner, E. F. Reaching a genetic and molecular understanding ofskeletal development. Dev. Cell 2, 389–406 (2002).

16. Krishnan, V., Bryant, H. U. & Macdougald, O. A. Regulation of bone mass by Wntsignaling. J. Clin. Invest. 116, 1202–1209 (2006).

17. Chen, J. M. Studies on the morphogenesis of the mouse sternum. I. Normalembryonic development. J. Anat. 86, 373–386 (1952).

18. Ito, Y. et al. Conditional inactivation of Tgfbr2 in cranial neural crest causes cleftpalate and calvaria defects. Development 130, 5269–5280 (2003).

19. Hentges, K. E. et al. FRAP/mTOR is required for proliferation and patterningduring embryonic development in the mouse. Proc. Natl Acad. Sci. USA 98,13796–13801 (2001).

20. Flake, A. W., Villa, R. L., Adzick, N. S. & Harrison, M. R. Transamniotic fetal feeding.II. A model of intrauterine growth retardation using the relationship of ‘‘naturalrunting’’ to uterine position. J. Pediatr. Surg. 22, 816–819 (1987).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank M. S. Dionne for critical reading of the manuscript;M. S. Dionne, M. M. Winslow, J. E. Gestwicki, J. H. Bayle, S. C. Kao and members of theLongaker and Crabtree laboratories for invaluable discussions; and J. Woodgett forthe gift of GSK-3b knockout mice. K.J.L and M.T.L. are supported by the NIH, M.T.L. isalso supported by the Oak Foundation and J.R.A. is a fellow of the Berry Foundation.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Correspondence and requests for materials should be addressed to K.J.L.(karen.liu@kcl.ac.uk) or M.T.L. (longaker@stanford.edu).

e13.5 e14.5 e15.5 e16.5e12.5

WT + Rap F*/F* + Rap WT + Rap –/– + Rap

Timelineof drug addition

Timelineof drug addition

e13.5 e14.5 e15.5 e16.5e12.5 e17.5 e18.5

WT + Rap F*/F* + Rap WT + Rap F*/F* + Rap

e17.5 e18.5

a

e

i

f g h

b c d

GSK-3β–/–

GSK-3β–/–

GSK-3β

GSK-3β

GSK-3β

GSK-3β–/–

GSK-3β

GSK-3ββ

Untreated Bifid Fused Delay Bifid Fused Delay+ Rapamycin

e17.5 e13.5–e15

e14.5–e16

e15.5–e17

e15.5–e17

e16.5–e18

e18.5

14/16 2/16 11/162/3

6/7

0/3

2/7

3/3

5/7

p0 3/60/66/6

e17.5

e18.5

11/11 0/11 3/112/4

1/1

0/4

0/1

4/4

4/4

4/4

3/7

1/3

0/4

0/4

4/7

2/3

4/4

1/4

0/7

2/3

2/2 0/2 2/2

GSK-3β–/–

Vehicle-treated animals

2/2 0/2 2/2

1/1

e18.5

p0

6/8 2/8 6/81/1

5/7

0/1

2/7

1/1

5/7

p0 2/60/66/6F*/F*

–/F*

F*/F*

Figure 4 | Reversal of sternal defects in chemically sensitive GSK-3bembryos. Embryos were treated in utero with four doses of rapamycin(5 mg kg21 every 12 h) as indicated, euthanized at e18.5, stained withalizarin red (ossified bone) and alcian blue (cartilage) and scored. In eachcase, mutant animals were compared with the littermate controls shown.a, c, e, g, Sternal defects in wild-type embryos are unaffected by treatmentwith rapamycin in utero. Timing of treatments is indicated by schematics.b, GSK-3bFRB*/FRB* mutants treated with rapamycin from e15.5 to e17.0showed improved sternal fusion and ossification. Note the fusedsternebrae (midline marked by orange arrow), robust ossification (fifthossification centre, black arrowhead) and some rescue of ossification of thexiphoid cartilage (red arrowhead). d, Treatment of GSK-3b2/2 embryoswith rapamycin at e15.5 to e17.0 failed to rescue the mutant phenotype.f, Treatment of GSK-3bFRB*/FRB* embryos from e13.5 to e15.0 failed torescue sternal bifurcation. h, Treatment of GSK-3bFRB*/FRB* embryos frome14.5 to e16.0 failed to rescue sternal bifurcation, although the delay inossification is not as severe (1/4 was delayed). Fifth ossification centre,black arrowhead. i, Distribution of sternal defects in mutant animals.Many GSK-3b2/2, GSK-3bFRB*/FRB* and GSK-3b2/FRB* mutants showedbifid sternebrae. Some animals also exhibited delayed onset ofossification. In untreated animals, totals are listed by genotype followed bybreakdown according to gestational age. In rapamycin treated animals, allanimals were scored at e18.5; gestational age listed indicates beginning ofin utero rapamycin treatment. Vehicle controls were treated from e15.5 toe17.0.

LETTERS NATURE | Vol 446 | 1 March 2007

82Nature ©2007 Publishing Group