The Hedgehog Response Network: Sensors, Switches, and Routers. Lawrence Lum and Philip A. Beachy*

Transcription

1 The Hedgehog Response Network: Sensors, Switches, and Routers Lawrence Lum and Philip. Beachy* The Hedgehog () signaling pathway is intimately linked to cell growth and differentiation, with normal roles in embryonic pattern formation and adult tissue homeostasis and pathological roles in tumor initiation and growth. Recent advances in our understanding of response have resulted from the identification of new pathway components and new mechanisms of action for old pathway components. The most striking new finding is that signal transmission from membrane to cytoplasm proceeds through, by the seven-transmembrane protein othened, of an atypical kinesin, which routes pathway activation by interaction with other components of a complex that includes the latent zinc finger transcription factor,. The expression and activity of Hedgehog proteins () exemplify a common strategy for pattern generation in metazoan embryos, namely, the specification of multiple cell fates through localized production and secretion of an instructive signal. In this manner, signals regulate cell proliferation and differentiation in a diverse array of essential patterning events ranging from embryonic segmentation and appendage development in insects (Fig. 1) to neural tube differentiation in vertebrates (1). But, signaling also assumes homeostatic roles in postembryonic tissues to maintain stem cells (2 6), and continuous pathway activity plays a pathological role in the growth of a group of endodermally derived human cancers that together account for 25% of human cancer deaths (7 9). Despite the importance of these processes in human health and disease, substantial gaps remain in our understanding of the mechanisms that mediate signal response. We review here recent fundamental advances in this area. Pathway Overview proteins enter the secretory pathway and undergo autoprocessing and lipid modification reactions that produce a signaling peptide Department of Molecular Biology and Genetics, Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, MD 21205, US. *To whom correspondence should be addressed. E- mail: pbeachy@jhmi.edu embryo dually modified at its N- and C-termini by palmitoyl and cholesteryl adducts, respectively (10). Despite its dual lipid modification and consequent tight association with membranes, the protein acts directly on distant cells in developing tissues. This remote action requires the transmembrane transporter-like protein Dispatched (Disp) for release of from secreting cells (11 P wing imaginal disc P Drosophila cell culture (RNi) Fig. 1. Hedgehog signal response. () Drosophila genetic systems used in the study of signaling. expressed in the posterior (P) compartment of the embryonic segment or the wing imaginal disc (green) induces transcriptional activation of target genes in a graded fashion in the corresponding anterior () compartments (red). We focus on signal response as elucidated by loss-of-function 14), the heparan sulfate proteoglycans Dally-like (Dlp) and Dally for extracellular transport of protein (15), and enzymes such as Sulfateless and Tout velu that are required for heparan sulfate biosynthesis [(10), and references therein]. pathway activity is triggered by stoichiometric binding of ligand to Patched B REVIEW target gene transcription studies using traditional mutational approaches in the embryo or imaginal disc or by disruption of gene expression using RNi (RN-mediated interference). (B) Overview of pathway response. utoprocessed, dually lipidated protein released from producing cells binds to and inactivates, thus permitting activation of in the responding cell. ctivated transmits a signal to the transcriptional effector, which induces target gene expression. () (16 18), also an apparent transmembrane transporter that in the absence of acts catalytically to suppress activity of the seventransmembrane protein othened [ (19)] (Fig. 1B). Inactivation of by binding to permits activation of, which in turn results in activation of latent cytoplasmic transcription factors, the protein in Drosophila and the homologous Gli proteins in mammals. These aspects of the signal response circuitry are well established and widely conserved from insects to mammals. Many of these findings, however, have their basis in genetic studies, which do not provide a mechanistic understanding of how the signal is sensed, how switches activity off, and how activation is routed from in the membrane to or Gli in the cytoplasm. Our consideration of these issues centers in great part on Drosophila, where recent advances have identified new components involved in sensing the signal and have implicated previously identified cytoplasmic components in routing the signal from to. Transcriptional Repression or ctivation in Response -responsive changes in gene expression are mediated by the zinc finger transcription factor SCIENCE VOL JUNE

2 R EVIEW Fig. 2. Hedgehog target gene regulation mediated by activator and repressor functions. () functional domains and motifs. nctional domains and motifs are labeled by amino acid numbers (in parentheses). lso indicated are phosphorylation sites that are required for initiation of proteolytic processing, as well as a protein region that encompasses a possible site of proteolytic cleavage (20). Su(fu), cmp response element binding protein (CBP) (64), and Cos2 interaction domains are labeled (41, 65). Pathway activating functions of are associated with full-length (green line) whereas repressor functions are associated with the proteolytically processed form (R, red line). ZnF, zinc finger DN binding domain; NES, putative nuclear export sequence (22, 66); NLS, putative nuclear localization sequences (48). (B) processing and functions of and R. (Left) Components required to initiate R formation. Cos2 is required to stimulate processing, possibly by scaffolding these kinases with, but this function in stimulating processing is blocked by to. Blue indicates components with dedicated pathway function; black, components with additional functions outside of pathway. (Right) Role of and R in regulating transcription of pathway targets. In the absence of pathway activity (a), is anchored in the cytoplasm and the presence of R in the nucleus represses target gene transcription. Loss of R by loss-of-function mutation of or partial activation of pathway (b) is sufficient to activate genes such as dpp but not ptc. ll activation of pathway response (c) results in the additional activation of genes such as ptc and entails elimination of R, suppression of Su(fu), and nuclear localization of., which can assume repressing and activating forms. The repressing form, R, comprises an N-terminal proteolytic fragment that retains the zinc finger mediated DN binding specificity of but lacks nuclear export signals, a cytoplasmic anchoring sequence, and a transcriptional activation domain (20 22) (Fig. 2, and B). stimulation blocks R formation and causes increased nuclear import of cytoplasmic, revealed inhibition of nuclear export by leptomycin B (LMB) (20, 22, 23). For some genes, such as decapentaplegic (dpp) in the wing imaginal disc, loss of R alone suffices for activation of expression, presumably because of the presence of an otherwise constitutively active promoter (24) (Fig. 2B). For other genes, such as the universal pathway target ptc, expression requires not only loss of R but also the positive action of (25). The expression of pathway targets thus depends on regulation of processing and localization. Proteolytic Processing of to Form R Formation of R requires phosphorylation of specific serine-threonine residues by cyclic adenosine monophosphate (cmp) dependent protein kinase [PK (26)]. ction of PK primes for further phosphorylation by glycogen synthase kinase 3 [GSK3 (27, 28)] and a member of the CK1 family of kinases (27, 29) (Fig. 2, and B). The phosphorylated form of appears to be a substrate for a proteolytic processing reaction that requires function of the proteasome and of Slimb (Slmb), an F-box containing E3 ubiquitin ligase component (30, 31). Direct ubiquitination of and cleavage by the proteasome to form R, however, have been difficult to demonstrate, and their contributions to R formation thus could be indirect. Neither these kinases (PK, GSK3, and CK1), the proteasome, nor Slmb are dedicated exclusively to signaling. Their roles in processing likely result from channeling of their activities to by Costal-2 (Cos2), a kinesinlike protein that stably associates with and is required for processing (32 34) (Fig. 2). phosphorylation and processing may be mediated by Cos2 scaffolding of kinases with, although direct associations of these kinases with Cos2 or have not yet been reported. Switches nction by Recruitment of Cos2 R formation appears to be regulated by interaction of Cos2 with (35, 36) (Fig. 3). On stimulation of cells by, the protein is stabilized and accumulates at least 10-fold (34, 37, 38) (Fig. 3B). Cos2 is recruited to via sequences in the cytoplasmic tail [C (34, 36, 39, 40)], resulting in a loss of processing (35, 36). Indeed, overexpression of C in a myristoylated, membrane-tethered form (myrc) suffices for loss of R, as indicated by accumulation of uncleaved and activation of dpp expression (35, 36) (Fig. 3C). Loss of R alone, however, is insufficient to activate the full range of pathway response, because expression of ptc and other target genes is not induced either by overexpression of myr- C (35, 36) or by loss of components required for R formation, such as Slmb or GSK3 (27, 28, 30, 31). In addition to promoting R formation, Cos2 also regulates by anchoring it in the cytoplasm (22, 41 43). Thus, in cells or tissues treated with LMB, nuclear accumulation of is observed upon loss of Cos2 activity. Nuclear accumulation of similarly results from stimulation or overexpression of myrc, indicating that cytoplasmic anchoring activity of Cos2, like its role in R formation, is abrogated upon by (35, 36). Cos2 as a Molecular Router That Channels ctivity lthough loss of Cos2 produces some pathway activity, the highest levels of response actually require the positive input of Cos2 (23, 41). One aspect of positive regulation by Cos2 is its absolute requirement for stability of the sed () serine-threonine kinase, which may relate to a tight association between these two proteins (34). -induced stabilization of thus results in of both and Cos2 (34, 40). hint as to the function of in pathway activation is its dispensability upon loss of Su(fu) [Suppressor of fused (44)], which exerts a negative regulatory effect on [ (42), see below]. Dispensability of with the loss of Su(fu) suggests a dedicated function for in Su(fu) inactivation, possibly by direct phosphorylation (34) JUNE 2004 VOL 304 SCIENCE

3 The activated switch thus is coupled to Cos2, which then routes the signal by ceasing production of R and cytoplasmic anchorage of, but also by inactivating Su(fu) through (Fig. 3). Loss of Cos2 thus produces incomplete pathway activation by lifting R repression and anchorage but without activating. Consistent with this model, all targets are activated upon loss of Cos2 if Su(fu) function is additionally removed (41). Similarly, the subset of targets activated by overexpression of myr- C is extended to include ptc upon additional loss of Su(fu) (36). In addition to stabilizing and mediating forward signaling events that affect, Cos2 is also required for the stabilization of activated, a critical aspect of a full response to the signal (34). Cos2 thus functions not just to route the signal forward to cytoplasmic components, but also in feedback amplification of the incoming signal through accumulation of activated in the membrane. Membrane Sensors of the Signal The activity state of the switch is controlled at the membrane by a series of sensors, most immediately the protein. Studies in mammalian cells have demonstrated that regulates activity indirectly and substoichiometrically. On the basis of primary sequence and predicted transmembrane topology, is a member of the resistance, modulation, division transporter family (19). These proteins export substrates across the bacterial membrane by a proton antiport mechanism, and their function requires a widely conserved motif in transmembrane span 4, which also is required for regulation of (19). These structural and functional homologies suggest that may transport an endogenous molecule that modulates activity, but such a molecule and transport activity have not yet been characterized (19). nother Drosophila membrane protein that appears to function as sensors at or upstream of is Dlp, a member of the glypican family of glycosylphosphatidylinositol-linked proteins (15, 29, 45). Dlp may bind directly to (29, 46), suggesting a possible role in delivering to (15, 29). Dlp is also involved in extracellular transport of the signal for distant action (10, 15). The Response Network in ction Several rapid changes in protein conformation or modification are elicited by stimulation, including increased phosphorylation of, Cos2,, and Su(fu) (32 34, 38, 47) and decreased phosphorylation of (22, 48). lthough phosphorylation of stimulates processing to form R, the residues targeted and the functional roles of phosphorylation of other pathway components are largely undefined. These changes in phosphorylation nevertheless may be functionally important, because they are regulated by and and are triggered by stimulation. With these phosphorylation events as benchmarks, we can surmise that pathway activation begins within minutes after stimulation (34, 47). These kinetics provide constraints in considering the physical mechanisms underlying pathway activation. Thus, for example, is proposed to function through transport of an endogenous molecule, perhaps a lipid, that modulates activity (19). This cellular state, in which is kept inactive, must dissipate rapidly upon -mediated inactivation of. Changes in the subcellular distribution of this proposed modulator upon -mediated loss B C Inactive / ctive Su(fu) (Inactive) Low Cos2 R formation cytoplasmic no target gene expression ctive / Inactive Su(fu) (ctive) High Cos2 no R formation nuclear dpp and ptc expression of function thus should be consistent with the kinetics of pathway induction. Similarly, critical events proposed to activate, such as changes in conformation (49), subcellular localization (38, 50), phosphorylation (38), or dimerization (35) must occur on a sufficiently rapid time scale to be consistent with the kinetics of pathway activation. One of the primary functions of upon activation would appear to be inhibition of Su(fu) activity through the activation of (44). It is tempting to speculate that Su(fu) is inactivated by phosphorylation and that this results from activation. However, only a small fraction of Su(fu) associates with the Cos2-- complex or is phosphorylated upon stimulation, suggesting that phosphorylation may result from a transient interaction between Su(fu) and the Cos2-- complex (34). It is less clear, however, whether activates kinase through direct contact or indirectly via Cos2. The roles of phosphorylation and/or conformational shifts of and Cos2 in activation of are not yet established. The function of Su(fu) in curtailing activity involves localization of, either through cytoplasmic anchoring or nuclear export activity, and possibly suppression of function in the nucleus (41, 42). Mammalian Su(fu) is capable of interacting with Gli proteins bound to DN (51) and also interacts with chromatin modulating factors (52), suggesting that Su(fu) also could affect transcriptional activity by recruiting myrc (Partially active) Inactive / ctive Su(fu) High Cos2 no R formation nuclear dpp expression Fig. 3. activation of target genes via and Cos2. () In the absence of stimulation, the action of keeps in a state of inactivity and low abundance. Consequently, little cytoplasmic complex (Cos2--) is recruited to, Cos2 anchors in the cytoplasm, and R formation proceeds unhindered. In addition, is inactive and Su(fu) is able to exert its negative regulatory effect on. The absence of and the presence of R in the nucleus thus result in suppression of target gene activation. (B) Upon stimulation, is activated and accumulates, thereby recruiting most or all of the cytoplasmic complex. The consequent loss of the Cos2 anchor and R formation and the activation of result in transcriptional activation of the full range of transcriptional targets including dpp and ptc. (C) Overexpression of a myristoylated C (myrc) results in of the cytoplasmic complex but an inability to suppress Su(fu) function (35, 36). s a result, R formation is blocked and dpp expression is activated (Fig. 2). Unrestrained activity of Su(fu), however, prevents activation of target genes such as ptc. R EVIEW chromatin modulating factors to the DN sites of Gli protein binding (52). Speed, Precision, and Versatility of the Response lthough rapid phosphorylation of pathway components upon stimulation indicates a rapid initiation of response, the overall kinetics of change in gene expression are considerably slower, because -induced expression of transcriptional targets requires elimination of R, which in turn requires cessation of R production and degradation of existing R. R production is blocked by of the Cos2-- complex to (34), which depends on accumulation of through new protein synthesis (34, 38). The kinetics of transcriptional SCIENCE VOL JUNE

4 R EVIEW Hedgehog Cos2 Cos2 rrow Wingless CK1α xin Wg Dsh rm GSK PC 3β Fz B Signal Sensor Switch Router Regulator Transcriptional effector Drosophila Ski Disp Dlp Cos2? Su(fu) response thus are not determined by rapid changes in protein modification and conformation, but rather by the slower processes of protein synthesis and R degradation. ll induction of a transcriptional response thus requires saturating stimulation over a time scale of several hours after initial exposure. Critical roles for the relatively slow processes of protein synthesis or degradation in activation of the pathway represent an opportunity to integrate signal strength over time. The increased precision inherent in this type of signal response comes at the expense of a rapid response ability typical of other signaling pathways but may be particularly useful in the context of embryonic development, where -induced cell differentiation has lasting consequences for the pattern and function of mature structures. This principle of precision in signal response through time integration may generalize to other pathways that specify cell fates. For example, although rapid signal-induced changes in protein modification also occur within the Wnt pathway, transcriptional activity is limited by the slower synthesis and accumulation of -catenin. In addition to transcriptional regulation, signaling has been implicated in axon guidance in vertebrate development (53). This chemoattractant activity for Sonic hedgehog (Shh) proceeds in a relatively short time scale with a local cell polarity that suggests a possible nonnuclear mechanism of response. lthough this relatively rapid cytoplasmic response is mediated by, other components have not yet been identified. Heterotrimeric guanine nucleotide binding proteins (G proteins) do not mediate regulated transcription in Drosophila (34), but a potential role for G proteins or other mediators in other responses to signaling cannot be ruled out, particularly in cytoplasmic or nuclear responses that do not involve transcriptional regulation via and Gli. In this regard, it is notable that Wnt proteins, which signal via the Frizzled (Fz) relatives of, also elicit a variety of responses that extend beyond transcriptional activation to include axonal guidance and Ca 2 mobilization (54). and Wnt: Sister Pathways? and Wnt signals and responses are similar in many respects. Both signals are lipidated, an important factor in their activity and tissue distribution [see (10) for a full discussion]. Both pathways rely on protein synthesis for full activation (see above) and use several related or identical components, including the GSK3 and CK1 kinases, the Slmb ubiquitin ligase subunit, and the related and Fz proteins (3, 55, 56). The fundamental logic underlying pathway activation is also similar, involving receptor of a multicomponent complex scaffolded by a protein with roles in cytoplasmic tethering and proteolysis of a key transcriptional effector (Fig. 4). The Wnt pathway component xin, in its cytoplasmic anchoring and Mammals Skn Disp Shh, Ihh, Dhh? h1, h2 (?) Su(fu) Gli1, Gli2, and Gli3 Fig. 4. Routing of and Wnt signals in Drosophila and in mammals. () ctivation of and Wnt transmembrane effectors recruits large cytoplasmic complexes to the membrane. Suppression of by binding to permits activation and consequent of Cos2,, and via an interaction between the cytoplasmic tail and Cos2, resulting in activation, inhibition of R formation, and loss of Cos2 anchoring activity. Wg forms a tertiary complex with rrow and Fz, resulting in of the xin-dsh complex to the membrane [reviewed in (57)]. xin interacts with the cytoplasmic tail of rrow and Dsh with the cytoplasmic tail of Fz. Recruitment of the xin-dsh complex suppresses rm phosphorylation by CK1 and GSK3, thus abrogating proteasomemediated destruction, and releases rm from its cytoplasmic anchor. The Wnt signaling complex model depicted here is a composite based on characterized individual protein interactions. (B) Drosophila components required for response and their mammalian orthologs. Mammalian genes that appear to function in signaling but that have either no identified Drosophila orthologs or no known function in signaling include SIL, HIP, Rab23, the intracellular flagellar transport (IFT ) proteins, and FKBP8 [discussed in (63)]. The mammalian ortholog of Cos2 has not been identified and the roles of the apparent orthologs of Dlp (glypican 4,6) in mammalian signaling (indicated by question marks) have not been determined. proteolysis of rmadillo [rm; reviewed in (57)], thus parallels Cos2. In both cases, signaldependent of these complexes to the membrane abrogates their anchoring and proteolytic functions, thus permitting nuclear accumulation of the rm and transcriptional effectors. It will be interesting to learn whether xin, heretofore considered exclusively as a negative regulator, may also similarly play a positive role in pathway activation similar to that of Cos2. Similarly, it will be interesting to learn whether Cos2, like xin, participates in scaffolding of the kinases that trigger proteolytic processing. Finally, the role of phosphorylation in the of Cos2 may parallel that of rrow phosphorylation in response to Wg that allows of xin (58). How Other nimals Respond to Hedgehogs The pathway in other animals to some extent parallels that in Drosophila, albeit with duplications of certain pathway components, with substantial gaps in our understanding, and with pathway roles for some proteins that do not function in the Drosophila pathway (Fig. 4B). Some duplication appears simply to provide diverse tissue-specific patterns of expression (e.g., the three mammalian proteins), whereas other duplications permit divergence of function, [e.g., the Gli homologs of, which subdivide the positive (Gli2) and negative (Gli3) transcriptional regulatory functions of among them (59, 60)]. Between and the Gli proteins there is a considerable gap in our understanding of mammalian signal response. The nearest homologs to Cos2 are not nearly as well conserved as,,, and Gli and have not been functionally linked to pathway regulation. Similarly, the putative homolog shows more limited conservation (61), and its pathway role is not clear because loss of function analyses have not been reported. Somewhat surprisingly, given its relatively mild mutant phenotype in Drosophila, the best JUNE 2004 VOL 304 SCIENCE

5 conserved component known to act in the interval between and is Su(fu) (51). Its relatively high conservation in mammals suggests that it may play an important role in pathway regulation, and loss of function analyses using morpholino oligonucleotides in the zebrafish suggest that Su(fu) may indeed have pathway regulatory functions (62). In addition, several mammalian components recently identified in genetic screens appear to function between and the Gli proteins, and some of these have orthologs that apparently do not contribute to signaling in Drosophila [(63) and references therein]. The biochemical functions of these proteins in mammalian signaling remain unknown. The function of these novel pathway components in mammalian but not Drosophila response is especially noteworthy in view of the high degree of functional conservation of other components of the signal response network. Hedgehog-regulated expression of transcriptional targets across evolution thus is similarly responsive to the sensor and to the switch and involves similar processing and/or activation of latent transcriptional regulators. But the unique roles of these novel mammalian components suggests that distinct mechanisms may be involved in routing signaling activity from to Gli. The presence of a clear mammalian ortholog of Su(fu) and of a putative ortholog suggest that these additional mechanisms may operate in tandem with some of the same routing mechanisms that operate in Drosophila, although the absence of a clear Cos2 ortholog also suggests that certain Drosophila routing mechanisms may have lesser importance in mammals. rther mechanistic investigation of signal routing from to transcriptional effectors will be critical to an understanding of how evolution has shaped signal response and of the manner in which this response can be disrupted to produce neoplasia or birth defects. References and Notes 1. T. M. Jessell, Nat. Rev. Genet. 1, 20 (2000). 2. Y. Zhang, D. Kalderon, Nature 410, 599 (2001). 3. J. Taipale, P.. Beachy, Nature 411, 349 (2001). 4. T. Reya, S. J. Morrison, M. F. Clarke, I. L. Weissman, Nature 414, 105 (2001). 5. R. Machold et al., Neuron 39, 937 (2003). 6. K. Lai, B. K. Kaspar, F. H. Gage, D. V. Schaffer, Nat. Neurosci. 6, 21 (2003). 7. D. M. Berman et al., Nature 425, 846 (2003). 8. S. P. Thayer et al., Nature 425, 851 (2003). 9. D. N. Watkins et al., Nature 422, 313 (2003). 10. R. K. Mann, P.. Beachy, nnu. Rev. Biochem. 73, 891 (2004). 11. R. Burke et al., Cell 99, 803 (1999). 12. T. Caspary et al., Curr. Biol. 12, 1628 (2002). 13. T. Kawakami et al., Development 129, 5753 (2002). 14. Y. Ma et al., Cell 111, 63 (2002). 15. C. Han, T. Y. Belenkaya, B. Wang, X. Lin, Development 131, 601 (2004). 16. V. Marigo, R.. Davey, Y. Zuo, J. M. Cunningham, C. J. Tabin, Nature 384, 176 (1996). 17. D. M. Stone et al., Nature 384, 129 (1996). 18. N. se et al., Proc. Natl. cad. Sci. U.S.. 96, (1999). 19. J. Taipale, M. K. Cooper, T. Maiti, P.. Beachy, Nature 418, 892 (2002). 20. P. za-blanc, F.. Ramirez-Weber, M. P. Laget, C. Schwartz, T. B. Kornberg, Cell 89, 1043 (1997). 21. H. kimaru et al., Nature 386, 735 (1997). 22. C. H. Chen et al., Cell 98, 305 (1999). 23. Q. T. Wang, R.. Holmgren, Development 127, 3131 (2000). 24. N. Methot, K. Basler, Cell 96, 819 (1999). 25. N. Methot, K. Basler, Development 128, 733 (2001). 26. Y. Chen, N. Gallaher, R. H. Goodman, S. M. lik, Proc. Natl. cad. Sci. U.S.. 95, 2349 (1998). 27. M.. Price, D. Kalderon, Cell 108, 823 (2002). 28. J. Jia et al., Nature 416, 548 (2002). 29. L. Lum et al., Science 299, 2039 (2003). 30. N.. Theodosiou, S. Zhang, W. Y. Wang, T. Xu, Development 125, 3411 (1998). 31. J. Jiang, G. Struhl, Nature 391, 493 (1998). 32. D. J. Robbins et al., Cell 90, 225 (1997). 33. J. C. Sisson, K. S. Ho, K. Suyama, M. P. Scott, Cell 90, 235 (1997). 34. L. Lum et al., Mol. Cell 12, 1261 (2003). 35. J. E. Hooper, Development 130, 3951 (2003). R EVIEW 36. J. Jia, C. Tong, J. Jiang, Genes Dev. 17, 2709 (2003). 37. J. lcedo, Y. Zou, M. Noll, Mol. Cell 6, 457 (2000). 38. N. Denef, D. Neubuser, L. Perez, S. M. Cohen, Cell 102, 521 (2000). 39. S. K. Ogden et al., Curr. Biol. 13, 1998 (2003). 40. L. Ruel, R. Rodriguez,. Gallet, L. Lavenant-Staccini, P. P. Therond, Nat. Cell Biol. 5, 907 (2003). 41. G. Wang, K. manai, B. Wang, J. Jiang, Genes Dev. 14, 2893 (2000). 42. N. Methot, K. Basler, Development 127, 4001 (2000). 43. M.. Lefers, Q. T. Wang, R.. Holmgren, Dev. Biol. 236, 411 (2001). 44. T. Preat, Genetics 132, 725 (1992). 45. S. C. Desbordes, B. Sanson, Development 130, 6245 (2003). 46. L. Lum, P.. Beachy, unpublished data. 47. P. P. Therond, J. D. Knight, T. B. Kornberg, J. M. Bishop, Proc. Natl. cad. Sci. U.S.. 93, 4224 (1996). 48. Q. T. Wang, R.. Holmgren, Development 126, 5097 (1999). 49. J. Taipale et al., Nature 406, 1005 (2000) J. Zhu, L. Zheng, K. Suyama, M. P. Scott, Genes Dev. 17, 1240 (2003). 51. R. V. Pearse, 2nd, L. S. Collier, M. P. Scott, C. J. Tabin, Dev. Biol. 212, 323 (1999). 52. S. Y. Cheng, J. M. Bishop, Proc. Natl. cad. Sci. U.S.. 99, 5442 (2002). 53. F. Charron, E. Stein, J. Jeong,. P. McMahon, M. Tessier-Lavigne, Cell 113, 11 (2003). 54. M. T. Veeman, J. D. xelrod, R. T. Moon, Dev. Cell 5, 367 (2003). 55. D. Kalderon, Trends Cell Biol. 12, 523 (2002). 56. R. Nusse, Development 130, 5297 (2003). 57. X. He, M. Semenov, K. Tamai, X. Zeng, Development 131, 1663 (2004). 58. K. Tamai et al., Mol. Cell 13, 149 (2004). 59. Y. Litingtung, C. Chiang, Nat. Neurosci. 3, 979 (2000). 60. H. L. Park et al., Development 127, 1593 (2000). 61. M. Murone et al., Nat. Cell Biol. 2, 310 (2000). 62. C. Wolff, S. Roy, P. W. Ingham, Curr. Biol. 13, 1169 (2003). 63. O. V. Bulgakov, J. T. Eggenschwiler, D. H. Hong, K. V. nderson, T. Li, Development 131, 2149 (2004). 64. Y. Chen, R. H. Goodman, S. M. lik, Mol. Cell. Biol. 20, 1616 (2000). 65. G. Wang, J. Jiang, Dev. Biol. 268, 493 (2004). 66. C. H. Chen, D. V. Kessler, P.. Beachy, unpublished data. 67. We would like to thank members of the Beachy lab and J. Taipale for critical comments on this manuscript. We apologize to the authors of original work not cited. Because of editorial limits on references, we were largely limited to citing review articles and more recent original articles. P..B. is an investigator of the Howard Hughes Medical Institute. This work was supported by a grant from the NIH. SCIENCE VOL JUNE