Notch signals control the fate of immatureprogenitor cells in the intestineSilvia Fre1,2, Mathilde Huyghe2, Philippos Mourikis1,3, Sylvie Robine2, Daniel Louvard2

& Spyros Artavanis-Tsakonas1,4

The Notch signalling pathway plays a crucial role in specifyingcellular fates in metazoan development by regulating communi-cation between adjacent cells1,2. Correlative studies suggested aninvolvement of Notch in intestinal development. Here, by modu-lating Notch activity in the mouse intestine, we directly implicateNotch signals in intestinal cell lineage specification. We also showthat Notch activation is capable of amplifying the intestinalprogenitor pool while inhibiting cell differentiation. We concludethat Notch activity is required for themaintenance of proliferatingcrypt cells in the intestinal epithelium.The Notch transmembrane receptor is the central element of a

signalling pathway that controls a broad spectrum of metazoan cellfates and developmental processes through local cell interactions1,2. Itis now well established that signals through the Notch receptor areinvolved in the development of several cell types and that themodulation of these signals can markedly affect differentiation,proliferation and apoptotic events3,4. Activation of the pathway hasbeen shown to be a potent inhibitor of differentiation in differentdevelopmental contexts and has been associatedwith the amplificationof some somatic stem cells, such as the neural and haematopoieticstem cells5,6.The intestinal epithelium is a model of tissue renewal from a

source of multipotent stem cells. Throughout adulthood, the pro-liferation, cell fate specification and differentiation of intestinal cellscorrelate with migration along the crypt–villus axis. The rate ofcell turnover in the gastrointestinal tract is remarkably rapid, withnew precursor cells being constantly generated in the crypts ofLieberkuhn. Once they become differentiated, these cells migrateupwards towards the apex of the villi and are eventually shed into thegut lumen.The mechanisms that control the differentiation of epithelial cells

in the intestine remain largely unknown. Nonetheless, the analysis ofmice deficient for the basic helix–loop–helix proteins Hes-1, Math-1and neurogenin-3, all of which are transcriptional targets of Notchsignalling in other tissues7–9, have indirectly implicated the Notchpathway in the regulation of the earliest intestinal cell fate decisions.If these transcription factors are indeed controlled by Notch in thistissue, we reasoned that the activation of the Notch receptor shouldaffect their expression. In addition, such an approach would allow usto evaluate the role of Notch signal activation on the differentiationand proliferation of intestinal precursors. We therefore used the villinpromoter to target the expression of a constitutively active form ofthe mouse Notch 1 receptor (N1ic) in all cells of the intestinalepithelium, including the stem cells of the crypts10–14.Doubly transgenic mice carrying both Villin-Cre14 and Rosa-

Notch15 (Rosa-Notch/Creþ) transgenes (see Supplementary Fig. S1)are born at mendelian ratios, but they die within 3 days of birth

(n ¼ 57). At birth, Rosa-Notch/Creþ mice are macroscopicallyindistinguishable from the Rosa-Notch/Cre2 littermates. Within24 h, however, the Rosa-Notch/Creþ pups become runted, showsigns of malnutrition and display a markedly altered architectureof the intestinal epithelium accompanied by an increase in apoptosis

LETTERS

Figure 1 | Activation of Notch signalling induces apoptosis and impairsgoblet and enteroendocrine cell differentiation. a, b, Paraffin sections of P0Rosa-Notch/Cre2 (a) and Rosa-Notch/Creþ (b) mice were stained with anantibody that recognizes cleaved caspase-3. In villi from the proximalduodenum of Rosa-Notch/Creþ mice several apoptotic figures can be seen(circled in b), whereas Rosa-Notch/Cre2 neonatal intestines do not showapoptotic figures at this developmental stage (a). c, d, Alcian bluestaining reveals several goblet cells in Rosa-Notch/Cre2 duodenum (stainedblue in c) but not in the Rosa-Notch/CreþP0 mice (d). e, f, Representativesections of P0 Rosa-Notch/Cre2 (e) and Rosa-Notch/Creþ (f) mice stainedwith the pan-endocrine marker chromogranin A/B. Several chromogranin-positive cells are evident in the small intestine of Rosa-Notch/Cre2 mice(circled in e), but not in the Rosa-Notch/Creþ mice (f). Scale bar, 80mm.

1Department of Cell Biology, Harvard Medical School, Massachusetts General Hospital Cancer Center, Charlestown, Massachusetts 02129, USA. 2Morphogenesis andIntracellular Signaling, Institut Curie-CNRS, 75248 Paris, France. 3Faculte des Sciences d’Orsay, Universite de Paris-Sud XI, 91405 Orsay, France. 4College de France, 75231 Paris,France.

Vol 435|16 June 2005|doi:10.1038/nature03589

964© 2005 Nature Publishing Group

(Fig. 1a, b). Using secretory cell differentiation markers, we find thatneonatal intestines from postnatal day 0 (P0) Rosa-Notch/Creþmicehave a complete depletion of mucosecreting goblet cells in allintestinal tracts (Fig. 1c, d). In addition, we notice a markedreduction in entero-endocrine cells, as judged by the lack of stainingwith the pan-endocrine marker chromogranin16 (Fig. 1e, f) or withGrimelius silver stain (data not shown). Last, the low expression ofcryptidin-1, as detected by reverse transcriptase polymerase chainreaction (RT–PCR), indicates that the differentiation of the Panethcells17 in the Rosa-Notch/Creþ intestines is also compromised(Fig. 2a). These observations indicate that the villin promotermight drive the expression of activated Notch in early progenitorcells and consequently that the differentiation of all secretory celllineages is inhibited.Microscopic examination of earlier developmental stages revealed

that already at embryonic day 18.5 (E18.5), N1ic expression affectsthe architecture of the villi, the differentiation of secretory celllineages along the duodenal–ileal axis and the cranial-to-caudalwave of intestinal differentiation (see Supplementary Fig. S2). Asnormal antero-posterior development proceeds, the absorptive surfaceof the small intestine is markedly increased by the presence ofnumerous villi. The large intestine, in contrast, lacks villi but ischaracterized by the occurrence of several crypts of Lieberkuhn,which develop as invaginations into the intestinal mucosa (Supple-mentary Fig. S2c). The proximal intestinal tracts of Rosa-Notch/Creþ

embryos have a decreased number of villi (Supplementary Fig. S2b),whereas the distal tracts, corresponding to the large intestine in awild-type animal, show several fingerlike disorganized villous struc-tures and seem clearly different from the normal flat colonicepithelium (Supplementary Fig. S2d).Transcriptional analysis revealed a direct correlation between N1ic

expression and elevated levels of Hes-1 transcription in the intestinalepithelium of Rosa-Notch/Creþ mice (Fig. 2a). The upregulation ofHes-1 is also reflected by the high concentrations of nuclear Hes-1protein seen in all cells lining the intestinal lumen (Fig. 2b, c). Incontrast, N1ic activation in the intestine does not influencethe expression of either Hes-5 or Hey-1, two additional HLHtranscription factor genes that are targeted by Notch signals inother developmental contexts18 (Fig. 2a).

Figure 2 | Notch activation upregulates Hes-1 and represses thetranscription of Math1 and ngn3. a, RT–PCR from P0 intestines showselevated expression of N1ic and Hes-1 in Rosa-Notch/Creþ intestines,compared with the Rosa-Notch/Cre2 controls. Consequently, the levels ofMath1 and ngn3 are repressed. Notch activation represses the Paneth celldifferentiation marker cryptdin-1, whereas the intestinal levels of Hes-5 andHey-1 mRNA are not significantly affected. The expression of the Wntsignalling target genes TCF-4 and LEF-1 is not affected by Notch activation.RNA extracted from the kidneys of the same mice was used to assessspecificity of expression for the intestine. Bottom, b-actin mRNA level wasused as a control. b, c, Immunohistochemical analysis of Hes-1 expression inP0 intestines from Rosa-Notch/Cre2 (b) and Rosa-Notch/Creþ (c) micereveals strong staining for Hes-1 (dark brown) in the nuclei of all cells in theintestinal epithelium of N1ic-expressing mice (c). Note that the goblet cellsin Rosa-Notch/Cre2 intestine show non-specific staining with this antibody(b). Scale bar, 80 mm.

Figure 3 | Activation of Notch expands the population of proliferatingintestinal progenitors. Immunohistochemical analysis of sections of smallintestine from Rosa-Notch/Cre2 (a, c, d) and Rosa-Notch/Creþ (b, e, f) P0mice stained with an antibody against the proliferating cell antigen Ki67(a, b) or with anti-BrdU antibody 24 h after administration of BrdU (c–f).a, b, The proliferating cells (brown) are confined to the crypt domain inRosa-Notch/Cre2 intestine (a), whereas they are detected all along the villusaxis in N1ic-expressing mice (b), where only terminally differentiated cells

normally reside. c–f, BrdU-positive cells (dark brown) have differentiatedand migrated upwards in Rosa-Notch/Cre2 mice (c, d), whereas inRosa-Notch/Creþ sections the number of positively stained cells is increased(e, f). Scale bar, 80 mm (a, b), 60mm (c–f). We note that the morphology ofthe Rosa-Notch/Creþ sample is inferior to that of the wild-type control,indicating a possibly increased sensitivity of the mutant tissue to theexperimental conditions used in these preparations.

NATURE|Vol 435|16 June 2005 LETTERS

965© 2005 Nature Publishing Group

Given these findings, we wished to explore the possibility that thedifferentiation defects observed in Notch transgenic mice wereunderlined by a mechanism whereby the upregulation of Hes-1resulted in the repression of the mouse atonal homologue Math-1(ref. 19) and neurogenin-3 (ngn3)20, both coded by essential genesfor secretory cell lineage specification. Consistent with this model isour finding that the expression of Math-1 and, to a smaller extent,ngn3 is repressed by Notch activation (Fig. 2a). We note that theintestinal phenotype we observe is reminiscent of Math-1 knockoutmice, which lack goblet and enteroendocrine cells8. In contrast,mice lacking Hes-1 activity show an ‘opposite’ phenotype to thatof Math-1 deficient mice, with an excess of secretory cells at theexpense of enterocytes7. Thus the gain of function phenotype wedescribe here provides direct evidence that Notch signals target Hes-1in the intestine, explaining mechanistically the differentiation defectsthat we observe.Because it is known that, depending on the cellular context, Notch

signalling can significantly influence cell proliferation21,22, we soughtto examine the effects of Notch signal activation on cell division inthe intestine. Staining intestinal sections from newborn transgenicmice with the proliferation marker Ki67 (ref. 23) reveals a markedincrease in proliferating cells in the intervillus regions accompaniedby a considerable expansion of the proliferating compartment. Infact, numerous Ki67-positive cells penetrated the villus domain(compare Fig. 3a and b) and we repeatedly detected cells that werestill dividing all along the vertical axis of the villi.This analysis was extended by examining the pattern of migration

of bromodeoxyuridine (BrdU)-labelled cells. The intestine of miceexpressing intracellular Notch1, killed 24 h after BrdU exposure,displays a marked increase in the number of BrdU-positive cellscompared with the Rosa-Notch/Cre2 littermates (Fig. 3c–f). Thedistribution of labelled nuclei in the Rosa-Notch/Cre2 mice reflectsthe apical migration of cells that have started differentiating after theBrdU pulse (Fig. 3c, d). The noticeable increase in the number ofmarked nuclei in the Rosa-Notch/Creþ mice (Fig. 3e, f) is correlatedwith the observed pattern of Ki67 expression and furthermore pointsto an expansion of the population of progenitor cells. However, theseexperiments do not allow us to determine whether the highernumber of BrdU-positive cells is a consequence of an acceleratedcell division rate or results from an inhibition of cell cycle arrest at theedge of the prospective crypt region.To further characterize the nature of the ectopically proliferating

cells in the Rosa-Notch/Creþ mice, intestinal sections of newbornP0 mice were examined by electron microscopy. Ultrastructuralexamination of the apical surface of Notch-expressing intestinalepithelial cells revealed a brush border with a lower density ofmicrovilli than in the wild type (Fig. 4). The microvilli of thecolumnar epithelial cells in Rosa-Notch/Creþ villi (Fig. 4b, d) arenot uniformly distributed and seem less rigid than those in theRosa-Notch/Cre2 littermates (Fig. 4a, c). This phenotype resemblesthe immature brush border of undifferentiated crypt cells24,25, whichis consistent with the notion that Notch activation inhibitsdifferentiation.Our study provides direct evidence that a major effector of Notch

signals in the intestine is Hes-1, in agreement with the previouslyreported loss of function studies of Hes-1 and its targets7,8. However,Hes-1-deficient mice do not show a change in the proliferative statusof the intestinal precursor pool7, whereas Notch activation pro-foundly affects the proliferation potential of intestinal progenitors.In this context, it is worth noting that disruption of Notch signallingby the deletion of the Notch downstream effector CSL (Su(H) inDrosophila) results in what can be described as a ‘reciprocal’ pheno-type. All proliferation ceases in the crypt cells, which differentiatetowards a secretory fate26. In addition, a recent study in zebrafishuncovered a critical role for Notch in the control of intestinal cellcommitment, demonstrating that loss of function mutations in theNotch ligand Delta and block of Notch signals in the mindbomb

mutant background result in a dramatic increase of secretory cells atthe expense of enterocytes27.Depletion of crypt epithelial progenitors in mice lacking TCF-4, a

downstream effector of the Wnt signalling cascade, have previouslyshown the importance of the b-catenin/TCF pathway in the main-tenance of proliferating crypt cells in the intestine28. Given that theWnt/Wingless and Notch pathways are known to cross-talk inDrosophila21, we examined the levels of the Wnt-responsive tran-scription factors TFC-4 and LEF1 in Rosa-Notch/Creþ intestine. Wefind that Notch activation affects neither the transcription of TCF-4or LEF1 (Fig. 2a) nor the nuclear localization of b-catenin in thecrypt cells (data not shown). These observations are consistent withthe epistatic relationship between the Notch and the Wnt pathwayspostulated in vertebrate somitogenesis, in which LEF1 was found toact upstream of Notch/Delta signalling29.We consider the observation that we can affect the proliferation

potential of all cells in the intestinal crypts to be of particularimportance, because it raises the possibility that Notch signalsmight serve as a tool to affect the number of stem cells known toreside in this region, while inhibiting their differentiation. However,because there are no distinguishing cellular markers for the pre-sumptive gastrointestinal stem cells30, nor do we know the molecular

Figure 4 | Apical defect in Rosa-Notch/Cre1 intestinal epithelialcells. Transmission electron microscopy of the apical surface of intestinefrom P0 mice. a, c, The brush border from Rosa-Notch/Cre2 epithelial cellsreveals densely packed, uniformly distributed microvilli. b, d, In Rosa-Notch/Creþ intestinal sections, in contrast, there are several areas, along thevertical axis of the villus, with a lower density of brush border microvilli,reminiscent of the immature microvilli seen in undifferentiated crypt cellsbefore their apical migration. Scale bar, 5mm (a, b), 0.5 mm (c, d).

LETTERS NATURE|Vol 435|16 June 2005

966© 2005 Nature Publishing Group

factors necessary for maintaining the intestinal stem cell niche, it isnot possible at this time to determine rigorously whether ourtransgenic model does indeed affect intestinal stem cells or earlyprecursor cells. Nevertheless, we do know that the villin promoterdriving Cre expression14 targets the intestinal stem cells. Using amouse that carried a tamoxifen-inducible Cre recombinase,expressed under the control of the villin promoter, we were able toinduce transiently the expression of the Cre-responsive Rosa-Notch-IRES-green fluorescent protein (GFP) transgene15 used in this study.Mice killed 4 months after the tamoxifen induction revealed that therecombined Notch-IRES-GFP allele had persisted in intestinal cells,showing that the stem cell compartment had been targeted14 (datanot shown).It is clear that a functional characterization of the cells that are

expanded after Notch activationwill eventually be essential if we are todetermine their true nature and developmental fate.We note, however,that developmental plasticity of cells has been intimately associatedwith proliferation potential and there is therefore the need to examinefurther the possibility that Notch signals might serve as a tool toinfluence the mechanisms affecting the number and fate of earlyprecursors, possibly the self-renewal of stem cells in the intestine.

METHODSTransgenic mice. The Rosa-Notch transgenic mice (provided by C. Murtaughand D. Melton prior to publication of ref. 15) harbour a construct targeting theintracellular domain of mouse Notch 1 (N1ic) to the ubiquitously expressedRosa26 locus15. The N1ic sequence is preceded by a STOP fragment flanked byloxP sites, blocking the expression of N1ic in the absence of Cre recombinase.The N1ic sequence is also followed by internal ribosome entry sequence (IRES)and nuclear enhanced green fluorescent protein (EGFP).Whenmice carrying thisconstruct are crossed with Cre transgenic mice, recombination occurs at the loxPsites, removing the STOPand allowing heritable coexpression of N1ic and nuclearEGFP. The Villin-Cre mice were generated by S.R. and D.L. and carry a 9-kilobaseregulatory region of the mouse villin gene upstream of the Cre recombinasesequence14. Rosa-Notch homozygous or heterozygous mice were crossed toVillin-Cre/þ mice to obtain Rosa-Notch/Creþ and Rosa-Notch/Cre2 progeny.Western blot analysis. Tissue extracts were prepared by homogenization inRIPA (radioimmunoprecipitation assay) buffer (1% Triton X-100, 0.5% deoxy-cholate, 0.1% SDS, 50mM Tris-HCl, 150mM NaCl, 1mM EDTA, 1mM EGTAand a protease/phosphatase inhibitor cocktail). Proteins (100–200mg) wereseparated on 6% SDS–PAGE gels, transferred to Immobilon (Millipore) andprobed with anti-Notch1 antibody (1/1,000 dilution, a gift from J. Aster).Horseradish peroxidase (HRP)-conjugated secondary antibodies were detectedby chemiluminescence.RT–PCR analysis. Total RNA was isolated from freshly dissected intestine andkidney with Trizol reagent (Life Technologies) and complementary DNAsynthesis was performed according to manufacturer’s instructions (SuperScriptkit; Invitrogen). For PCR amplification reactions, the thermocycler profile usedconsisted of an initial denaturation at 96 8C for 1min, followed by 22–28 cyclesof 96 8C for 30 s, 55 8C for 30 s and 72 8C for 40 s. The primers used were asfollows: mouse Notch1, sense 5 0 -GCTGACCTGCGCATGTCTGCCATG-3 0 andantisense 5 0 -CATGTTGTCCTGGATGTTGGCATCTG-3 0 ; mouse Hes-1, sense5 0 -ACACCGGACAAACCAAAGAC-3 0 and antisense 5 0 -GTCACCTCGTTCATGCACTC-3 0 ; Math1, sense 5 0 -GACCACCATCACCTTCGCACCG-3 0 and anti-sense 5 0 -AACTCTCCGTCACTTCTGTGG-3 0 ; crypt1, sense 5 0 -AAGAGACTAAAACTGAGGAGCAGC-3 0 and antisense 5 0 -CGACAGCAGAGCGTGTA-3 0 ;ngn3, sense 5 0 -CGGATGACGCCAAACTTACAAAG-3 0 and antisense 5 0 -CACAAGAAGTCTGAGAACACCAG-3 0 ; mouseHey1, sense 5 0 -GAGAAGGCTGGTACCCAGTG-3 0 and antisense 5 0 -TGGGATGCGTAGTTGTTGAG-3 0 ; mouseHes5, sense 5 0 -AGATGCTCAGTCCCAAGGAG-3 0 and antisense 5 0 -TAGCCCTCGCTGTAGTCCTG-3 0 ; b-actin, sense 5 0 -GACGGCCAGGTCATCACTAT-3 0

and antisense 5 0 -ACATCTGCTGGAAGGTGGAC-3 0 ; TCF-4, sense 5 0 -CGAGATATCAACGAGGCTTTCAAG-3

0and antisense 5

0-CATGTGATTCGCTG

CGTCTCC-30; LEF1, sense 5

0-CTCAACACGAACAGAGAAAGGAGCAGG-3

0

and antisense 50-GTACCTGAAGTCGACTCCTGTAG-3

0.

Histology and immunohistochemistry. Tissues were fixed overnight in 4%neutral-buffered paraformaldehyde at 4 8C, paraffin-embedded and sectioned at4 mm thickness. Sections were stained with haematoxylin and eosin or subjectedto immunohistochemistry with the following primary antibodies: anti-cleavedcaspase-3 (1:200 dilution; Cell Signalling), anti-chromogranin A þ B (1:100dilution; Progen), anti-Hes1 (1:500 dilution; a gift from T. Sudo31), anti-Ki67

(1:100 dilution; Novocastra), anti-GFP (1:200 dilution; Molecular Probes), anti-BrdU (1:200 dilution; Becton Dickinson) and anti-b-catenin (1:50 dilution;Transduction Laboratories). Antigen retrieval was achieved by boiling in 10mMcitrate buffer pH 6.0 (20min) for all antibodies, except anti-BrdU, which usedtreatment with 0.1% trypsin (20min at 37 8C) and anti-b-catenin, whichrequired boiling in Tris-EDTA (1mM Tris, 40mM EDTA) pH9.0 (20min).HRP-conjugated secondary antibodies were detected with the diaminobenzidineperoxidase substrate kit (Vector Laboratories). For the detection of nuclearEGFP, fresh tissues were embedded in OTC (optimum cutting temperature)compound (Tissue-Teck), snap-frozen and sectioned on a cryostat at 8mmthickness. Histochemical identification of intestinal cell types was performed onparaffin sections with periodic acid Schiff, Alcian blue and Grimelius silver stainreagents as recommended by the manufacturer (PolyScientific). BrdU incorpo-ration experiments were performed by injecting pregnant females at day 18.5 ofgestation with BrdU (Sigma) at 100mg per gram animal body weight, dissolvedin sterile PBS. After delivery the newborn mice were killed and processed forimmunohistochemistry as described.Electron microscopy. Tissue samples from the proximal tract of the smallintestine were fixed in 4% glutaraldehyde in sodium cacodylate buffer (0.1M,pH7.4) for 18 h at 4 8C. The tissues were then cut into 5-mm fragments, post-fixed in 1% osmium tetroxide, dehydrated in graded alcohols and embedded inEpon 812. After polymerization overnight at 60 8C, 1-mm sections were preparedand stained with toluidine blue. Representative areas were chosen, thin-sectioned, stained with lead citrate and examined on a Philips 301 electronmicroscope. Images were captured by an Advanced Microscopy Techniquesdigital imaging system.

Received 23 December 2004; accepted 4 April 2005.

1. Artavanis-Tsakonas, S., Rand, M. D. & Lake, R. J. Notch signaling: cell fatecontrol and signal integration in development. Science 284, 770–-776 (1999).

2. Greenwald, I. LIN-12/Notch signaling: lessons from worms and flies. Genes Dev.12, 1751–-1762 (1998).

3. Kimble, J. & Simpson, P. The LIN-12/Notch signaling pathway and itsregulation. Annu. Rev. Cell Dev. Biol. 13, 333–-361 (1997).

4. Weinmaster, G. The ins and outs of notch signaling. Mol. Cell. Neurosci. 9,91–-102 (1997).

5. Shen, Q. et al. Endothelial cells stimulate self-renewal and expand neurogenesisof neural stem cells. Science 304, 1338–-1340 (2004).

6. Varnum-Finney, B. et al. Pluripotent, cytokine-dependent, hematopoietic stemcells are immortalized by constitutive Notch1 signaling. Nature Med. 6,1278–-1281 (2000).

7. Jensen, J. et al. Control of endodermal endocrine development by Hes-1. NatureGenet. 24, 36–-44 (2000).

8. Yang, Q., Bermingham, N. A., Finegold, M. J. & Zoghbi, H. Y. Requirement ofMath1 for secretory cell lineage commitment in the mouse intestine. Science294, 2155–-2158 (2001).

9. Jenny, M. et al. Neurogenin3 is differentially required for endocrine cell fatespecification in the intestinal and gastric epithelium. EMBO J. 21, 6338–-6347(2002).

10. Pinto, D., Robine, S., Jaisser, F., El Marjou, F. E. & Louvard, D. Regulatorysequences of the mouse villin gene that efficiently drive transgenic expressionin immature and differentiated epithelial cells of small and large intestines.J. Biol. Chem. 274, 6476–-6482 (1999).

11. Madison, B. B. et al. Cis elements of the villin gene control expression inrestricted domains of the vertical (crypt) and horizontal (duodenum, cecum)axes of the intestine. J. Biol. Chem. 277, 33275–-33283 (2002).

12. Maunoury, R. et al. Developmental regulation of villin gene expression in theepithelial cell lineages of mouse digestive and urogenital tracts. Development115, 717–-728 (1992).

13. Maunoury, R. et al. Villin expression in the visceral endoderm and in the gutanlage during early mouse embryogenesis. EMBO J. 7, 3321–-3329 (1988).

14. el Marjou, F. et al. Tissue-specific and inducible Cre-mediated recombination inthe gut epithelium. Genesis 39, 186–-193 (2004).

15. Murtaugh, L. C., Stanger, B. Z., Kwan, K. M. & Melton, D. A. Notch signalingcontrols multiple steps of pancreatic differentiation. Proc. Natl Acad. Sci. USA100, 14920–-14925 (2003).

16. Hofer, D. & Drenckhahn, D. Cytoskeletal markers allowing discriminationbetween brush cells and other epithelial cells of the gut includingenteroendocrine cells. Histochem. Cell Biol. 105, 405–-412 (1996).

17. Bry, L. et al. Paneth cell differentiation in the developing intestine of normal andtransgenic mice. Proc. Natl Acad. Sci. USA 91, 10335–-10339 (1994).

18. Ohtsuka, T. et al. Hes1 and Hes5 as Notch effectors in mammalian neuronaldifferentiation. EMBO J. 18, 2196–-2207 (1999).

19. Zine, A. & de Ribaupierre, F. Notch/Notch ligands and Math1 expressionpatterns in the organ of Corti of wild-type and Hes1 and Hes5 mutant mice.Hear. Res. 170, 22–-31 (2002).

20. Lee, J. C. et al. Regulation of the pancreatic pro-endocrine gene neurogenin3.Diabetes 50, 928–-936 (2001).

NATURE|Vol 435|16 June 2005 LETTERS

967© 2005 Nature Publishing Group

21. Go, M. J., Eastman, D. S. & Artavanis-Tsakonas, S. Cell proliferation control byNotch signaling in Drosophila development. Development 125, 2031–-2040(1998).

22. Capobianco, A. J., Zagouras, P., Blaumueller, C. M., Artavanis-Tsakonas, S. &Bishop, J. M. Neoplastic transformation by truncated alleles of humanNOTCH1/TAN1 and NOTCH2. Mol. Cell. Biol. 17, 6265–-6273 (1997).

23. Gerdes, J., Schwab, U., Lemke, H. & Stein, H. Production of a mousemonoclonal antibody reactive with a human nuclear antigen associated withcell proliferation. Int. J. Cancer 31, 13–-20 (1983).

24. Fath, K. R., Obenauf, S. D. & Burgess, D. R. Cytoskeletal protein and mRNAaccumulation during brush border formation in adult chicken enterocytes.Development 109, 449–-459 (1990).

25. Louvard, D., Kedinger, M. & Hauri, H. P. The differentiating intestinal epithelialcell: establishment and maintenance of functions through interactions betweencellular structures. Annu. Rev. Cell Biol. 8, 157–-195 (1992).

26. van Es, J. et al. Notch pathway/g-secretase inhibition turns proliferative cells inintestinal crypts and neoplasia into Goblet cells. Nature doi:10.1038/nature03659 (this issue).

27. Crosnier, C. et al. Delta-Notch signalling controls commitment to a secretoryfate in the zebrafish intestine. Development 132, 1093–-1094 (2005).

28. Korinek, V. et al. Depletion of epithelial stem-cell compartments in the smallintestine of mice lacking Tcf-4. Nature Genet. 19, 379–-383 (1998).

29. Galceran, J., Sustmann, C., Hsu, S. C., Folberth, S. & Grosschedl, R.LEF1-mediated regulation of Delta-like1 links Wnt and Notch signaling insomitogenesis. Genes Dev. 18, 2718–-2723 (2004).

30. Sancho, E., Batlle, E. & Clevers, H. Live and let die in the intestinal epithelium.Curr. Opin. Cell Biol. 15, 763–-770 (2003).

31. Ito, T. et al. Basic helix–-loop–-helix transcription factors regulate theneuroendocrine differentiation of fetal mouse pulmonary epithelium.Development 127, 3913–-3921 (2000).

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

Acknowledgements We thank K. Isselbacher, A. McClatchey, M. Curto andI. Saotome for technical help, discussions and critical reading of the manuscript.S.A.-T. was supported by the National Institutes of Health. S.R. and D.L. weresupported by the Association pour la Recherche sur le Cancer and Biologie dudeveloppement et physiologie integrative.

Author Information Reprints and permissions information is available atnpg.nature.com/reprintsandpermissions. The authors declare no competingfinancial interests. Correspondence and requests for materials should beaddressed to S.A.-T. (tsakonas@helix.mgh.harvard.edu).

LETTERS NATURE|Vol 435|16 June 2005

968© 2005 Nature Publishing Group