AU728541B2 - Novel hedgehog-derived polypeptides - Google Patents

Description

WO 98/30576 PCT/US97/15753 -1- NOVEL HEDGEHOG-DERIVED POLYPEPTIDES BACKGROUND OF THE INVENTION 1. Field of the Invention This invention relates generally to the field of protein processing and protein signalling pathways and specifically to two novel proteins having distinct activities, which are derived from a common hedgehog protein precursor.

2. Description of the Related Art Embryologists have long performed experimental manipulations that reveal the striking abilities of certain structures in vertebrate embryos to impose pattern upon surrounding tissues. Speculation on the mechanisms underlying these patterning effects usually centers on the secretion of signaling molecule that elicits an appropriate response from the tissues begin patterned. More recent work aimed at the identification of such signaling molecules implicates secreted proteins encoded by individual members of a small number of gene families. One such family of proteins which may have an influential effect upon patterning activities are those proteins encoded by the hedgehog gene family.

The hedgehog (hh) gene was initially identified based on its requirement for normal segmental patterning in Drosophila (Nisslein-Volhard, C. Wieschaus, E, Nature 287:795-801, 1980). Its functions include local signaling to coordinate the identities of adjacent cells within early embryonic segments (Hooper, Scott, M.P. Early Embryonic Development of Animals, pp.1-48, 1992) and a later function in cuticle patterning that extends across many cell diameters (Heemskerk, J. DiNardo, Cell, 76:449-460, 1994). The hh gene also functions in the patterning of imaginal precursors of adult structures, including the appendages and the eye (Mohler, J. Genetics, 120:1061- 1072, 1988; Ma, et Cell, 75:927-938, 1993; Heberlein, et al., Cell, 75:913-926, 1993; Tabata, T. Kornberg, Cell, 76:89-102, 1992; Basler, K. Struhl, Nature,

I

WO 98/30576 PCTIUS97/15753 -2- 368:208-214, 1994). Genetic and molecular evidence indicates that hedgehog proteins are secreted and function in extracellular signaling (Mohler, supra; Lee, et al., Cell, 71:33-50, 1992; Taylor, et al., Mech. Dev., 42:89-96, 1993).

In vertebrates activities encoded by hh homologues have been implicated in anterior/posterior patterning of the limb (Riddle, et al., Cell, 75:1401-1416, 1993; Chang, et al., Development, 120:3339, 1994), and in dorsal/ventral patterning of the neural tube (Echelard, et al., Cell, 75:1417-1430, 1993; Krauss, et al., Cell, 75:1431-1444, 1993; Roelink, et al., Cell, 76:761-775, 1994).

The vertebrate ventral midbrain contains neurons whose degeneration or abnormal function are linked to a number of diseases, including Parkinson's disease and schizophrenia. It is known that motor neurons develop in close proximity to the floor plate in the ventral midbrain. Midbrain projections to the striatum are involved in the control of voluntary movement (Bjorklund and Lindvall, In: Handbook of Chemical Neuroanatomy, eds., Borklund, et al., Amsterdam: Elsevier, pp55-122, 1984) and loss of these neurons results in the motor disorders of Parkinson's disease (Hirsch, et al., Nature, 334:345, 1988). Midbrain dopaminergic neurons that innervate limbic structures and the cortex influence emotional and cognitive behavior, respectively, and abnormal function of these neurons has been associated with schizophrenia and drug addiction (Seeman, et al., Nature, 365:441, 1993).

While the molecular nature of the factors that specify neuronal cell fate have not been established, members of the transforming growth factor-p (TGF-3) (Lyons, et al., Trends in Genetics, 7:408, 1991) or the hedgehog protein family (Smith, Cell, 76:193, 1994) may possess the characteristics expected from such factors as they participate in specification of cell fate, mediate inductive interactions between tissues, and in many cases act at a distance of only a few cell diameters.

WO 98/30576 PCTIUS97/15753 -3- The present invention establishes that hh activities encoded by these genes play a crucial role in early patterning of the developing eye and in patterning of the brain. For the first time, the invention shows that internal cleavage of hedgehog protein product is critical for full function, and that the two novel products of this auto-proteolytic cleavage display distinguishable activities, thus demonstrating that hh signaling activity is a composite effect of two separate signaling proteins that derive from a common hh protein precursor.

In so doing, the invention provides the means for specific patterning and proliferation of desired neuronal cell types for addressing disorders which arise from neuronal degeneration or abnormal function.

SUMMARY OF THE INVENTION The present invention is based on the seminal discovery that hedgehog proteins undergo auto-proteolytic cleavage which results in two separate proteins having distinct functional and structural characteristics. The two polypeptides, referred to as the and fragments of hedgehog, or N-terminal and C-terminal fragments, respectively, are produced after specific cleavage at a G'CF site recognized by the autoproteolytic domain in the native protein. The fragment functions as a cholesterol transferase during autoproteolysis thus allowing cholesterol modification of the fragment.

Thus, in one embodiment, the invention provides a substantially pure polypeptide characterized by having an amino acid sequence derived from amino terminal amino acids of a hedgehog protein and having at its carboxy terminus, a G ICF cleavage site specifically recognized by a proteolytic activity of the carboxy terminal fragment of the native hedgehog polypeptide. The invention also provides a substantially pure polypeptide characterized by having an amino acid sequence of a hedgehog polypeptide or a fragment derived from amino terminal amino acids of a hedgehog polypeptide, wherein the polypeptide or fragment thereof comprises a sterol moiety. Fragments derived from a native hedgehog polypeptide are included and preferably include extracellular amino WO 98/30576 PCT/US97/15753 -4acid residues, such as those derived from the N fragment. In one embodiment of the invention, the sterol moiety is cholesterol.

In another embodiment, the invention provides a substantially pure polypeptide characterized by having an amino acid sequence derived from carboxy terminal amino acids of a hedgehog protein and having at its amino terminus, a G CF cleavage site specifically recognized by a proteolytic activity of the carboxy terminal fragment of the native hedgehog polypeptide.

The invention also provides a method for modulating proliferation or differentiation of neuronal cells, comprising contacting the cells with a hedgehog polypeptide. The native hedgehog polypeptide, the N, or the C fragment, or functional fragments derived therefrom, are most useful for the induction of proliferation or differentiation of neuronal cells substantially derived from floor plate neuronal cells.

In yet another embodiment, the invention provides a method for identifying a compound which affects hedgehog activity comprising incubating the compound with hedgehog polypeptide, or with biologically active fragments thereof, or with a recombinant cell expressing hedgehog, under conditions sufficient to allow the components to interact; and determining the effect of the compound on hedgehog activity or expression. For example, cholesterol level biosynthesis or transport) is measured as an inidicator of hedgehog activity. In one aspect of the invention, the method provides a means for affecting cholesterol biosynthesis or transport in a cell comprising contacting a cell with an effective amount of a compound that affects hedgehog, thereby affecting cholesterol biosynthesis or transport. The effect may be inhibition or stimulation of cholesterol biosynthesis or transport.

WO 98/30576 PCT/US97/15753 BRIEF DESCRIPTION OF THE DRAWINGS FIGURE 1 shows processing of the hh protein by immunoblots with antibodies against amino (Ab1) and carboxy-terminal (Ab2) epitopes. FIGURE 1B and D are blots of samples immunoprecipitated with Abl lanes Ab2 lanes 19-21), or preimmune serum lanes 10-12, and D, lanes 22-24).

FIGURES 1E and 1F show a schematic illustration of the hedgehog cleavage mechanism.

FIGURE 2 shows sequence similarity between hh proteins and serine proteases. hh protein sequences are aligned to residues 323 to 329 of the D. melanogaster protein and numbered as positions 1 to 7 (group The catalytic histidines of mammalian serine proteinases (group B) are aligned to the invariant histidine at position 7 in hh proteins.

FIGURE 3 shows autoproteolysis of the hh protein. 3A shows a coomasie blue stained polyacrylamide gel showing production and purification of His 6 -U and Hi -q1329A proteins from E. coli. Samples were molecular weight markers (lanes 1 and lysates ofE. coli cells carrying the His 6 -U expression construct without (lane 3) and with (lane 4) induction by IPTG; purified His 6 -U protein (lane lysates ofE. coli cells that carry the His6-UH 3 29 A expression construct without (lane 6) and with (lane 7) induction by IPTG; purified His 6

-UH

32 9 A protein (lane FIGURE 3(B) is an immunoblot detected with Ab2 showing transfected S2 cells induced to express hh (lane His 6 -U and His 6 -UH29A proteins incubated in cleavage reaction buffer for 0 hours (lanes 2 and for 20 hours (lanes 3 and and for 20 hours in the presence of 20 mM TAME (a serine protease inhibitor) (lanes 4 and 7).

FIGURE 4 shows autoproteolytic functions of Drosophila (4A-C) and zebrafish hh proteins map to the carboxy terminal fragments by in vitro translations of wild-type and mutant hh proteins. The locations of mutations and cleavage sites (arrows) in these proteins are illustrated schematically in 4E.

WO 98/30576 PCT/US97/15753 -6- FIGURE 5 shows immunoblots showing heat shock induced expression of wild type and H329A mutant hh proteins in Drosophilia embryos and are immunoblots developed using Abl and Ab2 antibodies, respectively. Lanes 1 and 6, induced untransfected S2 cells; lanes 2 and 7, transfected S2 cells induced to express hh; lanes 3 and 8, heat shocked wild-type embryos; lanes 4 and 9, heat shocked hshh embryos; lanes 5 and 10, heat shocked hshh H329A embryos.

FIGURE 6 shows in situ hybridization showing the embryonic effects of ubiquitously expressed wild type and H329A hh proteins. FIGURE 6 shows the embryonic distribution of wingless (wg) RNA as revealed by in situ hybridization is shown in (A) wild-type (homozygous y' w hshh, and hshh H329A embryos that were exposed to two 10 minute heat shocks separated by a 90-minute recovery period (33).

Wild-type embryos showed little change in wg expression, whereas the wild-type protein and, to a lesser extent, the H329A protein each induced ectopic wg expression (Table 1).

Panels and show the dorsal surfaces of y' hshh, and hshh H329A larvae, respectively, at the level of the fourth abdominal segment. These larvae were shocked for 30 minutes as embryos and allowed to complete embryogenesis. Cuticle cell types (1 o, 30, and are labeled as described Heemskerk and S. DiNardo, Cell 76,449, 1994). Note the expansion of 2 cell types (naked cuticle) at the expense of 3 and some 40 types in the hshh embryo under conditions where the phenotype of hshh H329A embryos is identical to that of control embryos FIGURE 7 shows X-gal staining to show imaginal disc effects of ubiquitous wild type and H329 hh proteins. X-gal staining was used to follow expression of wg or dpp in imaginal discs of late third-instar larvae that carry wg-lacZ or dpp-lacZ reporter genes. Leg wing and eye-antennal discs from control larvae D, G, larvae carrying the hshh transgene E, H, K) and larvae carrying the hshh H329A transgene F, I, L) are displayed. In all panels anterior is to the left.

WO 98/30576 PCT/US97/15753 -7- FIGURE 8 and are immunoblots of cell pellets (lane 1) or supernatants (lane 2) from transfected S2 cell cultures expressing HH protein, developed with Ab and Ab2 Samples in each lane were from the same volume of resuspended total culture.

Whereas N remained mostly associated with the cell pellet (compare lanes 1 and 2 in A), C was nearly quantitatively released into the supematant (compare lanes 1 and 2 in B).

U displayed partitioning properties in between those of N and C (A and (C) demonstrates the heparin binding activity of various HH protein species generated by in vitro translations with microsomes Samples were: total translation mix (lane 1); supernatant after incubation with heparin agarose or agarose (control) beads (lanes 2 and and material eluted from heparin agarose or agarose beads after washing (lanes 3 and F, U, Nss and N fragments are depleted from reactions incubated with heparin agarose but not agarose beads (compare lanes 2 and 4 to and the same species subsequently can be eluted from the heparin agarose but not the agarose beads (compare lanes 3 and with lane 1).

FIGURE 9 shows the differential localizations of N and C in embryos by in situ localization of the hh transcript. Fig. 9 is shown in comparison to the distribution of N and C epitopes detected with Abl and Ab2 in panels and respectively. Note that the distribution of N and C epitopes span approximately one-third and one-half of each segmental unit respectively, while the transcript is limited to approximately one-quarter of each unit. In the localization of C epitopes in embryos homozygous for the hh' 3 allele is detected with the use of Ab2. C epitopes in this mutant, which displays impaired auto-proteolytic activity (see text), are more restricted, and resemble the wild-type localization ofN. Homozygous hhl 3 E embryos were identified by loss of a marked balancer from a heterozygous parent stock. All embryos are at mid to late stage 9 (extended germ-band).

FIGURE 10 shows a signal relay versus dual function models for hh protein action. In Fig. 10 the long-range effects ofhh signaling are achieved indirectly through shortrange induction of a second signaling molecule Based on its biochemical properties WO 98/30576 PCT/US97/15753 -8and its restricted tissue localization, N is presumed to represent the active short-range signal while the role of C would be limited to supplying the catalytic machinery required for biogenesis of N. In the long- and short-range signaling functions of hh are supplied by the N and C proteins derived by internal auto-proteolysis of the U precursor.

N is implicated in short-range signaling by retention near its cellular site of synthesis, while C is less restricted in its distribution and would execute long-range signaling functions. In both models, auto-proteolysis is required to generate fully active signaling proteins.

FIGURES 10 C and D show an immunoblot of the N fragment synthesized from a wild type construct or a consruct lacking the C domain FIGURES 11 A and B show the nucleotide and deduced amino acid sequences for partial human hh clones.

FIGURE 12 A and B show in vitro cleavage reactions of a Drosophila hh protein produced in E. coli and purified to homogeneity. FIGURE 12, Panel A shows a time course of cleavage after initiation by addition of DTT. Panel B shows incubations of concentrations ranging over three order of magnitude for a fixed time period (four hours), with no difference in the extent of conversion to the cleaved form. Panel C shows the sequence around the cleavage site as determined by amino-terminal sequence of the cleaved fragment C. The cleavage site is denoted by the arrow, and the actual residues sequenced by Edman degradation of the C fragment are underlined. Panel C also shows an alignment of all published vertebrate hh sequences plus some of unpublished sequences from fish and Xenopus. The sequences shown correspond to the region of Drosophila hh where the cleavage occurs, and demonstrates the absolute conservation of the Gly-Cys-Phe sequence at the site of cleavage. Panel D shows a SDS-PAGE gel loaded with in vitro transcription/translation reactions as described in the previous Examples, using various hh genes as templates. dhh is Drosophila, twhh and zfshh are the twiggy-winkle and sonic hh genes of the zebrafish, and mshh is the shh/Hgh-1/vhh-1 WO 98/30576 PCT/US97/15753 -9gene of the mouse. Panel E shows that Edman degradation of the C fragments releases 3"S counts on the first but not subsequent rounds for all these proteins, indicating that the site of autoproteolytic cleavage for all of these hh proteins is the amide bond to the amino-terminal side of the Cys residue that forms the center of the conserved Gly-Cys- Phe sequence highlighted in panel C.

FIGURE 13 shows the predicted amino acid sequences are shown in single letter code.

13(a) shows sequences common to five distinct hh-like genes are shown with a dot indicating identity with the corresponding residue of zebrafish twiggy-winkle. 13(b) shows amino acid sequences of twhh and shh are aligned to those of the soniclvhh-l class from chick and mouse. The amino-terminal hydrophobic stretch common to all four hh genes is shaded. The asterisk denotes invariant amino acid residues associated with the proteolytic domain of C fragment from various species. 13(c) shows percent identity of residues carboxy-terminal to the hydrophobic region.

FIGURES 14A-S show a comparative expression of twhh, shh, and pax-2 during zebrafish embryogenesis.

FIGURE 15, panels 15A-15I, show the effects of ectopic hh on zebrafish development.

Wild type zebrafish, Danio rerio, Ekkwill Waterlife Resources) were maintained at 28.5"C, some embryos were then cultured overnight at RT. Zebrafish embryos were injected at the 1-8 cell stage with twhh, shh, or lacZRNA and examined at 28 h of development. Dorsal view of the midbrain-hindbrain region; anterior is left. (a) lacZ. twhh. shh. Frontal optical section of the forebrain region; anterior is up. lacZ. twhh. shh. Lateral view of the eye region; anterior is left. (g) lacZ. twhh. twhh.

FIGURE 16 is a table showing the effects of ectopic expression of shh, twhh and twhh mutants on zebrafish embryonic development.

WO 98/30576 PCT/US97/15753 FIGURE 17 shows zebrafish twiggy-winkle hedgehog derivatives. FIGURE 17A shows cartoons of various twhh open reading frames. SS (shaded) is the predicted N-terminal signal sequence for secretion of these proteins and encompasses the first 27 amino acids of each open reading frame. The arrow indicates the predicted internal site of autoproteolytic cleavage. Amino acid residue numbers are according to Figure 13b. The filled triangle denotes the normal termination codon for the twhh open reading frame.

Construct UHA contains a mutation that blocks auto-proteolysis (the histidine at residue 273 is changed to an alanine; see Lee, et al., supra.). Construct U 3 56A contains a stop codon in place of amino acid residue 357 as well as the H273A mutation in UHA.

Construct N encodes just the first 200 amino acids of twhh. Construct C has had the codons for residues 31-197 deleted.

FIGURE 17B shows in vitro translation of the expression constructs shown schematically in part a. Constructs were translated in vitro in the presence of "S methionine and analyzed by autoradiography after SDS-PAGE.

FIGURES 18A and 18B show Northern blot analysis of the effect of hedgehog on expression of various neural markers.

FIGURES 19A and 19B show hh synergy with naturally occurring neural markers or agents XAG-1, XANF-2, Otx-A, En-2, Krox-20, Xlh box-6, NCAM, and EF-la).

FIGURE 20A shows hh constructs including delta N-C.

FIGURE 20B shows a Northem blot analysis of the effect of hedgehog N or C on various neural markers.

FIGURE 21 shows AN-C interferes with X-bhh and N-activity in animal cap explants as shown by RT-PCR analysis.

WO 98/30576 PCT/US97/15753 11 FIGURE 22A is an illustration of lipid stimulation of hedgehog autoprocessing.

FIGURE 22B shows a Coomassie blue stained SDS-PAGE ofautocleavage reactions in bacterially expressed His 6 Hh-C protein.

FIGURE 23A is a thin layer chromatography (TLC) plate coated with silica gel G (Merck) showing the fractionation of bulk S2 cell lipids using a heptane:ether:formic acid solvent (80:20:2).

FIGURE 23B is a Coomassie blue-stained SDS-polyacrylamide gel showing in vitro autocleavage reactions of the bacterially expressed His6Hh-C protein incubated with 1 mMDTT plus either unfractionated S2 cell lipids (lane or spots A through F (lanes 2- 7, respectively).

FIGURE 23C is TLC of S2 cell lipids (lane 1) along with selected lipid standards: phosphatidylcholine (lane a diacylglycerol (lane cholesterol (lane stearic acid (lane a triacylglycerol (lane and cholesteryl ester (lane Lipid spot B comigrates with cholesterol, as also demonstrated by mixing radio-labeled cholesterol with S2 lipids before TLC fractionation.

FIGURE 23D is a Coomassie blue stained SDS-polyacrylamide gel showing that relative to 1 mMDTT alone (lane 1) cholesterol (0.35 mM) 1 mMDTT (lane 2) stimulates His2Hh-C autocleavage in vitro.

FIGURE 23E is an autoradiogram of electrophoretically-resolved products of His6Hh-C autocleavage reactions driven by 20 mMDTT (lane 1) or 1 mMDTT+0.35 mM cholesterol (lane 2).

WO 98/30576 PCT/US97/15753 12- FIGURE 24A shows Coomassie stained gels of His6Hh-C autocleavage reactions carried out in the presence of 20 mMDTT (lane or 1 mMDTT+0.35 mM cholesterol (lane 2).

Lane 3 contains a mixture of the samples loaded in lanes 1 and 2.

FIGURE 24B is Coomassie stained gels showing protein products of His6Hh-C autocleavage reactions carried out in the presence of 1 mM DTT+0.35 mM cholesterol (lanes 1 and 2) or with 20 mM DTT (lane 3).

FIGURE 24C is an autoradiogram of immunoblotted Hh amino-terminal domains purified from cultured S2 cells.

FIGURE 25A is an audioradiogram of a gel loaded with total cell proteins from S2 cells containing a stably integrated Cu++-inducible hedgehog gene.

FIGURE 25B is an HPLC profile of sterols separated on a C18 column by isocratic elution with a solvent containing methanol:ethanol:water (86:10:4).

FIGURE 25C shows HPLC analysis as in of the adduct released by base treatment of Hh-Np metabolically labeled with [3H]cholesterol FIGURE 25D shows metabolic labeling of vertebrate Sonic hedgehog protein with [3H]cholesterol. Autoradiogram of a gel loaded with total cell proteins from COS-7 cells transfected with a wild-type Sonic hedgehog expression construct (Shh, lane 1) or a construct that generates an unprocessed amino-terminal protein truncated after the conserved glycine at the site of autocleavage (Shh-N, lane 2).

FIGURE 26A is a schematic drawing of a two-step mechanism for Hh autoprocessing.

Aided by deprotonation by either solvent or a base the thiol group of Cys-258 initiates a nucleophilic attack on the carbonyl carbon of the preceding residue, Gly-257.

This attack results in replacement of the peptide bond between Gly-257 and Cys-258 by WO 98/30576 PCT/US97/15753 -13 a thioester linkage (step The emerging a-amino group of Cys-258 likely becomes protonated, and an acid is shown donating a proton. The Thioester is subject to a second nucleophilic attack from the 3p-hydroxyl group of a cholesterol molecule, shown here facilitated by a second base resulting in a cholesterol-modified amino-terminal domain and a free carboxy-terminal domain. In vitro cleavage reactions may also be stimulated by addition of small nucleophiles including DTT, glutathione, and hydroxylamine.

Figure 26B is a schematic drawing of a mechanism for intein self-splicing. A base (B 1') or solvent deprotonates a cysteine or serine residue at the N-extein/intein junction (shown here as a cysteine residue) for attack on the carbonyl group of the preceding amino-acid residue resulting in the formation of a thioester/ester intermediate. An acid may protonate the a-amino group of the cysteine/serine residue promoting its release. The thioester/ester is then subject to a second nucleophilic attack from a cysteine, serine, or threonine residue at the intein/C-extein junction (shown here as a cysteine residue). A second base is shown facilitating deprotonation of the second nucleophile, although this function may also be carried out by B1'. This reaction produces a branched protein intermediate that ultimately resolves to a free intein and ligated exteins.

Figure 27 is a Coomassie Brilliant Blue-stained SDS-polyacrylamide gel showing in vitro autocleavage reactions of bacterially-expressed His 6 Hh-C 25 (lanes 1-3) and His Hh-Ci7 (lanes 4-6) proteins. Proteins were incubated with 1 mM DTT (lanes 1 and 50 mM DTT (lanes 2 and 5) or 350 uM cholesterol/1 mM DTT (lanes 3 and The uncleaved His 6 Hh-C 2 5 protein migrates as a -29-kDa species, and the carboxy-terminal cleavage product of this protein migrates as a -25-kDa species (Porter et al., 1996). The uncleaved His 6 Hh-C 7 protein migrates as a -21-kDa species, and the carboxy-terminal product of this truncated protein migrates as a -14-kDa species. The amino-terminal product of the His 6 Hh-C,, and His 6 Hh-C,, proteins migrates as a -7-kDa species when DTT-modified or as a -5-kDa species when cholesterol-modified. His 6 Hh-C, 7 was also incubated with 46 LM 3 H]cholesterol/1 mM DTT, and no cholesterol-modified product WO 98/30576 PCT/US97/15753 14was detected by autoradiography. A cholesterol-transfer activity 1% of wildtype could have been detected by this radioassay.

Figure 28A is a ribbon diagram of The amino- and carboxy- termini are labeled. This panel was prepared with MOLSCRIPT (Kraulis, 1991).

Figure 28B is a topology diagram of Residues in P strands are in boxes with amino-acid type and number indicated. Residues in turns of 3,o helix are ovals with amino-acid type and number indicated. Other residues in the structure are in boxes with amino-acid number indicated. Hydrogen bonds between P strands are indicated with arrows. A pseudo two-fold axis of symmetry is indicated with a diamond. This panel was prepared using the output of the program PROMOTIF.

FIGURE 29A is a pseudo two-fold symmetry in Hh-C 7 A stereodiagram of a trace of the a-carbon backbone of residues 258-393 of Hh-C 17 viewed along the pseudo-twofold symmetry axis is shown. Equivalent loops are colored identically. Residues 258-276 and 324-347 are colored yellow, residues 276-301 and 347-373 are colored magenta, and residues 312-320 and 381-389 are colored cyan. The pseudo two-fold axis is indicated with a closed circle.

FIGURE 29B is a stereodiagram of a backbone trace of Hh-C17 is shown with residues 258-323 colored green and residues 324-395 colored yellow. The extended loops that make up the Hh-C, 1 structure are labeled in the order in which they appear in the aminoacid sequence, A1-A2-A3-B1-B2-B3. Two structurally cohesive subdomains are apparent, one comprising loops Al, A2, and B3 and another comprising loops B 1, B2, and A3. Hh-C 7 appears to have arisen from a tandem duplication of a primordial gene to produce the and sequence regions coupled with exchange of the homologous A3 (residues 310-323) and B3 (residues 379-395) loops to form structural subdomains that are hybrids of A' and sequences. A pivot about which exchange 9of these loops appears to have occurred is indicated by an arrow.

WO 98/30576 PCT/US97/15753 FIGURE 29C is a stereodiagram of backbone traces of the regions of Hh-C17 corresponding to the sequence duplication (residues 259-320 colored green and residues 325-389 colored yellow) following superposition is shown. The structures were aligned with the program QUANTA (Polygen). The r.m.s. deviation in a-carbon position for matched residues in the subdomains is 1.38 A. Conserved p turns (see below) are colored red. Panels A, B and C were prepared with MOLSCRIPT (Kraulis, 1991, supra).

FIGURE 29D is a structure-based alignment of the amino-acid sequences of the two subdomains of Hh-C 1 y. Conserved amino acids are highlighted with yellow. Active site residues are in red. P strands are indicated with arrows. P Ib and P2b are slightly longer than la and P2a, respectively, and are indicated with lighter green coloring. Fractional solvent accessibility (FSA) is shown in blue for each residue in the Hh-C, structure. The FSA is the ratio of the solvent accessible surface area of residue X in a Gly-X-Gly tripeptide vs. in the Hh-C17 structure. A value of 0 represents a value from 0.00 to 0.09, 1 represents 0.10 to 0.19, and so on. Type I P turns are conserved at homologous positions in both Hh-C, 1 subdomains at residues 260-263 (homologous to residues 326- 329) and residues 317-320 (homologous to residues 386-389). A type II p turn is conserved between both subdomains at residues 279-282 (homologous to residues 350- 353), and a type IV p turn is conserved between both subdomains at residues 288-291 (homologous to residues 359-362). p bulges are found at homologous positions in both Hh-C,, subdomains at residues 282 (homologous to residue 353) and 300 (homologous to 372).

FIGURE 30A is a stereodiagram of the nucleophilic residue, Cys-258, and nearby residues. Distances between atoms are indicated.

FIGURE 30B is a ribbon diagram of Hh-C, 7 with the side chains of Cys-258 and other putative active site residues indicated. Panels A and B were prepared with MOLSCRIPT (Kraulis, 1991, supra).

WO 98/30576 PCT/US97/15753 -16- FIGURE 30C shows a Coomassie Brilliant Blue-stained SDS-polyacrylamide gel showing in vitro autocleavage reactions of bacterially-expressed His 6 HhC wildtype (lanes 1-3) and mutant proteins, H329A (lanes T326A (lanes and D303A (lanes 10-12). Proteins were incubated with 1 mM DTT (lanes 1, 4, 7, and 10), 50 mM DTT (lanes 2, 5, 8, and 11) or 350 uM cholesterol/1 mM DTT (lanes 3, 6, 9, and 12).

The uncleaved protein migrates as a -29 kDa species. The carboxy-terminal cleavage product migrates as a -25-kDa species and the amino-terminal product migrates as a -7kDa species when DTT-modified or as a ~5-kDa species when cholesterol-modified.

The significant level of apparent cleavage seen with the D303A protein with 1 mM DTT results from preexisting cleavage products in the preparation; however, addition of mM DTT greatly increases the amount of cleavage products and addition of cholesterol does not produce a cholesterol-modified product (-5-kDa species). D303A was also incubated with 46 uM 3 H]cholesterol/1 mM DTT, and no cholesterol-modified product was detected by autoradiography (data not shown). A cholesterol-transfer activity 1% of wildtype could have been detected by this radioassay.

FIGURE 31A is an alignment of the Hh-C, 7 amino-acid sequence (residues 258-402) with other Hh sequences, with nematode sequences homologous to Hh-C, and intein sequences. The alignment was constructed by superimposing the Hh-C and intein alignments produced by the CLUSTALW program using the results of the PSI-BLAST analysis as a guide (Thompson et al., 1994). Additionally, the alignment was verified by analyzing a subset of the sequences containing fifteen diverse intein sequences and three Hh-C sequences with the MACAW program (Schuler et al., 1991, supra). In this analysis the alignment of the blocks containing the cysteine and histidine residues implicated in catalysis was significant with 1 0, and the block including p2b of Hh-C with p< 10- 4 The exact counterpart of P4a in the intein sequences remained uncertain; the respective region is replaced by the number of amino-acid residues. The position of the endonuclease domain (ENDO domain II according to Duan et al., 1997) inserted in the intein sequences is shown and the number of amino acid residues in these domains is WO98/30576 PCT/US97/15753 -17 indicated. A second inserted domain in the PI-SceI/YEAST intein thought to be involved in DNA recognition (DRR) is located between Plb and p2b. Three inteins, GYRA/MYCXE, DNAB/PORPU, and KLBA/METJA, contain a short insert replacing the endonuclease domain. The yeast HO endonuclease does not undergo self-splicing, but contains a vestigial, inactive intein domain. The KLBA/METJA intein homologue in which the amino-terminal nucleophile is replaced by alanine is likely inactive as well.

A consensus sequence is shown above the aligned sequences and shows amino acid residues conserved in at least one half of the sequences in each of the two aligned sets.

indicates a bulky hydrophobic residue L, V, M, F, Y, and indicates a negatively-charged residue (Dor Catalytic site residues are highlighted with red; hydrophobic residues are highlighted with yellow; other residues that conform with the consensus are highlighted with blue. The secondary structure elements and for Hh-C,7 are shown. Every tenth residue in the Hh-C 17 sequence is indicated with a dot. The leftmost column shows abbreviated protein and species names, and the second column shows the gene identification number in the NCBI protein database. Protein name abbreviations: CE(R084B4.1), F46B3 (F46B3.C), M75 (ZK678.5), M89 (C29F3.d), ZK (ZK1290.5), ZK377 (ZK377.1), M1 10 (T05C12.10) uncharacterized nematode proteins containing Hh carboxy-terminal domain homologues; HH hedgehog; EHH Echidna hedgehog; CHH Cephalic hedgehog; DHH Desert hedgehog; IHH Indian hedgehog; BHH Banded hedgehog; TWHH Tiggy-winkle hedgehog; XHH Xenopus hedgehog; SHH Sonic hedgehog; PI-Scel, PI-CtrI yeast intein endonucleases; GYRA, GYRB DNA gyrase A and B subunits; RECA recombinase; DNAB replicative DNA helicase; POLC DNA polymerase III a subunit; CLPP endopeptidase; IF-2 translation initiation factor 2; HELI putative helicase; RFC replication factor C; ORF uncharacterized open reading frame product; G6PT glucose-6-phosphate transaminase; RPO-A', PRO-A" DNA-dependent RNA -polymerase subunits; RGYR reverse gyrase; PEPS phosphoenolpyruvate synthase; UDGD uridine diphosphate glucose dehydrogenase; RNR ribonucleotide reductase; DPOL DNA polymerase, B family; TFIIB transcription factor IIB; KLBA predicted ATPase; HO homothallic endonuclease. Species abbreviations: CAEEL Caenorhabditis elegans; DANRE WO 98/30576 PCTIUS97/15753 -18 Danio rerio; XENLA Xenopus laevis; Cynpy Cynops pyrrhogaster; DROHY Drosophila hydei; DROME Drosophila melanogaster; CANTR Candida tropicalis; MYCLE Mycobacterium leprae; MYCXE Mycobacterium xenopi; MYCTU Mycobacterium tuberculosis; PORPU Porphyra purpurea; SYNSP Synechocystis sp; CHLEU Chlamydomonas; METJA Methanococcus jannaschii; PYRFU Pyrococcus furiosus; PYRSP Pyrococcus sp.; THELI Thermococcus litoralis. Several Hh and intein sequences closely related to those included were omitted.

FIGURE 31B is a stereo ribbon diagram of showing where the endonuclease domain and additional DNA recognition region of PI-Scel are inserted. The loop where the endonuclease domain is inserted is colored red and the loop where the additional DNA recognition region ("the arm of the self-splicing domain" is inserted is colored blue. The orientation of the Hh-C,7 in this view is the same as the orientation of the PI- Scel intein in Figure 2 of Duan, et al., 1997, supra. This panel was prepared with

MOLSCRIPT.

FIGURE 32 is a schematic drawing illustrating the duplication and insertion events that appear to have occurred during the evolution of Hh proteins and inteins. The insertion of the intein into a host protein is not shown. The order of some of these events is speculative. For example, dimerization through loop swapping may have preceded the gene duplication that produced an Hh-C,7-like protein. Abbreviations: Hh-C Hh carboxy-endonuclease domain, DRR DNA recognition region.

FIGURE 33 shows inhibition of cholesterol biosynthesis by the plant steroidal alkaloid, jervine.

WO 98/30576 PCT/US97/15753 -19 DETAILED DESCRIPTION OF THE INVENTION The present invention provides two novel polypeptides originally derived from a single precursor protein, both of which have distinct structural and functional characteristics.

The proteins are derived from a hedgehog protein and can be naturally produced by autoproteolytic cleavage of the full-length hedgehog protein. Based on evidence provided herein, which indicates that hedgehog precursor protein and the auto-proteolytic products of hedgehog precursor protein are expressed in the floorplate of the ventral midline of the neural tube and notochord, the invention now provides a method for the induction of proliferation or differentiation of neuronal cells associated with or in close proximity to the floorplate and notochord. The invention also provides cholesterol modified hedgehog polypeptides and function fragments thereof.

In a first embodiment, the invention provides a substantially pure polypeptide characterized by having an amino acid sequence derived from amino terminal amino acids of a hedgehog protein and having at its carboxy terminus, a glycine-cysteinephenylalanine (G ICF) cleavage site specifically recognized by a proteolytic activity of the carboxy terminal fragment of the native hedgehog polypeptide. This fragment is denoted the N-terminal fragment or polypeptide or herein. For example, in the case of the Drosophila hedgehog, the N fragment includes amino acids 1-257 of hedgehog protein, wherein amino acids 85-257 have a molecular weight of about 19 kD by nonreducing SDS-PAGE (Amino acid residue numbers 1-257 include non-structural features such as signal sequences.). The G I CF cleavage site in Drosophila hedgehog precursor protein occurs at amino acid residues 257-259. Those of skill in the art will be able to identify the G I CF cleavage site in other hedgehog genes, as the amino acid location will be similar and the site will be specifically recognized by the autoproteolytic activity of the corresponding C fragment.

The N-terminal polypeptide is also characterized by being cell-associated in cells expressing the polypeptide in vitro, and being specifically localized in vertebrate or WO 98/30576 PCT/US97/15753 Drosophila cells or embryos, for example. In other words, this N-terminal fragment of hedgehog, remains close to the site of cellular synthesis. The association of N with the cell is a result of the processing event which involves lipophilic modification of the amino terminal domain. (See Figure IE and Example 19) This modification is initiated by the action of the carboxy terminal domain, generating a thioester intermediate; the carboxy-terminal domain thus does not act simply as a protease, although cleavage of a peptide bond does ultimately result from its action. Specifically, the lipid modification is a cholesterol moiety. In addition, the N fragment binds to heparin agarose in vitro.

The N polypeptide of the invention is characterized by having an amino acid sequence derived from amino terminal amino acids of hedgehog protein, 1-257 in Drosophila, wherein amino acids 1-257 have a molecular weight of about 19 kD by non-reducing SDS-PAGE. The N polypeptide includes smaller fragments which retain the functional characteristics of full length N, bind to heparin. The hedgehog protein from which N is derived includes, but is not limited to Drosophila, Xenopus, chicken, zebrafish, mouse, and human. Crystallographic analysis shows the structure of SHH-N includes the presence of a zinc ion. While not wanting to be bound by a particular theory, the presence of the zinc ion is suggestive of zinc hydrolase activity. Zinc hydrolases include proteases such as carboxypeptidase A and thermolysin, lipases such as phospholipase C, and other enzymes such as carbonic anhydrase. Alterations in the zinc hydrolase site of the amino terminal signaling domain may be useful for modulating the range of diffusion of a hedgehog protein or to alter the signaling characteristics of the amino terminal signaling domain. For example, a mutation in the zinc hydrolase site may result in a tethered protein where ordinarily the protein is secreted at a distance. The result would be induction of a cell type not typically induced. Alteration in the zinc site may result in a molecule capable of inducing motor neurons and not floor plate, and vice versa.

The identification of a cell-surface, or extracellular matrix localization of N and its expression in notochord and floor plate-associated cells, provides a means for isolation or specific selection of cells expressing N, to isolate a notochord sample or to isolate WO 98/30576 PCT/US97/15753 -21 floor plate cells. In addition, antibodies directed to N are useful for histological analysis of tissues suspected of expressing N protein.

The invention also provides a substantially pure polypeptide characterized by having an amino acid sequence derived from carboxy terminal amino acids of a hedgehog protein and having at its amino terminus a GICF cleavage site specifically recognized by a proteolytic activity of the carboxy terminal fragment of the native hedgehog polypeptide.

This fragment is denoted the C-terminal fragment or polypeptide or herein. For example, in Drosophila this polypeptide derives from the C-terminal domain of hedgehog precursor protein beginning at amino acid residue 258, wherein the full length C-terminal domain has a molecular weight of about 25 kD by non-reducing SDS-PAGE, a histidine residue at position 72, and has protease activity. The GICF cleavage site specifically recognized by the proteolytic activity of the carboxy terminal fragment of the native hedgehog polypeptide is located at amino acid residues 257-259. As described above for the N fragment, now that the present invention has shown the precise cleavage recognition site for the autoproteolytic domain of hedgehog, those of skill in the art can readily discern the cleavage site in other hedgehog proteins thereby allowing the ready identification of any N or C polypeptide of any hedgehog precursor protein.

The polypeptide of the invention is derived from the C-terminus of a hedgehog precursor protein, beginning at the autoproteolytic cleavage site identified at the GCF amino acid sequence, which in Drosophila corresponds to amino acids 257-259. In Drosophila the histidine residue found invariably at amino acid residue 329 of the native hedgehog protein, and at amino acid residue 72 of the C polypeptide, is essential for auto-proteolytic cleavage between amino acids 257 and 258 (G and Corresponding C-polypeptides of the invention will likewise contain a similarly located histidine residue which can be readily identified, such as by comparison to the Drosophila C-polypeptide.

Among various species, the proteolytic domain can be characterized by the amino acid sequence -XTXXHLXX-.

WO 98/30576 PCTIUS97/15753 -22- The C polypeptide of the invention, unlike N, does not significantly bind to heparin agarose. C is characterized by being released into the culture supernatant of cells expressing C polypeptide in vitro and by being localized diffusely in cells and embryos.

Because C polypeptide diffuses freely, it would be detectable in various body fluids and tissues in a subject. Identification of C polypeptide expression near the midline of the neural tube, as described herein, provides a useful assay for neural tube closure in an embryo/fetus, for example. The presence of C polypeptide in amniotic fluid would be diagnostic of a disorder in which the neural tube may be malformed.

Altered levels of C polypeptide in cerebrospinal fluid may be indicative of neurodegenerative disorders, for example. Because C polypeptide is released from the cell after synthesis and autoproteolysis of native hedgehog precursor polypeptide, tumors synthesizing and releasing high levels ofC polypeptide would be detectable without prior knowledge of the exact location of the tumor.

C fragment is effective in inducing genes of the pituitary and anterior brain as well. In particular, induction is increased by the addition of a member of the TGF-P family of growth factors. For example, human activin in combination with C fragment may be effective in enhancing pituitary cell growth and activity or development. C fragment possesses cholesterol transferase activity thereby effecting precursor cleavage and transfer of a cholesterol moiety to N fragment, resulting in a biologically active N fragment.

C fragment is effective in inducing posterior markers of the brain by inhibiting N. Such a fragment is exemplified in Example 18 as AN-C. Therefore in another embodiment, the invention includes a polypeptide deleting amino acid residues 28-194 of X-bhh.

(Autoproteolysis gives a C domain of 198-409 as well as a seven amino acid peptide, representing aa 24-27 and 195-197). This polypeptide blocks the activity of X-bhh and N in explants and reduces dorsoanterior structures in embryos. Also included are polynucleotide sequences encoding AN-C. AN-C is useful for increasing expression of WO 98/30576 PCTIUS97/15753 -23 posterior neural markers En-2, Krox-20, Xlttbox-6) and decreasing expression of anterior neural markers XANF-2, XAG-1, Otx-A) when desirable to do so to modulate neural patterning.

The term "substantially pure" as used herein refers to hedgehog N or C polypeptide which is substantially free of other proteins, lipids, carbohydrates, nucleic acids or other materials with which it is naturally associated. One skilled in the art can purify hedgehog N or C polypeptide using standard techniques for protein purification. The substantially pure polypeptide will yield a single major band on a non-reducing polyacrylamide gel.

The purity of the hedgehog N or C polypeptide can also be determined by aminoterminal amino acid sequence analysis.

The invention includes a functional N or C polypeptide, and functional fragments thereof. As used herein, the term "functional polypeptide" or "functional fragment" refers to a polypeptide which possesses a biological function or activity which is identified through a defined functional assay and which is associated with a particular biologic, morphologic, or phenotypic alteration in the cell. Functional fragments of the hedgehog N or C polypeptide include fragments of N or C polypeptide as long as the activity, proteolytic activity or cholesterol transferase activity of C polypeptide remains. Smaller peptides containing the biological activity of N or C polypeptide are therefore included in the invention. The biological function, for example, can vary from a polypeptide fragment as small as an epitope to which an antibody molecule can bind to a large polypeptide which is capable of participating in the characteristic induction or programming of phenotypic changes within a cell. A "functional polynucleotide" denotes a polynucleotide which encodes a functional polypeptide as described herein.

Biologically active or functional fragments of hedgehog, as described herein, are included in the invention and can be identified as such by functional assays. For example, fragments of hedgehog are identified as inducing differentiation of neuronal cells; regulating differentiationof chondrocytes; able to complement a loss of function WO 98/30576 PCT/US97/15753 -24 mutation of hedgehog, for example in a transgenic Drosophila; binding to Patched (Ptc); or having cholesterol transferase activity C fragment). Fragments of the invention may be from about 30 to 450 amino acids in length; from about 50 to 300 amino acids in length; from about 75 to 250 amino acids in length; or from about 100 to 200 amino acids in length, as long as a biological activity of hedgehog is retained therein.

Minor modifications of the N or C polypeptide primary amino acid sequence may result in polypeptides which have substantially equivalent activity as compared to the N or C polypeptide described herein. Such modifications may be deliberate, as by site-directed mutagenesis, or may be spontaneous. All of the polypeptides produced by these modifications are included herein as long as the proteolytic activity of C polypeptide, for example, is present. Further, deletion of one or more amino acids can also result in a modification of the structure of the resultant molecule without significantly altering its activity. This can lead to the development of a smaller active molecule which would have broader utility. For example, it is possible to remove amino or carboxy terminal amino acids which may not be required for N or C polypeptide activity.

The N or C polypeptide of the invention also includes conservative variations of the polypeptide sequence. The term "conservative variation" as used herein denotes the replacement of an amino acid residue by another biologically similar residue. Examples of conservative variations include the substitution of one hydrophobic residue such as isoleucine, valine, leucine or methionine for another, or the substitution of one polar residue for another, such as the substitution of arginine for lysine, glutamic for aspartic acids, or glutamine for asparagine, and the like. The term "conservative variation" also includes the use of a substituted amino acid in place of an unsubstituted parent amino acid provided that antibodies raised to the substituted polypeptide also immunoreact with the unsubstituted polypeptide.

The N fragment of the invention includes both the active form of the polypeptide and the N fragment including the uncleaved signal sequence. For example, in Drosophila where WO 98/30576 PCT/US97/15753 the signal sequence is internal (at about amino acids 60-80), the entire uncleaved N fragment beginning at the initiating methionine is included in the invention. Those of skill in the art can readily ascertain the nature and location of the signal sequence by using, for example, the algorithm described in von Heijne, Nucl. Acids Res. 14:4683, (1986).

Hedgehog polypeptides of the invention include polypeptides having at least.about 100% homology with the hedgehog polypeptides provided herein, for example 52%, 64%, 68%, 70%, 75%, 80%, 85%, 90%, 95% and up to 100% homology. Preferably homologous polypeptides are derived from vertebrate species, most preferably mammalian species, such as humans.

The invention also provides an isolated polynucleotide sequence encoding a polypeptide having the amino acid sequence of N or C polypeptide of the invention. The term "isolated" as used herein includes polynucleotides substantially free of other nucleic acids, proteins, lipids, carbohydrates or other materials with which it is naturally associated. Polynucleotide sequences of the invention include DNA, cDNA and RNA sequences which encode N or C polypeptide. It is understood that all polynucleotides encoding all or a portion of N or C polypeptide are also included herein, as long as they encode a polypeptide with N or C polypeptide activity. Such polynucleotides include naturally occurring, synthetic, and intentionally manipulated polynucleotides. For example, N or C polypeptide polynucleotide may be subjected to site-directed mutagenesis. The polynucleotide sequence for N or C polypeptide also includes antisense sequences. The polynucleotides of the invention include sequences that are degenerate as a result of the genetic code. There are 20 natural amino acids, most of which are specified by more than one codon. Therefore, all degenerate nucleotide sequences are included in the invention as long as the amino acid sequence of N or C polypeptide polypeptide encoded by the nucleotide sequence is functionally unchanged.

In addition, the invention also includes a polynucleotide consisting essentially of a polynucleotide sequence encoding a polypeptide having an amino acid sequence of N or WO 98/30576 PCT/US97/15753 -26- C and having at least one epitope for an antibody immunoreactive with N or C polypeptide.

The polynucleotide encoding N or C polypeptide includes the entire polypeptide or fragments thereof, as well as nucleic acid sequences complementary to that sequence.

A complementary sequence may include an antisense nucleotide. When the sequence is RNA, the deoxynucleotides A, G, C, and T are replaced by ribonucleotides A, G, C, and U, respectively. Also included in the invention are fragments of the above-described nucleic acid sequences that are at least 15 bases in length, which is sufficient to permit the fragment to selectively hybridize to DNA that encodes the protein under physiological conditions.

Hedgehog encoding polynucleotides of the invention include nucleic acid sequences identified by hybridization to a hedgehog nucleic acid described herein. In nucleic acid hybridization reactions, the conditions used to achieve a particular level of stringency will vary, depending on the nature of the nucleic acids being hybridized. For example, the length, degree of complementarity, nucleotide sequence composition GC v. AT content), and nucleic acid type RNA v. DNA) of the hybridizing regions of the nucleic acids can be considered in selecting hybridization conditions. An additional consideration is whether one of the nucleic acids is immobilized, for example, on a filter.

An example of progressively higher stringency conditions is as follows: 2 x SSC/0.1% SDS at about room temperature (hybridization conditions); 0.2 x SSC/0.1% SDS at about room temperature (low stringency conditions); 0.2 x SSC/0.1% SDS at about 42°C (moderate stringency conditions); and 0.1 x SSC at about 68 0 C (high stringency conditions). Washing can be carried out using only one of these conditions, high stringency conditions, or each of the conditions can be used, for 10-15 minutes each, in the order listed above, repeating any or all.of the steps listed. However, as mentioned above, optimal conditions will vary, depending on the particular hybridization reaction involved, and can be determined empirically.

WO 98/30576 PCT/US97/15753 -27 DNA sequences of the invention can be obtained by several methods. For example, the DNA can be isolated using hybridization techniques which are well known in the art.

These include, but are not limited to: I) hybridization of genomic or cDNA libraries with probes to detect homologous nucleotide sequences; 2) antibody screening of expression to detect cloned DNA fragments with shared structural features; and 3) PCR amplification of a desired nucleotide sequence using oligonucleotide primers.

Preferably the hedgehog, N, or C polynucleotide of the invention is derived from a vertebrate organism, and most preferably from human. Screening procedures which rely on nucleic acid hybridization make it possible to isolate any gene sequence from any organism, provided the appropriate probe is available. Oligonucleotide probes, which correspond to a part of the sequence encoding the protein in question, can be synthesized chemically. This requires that short, oligopeptide stretches of amino acid sequence must be known. The DNA sequence encoding the protein can be deduced from the genetic code, however, the degeneracy of the code must be taken into account. It is possible to perform a mixed addition reaction when.the sequence is degenerate. This includes a heterogeneous mixture of denatured double-stranded DNA. For such screening, hybridization is preferably performed on either single-stranded DNA or denatured double-stranded DNA. Hybridization is particularly useful in the detection of cDNA clones derived from sources where an extremely low amount of mRNA sequences relating to the polypeptide of interest are present. In other words, by using stringent hybridization conditions directed to avoid non-specific binding, it is possible, for example, to allow the autoradiographic visualization of a specific cDNA clone by the hybridization of the target DNA to that single probe in the mixture which is its complete complement (Wallace, et al., Nucl. Acid Res., 9:879, 1981).

The development of specific DNA sequences encoding hedgehog can also be obtained by: 1)isolation of double-stranded DNA sequences from the genomic DNA; 2) chemical manufacture of a DNA sequence to provide the necessary codons for the polypeptide of interest; and 3) in vitro synthesis of a double-stranded DNA sequence by reverse WO 98/30576 PCT/US97/15753 -28 transcription ofmRNA isolated from a eukaryotic donor cell. In the latter case, a doublestranded DNA complement of mRNA is eventually formed which is generally referred to as cDNA.

Of the three above-noted methods for developing specific DNA sequences for use in recombinant procedures, the isolation of genomic DNA isolates is the least common.

This is especially true when it is desirable to obtain the microbial expression of mammalian polypeptides due to the presence of introns.

The synthesis of DNA sequences is frequently the method of choice when the entire sequence of amino acid residues of the desired polypeptide product is known. When the entire sequence of amino acid residues of the desired polypeptide is not known, the direct synthesis of DNA sequences is not possible and the method of choice is the synthesis of cDNA sequences. Among the standard procedures for isolating cDNA sequences of interest is the formation of plasmid- or phage-carrying cDNA libraries which are derived from reverse transcription of mRNA which is abundant in donor cells that have a high level of genetic expression. When used in combination with polymerase chain reaction technology, even rare expression products can be cloned. In those cases where significant portions of the amino acid sequence of the polypeptide are known, the production of labeled single or double-stranded DNA or RNA probe sequences duplicating a sequence putatively present in the target cDNA may be employed in DNA/DNA hybridization procedures which are carried out on cloned copies of the cDNA which have been denatured into a single-stranded form (Jay, et al., Nucl. Acid Res., 11:2325, 1983).

A preferred method for obtaining genomic DNA, for example, is Polymerase Chain Reaction (PCR), which relies on an in vitro method of nucleic acid synthesis by which a particular segment of DNA is specifically replicated. Two oligonucleotide primers that flank the DNA fragment to be amplified are utilized in repeated cycles of heat denaturation of the DNA, annealing of the primers to their complementary sequences, WO 98/30576 PCT/US97/15753 29 and extension of the annealed primers with DNA polymerase. These primers hybridize to opposite strands of the target sequence and are oriented so that DNA synthesis by the polymerase proceeds across the region between the primers. Since the extension products themselves are also complementary to and capable of binding primers, successive cycles of amplification essentially double the amount of the target DNA synthesized in the previous cycle. The result is an exponential accumulation of the specific target fragment, approximately where n is the number of cycles of amplification performed (see PCR Protocols, Eds. Innis, et al., Academic Press, Inc., 1990, incorporated herein by reference).

A cDNA expression library, such as Xgtl 1, can be screened indirectly for hedgehog, N, or C polypeptides having at least one epitope, using antibodies specific for hedgehog, N, or C. Such antibodies can be either polyclonally or monoclonally derived and used to detect expression product indicative of the presence of the desired hedgehog cDNA.

The polynucleotide sequence for hedgehog, N, or C, also includes sequences complementary to the polynucleotide encoding hedgehog, N or C (antisense sequences).

Antisense nucleic acids are DNA or RNA molecules that are complementary to at least a portion of a specific mRNA molecule (Weintraub, Scientific American, 262:40, 1990).

The invention embraces all antisense polynucleotides capable of inhibiting production of hedgehog, N, or C polypeptide. In the cell, the antisense nucleic acids hybridize to the corresponding mRNA, forming a double-stranded molecule. The antisense nucleic acids interfere with the translation of the mRNA since the cell will not translate a mRNA that is double-stranded. Antisense oligomers of about 15 nucleotides are preferred, since they are easily synthesized and are less likely to cause problems than larger molecules when introduced into the target hedgehog, N, or C-producing cell. The use of antisense methods to inhibit the translation of genes is well known in the art (Marcus-Sakura, Anal.

Biochem., 172:289, 1988). Inhibition of target nucleotide would be desirable, for example, in inhibiting cell-proliferative disorders, such as certain tumors, which are mediated by hedgehog, N or C.

WO 98/30576 PCT/US97/15753 30 In addition, ribozyme nucleotide sequences for hedgehog, N or C are included in the invention. Ribozymes are RNA molecules possessing the ability to specifically cleave other single-stranded RNA in a manner analogous to DNA restriction endonucleases.

Through the modification of nucleotide sequences which encode these RNAs, it is possible to engineer molecules that recognize specific nucleotide sequences in an RNA molecule and cleave it (Cech, J. Amer. Med. Assn., 260:3030, 1988). A major advantage of this approach is that, because they are sequence-specific, only mRNAs with particular sequences are inactivated.

There are two basic types of ribozymes namely, tetrahymena-type (Hasselhoff, Nature, 334:585, 1988) and "hammerhead"-type. Tetrahymena-type ribozymes recognize sequences which are four bases in length, while "hammerhead"-type ribozymes recognize base sequences 11-18 bases in length. The longer the recognition sequence, the greater the likelihood that sequence will occur exclusively in the target mRNA species. Consequently, hammerhead-type ribozymes are preferable to tetrahymena-type ribozymes for inactivating a specific mRNA species and 18-based recognition sequences are preferable to shorter recognition sequences.

DNA sequences encoding hedgehog, N or C can be expressed in vitro by DNA transfer into a suitable host cell. "Host cells" are cells in which a vector can be propagated and its DNA expressed. The term also includes any progeny of the subject host cell. It is understood that all progeny.may not be identical to the parental cell since there may be mutations that occur during replication. However, such progeny are included when the term "host cell" is used. Methods of stable transfer, meaning that the foreign DNA is continuously maintained in the host, are known in the art.

In the present invention, the hedgehog, N or C polynucleotide sequences may be inserted into a recombinant expression vector. The term "recombinant expression vector" refers to a plasmid, virus or other vehicle known in the art that has been manipulated by insertion or incorporation of the hedgehog, N or C genetic sequences. Such expression WO 98/30576 PCT/US97/15753 31 vectors contain a promoter sequence which facilitates the efficient transcription of the inserted genetic sequence of the host. The expression vector typically contains an origin of replication, a promoter, as well as specific genes which allow phenotypic selection of the transformed cells. Vectors suitable for use in the present invention include, but are not limited to the T7-based expression vector for expression in bacteria (Rosenberg, et al., Gene, 56:125, 1987), the pMSXND expression vector for expression in mammalian cells (Lee and Nathans, J Biol. Chem., 263:3521, 1988) and baculovirus-derived vectors for expression in insect cells. The DNA segment can be present in the vector operably linked to regulatory elements, for example, a promoter T7, metallothionein I, or polyhedrin promoters).

Polynucleotide sequences encoding hedgehog, N or C can be expressed in either prokaryotes or eukaryotes, although post-translational modification of eukaryotically derived polypeptides, such as carboxylation, would occur in a eukaryotic host. Hosts can include microbial, yeast, insect and mammalian organisms. Methods of expressing DNA sequences having eukaryotic or viral sequences in prokaryotes are well known in the art.

Biologically functional viral and plasmid DNA vectors capable of expression and replication in a host are known in the art. Such vectors are used to incorporate DNA sequences of the invention.

Methods which are well known to those skilled in the art can be used to construct expression vectors containing the hedgehog, N or C coding sequence and appropriate transcriptional/translational control signals. These methods include in vitro recombinant DNA techniques, synthetic techniques, and in vivo recombination/genetic techniques.

See, for example, the techniques described in Maniatis, et al., 1989 Molecular Cloning A Laboratory Manual, Cold Spring Harbor Laboratory, N.Y.

A variety of host-expression vector systems may be utilized to express the hedgehog, N or C coding sequence. These include but are not limited to microorganisms such as bacteria transformed with recombinant bacteriophage DNA, plasmid DNA or cosmid WO 98/30576 PCT/US97/15753 32 DNA expression vectors containing the hedgehog, N or C coding sequence; yeast transformed with recombinant yeast expression vectors containing the hedgehog, N or C coding sequence; plant cell systems infected with recombinant virus expression vectors cauliflower mosaic virus, CaMV; tobacco mosaic virus, TMV) or transformed with recombinant plasmid expression vectors Ti plasmid) containing the Hedgehog, N or C coding sequence; insect cell systems infected with recombinant virus expression vectors baculovirus) containing the hedgehog, N or C coding sequence; or animal cell systems infected with recombinant virus expression vectors retroviruses, adenovirus, vaccinia virus) containing the hedgehog, N or C coding sequence, or transformed animal cell systems engineered for stable expression.

Depending on the host/vector system utilized, any of a number of suitable transcription and translation elements, including constitutive and inducible promoters, transcription enhancer elements, transcription terminators, etc., may be used in the expression vector (see Bitter, et al., 1987, Methods in Enzymology, 153:516-544). For example, when cloning in bacterial systems, inducible promoters such as pL of bacteriophage y, plac, ptrp, ptac (ptrp-lac hybrid promoter) and the like may be used. When cloning in mammalian cell systems, promoters derived from the genome of mammalian cells metallothionein promoter) or from mammalian viruses the retrovirus long terminal repeat; the adenovirus late promoter; the vaccinia virus 7.5K promoter) may be used.

Promoters produced by recombinant DNA or synthetic techniques may also be used to provide for transcription of the inserted hedgehog, N or C coding sequence.

In bacterial systems a number of expression vectors may be advantageously selected depending upon the use intended for the expressed. For example, when large quantities of hedgehog, N or C are to be produced, vectors which direct the expression of high levels of fusion protein products that are readily purified may be desirable. Those which are engineered to contain a cleavage site to aid in recovering are preferred. Such vectors include but are not limited to the E. coli expression vector pUR278 (Ruther, et al., EMBO 2:1791, 1983), in which the Hedgehog, N or C coding sequence may be ligated into WO 98/30576 PCT/US97/15753 -33 the vector in frame with the lac Z coding region so that a hybrid -lac Z protein is produced; pIN vectors (Inouye and Inouye, Nucleic Acids Res., 13:3101, 1985; Van Heeke and Schuster, J. Biol. Chem. 264:5503, 1989) and the like.

In yeast, a number of vectors containing constitutive or inducible promoters may be used.

For a review see, Current Protocols in Molecular Biology, Vol. 2, 1988, Ed. Ausubel, et al., Greene Publish. Assoc. Wiley Interscience, Ch. 13; Grant, et al., 1987, Expression and Secretion Vectors for Yeast, in Methods in Enzymology, Eds. Wu and Grossman, 31987, Acad. Press, Vol. 153, pp.516-544; Glover, 1986, DNA Cloning, Vol. II, IRL Press, Wash., Ch. 3; and Bitter, 1987, Heterologous Gene Expression in Yeast, Methods in Enzymology, Eds. Berger and Kimmel, Acad. Press, Vol. 152, pp. 673-684; and The Molecular Biology of the Yeast Saccharomyces, 1982, Eds. Strathem, et al., Cold Spring Harbor Press, Vols. I and II. A constitutive yeast promoter such as ADH or LEU2 or an inducible promoter such as GAL may be used (Cloning in Yeast, Ch. 3, R. Rothstein In: DNA Cloning Vol.11, A Practical Approach, Ed. DM Glover, 1986, IRL Press, Wash., Alternatively, vectors may be used which promote integration of foreign DNA sequences into the yeast chromosome.

In cases where plant expression vectors are used, the expression of the hedgehog, N or C coding sequence may be driven by any of a number of promoters. For example, viral promoters such as the 35S RNA and 19S RNA promoters of CaMV (Brisson,. et al., Nature, 310:511, 1984), or the coat protein promoter to TMV (Takamatsu, et al., EMBO 6:307, 1987) may be used; alternatively, plant promoters such as the small subunit of RUBISCO (Coruzzi, et al., EMBO 3:1671-1680, 1984; Broglie, et al., Science, 224:838, 1984); or heat shock promoters, soybean hspl7.5-E or hspl7.3-B (Gurley, et al., Mol. Cell. Biol., 6:559, 1986) may be used. These constructs can be introduced into plant cells using Ti plasmids, Ri plasmids, plant virus vectors, direct DNA transformation, microinjection, electroporation, etc. For reviews of such techniques see, Sfor example, Weissbach and Weissbach, 1988, Methods for Plant Molecular Biology, WO 98/30576 PCT/US97/15753 34 Academic Press, NY, Section VIII, pp. 421-463; and Grierson and Corey, 1988, Plant Molecular Biology, 2d Ed., Blackie, London, Ch. 7-9.

An alternative expression system which could be used to express is an insect system.

In one such system, Autographa californica nuclear polyhedrosis virus (AcNPV) is used as a vector to express foreign genes. The virus grows in Spodoptera frugiperda cells.

The hedgehog, N or C coding sequence may be cloned into non-essential regions (for example the polyhedrin gene) of the virus and placed under control of an AcNPV promoter (for example the polyhedrin promoter). Successful insertion of the hedgehog, N or C coding sequence will result in inactivation of the polyhedrin gene and production of non-occluded recombinant virus virus lacking the proteinaceous coat coded for by the polyhedrin gene). These recombinant viruses are then used to infect Spodoptera frugiperda cells in which the inserted gene is expressed. see Smith, et al., J. Viol., 46:584, 1983; Smith, U.S. Patent No. 4,215,051).

Eukaryotic systems, and preferably mammalian expression systems, allow for proper post-translational modifications of expressed mammalian proteins to occur. Eukaryotic cells which possess the cellular machinery for proper processing of the primary transcript, glycosylation, phosphorylation, and advantageously, secretion of the gene product may be used as host cells for the expression of hedgehog, N or C. Mammalian cell lines may be preferable. Such host cell lines may include but are not limited to CHO, VERO, BHK, HeLa, COS, MDCK, -293, and WI38.

Mammalian cell systems which utilize recombinant viruses or viral elements to direct expression may be engineered. For example, when using adenovirus expression vectors, the hedgehog, N or C coding sequence may be ligated to an adenovirus transcription/translation control complex, the late promoter and tripartite leader sequence. This chimeric gene may then be inserted in the adenovirus genome by in vitro or in vivo recombination. Insertion in a non-essential region of the viral genome region El or E3) will result in a recombinant virus that is viable and capable of expressing the WO 98/30576 PCT/US97/15753 35 protein in infected hosts see Logan and Shenk, Proc. Natl. Acad. Sci. USA, 81:3655, 1984). Alternatively, the vaccinia virus 7.5K promoter may be used. see, Mackett, et al., Proc. Natl. Acad. Sci. USA, 79:7415, 1982; Mackett, et al., J. Virol., 49: 857, 1984; Panicali, et al., Proc. Natl. Acad. Sci. USA, 79:4927, 1982). Of particular interest are vectors based on bovine papilloma virus which have the ability to replicate as extrachromosomal elements (Sarver, et al., Mol. Cell. Biol., 1:486, 1981). Shortly after entry of this DNA into mouse cells, the plasmid replicates to about 100 to 200 copies per cell. Transcription of the inserted cDNA does not require integration of the plasmid into the host's chromosome, thereby yielding a high level of expression. These vectors can be used for stable expression by including a selectable marker in the plasmid, such as, for example, the neo gene. Alternatively, the retroviral genome can be modified for use as a vector capable of introducing and directing the expression of the hedgehog, N or C gene in host cells (Cone and Mulligan, Proc. Natl. Acad. Sci. USA, 81:6349, 1984). High level expression may also be achieved using inducible promoters, including, but not limited to, the metallothionine IIA promoter and heat shock promoters.

For long-term, high-yield production of recombinant proteins, stable expression is preferred. Rather than using expression vectors which contain viral origins of replication, host cells can be transformed with the hedgehog, N or C cDNA controlled by appropriate expression control elements promoter, enhancer, sequences, transcription terminators, polyadenylation sites, etc.), and a selectable marker. The selectable marker in the recombinant plasmid confers resistance to the selection and allows cells to stably integrate the plasmid into their chromosomes and grow to form foci which in turn can be cloned and expanded into cell lines. For example, following the introduction of foreign DNA, engineered cells may be allowed to grow for 1-2 days in an enriched media, and then are switched to a selective media. A number of selection systems may be used, including but not limited to the herpes simplex virus thymidine kinase (Wigler, et al., Cell, 11: 223, 1977), hypoxanthine-guanine phosphoribosyltransferase (Szybalska and Szybalski, Proc. Natl. Acad. Sci. USA, 48:2026, 1962), and adenine phosphoribosyltransferase (Lowy, et al., Cell, 22: 817, WO 98/30576 PCT/US97/15753 -36- 1980) genes can be employed in tk', hgprt- or aprt' cells respectively. Also, antimetabolite resistance can be used as the basis of selection for dhfr, which confers resistance to methotrexate (Wigler, et al., Natl. Acad Sci. USA, 77: 3567, 1980; O'Hare, et al., Proc.

Natl. Acad. Sci. USA, 78: 1527, 1981); gpt, which confers resistance to mycophenolic acid (Mulligan and Berg, Proc. Natl. Acad Sci. USA, 78: 2072, 1981; neo, which confers resistance to the aminoglycoside G-418 (Colberre-Garapin, et al., J. Mol. Biol., 150:1, 1981); and hygro, which confers resistance to hygromycin (Santerre, et al., Gene, 30:147, 1984) genes. Recently, additional selectable genes have been described, namely trpB, which allows cells to utilize indole in place of tryptophan; hisD, which allows cells to utilize histinol in place of histidine (Hartman and Mulligan, Proc. Natl. Acad. Sci. USA, 85:8047, 1988); and ODC (omithine decarboxylase) which confers resistance to the ornithine decarboxylase inhibitor, 2-(difluoromethyl)-DL-omithine, DFMO (McConlogue 1987, In: Current Communications in Molecular Biology, Cold Spring Harbor Laboratory ed.).

Transformation of a host cell with recombinant DNA may be carried out by conventional techniques as are well known to those skilled in the art. Where the host is prokaryotic, such as E. coli, competent cells which are capable of DNA uptake can be prepared from cells harvested after exponential growth phase and subsequently treated by the CaCI, method using procedures well known in the art. Alternatively, MgCI, or RbCl can be used. Transformation can also be performed after forming a protoplast of the host cell if desired.

When the host is a eukaryote, such methods of transfection of DNA as calcium phosphate co-precipitates, conventional mechanical procedures such as microinjection, electroporation, insertion of a plasmid encased in liposomes, or virus vectors may be used.

Eukaryotic cells can also be cotransformed with DNA sequences encoding the hedgehog, N or C of the invention, and a second foreign DNA molecule encoding a selectable phenotype, such as the herpes simplex thymidine kinase gene. Another method is to use a eukaryotic viral vector, such as simian virus 40 (SV40) or bovine papilloma virus, to WO 98/30576 PCT/US97/15753 37 transiently infect or transform eukaryotic cells and express the protein. (see for example, Eukaryotic Viral Vectors, Cold Spring Harbor Laboratory, Gluzman ed., 1982).

Isolation and purification of microbial expressed polypeptide, or fragments thereof, provided by the invention, may be carried out by conventional means including preparative chromatography and immunological separations involving monoclonal or polyclonal antibodies.

The invention includes antibodies immunoreactive with or which bind to hedgehog, N or C polypeptide or functional fragments thereof. Antibody which consists essentially of pooled monoclonal antibodies with different epitopic specificities, as well as distinct monoclonal antibody preparations are provided. Monoclonal antibodies are made from antigen containing fragments of the protein by methods well known to those skilled in the art (Kohler, et al., Nature, 256:495, 1975). The term antibody as used in this invention is meant to include intact molecules as well as fragments thereof, such as Fab and which are capable of binding an epitopic determinant on hedgehog, N or C.

The antibodies of the invention include antibodies which bind to the N or C polypeptide and which bind with immunoreactive fragments N or C.

The term "antibody" as used in this invention includes intact molecules as well as fragments thereof, such as Fab, F(ab')z, and Fv which are capable of binding the epitopic determinant. These antibody fragments retain some ability to selectively bind with its antigen or receptor and are defined as follows: Fab, the fragment which contains a monovalent antigen-binding fragment of an antibody molecule can be produced by digestion of whole antibody with the enzyme papain to yield an intact light chain and a portion of one heavy chain; WO 98/30576 PCTUS97/15753 -38 Fab', the fragment of an antibody molecule can be obtained by treating whole antibody with pepsin, followed by reduction, to yield an intact light chain and a portion of the heavy chain; two Fab' fragments are obtained per antibody molecule; the fragment of the antibody that can be obtained by treating whole antibody with the enzyme pepsin without subsequent reduction; is a dimer of two Fab' fragments held together by two disulfide bonds; Fv, defined as a genetically engineered fragment containing the variable genetically fused single chain molecule.

Methods of making these fragments are known in the art. (See for example, Harlow and Lane, Antibodies: A Laboratory Manual, Cold Spring Harbor Laboratory, New York (1988), incorporated herein by reference).

As used in this invention, the term "epitope" means any antigenic determinant on an antigen to which the paratope of an antibody binds. Epitopic determinants usually consist of chemically active surface groupings of molecules such as amino acids or sugar side chains and usually have specific three dimensional structural characteristics, as well as specific charge characteristics.

Antibodies which bind to the hedgehog, N or C polypeptide of the invention can be prepared using an intact polypeptide or fragments containing small peptides of interest as the immunizing antigen. The polypeptide such as N or C, or fragments thereof used to immunize an animal can be derived from translated cDNA or chemical synthesis which can be conjugated to a carrier protein, if desired. Such commonly used carriers which are chemically coupled to the peptide include keyhole limpet hemocyanin (KLH), thyroglobulin, bovine serum albumin (BSA), and tetanus toxoid. The coupled peptide is then used to immunize the animal a mouse, a rat, or a rabbit).

WO 98/30576 PCT/US97/15753 -39 If desired, polyclonal or monoclonal antibodies can be further purified, for example, by binding to and elution from a matrix to which the polypeptide or a peptide to which the antibodies were raised is bound. Those of skill in the art will know of various techniques common in the immunology arts for purification and/or concentration of polyclonal antibodies, as well as monoclonal antibodies (See for example, Coligan, et al., Unit 9, Current Protocols in Immunology, Wiley Interscience, 1991, incorporated by reference).

It is also possible to use the anti-idiotype technology to produce monoclonal antibodies which mimic an epitope. For example, an anti-idiotypic monoclonal antibody made to a first monoclonal antibody will have a binding domain in the hypervariable region which is the "image" of the epitope bound by the first monoclonal antibody.

Antibodies as described herein as having specificity for N polypeptide, Abl (residues 83-160), are useful for specific identification of cells or tissues expressing the N fragment of hedgehog. Similarly, antibodies described herein as having specificity for C polypeptide, Ab2 (residues 300-391), are useful for specific identification of cells or tissues expressing the C fragment of hedgehog. Both antibodies, naturally, will also detect native hedgehog polypeptide.

The N and C-specific antibodies of the invention are useful for purification of N and C polypeptide, respectively, especially using the antibodies immobilized on solid phase.

By contacting a sample with anti-N antibody, both N and native hedgehog polypeptides can be isolated. By next contacting the sample removed by anti-N antibodies, with anti- C antibodies, the native hedgehog polypeptide is removed, thus allowing purification of N polypeptide. In a similar manner, C polypeptide can be antibody purified from a sample.

Monoclonal antibodies of the invention are suited for use, for example, in immunoassays in which they can be utilized in liquid phase or bound to a solid phase carrier. In addition, the monoclonal antibodies in these immunoassays can be detectably labeled in various ways. Examples of types of immunoassays which can utilize monoclonal WO 98/30576 PCT/US97/15753 antibodies of the invention are competitive and non-competitive immunoassays in either a direct or indirect format. Examples of such immunoassays are the radioimmunoassay (RIA) and the sandwich (immunometric) assay. Detection of the antigens using the monoclonal antibodies of the invention can be done utilizing immunoassays which are run in either the forward, reverse, or simultaneous modes, including immunohistochemical assays on physiological samples. Those of skill in the art will know, or can readily discern, other immunoassay formats without undue experimentation.

The term "immunometric assay" or "sandwich immunoassay", includes simultaneous sandwich, forward sandwich and reverse sandwich immunoassays. These terms are well understood by those skilled in the art. Those of skill will also appreciate that antibodies according to the present invention will be useful in other variations and forms of assays which are presently known or which may be developed in the future. These are intended to be included within the scope of the present invention.

Monoclonal antibodies can be bound to many different carriers and used to detect the presence of N or C polypeptide. Examples of well-known carriers include glass, polystyrene, polypropylene, polyethylene, dextran, nylon, amylases, natural and modified celluloses, polyacrylamides, agaroses and magnetite. The nature of the carrier can be either soluble or insoluble for purposes of the invention. Those skilled in the art will know of other suitable carriers for binding monoclonal antibodies, or will be able to ascertain such using routine experimentation.

For purposes of the invention, N or C polypeptide may be detected by the monoclonal antibodies when present in biological fluids and tissues. Any sample containing a detectable amount ofN or C can be used. A sample can be a liquid such as urine, saliva, cerebrospinal fluid, blood, serum and the like, or a solid or semi-solid such as tissues, feces, and the like, or, alternatively, a solid tissue such as those commonly used in histological diagnosis. C polypeptide in particular is detectable in. biological samples, since it tends to diffuse more readily than N polypeptide.

WO 98/30576 PCTIUS97/15753 -41 In performing the assays it may be desirable to include certain "blockers" in the incubation medium (usually added with the labeled soluble antibody). The "blockers" are added to assure that non-specific proteins, proteases, or anti-heterophilic immunoglobulins to anti-C or N immunoglobulins present in the experimental sample do not cross-link or destroy the antibodies on the solid phase support, or the radiolabeled indicator antibody, to yield false positive or false negative results. The selection of "blockers" therefore may add substantially to the specificity of the assays described in the present invention.

The invention also provides a method for modulating proliferation or differentiation of neuronal cells comprising contacting the cells with a hedgehog polypeptide. The hedgehog polypeptide may be a native hedgehog polypeptide, or a N or C polypeptide, or functional fragments thereof. Preferably, the modulation is induction of proliferation or differentiation of a particular cell type. This can involve either synergistic positive induction of neuronal cells by N, or negative modulation by delta N-C for example (Lai, et al., Development 121:2349, 1995). Delta N-C enhances expression of posterier relative to anterior neural genes and does so through inhibition of N (see EXAMPLE 18 and Figure 18D). In addition to hedgehog polypeptide, a TGF-P factor may also be utilized in the method of the invention.

Previous studies with the rat hedgehog gene showed that co-culture of cells expressing rat hedgehog precursor gene, with explant from neural tube, was sufficient to induce formation of motor neurons and floor plate from the explant (Jessesl, and Dodd, J., In Cell-Cell Signaling in Vertebrate Development (ed. E.J. Robertson, et al., pp 139-155, San Diego, 1993). Therefore, based on the Examples herein showing that hedgehog is expressed near the floorplate of the ventral midline of the neural tube and notochord, neuronal cells substantially derived from floor plate neuronal cells can be induced by contacting the cells with hedgehog, N or C polypeptide. As used herein, the term "substantially derived", refers to those cells from the floor plate or proximate to the floor plate. For example, such cells include motor neurons and dopaminergic neurons. Those WO 98/30576 PCT/US97/15753 -42of skill in the art will be able to identify other neuronal cells substantially derived from the floor plate. Preferably the cells are vertebrate cells and most preferably, human cells.

In addition, as described herein in the Examples, hedgehog, and particularly C fragment, induces the expression of pituitary genes. Hedgehog is also effective in inducing anterior brain gene expression as exemplified by the OTX-A marker. Further, the addition of a TGF-P family member, for example activin, may be used to further induce expression of such genes. Other TGF-P family members will be known to those of skill in the art.

This apparent synergy of hh fragments with TGF-P family members occurs through the TGF-P protein inducing expression of neural inducers such as noggin and follistatin. The hh fragment then synergizes with these inducers to pattern neural gene expression.

hh fragments may also be useful as nerve-sparing agents or in restoring or promoting appropriate patterning during the healing of major limb trauma. In addition, the N and C fragments may be useful in the area of genetic counseling. Specifically, familial midline defects such as cyclopia, polydactyly or neural tube defects may be diagnosed by mapping close to hh. Since autoproteolytic defects may be responsible for the disorders, N or C therapy could be provided.

The invention also provides an autoproteolytic fusion protein comprising a first polypeptide including the proteolytic domain of the C polypeptide of the invention, a cleavage site recognized by the first polypeptide, and a second polypeptide. (It is understood that the first and second polypeptides can be reversed.) The auto-proteolytic activity of the native hedgehog protein is found entirely within the C polypeptide, therefore, the C polypeptide is useful for producing a fusion polypeptide which can then be cleaved at the junction of the C polypeptide and the second polypeptide. The fusion protein may optionally have a purification tag, such as a poly-histidine tag for isolation on a nickel column, or an antibody epitope tag, preferably on the C fragment. The cleavage site includes the sequence "GCF", which is recognized by the proteolytic WO 98/30576 PCT/US97/15753 43 domain of the C polypeptide and is utilized to cleave the second polypeptide from the C fragment. Also included in the invention is a polynucleotide encoding the fusion protein of the invention.

The invention also provides a method for producing an autoproteolytic fusion protein comprising operably linking a first polynucleotide, wherein the first polynucleotide encodes a first polypeptide including the proteolytic domain of the C polypeptide of the invention and the cleavage site recognized by the proteolytic domain, and a second polynucleotide encoding a second polypeptide. As described above, the fusion protein may also include a carrier peptide and/or a purification tag.

The C polypeptide or functional fragment thereof is useful as a fusion partner to cause lipophilic modification and tethering of other proteins in vivo or in vitro. Such fusion proteins may be desirable for factors whose activity is required in a localized manner, either by targeting DNA constructs to specific cells or by introducing cells transfected with specific DNA constructs, for example. It may be desirable to lipid-modify a normally secreted protein in order to produce a cell-associated protein. For example, it may be desirable to produce a viral antigen that remains cell associated. Specifically, cholesterol is covalently attached to the N-terminal protein during autoprocessing and the C polypeptide acts as an intramolecular cholesterol transferase.

Alternatively, the C polypeptide or functional fragments thereof can be used as a fusion partner with a protein of interest Protein X fused to hh-C domain). Such fusions form thioesters at the junction between Protein X and hh-C (via an S to N shift). The thioesters are then available as substrates for a peptide ligation reaction in which any peptide or protein having an amino terminal cysteine (Peptide Y) is added and undergoes spontaneous rearrangement (S to N shift) that generates a stable peptide bond between Protein X and Peptide Y (Protein X-peptide bond-Peptide For example, a protein that is toxic when produced in vivo could be produced in vitro using the hh-C domain fusion protein method.

WO 98/30576 PCT/US97/15753 44- The fusion polypeptide may also include an optional carrier peptide. The "carrier peptide", or signal sequence, is located at the amino terminal end of the fusion peptide sequence. In the case of eukaryotes, the carrier peptide is believed to function to transport the fusion polypeptide across the endoplasmic reticulum. The secretory protein is then transported through the Golgi apparatus, into secretory vesicles and into the extracellular space or, preferably, the external environment. Carrier peptides which can be utilized according to the invention include pre-pro peptides which contain a proteolytic enzyme recognition site. Acceptable carrier peptides include the amino terminal pro-region of calcitonin or other hormones, which undergo cleavage at the flanking dibasic sites. However, it should be noted that the invention is not limited to the use of any particular peptide as a carrier. Other carrier peptides are known to those skilled in the art or can be readily ascertained without undue experimentation.

In one embodiment of the invention, a carrier peptide which is a signal sequence is included in the expression vector, specifically located adjacent to the N-terminal end of the fusion polypeptide. This signal sequence allows the fusion protein to be directed toward the endoplasmic reticulum. Typically, the signal sequence consists of a leader of from about 16 to about 29 amino acids, starting with two or three polar residues and continuing with a high content of hydrophobic amino acids; there is otherwise no detectable conservation of sequence known. Such signal sequences are known to those of skill in the art, and include the naturally occurring signal sequence derived from a hedgehog protein.

The fusion polypeptide of the invention includes a polypeptide encoded by a structural gene, preferably at the amino-terminus of the fusion polypeptide. Any structural gene is expressed in conjunction with the C-polypeptide (polynucleotide) and optionally a carrier peptide. The structural gene is operably linked with the carrier in an expression vector so that the fusion polypeptide is expressed as a single unit.

WO 98/30576 PCT/US97/15753 The identification of the autoproteolysis of hedgehog into the N and C domains is useful in a screening method to identify compounds or compositions which affect this processing activity. Thus, in another embodiment, the invention provides a method for identifying a composition which affects hh processing, which can be determined by S activity or gene expression, comprising incubating the components, which include the composition to be tested a drug, a small molecule, a protein) and a hh polypeptide or a recombinant cell expressing hedgehog or a gene encoding a C domain or functional fragment thereof operably linked to an N domain or functional fragment thereof, under conditions sufficient to allow the components to interact, then subsequently measuring the effect the composition has on hedgehog activity or expression. Fragments of hedgehog polypeptide or polynucleotide can be used in the method of the invention as long as autoproteolytic activity remains the construct exemplified in Figure 12a and 12b, Example 10). The observed effect on hh may be either inhibitory or stimulatory.

For example, one can determine whether the N domain is associated with the cell, or whether the N domain is secreted into the medium, in other words, whether incomplete processing has occurred. Such methods for determining the effect of the compound or composition on hh processing include those described herein (see Example 10, Figure 12a and 12b) such as time course of autoproteolytic cleavage or course of cleavage based on concentration ranges. Alternatively, the effect of the composition on hh can be determined by the expression of anterior or posterior neural markers. Other methods for determining the effect of a composition on processing of N and C will be known to those of skill in the art. Various labels can be used to detect the N and C domains, for example, a radioisotope, a fluorescent compound, a bioluminescent compound, a chemiluminescent compound, a metal chelator or an enzyme could be used. Those of ordinary skill in the art will know of other suitable labels or will be able to ascertain such, using routine experimentation.

The identification of the lipid modification of the N domain of hedgehog by the C domain, resulting in a biologically active N domain, is useful in a screening method to identify compounds or compositions which affect the cholesterol transferase/processing WO 98/30576 PCT/US97/15753 -46activity of hedgehog. In a broader aspect, the modification may be a general sterol or lipid modification, and not limited to cholesterol. Thus, in another embodiment, the invention provides a method for identifying a composition which affects hh biological activity, which can be determined by activity or lipid modification cholesterol), comprising incubating the components, which include the composition to be tested a drug, a small molecule, a protein) and a hh polypeptide or a recombinant cell expressing hedgehog or a gene encoding a C domain or functional fragment thereof operably linked to an N domain or functional fragment thereof, under conditions sufficient to allow the components to interact, then subsequently measuring the effect the composition has on hedgehog activity. Fragments of hedgehog polypeptide or polynucleotide can be used in the method of the invention as long as cholesterol transferase activity remains, for example. The effect on hh may be either inhibitory or stimulatory. For example, one can determine whether the N domain is associated with the cell, or whether the N domain is secreted into the medium, in other words, whether incomplete processing and modification has occurred. Such methods for determining the effect of the compound or composition on hh processing include those described herein (see Example 10, Figure 12a and 12b) such as time course of autoproteolytic cleavage or course of cleavage based on concentration ranges. Alternatively, the effect of the composition on hh can be determined by the level of cholesterol modification as determined by thin layer chromatography Example 19, Figure 23) or incorporation of labeled cholesterol into hh protein Example 19, Figure 25) or into a fragment appended to the transferase domain.. Other methods for determining the effect of a composition on processing and cholesterol modification of N and C will be known to those of skill in the art. Various labels can be used to detect the N and C domains, for example, a radioisotope, a fluorescent compound, a bioluminescent compound, a chemiluminescent compound, a metal chelator or an enzyme could be used. Those of ordinary skill in the art will know of other suitable labels or will be able to ascertain such, using routine experimentation.

WO 98/30576 PCT/US97/15753 -47 As used herein, "hh activity" as described in the screening method refers preferably to autoproteolytic activity. However, it is understood, that one of skill in the art could use the above-described screening assay to identify a composition having an affect on other hh activities, for example, zinc hydrolase activity or cholesterol transferase activity; or inductionor regulatoin of differentiation of neuronal cells or chondrocytes. Appropriate assays for determining the effect on such activities will be known to those of skill in the art. Example 19 provides lipophilic modification assays useful in the described screening methods above.

Now that the present invention describes the cholesterol modification of N by C, it is possible to design various diagnostic and therapeutic approaches for treatment of hh associated disorders due to defective or altered sterol modification. For example, Smith- Lemli-Optiz syndrome (SLOS) is characterized by a loss of hh function and a sterol profile indicating a cholesterol deficiency. Therefore, SLOS may be diagnosed and/or treated based on the cholesterol profile. Further, a defect in Desert hh in the testes is associated with male sterility (M.Bitgood, L.Shen, A.P. McMahon, Current Biology 6, 298, 1996; A. Vortkamp et al., Science 273, 613, 1996), consequently, it may be possible to design male contraceptives based on defective cholesterol modification of hh. On the other hand, if sterility or decreased fertility was desirable, hh cholesterol transferase activity could be altered to reduce cholesterol modification. Processing of the C and N fragments of hh is required for hh activity, therefore alterations in cholesterol modification of the amino terminal fragment may also be related to developmental defects in vertebrate embryos.

Another aspect of the present invention concerns three-dimensional molecular models of the subject hedgehog proteins, and their use as templates for the design of agents able to inhibit or potentiate at least one biological activity of the hedgehog, particularly the autoproteolytic. An integral step to our approach to designing inhibitors of the subject hegehog proteins, for example, involves construction of computer graphics models of the hegehog protein which can be used to design pharmacophores by rational drug design.

WO 98/30576 PCT/US97/15753 -48 For instance, for an inhibitor to interact optimally with the subject proteolytic domain of hedgehog, it will generally be desirable that it have a shape which is at least partly complimentary to that of a particular binding site of the enzyme, as for example those portions of the human hegehog protein which are involved in the autoproteolytic activity.

Additionally, other factors, including electrostatic interactions, hydrogen bonding, hydrophobic interactions, desolvation effects, and cooperative motions of ligand and enzyme, all influence the binding effect and should be taken into account in attempts to design bioactive inhibitors.

A computer-generated molecular model of the subject hedgehog proteins can be created.

In preferred embodiments, at least the Ca-carbon positions of the hedgehog sequence of interest are mapped to a particular coordinate pattern, such as the coordinates for hedgehog determined by x-ray crystallography, by homology modeling, and the structure of the protein and velocities of each atom are calculated at a simulation temperature (To) at which the docking simulation is to be determined. Typically, such a protocol involves primarily the prediction of side-chain conformations in the modeled protein, while assuming a main-chain trace taken from a tertiary structure such as provided in xcrystallographic model described herein. Computer programs for performing energy minimization routines are commonly used to generate molecular models. For example, both the CHARMM (Brooks et al. (1983) J Comput Chem 4:187-217) and AMBER (Weiner et al (1981) J. Comput. Chem. 106: 765) algorithms handle all of the molecular system setup, force field calculation, and analysis (see also, Eisenfield et al. (1991) Am JPhysiol 261:C376-386; Lybrand (1991) JPharm Belg 46:49-54; Froimowitz (1990) Biotechniques 8:640-644; Burbam et al. (1990) Proteins 7:99-111; Pedersen (1985) Environ Health Perspect 61:185-190; and Kini et al. (1991) JBiomol Struct Dyn 9:475- 488). At the heart of these programs is a set of subroutines that, given the position of every atom in the model, calculate the total potential energy of the system and the force on each atom. These programs may utilize a starting set of atomic coordinates, such as the model coordinates provided in crystallographic-derived models, the parameters for the various terms of the potential energy function, and a description of the molecular WO 98130576 PCT/US97/15753 -49topology (the covalent structure). Common features of such molecular modeling methods include: provisions for handling hydrogen bonds and other constraint forces; the use of periodic boundary conditions; and provisions for occasionally adjusting positions, velocities, or other parameters in order to maintain or change temperature, pressure, volume, forces of constraint, or other externally controlled conditions.

Most conventional energy minimization methods use the input data described above and the fact that the potential energy function is an explicit, differentiable function of Cartesian coordinates, to calculate the potential energy and its gradient (which gives the force on each atom) for any set of atomic positions. This information can be used to generate a new set of coordinates in an effort to reduce the total potential energy and, by repeating this process over and over, to optimize the molecular structure under a given set of external conditions. These energy minimization methods are routinely applied to molecules similar to the subject hedgehog proteins as well as nucleic acids, polymers and zeolites.

In general, energy minimization methods can be carried out for a given temperature, Ti, which maybe different than the docking simulation temperature, T o Upon energy minimization of the molecule at T i coordinates and velocities of all the atoms in the system are computed. Additionally, the normal modes of the system are calculated. It will be appreciated by those skilled in the art that each normal mode is a collective, periodic motion, with all parts of the system moving in phase with each other, and that the motion of the molecule is the superposition of all normal modes. For a given temperature, the mean square amplitude of motion in a particular mode is inversely proportional to the effective force constant for that mode, so that the motion of the molecule will often be dominated by the low frequency vibrations.

After the molecular model has been energy minimized at T i the system is "heated" or "cooled" to the simulation temperature, To, by carrying out an equilibration run where the velocities of the atoms are scaled in a step-wise manner until the desired temperature, WO 98/30576 PCT/US97/15753 To, is reached. The system is further equilibrated for a specified period of time until certain properties of the system, such as average kinetic energy, remain constant. The coordinates and velocities of each atom are then obtained from the equilibrated system.

Further energy minimization routines can also be carried out. For example, a second class of methods involves calculating approximate solutions to the constrained EOM for the protein. These methods use an iterative approach to solve for the Lagrange multipliers and, typically, only need a few iterations if the corrections required are small.

The most popular method of this type, SHAKE (Ryckaert et al. (1977) J Comput Phys 23:327; and Van Gunsteren et al. (1977) Mol Phys 34:1311) is easy to implement and scales as O(N) as the number of constraints increases. Therefore, the method is applicable to macromolecules such as the Hedgehog proteins of the present invention.

An alternative method, RATTLE (Anderson (1983) J Comput Phys 52:24) is based on the velocity version of the Verlet algorithm. Like SHAKE, RATTLE is an iterative algorithm and can be used to energy minimize the model of the subject hedgehog protein.

The increasing availability of biomacromolecule structures of potential pharmacophoric molecules that have been solved crystallographically has prompted the development of a variety of direct computational methods for molecular design, in which the steric and electronic properties of catalytic and substrate recognition sites are use to guide the design of potential inhibitors (Cohen et al. (1990) J. Med. Cam. 33: 883-894; Kuntz et al. (1982) J. Mol. Biol 161: 269-288; DesJarlais (1988) J. Med. Cam. 31: 722-729; Bartlett et al. (1989) (Spec. Publ., Roy. Soc. Chem.) 78: 182-196; Goodford et al. (1985) J. Med. Cam. 28: 849-857; DesJarlais et al. J Med. Cam. 29: 2149-2153). Directed methods generally fall into two categories: design by analogy in which 3-D structures of known molecules (such as from a crystallographic database) are docked to the enzyme structure and scored for goodness-of-fit; and de novo design, in which the ligand model is constructed piece-wise in the enzyme. The latter approach, in particular, can facilitate the development of novel molecules, uniquely designed to bind to, and, e.g., inhibit the proteolytic activity of a hegehog protein.

WO 98/30576 PCT/US97/15753 -51 In an illustrative embodiment, the design of potential hedgehog inhibitors begins from the general perspective of shape complimentary for the active site and substrate specificity subsites of the enzyme, and a search algorithm is employed which is capable of scanning a database of small molecules of known three-dimensional structure for candidates which fit geometrically into the target protein site. It is not expected that the molecules found in the shape search will necessarily be leads themselves, since no evaluation of chemical interaction necessarily be made during the initial search. Rather, it is anticipated that such candidates might act as the framework for further design, providing molecular skeletons to which appropriate atomic replacements can be made.

Of course, the chemical complimentary of these molecules can be evaluated, but it is expected that atom types will be changed to maximize the electrostatic, hydrogen bonding, and hydrophobic interactions with the enzyme. Most algorithms of this type provide a method for finding a wide assortment of chemical structures that are complementary to the shape of a binding site of the subject enzyme. Each of a set of small molecules from a particular data-base, such as the Cambridge Crystallographic Data Bank (CCDB) (Allen et al. (1973) J. Chem. Doc. 13: 119), is individually docked to the binding site of the hedgehog proteolytic domain in a number of geomietrically permissible orientations with use of a docking algorithm. In a preferred embodiment, a set of computer algorithms called DOCK, can be used to characterize the shape of invaginations and grooves that form the active sites and recognition surfaces of the subject protein (Kuntz et al. (1982) J. Mol. Biol 161: 269-288). The program can also search a database of small molecules for templates whose shapes are complementary to particular binding sites of the enzyme (DesJarlais et al. (1988) J Med Chem 31: 722- 729). These templates normally require modification to achieve good chemical and electrostatic interactions (DesJarlais et al. (1989) ACS Symp Ser 413: 60-69). However, the program has been shown to position accurately known cofactors for inhibitors based on shape constraints alone.

The orientations are evaluated for goodness-of-fit and the best are kept for further examination using molecular mechanics programs, such as AMBER or CHARMM.

WO 98/30576 PCT/US97/15753 -52 Such algorithms have previously proven successful in finding a variety of molecules that are complementary in shape to a given binding site of a receptor-enzyme, and have been shown to have several attractive features. First, such algorithms can retrieve a remarkable diversity of molecular architectures. Second, the best structures have, in previous applications to other proteins, demonstrated impressive shape complementarity over an extended surface area. Third, the overall approach appears to be quite robust with respect to small uncertainties in positioning of the candidate atoms.

Goodford (1985, J Med Chem 28:849-857) and Boobbyer et al. (1989, J Med Chem 32:1083-1094) have produced a computer program (GRID) which seeks to determine regions of high affinity for different chemical groups (termed probes) on the molecular surface of the binding site. GRID hence provides a tool for suggesting modifications to known ligands that might enhance binding. It may be anticipated that some of the sites discerned by GRID as regions of high affinity correspond to "pharmacophoric patterns" determined inferentially from a series of known ligands. As used herein, a pharmacophoric pattern is a geometric arrangement of features of the anticipated ligand that is believed to be important for binding. Attempts have been made to use pharmacophoric patterns as a search screen for novel ligands (Jakes et al. (1987) JMol Graph 5:41-48; Brint et al. (1987) JMol Graph 5:49-56; Jakes et al. (1986) JMol Graph 4:12-20); however, the constraint of steric and "chemical" fit in the putative (and possibly unknown) receptor binding site is ignored. Goodsell and Olson (1990, Proteins.

Struct Funct Genet 8:195-202) have used the Metropolis (simulated annealing) algorithm to dock a single known ligand into a target protein. They allow torsional flexibility in the ligand and use GRID interaction energy maps as rapid lookup tables for computing approximate interaction energies. Given the large number of degrees of freedom available to the ligand, the Metropolis algorithm is time-consuming and is unsuited to searching a candidate database of a few thousand small molecules.

Yet a further embodiment of the present invention utilizes a computer algorithm such as CLIX which searches such databases as CCDB for small molecules which can be WO 98/30576 PCTIUS97/15753 -53 oriented in the receptor binding site in a way that is both sterically acceptable and has a high likelihood of achieving favorable chemical interactions between the candidate molecule and the surrounding amino acid residues. The method is based on characterizing the receptor site in terms of an ensemble of favorable binding positions for different chemical groups and then searching for orientations of the candidate molecules that cause maximum spatial coincidence of individual candidate chemical groups with members of the ensemble. The current availability of computer power dictates that a computer-based search for novel ligands follows a breadth-first strategy. A breadth-first strategy aims to reduce progressively the size of the potential candidate search space by the application of increasingly stringent criteria, as opposed to a depth-first strategy wherein a maximally detailed analysis of one candidate is performed before proceeding to the next.

CLIX conforms to this strategy in that its analysis of binding is rudimentary -it seeks to satisfy the necessary conditions of steric fit and of having individual groups in "correct" places for bonding, without imposing the sufficient condition that favorable bonding interactions actually occur. A ranked "shortlist" of molecules, in their favored orientations, is produced which can then be examined on a molecule-by-molecule basis, using computer graphics and more sophisticated molecular modeling techniques. CLIX is also capable of suggesting changes to the substituent chemical groups of the candidate molecules that might enhance binding.

The algorithmic details of CLIX is described in Lawerence et al. (1992) Proteins 12:31- 41, and the CLIX algorithm can be summarized as follows. The GRID program is used to determine discrete favorable interaction positions (termed target sites) in the binding site of the protein for a wide variety of representative chemical groups. For each candidate ligand in the CCDB an exhaustive attempt is made to make coincident, in a spatial sense in the binding site of the protein, a pair of the candidate's substituent chemical groups with a pair of corresponding favorable interaction sites proposed by GRID. All possible combinations of pairs of ligand groups with pairs of GRID sites are considered during this procedure. Upon locating such coincidence, the program rotates the candidate ligand about the two pairs of groups and checks for steric hindrance and WO 98/30576 PCTIUS97/15753 -54coincidence of other candidate atomic groups with appropriate target sites. Particular candidate/orientation combinations that are good geometric fits in the binding site and show sufficient coincidence of atomic groups with GRID sites are retained.

Consistent with the breadth-first strategy, this approach involves simplifying assumptions. Rigid protein and small molecule geometry is maintained throughout. As a first approximation rigid geometry is acceptable as the energy minimized coordinates of the hedgehog deduced structure, describe an energy minimum for the molecule, albeit a local one. If the surface residues of the site of interest are not involved in crystal contacts then the crystal configuration of those residues. We believe that the deduced crystal structure described in herein should reasonably mimic the mean solution configuration.

Moreover, the equivalent models of of hedgehog isoforms (Ihh, Dhh, etc) can be derived by the same method.

A further assumption implicit in CLIX is that the potential ligand, when introduced into the active site of hegehog protein, does not induce change in the protein's stereochemistry or partial charge distribution and so alter the basis on which the GRID interaction energy maps were computed. It must also be stressed that the interaction sites predicted by GRID are used in a positional and type sense only, when a candidate atomic group is placed at a site predicted as favorable by GRID, no check is made to ensure that the bond geometry, the state of protonation, or the partial charge distribution favors a strong interaction between the protein and that group. Such detailed analysis should form part of more advanced modeling of candidates identified in the CLIX shortlist.

Yet another embodiment of a computer-assisted molecular design method for identifying inhibitors of the subject hegelh.og protein comprises the de novo synthesis of potential inhibitors by algorithmic connection of small molecular fragments that will exhibit the desired structural and electrostatic complementarity with the active site of the enzyme.

The methodology employs a large template set of small molecules with are iteratively pieced together in a model of the hiedgeho active site. Each stage of ligand growth is WO 98/30576 PCTI/US97/15753 evaluated according to a molecular mechanics-based energy function, which considers van der Waals and coulombic interactions, internal strain energy of the lengthening ligand, and desolvation of both ligand and enzyme. The search space can be managed by use of a data tree which is kept under control by pruning according to the binding criteria.

In an illustrative embodiment, the search space is limited to consider only amino acids and amino acid analogs as the molecular building blocks. Such a methodology generally employs a large template set of amino acid conformations, though need not be restricted to just the 20 natural amino acids, as it can easily be extended to include other related fragments of interest to the medicinal chemist, e.g. amino acid analogs. The putative ligands that result from this construction method are peptides and peptide-like compounds rather than the small organic molecules that are typically the goal of drug design research. The appeal of the peptide building approach is not that peptides are preferable to organics as potential pharmaceutical agents, but rather that: they can be generated .relatively rapidly de novo; their energetics can be studied by wellparameterized force field methods; they are much easier to synthesize than are most organics; and they can be used in a variety of ways, for peptidomimetic inhibitor design, protein-protein binding studies, and even as shape templates in the more commonly used 3D organic database search approach described above.

Such a de novo peptide design method has been incorporated in a software package called GROW (Moon et al. (1991) Proteins 11:314-328). In a typical design session, standard interactive graphical modeling methods are employed to define the structural environment in which GROW is to operate. For instance, environment could be the active site cleft of hedgehog, or it could be a set of features on the protein's surface to which the user wishes to bind a peptide-like molecule, a peptide sequence based on the cleavage site of hedgehog itself to represent the autoproteolytic event). The GROW program then operates to generate a set of potential ligand molecules. Interactive WO 98/30576 PCT/US97/15753 -56 modeling methods then come into play again, for examination of the resulting molecules, and for selection of one or more of them for further refinement.

To illustrate, GROW operates on an atomic coordinate file generated by the user in the interactive modeling session, such as the coordinates provided in the crysrallograph-,icderived models, plus a small fragment an acetyl group) positioned in the active site to provide a starting point for peptide growth. These are referred to as "site" atoms and "seed" atoms, respectively. A second file provided by the user contains a number of control parameters to guide the peptide growth (Moon et al. (1991) Proteins 11:314-328).

The operation of the GROW algorithm is conceptually fairly simple. GROW proceeds in an iterative fashion, to systematically attach to the seed fragment each amino acid template in a large preconstructed library of amino acid conformations. When a template has been attached, it is scored for goodness-of-fit to the receptor site, and then the next template in the library is attached to the seed. After all the templates have been tested, only the highest scoring ones are retained for the next level of growth. This procedure is repeated for the second growth level; each library template is attached in turn to each of the bonded seed/amino acid molecules that were retained from the first step, and is then scored. Again, only the best of the bonded seeddipeptide molecules that result are retained for the third level of growth. The growth of peptides can proceed in the N-to-C direction only, the reverse direction only, or in alternating directions, depending on the initial control specifications supplied by the user. Successive growth levels therefore generate peptides that are lengthened by one residue. The procedure terminates when the user-defined peptide length has been reached, at which point the user can select from the constructed peptides those to be studied further. The resulting data provided by the GROW procedure include not only residue sequences and scores, but also atomic coordinates of the peptides, related directly to the coordinate system of the receptor site atoms.

In yet another embodiment, potential pharmacophoric compounds can be determined using a method based on an energy minimization-quenched molecular dynamics WO 98/30576 PCT/US97/15753 57 algorithm for determining energetically favorable positions of functional groups in the binding sites of the subject hegehog protein. The method can aid in the design of molecules that incorporate such functional groups by modification of known ligands or de novo construction.

For example, the multiple copy simultaneous search method (MCSS) described by Miranker et al. (1991) Proteins 11: 29-34. To determine and characterize a local minima of a functional group in the forcefield of the protein, multiple copies of selected functional groups are first distributed in a binding site of interest on the hedgehog protein. Energy minimization of these copies by molecular mechanics or quenched dynamics yields the distinct local minima. The neighborhood of these minima can then be explored by a grid search or by constrained minimization. In one embodiment, the MCSS method uses the classical time dependent Hartee (TDH) approximation to simultaneously minimize or quench many identical groups in the forcefield of the protein.

Implementation of the MCSS algorithm requires a choice of functional groups and a molecular mechanics model for each of them. Groups must be simple enough to be easily characterized and manipulated (3-6 atoms, few or no dihedral degrees of freedom), yet complex enough to approximate the steric and electrostatic interactions that the functional group would have in binding to the site of interest in the hedgehog protein.

A preferred set is, for example, one in which most organic molecules can be described as a collection of such groups (Patai's Guide to the Chemistry of Functional Groups, ed.

S. Patai (New York: John Wiley, and Sons, (1989)). This includes fragments such as acetonitrile, methanol, acetate, methyl ammonium, dimethyl ether, methane, and acetaldehyde.

Determination of the local energy minima in the binding site requires that many starting positions be sampled. This can be achieved by distributing, for example, 1,000-5,000 groups at random inside a sphere centered on the binding site; only the space not WO 98/30576 PCT/US97/15753 -58occupied by the protein needs to be considered. If the interaction energy of a particular group at a certain location with the protein is more positive than a given cut-off kcal/mole) the group is discarded from that site. Given the set of starting positions, all the fragments are minimized simultaneously by use of the TDH approximation (Elber et al. (1990) JAm Chem Soc 112:9161-9175). In this method, the forces on each fragment consist of its internal forces and those due to the protein. The essential element of this method is that the interactions between the fragments are omitted and the forces on the protein are normalized to those due to a single fragment. In this way simultaneous minimization or dynamics of any number of functional groups in the field of a single protein can be performed.

Minimization is performed successively on subsets of, e.g. 100, of the randomly placed groups. After a certain number of step intervals, such as 1,000 intervals, the results can be examined to eliminate groups converging to the same minimum. This process is repeated until minimization is complete RMS gradient of 0.01 kcal/mole/A). Thus the resulting energy minimized set of molecules comprises what amounts to a set of disconnected fragments in three dimensions representing potential pharmacophores.

The next step then is to connect the pharmacophoric pieces with spacers assembled from small chemical entities (atoms, chains, or ring moieties). In a preferred embodiment, each of the disconnected can be linked in space to generate a single molecule using such computer programs as, for example, NEWLEAD (Tschinke et al. (1993) JMed Chem 36: 3863,3870). The procedure adopted by NEWLEAD executes the following sequence of commands connect two isolated moieties, retain the intermediate solutions for further processing, repeat the above steps for each of the intermediate solutions until no disconnected units are found, and output the final solutions, each of which is single molecule. Such a program can use for example, three types of spacers: library spacers, single-atom spacers, and fuse-ring spacers. The library spacers are optimized structures of small molecules such as ethylene, benzene and methylamide. The output produced by programs such as NEWLEAD consist of a set of molecules containing the WO 98/30576 PCT/US97/15753 -59 original fragments now connected by spacers. The atoms belonging to the input fragments maintain their original orientations in space. The molecules are chemically plausible because of the simple makeup of the spacers and functional groups, and energetically acceptable because of the rejection of solutions with van-der Waals radii violations.

The three-dimensional structure of hedgehog is useful to aid in screening and development of diagnostic and therapeutic protein fragments as in rational drug design, to search for structural analogs of known protein structures, or to aid in an analysis of biological function and activity. Also, the method may be used to predict protein secondary structures and protein subsecondary structures from amino acid sequences alone, and to predict those regions of a protein molecule that are on the outside and those that are on the inside.

Compounds can also be prepared using the three-dimensional structure provided herein and tested using assays known to those of skill in the art. For example, compounds can be synthesized and screened for hedgehog autoproteolytic activity by cleavage assays (see for example, Porter et al., Cell 86:21, 1996; W096/17924, herein incorporated by reference).

Compounds of the invention include drugs, small molecules, peptides, peptidomimetics, polypeptides, chemical compounds and biologic agents. For example, peptidomimetics are synthetic compounds having a three-dimensinal structure a "peptide motif') based upon the three-dimensional structure of a selected peptide. The peptide motif provides the peptidomimetic compound with Hedgehog agonist or antagonist activity that is substantially the same as, or greater than, the Hedgehog agonist or antagonist activity of the peptide from which the peptidomimetic was derived. Peptidomimetic compounds can have additional characteristics that enhance their therapeutic application, e.g., enhanced cell permeability, increased receptor or polypeptide binding affinity and/or avidity, and prolonged biological half-life. The design of peptidomimetic compounds WO 98/30576 PCT/US97/15753 having agonist or antagonist activity can be aided through computer modeling techniques well known in the art. Other methods for the design, as well as the preparation of, peptidomimemtic compounds are well known in the art.

Atomic coordinates and structure factors have been deposited in the Brookhaven Protein Data Bank. Applicant assures complete access and disclosure of these coordinates and factors upon issuance of a patent.

The following examples are intended to illustrate but not limit the invention. While they are typical of those that might be used, other procedures known to those skilled in the art may alternatively be used.

EXAMPLE 1 HEDGEHOG PROTEIN PROCESSING The full length form of the hh protein migrates with a mobility corresponding to a relative molecular mass of 46 kD. FIGURES 1 and are immunoblots with antibodies against amino- (Abl) and carboxy-terminal (Ab2) epitopes. GST fusion proteins containing either residues 83 to 160 or 300 to 391 from HH protein were expressed in Escherichia coli, purified as recommended M. Ausubel, et al., Current Protocols in Molecular Biology (Greene and Wiley-Interscience, New York, 1991)], and used to immunize rabbits by standard methods. The antibodies were affinity purified on a column of His 6 -U protein Harlow and D. Lane, Antibodies: A Laboratory Manual (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, 1988)] linked to Affi-Gel beads (Bio-Rad). The purification was performed as described (Harlow and Lane, supra) except that the acid and base elutions contained 10 percent dioxane. Biotinylated hh antibodies were prepared by purifying the rabbit antisera over a protein A column, followed by biotinylation with the use of the Immunoprobe biotinylation kit (Sigma).

Immunoprecipitations were performed as described [Harlow and Lane] with the use of cold RIPA lysis buffer containing 0.25 mM phenylmethylsulfonyl fluoride (PMSF) and WO 98/30576 PCT/US97/15753 61 mM EDTA for tissue homogenization. Lysates were precleared twice with pre-immune rabbit serum plus protein A beads (Gibco-BRL). Affinity-purified antibodies or preimmune serum was then added, and the immunoprecipitation was performed with protein A beads, with the use ofNP-40 lysis buffer for the washes.

Immunoblots were performed with affinity purified Abl or Ab2 by either of two chemiluminescence based protocols. In the first protocol (used in Figures 1, 3, and samples were resolved on 15 percent or 12 percent SDS-polyacrylamide gels M.

Ausubel et al., supra) and transferred to Magnagraph nylon membranes (MSI) by electroblotting. Blots were developed with the use of an alkaline phosphatase conjugated donkey anti-rabbit IgG secondary antibody and Lumi-Phos 530 (Boehringer Mannheim) under recommended conditions. In the second protocol (used in FIGURE samples were transferred to nitrocellulose filters (Schleicher and Schuell), and blots were developed using ECL reagents (Amersham) as recommended. The secondary antibody in this case was horseradish peroxidase conjugated goat anti-rabbit IgG (Jackson ImmunoResearch). Lanes contain protein from induced untransfected S2 cells (lanes 1 and 13), transfected S2 cells induced to express hh (lanes 2 and 14), imaginal discs (lanes 3 and 15), wild type embryos (lanes 6 and 18), and in vitro translations of synthetic h mRNA both in the presence (lanes 5 and 17) and absence of microsomes (lanes 4 and 16).

cDNAs encoding various hh protein species were cloned into the pMK33 vector,.which allows for inducible expression under metallothionein promoter control R. Koelle et al., Cell 67:59,1991). Stable S2 cell lines were made by transfection of the hh/pMK33 plasmids with constant selection for hygromycin resistance. Proteins were expressed by plating a log phase culture of cells diluted to 0.1 A5 95 units, waiting 48 hours, inducing with CuSO 4 at 0.2 mM final concentration, and harvesting the cells and/or supernatant 24 hours later. Cell samples for immunoblotting were made by adding 10 volumes of IX SDS PAGE loading buffer to pelleted cells.

WO 98/30576 PCTIUS97/15753 -62- In vitro translations were performed with the use of the TNT coupled transcription-translation system (Promega). "S methionine (DuPont NEN) was used for detection by autoradiography. In the heparin binding experiment in vitro translation lysate with microsomes that produce wild-type hh protein was added to heparin agarose (Sigma) or Sepharose CL-4B (Pharmacia) beads pre-equilibrated with heparin binding buffer (HBB; mM Tris 150 mM NaC1, 0.1 percent Triton X-100). Samples were incubated at C for four hours with gentle rocking. After pelleting the beads, supematants in some samples were analyzed (lanes 2 and The beads were then washed 5 times with chilled HBB and samples (lanes 3 and 5) were subsequently eluted at 80° C for 10 minutes in SDS PAGE loading buffer M. Ausubel et al., supra).

Embryos from the wild-type Canton-S line and from the matings, hshh/hshh or hshh H329A/hshh H329A X y; Sco/CyO, enlacZ 11.-wg (Kassis, et al., Proc. Natl. Acad. Sci.

U.S.A. 89: 1919, 1992), were collected 0 to 16 hours after egg laying (AEL) at 25° C.

They were heat shocked for 30 minutes at 37° C and allowed to recover for 1 hour at 25 o C. Embryos in FIGURE 1 (Canton-S) were collected 4 to 8 hours AEL at 25' C. In preparation for immunoblotting, all embryos were dechorionated in 2.6 percent sodium hypochlorite and homogenized in 10 volumes of IX SDS PAGE loading buffer.

Multiple species were detected and minor cross reactive bands are seen in most samples including extracts of induced untransfected S2 cells (lanes 1 and 13). One of these bands (occurring in both panels) co-migrates with U (at 39 kD) and is particularly abundant in lane 6 of FIGURE 1 FIGURES 1 and are blots of samples immunoprecipitated with Abl lanes Ab2 lanes 19-21), or pre-immune serum lanes 10-12 and D, lanes 22-24).

Detection was with biotinylated derivatives of Abl and Ab2 Samples used were: induced untransfected S2 cells, lanes 7, 10, 19 and 22; transfected S2 cells induced to express hh, lanes 8, 11, 20 and 23; and embryos, lanes 9, 12, 21 and 24. For either antibody, hh protein fragments were specifically immunoprecipitated from hh expressing WO 98/30576 PCT/US97/15753 -63 cells and embryos, but not from untransfected cells. In the schematic diagram, cleavage sites are denoted by arrows. The cleavage site marked by the asterisk is inferred by identification of only one cleavage product and may therefore occur at another location within the C fragment. The first two columns to the right of the diagram indicate the reactivity of Abl and Ab2 to each hh fragment. The other columns indicate the presence or absence of each hh fragment in the various samples. Parentheses around F and Nss indicate that these species are detected in in vitro translation reactions but not in vivo.

The 46kD species was detected from in vitro translation extracts by Abl and Ab2 (FIGURE 1, lanes 4 and 16), and was partially converted to a species of 39 kD when translation occurred in the presence of microsomes (FIGURE 1, lanes 5 and 17). A 39 kD species co-migrating with U is also present in extracts from all in vivo sources, but none of these extracts contain detectable levels of F. U represents the signal-cleaved form of F; signal cleavage thus appears to be relatively inefficient in vitro, as reported previously, J. Lee, et al., Cell, 71:33, 1992), but is highly efficient in vivo. To confirm that signal cleavage indeed is occurring at this unusual internal location, a mutation that changes residue S84 to N at the predicted signal cleavage site was introduced. This mutation prevented conversion by microsomes of F to U and also produced a species that comigrated with F upon transfection into cultured S2 cells. The effects of independently mutating the two methionine codons present upstream of the signal sequence were also examined. In vitro translation of the sequence in which the first methionine is removed produces a protein species intermediate in mobility between F and U, and this species is converted to a species that comigrates with U in the presence of microsomes or when produced in vivo. Alteration of the second methionine codon caused no change in the electrophoretic mobility of Hh protein produced in vivo or in vitro.

Smaller species of Hh proteins from in vivo sources have been reported previously (T.

Tabata and T. B. Komberg, Cell 76: 89, 1994). The latter study examined not endogenous proteins, but proteins induced to express at high levels from exogenously WO 98/30576 PCT/US97/15753 -64introduced constructs. The antibody used did not distinguish epitopes from distinct portions of the molecule.

In addition to signal cleavage, a further cleavage of the U precursor is responsible for generating other forms of hh protein observed in vivo. This was deduced from the observation that AbI and Ab2 both detected the U (uncleaved) species, but also interacted individually with smaller protein species expressed endogenously in embryos and imaginal discs or with species expressed upon introduction of the hh gene into S2 cells. Abl thus interacts with a 19kD species from all of these tissues (FIGURE 1, lanes 2, 3, 6, 8, while Ab2 interacts with a 25 kD species and a 16 kD species (FIGURE 1, lanes 14, 15, 18, 20, 21). The 19 kD species hereafter is referred to as N (N-terminal fragment), the 25 kD species as C (C-terminal fragment) and the 16 kD species as C'; these species represent the major forms of endogenous hh protein present in vivo.

The proposed cleavages by which these species arise are shown schematically in the bottom portion of FIGURE 1. The N and C species are uniquely detected by Abl and Ab2, respectively, and the sum of the relative masses of the two smaller species is roughly equivalent to the relative mass of U. The electrophoretic mobilities of the F and U species are somewhat at variance with their predicted relative masses (52.1 kD and 43.3 kD, respectively). The identities of these species were confirmed by in vitro translation of a variety of hedgehog open reading frames modified to contain different extents of sequence at the NH2- or COOH- terminus, and by insertion of epitope tags.

The migration anomalies appear to be associated with protein species in which sequences from both the NH,- and COOH-terminal fragments are simultaneously present. The mobilities of the NH,- and COOH-terminal fragments, in contrast, correspond to relative masses (19 kD and 25 kD, respectively) that sum to yield 44 kD, roughly equivalent to the expected relative mass of U.

A simple mechanism that could account for the derivation of the two smaller species therefore would be a single internal cleavage of the U precursor. Processing of the hh WO 98/30576 PCT/US97/15753 65 protein when translated in vitro also yields a 25 kD species lanes 16 and 17) and either a 29 kD or 19 kD species (lanes 4 and The 19 kD species comigrates with N, and its formation depends upon the presence of microsomes, consistent with the proposal that N derives from F by signal cleavage and a further internal cleavage. The overall pathway for formation of the predominant forms of hh protein observed in vivo thus appears to involve signal cleavage ofF to generate U. U is then cleaved internally to form N and C, which are the predominant forms found in vivo. Further processing of the 25 kD C species might then generate the 16 kD C* species, but whether this processing is a single cleavage event or not is not clear since Ab2 does not recognize the smaller 9 kD fragment that would result. The processing of C to generate C" appears to occur with greater efficiency in imaginal discs as compared to embryos (compare lanes and 18); this may be caused by the more extended mass isolation procedure of imaginal discs M. Eugene, et al., Tissue Culture Assn. Man., 5: 1055, 1979).

EXAMPLE 2 AUTO-PROTEOLYSIS OF THE HEDGEHOG PROTEIN The comigration of endogenous and in vitro-generated hh protein species suggested that in vitro processing is similar to that observed in vivo. FIGURE 2 shows limited sequence similarity between hh proteins and serine proteinases. hh protein sequences are aligned to residues 323 to 329 of the D. melanogaster protein and numbered as positions 1 to 7 (group Conserved hh residues are in bold letters. The catalytic histidines.(A. J.

Barrett, in Proteinase inhibitors A. J. Barrett, G. Salvesen, Eds. (Elsevier, Amsterdam, 1986) pp. 3-22) of mammalian serine proteinases (group B) are aligned to the invariant histidine at position 7 in Hh proteins. Abbreviations are as follows: C-Shh, chicken Sonic hh D. Riddle, et al., Cell 75: 1401, 1993); M-Shh, mouse Sonic hh Echelard et al., Cell 75: 1417, 1993) (identical to Hhg-1; R vhh-1, rat vhh-1 Roelink et al., Cell 76: 761, 1994); Z-Shh, zebrafish Sonic hh Krauss, et al., Cell 75: 1431, 1993) (identical to shh) and zebrafish vhh-1, Roelink et al., supra); twhh, no other abbreviation; M-Dhh, mouse Desert hh Echelard et al., Cell 75: 1417, 1993); M-Ihh, WO 98/30576 PCT/US97/15753 66mouse Indian hh Echelard et al., supra); CHT, bovine chymotrypsin; TRP, bovine trypsin; ELA, porcine elastase; UKH, human urokinase; C1R, human complement factor 1R; C S, human complement factor 1S; MCP, rat mast cell protease; FAX, human blood clotting factor X; TPA, human tissue plasminogen activator.

Figure 2 shows that a seven residue region of hh coding sequence (residues 323 to 329 in the Drosophila protein) displays some similarity to the sequences of serine proteases.

This region lies approximately two thirds of the distance from the signal cleavage site to the carboxy-terminus, and includes Thr and His, residues (positions 4 and 7 in FIGURE 2) that are invariant among all hh sequences from all species. In the serine proteases, this conserved sequence contains an invariant His that acts as a general base in catalysis (A.

J. Barrett, in Proteinase inhibitors A. J. Barrett, G. Salvesen, Eds. Elsevier, Amsterdam, 1986, pp. 3-22).

To determine whether this invariant His residue in the hh protein indeed plays a role in auto-proteolysis, two proteins from E. coli were purified: one carried the wild type sequence and the other a substitution of an Ala codon for the His codon at position 329 (H329A). Both of these proteins were engineered to contain a hexa-histidine tag at the amino terminus fused to Drosophila sequences extending from a residue just before the signal cleavage site to the carboxy-terminus (residues 83 to 471; the wild type form of this protein is referred to as His 6 Both proteins were extensively purified under denaturing conditions using a Ni" -chelating matrix. FIGURE 3(A) is a coomasie blue stained polyacrylamide gel that shows the production and purification of His 6 -U and His 6

-UH

3 29 A proteins from E. coli. Samples were molecular weight markers (lanes 1 and lysates ofE. coli cells carrying the His 6 -U expression construct without (lane 3) and with (lane 4) induction by IPTG; purified His 6 -U protein (lane lysates ofE. coli cells that carry the His 6

-UH

329 A expression construct without (lane 6) and with (lane 7) induction by IPTG; purified His 6

-UH

329 A protein (lane Purified proteins were essentially homogeneous except for several minor species of lower relative mass; these species are endogenous breakdown products of the full-length proteins since they were WO 98/30576 PCT/US97/15753 -67absent in uninduced extracts and were detectable with hh antibodies. FIGURE 3 is an immunoblot detected with Ab2 showing transfected S2 cells induced to express hh (lane His 6 -U and Hi% 4 329A proteins incubated in cleavage reaction buffer for 0 hours (lanes 2 and for 20 hours (lanes 3 and and for 20 hours in the presence of 20 mM TAME (a serine protease inhibitor) (lanes 4 and Upon incubation the His 6

-U,

but not the His6-UH32 9 A protein, released a fragment presumed to be C on the basis of reactivity with Ab2 and co-migration with C produced in S2 cells. Release of C (lane 3) was only partially inhibited by TAME.

Preliminary proteinase inhibitor studies have been performed on in vitro translated Hh protein by adding various inhibitors at the start of the translation reaction. These studies have been complicated by the fact that numerous protease inhibitors lower or block translation efficiency. In some cases the effectiveness of an inhibitor was assayed by determining if addition of an inhibitor to a completed translation reaction will inhibit the self-processing that normally continues to occur. At this time we can only state the following with certainty: the serine protease inhibitor TAME p-toluenesulfonyl-L-arginine methyl ester) inhibits auto-proteolysis of in-vitro translated Hh protein; (ii) soybean trypsin inhibitor, a, anti-trypsin, aprotitin, leupeptin, and E-64 do not block auto-proteolysis of translated Hh protein; and (iii) TAME partially inhibits auto-proteolysis of purified His 6 -U protein (FIGURE 3, panel B).

As seen in FIGURE 3B, upon dilution of denaturant the wild type protein but not the H329A mutant protein released a 25 kD species detectable by Ab2 and identical in mobility with the C species produced from in vitro translations and various in vivo sources. This cleavage was also observed when the wild type protein was purified and renatured by other protocols and cleaved under distinct conditions. Plasmids encoding the His 6 -U and His 6

-UH

32 9 A proteins were generated by inserting sequences corresponding to residues 83 to 471 from the wild-type or hh H329A ORF into the pRSETB expression vector (Invitrogen). Proteins were induced in BL21(DE3)/pLysS E. coli cells as described M. Ausubel et al., supra). The basic purification was performed on WO 98/30576 PCT/US97/15753 68 Ni-NTA agarose beads (Qiagen) by a denaturing protocol with the use of 6 M guanidinium HCI and 8 M urea essentially as recommended (a detailed protocol of exact conditions used is available upon request). Washes contained 0.2 percent Tween 20 and mM b-mercaptoethanol. The final wash buffer was: 6 M urea, 100 mM Tris, 500 mM NaC1, 20 percent glycerol, (pH Elutions were with the final wash buffer containing 250 mM imidazole. In vitro cleavage reactions were performed by incubating the purified protein (diluted 1:30 in the final mix) in cleavage buffer [50 mM Tris, 500 mM NaCI, percent glycerol, 0.2% Triton X-100, 50 mM DTT, (pH To isolate soluble full-length His 6 -U protein free from denaturants or detergents, additional steps were taken (this refers to the other renaturation protocols mentioned in the text). Full-length protein from the eluate described above was further purified from breakdown products by precipitation, by urea removal through dialysis. The precipitate was then re-solubilized in a buffer containing guanidinium HCI and loaded onto another Ni-NTA agarose column. After washing as described, the protein was re-folded (while attached to the beads) by gradual dilution of urea (from 6M to 0.5M) with dilution buffer [(100 mM Tris, 500 mM NaC1, 20 percent glycerol, (pH over an 8 hour period at 4° C.

The protein.was eluted with dilution buffer containing 250 mM imidazole and 0.5M urea.

The eluate was dialyzed in 100 mM Tris, 150 mM NaCI, 10 percent glycerol, (pH 7.4) at 4 °C and stored at -70 °C.

EXAMPLE 3 MAPPING THE AUTO-PROTEOLYTIC FUNCTIONS OF hh To more precisely define the domain of the hh protein responsible for this autoproteolytic event, the effects of several distinct types of mutations upon in vitro processing were examined. The most informative mutation was a deletion that removes residues 89 to 254 (A89-254), which together constitute most of the amino acids within the portion of the molecule presumed to form the N fragment. In vitro translations of wild-type and mutant Hh proteins from Drosophila (FIGURES 4 A-C) and zebrafish WO 98/30576 PCT/US97/15753 69 (FIGURE 4D) are shown. The locations of mutations and cleavage sites (arrows) in these proteins are schematically illustrated (FIGURE 4E). In the Drosophila protein (FIGURES 4A, B, and auto-proteolysis is blocked or severely inhibited by several mutations in the COOH-terminus (H329A, 294 trunc, 410 trunc, flu408 and 456 trunc), but is unaffected by a large deletion (A89-254) or insertion of a flu-tag epitope trimer (flu227) in the NH,-terminus. Auto-proteolysis thus depends primarily on residues within the C fragment (sequences to the right of the cleavage site in the diagram below; see FIGURE Furthermore, the H329A/flu227 double mutant is not cleaved by wild-type protein in a mixing experiment (lane 11), suggesting an intramolecular mechanism for auto-proteolysis. Hh proteins encoded by the zebrafish genes twhh and shh display a pattern of processing similar to that of the Drosophila protein although the NH,-terminal fragment of each zebrafish protein (23 kD for twhh and 22 kD for shh) has a lower apparent mass than the COOH-terminal fragment (25 kD for twhh and shh). This is the result of a shorter stretch of residues that precedes the signal sequences as compared to the Drosophila protein. Processing is blocked by H273A and H270A mutations in twhh and shh proteins respectively (analogous to the H329A mutation in the Drosophila protein), which suggests an auto-proteolytic processing mechanism is used similar to that observed for the Drosophila protein.

In vitro translations were performed with the use of the TNT coupled transcription-translation system (Promega). 3 S methionine (DuPont NEN) was used for detection by autoradiography. In the heparin binding experiment (FIGURE 8C), in vitro translation lysate with microsomes that produce wild-type Hh protein was added to heparin agarose (Sigma) or Sepharose CL-4B (Pharmacia) beads pre-equilibrated with heparin binding buffer (HBB; 20 mM Tris 150 mM NaC1, 0.1 percent Triton X-100). Samples were incubated at 4' C for four hours with gentle rocking. After pelleting the beads, supernatants in some samples were analyzed (lanes 2 and The beads were then washed 5 times with chilled HBB and samples (lanes 3 and 5) were subsequently eluted at 80° C for 10 minutes in SDS PAGE loading buffer M. Ausubel et al., supra).

WO 98/30576 PCT/US97/15753 70 All mutations in the hh gene were generated in the plasmid pFl J. Lee, et al., supra).

Mutations in the zebrafish twhh and shh genes were generated with the original cDNA clones as described (Ekker, et al., Current Biology, 944,1995). All point mutations were generated with the use of recombinant circle PCR H. Jones and S. C.

Winistorfer, Biotechniques 12: 528, 1992). The flu408 and flu227 mutations were generated by inserting a trimer of the influenza hemagglutinin antigen (42 residues for flu408 and 43 residues for flu227) into the AlwN I and Bgl I sites present in the hh ORF (nucleotide positions 1604 and 1058 respectively) J. Lee, et al., supra). The A89-254 mutation was generated by removing sequences between the EcoN I site (644) and the Pml I site (1145). The 294 trunc mutation was generated by removing sequences between the Acc I site (1265) and the Xcm I site (1792). The 410 trunc mutation was previously generated and identified as Hh 41 o J. Lee, et al., supra). To map the mutation in the hh 13E allele (base change 7 56 to A; coding change Ty$ 57 to STOP), DNA isolated from hh 13 '/TM3 was used to seed PCR reactions generating regions of the hh ORF and flanking sequences, which were subcloned into Bluescript KSM (Stratagene). Six clones each, derived from two different PCR amplifications were sequenced.

As seen in lanes 1 and 2 of FIGURE 4A, this construct generates a full length species of a mobility corresponding to the expected relative mass of 33 kD, and two cleaved products whose apparent relative masses (25 and 9 kD) sum to give the relative mass of the larger species. The smaller of the cleaved products will occasionally migrate as two bands as seen in Fig 4A. We have chosen the lower of the two bands between the 14.3-kD and 6.2-kD markers for our molecular weight measurement. The larger of the two cleaved products comigrates with the C species produced from the wild type protein, suggesting that the A89-254 hh protein contains the residues normally present in C and all of the determinants required for auto-proteolysis, including the normal cleavage site; most of the residues within N are dispensable for auto-proteolytic activity.

In contrast, lesions affecting residues presumed to lie within C block auto-proteolysis in vitro. All mutations tested by in vitro translation were also examined in S2 cells by WO 98/30576 PCT/US97/15753 -71 immunoblotting. In all cases the patterns of cleavage in S2 cells were identical to those observed in translations except that C* was always present whenever C was formed. The former fragment was not observed in translations. These include the H329A mutation described above, a mutation that inserts an influenza virus epitope between residues 408 and 409 (flu408), and three mutations that cause premature termination of the protein at the carboxy terminus. The two most severe truncations, 294 trunc and 410 trunc, are mutations generated in vitro. They cause a loss of 177 and 61 residues, respectively, from the carboxyl-terminus of the protein, and neither undergoes proteolysis. The 456 trunc hh protein is like that encoded by the EMS-induced hh' 3 E mutant allele, which results in the loss of 15 residues from the carboxy-terminus of the protein. This protein undergoes auto-proteolysis, as demonstrated by the appearance of a 24 kD band in place of C, but the efficiency of the reaction is much impaired in vitro (FIGURE 4B). Autoproteolysis of the hh protein relies mainly upon residues within C; deletion or alteration of residues within this domain is associated with reduced efficiency of processing, and one such deletion appears to be the cause of the hh'3E mutation.

The sequence homology and auto-proteolytic function of the full length hh protein suggested the possibility that F or the C fragment is a sequence-specific protease. As a first step in clarifying the mechanism ofauto-proteolysis, an influenza virus epitope tag was introduced into the N-terminus of a hh open reading frame that also carried a H329A mutation. FIGURE 4C shows that the insertion of the epitope tag alone does not interfere with auto-proteolysis (lane and yields a normal C fragment and an N fragment of increased relative mass (compare to wild type in lane 12). The protein carrying both mutations does not undergo proteolysis (lane 10), and since the epitope-tagged N fragment migrates differently from N, this double mutant provides an ideal substrate to look for intermolecular cleavage upon mixture with a wild type sequence. Lane 11 shows that in such a mixture, although normal N is formed, no tagged N can be detected.

Thus, in this experiment, no appreciable intermolecular cleavage occurs. We also failed to detect intermolecular cleavage in the following two experiments: co-transfection of wild type and 410 trunc sequences into S2 cells (the cleaved 410 trunc protein would WO 98/30576 PCT/US97/15753 72 yield a smaller and therefore identifiable form of (ii) mixing of excess unlabelled, purified His 6 -U protein with labelled, in vitro translated H329A mutant protein. Thus, although an intermolecular mechanism for regulation of auto-proteolysis or for cleavage of other proteins can not be ruled out, the current evidence suggests that cleavage of the hh protein occurs predominantly by an intramolecular mechanism.

The hh gene has been broadly conserved in evolution, with single homologues unidentified in a wide variety of invertebrate species and multiple distinct homologues in each of several vertebrate species Echelard et al., Cell 75: 1417, 1993; S. Krauss, et al., Cell 75: 1431, 1993; H. Roelink et al., Cell, supra). As seen in FIGURE 2, all of these coding sequences contain an invariant histidine and other conserved residues at a position corresponding to H329 in the Drosophila protein. In addition, the protein encoded by at least one of the mouse genes appears to be processed in vivo to yield two smaller species in a manner resembling the in vivo processing of the Drosophila protein.

To determine whether auto-proteolysis may also play a role in vertebrates we examined the behavior of proteins encoded by two distinct hh homologues from the zebrafish, twhh and shh. FIGURE 4D demonstrates that when these sequences are translated in vitro, smaller species are generated whose relative masses sum to yield approximately the relative mass of the full length protein (lanes I and As seen in lanes 2 and 4, this cleavage reaction is blocked by substitution of Ala codons for the His codons at positions corresponding to H329 in Drosophila (see FIGURE Vertebrate hh proteins thus appear to be processed by a similar mechanism as the Drosophila protein.

EXAMPLE 4 ROLE OF AUTO-PROTEOLYSIS IN EMBRYOS Numerous functions for the hh gene have been described in Drosophila. At the morphological level these include a role in patteming of larval cuticular structures and adult structures such as the eye and appendages Niisslein-Volhard and E. Wieschaus, WO 98/30576 PCT/US97/15753 -73 Nature 287: 795, 1980; and J. Mohler, Genetics 120: 1061, 1988).; the mechanistic basis for thee morphological effects involves signaling for maintenance or induction of gene expression in embryos and imaginal discs J. Lee, supra; T. Tabata and T. B.

Komberg, Cell 76: 89, 1994; and K. Basler and G. Struhl, Nature 368: 208, 1994). To ascertain the importance of auto-proteolysis for these functions, the H329A mutant gene under control of the hsp 70 promoter was introduced by P element-mediated transformation into the Drosophila germline. The hshh H329A construct was made identically to the hshh construct with the use of a hh ORF fragment containing the H329A mutation.

Transgenic flies were generated from ay' w 1 8 parental strain using standard methods of P element mediated transformation C. Spradling and G. M. Rubin, Science 218: 341 1982). A line, HA3, carrying the hshh H329A P element on the second chromosome was maintained as a homozygous stock. To assay for expansion of wg stripes, embryos collected at 4 to 6 hours after egg laying (AEL) at 250 C were subjected to the following heat shock protocols prior to fixation. Embryos receiving single shocks (10 or 30 minutes at 37' C) were allowed to recover for 1 hour at 250 C. Embryos receiving double shocks (two 10 minute or two 30 minute shocks at 370 C) were allowed to recover 90 minutes after the first shock and 40 minutes after the second (Both recoveries were at 250 C. The double 30 minute protocol was as previously described, Krauss, supra). In situ hybridizations were performed as described Tautz, Chromosoma 98: 81, 1989) using a wg specific probe T. Chang et al., supra). Embryos assayed for cuticle phenotype were heat shocked 6 to 8 hours AEL for 30 minutes at 370 C, allowed to develop at 250 C for 36 hours and then processed and mounted as described Ashburer, Drosophila: A Laboratory Manual, Cold Spring Harbor Laboratory Press, New York, 1989).

Immunolocalizations (single or double stains) were performed as described. With the use of affinity purified Abl or Ab2 for the primary antibody and alkaline phosphatase (AP) or horseradish peroxidase (HRP) conjugated anti rabbit or mouse IgG (Jackson Immunoresearch) for the secondary. Embryos from a hhl 3 E/TM3 ftz-lacZ (the balancer chromosome was from the Bloomington Stock Center, strain 3218) stock homozygous for the hhl 3 E allele were identified by the lack of staining with an anti b-galactosidase antibody (Promega) in a double stain with Ab2 (FIGURE 9, panel Staining in WO 98/30576 PCT/US97/15753 74 FIGURE 9, panels B and C were performed formaldehyde fixed Canton-S embryos with the use of an AP conjugated anti-rabbit IgG secondary. Although standard formaldehyde fixation was generally used, heat and acid-formaldehyde fixation also gave similar results. GST fusion proteins containing either residues 83 to 160 or 300 to 391 from the Hh protein were expressed in E. coli, purified as recommended M. Ausubel et al., supra), and used to immunize rabbits by standard methods. The antibodies were affinity purified on a column of His 6 -U protein (Harlow and Lane, supra) linked to Affi-Gel beads (Bio-Rad). The purification was performed as described (Harlow and Lane, supra) except that the acid and base elutions contained 10 percent dioxane. Biotinylated hh antibodies were prepared by purifying the rabbit antisera over a protein A column, followed by biotinylation with the use of the Immunoprobe biotinylation kit (Sigma).

Immunoprecipitations were performed as described (Harlow and Lane, supra) with the use of cold RIPA lysis buffer containing 0.25 mM PMSF and 5 mM EDTA for tissue homogenization. Lysates were precleared twice with pre-immune rabbit serum plus protein A beads (Gibco BRL). Affinity purified antibodies or pre-immune serum was then added, and the immunoprecipitation was performed with protein A beads, with the use ofNP-40 lysis buffer for the washes.

FIGURE 5 and are immunoblots developed with the use of Abl and Ab2 antibodies respectively. Lanes 1 and 6, induced untransfected S2 cells; lanes 2 and 7, transfected S2 cells induced to express hh; lanes 3 and 8, heat shocked wild-type embryos; lanes 4 and 9, heat shocked hshh embryos; lanes 5 and 10, heat shocked hshh H329A embryos. In heat shocked hshh embryos, the wild-type Hh protein is both induced and properly processed to generate the U, N C and C* species seen in other expression contexts. In contrast, the H329A is induced but not appreciably processed in hshh H329A embryos (the low levels of processed species in lanes 5 and 10 are probably from endogenous hh expression since they are seen at identical levels in heat shocked wild-type embryos in lanes 3 and 8).

WO 98/30576 PCT/US97/15753 FIGURE 5 shows that heat shock induction results in the formation of an abundant species that corresponds to U based on its mobility and its interaction with Abl and Ab2 (lanes 5 and 10). In contrast, induction of wild type hh protein using a similar contruct resulted in similar levels of the N and C processed products (lanes 4 and with very little uncleaved U. Thus, as observed in vitro and in S2 cells, the H329A mutation in embryos appears to greatly reduce the efficiency of auto-proteolytic cleavage of the hh protein.

In FIGURE 6, the embryonic distribution of wingless (wg) RNA as revealed by in situ hybridization is shown in FIGURE 6 wild-type (homozygousy' w' 1 i hshh, and hshh H329A embryos that were exposed to two 10 minute heat shocks separated by a 90-minute recovery period. Wild-type embryos showed little change in wg expression, whereas the wild-type protein and, to a lesser extent, the H329A protein each induced ectopic wg expression (Table Panels and show the dorsal surfaces ofy' hshh, and hshh H329A larvae, respectively, at the level of the fourth abdominal segment. These larvae were shocked for 30 minutes as embryos and allowed to complete embryogenesis. Cuticle cell types 30, and 40) are labeled as described (J.

Heemskerk and S. DiNardo, Cell 76: 449, 1994). Note the expansion of 20 cell types (naked cuticle) at the expense of 30 and some 40 types in the hshh embryo under conditions where the phenotype of hshh H329A embryos is identical to that of control embryos Perhaps the earliest known requirement for Hh protein is in maintenance of an adjacent stripe of wingless (wg) gene expression in each embryonic segment Martinez Arias, et al., Development 103: 157, 1988; and S. DiNardo, et al., Nature 332: 604, 1988). This requirement is deduced from the loss of wg expression when hh function is absent; in addition, the ubiquitous expression of wild-type Hh protein induces expansion of the domain of wg gene expression W. Ingham, Nature 366: 560, 1993). The effects of the H329A mutation upon wg expansion were examined by heat shocking embryos carrying the H329A mutant construct in parallel with embryos containing the wild-type construct.

WO 98/30576 PCT/US97/15753 -76 Although the H329A mutant protein is able to induce some expansion of the wg domain, the efficiency of this activity is impaired relative to that of the wild-type protein (FIGURE 6, B and C; Table The difference in efficiency ranges nearly as high as threefold depending upon the heat shock regime, and these results suggest that a uto-proteolysis of the Hh protein is important for optimal activity in embryonic signaling to induce wg expression.

TABLE 1 Wild-type and mutant hh activity in embryonic induction of wg expression* minutes of heat shock 10 30 10/10 30/30 hshh 1.0±0.3 (93) 1.5 0.6(120) 2.9 0.3 (41) 2.8 0.4 (54) hshh H329A 0.7±0.5 (190) 0.9 ±0.4(111) 1.1 ±0.4(145) 1.9±0.5(93) SExpansion of wg expression beyond wild-type controls is given as average number of cell diameters standard deviation with number of embryos scored in parentheses.

The effects of Hh protein on the patterning of cuticular structures are most clearly visible on the dorsal surface of the larva, where four distinctive cell types can be identified in each parasegment. These cell types have been designated and from anterior to posterior, with hh transcription occurring in precursors of the 1 0 cells (J.

Heemskerk and S. DiNardo, supra). Differentiation of the first three cell types was shown to be dependent upon hh gene function, and it has been proposed that the fates of these cells are determined by the concentration of Hh protein, with highest concentra- WO 98/30576 PCT/US97/15753 -77 tions producing the l1 fate, intermediate concentrations producing the 20 fate, and the lowest concentrations producing the 3 fate Heemskerk and S. DiNardo, supra). This proposal was supported by observations that the most anterior cell types display the greatest sensitivity to a reduction of hh expression, and that all of the 3 and some of the 4° bristles are replaced by naked cuticle characteristic of the more anterior 20 cell type when hh is expressed ubiquitously at high levels. We have reproduced suppression 3 and some 40 fates by heat shock induction of embryos that carry our wild-type construct (FIGURE 6E), but find that the H329A mutant is unable to alter cell fates in the dorsal cuticle of the larva (FIGURE 6F). Auto-proteolysis, or perhaps some other function blocked by the H329A mutation, thus appears to be essential for the patterning influence of Hh protein upon the dorsal cuticle.

EXAMPLE EFFECTS OF THE H329A MUTATION UPON SIGNALING IN IMAGINAL DISCS Studies of H329A mutant protein were extended to the function to the patterning of adult structures and signaling within imaginal discs. In the eye imaginal disc hh function is required for appropriate development of pattern Mohler, Genetics 120: 1061, 1988; J. J. Lee, supra; and J. Mohler and K. Vani, supra) and more recently has been shown to control progression of a wave of differentiation via induction of decapentaplegic (dpp) gene expression in the morphogenetic furrow of the eye Heberlein, et al., Cell 913, 1993; and C. Ma, et al., Cell 75: 927, 1993). In leg and wing discs, ectopic expression of hh has also been shown to yield pattern duplications and defects and is associated with induction of ectopic expression of other signaling molecules normally expressed in a zone along the anterior/posterior compartment boundary Tabata and T. B. Komberg, Cell 76: 89, 1994; and K. Basler and G. Struhl, Nature 368: 208. 1994).

WO 98/30576 PCT/US97/15753 78- For studies of signaling in imaginal discs, a thermal cycler was utilized to subject larvae carrying heat shock-inducible hh constructs to successive rounds of heat shock and recovery. The effects of temperature cycling upon expression of dpp and wg in imaginal discs was examined by monitoring B-galactosidase expression from a reporter gene carrying dpp promoter sequences or from an enhancer detector P element inserted in the wg gene. In FIGURE 7, X-gal staining was used to follow expression of wg FIGURE 7 or dpp FIGURE 7 in imaginal discs of late third-instar larvae that carry wg-lacZ or dpp-lacZ reporter genes. Leg wing and eye-antennal discs (J-L) from control larvae D, G, larvae carrying the hshh transgene E, H, K) and larvae carrying the hshh H329A transgene F, I, L) are displayed. In all panels anterior is to the left. Arrows highlight the following features: an ectopic patch of dpp expression in the anterior compartment of wing discs in hshh H329A larvae and an ectopic band of dpp expression in eye portion of the eye-antennal disc anterior to the morphogenetic furrow (marked by the other band ofdpp expression more posteriorly) in hshh larvae Expansion into the anterior compartment of wg expression in leg discs, and dpp expression in leg and wing discs in hshh larvae is similar to that described for the ectopic expression of hh. Morphological changes in the anterior compartment of leg (B and E) and wing discs were also as described Basler and G. Struhl, supra). In contrast, discs from hshh H329A and control larvae showed very little change in wg and dpp expression, even under prolonged heat shock conditions and morphological changes were never observed. The eye phenotypes of adult control hshh and hshh H329A flies that were shocked during larval development in a manner similar.to that of the imaginal disc experiments above. Duplicated eye structures were observed in hshh flies, but never in hshh H329A flies. The arrow in points to a thin strip of cuticle between the two eye structures. Other deformities were also seen in hshh flies (for example, compare the thorax in N to M) Virgin female flies from the homozygous lines hshh T. Chang et al., Development, 1994, in press), hshh H329A, and y' ,w, 18 were crossed to males from the homozygous line (bearing a P element dpp reporter construct on the 2nd chromosome, referred WO 98/30576 PCT/US97/15753 79to as dpp-lacZ) K. Blackman, el al., Development 1 1 1: 657, 1991) or the line y; Sco/ CyO, enlacZll:.wg (bearing a wg reporter P element enhancer trap on a second chromosome balancer; called wg-lacZ) A. Kassis, et al., Proc. Natl. Acad. Sci. U.S.A.

89: 1919, 1992). Progeny were grown at 25 C in aerated 0.5-ml microcentrifuge tubes containing yeast paste until the late second instar or early third instar stage of larval development. The larvae were then cycled continuously at 370 C for 30 minutes followed by 250 C for 90 minutes in a Perkin-Elmer thermal cycler until they reached the late third instar stage. They were subsequently dissected and stained with X-gal as described (M.

Ashbumer, supra) or allowed to grow to adulthood for phenotypic analysis.

As shown in FIGURE 7A, wg expression normally occurs in a ventral sector of the leg disc along the anterior/posterior compartment boundary while dpp is expressed in the dorsal portion of the disc along this boundary (FIGURE 7D). Although thermal cycling of larvae carrying the wild-type hh gene produced abnormal leg disc morphology and extensive ectopic expression of both target genes, as previously reported for ectopic hh expression (FIGURE 7B and the H329A construct produced little if any detectable difference in these patterns of expression (FIGURE 7, C and Ectopic hh expression in the wing disc also leads to morphological changes and expanded expression of dpp (compare FIGURE 7, G and but the H329A construct produced only an occasional small patch of anterior ectopic expression (FIGURE 71).

Ubiquitous expression of wild-type hh also leads to ectopic expression of dpp. in the eye-antennal disc (compare FIGURE 7, J and In the antennal portion of this disc the expansion of dpp expression resembles that observed in leg discs. In the eye portion of the disc dpp expression is observed at its normal location in the furrow; however, ectopic expression also occurs in the form of a second dorso-ventral band at a location somewhat anterior to the furrow, thus giving the appearance of an eye disc with two morphogenetic furrows (FIGURE 7K). Indeed, in adults derived from temperature-cycled larvae that carry the wild-type hh construct, an apparently duplicated eye structure such as that in FIGURE 7N can be observed, with two eye structures separated by a thin strip of cuticle WO 98/30576 PCT/US97/15753 (arrow). The H329A mutant protein, in contrast, did not induce expansion of dpp expression in either portion of the eye-antennal disc (FIGURE 7L), and does not induce eye duplications or cuticle defects in the adult (FIGURE The experiments described thus far comprise multiple series of larvae subjected to two days of thermal cycling followed by immediate dissection for analysis of imaginal structures or further incubation at constant temperature for analysis of adult structures.

Although the H329A protein appeared to have little activity in these experiments, the small patch of ectopic dpp expression induced in the wing disc (FIGURE 71, arrow) suggested that some residual activity remained. This suggestion was borne out in a similar experiment involving three days of cycling prior to dissection: the H329A protein clearly displayed some dpp-inducing activity in this experiment, presumably as a result of the higher amounts of protein that accumulated during the longer cycling period. The wing in particular, but also other imaginal discs, displayed low and variable amounts of ectopic dpp expression. This expression in all cases was far less extensive than that observed for the wild-type construct examined in parallel; furthermore, morphological deformations of the imaginal discs, although quite common with the wild-type protein, were extremely rare with the H329A protein. Although its potency is greatly reduced relative to wild-type, the H329A protein retained at least some activity in early embryonic and imaginal disc induction of wg and dpp expression; in contrast, even under heat shock conditions far more severe than those required for effects by the wild-type protein, the H329A mutant remained completely inert with respect to the re-specification of cell fates in the dorsal cuticle of the larva.

WO 98/30576 PCT/US97/15753 -81 EXAMPLE 6 DIFFERENTIAL RELEASE OF N AND C INTO CULTURED CELL SUPERNATANTS A puzzling feature of hh function is its apparent short-range action in settings such as embryonic and imaginal disc signaling to wg and dpp, and longer-range action in other settings, such as patterning of the dorsal larval cuticle. These observations and the existence of two major protein products in vivo prompted us to look for differences in the solubility or diffusibility of N and C expressed in S2 cultured cells. FIGURES 8 (A) and are immunoblots of cell pellets (lane 1) or supematants (lane 2) from transfected S2 cell cultures expressing Hh protein, developed with Abl and Ab2 Samples in each lane were from the same volume of resuspended total culture. Whereas N remained mostly associated with the cell pellet (compare lanes 1 and 2 in C was nearly quantitatively released into the supernatant (compare lanes land 2 in U displayed partitioning properties in between those of N and C (A and (8C) demonstrates the heparin binding activity of various Hh protein species generated by in vitro translations with microsomes. Samples were: total translation mix (lane 1); supematant after incubation with heparin agarose or agarose (control) beads (lanes 2 and and material eluted from heparin agarose or agarose beads after washing (lanes 3 and F, U, Nss and N fragments are depleted from reactions incubated with heparin agarose but not agarose beads (compare lanes 2 and 4 to and the same species subsequently can be eluted from the heparin agarose but not the agarose beads (compare lanes 3 and with lane FIGURES 8, A and B indeed show that these proteins behave differently, with most of the N fragment remaining cell-associated and all, or nearly all, of C being released into the culture supernatant.

One possible explanation for this differential behavior might be association of the N fragment with extracellular matrix proteins on the surfaces of the S2 cells. Accordingly, the relative affinity of these two proteins for heparin agarose was examined, since heparin binding is a common property of proteins that associate with the extracellular WO 98/30576 PCT/US97/15753 82 matrix. Given the obvious difficulty in obtaining soluble N from cultured cells, in vitro translation in the presence of microsomes was used to generate soluble. labelled N and C. As shown in FIGURE 8C, N but not C is depleted from these translation extracts by treatment with heparin agarose beads, while treatment with unmodified agarose beads did not deplete either fragment. Furthermore, N but not C was retained upon the heparin agarose beads upon extensive washing with a solution that contains 0.1% Triton X-100 and 150 mM NaCl; in contrast, neither fragment was retained by unmodified agarose. N, but not C, binds tightly to heparin, and this behavior suggests that the low concentration of N released into culture supematants may be the result of binding to the extracellular matrix. Another mechanism that might contribute to the differential release of N and C into culture supernatant would be the expression in S2 cells of a receptor for N but not for C. Our current data can not distinguish these possibilities.

EXAMPLE 7 DISTINCT EMBRYONIC LOCALIZATIONS OF N AND C The differential release of N and C into cultured cell supernatants suggested the possibility that these fragments might also be differentially localized in embryos.

Previously reported hh protein localizations utilized either antibodies specific for N epitopes or antibodies unable to distinguish between N and C. FIGURE 9 shows the differential localizations of N and C in embryos by in situ localization of the hh transcript. FIGURE 9 is shown in comparison to the distribution of N and C epitopes detected with Abl and Ab2 in panels (9B) and respectively. Note that the distribution of N and C epitopes span approximately one-third and one-half of each segmental unit respectively, while the transcript is limited to approximately one-quarter of each unit. In the localization of C epitopes in embryos homozygous for the hhl 3

E

allele is detected with the use of Ab2. C epitopes in this mutant, which displays impaired auto-proteolytic activity are more restricted, and resemble the wild-type localization of N. Homozygous hhl 3 E embryos were identified by loss of a marked balancer from a heterozygous parent stock. All embryos are at mid to late stage 9 (extended germ-band).

WO 98/30576 PCT/US97/15753 83 FIGURE 9B shows in accordance with these reports, Abl, which is specific for N epitopes, reveals a segmentally localized distribution that is slightly broader than that of the hh transcript at the same stage (FIGURE 9A). Also consistent with these reports, we observed that N epitopes at later stages accumulate in large punctate structures. Our analysis here concentrates on the earlier stage, when antibody staining is weaker but before formation of the invaginations and grooves that later crease the epidermis and thereby complicate the interpretation. Ab2 was also utilized to detect C-specific epitopes with a variety of fixation and staining procedures. Although detection of C epitopes above background is more difficult than for N, we consistently observed a segmentally modulated pattern, albeit with a broader distribution than N (FIGURE 9C). This localization is also distinctive in that C epitopes at early or late stages are not found in the punctate structures characteristic of N.

The hhl 3 E mutation encodes a prematurely truncated protein that is missing 15 residues normally present at the COOH-terminus. Because this protein displays a much reduced efficiency in auto-proteolysis the distribution of C in this mutant background was examined. FIGURE 9D shows that C epitopes in a homozygous hh 3 E embryo (identified by absence of a marked balancer) are distributed in a much tighter segmental pattern than in wild-type. This localization resembles that of N, and we thus conclude that the broad distribution of C epitopes normally seen is altered in hhl 3 E by retention of the uncleaved precursor near the site of synthesis.

EXAMPLE 8 THE ROLE OF AUTO-PROTEOLYSIS IN BIOGENESIS OF ACTIVE HEDGEHOG PROTEIN In addition to signal cleavage, the hh protein undergoes auto-proteolysis at an internal site to generate the predominant protein species observed in vivo. All or most of the amino acid residues required for this auto-proteolysis function map to C, the carboxy-terminal product of this internal cleavage. In an effort to determine the WO 98/30576 PCT/US97/15753 -84importance of auto-proteolysis for function, we introduced a single residue mutation (H329A) that blocks auto-proteolysis of the hh protein in vitro and demonstrated that both processing and function of this protein is impaired in vivo. Since similar levels of induced protein were detected from a strain carrying the wild-type construct or from several strains carrying independent insertions of the mutant construct (FIGURE the impaired function of the H329A protein relative to wild-type is not the result of reduced levels of expression. Further evidence in support of a role for auto-proteolysis derives from the effect of the hh' 1 3E mutation, which reduces but does not eliminate auto-proteolysis of the hh protein in vitro (FIGURE Correspondingly, the hh 3

E

mutation is associated with a phenotype of intermediate strength in vivo Mohler, supra).

Curiously, the H329A Hh protein appears to retain weak activity in embryonic signaling to induce ectopic wg expression and, to a lesser degree, can function in imaginal disc signaling for induction of ectopic wg and dpp expression. In contrast to its retention of at least some signaling functions in embryonic and imaginal tissues, the H329 protein is completely inert when assayed for the ability to reprogram cell fates in the dorsal cuticle ofthe larva.

The assays in which the H329A protein is active or partially active involve short-range signaling that normally occurs across one or at most several cell diameters; in contrast, the H329A protein fails to exert any effect upon patterning of the dorsal cuticle, a long-range activity that normally operates across most of the segment. Previous proposals to account for long-range patterning activities have suggested that hh expression induces other signaling molecules which are then responsible for executing the patterning functions (the signal relay model; see FIGURE 1 OA). FIGURE 10 shows a signal relay versus dual function models for hh protein action. In FIGURE 10 the long-range effects of hh signaling are achieved indirectly through short-range induction of a second signaling molecule Based on its biochemical properties and its restricted tissue localization, N is presumed to represent the active short-range signal WO 98/30576 PCT/US97/15753 while the role of C would be limited to supplying the catalytic machinery required for biogenesis of N. In (10B), the long- and short-range signaling functions of hh are supplied by the N and C proteins derived by internal auto-proteolysis of the U precursor.

N is implicated in short-range signaling by retention near its cellular site of synthesis, while C is less restricted in its distribution and would execute long-range signaling functions. In both models, auto-proteolysis is required to generate fully active signaling proteins. See text for further discussion.

These proposals seek to maintain a consistent mode of hedgehog action by rationalizing the apparent long-range activities of hh products as indirect consequences of short-range signaling. Based on the distribution observed, the active molecule in this model might be N and the role of C would then be limited to supplying the catalytic machinery required for biogenesis of N.

Our evidence suggests an alternative model, the dual function model (FIGURE 10B), in which long- and short-range activities of the hh protein might be executed by N and C, the two predominant forms of the molecule observed in vivo. The nearly quantitative release of C fragment into the culture medium of hh-expressing S2 cells and its broad, though segmentally modulated distribution within embryos suggests that C might execute or contribute to long-range signaling functions. The N fragment, on the other hand, predominantly remains associated with the expressing S2 cells and also binds to heparin, which suggests a possible association with the extracellular matrix. These properties and the segmentally restricted embryonic distribution of N are suggestive of a role in the execution of short-range hh signaling activities. Since the vertebrate Hh proteins we tested also appear to be auto-processed and also carry predicted heparin binding sites just carboxy-terminal to their signal sequences Roelink et al., supra), many aspects of the dual function model discussed here in the context of Drosophila development may also apply to hh protein function in vertebrate development.

WO 98/30576 PCT/US97/15753 86 Execution of short-range functions by N would be consistent with the observation that the H329A mutant protein has at least partial function in signaling for the induction of wg and dpp, since this mutation does not alter residues located in the amino-terminal portion of the protein that normally would give rise to N. The uncleaved H329A protein thus would carry all the residues that normally interact with a presumed receptor for N, although there might be some effect on the affinity of the interaction due to the presence of carboxy-terminal sequences, thus accounting for the decreased potency of the H329A protein. Alternatively, the partial function of H329A protein may derive from an extremely small fraction of protein that appears to be cleaved, a very faint band with identical mobility to C appears in in vitro translations with the H329A protein (FIGURE 4, lane Execution of long-range functions by C is also consistent with our observations because long-range signaling might require the release of the C fragment or otherwise require the H329 residue for some function other than for cleavage.

When N is synthesized from a native construct (wild type hh), it remains primarily cellassociated (FIGURE 10C), however, N generated from a truncated construct in cultured cells predominantly enters the culture medium (FIGURE 10D) (For constructs, see Porter, et al., Nature, 374:363, 1995). These results further confirm that autoprocessing by fragment C may regulate the degree of N association with the cell surface and therefore its range of action.

EXAMPLE 9 ISOLATION OF HEDGEHOG HOMOLOGUES The mouse and human hh-like sequences were isolated by polymerase chain reaction (PCR) using primers degenerate for all possible coding combinations of the sequences underlined in FIGURE 1 of Chang, et al., (Development, 120: 1994). PCR amplifications contained from 100 ng to 2 pg genomic DNA (depending upon the genome size of the species), 2 pM of each primer, 200 1M dNTPs (Pharmacia), IX reaction buffer (Boehringer-Mannheim) and 2.5 units Taq polymerase (Boehringer-Mannheim) in 50 pl WO 98/30576 PCTIUS97/15753 -87 reactions. Amplification was as follows: 94°C 5 min, addition ofTaq polymerase at 0 C, followed by 94°C 1 min, 52°C 1.5 min and 72°C 1 min for 30 cycles and a final extension of 72°C for 5 min. All PCR products were cloned into pBluescript (Stratagene) prior to sequence determination.

Mouse clones obtained in this manner contained 144 bases of sequence between the primer ends and were labelled with [ac 3 2 P]dATP and used for high stringency screens of mouse cDNA libraries made from whole 8.5 dpc embryonic RNA and from 14.5 dpc embryonic brain in the XZAP vector (a gift from A. Lanahan). Several clones corresponding to Hhg-l were isolated and the largest, 2629 bp in length (pDTC8.0), was chosen for sequence analysis using dideoxy chain termination (Sanger, el al., 1977) and Sequenase v2.0 (US Biochemicals). Compressions were resolved by using 7-deazaguanosine (US Biochemicals). Sequence analysis made use of the Geneworks (IntelliGenetics) and MacVector 3.5 (IBI) software packages.

One of the three mouse clones, Hhg-1, when used as a probe, yielded a 2.0 kb clone from a 8.5 dpc mouse embryonic cDNA library and a 2.7 kb clone from a 14.5 dpc embryonic cDNA library. The 2.7 kb cDNA appears to represent a nearly full length mRNA because it corresponds to a 2.7 kb band detected by hybridization on a Northern blot.

The largest methionine-initiated open reading frame within this cDNA encompasses 437 codons, and is preceded by one in frame upstream stop codon. Sequence comparisons indicate that the protein encoded by Hhg-1 is identical to the independently characterized mouse Shh (Echelard, et al., Cell, 75:1417-1430,1993) except for an arginine to lysine difference at residue 122. Hhg-1 also corresponds closely to the rat vhh-1 gene (97% amino acid identity; Roelink, et al., Cell, 76:761-775, 1994), the chicken Sonic hedgehog (81% identity; Riddle, et al., Cell, 75:1401-1416, 1993) and Shh from the zebrafish (68% identity; Krauss, et al., Cell, 75:1431-1444, 1993; Roelink, et al., Cell, 76:761-775, 1994). The PCR-generated fragments Hhg-2 and Hhg-3 appear to correspond to the Indian and Desert classes of mouse hedgehog genes, respectively (Echelard, et al., Cell, 75:1417-1430, 1993).

WO 98/30576 PCT/US97/15753 -88 Alignment of the Hhg-I open reading frame with the two Drosophila hh sequences showed that all three proteins contain hydrophobic amino acid sequences near their amino-termini; the hydrophobic stretches within the D. melanogaster protein (residues 64 to 83) and within the mouse protein are known to act efficiently as signal sequences for cleavage (Lee, et al, Cell, 71:33-50, 1992). Both Drosophila signal sequences are unusual in their internal locations, while the hydrophobic stretch of the mouse gene occurs at the extreme amino-terminus, a more conventional location for cleaved signal sequences. Although portions of sequence N-terminal to the Drosophila signal sequences are conserved, suggesting a functional role, the mouse gene lacks this region.

The overall level of amino acid identity between Hhg-I and hh carboxy-terminal to the signal sequences is 46%. A closer examination shows that the amino terminal portion, from residues 25 to 187, displays 69% identity, while remaining residues in the carboxyterminal portion display a much lower 31% identity. Like hh, the Hhg-1 coding sequence is divided into three exons, and the boundaries of these exons are at the same positions within coding sequence as those of the three Drosophila hh exons. Curiously, the boundary between coding sequences of the second and third exons occurs near the transition from high to low levels of overall sequence conservation. The coincidence of these two boundaries suggests a possible demarcation of functional domains within these proteins. This location within Hhg-l coding sequence also coincides approximately with the site of a presumed proteolytic cleavage.

EXAMPLE HUMAN CLONING OF hh GENES Partial sequence for two human hh genes has been obtained by DNA sequencing of clones derived by PCR amplification from genomic DNA with hh-specific degenerate primers as outlined in Chang, et al., (Development, 120:3339, 1994) and EXAMPLE 9 (FIGURE 11A and More extensive screening by the same approach, either with the WO 98/30576 PCTIUS97/15753 -89same primers or with other primers from the hh coding region or with the human hh fragments seen in FIGURES 11 A and B, is expected to yield at the least a third gene, and possibly more, since at least three genes are found in the mouse. These segments of human hh genes can be used to obtain full coding sequences for human proteins by the following cloning method commonly used by those of skill in the art and which are extensively described in the literature.

For example, ready-made cDNA libraries or RNAs from a variety of human sources, including various fetal stages and organs (from abortuses) and specific infant or adult organs (from pathological or autopsy specimens), are being tested for the presence of hh sequences by PCR or RT-PCR using the primers described in Chang, et al., supra, and other primers derived directly from the sequence of the human fragments. Ready-made libraries containing hh sequences are being screened directly and, where necessary, new libraries are being constructed by standard methods from RNA sources containing hh sequences. The probe for these screens is a mixture of all the distinct human hh fragments. Sequences of cDNA clones can then be determined. Most clones containing the probe sequences, which are located in the N region, will also include a full C coding region since standard methods of library construction result in cDNA clones that are most complete at their 3' ends. All full length hh-coding sequences obtained previously in vertebrates and invertebrates contain N and C sequences encoded in a single RNA.

Screening is continued until complete open reading frames that correspond to all of the fragments of human hh genes are obtained. Specifically, 1.2 x 106 clones from a human fetal brain library (Stratagene, La Jolla, CA) was screened using a mixture of the two human hh fragments (FIGURE 11A and B) as probes. Twenty-nine clones were identified as specifically hybridizing with these probes.

Second, the RNA sources identified as containing hh sequences can be used as templates from anchored PCR (also referred to in the literature as RACE, for rapid amplification of cDNA ends). Briefly, this method provides a means to isolate further mRNA WO 98/30576 PCT/US97/15753 sequence in either the 5' or 3' direction provided that sequence is known from an internal starting point. Anchored PCR can also be used to isolate sequences from cDNA library.

Third, genomic libraries can be screened with the probes described in the first technique.

Where necessary, human hh exons and coding sequences are being identified by hybridization to previously isolated human and mouse coding sequences by sequence determination, and by exon-trapping methods to identify all hh coding sequences within genomic clones; these coding sequences can be "stitched" together by standard recombinant DNA methods to generate complete hh open reading frames.

FIGURE 12 A and B show in vitro cleavage reactions of a Drosophila hh protein produced in E. coli and purified to homogeneity. This protein has residues 89-254 deleted, rendering it more soluble and easier to purify. It also contains a His 6 purification tag appended to the N-terminus. Autoproteolysis of this protein is triggered by the addition of reducing agents (DTT), and the resulting product corresponds to the C fragment identified in vivo. FIGURE 12, Panel A shows a time course of cleavage after initiation by addition of DTT. Panel B shows incubations of concentrations ranging over three order of magnitude for a fixed time period (four hours), with no difference in the extent of conversion to the cleaved form. This concentration-independent rate of cleavage indicates an intramolecular mechanism of cleavage. Panel C shows the sequence around the cleavage site as determined by amino-terminal sequence of the cleaved fragment C. The cleavage site is denoted by the arrow, and the actual residues sequenced by Edman degradation of the C fragment are underlined. Panel C also shows an alignment of all published vertebrate hh sequences plus some of unpublished sequences from fish and Xenopus. The sequences shown correspond to the region of Drosophila hh where the cleavage occurs, and demonstrates the absolute conservation of the Gly-Cys-Phe sequence at the site of cleavage. Panel D shows a SDS-PAGE gel loaded with in vitro transcription/translation reactions as described in the previous Examples, using various hh genes as templates, dhh is Drosophila, twhh and zfshh are the twiggy-winkle and sonic hh genes of the zebrafish, and mshh is the shh/Hgh-1/vhh-I WO 98/30576 PCT/US97/15753 -91 gene of the mouse. The translation mix included "S-labelled cysteine, used to visualize the resulting products by autoradiography. Note that each gene give a larger product (the precursor or U) and two smaller products of cleavage (N and The larger species is C for each of the vertebrate genes, whereas the Drosophila N is larger than C due to the presence of -60 residues occurring amino-terminal to the signal sequence that are present in the vertebrate open reading frame. This panel shows that vertebrate hh proteins are processed similarly to the Drosophila protein. Panel E shows that Edman degradation of the C fragments releases 35 S counts on the first but not subsequent rounds for all these proteins, indicating that the site of autoproteolytic cleavage for all of these hh proteins is the amide bond to the amino-terminal side of the Cys residue that forms the center of the conserved Gly-Cys-Phe sequence highlighted in panel C. This is a generalizable approach to establish the composition of protein fragments from any other hh family members.

EXAMPLE 11 DIFFERENTIAL EXPRESSION OF TWO hh GENES IN AXIAL MESODERM AND IN NEURAL PROGENITORS.

Partial sequences corresponding to five distinct zebrafish hh-like genes were isolated and the complete coding sequences for two of these genes were obtained from an embryonic cDNA library. One of these two sequences is identical to that of the zebrafish nhh-I gene (Roelink, et al., Cell, 76:761, 1994), and appears to correspond to the shh gene reported by Krauss, et al., (Cell, 75:1431, 1993) (See FIGURE 13 description); the other gene, tiggy-winkle (Potter, The Tale of Mrs. Tiggy-Winkle, The Penguin Group, London, 1905), represents a novel vertebrate hh. Coding sequences for both are shown in alignment to mouse and chicken sequences of the sonic/vhh-1 class (FIGURE 13b). Like other vertebrate hh homologues, the twhh and shh proteins contain an amino-terminal stretch of hydrophobic residues. These residues function as signal sequences since cleavage is observed when coding sequences are translated in the presence of micro- WO 98/30576 PCT/US97/15753 -92 somoses; vertebrate hh genes thus appear to encode secreted proteins, as previously reported for Drosophila hh (Kimmel C.B. Warga, Developmental Biology, 124:269-280, 1987; Warge, Kimmel, Development, 108:569-580, 1990).

The first four sequences were isolated from zebrafish genomic DNA (a gift from J.

Pellegrino) using degenerate primers in polymerase chain reactions as described (Chang, et al., supra). twhh and shh clones were isolated from a 20-28 hour cDNA library (a gift from R. Riggleman, K. Helde, D. Grunwald and J. Pellegrino) using the first three sequences as probes. The translational reading frames for twhh and shh were closed 12 and 16 codons, respectively, upstream of the putative initiating methionine.

Figure 13 shows the predicted amino acid sequences are shown in single letter code.

13(a) shows sequences common to five distinct hh-like genes are shown with a dot indicating identity with the corresponding residue of zebrafish twiggy-winkle (twhh; Potter 1905; supra), hh[zfB] and hh[zfC] is more diverged and appears to represent a novel class. 13(b) shows amino acid sequences of twhh and shh are aligned to those of the soniclvhh-1 class from chick and mouse (Riddle, et al., Cell, 75:1401-1416, 1993; Chang, et al., Development, supra; Echelard, et al., Cell, 75:1431-1444, 1993).

Zebrafish sonic hedgehog (shh) is identical in sequence to z-vhh-1 reported by Roelink, et al., Cell, 76761-775, 1994. Based on expression and extensive sequence identity throughout most of the coding region, vhh-1 and the sonic sequence reported here probably correspond to shh of Krauss, et al., Cell, 75:1431-1444, 1993, diverges dramatically throughout a 26 residue stretch near the carboxy-terminus. Rat vhh-1/sonic hh (Roelink, et al., supra.) was excluded in this alignment because of its 97% sequence identity to the predicted mouse protein. Residues identical in all four sequences are boxed, and a dash indicates a gap in the alignment. The arrow indicates the predicted signal sequence cleavage site (von Heijine, Nucleic Acids Res., 14, 4683-4690, 1986) for twhh. The amino-terminal hydrophobic stretch common to all four hh genes is shaded. 13(c) shows percent identity of residues carboxy-terminal to the hydrophobic region.

WO 98/30576 PCT/US97/15753 93 Figure 14 shows a comparative expression of twhh, shh, and pax-2 during zebrafish embryogenesis. Whole mount in situ hybridizations on 0-36 hour embryos were performed using a modification of the procedure of Tautz and Pfeifle, Chronosoma, 98:81-85, 1989, with antisense probes. Transcript localization is revealed by the purple product of an alkaline phosphatase enzymatic reaction. Staging of the embryos is according to Westerfield, (The Zebrafish Book, University of Oregon Press, Eugene, 1993). Transcripts were visualized by in situ hybridization to whole embryos. b) twhh expression in a single late shield stage embryo. Dorsal view, animal pole is to the top. The triangular shape of expression is characteristic of axial mesoderm-forming cells of the hypoblast (Statchel, et al., Development, 117:1261-1274, 1993). (b) Lateral view: the thicker layer of cells on the left (dorsal) side of the embryo is the embryonic shield; the two arrows indicate the twhh-expressing hypoblast cells and the non-expressing epiblast. Antcrior is to the left in all subsequent embryos. Dorsal is to the top in all lateral views. d) A single embryo at the end of gastrulation (100% epiboly) with twhh-expressing cells. Caudal-dorsal view. Note the wide patch of stain in the presumptive tailbud which narrows anteriorly. j) Early somitogenesis (11.5 hour, 3-4 somite) embryos; optic vesicles have not begun to evaginate from the wall of the diencephalon. h, k) Lateral views of developing brain. i, 1) Dorsal views of developing brain. f, g) Localization of twhh-expressing in a single row of cells that will form the flood plate. The arrowhead marks a parch of twhh-expressing cells lateral to the tailbud. i, j) Localization of shh. shh is also expressed strongly in the protuberance. Lateral view of the developing tail. shh is also expressed strongly in the protuberance. Lateral view of developing tail. shh is expressed in cells that will form both floor plate and notochord. 1, m) Localization of pax-2 during early optic vesicle formation; also shows twhh expression. 12 hour (4-5 somites) embryo. 12.5 hour (5-6 somites) embryo. Expression of pax-2 in the developing optic vesicle is in a gradient away from the protuberance. Note the expression of pax-2 (asterisk) at the future midbrain-hindbrain border. twhh (arrow) and pax-2 expression in a 6-7 somite (13 hour) stage embryo. Note differential expression of twhh in ventral neural keel (corresponding to neural tube in other vertebrates). Embryos WO 98/30576 PCT/US97/15753 -94 at end ofsomitogenesis (22-24 hours). o, p) Localization oftwhh. o) Developing brain. Note isolated groups of cells staining in the diencephalon (filled triangles) and the protuberance (arrowhead), and floor plate expression underlying the midbrain and hindbrain. The floor plate expression is contiguous caudally along the axis. Lateral view. Dorsal view. Lateral view of tail. Expression is restricted to the floor plate. r, s) Localization of shh. r) Developing brain Lateral view. pax-2 expression in the otic vesicle is indicated. Dorsal view. Expression in the protuberance (arrowhead) and in the neural keel. Lateral view of tail. Expression is strongest in the floor plate, but contrary to the report ofKrauss, et al., supra., is still also in the notochord. Abbreviations: white e epiblast; h hypoblast; tb tailbud; p protuberance; c eye; ov optic vesicle; ot otic vesicle; fp floor plate; nc notochord; asterisk midbrain-hindbrain boundary or pax-2-labeled prospective midbrain-hindbrain boundary; t telencephalon.

Comparison of twhh and shh expression patterns (Krauss, et al, supra), reveals that both gene are predominantly expressed in midline structures, albeit with notable differences in regard to timing, rostra-caudal extent, and tissue restriction. Expression of twhh is first detected during gastrulation in the dorsal mcsoderm (FIGURE 14a, this expression occurs in a band corresponding to a subset of the embryonic shield, a structure, analogous to Spemann's organizer in Xenopus (Stachel, et al., Dev., 117:1261- 1274, 1993, and reference therein; Ho., Seminars in Developmental Biology, pg.3, 1992). In concert with the movements of convergence and extension, this band of twhh expression shortens along the equatorial plane and extends along the incipient embryonic axis until, by the end of gastrulation, expression occurs throughout the entire axis (FIGURE 14c,d). Early in somitogenesis, twhh RNA is found restricted to presumptive ventral neural tissue along the entire body (FIGURE 14e, f, the only exception being cells in and near the tailbud (FIGURE 14g). In contrast to the neural restriction of twhh, shh is localized both to presumptive neural and notochordal cells (FIGURE 14j).

WO 98/30576 PCT/US97/15753 95 As somitogenesis proceeds, ventral midline expression of shh and twhh is reduced in most of the prospective forebrain, but remains strong in an anterior patch of midline cells within the floor of the prospective diencephalon (FIGURES 14e, f, for twhh; FIGURES h, i for shh), This patch later will give rise to the protuberance (Schmitt, E.A. and Dowling, J. Comp. Neur., 344:532-542, 1994), an anterior extension of the diencephalon. This structure, which is medial and just rostral to the developing optic stalks, is the site we propose as the focus of early patterning activity for the developing eyes (see below). By the end of somitogenesis, both twhh and shh are strongly expressed in the floor plate (FIGURES 14p, although shh transcripts remain detectable in the notochord at this stage and at 36 hours of development (FIGURES 14s; later stage not shown). At 28 hours, twhh transcripts are also found in a small cluster of cells within the first gill arch (not shown), as also reported for shh at 33 hours of development (Krauss, et al., supra).

Differences between twhh and shh expression are apparent from the beginning of gastrulation, since twhh RNA can be detected as early as the shield stage while shh is first detected later,-at about 60% epiboly (not shown; (Krauss, et al., supra). In addition, twhh transcripts are restricted to neural tissues early in development, and are never detected in the notochord (compare FIGURE 14g to FIGURE 14j). Later differences in expression include differential rostra-caudal restriction within the diencephalon and midbrain and weaker and more restricted expression of twhh in the protuberance (compare FIGURES 14n and 14q), such that the later domain of twhh expression in the brain appears to constitute a subset of the shh domain. In addition, shh but not twhh is expressed in the developing fin bud (Krauss, et al., supra). Comparison of shh and twhh expression patterns to this previously reported for hh homologues in zebrafish and other vertebrate species indicates that shh is the zebrafish homologue of the sonic/vhh-I class while twhh represents a novel class of vertebrate hh.

WO 98/30576 PCT/US97/15753 -96- EXAMPLE 12 DEVELOPMENTAL CONSEQUENCES OF ECTOPIC hh EXPRESSION DURING ZEBRAFISH EMBRYOGENESIS To gain insight into the potential roles of hh products in development, synthetic twhh and shh mRNA was injected into 1-8 cell embryos. This technique yields a mosaic but fairly uniform pattern of expression, as determined for the control mRNA encoding 3galactosidase (not shown). Uniformity of expression is in good agreement with fate mapping studies of the early zebrafish embryo (Kimmel Warga, supra; Warga Kimmel, supra; Heide, et al., Science, 265:517-520, 1994), which indicate that blastomeres undergo extensive cell mixing during the cleavages prior to gastrulation. We note that mosaicism of expression caused surprisingly little variation in the phenotypes of the hh injected embryos, possibly due to secretion of hh gene products.

Embryos injected with synthetic twhh or shh mRNA (hh RNA) exhibited numerous yet highly reproducible abnormalities in comparison to control embryos injected with lacZ mRNA. These abnormalities, discussed below, are primarily defects in the brain and eyes. Although the effects of ectopic twhh and shh expression were qualitatively similar, the incidence and severity were greater with twhh RNA (see text below, FIGURE 15 and FIGURE 16). The proteins encoded by these two genes have qualitatively similar biological activities, but apparent differences in potency.

FIGURE 15 shows the effects of ectopic hh on zebrafish development. Wild type zebrafish, Danio rerio, Ekkwill Waterlife Resources) were maintained at 28.5"C, some embryos were then cultured overnight at RT. Zebrafish embryos were injected at the 1-8 cell stage with twhh, shh, or lacZRNA and examined at 28 h of development. (a-c) Dorsal view of the midbrain-hindbrain region; anterior is left. lacZ. twhh. shh.

Frontal optical section of the forebrain region; anterior is up. lacZ. twhh.

shh. Lateral view of the eye region; anterior is left. lacZ. twhh. twhh.

At levels caudal to the prospective brain, the notochord, somites, and neural keel formed WO 98/30576 PCT/US97/15753 -97 by most hh-injected embryos appeared grossly normal except for an overall shortening and dorsal curvature of the axis. A minority of hh-injected embryos (15% are not shown) displayed partially bifurcated axes, containing duplicated axial mesoderm and parallel neural keels, each neural keel comprising ventral midline cells and some bilaterally symmetric lateral cells (not shown). Although we have not determined the primary cause of these axial defects, analysis of late gastrulation stage embryos suggests hat the bifurcation may result from difficulties in epiboly and convergence. Abbreviations: my mesencaphalic ventricle; rv rhombencephalic rentricle; asterisk midbrain-hindbrain boundary; ot otic vesicle; tv third (diencephalic) ventricle; r retina or retina-like structure; 1 lens or lens-like structure; pe pigmented retinal epithelium.

Morphological defects in the brain and other rostral neural derivatives occur at high frequency in hh-injected embryos. The three ventricles of the fish brain normally apparent at 28 hours of development the rhombencephalic, mesencephalic (FIGURE and diencephalic (third ventricle; FIGURE 15d) are not formed in the brains of hh injectees (FIGURES 15b, c; FIGURES 15e, despite the obvious presence of a lumen. The prominent construction normally present at the midbrain-hindbrain boundary also is absent (compare FIGURE 15a to FIGURES 15b, Formation of this constriction requires function of pax-2 (Krauss, et al., Nature, 353:267-270, 1991; Krauss, et al., Nature, 360:87-89, 1992), which normally is expressed in a band at the midbrain-hindbrain boundary (Krauss, et al., supra; Krauss, et al., Development, 113:1193-1206, 1991) pax2 expression at this boundary is not disrupted by hh RNA injection, however, indicating that this phenotype does not result from disruption of rostra-caudal information.

Defects in eye development also occur at high frequency in embryos injected with hh RNA. Thus, while at 28 hours the normal zebrafish eye has a lens and a retina with pigmented epithelium (FIGURE 15d, hh-injected embryos usually fail to develop lenses and retinal pigmentation (FIGURE 15e, Eye duplications are also observed at low frequencies (FIGURE 15i). The poorly developed eyes do not appear to result WO 98/30576 PCT/US97/15753 -98 from a simple delay in development since pigmentation elsewhere in injected embryos appears in its normal time course. Examined at three days of development, the consequences ofhh RNA injection include defects that range from complete absence of eyes to partially formed eyes lacing a ventral portion of the retina.

The eye phenotypes caused by hh RNA injection resemble those produced by treatment of zebrafish and Xenopus laevis embryos with retinoic acid. In Xenopus, phenotypes range from reduction of the eye and absence of the lends to eyes with retinal folds (resembling duplicated dyes) and multiple small lenses (Manns, M. Fritzsch, B., Neurosci. Lett., 127:150-154, 1991). In zebrafish, exposure to retinoic acid during gastrulation interferes with the formation of the eye (Holder, N. Hill, Development, 113:1159-1170, 1991), while exposure during formation of the optic primordia induces formation of duplicated retinas and extra lenses (Hyatt, et al., Proc. Natl. Acad. Sci. USA, 89:8293-8297, 1992). Patterning effects of retinoic acid upon the developing chick limb appear to be mediated through ectopic activation of the endogenous sonic hh gene (Riddle, et al., supra), these results with ectopic hh expression suggest the possibility of a similar mechanism underlying the patterning effects of retinoid acid treatment in the vertebrate eye.

EXAMPLE 13 hh EXPRESSION IN THE OPTIC VESICLE SPECIFIES PROXIMAL FATES AT THE EXPENSE OF DISTAL FATES To further elucidate the role of hh in eye development we utilized pax-2 and pax-6 (Krauss, et al., EMBO 10:3609-3619, 1991; Pitischel, et al., Development, 114:643- 651, 1992) were utilized as positional markers to examine the effects of ectopic hh expression on the optic vesicle. As the optic vesicle evaginates from the lateral walls of the zebrafish forebrain (Schmitt, E.A. Dowling, J. Comp. Neur., 344:532-542, 1994), pax-2 is expressed in a gradient, with highest RNA levels in the anterior and WO 98/30576 PCTIUS97/15753 99 ventral regions of the optic vesicle (Krauss, et al., supra; FIGURE 14k, 1, m).

Immediately adjacent to the maximum of this pax-2 expression gradient is the region of the dicnecphalon termed the protuberance (Schmitt Dowling, supra), where both twhh and shh but notpax-2 are strongly expressed (FIGURES 14e, f, h, i, The concentration gradient of pax-2 expression in the eptic vesicle thus appears to incline downward from its maximum at a location adjacent to the site of twhh and shh expression in the protuberance. Superposition of developmental fate within the optic vesicle (Schmitt, et al., supra), upon the pattern of pax-2 expression suggests that the gradient of pax-2 RNA prefigures the future proximal/distal axis of the eye.

Ectopic hh alters the expression ofpax-2, pax-6, and F-spondin. Zebrafish embryos were injected at the 1-8 cell stage with twhh or shh RNA and the pattern of pax-2, pax-6, or F-spondin expression was examined by whole mount in situ hybridization. Control embryos injected with lacZ RNA were performed in every case and displayed wild-type expression patterns. At embryo stage, the anterior-posterior axis of the optic vesicle corresponds to the future proximal-distal axis of the eye. During the next hour of development, the posterior edge of the optic vesicle will separate from the diencephalon (Schmitt and Dowling, Comp. Neur., 344:532-542, 1994).

Injection of either hh RNA causes uniform initiation ofpax-2 expression along both the proximal-distal and dorsal-ventral axes of the optic vesicle as it begins to evaginate. The ectopicpax-2 expression appears at the same time as normal pax-2 expression is initiated in the eye, and in some cases, is also seen in the diencephalon between the optic vesicles.

At the end of somitogenesis, a time when pax-2 would normally be restricted to the optic stalk, pax-2 RNA in hh injected embryos is detected in all but the most distal portion of the optic vesicle.

The effects of ectopic hh on expression of pax-6, which encodes a transcription factor critical for eye development was also studied. At 22 hours of zebrafish development, pax-6 is normally expressed in the lens and in most of the distal part of the optic cup WO 98/30576 PCT/US97/15753 100 (Krauss, et al., supra; Puschel, et al., Development, 114:643-651, 1992). In hh-injected embryos, pax-6 is repressed in the optic vesicle, although many embryos retain pax-6 expression in the most distal cells. With regard to pax-2 and pax-6 as markers of positional identity, hh expression in the optic vesicle can be characterized as inducing proximal fates and repressing distal fates.

The distal part of the optic vesicle is the most refractory to hh-induced changes in both pax-2 and pax-6 gene expression. Due to a later rotation, this distal portion of the optic vesicle will give rise to the dorsal portion of the mature eye (Schmitt, et al., supra); interestingly, this is the portion of the eye that remains in 3-day old injected embryos with intermediate phenotypes (see above).

Lesions in the pax-6 gene have been assigned as the basis for the Aniridia (Ton, et al., Cell, 67:1059-1074, 1991; Glaser, et al., Nat. Genetics, 2:232-239, 1992), Small eye (Hill, et al., Nature, 354:522-525, 1992), and eyeless mutations (Quiring, et al., Science 265:785-789, 1994), in humans, mice and Drosophila, respectively; pax-6 function thus appears to be critically required for eye development in Drosophila and mammals. As we argue here, hh-encoded activities also appear to play a role in vertebrate eye development, and this suggests a further molecular parallel between vertebrates and insects, since the role of hh in Drosophila eye development is well established (Mohler, et al., supra; Ma, et al., supra; Heberlein, et al., supra; Lee, et al., supra). The reciprocal and non-overlapping patterns of hh and pax-6 expression in the developing Drosophila eye (Ma, et al., supra; Quiring, et al., Science, 265:785-789, 1994), suggest the possibility of pax-6 repression by hh, but whether hh functions by similar mechanisms in vertebrate and Drosophila eye development is a questions that requires further investigation.

In mice, the dosage of pax-6 protein is crucial for normal eye development (Hill, et al., supra). Small eye heterozygotes develop an abnormally small lens (Hogan, et al., J.

Embryol. Exp. Morph., 97:95-110, 1986; Hogan, et al., Development, 103 Suppl., 115- WO 98/30576 PCT/US97/15753 101 119, 1988), as do hh-injected embryos with weaker phenotypes (FIGURE 14f). Small eye homozygotes lacking lenses eventually generate and the animals lack eyes at birth (Hogan, et al., supra; Hogan, et al., supra), as do many of the hh-injected embryos at three days of development. These parallels suggest that many of the later eye defects observed in hh-injected zebrafish may be caused by partial or complete repression of pax-6 during eye development.

EXAMPLE 14 GENETIC ABLATION OF hh FOREBRAIN EXPRESSION CAUSES LOSS OF PROXIMAL FATES IN THE OPTIC VESICLE The patterns of twhh and shh expression (FIGURE 14) and the effects of ectopic hh expression (FIGURE 15) are consistent with a normal role for shh and twhh in eye development. If hh activities indeed play a normal role in promoting proximal fates within the developing eye, removal of hh activities would be expected to result in a loss of proximal fates. In embryos homozygous for the cyclops mutation ventral neural structures fail to form and the developing eyes fuse at the midline,, yielding an embryo with a single eye (Hatta, et al., Nature, 350:339-341, 1991). The missing ventral structures in cyclops mutants include the regions where we observe expression of twhh and shh, and we therefore examined the effects of the cyclops mutation on hh expression.

cycbl 6 (Hatta, et al., Nature, 350:339-341, 1991), heterozygous adults (a kind gift of R.

Riggleman) were spawned and their offspring analyzed by whole mount in situ hybridization. Detection of pax-2 and either twhh or shh RNAs in embryos homozygous for the cyc mutation or their wild-type siblings. twhh RNA is only expressed in the presumptive tailbud (caret) of cyc embryos. As reported by Krauss, et al., Cell, supra, neural expression of shh is abolished in cyc embryos. Strong pax-2 expression was observed in the optic vesicles of wild-type embryos which is significantly reduced in cyc mutant embryos.

WO 98/30576 PCT/US97/15753 102 twhh RNA in cyclops embryos is found only in a small patch of cells at the presumptive tailbud and neural expression was not detected at any later stage examined. Neural expression of shh is also lost in eye mutants, although expression in the notochord is reunited (Krauss, et al., supra; data not shown).

Since the eye mutation appears to ablatc hh-expressing cells in the developing brain, this mutation can be used as a genetic tool to examine the requirement for hh function in eye development. liatta, et al.; Hatta, et al., Proc. Natl. Acad. Sci. USA, 91:2061-2065, 1994), recently demonstrated that pax-6 expression is fused at the midline due to loss of ventral midline cells that normally do not express pax-6 and, in addition, pax-2 expression in the fused eye of eye mutant embryos is reduced. We extended these observations to an earlier stage when the optic vesicles first form and found that pax-2 expression is weak and fails to extend within the vesicles in eye mutants. In conjunction with the results of ectopic hh expression, these observations suggest that hh signaling that activity promotes and is required for the induction of proxima fates within the eye vesicle. In this model, we propose that the protuberance acts as a proximal patterning center for the developing zebrafish eye by providing a localized source of hh activity.

EXAMPLE hh ACTIVITY VENTRALIZES THE DEVELOPING BRAIN Previous work has established an important role of signals from the floor plate and notochord in ventral patterning of the neural tube (Jessell, Dodd, Cell, 69:95- 110, 1992). For example, Goulding, et al., Development, 117:1001-1016, 1993, recently demonstrated that notochord and floor plate grafts can repress the normal lateral expression of pax-6 in the neural tube. Other recent work has implicated hh activity in at least some aspects of ventral neural tube patterning (Echelard, et al., Cell, 75:1417- 1430, 1993; Krauss, et al., supra; Roelink, et al., supra); consequently, we examined hhinjected embryos for effects on pax-6 expression in the brain.

WO 98/30576 PCT/US97/15753 103 In the zebrafish at 22 hours of development, pax-6 is expressed in dorso-lateral regions of the diencephalon and in a ventro-lateral domain of the hindbrain and spinal cord that excludes the floor plate and adjacent cells (Krauss, et al., supra; Puschel, et al., supra).

This pattern of expression is reciprocal to that of both twhh and shh in the diencephalon (compare FIGURES 14q and 14i) and in the hindbrain. hh RNA injection caused repression of pax-6 in the more ventral domain in the diencephalon, while more dorsal expression persisted. In addition, pax-6 expression was significantly reduced ventrally in rhombomeres 1, 2, and 4 and, in some cases, was completely abolished in these rhombomeres. The repressing effect of ectopically expressed hh and pax-6 in normal embryos are due to repression of pax-6 by nearby hh expressing cells.

Since absence of pax-6 expression is a feature of the ventral midline, repression of pax-6 in lateral positions suggests ventralization. Consequently, twhh was injected into embryos for analysis of induction of a floor plate marker, F-spondin (Riddle, et al., supra). As described above, ectopic twhh induces F-spondin expression at more dorsal levels in the midbrain and anterior hindbrain. The effects of hh upon expression of both pax-6 and F-spondin indicate a ventralization of the brain. Adoption of ventral cell identity by lateral cells might explain their failure to form ventricles (FIGURE The ventralizing activities of twhh confirm and extend those previously reported for shh/vhh-1 class genes of chicken, zebrafish, and rat (Echelard, et al., supra; Krauss, et al., supra; Roelink et al., supra). The early restriction of twhh to midline -neural progenitors, however, suggests that it may play a specific role in the homeogentic mechanisms of floor plate maintenance and expansion (Placzek, et al., Dev., 117:205- 218, 1993). In the zebrafish, wild type cells in cyclops hosts can contribute to and induce adjacent cells to form floor plate, but only when the transplanted cells populate the neural plate and not the notochord (Hatta, et al., Nature, 350:339-341, 1991). We have demonstrated that, in cyclops mutants, midline expression of twhh is lost while shh expression is maintained in the notochord (FIGURE 18; Krauss, et al., supra for shh); taken together, these results suggest that the homogenetic floor plate signal lost in the WO 98/30576 PCT/US97/15753 104 cyclops mutant may be encoded by the twhh gene. In the chick and rat, the floor plate retains auto-inductive potential long after the loss of floor plate inducing properties by the notochord, despite continued expression ofshh/vhhl in the notochord (Roelink, et al., supra; Placzek, et al., supra; Yamada, et al., Cell, 73:673-686, 1993). Although no homologues of the twhh class have been reported in other vertebrates, expression of other hh homologues in patterns more like those of twhh might help explain these discrepancies.

EXAMPLE 16 TWO DISTINCT SIGNALING PROTEINS DERIVE FROM THE twhh-ENCODED PRECURSOR Endogenous hh protein in Drosophila is fund predominantly as an amino- and a carboxyterminal fragment (N and C, respectively) derived by an internal auto-proteolytic cleavage of a larger precursor (U for uncleaved), which also occurs in vivo but at lower levels (Lee, et al., supra). Determinants within the amino-terminal domain appear not to be required for auto-proteolytic activity, whereas mutations affecting the carboxyterminal domain can block auto-proteolysis and reduce activity in vivo (Lee, et al., supra). The auto-proteolysis is blocked by a substitution of alanine for the histidine normally present at position 329. This histidine is absolutely invariant in alignments of all known hh genes, and its sequence context suggests a catalytic role in auto-proteolysis (Lee, et al., supra).

FIGURE 17 shows zebrafish twiggy-winkle hedgehog derivatives. 17(a) Cartoons of various twhh open reading frames. SS (shaded) is the predicted N-terminal signal sequence for secretion of these proteins and encompasses the first 27 amino acids of each open reading frame. The arrow indicates the predicted internal site of auto-proteolytic cleavage. Amino acid residue numbers are according to Figure 13b. The filled triangle denotes the normal termination codon for the twhh open reading frame. Construct UHA contains a mutation that blocks auto-proteolysis (the histidine at residue 273 is changed to an alanine; see Lee, et al., supra.). Construct U 3 56HA contains a stop codon in WO 98/30576 PCT/US97/15753 -105 place of amino acid residue 357 as well as the H273A mutation in UHA. Construct N encodes just the first 200 amino acids of twhh. Construct C has had the codons for residues 31-197 deleted. 17(b) shows in vitro translation of the expression constructs shown schematically in part a. Constructs were translated in vitro in the presence of 5

S

methionine and analyzed by autoradiography after SDS-PAGE. The protein products are shown schematically to the left. Lanes 1 and 6: Auto-proteolysis of the full-length (Uss) protein creates two fragments, an N-terminal fragment (Nss) and a C-terminal fragment Lane 2: Construct UHA only makes an uncleaved form of twhh protein that comigrates with Uss twhh via auto-cleavage. Lane 5: Construct C encodes processed and unprocessed forms which are visible as two bands migrating closely together. The bottom band is the C protein made from auto-proteolysis of the Uss (N31-197). All constructs were made by in vitro mutagenesis of expression construct T7TStwhh (see FIGURE 15) using the method of RPCR. The sequence of all constructs were confirmed by dideoxy sequencing. In vitro translations were performed according to manufacturer's instructions (Promega).

The vertebrate hh proteins encoded by shh, twhh and mouse-shh/Hhg-1 also undergo auto-proteolysis to yield two smaller species from a single larger precursor (Lee, et al., supra; Chang, et al., supra; see lanes 1 and 6 in FIGURE 17b). The invariant histidine to alanine mutation to generate a construct encoding a form of the tvvhh protein that is not auto-proteolytically cleaved We have also introduced a nonsense codon and deleted a segment of coding sequence to generate constructs that produce either the amino- or the carboxy-terminal domains of twhh (N and C, respectively; see lanes 4 and in FIGURE 17b); constructs are schematically diagrammed in FIGURE 17a). To target these proteins to the secretory pathway, all constructs retained the normal twhh signal sequence.

Synthetic mRNAs transcribed from these constructs were injected to examine the role of processing and to assay the activities of individual protein fragments; the results are WO 98/30576 PCT/US97/15753 -106 summarized in Table I and are based on the activities presented in FIGURE 15. The most striking conclusion from these experiments is that N and C both exhibit activity, and that these activities are distinguishable. Thus, although both N and C are capable of ectobpically activating pax-2 in the developing eye, thereby providing an internal injection control, only N was capable of efficiently repressing pax-6 (FIGURE 16). Later effects on lens development were also more extreme for N, consistent with the role of pax-6 in lens development suggested by its mutant phenotypes in mice. (See Ton, C.C., et al., Cell 67:1059-1074, 1991; Glaser, et al., Nat. Genetics 2:232-239, 1992; Hill, et al., Nature 354:522-525, 1991; Hogan, et al., J. Embryol. Exp. Morph., 97:95-110, 1986; and Hogan, etal., Development, 103Suppl.:115-119, 1988.) In considering the activity of delta N-C, it is important to recognize the activity of endogenous hh genes in these experiments, which are inhibited by delta N-C and fragments thereof. (see Example 18 and FIGURE 18 for further discussion) The uncleaved UHA protein is only somewhat less active than C in inducing pax-2, but it also was not able to repress pax6 efficiently (FIGURE 16). The latter is particularly notable since the UH protein (U35A see FIGURE 17a, b) has activities not significantly different from N (FIGURE 16). Thus, in addition to carrying determinants important for auto-proteolysis andpax-2induction, the C-terminus also contains a domain inhibitory to N-terminal function when in the context of the uncleaved hh protein. The C-terminus can also inhibit N action by an intermolecular mechanism (Lai, et al., supra).

The existence of such an inhibitory domain in C suggests that if autoproteolyis can be modulated, such modulation might regulate the activity of hh in vivo. This possibility highlights the importance of ascertaining the processed state of hh proteins expressed in any particular patterning center to understand the potential hh activities generated.

WO 98/30576 PCT/US97/15753 107 EXAMPLE 17 DUAL ROLES OF hh SIGNALING PROTEINS IN EARLY EYE AND BRAIN PATTERNING In understanding the normal roles of N and C in eye and brain patterning, the N and C derivatives of the Drosophila hh gene may offer some insight. The Drosophila N derivative is retained close to its embryonic site of synthesis in a segmentally striped pattern (Tabata and Kornberg, Cell, 76:89-102, 1994; Taylor, et al., Mech. Dev., 42 89- 96, 1993), is cell-associated when expressed in cultured cells, and is effectively bound by heparin agarose in vitro, suggesting the possibility of extracellular matrix association.

The C-terminal fragment, in contrast, is not bound effectively by heparin agarose, is almost quantitatively released into the culture supernatant of expressing cultured cells, and is only diffusely localized in embryos. Although the activities of individual fragments have not been assayed, the biochemical differences and tissue distributions of Drosophila N and C may account for the short and long range nature of the functions associated with hh during Drosophila development.

Although the tissue distributions of zebrafish N and C are not known, their activities in ectopic expression assays are also suggestive of short- and long-range functions when considered in the context of normal expression patterns of hh, pax-2 and pax-6. The normal gradient of pax-2 expression in the optic vesicle extends a substantial distance from its maximum adjacent to the site of hh expression in the protuberance; the ability of ectopic C to activate pax-2 therefore suggests that, consistent with the distribution of C in Drosophila, zebrafish C may carry out a long-range function. Repression of endogenous pax-6 expression, in contrast, appears to be a short-range function since pax- 6 expression occurs close to endogenous hh expression. Efficient repression ofpax-6 is an attribute of constructs producing N, and a short-range function for N would be consistent with the distribution of N in Drosophila.

WO 98/30576 PCT/US97/15753 108 Two types ofhh-dependent activity have been reported for hh-transfected cultured cells.

One is the apparent contact-dependent induction of floor plate markers (Roelink, et al., Cell 76:761-775, 1994); the second induction of sclerotome markers in presomitic mesoderm, is diffusible and acts at long-range.

EXAMPLE 18 CHARACTERIZATION OF XENOPUS hh 1. Materials and Methods cDNAs encoding full-length Xenopus hedgehogs, or encoding amino terminal or carboxy terminal domains linked to secretory leader sequences were transcribed in vitro to yield translatable messenger RNA. The synthetic messenger RNAs, and control mRNAs, were microinjected into the animal poles of cleavage stage Xenopus embryos, which were allowed to develop to the blastula stage, at which time the animal cap explants were prepared from the upper one fourth of the embryo. These blastula cap explants were then cultured in vitro in physiological saline in the presence or absence of the transforming growth factor beta family member, recombinant human activin A. All explants were allowed to develop until control embryos had grown to neurula stage, or to tadpole stage.

Importantly, blastula caps left untreated differentiate from ectoderm into atypical epidermis. Blastula caps treated with activin differentiate into mesodermal and neural cell types. Thus, the question was whether hedgehog, or its proteolytic derivatives, would change the differentiation of cells away from becoming epidermis, and into another cell type. A second question was whether hedgehog can work with activin to alter the normal response of the tissue to either factor by itself.

Explants were then extracted to yield mRNA by methods commonly used by those of skill in the art, which was used as template with reverse transcriptase to yield cDNA.

The cDNA was then used as template with various sets of primers for PCR for specific genes, reverse-transcriptase-polymerase chain reaction, or RT-PCR. This results in WO 98/30576 PCT/US97/15753 -109specific amplification of radioactive products which are diagnostic for the presence and level of the messenger RNAs which were present in the explants. Samples were separated on polyacrylamide gels, which were exposed to X-ray film to yield the bands shown in the figures. Thus, the darker bands correspond to a greater level of the specific mRNA.

FIGURE 18A and B demonstrate that hedgehog induces pituitary and anterior brain genes, and can cooperate with activin or with neural inducers such as noggin and follistatin which are induced by activin to elevate expression of these genes in explanted embryonic tissue. All odd numbered lanes lack reverse transcriptase in the RT-PCR reaction and are negative controls. All even numbered lanes have this enzyme, and thus give specific bands to mRNA. In Panel A, Lanes 1-2 are control blastula caps, lanes 3-4 are Xenopus hedgehog-expressing blastula caps, lanes 5-6 are control blastula caps treated with activin, lanes 7-8 are hedgehog-expressing blastula caps treated with activin, and 9-10 are prolactin-expressing blastula caps treated with activin to serve as a control for simply expressing a secreted protein in the blastula cap. The primers used for the assay are shown to the left of each panel, XAG 1 is a cement gland marker, XANF1B is a pituitary marker, otx-A is an anterior brain marker, en-2 is a midbrainhindbrain boundary marker, krox 20 is a rhombomere-specific hindbrain marker, HIHbox 6 is a posterior hindbrain marker, NCAM is a general neural marker, activin is a control for mesoderm induction, and elongation factor is a positive control to shown that all even numbered lanes did in fact have cDNA present.

The panel labelled XANF1B detects a pituitary gene. Lane 4 (panel A) shows that hedgehog induces this pituitary marker, and thus likely pituitary cell types, in blastula cap explants (see also FIGURE 20, lane 6, for a stronger signal showing this), when compared to control explants in the absence of hedgehog (lane which do not express this gene. Lane 6 shows that explants treated with activin, in the absence of hedgehog, also express the pituitary gene. Lane 8 shows that explants treated with both hedgehog, and with activin, give highest levels of the pituitary gene. Lane 10 proves that this effect WO 98/30576 PCT/US97/15753 -110of hedgehog is specific, since prolactin, another secreted protein, does not lead to this elevated level of pituitary gene.

The panel labelled OTX-A detects this anterior brain gene. Lane 4 (and 6 in Figure shows that hedgehog can induce this neural-specific gene. Lane 8 shows that the level of this neural gene is highest in tissue treated with both activin and hedgehog, relative to hedgehog alone (lane or activin along (lane and control explants do not express this gene (lane Again, this effect is specific to hedgehog, since prolactin (lane 10) did not lead to elevated expression of this gene. The panel labelled XAG-1 detects a cement gland-specific gene, and lane 4 shows that hedgehog induces this gene at high level.

In panel 18B, embryos were injected with N or AN-C, and some animal cap explants were treated with activin before culturing until sibling embryos reached tailbud stage.

Lanes 1, 2: control animal caps from uninjected embryos. Lanes 3, 4: control animal caps from uninjected embryos, treated with activin. Lanes 5, 6: animal caps from embryos injected with N and treated with activin. Lanes 7, 8: animal caps from embryos injected with AN-C and treated with activin. Whereas N displays activities in activin-treated explants similar to those of X-bhh (see B) AN-C produces the opposite effect, decreasing anterior and increasing posterior neural marker expression. As shown in Figure 18B, N behaves like X-bhh in that it induces elevated levels of XANF-2 and Otx-A (lane 6) relative to control activin-treated animal caps (lane Moreover, N also leads to a decrease in the expression of more posterior markers, such as krox-20 and XlHbox-6, as observed following injection of X-bhh. In contrast to the activity of N (Fig. 4C, lane 6), AN-C decreases the expression of the anterior neural genes XANF-2 or Otx-A (Fig. 4C, lane 8) in activin-treated animal caps when compared to uninjected controls (lane 4).

Moreover, AN-C also leads to an increase in the expression of more posterior markers, such as En-2 and Xlhbox-6.

FIGURE 19 shows X-bhh modifies the anteroposterior pattern of neural gene expression in explants under the influence of endogenous neural inducers. Isolation of dorsal WO 98/30576 PCT/US97/15753 111 explants from injected embryos for the preparation of Keller sandwiches (Keller and Danilchik, 1988; Doniach, et al., 1992; redrawn from Doniach, 1993). Keller sandwiches were made from uninjected (lanes 1 and 2) and X-bhh-injected (lanes 3 and 4) embroys, total RNA was isolated when control embryos reached stage 20, and RT- PCR was used to analyze the expression of XAG-1 and neural markers. XAG-1 is a cement gland marker, XANF-2 is an anterior pituitary marker, Otx-A is a forebrain marker, En-2 demarcates the midbrain-hindbrain boundary, Krox-20 marks rhombomeres 3 and 5 of the hindbrain and XlHbox-6 is a spinal cord marker. N-CAM is a general neural marker whose expression is not restricted along the anteroposterior axis. The EFla control demonstrates that a comparable amount of RNA was assayed in each set. Note that expression of XAG-1 and anterior neural markers is stimulated by X-bhh treatment, whereas expression of posterior neural markers is suppressed.

FIGURE 20 demonstration of differential activities of N and C domains of hedgehog proteins. As in FIGURE 18 above, odd numbered lanes are negative control lanes, and positive numbered lanes show specific gene expression for the markers described above.

The N domain of hedgehog is encoded in the construct called Xhhl208 (lane and the C domain is encoded in the construct called Xhhldelta 27-208 (lane 10). The construct Xhh 1l-1270A (lane 12) is specifically mutated so that it is unable to undergo selfprocessing. The ability of the N and C domains to induce the genes described above is compared to control blastula cap explants (lane entire embryos as a positive control (lane blastula cap explants expressing a mutated hedgehog as a negative control (lane 14), blastula caps expressing the entire hedgehog 1 (lane and blastula cap explants treated with an independent neural inducer, noggin (lane 16) (discovered by Richard Harland at University of California at Berkeley).

Examining the first panel for the cement gland marker XAG-1 clearly shows that intact hedgehog (lane 6) and the N domain (lane 8) and the processing defective hedgehog (lane 12) are much better than inducing the cement gland than is the C domain (lane 1).

Examining the second panel demonstrates that the C domain (lane 10) is better at WO 98/30576 PCT/US97/15753 -112inducing the pituitary gene XANF1B than is the N domain (lane Since the N domain induces the XAG-1 marker better, described in point A above, the two results together clearly demonstrate that the N and C domains have distinguishable activities.

Examination of the remaining panels shows that all described activities of the normal hedgehog (lane 6 can be defined in terms of the activities of the N and C domain.

Examining the third panel, for the forebrain gene otx-A, shows that both the N domain (lane 8) and C domain (lane 10) induce similar levels of this gene, but the processing defective hedgehog (lane 12) is better than either at inducing this gene.

Examining the fourth panel of this figure (NCAM), (as well as the FIGURE 18 panels EN-2, krox20, XIHbox6, and NCAM), shows that hedgehog does not induces these more posterior neural genes. Notably, noggin (lane 16) is able to induce pituitary gene and forebrain gene, but it also induces the general neural gene, NCAM, which hedgehog does not. This clearly shows that hedgehog is a distinct activity from the neural inducer noggin, and has a more restricted ability to induce neural genes.

Experiments in the Xenopus embryo were conducted by injecting full-length hedgehog RNA, and immunoprecipitating with a C-domain specific antibody, which proves that full length hedgehog does in fact get processed in vivo in vertebrates, consistent with the data shown in earlier Examples in Drosophila. Thus, the ideas for the utility of detecting hedgehog N and C domains is based on knowledge that such domains do appear through hedgehog processing in vertebrates. Moreover, the knowledge that hedgehog processing does occur in vivo naturally raised the question of whether the resulting N and C domains have independent activity.

The results in FIGURE 18 are novel insofar as they establish that the activity of hedgehog in inducing a pituitary gene, and an anterior brain gene, may be enhanced by the TGFP family of growth factors. This enhancement likely applies to the N and C WO 98/30576 PCTIUS97/15753 -113 domains described in FIGURE 20, since the genes analyzed are the same. This enhancement is due to hh synergizing with neural inducing factors which are themselves induced by TGF-P family members, including but not limited to such molecules as noggin and follistatin.

The data in FIGURE 20 makes several important points. First, the data show that the N and C domains have different though somewhat overlapping activities, and that the N and C activities added together account for all of the observed activity of the intact hedgehog protein. Thus, any clinical or diagnostic uses of hedgehog might be improved by use of the N or C domain, as one generally wishes to use the smallest protein which has an activity for clinical work, as it is less likely to evoke adverse immune responses, or other adverse side effects. Second, the data show that the C domain is better than the N domain in inducing pituitary gene expression and, since it has less induction of cement gland genes that intact hedgehog, or N domain, it suggests that the C domain might be useful in clinical situations where one wishes to enhance the development or expression of the pituitary as specifically as possible. As the pituitary is the source of a number of hormones, any treatment for enhancing pituitary cell growth and activity would ideally have as few side effects as possible, and the C domain is thus a viable candidate for therapies with enhanced pituitary cell growth and function in mind. Third, relating to studies regarding noggin, FIGURE 20 shows clearly that while both hedgehog and noggin can induce pituitary gene expression, hedgehog is more specific, since hedgehog does not induce the general neural marker NCAM, whereas noggin induces NCAM as well as pituitary. Fourth, the hedgehog which was mutated to prevent processing (lane 12) is as active as full-length and wild-type hedgehog (lane 6) in inducing pituitary gene expression, but the processing defective hedgehog is better at inducing the forebrain marker otx-A. Thus, for some clinical applications of hedgehog in inducing specific cell types, it is possible that the processing-defective hedgehog will be superior compared to normal hedgehog.

WO 98/30576 PCT/US97/15753 -114- FIGURE 21 shows AN-C interferes with X-bhh and N activity in animal cap explants.

Embryos were injected with various RNAs, animal cap explants were cultured until sibling embryos reached tailbud (stage 25), at which time RT-PCR was used to analyze the expression of the cement gland marker XAG-1 and the control RNA, EF-la. Lanes 1, 2: control animal caps from uninjected embryos. Lanes 3, 4: animal caps from embryos injected with both X-bhh and prolactin RNAs. Lanes 5, 6: animal caps from embryos injected with box X-bhh and AN-C. Lanes 7, 8: animal caps from embryos injected with both N and prolacting RNAs. Lanes 9, 10: animal caps from embryos injected with both N and AN-C. The N and X-bhh experiments were conducted independently and thus absolute levels in lanes 3-6 should not be compared to those in lanes 7-10. Note that the induction of XAG-1 expression by X-bhh or N is reduced by co-injection of AN-C.

An internal deletion ofX-bhh (AN-C) blocked the activity of X-bhh and N in explants and reduced dorsoanterior structures in embryos. As elevated hh activity increases the expression of anterior neural genes, and as AN-C reduces dorsoanterior structures, these complementary data support a role for hh in neural induction and anteroposterior patterning.

AN-C deletes amino acids 28-194 of X-bhh. The primary translation product is predicted to undergo signal sequence cleavage removing amino acids 1-23, and to undergo autoproteolysis. Based on the cleavage site in Drosophila hh (Porter, et al., Nature, 374:363, 1995) autoproteolysis would generate a C domain of X-bhh amino acids 198- 409, as well as a predicted seven amino acid polypeptide, representing amino acids 24- 27, and 195-197 (Lai, et al., Development 121:2349, 1995). Analysis of the effect of AN- C on neural markers was by standard methods including Northern blot analysis and in situ hybridization (Lai, et al., supra, incorporated herein by reference).

Although AN-C does not induce the cement gland marker XAG-1, it decreases the expression of anterior ectodermal and neural markers in activin-treated animal caps.

WO 98/30576 PCT/US97/15753 -115 Thus, aN-C has the capacity to affect neural patterning. AN-C also promotes an increase.

in posterior neural markers in activin-treated animal caps. Mixing AN-C with N or full length X-bhh at a 1:1 ratio led to a dramatic inhibition of the induction of cement gland in animal cap assays, supporting the hypothesis that AN-C interfered with X-hh.

EXAMPLE 19 CHOLESTEROL MODIFICATION OF HEDGEHOG POLYPEPTIDE In addition to peptide bond cleavage, Hh autoprocessing causes the covalent attachment of a lipophilic adduct to the COOH-terminus of Hh-Np (J.A.Porter et al., Cell 86, 21, 1996). This modification is critical for the spatially restricted tissue localization of the Hh signal; in its absence, the signaling domain exerts an inappropriate influence beyond its site of expression (J.A.Porter et al., Cell 86, 21, 1996). Physical and biochemical characterization of this lipophilic adduct indicates that it is not the glycosyl phosphatidyl inositol (GPI) anchor, the only other known lipophilic modification associated with secreted cell surface proteins in eukaryotes Udenfriend and K. Kodukula, A nnu.Rev.Biochem. 64, 563, 1995; and P.J.Casey, Science 268, 221, 1995).

In vitro studies of Hh autoprocessing were performed using a bacterially expressed derivative of the Drosophila protein, His 6 Hh-C, in which the majority of the NH terminal signaling domain and the signal sequence are replaced by a hexa-histidine tag.

Cleavage of this protein occurs between residues corresponding to Gly 257 and Cys 258 Porter et al., Nature 374, 363, 1995) and likely proceeds through a labile thioester intermediate formed by the cysteine thiol and the glycine carbonyl carbon. In the presence of high concentrations of thiols or other small molecules with strongly nucleophilic properties at neutral pH, cleavage of the peptide results from nucleophilic attack upon the thioester carbonyl, causing displacement of the thiol group and formation of an adduct to Gly 257 by the attacking nucleophilic (Fig 22A). Thus, in reactions with mM dithiothreitol, in vitro cleavage of His 6 Hh-C proceeded to greater than WO 98/30576 PCT/US97/15753 116completion within three hours at 30°C (Fig. 22B). At ImM dithiothreitol, however, the reaction yielded no visible cleavage product (Fig 22B).

Figure 22 shows lipid stimulation of Hh autoprocessing in vitro. Panel A illustrates the mechanism of Hh processing. The reaction is initiated by formation of a thioester between the thiol side chain of cysteine 258 and the carbonyl carbon of glycine 257, and N to S shift. This activated intermediate then undergoes a nucleophilic attack by DTT in vitro or by a piophilic nucleophilic in vivo resulting in cleavage as well as a formation of a covalent adduct at the carboxy-terminus of the amino-terminal product, X denotes the attacking nucleophilic. Panel B shows a coomassie blue stained SDS-polyacrylamide gel showing in vitro autocleavage reactions of the bacterially expressed His6Hh-C protein (~29kD) incubated for 3 hours at 30 0 C with no additions (lane 50 mMDTT (lane 1 mMDTT (lane or 1 mMDTT plus bulk S2 cell lipids (lane The Hh-C product of the autoprocessing reaction migrates as an -25kD species (lanes 2 and the NH2-terminal product is not resolved in this gel.

The in vivo reaction resulted in lipophilic modification of the NH,-terminal signaling domain. The most direct mechanism by which this could occur, by analogy to the in vitro mechanism (Fig 22A), would be for a lipid to function as the displacing nucleophilic in attack of the thioester. To explore this possibility, bulk lipids extracted from Drosophila S2 cultured cells (I.Schneider, JEmbryol Exp Morph 27, 353 (1972); and F.M. Ausubel et al., Current protocols in molecular biology (Greene Publishing Associates and Wiley- Interscience, New York, 1995) were added to the in vitro processing reaction in the presence of 1 mM dithiothreitol. Cleavage was observed and the reaction proceeded to completion in a three hour period (Fig 22B). The reaction continues beyond this time and reaches -50% completion by 18 hours.

To identify the components active in the reaction, the bulk S2 lipids were separated into two classes, neutral and complex, by silicic acid column chromatography Christie, Lipid analysis (Pergamon, Oxford, ed.2nd, 1982). Figure 23A is a thin layer chromatog- WO 98/30576 PCT/US97/15753 -117 raphy (TLC) plate coated with silica gel G (Merck) showing the fractionation of bulk S2 cell lipids using a heptane:ether:formic acid solvent (80:20:2). Six major spots are visualized by acid charring and are indicated by letters A-F. Figure 23B is a Coomassie blue-stained SDS-polyacrylamide gel showing in vitro autocleavage reactions of the bacterial expressed His 6 Hh-C protein incubated with 1 mMDTT plus either unfractionated S2 cell lipids (lane or spots A through F (lanes 2-7, respectively).

Addition of lipid spot B but no other resulted in processing of His 6 Hh-C protein. Figure 23C is TLC of S2 cell lipids (lane 1) along with selected lipid standards: phosphatidylcholine (lane a diacylglycerol (lane cholesterol (lane stearic acid (lane a triacylglycerol (lane and cholesteryl ester (lane Lipid spot B comigrates with cholesterol, as also demonstrated by mixing radio-labeled cholesterol with S2 lipids before TLC fractionation. Figure 23D is a Coomassie blue stained SDS-polyacrylamide gel showing that relative to 1 mMDTT alone (lane 1) cholesterol (0.35 mM) 1 mMDTT (lane 2) stimulates His,Hh-C autocleavage in vitro. Figure 23E is an autoradiogram of electrophoretically-resolved products of His 6 Hh-C autocleavage reactions driven by mMDTT (lane 1) or 1 mMDTT+0.35 mM cholesterol (lane For lane 1 3 H]cholesterol (3/zCi) was added at the end of the incubation period just prior to electrophoresis; for lane 2[ 3 H]cholesterol was present throughout the incubation period and is incorporated into the amino-terminal product of the reaction. To resolve the 5 kD product of His6Hh- C autocleavage, reaction products were separated in 17% SDS-polyacrylamide gels.

The activity was found exclusively in the neutral class, so the lipids were subjected to preparative thin layer chromatography (TLC) using a solvent system that resolved neutral lipids Christie, Lipid analysis (Pergamon, Oxford, ed.2nd, 1982) (Fig 23A). Lipid spots were visualized with iodine vapor or acid charring, and adsorbent at the corresponding positions of identical uncharred plates was excised and extracted with chloroform/methanol/water. Only lipids extracted from spot B displayed stimulatory activity in the in vitro cleavage reaction (Fig 23B).

WO 98/30576 PCT/US97/15753 -118 With the use of various lipid standards, it was found that spot B comigrated with cholesterol (Fig 23C). In addition, the active S2 cell-derived lipid displayed the same mobility as cholesterol in two other solvent systems and gave a positive color test when sprayed with a specific reagent that reacts with sterols;W.W. Christie, Lipid analysis (Pergamon, Oxford, ed.2nd, 1982); and R.R.Lowry, Journal of Lipid Research 9, 397, 1968). Taken together these results imply that the active lipid component is in the sterol fraction of the S2 lipids. Indeed, it was found that cholesterol, which is the principal sterol in eukaryotic cell membranes Christie, Lipid analysis (Pergamon, Oxford, ed.2nd, 1982)), displayed stimulatory activity similar to that observed with lipids extracted from spot B when added in pure form to the in vitro processing reaction (Fig 23D). To establish that the stimulatory activity of cholesterol is a result of its participation as a modifying group, it was shown that 3 H-labeled cholesterol added to the 1 mM dithiothreitol reaction was incorporated into the NH,-terminal product (Fig 23E). No incorporation was seen, however, when 3 H]cholesterol was added just prior to electrophoresis to a reaction incubated for 3 hours with 20 mM dithiothreitol (Fig 23E).

Also consistent with covalent cholesterol addition, the NH 2 -terminal fragment of HisHh- C generated by the cholesterol-driven reaction migrated just beneath the 6 kD marker, whereas the product of the reaction driven by 20 mM dithiothreitol migrated just above this marker (Fig 24A). Such a shift in mobility, thought to result from an increase capacity for SDS binding to the covalently linked lipid (M.L.Cardoso de Almeida and M.J. Turner, Nature 302, 349, 1983),was also noted for Hh-Np as compared to the precisely truncated NH2-terminal fragment (Hh-N, truncated following Gly 257).

The part of the sterol most likely to act as attacking is the 3P hydroxyl. Such an attack would leave cholesterol as a covalent adduct in ester linkage to the carboxylate of the terminal residue of the NH2-terminal fragment (GLY 257).

Figure 24A shows Coomassie stained gels of His 6 Hh-C autocleavage reactions carried out in the presence of 20 mMDTT (lane or 1 mMDTT+0.35 mM cholesterol (lane 2).

Lane 3 contains a mixture of the samples loaded in lanes 1 and 2. The amino-terminal WO 98/30576 PCT/US97/15753 119product of the cholesterol driven reaction migrates approximately 2 kD faster than the DTT-driven reaction fragment. Figure 24B is Coomassie stained gels showing protein products of His 6 Hh-C autocleavage reactions carried out in the presence of 1 mM DTT+0.35 mM cholesterol (lanes 1 and 2) or with 20 mM DTT (lane Prior to loading the gel, samples in lane 2 and 3 were incubated for 60 minutes with 50 mMKOH in methanol Field and A.K. Menon, in Lipid modification.ofproteins N.M. Hooper, A.J. Turner, Eds. (Oxford University Press, New York 1992) pp.

1 55). Base treatment causes the cholesterol-driven amino-terminal reaction product to comigrate with the corresponding DTT-driven reaction product. Figure 24C is an autoradiogram of immunoblotted Hh amino-terminal domains purified from cultured S2 cells. Aminoterminal domains were derived either from a construct truncated after glycine 257 (Hh-N lanes 1,3,4,8, and 9) or from a construct encoding wild-type Hh that produces the aminoterminal domain via the processing reaction (Hh-Np, lane 2,3,5,6,7,8, and Proteins were either directly loaded (lanes 1 and 2) or base-treated Field and A.K. Menon, in Lipid modification of proteins N.M. Hooper, A.J. Turner, Eds. (Oxford University Press, New York 1992) pp.155) for 5 minutes (lane 20 minutes (lane 6) or 1 hour (lanes 7 and 4) prior to electrophoresis. Lane 3 contains a mixture of the samples loaded in lanes 1 and 2, lane 8 contains a mixture of the samples loaded in lanes 7 and 4, and lane 9 contains a mixture of the samples loaded in lanes 7 and 2. Upon base treatment, Hh-N undergoes a shift in mobility from 18.5 kD to 19.5 kD, the mobility of the unmodified Hh-N protein.

Ester bonds are subject to hydrolysis in alkaline conditions and base treatment prior to electrophoresis indeed reduced the migration of the cholesterol-driven reaction product to a position coinciding with that of the dithiothreitol-driven reaction product. These results are consistent with stimulation of the in vitro processing reaction by direct nucleophilic attack of cholesterol on the thioester intermediate to form an ester-linked adduct. If processing of Hh also results in formation of an ester-linked cholesterol adduct in vivo, then the protein-lipid linkage should be subject to base hydrolysis with a concomitant shift in electrophoretic mobility of the protein (normally 18.5 kD). The WO 98/30576 PCT/US97/15753 -120 immunoblot in Fig 24C shows the base-induced appearance of a species of reduced mobility (19.5 kD), which increased in abundance from -1/3 of the total after five minutes of treatment to most of the immunoreactive protein after one hour. This novel species comigrated with truncated, unprocessed Hh-N, which is not affected by base treatment. These data are consistent with an ester bond as the protein-lipid linkage in Hh- Np.

To confirm the involvement of cholesterol in formation of the Hh-Np adduct in vivo, S2 cells containing an inducible wild-type Hh construct were metabolically labeled with 3 H]cholesterol. Figure 25A is an audioradiogram of a gel loaded with total cell proteins from S2 cells containing a stably integrated Cu++-inducible hedgehog gene. Prior to harvesting, these cells were grown in media supplemented with ['H]cholesterol in the absence (lane 1) or presence (lane 2) of 1 mM CuSO4. 3 H]cholesterol incorporation is dependent upon Cu++ induction (lane 2) and is restricted to a single protein species migrating at a position corresponding to Hh-N,. Figure 25B is an HPLC profile of sterols separated on a C18 column by isocratic elution with a solvent containing methanol:ethanol:water (86:10:4) Rodriguez and L.W. Parks, Methods of Enzymology 111, 37, 1985). -5 gtg of each sterol was mixed, loaded, and elution monitored by absorbance at 210 nM. The structure of cholesterol is shown above cholesterol peak. Other sterols include: 1) 10 desmosterol, which contains one additional double bond between carbon 24 and 25; 2) 20 7-dehydrocholesterol, which contains one additional double bond between carbon 7 and 8; 3) campesterol which contains an additional methyl group on carbon 24; and 4) sitosterol, which contains an additional ethyl group on carbon 24.

Figure 25C shows HPLC analysis as in of the adduct released by base treatment of Hh-Np metabolically labeled with 3 H]cholesterol The radioactive species recovered from the metabolically labeled protein collates with cholesterol. Figure 25D shows metabolic labeling of vertebrate Sonic hedgehog protein with 3 H]cholesterol.

Autoradiogram of a gel loaded with total cell proteins from COS-7 cells transferred with a wild-type Sonic hedgehog expression construct (Shh, lane 1) or a construct that generates an unprocessed amino-terminal protein truncated after the conserved glycine WO 98/30576 PCT/US97/15753 121 at the site of autocleavage (Shh-N, lane The COS-7 cells were incubated in culture medium supplemented with 3 H]cholesterol for 24 hours prior to and 36 hours after transfection (COS-7 cells grown at 37 0 C in DMEM supplemented with 10% fetal calf serum were plated at -35% confluence onto two 35 mm dishes in 1 ml of Optimem media (Gibco) containing 1.5% fetal bovine sera and 25pCi of 3 H]cholesterol, giving as the final concentration of cholesterol with a specific activity of 2Ci/mmol (labeling medium). After 24 hours the labeling medium was removed and the cells were transfected for 6 hours with Shh or Shh-N expression constructs using lipofectamine (Gibco) and serum-free DMEM media. After transfection, 1 ml of fresh labeling medium was added to each dish and the cells were incubated for 36 hours at 37°C. The cells were then harvested without washing, lysed on the plate with Tris buffered saline plus 1% Triton X-100 and the total cell proteins were precipitated with acetone, washed and analyzed as described above for the S2 cell proteins). A strongly labeled species with the Shh but not Shh-N construct. Several other less heavily labeled species are apparent in both lanes, and may represent other cholesterol-modified proteins.

After 48 hours of growth in the presence of 3 H]cholesterol, induced and uninduced cultured cells were detergent extracted and total cell proteins were subjected to SDS- PAGE followed by fluorography (Metabolic labeling of S2 cultured cells with 3 H]cholesterol was performed essentially as described (Silberkang, et al., J Biol. Chem.

258:8503, 1983). Briefly, cells containing a stably integrated Cu"-inducible hedgehog gene were grown at 23 C for two weeks in Schneider cell media (Gibco) containing a fetal bovine serum depleted of lipoprotein (low cholesterol media, -20 Akg/ml cholesterol). These cells were then plated at 40% confluence onto two 35 mm tissue culture dishes (Nunc) in 1 ml of low cholesterol media supplemented with 300 /Ci of labeled cholesterol, [1.2.6.7- H 65 Ci/mM (NEN) giving a specific activity for cholesterol in this medium of -5 Ci/mmol. After 24 hours (1 doubling time) one plate of cells was induced to express Hh protein by the addition of CuSO, (1 mM final concentration). After an additional 24 hours the cells from both dishes were harvested, lysed in Tris buffered saline containing 1% Triton X-100, and total cell protein was WO 98/30576 PCT/US97/15753 122 precipitated with 5 volumes of cold acetone. The protein pellet was resuspended in 2% SDS in H,0 and reprecipitated with acetone several times to remove unincorporated radioactivity prior to loading onto SDS polyacrylamide gels for analysis. Initial labeling experiments in which 25 uCi of cholesterol was added resulted in 10 fold decrease in extent of label incorporated into the inducible Hh-Np protein). Whereas uninduced cells showed no incorporation of 3 H]cholesterol into cellular proteins, cells induced to express Hh showed a single strong band with a mobility corresponding to that of Hh-N (Fig 25A). Given the hydrophobic character of Hh-NP, these results suggest that either cholesterol itself or a sterol derivative constitutes the lipophilic adduct of Hh-N,. To determine whether cholesterol is the final form of the adduct, radio-labeled Hh-N protein excised from a gel was base-treated to release the adduct, which was then isolated by either extraction (HPLC analysis of the Hh-Np adduct involved gel isolation of the radioactive band, KOH/methanol treatment of the band to break the ester linkage as described, followed by neutralization of the solution with acetic acid, drying in a speedvac, resuspension in H,0 and extraction of the hydrophobic radioactivity with ether.

After evaporation of the ether the sample was resuspended in isopropanol and applied to the C 18 column for analysis. Radio-labeled adduct was then subjected to analysis by HPLC with a method specifically designed to resolve various sterols Rodriguez and L.W. Parks, Methods ofEnzymology 111, 37, 1985) (Fig 25B). The radioactive adduct released from Hh-N, eluted at the same position as the cholesterol standard, and no radioactivity was detected in any other fraction (Fig The amount of radioactive cholesterol incorporated is consistent with that expected if all of the Hh-Np synthesized upon induction received a cholesterol adduct (The specific activity of 3 H]cholesterol in the S2 cell labeling medium was -5 Ci/mmol. Assuming after a 24 hour doubling time that this concentration approximately represents that within the S2 cell membrane, then any protein subsequently expressed and receiving cholesterol as an adduct would also be labeled at the same specific activity. As determined by standardized coomassie blue staining, -50-100 ng or 2.5 to 5 picomoles of Hh-Np is produced by one 35 mm dish of S2 cells containing the Cu"-inducible Hh construct WO 98/30576 PCT/US97/15753 123 during 24 hours of induction with ImM CuSO 4 This predicts 12.5 to 25 nCi or 2.75 x 104 to 5.5 x 10' dpm of radioactivity would be incorporated into Hh-Np protein produced in our labeling experiment assuming it is cholesterol modified. Total incorporation of radioactivity into Hh-N during the in vivo labeling experiment described above was measured at -5 x 104 dpm by excision and scintillation counting of an Hh-Np gel band), suggesting that other cellular components do not complete effectively as nucleophilic adducts in the in vivo autoprocessing reaction. Also consistent with a homogenous adduct, the mass of cholesterol is consistent with the mass previously measured by mass spectrometry of processed protein purified from cultured cells. A recent MALDI mass spectral analysis gave a mass of -430 daltons for the Hh- N, adduct, larger than the mass of cholesterol (386.6). Detection of this modification required that Hh-N be treated with CNBr/70% formic acid, i.e. full length Hh-N could not be detected. The mass discrepancy noted above could be accounted for by the net addition of formic acid (45 daltons) during CNBr digestion. This reaction could involve the addition of H 2 0 across the 5,6 double bond of cholesterol, a common reaction of secondary alkenes in strong acids Morrison, R.N. Boyd, Organic Chemistry (Allyn and Bacon, Boston, ed.3rd, 1973)], followed by esterification of formate via this newly formed alcohol Cohen, G.S. Tint, T.Kuramoto, E.H.

Mosbach, Steroids 25, 365-378, 1975. To test whether the sterol backbone could be modified by the CNBr treatment, a positively charged cholesterol derivative (3p(N- (N',N'-dimethylamino) ethanecarbamoyl)-cholesterol, Sigma) detectable by MALDI was examined. It was found that incubation of this sterol derivative in 70% formic acid alone resulted in the addition of 45 mass units to the sterol a mass consistent with the net addition of a formic acid molecule). These in vitro and in vivo results show that the Hh-C processing domain functions as a cholesterol transferase; as a result of this activity, a cholesterol adduct is attached via an ester linkage to the COOH-terminus of the NH,terminal signaling domain of the Hh protein.

To test whether processing of vertebrate hedgehog proteins results in the incorporation of cholesterol as a covalent adduct to the signaling domain, cultured green monkey WO 98/30576 PCT/US97/15753 124kidney cells (COS-7) were metabolically labeled with ['H]cholesterol and transfected with expression constructs containing the full length murine Sonic hedgehog (Shh) open reading frame, leading to production of an autocatalytically processed signaling domain (Shh-N) or (ii) Shh coding sequences precisely truncated at the site of cleavage, thus producing an unprocessed amino terminal signaling domain (Shh-N) (COS-7 cells grown at 37 0 C in DMEM supplemented with 10% fetal calf serum were plated at confluence onto two 35 mm dishes in 1 ml of Optimem media (Gibco) containing fetal bovine sera and 25"Ci of 3 H]cholesterol, giving ~40,g/ml as the final concentration of cholesterol with a specific activity of 2Ci/mmol (labeling medium). After 24 hours the labeling medium was removed and the cells were transfected for 6 hours with Shh or Shh-N expression constructs using lipofectamine (Gibco) and serum-free DMEM media. After transfection, 1 ml of fresh labeling medium was added to each dish and the cells were incubated for 36 hours at 37 0 C. The cells were then harvested without washing, lysed on the plate with Tris buffered saline plus 1% Triton X-100 and the total cell proteins were precipitated with acetone, washed and analyzed as described above for the S2 cell proteins). Cells expressing the full length construct contained a prominent radio-labeled species migrating at -19 kD, suggesting that cholesterol is covalently added to Shh-N, (Fig. 25D). This band was not present in cultures expressing the truncated Shh-N protein (Fig. 25D), indicating that the incorporation of ['H]cholesterol is dependent on the presence of the Shh processing domain. These data strongly suggest that the ability to attach cholesterol as a covalent adduct during autocatalytic processing and cleavage is a universal property of Hh proteins. Several other protein species in addition to the Shh amino terminal domain also appeared to incorporate cholesterol in cells transfected with either construct, suggesting that covalent modification by cholesterol extends to proteins beyond the Hh family. This possibility is consistent with the recently reported occurrence of several sequences homologous to the Hh processing domain in association with amino terminal sequences distinct from hedgehog.

WO 98/30576 PCT/US97/15753 -125 EXAMPLE An experimental model for holoprosencephaly derives from the occurrence of epidemics of congenital craniofacial malformations among newborn lambs on sheep ranches in several National Forests of the western United States. The most dramatically affected lambs showed severe holoprosencephaly, including true cyclopia and other craniofacial malformations characteristic of holoprosencephaly. The occurrence of these defects was traced to grazing by pregnant ewes on the range plant Veratrum califoricum. The compounds responsible were identified as a family of steroidal alkaloids; the structures of two of these, cyclopamine andjervine, are shown as compared to cholesterol in Figure 33. In Figure 33, sterols were extracted and analyzed by HPLC from COS7 cells metabolically labelled with 3 H]-mevalonic acid in the presence or absence ofjervine, a teratogenic plant steroidal alkaloid. In the presence of 28mM jervine, radiolabelled cholesterol levels were reduced and another radiolabelled sterol was found to accumulate. On the basis of its retention time in this reverse phase HPLC method, this abnormal sterol is tentatively identified as zymosterol, an intermediate in the cholesterol biosynthetic pathway.

Given the structural similarities of these compounds to cholesterol and the similar teratogenic effects of cholesterol synthesis inhibitors upon the offspring of pregnant rats, a reasonable mechanism to consider for the effects of these plant sterol derivatives was the inhibition of cholesterol biosynthesis. Accordingly, COS7 cultured cells treated with jervine were tested for defects in cholesterol biosynthesis by labelling with [3H]-mevalonic acid and then extracting and analyzing radiolabelled, non-saponifiable lipids.

Metabolic labeling and sterol analysis was essentially as described (Popjak et al. J. Biol.

Chem. 264: 630-6238.1989; Rilling et al. 1993 Arch. Biochem. Biophys. 301: 210-215.), with minor modifications. Briefly, COS-7 cells were plated at -35% confluence into two mm dishes at 37 0 C in 4 ml each of Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). After 24 hr of growth the medium in WO 98/30576 PCT/US97/15753 126each dish was replaced with 2 ml fresh medium with 10% FBS; ['H]-mevalonic acid (NEN #NET 176) brought to a specific activity of 0.8 Ci/mmol in a 1% solution of bovine serum albumin was added to this medium to a final concentration of20mM. At this time, one dish received 6ml of a 4 mg/ml solution of jervine in ethanol (final concentration 28 mMjervine), and the other received 6 ml of ethanol. After 24 hr further incubation, cells were washed in PBS, extracted with methanol, and I M potassium hydroxide (KOH) added to 10%. Following a three hour incubation at 60 0 C, the methanol/KOH mixture was extracted with diethyl ether, the extract dried down, resuspended in isopropanol, and subjected to reverse phase HPLC analysis by the method of Rodriguez and Parks (Methods in Enzymology 111:37-511985).

Treated cells synthesized reduced levels of cholesterol and accumulated increased levels of another sterol that we have provisionally identified as the cholesterol precursor, zymosterol. The natural product jervine at these concentrations thus inhibits cholesterol biosynthesis in cultured cells in much the same manner as the synthetic drugs discussed above, although the specific enzyme(s) affected appear to differ. Given the similarities in their teratogenic effects, this inhibition seems likely to underlie the teratogenic effects of both the synthetic and natural compounds.

WO 9850576 PCTIUS97/15753 127

EXAMPLE

Protein exprcssion and Purification: Drosophila mnelanogasrer Hh oneiin which most of the! arrino-te rm, Mal signaling domnain and signal sequencc have bcen replaced by a hexa-hiscidine msr (His 6 was expressed as previously described (Porter ei al., 1995). SeMet His 6 Hrh-C, was prcpared by expression in E. cali strain B83' (DE3) pLysS, a rnrethionine auxorp, ,and growth in rninimnal media as previously described (Le-ahy et al., 19941). This His 6 -cagacd protein was purified on a Ni'-qT agaose column and autocicavage stimu la ted by addicicn of 50 mnM DTT. Aftr r,nioval of the DTT1 by dialysis, the cleaved prz)Lein was passed over a Ni?--NT1A agarose column znd the -hn czrooxy-terminal domain, Hh-C-!s, collec:ed in che column ran through. Hh-C25 was subjected to limited proteolysis by overnight irCLtibacion with 1:500 (w:wL) subtilisin (Bc.-Iriinger Ma-rnheim) at 4'C. A proteaste-stablc ragzment of aptproxirrately 17 k-Da, Hh-Cl- 7 was idcntifed by S DS-i'AGE and puriied by anion-exchange ch~ornacography utmlzing a Mon o-Q colunwi (?harrriacia). The amino- and carboxv-terrnza residues of H1h-C 17 were detertnined to be Cys-258 and Ser-4OS. rcspectively, by mass spectral analysis of cyanogen brorride-cleaved fragments. Mass spectrzl vnalysis was per-fo mn ed previousi y described (Ponte: cc al., 1996a).

Q ,stallization: Crystals were grown from hanging drops by the mnethod of vaccr diffusion M'lodawar et al., 1975). 6 4" Cf Z .A MEY/:51LI.~l i*HL-h-Cl 7 in 1.4 UM p-nicrcaotoethano1 2e~ v..~dih2 4cf a 1:1 diluti-n cif resz :vz: SoIIutEon (20% Po 2-35.0, EO m-M an-wnonium sulfate, and 10 rrJA sodium,,- cacod-A.a~e, -oH 5.2) wYi::- diszilHed %vat.e: an~d ciira~ ver thc reservcir Sai~tio. cryal'~s tvycallv grew co a final sizLe ci 0.2 mn 0.2 rnm x 0. 1 =i over days. Crystals are in group 1213 w,;ith unit cell dimension 0 1ciO.54 Ak.

Da a collecin -and roc ng: All data were collected fromn c:-vswais soaked in mothcer licuor made 10%7 ethyiene glycol and flash frozen in a zaseous Pi~r-gn strezm at -180-C. MkA3 data were collected at four wavelcniaths from a single Selvet c--'staId at bearnline X-A of Lhe National Synchrotron Light Source at Brookhaven National Laboratry. Data were collcted using Fuji HR-V phosphor-imaging places and digitized using a Fuji BA-3000 scanner. 20 oscillations at 0 and 0+180' wer-e Collected with no ovcriap for each oscillation range at ecUh wavelength. -'Ill diffraction innages were processcd using the progr-m DENZO and scaled wvith Le proz:am SCALEPACK (Orwinowskj and.Minor, 1997). and were- used fcr MAL~D phiase determidnacion and parialfly recordcd reflections wert used in all cases. Diffractionr data fromn different wavelengths were scaled with WVLSCL. aund values for FA and optimllal F Ind fWer= calculated with M--ADLSQ (Hendrickson, 199 Data collection slani'StICS are shown in Table L.

WO 98J30576 PCTIUS97/15753 128 Srmicturt detc~riantign: Tnree selenium sites were deduced from FA am:lI;Eudcs using' both the prog~ramn Sh7ELXS (Sheldrick, 1991) and Patterson methods. MAD phase: determrinaionS Were inadc with the progra.m~FP~ (Ccllaboracjve Computatiocnal ?rcet 1994, Rznzkrshn and Bicu. 1997), and solveri-flatre-ning and histogram--matching wert :ertor-ncd with zhe przararn DIM (Coilabcrative Compucational Project, 1994). An aiornic modtl ccnsiszing of Cys-258 to Tyr- A01 was reaidily bui!t into electron density maps computed wic.h MAD -derved phases for reflIections it. ,he range 20.0-2.0A. using- the prograrn (Jonies el: 199 One round of simulated annealing and sevcral rounds of Powell minimization using X-?LOR (Bruintgr, 1992) alternated with model buildinz with yielded the current model of HhCJAnjsi 2 o 4 residues, Cys-258 to Ala-402, and 126 water molecules. The mode! was refined usinz the data collected at 0.99 19 A. Onc rnclecule is present in the asyrnnimeL.rc unit, and che solvent content is avoroximaiely 59%. 4-ll bacikbone torsion angles are withi;n energetically acctptable regions. No electron density was observed for residues 403 to 408, but additional e-lectron density was observed ricar the thiol group cf Cys-258. As th'e c-'~icin~fe:~~n cacccyl5o acid, both AsO(CH 3 2 or an As at_-m were- modeled in this dcsbut neither thecstigahc- _Iactor ncr the irec R-fa::zr ncd with these atcrns a.dded to thc -_-_'ncmenc and tic atoms have beeni included in this region in the_ finial aiornruc model.

each mutated to alanrine (Hi329A, T326A, and TD303A, r:s-oe=.ivelv), and L--u-409 was =.taed to a stop codcr (His 6 Hh-C p) by the- method of Tecornbinant circle PC-~ (Jones a-rd Winstorfer, 1992). His 6 -tagged protei;ns contaiLnng resiusS 71oDroplaK ocnwthrsus 89-254 deleted and with Che mutated residues were exopressed in E. crii and ounnefid to near homogeneity as previously described (Porter et z-1, 1995). Autocleavage activity of the muIt proteins was assessed in 15 pI reactions by incubating I gg of protein in 150 n-LM MaCI, 100 M Th's-HCI (pH 0.05%I,, Triton X- 100, 1.25 mM P-mercaptoethanol, and 2.5%7- glycerol with either 50 rmM D77 or 35042M cholesterolll m-M DTfl for 6 hours at 30*C. T'he cleavage products were then fracdionared by SDS-PAGE and detected by Coomassie Brilliant Blue staining. The activity of D303A and ?iis 6 Hh-Cl 7 were also assessed by incubating I g of-protein with 46 pMf 3 H]cholestero! (11.6 Ci/rrmo1c)/1 cmuM DFF for 6 hours at 30 0 C. The vroteins were tnen suicjected to SDS-PAGEz and labeled proteins detected by autoradiogrza,,v.

Database searching and secuence alignment: Screening of the non-redundant protein sequence database at the NacionalI Center for Biotechnology Informaton (NTH) was Dcrformed usim.g the BLASTPGP program, which is xn enhanced vcrs ion of BLAST that Produces ~ed alignments WO 98/30576 PCT1US97/15753 129 (Altschul and Gish, 199 Additional searches were performed using the PSI-BLAST (Position- Specific Iterative BLAST) program, which constructs position-speific weight mnatrices fromn the BLASTPGP output arnd employs them for subsequenr iterations of database screening Using a modification of the BLAST statistics (Altschul et al., in. press). Alignm ents of multiple protein seqiuences were constructed using the CLUSTALW program (Thcxnpson er al., 1994) or the MACAW programn (Schuler ce al-. 199 1).

WO 98/30576 PCTIUS97/15753 130 Domain Identification arnd Structure Determination D~rosoohila niclanogaster Hh in which the sig-nal seauencc -rdMost of Chic arnino-tCe .nal s~cgnpIin domain havc been rz-piaced by a hexahiszidine tag was e: tssed in cri as previously described (Forcr et al., 1995). Following Duniication with Ni -NTA I -!arcsc, this protein cleaVeS itSelf: in vi=o in the Drsenc: of eihe !DTT or chlesLe.rol t- t'h- 25 kDa H h-C fragment (Hih-C 25 rtsidues Cys-258 to As-.17 Hn-C 2 5 reedby Uhis mnethod was found to be poorly soluble in thne absencc of deterzc-rs and' susccotible to Tun",r protColyLIc brzakdown when concentraced to 1 mzlnl or zreate. Trz:Menc of Hli-C-; N-:h subuiwS.;., n.owev~t-, resufted iM a i rutease-stable fragment of 17 kDa MOleCular weizht (Hh-C: 7 with impro-ved solubility.

Mass svectrcMetricl analysis of cyanogen brocmde cleavagc L 7*nmenzs of Hhl-C 17 showed it tcc c ,ss o s rsdues Cys-258 to Ser-408 (datu ncrL shown)1. Al~sd~ houev vd Hb-C hornclcgues (?zi-.er e( al., 19964), LcluC'ing :he nemna~tc s :utnczzs, are contained in Hri-

C

17 To dereine if Hh-C 17 retained iutoproctssing activity, a mutanC~rt verRIOn of Hi5--uagged H'h- C conca;nL;-. a t:rna.aion codon at rcsidue prsition 409 (HsR-;)was cxprcssed! and assayed for autocle-avage in the pres Once of DTT and cholesterol. As show~nn Fzourt 2. is 6 Hh-C 17 is cz-oable of :leaiin- >.in thco rtestncz DT utr nc: LHatS -"--CIi able to for-;. Uhc cLnicester in.tcrmediate (s;co iuz but th:, son-.- pcnicn Of LhC carbcxvtern-dnal 63 rosidues of I-Hh-Ci)5 (Leu-_409 to As:)47 1) is reouirtd for cbolesterol transfer.

C.vs~".s of-h-C11 7 t.xa h2C~ r. to =s 1.9 1 pa 072.- vc,' w -tzzl _aU fcc bct'- nlativC soeoehov- rtrc(Sci~dt) prcteir. Tecir scu:Zof 'l.

C

17 was Ic~ne y the method of nmtiWavcicng-th anoma:Ious -'~con (MAD) using Se-Met c,-ys:Is (H ri4Ckson et a !9 30; H.-ndrickscr., l131). H"igh cuv e LI!era e~c =ron density map:s allowed consu-jccior. of un atoauc model for Hh-C: -tsid-ci Cys-258 Cc) Ala-402 that readiy reine low R-factor wirh good ;,,cr--_herntisir/. Pini rtrmcen: und stcrcocne- szai~istics are sumrnazed in Tz,)I- 1 Description (if Hh-C17 Structure ?H-Ci- 7 possesses an a-bsu-jczr'-,e that Is roughly disk-s'.a~td with a Z- and width -20 A. Tne %-,uno and caroxyL=1:1c Mer in'eee fro=neI~ same su :fa o J-CIt 7 -6 aoar. A ribbon drzawing I.nd topology dlagvm of the H- screcsow inFigur: 3.

An unexcect:ed feavrtr Of tha_ Hh-C 17 stnjcture is Lhe presence or zwo hooussu;-doMaIns related by a pseudo-tw.;ofold axis of symmetry (Figures 4A and 43). The subdomains adopt an WO 98/30576 PCTIUS97/15753 131 ir-egular fold ch-anctrized by three extended p-hairpin loops and are ;inately associated, burying 1372 A2 of sur-ac: =ra at a hydrophobic interface such thiat a, sintzle hydrophobic core exists for che entire ?Eh-Cj-, rnoitcult. 7Tac topology of che Hh-Cj-, subdornairs rnaiches that of snak-. toxins Luch as cardiclCx-if V114 (Rees et al., 1990) and c-bungarotoxin (Lcve and Stroud, 1986), but the toxin and ;zh-Ci-, stt'icrres do nct supe.nmose well and thes.- stucnirts do not SeeCM C[tne ~iSe relatcd. A^s diSzcussed below, the full Hh-C I- fcld can be detEce-d in heself-s~licinz rezion of jnceinis (Duan et al., 1997), and the evidence for a divergent e'ioliutcrz~~y relationship in Etis case is s rrong.

Despit -a low level of sequence conseivaLion, the two Drosophila Hh-C 17 subdorrains are superimposabie with r.rn.s. deviation in a-carboti positions of 1.38 A, asid several notable structural features. including O-bulges and specific P-turn types. are conserved be~en the subdomains (Figure JPC). A sLucturet-'osed alignrn ol the Drosc#hila Hh-C,7 subdornain secunce issL1~W1in irc While 8 out of- 50 amino-acid residues ia rhis alignment art conser/ed, none, OF these 9 residues is absolutely conserved in both subdcrnains ofL* all Hh-C hornologues. A Characteristic or conscrved arnino-acid ry-pes, mostly hydrophobic reiues. d5 ct~mible in an enz of thcsc hornoica,.es, howe'.- I e !z:Vti Of s zr:ra 1 "!i1VV cLe two INn-C,- szi-dom a.;ns sugests th atc Hh- C17 could have zr~sci by tandem duplication off a primoran'l zene. 7re duplicated S~uencts do n~ot, however, ccze_soc-.d direct!v t*he compact subdomndn-s oose-ve!4i the Hh-C 1 structure.

sC- c. 2 aC V- C-hC z~on: Zav rc.. C) 0us regicnts. E~iac,;cfFivurc 43 s-cw.s how the loot) e.-chang: in nH.1-Cl- could be acbee by a siMDIC Divc: CJ: the ioo0-s abo-i.A a sing.!- flex Zoi-nL TheC struLC:Ur_1'iiy :31beSiVe subdL-oa': s Cf -h-

C

17 are thus mc-saics comy)osed ol elements trOm both units ofI the tand=-, scavence dualication.

To illustrate. te 1hr7= suc-essiVe_ loops in each 11L-i-C 17 subdomnain -elabeled 1-2-3 and A]-A_2- A3-B l-B2-33 in the duolicatd molecule prior to loop swapinhe cecageo the third loop) becten subdomais can~be represcnited as Al-A2-(A3-B l-B2)-33 where L the distinct subdornains are composed of loops either- inside or outside of the parenthtses (see Figure 4B). We not that duoiicaton cuoled with arr interdomain strucnra~l exchange such as atpears to haveo.ccurred in Hh-C 1 provides a mechanism to generate peruaticns in t-he order in which spDecific structural elements, occ-u- in the ariflno-acid sequence. Such V;em 1 uzaticris have beeni noteO in othner s i nclu din g sapos in h-c.-o lo gu es (Pon ti ng and RLs sell. !,995) an d b acterialslucanases (FKeinema-nn and Hahn. 1995).

The exchange c; domains cr elerrintis cf seccndmay stz-ucnire has betn observz.'d in se-veralj proteins and isbelieved to result in a.-more stable association ofsuburics in -nultidomain proteins (Beionert et aL., 1995). Exchange ofr structural rtgions has principally bcen observed between independent pclypeptidc chains w;ithin homodimers, but the lac operon. reprcs.sor and hornologJes WO 98/30576 PCT[US97/15753 132 also appear to repreSent a case of exchange between duplicared domains within a single poly-jeptide chain (Schumacher et al., 1994: Lewis et al., 1996).

Active Site Residues The ainro-terminal residue of Hh-C 17 Cys-258, is involved in bcth the thicester formation and cholesterol rninsfc.r steps of T-h autoprocessing (see Figure IA). Arnino-acid side chainis par-Icipacing directly in R~h autoprocessing chemistry will most likely pcssess polaz groups, and the only such residues near Cys-258 in the Hh-C: 17 smacture are THis-32 9, T'nr-32 6, and Asp- 303. The anangzement of thesc three amino acids in relation to Cys-258 is shown in, Figures SA and 5B. His-3 29 and Thr-326 arc absolutcly conserv~ed in all Hh-C hornologues, and the side chains of both of these residues are within hydrogen bonding distance of the =-armno group of Cys-258 in the Rh-Ci? Sit-Jcrure. Asp-303 is invariably asparic acid or histidine in Hh-C domains, ?-rd the side chain of Asp-303 is exposed to solvent .4.2-4.5 A away frnm the Cys-2SS tla grouc. Significant structural rearnangements would appearrnecessary for additional residues in Hh-C 1 7 to rarricipatt directly in Hh autoprocessing. Cys-258, His-329, and Asp-2S2 do noz 107M. a ;r.zaelk ca.i'itI,; mad as haaee vr,-ocse,. (e cca., 19,1; rin-e al., 1 996 a).

To assess the involvement of His-329, 7hr-325, and As-303 ir R~h autozrocessing, each of these r-esiduts was mutated to ;laniine within the of thetlcnc I-is 6 Hh-C-. protein.

and the mucan, prozteins were :n~s and a--ayeA 7" uor~esn autocicavirg activity of thet rn.:a-_nt -,roteins in the oresence of lhigh =nrctntraticn.of D77I wvas: used_ as an assay for z1h-oestc.- frrcicn, the f~sep in, the Kri aurtcproczssin; reaction, wimile the autocleaving activity in the prestence of cholesterol was used to assay for cholesterol cranISfer. the second szen in the autourccssing reaction. The r.esults or these assays are shown I- Fiz'r= 14C.

F..is-329 is known from earlier e:.oeriments to be essential for Hh auroprocessing a clivit (Lee e al., 1994), and the His-329 to alanine mutant (H329A) was inactive in bocth the D TIT- aid cholesterol-stir.-ulated-rctions. Tne Thr-326 to Jlanine mutant (T326A) also showed greatly reduced activity in both assays. By contrast, the Asp-3O3 to alanine muLanc (D303A) was active in the DTT-stimulated reaction but inactive in the choles terol-s-zinu lated reaction.

7he loss or draarc reduction of autocleavirig activity in the presence of coth DTT and cholesterol for .=329A and T326A imnplicates both His-329 and Th-326 in forrrauion of the iner-nal thi-oesr=: during R~h autop roccssing. The nccrac~ion of thc side chains otfbotln of these residues with the a-amnino group o f Cys-258, a coi-.poncnt of the cleaved Deptide bond.- szronglZy implies a direct role for these re-sidues in thiocster formation. Possible rolets for KIs--3-29 during, chioester form, ation include stabilization of negative charge on thc carbonyl oxytgen o- Glv-257, WO 98/30576 PCT[US97/15753 133 donation of a Droton o, the free a-a.-iLno grcup of Cys-2-58, and maintenancz cf an appDropriate on;iencadti Of reaction Comnponents t.hrLough polar interactions. His-329 may also deorotonate the rirt hostcr formation, but if zhis is the case some rearr znemer-t of Cys-258 e.1Itive cc i:s position in Hh-C, 7 crystal strucm=r would be rcquired to- brn the thiol grou of ys-23 ino proximity witlh His-3 29. As the- pK, of the thiol groupinrecsetes a base may not be necded to cata-lyze :tidol deprotonaticn. Possible roles for- Tr36 in thicester fotmation se=r rnorc limited. The ;hig-h ;)K3 of a threonirie hydroxyl grouo makes T-ar-326 =t =nlkely ctuncicaie -rr proton transfers, suggesting that this residue i;s needed to frm~ p~olar :nt-:-.cuions tUhat, stabilize reactive, conformations within the .11 crotein.

The activity of the D303A mutant in DTTI- but noc cholesterol-stiniulatedZ auroprocessing shows that Aso,;-303 is not nezedad-- for thicester formation but is required for cholesteroCl transfer.

The negratveiy-charged apazic acid rtsid,,c sceems unikely to be involved inbinding a hydrophobic chcieszt-rOl uiolecula. A roic in activating the cholesterzi molecule for eonc attack of tE thioester appears mort piausible. For cholesterol to becomne ar. effoctive nuciecohiie, the 3.I-hydrox yl group must become Aoprctonaied, and Asp-303 is a good cazndidate f;or the gcecal astha-,aa :stsd n n~r:o of *A 303 -'is-idinc in c moicgues ;s Consistent wita hysp~cthesi5s hisztine is a-so ca~abie orf :ici~ as a gencrai base.

As indicated by the inacti.viry of Hh-CI7 in chclestErol Lserassays, residues in the 63 acids removed -foMM the Fih-C~r carboxv r,:us -ealso involved in cholesterol tnsfe.

residues i- ch oltsnzrol inngor actvaticn. The dcreased solubility of mh-~ latve to C~SUTZ:Ssst thet carox- A- s c' K-h-Cls ccsscss -1 e,-:o.Se hydrophobic region that cculd serve as a choicsterol bin(Lnz site.

Relationship between Hh-Cj-, and Self-Splicing Proteins An ezarlier an~alysis identifed a 36 amino acid conserve,' Motif :n the a.mino-Cerm.-r. 2 reonIs of Hh-C homoloc-ies and inteins (Koonin. 1995). A greatly expanded databasc of h- and intein sequences coied with recentn enhanc~reets of the BLAST melhod for dam'abasc seac.-Jn2 enabled e-xten~sicn of !he detec-taie r-gion of sequence si-mlarty to the: 00iotc~ia ==0aio acids of I.Eh-C and inmto sequences (p The improve-- me.thods for d~b 'n includ-. szatisticai1 analyis of ga: znrimets and iterative database scanning vith*- pcsitlnspeiffc -atrices dcerve d frc:-m pre vious BL AST outputs (Altschui et all, in press) When a database search was initiated with any of (he Frh-C scaucnces or with m.-ost of the inteirin scquezCts.

mem-,bers Of th: respective second croteinifamnily were the only additional seouco-ccsrenvatm the database at a s-,atisr~icailv si-nificant level.

WO 98/30576 PCTIUS97/15753 134 SO'olof tile Rh-C 17 Crystal structure showed the expanded rezion of Hh-Clintein secuence homnology Co ter Miate halfway through one Of the subdornains in (he turn region of an exposed loop between 0 sZtrands 3b and 4b (.sce Figure 3A). This cbservation, coupled with the presence of charcteristic endonuiclcase motifs in inrein sequences sihon-1v after teedo h detectable-1110E ShCitinhml g 1seted that the Iicin endonuclease doniain had been inserted into thc P-b-pr4b Of an Hh-C 1 7 -like structurc. This hypothesis focused the se-: for resurnpfion of the S-/n ejnseuence Sirndlaxity in intcin sequences likely to ollcw the endonucleas: rezion. The recantly determined crystal scructure of the PI-Scel ince-In Indeed shows the insenicti cf the endonuclease- region of tile intein in the 03b- 4b lcop: Of an ull-C 17 Ilke szrutu-e and indicates that the re-gicnl of the inteinsqec in whlich Ct to Hh-C 17 MUSt resume is ir.==eiate iy ina1 to the -second extein (Duan a-I, 1997). An aLgnment of RhB-C 17 &-id iniein sequencets is shown in F;Iuro- 6A. A fully structure-based LIgnmcnt of the Rh-

C

17 and inceirn s-cu-cn-ce-s a waits direct comrn snofteati coordinates of Hh-C 17 and P1- S ce!.

As can be seen in Figure 6A, asidie fr-om sites with conserved hydrophobic character the only residucs absoluely or nearly absolutely conser-ved between Rh-C homologues and irnteins are those identified in Ehe active site C-LRh-C 17 and shown to be inmportz-nt for th ,ioester f",oT-ation by muaeei.Te n-e .InL cvtimrsid-= dircc:ly nvle in1 thioester formaLiOn in Hh-C homolceucs zsrep laced by senne in som ienad tlese i~e oma cs' er rather than a thioestcr ineneiae.Te ycast HO endoruclease. whlich lacks an ami~note=inal serir-. or cysttine :sedces rnct havc scff1--)1-c-rn- a:'v'v x C:L I97) n th cnly intein horncioguc in Mer..zoc~:~jcnre~J~i).s 5Spsecr~ed to beI,, i as weU.

The only residue zbsolutely conserved between-^ 1-C hornolo u.S a neisi ahsiine corresoonding to His-329 inDrosophila Rh-C. The presencc of His-329 in the act-ive site of Drosophila Rhi-C and the loss of tliioester form ation activity When H15--329 is mutated strongly imply that tis histidine is conserved because it perfo rms a vital role in thioester formation and that it functions similarly in inreiris and Hh-C homnologues. The only other residue conIScrvcd in (he active Site Of Hlh-C homologu,.es and shown by mutagcnesis to be required for eff-icient thioester formaon, Th:-326, is also extremely conserved in inrein sequences. Of thie 39 incein sequences in the database at -the time oi our comparison, 34 -Sequences Contain a th -onine at 3 homologous position to Tnr-326, while t-eInin have serinc, and one each have nsp~jragint r zlura-,C acid at this position. Tne high levz! o-ffconservation of threonine at this active site position, and its substitution with sirm-ia. ai-.j-no acids suggrests a conserved role for this threanine in in reins and R-h-C homologues. A conser.,ed residue homologous to A~p-303. also found In theL- .h-C 17 active WO 98/30576 WO 9830576PCT/US97/15753 135 site, is not found in intein sequences, consistent with its role in cholesterol activation rather than tbioester formation.

As -xpct-ed from the sequence homology, the Structures of the se! f-sp licin g re-gion of the P1-Sce:! intein (Dunn et. zl. 1997) and Hh-C 1 7, are clearly homologous. Ahhiough not previcusly noted, the self-splicing region or PI-Scel conrams homologous subdoinains reclated by pseudosyrnumet-y. The PI-ScelI subdomains are homologous to the Hh-C 1 7 subdormains anid possess the same loop exchanae observed in Hh-C i However, these features 2are obscured by insertion of endonucl ease- associated seque-nces. In addition to insert.ion of the core e-ndonuclease domain in th~e region hor-ologous to the P3b-f4b loop, the PI-Scel iintein contains an additional insertion of amino acids relative to the Hh-C 17 structure. The site or this insertion occurs in the turn between A strands 1b and 2b in the Hh-C 1 structure (see Figure-- and this inserted region is believed to be involved in aiding, DNA recognition by the PI-Sccl inzein (Duan et a,1997).

Figure 6B shows a stereodiagran, of the Hh-C 1 7 smructurz depicted in the same orientation as the PI-Scel intein Structure in Duan, ct al. (1997) with the sites of the nnuiaesoctd insertions indicated.

The ccrse-vation o lsL~jcftuC, sceouence:, and cleavas: mehnsr.Htwe h-C hom.o Icg an d tE- intec-n rc z; rs c f SelIf- szlici vrote ins firrny es zbliHs'&-s Lhe divergence of these two protein famil-Lies from a common precu-.rsor. F; pre 7 shows a pinusible evolutionary scm-naio for the deveicoment, of the Hh-C and inre-in protein famnilies from a1 primordial domain of =Z mocul: is sufr.icien: :-ly fc e a2 :e2iaCc~se~r 0 i bond with a thicester or ese nboh th =Eauoprocessn gandse -a~ngecions. In both proteinflis residues ca-&bocxy te=,inal to the Hh-Cj, modulear nec-cdfo: scit-ing or contributing th*e second nucicocitile thiat resolves the initial e-stei/thicester and cet'=rrtunes the products of the overal-l reaction. 7he loss of detectble sequcence s~m-i'alty in the regzion of the C.

elegans Nh-C hornologues following the Hh-C, rnoduie Mann. personal cornrnun~icadon) raises the possibility that these residues mnay tranisfer a moclecule otecr trian chiolesterol. The ongoing expansion of sequence databases provides the prospect of additional H-h-C 1 modules being discovc.-ed- that initiate novel splicing or trasfer reactions by form.ation of ester or thioester intermediates.

WO 98/30576 PCTIUS97/15753 136 Table Legend Table I. statiszics for data collection, phase determrination and refinement Rsym and completeness values were calculated considering Bijvoets equivalent. Values in parentheses for <cI/al, are for the highest resolution shell (1.98-1.9 R 5 yn 100 X Ihi [Ii(h) ZhIi Ii(h). r.rn.s. (NJFl) r.m.s. (IFI) where 6F is the BijvoeE difference at one wavelengrth (values onl the diagonal) or Ehc dispersive difference bezweci two waveleng~ls (values off the diagonal). Also shown a:e. the anoA.'alous components of the Se scattering factors as a function of wavelength as detet-rained by MADLSQ (Hendrickson, 199 All data for which IFL>2a were used in the refincezent. A subset of the data (10%7) was egcluded from the rzfinement and used to calculate the free- R-valuc, (Bri~nger, 1992). A final round of relfinement including this data was per-formned to produce the Final set of coordinates and cxystallcgraDhic R-value. R-value ZllFol-lFcII I IFol.

WO 98/30576 PCT/US97/15753 137 Table I Statistics for data collection, phase determination and refinement Data Collection Statistics (30,0 to 1.9 A) Wavelength

(A)

Reflections

(N)

0.9919 0.9793 0.9791 0.9686 26,790 26,791 e-6,792 26,792 Redundancy Completeness 10.3 1(0.

10.3 100.0 10G.6 100.0 10.4 100.0 Signal 19.9 (4.2) 19.4 (3.8) 18.8 (3.6) 18.7 (3.5) 9.2 9.9 10.4 10.3 MAD Structure Factor Ratios and Anomalous Scattering Factors Wavelength

(A)

0.99 19 0.9793 0.9791 0.9686 0.9919 0.9793 0.9791 0.9686 f' 0.041 0.064 0.059 0.055 0.049 0.076 0.053 0.062 0.058 0.063 (e) -3.94 -9.45 -8.05 -4.15 (e) 0.51 3.28 6.03 4.12 Refinement and stereochemical statistics R-value freze R-value Average a Rrns deviations Bonds (A) Angles B-values (A42) 0.218 2a, 6. 0- 1. 0.222 (all F, 6.0-1.9A4) 0.275 (F>2c, 6.0-1 .9A) 0.2E3 (all F, C.-1 .91) 21.5 for protein, 39.6 for solvent 0.008 1.97 1 .30/1-45 bonds/angles of main chain 2.83/3.20 bonds/angles of side chains WO 9W30576 PCTIUS97/15753 138 References AlLschul. S. F. and Gish, W. (1996). Local alienruerit statistics. Mct. Enzyniol. 266, 460-481.

Altschnul, S. Madden. T. *Schaffer. A. Zhang, Zharg. Miller, W. and Livrnan, D. 1. Gapped BLAST and PSI-BLAST a new venecration of o~roEtc-,n dtaba.se sczrch pron'ams.

Nucleic A cids Res., in press.

Bernnen, M. Sch-lunegger, M. P. and Eisenberg, D. (1995). 3D doan swzconi::=: A -meChanism fOr cli gomer atssembly. Prot. Sci. 4. 2455-2468.

Brannigan, J. Dodson, DugCleby. H. Moody. C. Snuth1, J. Tc-mchick. D.

R. and Murzin. A. G. (1995). A protein catalytic frarnewor. with an N-te.-rrnna nucicaphile is capable of self-activation. 'Nature 378, 416-419.

Briinger, A. T. (1992). X-PLOR, Version 3.1 A systcrn for x-ray crystallography and NMR.

New Haven. C Yzie University.

Bumcrct, D. Takada, R. and McMahon, A. P. (1995). Proteciyric _Ccessirg yields two se-crercd, forms of .ronic hedgehog. Mol. Cell. Biol. iS, 2294-2303.

Burglin, T. R. (1996). Warthog and Groundhog, novel farniiics related to HedgehoE. Cur. Bial.

6, 1047-1050.

Collaborati1VC Cc pu tatioanal P.rojct. N. 4. (1994). The CC?4 Suit: Por~ o rt cystal~orapi~'.%czz Crystaliog:r. D-50. 7-16 Cooper. A. and Stevens, T. H. (1995). Pr.tEdin splicing: Self-splicing of genericafly mobile elements at tE1- protein levcl. Trends Z iochem. Sci. 20. 351l-356.

Du. Fi~le. S. a-Z: Qui ccho, F. 15997). Cr St Of-.7Tr t ;;Omirgc Endonuclease with ?P:olcin Solicing Activit. Cell 059, 555-564.

Ekker. S. NMcGrew. L. Lai, LTe 3. von Kcssier, D. FF. ccn' R. 7r.an Beachy, Ditrc xrsio n h-datviiso nesoftenCevggnr family of Xenapus Lewvis, Development 121, 2237-2347.

Ekker. S. "Ungar, A. C':ccnsteln, von Kessle., D. Porttr. J. AMoor., PR. T7. and BchP. A. (1995). Patrning, activities of ve.-ebrate hedge,'g prins in tedvlon v and brain. Bic]. 5, 94.4-955.

Fan, Porter. J. Chiang, Chan-, D. Beochy, P. A. and Tessi.-r-Lavigne, M.

(1995). Long-range scierotomei induction by sonic hedgehog: Dire:: role of the annino-te~nrinal plaag roduc: and modulatior by the cyclic AMP signaling parhwav. Ce!1 457-465.

Fietz, MK jacinito, Taylor, A. Alexandre-. C. and Inaa., F. VW,. (1095). Secreticon of the amino-remnal "ararent of the Hedgehco protein isLOs~ ~d.~:ceTfr hedgehlog signalling in Drosopohila. Cu=. Bioi. 5, 643-650.

~a~neschn~tM.. Brook, ard McMahon, A. (1997). The wu Ccording to hedgebog.

T::nds Genet. 13, 141-2 1.

WO 98/30576 PCT[US97/15753 139 Heinemnarn, U. and Hahn, M. (1995). Circular per1nuEton of0 oti1e1ec:1oSo ae Trends in Biochemnical Sciences 20, 349-350.

Hendrickson, W. A. (199 DeterMination of macrornolecula- s-,ucrireS from anomalous diffraction of syncnhrocrofl radiation. Scie-nce- 254, 51 -58.

Hendrickson, W. Horton. J. R. and LeMaster, D. M. (1990). Scienome-.hionyl -nroteins oroduced for anialysiS by rnultiwavelength anomalous diffraction (MAD): a vehcle ror direct deterrniafl of three-dimensional s~ruccure. EMBO J. 9, 1665-1672.

Hutchinson, E. 0. and Thcron, J. M. (1996). PROMOTIF--a program cc; 16enti ry anc. analyze structural rnocifs in proteins. Prot. Sci. 5. 2 12-220.

Hynes, Porter, 3. Chian-, Chana, Tcssier-Lavigne, IM., Scachy, P. A. and Rosenthal, A. (1995). Induction of mnidibrain dooaminergic necurons by Sonic hedgehog. Neuron 1j, 35-44.

Jones, D. H. and Winistorfe-r. S. C. (1992). Recombinant circle PCR and recombination PCR for site-specific miutagenesis without PCR product purification. B iotechniques 12. 528-530.

Jones,'T. Zou, J. Cowan. S. W. and Kjeldsaard. M. (1991). In! roved mcrhods for the building of protein rnethods..in electron density maps and the locaticn of errors in these mnodels.

Acta Crystallographica, A 47, 110-119.

Kocnin. E. V. (9 9 Ak 'c zj uncLtion modf in ndeo~ f an'1y prcte ns. Trtenas BiLocnern. Sci. 20. 4-12 Kraulis, P.3J. (1991). MOLSCRIFT: a program to produce both deta.iled and sche-matic )lts of protein structures. L. 2 4, 9 4-950 Lai, Ekker. S. Beachy, P. A. and Moon, R. 7. (1995). Patterming or the neural ectoderm of Xviaopus laevis by the mi.no-Lte nal product of hLed~thc- autooroteclytic clcavase.

Deve-ODna .12'i. 23e49-2360.

Leahy, D. E-ickson. H. Aukh_,!, Joshi, P. and Hendr-ickson. W. A. (1994)j.

Crys cali7ation of a frag.-entc of hurnan fibronecrin: Int~roduction of =Lhiorine by SiLe-cirected rnutagenesis to allow phasingr by selenomethionifle. Proteins 19. 48-54.

Lee, 1. Ekker. S. von Kessler, D. Porter, J. Sun. B. 1. and Beachy, P. A. (1994).

Auco-proteolysis in hedgehog protein biogenesis. Science 266. 1528-1537.

Lewis, Chang. Horton, N. Kercher, M. Pace, H. Schumacher, M. A., Brennan, R. G. and Lu, P. (1996). Crystal stnucture Of tEn laccose oneron repressor and its corzplexecs wit.) DNA and inducer. Science 2771, 124 7-1254.

Lopez-M?_tinez, Chang, D. Chiang, Porter. J_ Ros, M. Simnandl. B. K., -eahy, P. A. a llon. J. F. (1995). Limb-patterniflg activity and reszricved oosterior localization of the amino-terminal product of Sornic he-dgehog cleavaec. Curr. Bijol. 5(79 1-796), Love, A. and Suroud, R. K. (1986). TJhec crystal stricture of alpha-bun garotoxin at 2.5 A resolution.: rclation to solution strcture and binding to acetyicholine rcccpcor. Prot. Ens. 1, 37- 46.

WO 98/30576 PCT[US97/15753 140 Marti, Bumcrot, D. Takada., R. and McMahon, A. P. (1995). Requilrmenr of 19K form of sonic hedgehog for induction of distinct ventral Cell types in CNS explants. Nature 375, 322- 325.

Murzin, A. G. (1996). Structural classification cf proteins: new superfarauies. Curr. Opin.

ScruCt. Biol. 6, 386-394.

Otwinowski. Z. and Minor, V. (1997). Processing of x-r-.v diffira...dc n daza collected in oscillation' ,node. Meth. Enzymol. 276, 307-326.

IPerler, F. Olsen. G. J. and Adamn, E. (1997). Compoilation and analysis cf inteinecnc.

Nucleic Acids Res. 25, 1087-1093.

?ietrokovski.S. (1994). Conserved sea uence- fetureS of intceins (pro tein intronis) and their use in identifying new inceins and related proteins- Prot. Sci. 3, 2340-23'50.

Ponting, C. P. and Russell. R. B. (1995). Swaposins: Circular pc~utations within genes ecoding saposin homologues. Trends Bioche-a. Sci. 20. 179-180- Porter, E-kk^e:, S. Par k, von Kessler, D. YLoung, K. Chen, 3\a.

Woods, A. Cotter, R. Koonin, E. V. and Beachy. P. A1. (1996a). Hedgehog patterning activity: Role of a lipophilic modification mediated by the axycrnl autoprocessing domain. Cell 86. 2 1-34.

Porter, J. von Kessler,*-D. Ekker. S. Young, K. Lee, 1. Moses, K. and Beachy, P. A. (1995). The product of he.geho- auto~roceo!-ytic cicavc-3 a::ive i c-cal zrange signalln-i. -Natue 374, ~-aca Porter, 1. Young, K B. ard Beachy.?P. A. (1996b). Cholesterol eodificarion of:edo signaling proteins in animal development. Science 274, 255-259.

?ashaV. 1and '=iec1, V. 199) O T~ne Lf~~~c as a snecial case of rnultrle iSOrrorvbous re-placement. Meth--.vi. 2 76. -5323-357.

Rees, Biw, Sarnazma, J. P. and Moras, D. (1990). C-'\a~xi a inossamoia s~bic a.Tereze crystal szz-ucture. 3. M.oi. 214 3 1-297.

Roelink, Porter, J. Chiang. Tanabe, Chang, D. 3eachy, P. A. and esi,7 A. (1995). Floor plate and mnotor neuron induc-ion by different concznraions of the anminoterminal cleavage product of sonic hed-geho- autoproeoyi.Cl 5A Schuler, G. Altschul, S. F. and Lininan, D. J. A work*e for multiple alignment construction and analysis. Proteins 180-190.

Schumacher-, M. Choi, K. Zalkin. H. and Brennan. R. CG. (1994). Crystal struct-ure of Ladl =~mber, Pur-R, bound to DINA: Minor groove binding by r- he.-ices. Scie-nce- 266, 763-770.

She.; drck, 0. (199 Parte rscn Icreaonand the use of ma oeirdcltz-F data.

Dzaresboury, UK- Thompson, J. Higgins. D. G. and Gibson, T. J. (1994). CL-USTA-L imrroving te sensitivity of progressive rnultcicie seauencz alignment throug-h se.-ucrce weiehtig "osiion specific gap penalties and weiihr rnatrLx choice. Nucleic Acids Rcs. 22; 4673-4680.

WO 98/30576 PCT/US97/15753 Wiodawer, Hodgson, K. 0. and Shooter, E. M. (1975). Crystallization of nerve growth factor from mouse subrnzxiflary glands. Proc. Nat]. Acad. Sci. USA 72. 777-779.

Xu, and Perler, F. B. (1996). The mnechanismi of protein splicing and its modulation by mutation. BO~O J. 15, 5146-5 153.

WO 98/30576 PCT/US97/15753 142- The above disclosure generally describes the present invention. A more complete understanding can be obtained by reference to the following specific examples which are provided herein for purposes of illustration only and are not intended to limit the scope of the invention.

WO 98/30576 PCT/US97/15753 143- SEQUENCE LISTING GENERAL INFORMATION: APPLICANT:The Johns Hopkins University School of Medicine et al.

(ii) TITLE OF INVENTION: NOVEL HEDGEHOG-DERIVED POLYPEPTIDES (iii) NUMBER OF SEQUENCES: (iv) CORRESPONDENCE ADDRESS: ADDRESSEE: Fish Richardson P.C.

STREET: 4225 Executive Square, Suite 1400 CITY: La Jolla STATE: CA COUNTRY: U.S.A.

ZIP: 92037 COMPUTER READABLE FORM: MEDIUM TYPE: Floppy disk COMPUTER: IBM PC compatible OPERATING SYSTEM: PC-DOS/MS-DOS SOFTWARE: PatentIn Release Version #1.30 (vi) CURRENT APPLICATION DATA: APPLICATION NUMBER: FILING DATE: 07-OCT-1997

CLASSIFICATION:

(vii) PRIOR APPLICATION DATA: APPLICATION NUMBER: 08/729,743 FILING DATE: 07-OCT-1996

CLASSIFICATION:

(vii) PRIOR APPLICATION DATA: APPLICATION NUMBER: FILING DATE: 02-OCT-1997

CLASSIFICATION:

(viii) ATTORNEY/AGENT INFORMATION: NAME: Haile, Lisa A.

REGISTRATION NUMBER: 38,347 REFERENCE/DOCKET NUMBER: 07265/099W01 (ix) TELECOMMUNICATION INFORMATION: TELEPHONE: 619/678-5070 TELEFAX: 619/678-5099 INFORMATION FOR SEQ ID NO:1: SEQUENCE CHARACTERISTICS: LENGTH: 144 base pairs TYPE: nucleic acid STRANDEDNESS: both TOPOLOGY: both (ii) MOLECULE TYPE: cDNA (ix) FEATURE: NAME/KEY: CDS WO 98/30576 PCT/US97/15753 144 LOCATION: 1..142 (xi) SEQUENCE DESCRIPTION: SEQ ID NO:1: GTG AAA CTG CGG Val Lys Leu Arg 1 ACC GAG CCC TGG Thr Glu Pro Trp

GAC

Asp 10 GAA GAT GGC CAC Glu Asp Gly His CAC TCA His Ser CAG GAG TCT Gin Glu Ser GAC CGC GAC Asp Arg Asp

CTG

Leu CAC TAC GAG GGC His Tyr Glu Gly GCA GTG GAC Ala Val Asp ATC ACC ACG TCT Ile Thr Thr Ser CTG GCG GTG G Leu Ala Val CGC AGC AAG TAC Arg Ser Lys Tyr ATG CTG GCC CGC Met Leu Ala Arg INFORMATION FOR SEQ ID NO:2: SEQUENCE CHARACTERISTICS: LENGTH: 144 base pairs TYPE: nucleic acid STRANDEDNESS: both TOPOLOGY: both (ii) MOLECULE TYPE: cDNA (ix) FEATURE: NAME/KEY: CDS LOCATION: 1..142 (xi) SEQUENCE DESCRIPTION: SEQ ID NO:2: GTG AAG CTG Val Lys Leu CGG GTG ACC GAG Arg Val Thr Glu

GGC

Gly 55 TGG GAC GAG GAC Trp Asp Glu Asp

GGC

Gly CAC CAC TCA His His Ser GAG GAG Glu Glu TCC CTG CAT TAT Ser Leu His Tyr GAC CGC AAT AAG Asp Arg Asn Lys GGC CGC GCG GTG Gly Arg Ala Val GAC ATC ACC ACA TCA Asp Ile Thr Thr Ser CGC TTG GCA GTG G Arg Leu Ala Val

GAC

Asp

CGC

Arg TAT GGA CTG CTG Tyr Gly Leu Leu 142 INFORMATION FOR SEQ ID NO:3: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein WO 98/30576 PCT/US97/15753 145 (xi) SEQUENCE DESCRIPTION: SEQ ID NO:3: Ile Ser Ser.His Val His Gly Cys Phe Thr Pro Glu Ser Thr 1 5 INFORMATION FOR SEQ ID NO:4: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:4: Ser Ile Ser His Met His Gly Cys Phe Thr Pro Glu Ser Thr 1 5 INFORMATION FOR SEQ ID SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID Val Ala Ala Lys Ser Gly Gly Cys Phe Pro Gly Ser Ala Thr 1 5 INFORMATION FOR SEQ ID NO:6: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:6: Val Ala Ala Lys Ser Asp Gly Cys Phe Pro Gly Ser Ala Thr 1 5 INFORMATION FOR SEQ ID NO:7: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid WO 98/30576 PCT/US97/15753 -146 STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:7: Val Ala Ala Lys Ser Gly Gly Cys Phe Pro Gly Ser Ala Leu 1 5 INFORMATION FOR SEQ ID NO:8: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:8: Val Ala Ala Lys Ser Gly Gly Cys Phe Pro Gly Ser Gly Thr 1 5 INFORMATION FOR SEQ ID NO:9: SEQUENCE CHARACTERISTICS: LENGTH: 15 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:9: Val Ala Ala Lys Ser Gly Gly Cys Phe Pro Ala Gly Ala Arg Thr 1 5 10 INFORMATION FOR SEQ ID SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID WO 98/30576 PCT/US97/15753 -147- Val Ala Ala Lys Thr Gly Gly Cys Phe Pro Ala Gly Ala Gln 1 5 INFORMATION FOR SEQ ID NO:11: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:11: Val Ala Ala Lys Thr Gly Gly Cys Phe Pro Gly Glu Ala Leu 1 5 INFORMATION FOR SEQ ID NO:12: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:12: Leu Gly Val Arg Ser Gly Gly Cys Phe Pro Gly Thr Ala Met 1 5 INFORMATION FOR SEQ ID NO:13: SEQUENCE CHARACTERISTICS: LENGTH: 14 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:13: Leu Ala Val Arg Ala Gly Gly Cys Phe Pro Gly Asn Ala Thr 1 5 INFORMATION FOR SEQ ID NO:14: SEQUENCE CHARACTERISTICS: LENGTH: 8 amino acids TYPE: amino acid STRANDEDNESS: not relevant WO 98/30576 PCT/US97/15753 148 TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:14: His Gly His Gly Cys Phe Thr Pro 1 INFORMATION FOR SEQ ID SEQUENCE CHARACTERISTICS: LENGTH: 8 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID His Gly His Gly Cys Phe Thr Pro 1 INFORMATION FOR SEQ ID NO:16: SEQUENCE CHARACTERISTICS: LENGTH: 8 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) Lys 1 SEQUENCE DESCRIPTION: SEQ ID NO:16: Ser Gly Gly Cys Phe Pro Gly INFORMATION FOR SEQ ID NO:17: SEQUENCE CHARACTERISTICS: LENGTH: 416 amino acids TYPE: amino acid TOPOLOGY: linear (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:17: Met Asp Val Arg Leu His Leu Lys Gln Phe Ala Leu Leu Cys Phe Ile 1 5 10 WO 98/30576 PCTIUS97/15753 149- Ser Leu Leu Leu Thr Pro Cys Gly Leu Ala Cys Gly Pro Tyr Gin Tyr Pro Ala Ala Glu Glu 145 Tyr Tyr Ser Thr Asp 225 Phe Val His Thr Asp 305 Giy I Phe 1 21u Asn Asp Ile Gly 130 Gly Gly Tyr Val Leu 210 Arg Ile Ile Leu Phe 290 Thr .ys le ily ryr %rg Ser 115 rrp Arg Met Glu Ala 195 Gly Val Met Glu Val 275 Ala Cys Arg Pro Lys Asn Leu 100 Val Asp Ala Leu Ser 180 Ala Asp Leu Phe Thr 260 Phe Se2 G1 Arg F Asn Ile '1 Pro Z Met Met Glu Val Ser 165 Lys Lys Gly Ala Ile 245 Ser Val Asn i Ser i Gly 325 is Tal hr 70 ksp rhr %sn ksp Asp 150 Arg Ala Ser Thr Ala 230 Asp Glu Gly Val Lei 31( Pro I Ala C 55 Arg I Ile Lys His Gly 135 lie Leu His Gly Arg 215 Asp His Pro Asn Lys 295 i Lys ys 40 lu ksn lie krg rrp 120 His Thr Ala Ile Gly 200 Lys Glu Asp Phe Sez 28C Prc Sei 25 Lys Lys Ser Phe Cys 105 Pro His Thr Val His 185 Cys Pro Lys Pro Thr 265 Ser Gl Val Leu Thr Glu Lys 90 Lys Gly Leu Ser Glu 170 Cys Phe lie Gly Thr 250 Lys Ala Asr Th Va Thr P Leu C Arg 1 75 Asp C Asp I Val Glu Asp 155 Ala Ser Pro Lys Asn 235 Thr Leu Ala Thr Val 315 L Thr 'ro ;ly ?he lu -ys .ys 3lu krg Gly Val ly Asp 220 Val Arg Thr Ser Val 300 Lys Alz Leu Ala Lys Glu Leu Leu 125 Ser Asp Phe Lys Ser 205 Leu Leu Arg Leu Gly 285 LeL Arc Hi Gly I Ala Ser Glu Asn Asn 110 Arg Leu Lys Asp Ala 190 Gly Lys Ile Gin Thr 270 Ile Val Ile 3 Gly krg ryr Gly eu rhr* Ser Val His Ser Trp 175 Glu Thr Val Ser Phe 255 Ala Thr Trp Tyx Thi Gly Lys Lys Ile Asn Leu Thr.

Tyr Lys 160 Val Asn Val Gly Asp 240 Ile .Ala Ala Glu Thr 320 Ile Glu Giu His G1i Ser Phe Ala Prc 330 335 Ile Val Asp Gin 340 Val Leu Ala Ser Cys Tyr Ala Val Ile 345 Glu Asn His 350 WO 98/30576 WO 9830576PCJ21US97/15753 -150 Lys Trp Ala His Trp Ala Phe Ala Pro Val Arg Leu Cys His Lys Leu 355 360 365 Met Thr Trp Leu Phe Pro Ala Arg Glu Ser Asn Val Asn Phe Gin Glu 370 375 380 Asp..Gly Ile His Trp, Tyr Ser Asn Met Leu Phe His Ile Giy Ser Trp 385 390 395 400 Leu Leu Asp Arg Asp Ser Phe His Pro Leu Gly Ile Leu His Leu Ser 405 410 415 INFORMATION FOR SEQ ID NO:i8: SEQUENCE CHARACTERISTICS: LENGTH: 418 amino acids TYPE: amino acid TOPOLOGY: linear (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID NO:18: Met Arg Leu Leu Thr Arg Val Leu Leu Val Ser Leu Leu Thr Leu Ser Leu Arg Pro Lys Asn Leu Val Asp Ala 145 Leu Ser Val Arg Asn so Ile Pro Met Met Giu 130 Val Ser Lys Val His Val Thr Asp Thr Asn 115 Asp Asp Arg Ala Ser Pro Ala Arg Ile Gin 100 His Gly Ile Leu His 180 Gly Lys Glu Asn Ile Arg Trp His Thr Ala 165 Ile Leu Lys Lys Ser 70 Phe Cys Pro His Thr 150 Val His Ala Leu Thr 55 Giu Lys Lys Gly Phe 135 Ser Giu Cys Thr 40 Leu Arg Asp Asp Val 120 Glu Asp Ala Gly 25 Pro Gly Phe Giu Lys 105 Lys Giu Arg Gly Pro Leu Ala Lys Giu 90 Leu Leu Ser Asp Phe 170 Gly Arg Gly Ala Tyr Lys Ser Gly Arg Giu Leu Thr 75 Asn Thr Gly Asn Ser Leu Arg Val Thr 125 Leu His Tyr 140 Lys Ser Lys 155 Asp Trp Val Ala Giu Asn Tyr Gin Tyr Pro Ala Ala 110 Glu Giu Tyr Tyr Ser 190 Gly Phe Giu Asn Asp Ile Gly Gly Gly Tyr 175 Val Arg Ile Gly Tyr Arg Ser Trp Arg Thr 160 Giu Ala Cys Ser Val Lys 185 Ala Lys Ser Gly Gly Cys Phe 195 Pro Gly Ser Ala Leu Val Ser Leu Gin 200 205 WO 98/30576 PCT/US97/15753 151 Asp Gly 210 Leu Ala 225 Phe Thr Thr Gin Phe Val Tyr Ala 290 Ser Gly 305 Gin Arg Asp Arg Ala His Phe Leu 370 Gin Glu 385 Gly Gin Lys Ala Ala Asp Asp Arg Glu Pro 260 Leu Asp 275 Ser Ser Gin Leu Gly Ser Ile Leu 340 Leu Ala 355 Phe Pro Gly Val Ser Asp 245 Val Asn Val Lys Phe 325 Ala Phe Gin His Ala 230 Ser Glu Ser Arg Ser 310 Ala Ser Ala Asn Trp 390 Val Lys 215 Gly Asn Thr Thr Lys Ile Thr Glu 280 Ala Gly 295 Val Ile Pro Val Cys Tyr Pro Ala 360 Ser Ser 375 Tyr Ser Leu Arg Thr 265 Asp Gin Val Thr Ala 345 Arg Ser Arg Val Arg 250 Leu Leu Lys Gin Ala 330 Val Leu Arg Leu Asp Leu Asn Pro Gly Asp Lys Val Phe 235 Val Thr His Val Arg 315 His Ile Tyr Ser Leu 395 220 Ser Phe Ala Thr Met 300 Ile Gly Glu Tyr Asn 380 Tyr Asp Phe Tyr Val Ala His 270 Met Thr 285 Val Val Tyr Thr Thr Ile Asp Gln 350 Tyr Val 365 Ala Thr Gin Met Ile Ile 255 Leu Ala Asp Glu Val 335 Gly Ser Leu Gly Met 240 Glu Leu Ala Asp Glu 320 Val Leu Ser Gln Thr 400 Trp Leu Leu Asp Ser Asn Met Leu His Pro Leu Gly Met Ser Val Asn 405 410 Ser Ser INFORMATION FOR SEQ ID NO:19: SEQUENCE CHARACTERISTICS: LENGTH: 425 amino acids TYPE: amino acid STRANDEDNESS: not relevant TOPOLOGY: both (ii) MOLECULE TYPE: protein (xi) Met 1 Cys SEQUENCE DESCRIPTION: SEQ ID NO:19: Val Glu Met Leu Leu Leu Thr Arg Ile Leu Leu Val Gly Phe Ile 5 10 Ala Leu Leu Val Ser Ser Gly Leu Thr Cys Gly Pro Gly Arg Gly 25 WO 98/30576 PCT/US97/15753 152- Ile Gly His Gin Phe Ile Tyr Glu Gly Pro Asn Tyr Ala Asp Arg Ala Ile Ser 115 Glu Gly Trp 130 Glu Gly Arg 145 Tyr Gly Met Tyr Tyr Glu Val Ala Ala 195 Leu Glu His 210 Arg Val Leu 225 Leu Leu Thr Val Ile Glu His Leu Leu 275 Ser Thr Ser 290 Pro Val Val 305 Val His Ser Thr Thr Ala Tyr Ala Val 355 Arg Arg His Pro Pro Lys Asn Leu 100 Val Asp Ala Leu Ser 180 Lys Gly Ala Phe Thr 260 Phe Gly Val Val Cys 340 Ile Asn Ile Pro I Met Met Glu Val Ala 165 Lys Ser Gly Ala Leu 245 Arg Val Gin Leu Ser 325 Gly Glu Val Thr 70 Asp Thr Asn Asp Asp 150 Arg Ala Gly Thr Asp 230 Asp Gin Ala Ala Gly 310 Leu Thr Glu Ala 55 Arg Ile Cys Cys Gly 135 Ile Leu His Gly Lys 215 Ala Arg Pro Pro Leu 295 Glu Arc Ile Hi Lys 40 Glu Asn Ile Arg Trp 120 His Thr Ala Ile Cys 200 Leu Asp Met Arg Gin 280 Phe Gly Glu Leu s Ser 360 Lys Ser Phe Cys 105 Pro His Thr Val Cys 185 Phe Val Gly Asp Ala 265 His Ala Gly Glu Ile 345 Trp Lys Leu Thr Pro Leu Ala Tyr Lys Thr Glu Lys 90 Lys Gly Ser Ser Glu 170 Ser Pro Lys Arg Ser 250 Arg Asn Ser Gin SAla 330 Asr SAla Leu Arg 75 Asp Asp Val Lys Asp 155 Ala Val Gly Asp Leu 235 Ser Leu Gin Asn SGin 315 Ser Arg His 3ly ?he 3lu Lys Met Glu 140 Arg Gly Lys Ser Leu 220 Leu Arg Leu Ser Val 300 Leu Gly Val Ala Ala Lys Glu Leu Leu 125 Ser Asp Phe Ala Ala 205 Ser Val Lys Leu Glu 285 Lys SLeu Ala Leu Ala 365 Ser Glu Asn Asn 110 Arg Leu Arg Asp Glu 190 Thr His Ser Leu Thr 270 Ala Pro Pro Tyr Ala 350 Phe Gly Leu Thr Ala Val His Ser Trp 175 Asn Val Gly Asp Phe 255 Ala Thr Gly Ala Ala 335 Ser Ala Arg Ile Gly Leu Thr Tyr Lys 160 Val Ser His Asp Phe 240 Tyr Ala Gly Gin Ser 320 Pro Cys Pro WO 98/30576 WO 9830576PCTIUS97/15753 -153 His Arg Leu Ala Gin Gly Leu Leu Ala Ala Leu Cys Pro Asp Gly Ala 370 375 380 Ile Pro Thr Ala Ala Thr Thr Thr Thr Gly Ile His Trp Tyr Ser Arg 385 390 395 400 Leu Leu Tyr Arg Ile Gly Ser Trp Val Leu Asp Gly Asp Ala Leu His 405 410 415 Pro Leu Gly Met Val Ala Pro Ala Ser 420 425 INFORMATION FOR SEQ ID SEQUENCE CHARACTERISTICS: LENGTH: 437 amino acids TYPE: amino acid TOPOLOGY: linear (ii) MOLECULE TYPE: protein (xi) SEQUENCE DESCRIPTION: SEQ ID Met Leu Leu Leu Leu Ala Arg Cys Phe Leu Val Ile Leu Ala Ser Ser 10 Leu Leu Val Lys Ile Gly Tyr Arg Ser Trp Arg 145 Met Glu Arg Pro Lys Asn Leu Val Asp 130 Ala Leu Ser Arg Asn Ile Pro Met Met 115 Glu Val Ala Lys Cys His Val Thr Asp Thr 100 Asn Asp Asp Arg Ala Pro Pro Ala Arg Ile Gin Gin Gly Ile Leu 165 His Gly Lys Glu Asn 70 Ile Arg Trp His Thr 150 Ala Ile Leu Lys Lys 55 Ser Phe Cys Pro His 135 Thr Val His Ala Leu 40 Thr Giu Lys Lys Gly 120 Ser Ser Giu Cys Cys 25 Thr Leu Arg Asp Asp 105 Val Giu Asp Ala Ser Gly Pro Gly Phe Giu 90 Lys Lys Giu Arg Gly 170 Val Pro Leu Ala Lys 75 Giu Leu Leu Ser Asp 155 Phe Lys Gly Ala Ser Glu Asn Asn Arg Leu 140 Arg Asp Al a Arg Tyr Gly Leu Thr Al a Val 125 His Ser Trp Giu Thr 205 Gly Lys Arg Thr Gly Leu 110 Thr Tyr Lys Val Asn Phe Gin Tyr Pro Ala Ala Glu Glu Tyr Tyr 175 Ser Gly Phe Glu Asn Asp Ile Gly Gly Gly 160 Tyr Val1 185 190 Val His Leu Ala Ala Lys 195 Ser Gly Gly Cys Phe Pro Gly Ser Ala 200 WO 98/30576 WO 9830576PCTIUS97/15753 154 Giu Val 225 Thr Glu Leu Ala Al a 305 Val His Ile Ala Gly 385 Al a Ile Ala Gin 210 Leu Phe Thr Phe Leu 290 Giu Thr Gly Glu His 370 Gly Giu Gly Val Gly Ala Leu Leu Val1 275 Phe Arg Leu Thr Giu 355 Ala Gly Pro Thr Lys 435 Gly Ala Asp Giu 260 Ala Ala Gly Arg Ile 340 His Leu Ser Thr Trp 420 Ser Thr Asp Arg 245 Pro Pro Ser Gly Glu 325 Leu Ser Leu Ile Al a 405 Leu Ser Lys Asp 230 Asp Arg His Arg Asp 310 Glu Ile Trp Ala Pro 390 Gly Leu Leu 215 Gin Glu Giu Asn Val 295 Arg Giu Asn Ala Ala 3'75 Ala Ile Asp Val Gly Gly Arg Asp 280 Arg Arg Ala Arg His 360 Leu Ala His Ser Lys Arg Ala Leu 265 Ser Pro Leu Gly Val 345 Arg Ala Gin Trp Giu 425 Asp Leu Leu Leu 235 Lys Lys 250 Leu Leu Gly Pro Gly Gin Leu Pro 315 Ala Tyr 330 Leu Ala Ala Phe Pro Ala Ser Ala 395 Tyr Ser 410 Thr Met Arg 220 Tyr Val Thr Thr Arg 300 Al a Al a Ser Ala Arg 380 Thr Gin His Pro Ser Phe Ala Pro 285 Val Ala Pro Cys Pro 365 Thr Glu Leu Pro Gly Asp Tyr Al a 270 Gly Tyr Val Leu Tyr 350 Phe Asp Ala Leu Leu 430 Asp Arg Phe Leu 240 Val Ile 255 His Leu Pro Ser Val Val His Ser 320 Thr Ala 335 Ala Val Arg Leu Gly Gly Arg Gly 400 Tyr His 415 Gly Met

Claims (7)

1. A substantially pure polypeptide characterized by having an amino acid sequence of a hedgehog polypeptide or a fragment derived from amino terminal amino acids of a hedgehog polypeptide, wherein the polypeptide or fragment thereof comprises a sterol moiety.

2. The polypeptide of claim 1, wherein the fragment is an extracellular hedgehog polypeptide fragment.

3. The polypeptide of claim 2, wherein the fragment has at its carboxy Si:i* terminus, a G CF cleavage site specifically recognized by a proteolytic 10 activity of the carboxy terminal fragment of a native hedgehog polypep- tide S4. The polypeptide of claim 1, wherein the sterol is cholesterol. The polypeptide of claim 1, wherein the hedgehog polypeptide is selected from the group consisting of Drosophila, Zebrafish, Xenopus, chicken, 15 murine and human hedgehog.

6. The polypeptide of claim 1, wherein the hedgehog fragment is about to 450 amino acids in length.

7. The polypeptide of claim 1, wherein the hedgehog fragment is about to 300 amino acids in length.

8. The polypeptide of claim 1, wherein the hedgehog fragment is about to 250 amino acids in length. 156

9. The polypeptide of claim 1, wherein the hedgehog fragment is about 100 to 200 amino acids in length. Dated this 31st day of October 2000 The Johns Hopkins University School of Medicine Patent Attorneys for the Applicant: F B RICE CO o *o