Cyclopamine Inhibition of Sonic Hedgehog Signal Transduction Is

Developmental Biology 224, 440–452 (2000) doi:10.1006/dbio.2000.9775, available online at http://www.idealibrary.com on

View metadata, citation and similar papers at core.ac.uk brought to you by CORE Cyclopamine Inhibition of Sonic Hedgehog Signalprovided by Elsevier - Publisher Connector Transduction Is Not Mediated through Effects on Cholesterol Transport

John P. Incardona,* William Gafﬁeld,† Yvonne Lange,‡ Adele Cooney,§ Peter G. Pentchev,§ Sharon Liu,¶ John A. Watson,¶ Raj P. Kapur,ሻ and Henk Roelink*,1 *Department of Biological Structure and Center for Developmental Biology and ࿣Department of Pathology, University of Washington, Seattle, Washington 98195; †Western Regional Research Center, ARS, United States Department of Agriculture, Albany, California 94710; ‡Department of Pathology, Rush–Presbyterian–St. Luke’s Medical Center, Chicago, Illinois 60612; §Developmental and Metabolic Neurobiology Branch, NINDS, National Institutes of Health, Bethesda, Maryland 20892; and ¶Department of Biochemistry and Biophysics, University of California at San Francisco, San Francisco, California 94143

Cyclopamine is a teratogenic steroidal alkaloid that causes cyclopia by blocking Sonic hedgehog (Shh) signal transduction. We have tested whether this activity of cyclopamine is related to disruption of cellular cholesterol transport and putative secondary effects on the Shh receptor, Patched (Ptc). First, we report that the potent antagonism of Shh signaling by cyclopamine is not a general property of steroidal alkaloids with similar structure. The structural features of steroidal alkaloids previously associated with the induction of holoprosencephaly in whole animals are also associated with inhibition of Shh signaling in vitro. Second, by comparing the effects of cyclopamine on Shh signaling with those of compounds known to block cholesterol transport, we show that the action of cyclopamine cannot be explained by inhibition of intracellular cholesterol transport. However, compounds that block cholesterol transport by affecting the vesicular trafficking of the Niemann–Pick C1 protein (NPC1), which is structurally similar to Ptc, are weak Shh antagonists. Rather than supporting a direct link between cholesterol homeostasis and Shh signaling, our findings suggest that the functions of both NPC1 and Ptc involve a common vesicular transport pathway. Consistent with this model, we find that Ptc and NPC1 colocalize extensively in a vesicular compartment in cotransfected cells. © 2000 Academic Press Key Words: neural development; holoprosencephaly; teratogen; U18666A; progesterone; membrane trafficking.

INTRODUCTION required for the induction of ventral cell types in the forebrain (Tanabe and Jessell, 1996; Dale et al., 1997). Loss Hedgehog family members are extracellular signaling of normal Shh function leads to a loss of ventral cell types molecules involved in embryonic patterning in many ani- all along the neural tube. This has signiﬁcant conse- mal phyla. In vertebrates, Sonic hedgehog (Shh) plays a quences, especially in the forebrain, which fails to divide critical role in patterning of the neural tube. Notochord- into hemispheres (Chiang et al., 1996; Roessler et al., 1996). derived Shh mediates both the induction of ventral cell An undivided forebrain, or holoprosencephaly, in its most types in the neural tube and the ventral exclusion of dorsal extreme form is associated with cyclopia (for review, see cell types, and Shh released from the prechordal plate is Siebert et al., 1990; Roessler and Muenke, 1998). Cyclopamine is one of a small group of steroidal alka- 1 To whom correspondence should be addressed at the Depart- loids, produced by Veratrum lily species, that induce holo- ment of Biological Structure, University of Washington at Seattle, prosencephaly and cyclopia in mammalian and avian em- Box 357420, HSB G527, Seattle, WA 98195. Fax: (206) 543-1524. bryos. Cyclopamine causes holoprosencephaly by blocking E-mail: [email protected]. the response to Shh of cells normally sensitive to Shh. The

0012-1606/00 $35.00 Copyright © 2000 by Academic Press 440 All rights of reproduction in any form reserved. Cyclopamine, Cholesterol Transport, and Shh Signaling 441 loss of Shh responsiveness within the developing neural tube, due to early embryonic exposure to cyclopamine, results in a phenocopy of the Shh null mutant (Cooper et al., 1998; Incardona et al., 1998). Our previous studies demonstrated that a suite of Shh-dependent cell types in the neural tube was affected in embryos with cyclopamine- induced holoprosencephaly. Furthermore, cyclopamine inhibits multiple Shh-mediated signaling events in other tissues at very different times of development, including somites (Incardona et al., 1998), heart (J.P.I., unpublished observation), gut (Kim and Melton, 1998), limb/ﬁn (Neu- mann et al., 1999), and skin (Chiang et al., 1999). Two decades of work by Keeler and colleagues demonstrated that unique structural features of cyclopamine and the related teratogenic compound jervine (Fig. 1) are required for the induction of holoprosencephaly in whole animal studies (reviewed in Gafﬁeld and Keeler, 1996a,b). Subtle changes in the structures of cyclopamine and jervine dramatically affect their teratogenic potency, suggesting that a precise interaction with a target is required for their inhibitory action. Since cyclopamine blocks the response to Shh, the cyclopamine target is likely to be involved in Shh signal transduction. Given the extreme hydrophobicity of cyclopamine and related compounds, we expect that their target is a membrane-associated protein, possibly an element of the Shh receptor complex. Elements of the Shh receptor are the multiple membrane spanning proteins Patched (Ptc) and Smoothened (Smo), which act in concert to mediate the Shh signal (Marigo et al., 1996; Stone et al., 1996; Chen and Struhl, 1998; for review, see Ingham, 1998; Murone et al., 1999). Shh binds to Ptc, which in turn releases the inhibition of Smo by Ptc, thus activating the FIG. 1. The structures of steroidal alkaloids and steroids tested for Shh response. inhibition of Shh signaling. The structure of cholesterol is shown at The structural similarity between cyclopamine and cho- top left and the generic jervane-type steroidal alkaloid structure is lesterol suggests a tantalizing potential mode of action of at top right. Natural jervane alkaloids and synthetic derivatives are cyclopamine, especially in light of the association of holo- listed below the generic structure, with differences indicated for prosencephaly, and thus attenuation of Shh signaling, with the functional groups RЈ and RЉ, and for the type of C–C bond at genetic or pharmacologic impairment of enzymes involved three variant positions. Cycloposine has a single glucose (glc) in the terminal steps of cholesterol biosynthesis. These residue in a glycosidic linkage to the 3-OH. Veratramine (middle, latter conditions produce both reduced levels of tissue left) occurs naturally and also is derived from cyclopamine by cholesterol and high levels of sterol precursors (Tint et al., electrophilic attack on the furan (E) ring under acidic conditions. The spirosolane-type steroidal alkaloids solasodine and tomatidine 1994; Kolf-Clauw et al., 1996). In particular, mutations in ⌬ are represented by a generic structure (middle, right). These two the gene encoding 7-dehydrocholesterol (7-DHC) reduc- alkaloids differ only in the position of the N in the F ring (RЈ and RЉ) tase (Fitzky et al., 1998) or inhibition of 7-DHC reductase and in the level of saturation at C-5, C-6. Note the differences in by the inhibitor AY-9944 promote mild forms of holo- the E–F ring attachment to ring D between jervanes and spi- prosencephaly (Roux and Aubry, 1966; Kelley et al., 1996). rosolanes, which results in an orientation of the spirofuranopiperi- The simple hypothesis that both cyclopamine and AY-9944 dine moiety of cyclopamine perpendicular to the planar steroidal inhibit the response of cells in the developing neural tube to framework. The glycoalkaloid saponins tomatine and solasonine Shh through the inhibition of cholesterol synthesis was possess 4- and 3-residue oligosaccharides attached to the 3-OH refuted, however, since AY-9944 is effective only when group of tomatidine and solasodine, respectively. The steroidal ␤ responding cells are deprived of exogenous cholesterol and amine U18666A [3 -(2-diethylaminoethoxy)androstenone] is represented at bottom left and progesterone at bottom right. therefore have to rely on endogenously synthesized cholesterol, while the action of cyclopamine is independent of the activity of the cholesterol synthetic pathway (Incardona et al., 1998). Accumulation of putatively teratogenic choles- holoprosencephaly-inducing activity of cyclopamine is in- terol precursors, rather than cholesterol depletion, may dependent of cholesterol supply and is most likely mediated underlie Shh-response inhibition by AY-9944. The by another mechanism.

In addition to cyclopamine and AY-9944, other com- dently on targets within the Shh signaling pathway and the pounds disturbing cholesterol homeostasis reportedly in- NPC1-mediated cholesterol transport pathway with vary- hibit the response to Shh (Cooper et al., 1998). These ing efficiency. Our data suggest that the remarkable sensi- compounds include other hydrophobic amines and steroids tivity of Shh signal transduction to cyclopamine and the that interfere with intracellular cholesterol trafficking, less potent effects of other compounds we tested reflect a such as imipramine, progesterone, and U18666A (Liscum primary effect on Ptc function. The structural similarities and Munn, 1999). This suggested the possibility that Shh of Ptc and NPC1 lead us to predict that Ptc mediates a signaling through Ptc is linked to cellular cholesterol ho- vesicular trafficking process that is differentially sensitive meostasis, in particular to intracellular cholesterol trans- to many of the same pharmacological manipulations and is port (Cooper et al., 1998). A target for the hydrophobic crucial for Shh signal transduction. amines and steroids that block intracellular cholesterol transport may be the Niemann–Pick C1 (NPC1) protein (Neufeld et al., 1999). Treatment of cells with these com- MATERIALS AND METHODS pounds induces a cellular cholesterol phenotype indistin- guishable from that of cells from Niemann–Pick type C Materials. Veratrum steroidal alkaloids and their synthetic (NP-C) patients, which lack the NPC1 protein (Liscum and derivatives were obtained as described previously (Brown and Faust, 1989; Rodriguez-Lafrasse et al., 1990; Roff et al., Keeler, 1978; Gaffield et al., 1986). Progesterone was obtained from 1991; Yoshikawa, 1991; Butler et al., 1992; Lange and Steck, Sigma (St. Louis, MO) and U18666A was obtained from BioMol 1994). NP-C disease is a neurodegenerative lysosomal stor- (Plymouth Meeting, PA). Steroidal alkaloids and progesterone were age disorder, with an underlying and severe defect in dissolved in 95% ethanol; U18666A was dissolved in water. intracellular transport of cholesterol and sphingolipids. Yet, Intermediate neural plate explants. Explants of intermediate neural plate regions were isolated from Stage 9–10 chick embryos loss of NPC1 function by mutation is not associated with and cultured in collagen gel as previously described (Yamada et al., affects on Shh signaling: NP-C patients and NPC1-null 1993; Incardona et al., 1998). Explants were distributed in batches mice do not have congenital malformations. This suggests of 10 onto pregelled collagen pillows and cultured at 37°C in that the inhibition of Shh signaling by cyclopamine, hydro- Neurobasal medium supplemented with N3 mixture [N2 plus 10 phobic amines, and steroids is not related to their effects on nM hydrocortisone (Bottenstein, 1992)], MEM nonessential amino intracellular cholesterol transport. Comparison of Ptc and acids, penicillin/streptomycin, and 10 mM glucose. Test com- NPC1 protein sequences demonstrates a significant simi- pounds were diluted in Neurobasal medium. In general compounds larity, largely, but not completely, confined to the were titrated to determine the highest concentration in which membrane-spanning regions (Carstea et al., 1997; Loftus et explants maintained viability for 12–14 h. Shh signaling was then al., 1997). This similarity includes a so-called “sterol- assayed at the concentrations just below those causing complete cell death within 12–14 h, but allowing viability to 30 h. Control sensing domain,” a pattern of conserved transmembrane explants were cultured with an equivalent concentration of solvent regions present in several other proteins involved in cho- (0.1% or less). For assessment of the Shh response, explants were lesterol homeostasis (Kumagai et al., 1995; Hua et al., 1996; harvested after 30 h of incubation, fixed in 4% paraformaldehyde, Carstea et al., 1997; Loftus et al., 1997; Bae et al., 1999). and processed for HNF-3␤, Isl1/2, and Pax7 immunofluorescence as The similarity between Ptc and NPC1 suggests another described previously (Incardona et al., 1998). The response to Shh mechanism by which hydrophobic amines and steroids was quantified by tallying positive nuclei with a hand counter for interfere with Shh reception: not through their documented at least 8 explants for each individual treatment. action on NPC1, but rather by acting on the same target as For 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGR) cyclopamine, a critical element of the Shh signal transduc- assays, 15 explants per well were tested in duplicate or triplicate. tion cascade. It remains a good possibility that this target is So that more cellular material was assayed, slightly larger explants which included some dorsal and ventral regions were excised. After Ptc, given its homology to NPC1. A converse prediction of 24 h of incubation with the test compounds in the same medium as this model is that cyclopamine and related compounds have above, explants were washed with PBS, then quickly frozen in a dry an effect on cholesterol homeostasis through their interac- ice/ethanol bath and stored at Ϫ80°C until assayed. tion with NPC1 comparable to that of progesterone or Shh-N preparation. Assays utilized recombinant Shh-N, either U18666A. obtained as conditioned medium from human 293T cells tran- To test these hypotheses we evaluated whether the siently transfected with pShh-N as described previously (Roelink et relative potencies of cyclopamine, structurally related ste- al., 1995) or partially purified from High-Five (Invitrogen, Carlsbad, roidal alkaloids that do not cause holoprosencephaly, CA) cells infected with a recombinant baculovirus expressing U18666A, and progesterone to inhibit Shh signaling corre- Shh-N (H.R., unpublished). The specific activity of each prepara- late with their effects on cholesterol transport. The results tion was determined by titration in the explant assay, and the Shh-N protein content was estimated by comparison to a standard show a correlation between the attenuation of the Shh after SDS–PAGE and Coomassie blue staining. Some preparations response and the disruption of NPC1-mediated vesicular of Shh-N were shown to contain a fraction of material acylated at trafficking. However, the relative concentrations of these an N-terminal Cys, which has a potency 30-fold higher than compounds sufficient to perturb cholesterol trafficking and unmodified Shh-N (Pepinsky et al., 1998). Our preparations most concentrations sufficient to block Shh signaling are very likely consisted of such a mixture, as suggested by the appearance different. This suggests that the compounds act indepen- of a doublet band on Western blots (data not shown). Also consis-

Copyright © 2000 by Academic Press. All rights of reproduction in any form reserved. Cyclopamine, Cholesterol Transport, and Shh Signaling 443 tent with this, the baculovirus preparation used for most of the RESULTS experiments described here proved to be 10-fold more potent than the bacterially expressed, unmodified protein originally used to The Neural Plate Explant Assay establish the Shh-N dose–response curve (Roelink et al., 1995; Ericson et al., 1997). However, neither the N-terminal acylation Chick embryo neural plate explants were cultured in the nor the C-terminal cholesterol modification of wild-type Shh is presence of Shh-N, which leads to the induction of markers ␤ required to assess the inhibitory activity of the compounds tested, for floor plate (HNF3 ) and motor neuron (Isl1/2) differen- as similar results were obtained using bacterially expressed Shh-N tiation, as well as repression of a marker for dorsal precursor (Cooper et al., 1998). cells (Pax7) (Roelink et al., 1995; Ericson et al., 1997; HMGR and cholesterol esterification assays. Neural plate ex- Briscoe and Ericson, 1999). In a typical Shh-N dose re- plants in collagen gels were processed and assayed for HMGR sponse, HNF-3␤ϩ floor plate cells are maximally induced activity as described previously (Plemenitas et al., 1990). Separate by 1–2 nM Shh-N, Isl1/2ϩ motor neurons are maximally aliquots of the cleared lysate were used for protein determination induced by 0.3–0.5 nM Shh-N, and Pax7ϩ dorsal precursor (Giron et al., 1993). For measurement of cholesterol esterification, cells are repressed by 0.05–0.1 nM Shh-N. Thus, at the 3 rat hepatoma cells at confluency were pulse-labeled with [ H]cho- highest Shh-N concentrations, HNF-3␤ induction predomi- lesterol and rinsed (Lange et al., 1993). Fresh medium containing nates, although this is at the expense of Isl1/2ϩ cells 10% fetal bovine serum plus the test compounds or a solvent (typically 60–100 Isl1/2ϩ cells and 1000–1200 HNF-3␤ϩ control was added and the cells were incubated for3hat37°C. The cells). Isl1/2ϩ cells are maximally induced (typically 200– cells were dissociated with trypsin, washed, extracted, and ana- lyzed by TLC for the radioactivity in cholesterol and cholesteryl 300 cells in a 1500-cell explant) at concentrations of Shh-N ␤ϩ esters (Lange et al., 1993). that induce few or no HNF-3 cells. Even lower concen- Assay of accumulation of low-density lipoprotein (LDL)- trations of Shh-N repress Pax7 in the absence of Isl1/2 or cholesterol in lysosomes. Normal human fibroblasts were seeded HNF-3␤ induction. Therefore, the number of HNF3␤ cells into two well chamber slides at 20,000 cells/well in McCoy’s is a straightforward, but somewhat less sensitive indicator medium supplemented with 5% lipoprotein-deficient serum and of Shh signaling. The Shh-dependent induction of Isl1/2 is antibiotics (LPDS-McCoy’s). After 4 days the cells were cultured more sensitive, but maximum response occurs in a rela- for 24 h in LPDS-McCoy’s containing the steroidal alkaloids or tively narrow dose range, with the number of Isl1/2ϩ cells progesterone (all dissolved in ethanol; final ethanol concentration per explant decreasing both at higher and at lower Shh was Ͻ0.1%) and LDL. The cells were then fixed and processed for concentrations and can thus be used as a reliable quantita- filipin/cholesterol histochemistry as previously described tive indicator of the Shh response only in conjunction with (Blanchette-Mackie et al., 1988). LPDS and purified human LDL Pax7 or HNF3␤ induction. were obtained from Intracel (Rockville, MD). At a constant high Shh-N concentration (2.5 nM), the DNAs, transfections, and immunofluorescence. The full- cyclopamine dose response mimics a reverse of the Shh-N length cDNA encoding chick Ptc-1 tagged with the influenza HA et al., et al., epitope (PtcHA) has been described previously (Marigo et al., 1996). dose response (Roelink 1995; Ericson 1997). ␤ϩ ϩ The PtcHA coding sequence was subcloned into pMT21 for expres- Induction of HNF-3 cells is reduced and more Isl1/2 sion in mammalian cells and was shown to function in endocytosis cells appear at low concentrations of cyclopamine (12–24 of Shh (Incardona et al., submitted for publication). The full-length nM); at intermediate cyclopamine concentrations HNF-3␤ human NPC1 cDNA contained in pSV-SPORT-1 was described induction is completely blocked (24–48 nM), while Isl1/2 previously and shown to restore normal cholesterol transport in induction persists; at higher cyclopamine concentrations NPC1-deficient cells (Watari et al., 1999). COS-1 cells, cultured in both HNF-3␤ and Isl1/2 induction is blocked, while some DMEM ϩ 10% fetal bovine serum, were transfected with PtcHA Pax7 repression persists (60 nM); all signaling is blocked by and/or NPC1 constructs using Lipofectamine (Gibco BRL, Be- 120 nM cyclopamine (Fig. 3, data not shown and Incardona thesda, MD) according to the manufacturer’s directions. Twenty- et al., 1998). Because each marker (HNF3␤, Isl1/2, and Pax7) four hours after transfection cells were trypsinized and seeded onto responds maximally to a different threshold Shh-N concen- poly-D-lysine-coated coverslips. At 48 h after transfection cells tration, comparisons of the inhibitory activity of the differ- were washed with PBS and fixed in 4% MEM-buffered paraformal- ent compounds are most straightforward using a single dehyde. After being washed with PBS and blocked in PBS ϩ 0.1% ϩ marker. Although in general we tested the effects of most of Triton X-100 5% normal serum, cells were incubated for 1 h at ␤ room temperature in primary antibodies, washed with PBS ϩ 0.1% the compounds on all three markers, only data for HNF3 Triton X-100, then incubated for 1 h at room temperature in are shown. secondary antibodies. Primary antibodies were FITC-conjugated anti-HA monoclonal 12CA5 (Boehringer Mannheim) at 1:50 or Steroidal Alkaloids Have Specific Structural anti-HA rat monoclonal (Boehringer Mannheim) at 1:250 and rabbit Requirements for Inhibition of Shh Signaling anti-human NPC1 polyclonal antibodies at 1:500 (Neufeld et al., 1999). Secondary antibodies (Jackson ImmunoResearch) were To determine if Keeler and colleagues’ results reflected an Texas red-conjugated anti-rat IgG and FITC-conjugated anti-rabbit interaction with a proximate target within the Shh signal- IgG. After final washes cells were mounted in 50% glycerol in 0.1 ing pathway, or differences in whole-animal pharmacoki- M sodium carbonate, pH 9.5, with a trace of phenylenediamine to netics, we tested many of their same collection of steroidal prevent bleaching. Confocal images were obtained on a Bio-Rad alkaloids both in the neural plate explant assay and by MRC600. direct application to chick embryos (data not shown). To

alkaloids thus affect Shh signaling at a concentration 100- fold above that sufficient to obtain comparable inhibition by cyclopamine, but are unable to completely block signaling. Similarly, application of 5 ␮g tomatidine or solasodine to chick embryos in ovo resulted in 1or 2 of 60 embryos with mild holoprosencephaly and minor defects in the floor plate at cervical levels (data not shown), suggesting that teratogenic concentrations are achieved in embryos only rarely. In contrast, malformations in cyclopamine-treated embryos are much more common and show greater severity both grossly and at the molecular level (Incardona et al., 1998). Therefore, the in vitro and in ovo effects on Shh signaling for steroidal alkaloids with E–F ring structures different from cyclopamine are consistent with their weaker teratogenicity. In addition to the E–F ring system, changes in both polarity and steric bulk of the planar steroid ring system also affect teratogenic potency. When administered by gavage to hamsters, jervine was consistently more potent than cyclopamine (Gaffield and Keeler, 1996b). This is FIG. 2. Structure/activity correlation for the inhibition of Shh likely due to the greater resistance of jervine to stomach signaling by steroidal alkaloids. VER, veratramine; TOM, tomati- acid-catalyzed opening of the furan ring than cyclopamine dine; SOL, solasodine; JER, jervine; 4-en-3-one, cyclopamine-4-en- (Gaffield and Keeler, 1996a), resulting in a higher effective 3-one; THJ, tetrahydrojervine; CpO, cycloposine. Chick neural plate explants were cultured with 2.5 nM Shh-N and the indicated dose in whole animals (Fig. 1). However, jervine is a weaker concentrations of steroidal alkaloids and processed for HNF-3␤ and teratogen in chick embryos, producing fewer holoprosence- Isl1/2 (data not shown) immunofluorescence. Values represent the phalic embryos (data not shown), and 5- to 10-fold less percentage of the mean HNF-3␤ϩ nuclei and SEM counted in potent than cyclopamine in the explant assay, producing an control explants treated with 2.5 nM Shh-N and solvent alone. incomplete (90%) block of HNF-3␤ induction (Fig. 2) at 240 Control explants had (mean Ϯ SE) 765 Ϯ 62 HNF3␤ϩ nuclei and nM. In contrast, oxidation of the 3-OH in cyclopamine to 77 Ϯ 11 Isl1/2ϩ nuclei. produce cyclopamine-4-en-3-one (Fig. 1) resulted in en- hanced potency in the hamster assay (Brown and Keeler, 1978), and this derivative produced a near-complete inhibition of Shh signaling in explants at 48 nM (Fig. 2). make structure/activity correlations with derivatives of Cyclopamine-4-en-3-one appears to be at least 2-fold more cyclopamine and other steroidal alkaloids, explants were potent than cyclopamine in the explant assay. challenged with increasing amounts of the compounds, and Other structural changes of the planar ring system also a block in the response to Shh-N was measured by the loss produced a loss of potency in the hamster assay (Brown and of HNF3␤ϩ (Fig. 2) and Isl1/2ϩ (data not shown) cells in Keeler, 1978; Gaffield and Keeler, 1993). Reduction of the explants. The effects of the compounds on Isl1/2 induction C-5, C-6 and C-12, C-13 double bonds of jervine produced were consistent with that observed for HNF3␤ induction, the weaker teratogen, tetrahydrojervine (Fig. 1). Tetrahy- described below. drojervine is about threefold weaker than jervine at block- A critical structural feature associated with the induction ing HNF-3␤ induction in explants (Fig. 2), producing 43% of holoprosencephaly is the E ring (Fig. 1), which is at a right inhibition at 240 nM. Finally, cycloposine, a glycoalkaloid angle in relation to the A–D rings. This furan E ring is easily analog of cyclopamine bearing a single glucose bonded to opened under acidic conditions, yielding veratramine (Fig. the 3-OH (Fig. 1), was found to be considerably weaker than 1), which does not induce holoprosencephaly. Steroidal cyclopamine in the explant assay (Fig. 2), but comparable to alkaloids such as tomatidine and solasodine (Fig. 1), which jervine. The reduction in potency secondary to the glyco- contain an E ring in the same plane as the A–D rings, have sidic attachment at the 3-OH suggests that cyclopamine very low teratogenicity and also do not induce holoprosen- does not act by binding to sterols: removal of sugar residues cephaly in hamsters (Gaffield and Keeler, 1996b). In con- is known to prevent the interaction of steroidal glycoalka- trast to the complete block in HNF3␤ induction by 120 nM loids with sterols (Roddick, 1979). cyclopamine, only partial loss of HNF3␤ϩ cells is achieved Overall, these data show that the relative teratogenic at 2.4 ␮M veratramine, tomatidine, and solasodine (Fig. 2). potencies for steroidal alkaloids in the hamster assay (Gaf- Concentrations of solasodine and veratramine higher than field and Keeler, 1996a,b) are very similar to their relative 10 ␮M were toxic to explant cells, and tomatidine at 24 ␮M abilities to block Shh signaling in explants, with a trivial produced only minimal inhibition of Isl1/2 induction and exception, the acid-catalyzed conversion of cyclopamine to Pax7 repression (data not shown). These latter steroidal veratramine in the nonruminant mammalian stomach. The

hibit responses to Shh-N, even at higher concentrations near the toxic threshold (Fig. 3). Although inhibition of HNF3␤ induction is detected with low concentrations of U18666A (80% of control at 0.048 ␮M), HNF3␤ϩ cells were detected in explants treated with 0.48 ␮M U18666A (19% of control), and Isl1/2 induction persisted at high levels (443% of control; data not shown). At U18666A concentrations of 2.4 ␮M HNF3␤ induction was completely blocked, but Isl1/2 induction still persisted at 33–66% of controls (data not shown). Concentrations higher than 2.4 ␮M were too toxic for explants. Progesterone proved to be an even weaker inhibitor of Shh signaling at very high concentrations relative to cyclopamine (Fig. 3). At 10 ␮M progesterone had only a moderate inhibitory effect on HNF-3␤ induction (68% of control), while increasing Isl1/2 induction (data not shown). Proges- terone at 30 ␮M produced a more significant inhibition of FIG. 3. Comparison of Shh-response inhibition by cyclopamine, HNF-3␤ (5%), but significant Isl1/2 induction remained U18666A, and progesterone. Chick neural plate explants were (67% of control, data not shown). Higher concentrations of cultured with 2.5 nM Shh-N and the indicated concentrations of progesterone killed the explants within 12 h. Progesterone cyclopamine, U18666A, and progesterone and processed for and U18666A were, thus, unable to block all of the Shh ␤ HNF-3 and Isl1/2 immunofluorescence as described under Mate- responses without significant toxicity, and the concentra- rials and Methods. Values represent the mean numbers and SEM of tions of progesterone and U18666A that maximally inhib- nuclei counted for HNF3␤, expressed as the percentage of the mean ited Shh signaling (30 and 2.4 ␮M, respectively) produced nuclei counted in control explants treated with 2.5 nM Shh-N and solvent alone. In the U18666A experiments, control explants had considerable cell death at the periphery of the explants (data (mean Ϯ SEM) 1552 Ϯ 100 HNF3␤ϩ nuclei and 71 Ϯ 8 Isl1/2ϩ not shown). In contrast, the cell death observed in explants nuclei; in the progesterone experiments, control explants had treated with 0.12 ␮M cyclopamine was comparable to that 750 Ϯ 47 HNF3␤ϩ nuclei and 28 Ϯ 5 Isl1/2ϩ nuclei. of explants cultured without Shh-N (data not shown), consistent with the complete block in Shh signaling. As

summarized in Table 1, we estimated the IC50s for inhibition of HNF3␤ induction by many of the compounds. significant teratogenic potencies and precise structural requirements of certain steroidal alkaloids likely reflect a specific interaction with a component of the Shh signaling pathway. TABLE 1 To further exclude the possibility that the Shh-inhibiting ␤ actions of cyclopamine and related compounds are indi- Comparison of IC50s for Inhibition of HNF3 Induction and Esterification of Cholesterol at the Endoplasmic Reticulum rectly caused by interference with cholesterol homeostasis,

we tested well-known cholesterol transport blockers for IC50 HNF3␤ IC50 esteriﬁcation their capacity to inhibit the Shh response in neural plate Drug inhibitiona inhibitionb explants in vitro. Cyclopamine 0.01 ␮M9␮M Jervine 0.08 ␮M7␮M The Prototypical Cholesterol Transport Blockers Veratramine 3 ␮M3␮M U18666A and Progesterone Are Weak Tomatidine 2 ␮M15␮M Shh-Response Inhibitors Solasodine 1 ␮M15␮M U18666A 0.1 ␮M 0.1–0.5 ␮Mc Using the explant assay, we compared the effects of Progesterone 15 ␮M3␮Md cyclopamine to those of two compounds, U18666A and a progesterone, with known effects on cholesterol transport, IC50s were estimated from plots similar to that shown in Fig. 3. b 3 most likely as a consequence of their blocking of normal Incorporation of [ H]cholesterol into esters was measured in rat NPC1 function (Higgins et al., 1999; Neufeld et al., 1999; hepatoma cells as described under Materials and Methods. IC50s were estimated by plotting the percentage of esteriﬁed cholesterol and Y. Lange, unpublished observations). U18666A and in cells treated with concentrations of steroidal alkaloids ranging progesterone were reported to block the response to Shh-N from 0.1 to 40 ␮M, relative to control cells incubated with ethanol ␮ at concentrations of 0.25 and 20 M, respectively (Cooper solvent only. Values were obtained from triplicate measurements et al., 1998). A more detailed analysis revealed that with SEM within 10%. U18666A and progesterone inhibit Shh signaling only par- c Underwood et al., 1996. tially at those concentrations and cannot completely in- d Lange and Steck, 1994.

Veratramine and the spirosolane steroidal alkaloids are roughly 100-fold weaker inhibitors than cyclopamine, U18666A is about 10-fold weaker, and progesterone is about 1000-fold weaker. If these compounds act indirectly on Shh signaling by interfering with cholesterol homeostasis in responding cells, there should be a similar correlation between the ability of each compound to inhibit the Shh response and the ability to block cholesterol transport.

The Ability of Steroidal Alkaloids and Steroids to Block Cholesterol Transport Does Not Correlate with Their Ability to Inhibit Shh Signaling

Under normal conditions, LDL-associated cholesterol is internalized by cells and subsequently distributed to various parts of the cell through an active process. Inhibition of intracellular cholesterol transport is manifested by several FIG. 4. Effects of steroidal alkaloids, U18666A, and progesterone changes in cholesterol homeostasis that can be quantified: compounds on HMGR activity in neural plate cells. Chick neural (1) increased activity of HMGR, the rate-limiting enzyme in plate explants were cultured for 24 h in 2.5 nM Shh-N and the cholesterol biosynthesis, due to reduced traffic of choles- indicated concentrations of cyclopamine (CpA), veratramine (VER), terol to the ER, and thus ER cholesterol levels; (2) inhibition U18666A (U), and progesterone (PG). HMGR activity was deter- of esterification of cholesterol in the ER, also reflecting mined as described under Materials and Methods. The values represent the means Ϯ SEM HMGR specific activity. reduced cholesterol trafficking to the ER; and (3) accumulation of internalized LDL-cholesterol in lysosomes, due to the failure of normal redistribution of internalized LDL- cholesterol from late endosomes. Although the precise inhibition of Shh signal response are mediated via distinct cholesterol transport pathway(s) measured by these assays targets. are incompletely characterized, the assays provide comple- To determine if the observed IC50s for HNF3␤ induction mentary information and overall produce similar results. correlated with IC50s for inhibition of cholesterol transport, Using these three assays, we found that at relatively high we utilized an assay of ER cholesterol esterification in a concentrations, all steroidal alkaloids have similar effects mammalian cell line (Table 1), since neural plate explants on cholesterol homeostasis and that a concentration of provide only very small amounts of tissue. At a cyclopamine cyclopamine sufficient to completely block Shh signaling concentration sufficient to completely block Shh response, does not disrupt cholesterol transport in neural plate cells. there was no effect on esterification of plasma membrane- HMGR levels were measured in chick neural plate ex- derived cholesterol (data not shown). At much higher con- plants treated with 2.5 nM Shh-N and various compounds. centrations cyclopamine and jervine had very similar ef-

Treatment with Shh-N alone (data not shown), plus 120 nM fects on cholesterol esterification, with IC50sof9and7␮M, cyclopamine, sufficient to completely block Shh respon- respectively. Veratramine, which does not induce holo- siveness, or 120 nM veratramine had no significant effect prosencephaly, was slightly more potent, with an IC50 near on explant HMGR levels (Fig. 4). Therefore, the complete 3 ␮M. All three jervane steroidal alkaloids produced an inhibition of Shh signaling in explants is not coupled to a 80–90% block of cholesterol esterification at 12 ␮M (data corresponding block in cholesterol transport to the ER. In not shown), which is near the toxic threshold for most cell contrast, treatment of explants with 120 nM U18666A types. The spirosolane steroidal alkaloids (tomatidine and resulted in a 4-fold increase in HMGR activity (Fig. 4), solasodine) were slightly weaker inhibitors of cholesterol although this concentration of U18666A induces only a esterification, with virtually identical IC50sof15␮M. partial block of Shh signaling (Fig. 3). Progesterone at 30 ␮M However, these compounds are orders of magnitude weaker produced a nearly 10-fold increase of HMGR activity (Fig. than cyclopamine in Shh signaling inhibition. The signifi- 4), but likewise, this concentration of progesterone pro- cant differences in teratogenicity among the steroidal alka- duced only an incomplete block of Shh signaling (Fig. 3). loids are not reflected in their capacity to block esterifica-

Thus even when movement of cholesterol to the ER is tion of cholesterol. However, the published IC50s for significantly blocked by U18666A or progesterone, some inhibition of cholesterol esterification by U18666A and Shh signaling still occurs. Overall, there is little correlation progesterone (Table 1) are relatively close to the concentra- between a compound’s ability to inhibit cholesterol trans- tions at which they inhibit Shh signaling, although proges- port, as it relates to upregulation of HMGR activity in terone appears to be weaker at inhibition of Shh signaling neural plate cells, and its ability to block Shh signaling, than inhibition of cholesterol esterification. indicating that the inhibition of cholesterol transport and Similar results were observed for accumulation of LDL-

FIG. 5. Accumulation of lipoprotein-derived cholesterol in lysosomes of cultured fibroblasts induced by steroidal alkaloids. Lysosomal cholesterol was visualized by fluorescent filipin staining as described under Materials and Methods. The images show low-power magnifications of fields of cells. Accumulation of cholesterol is apparent by the fluorescent signal in lysosomes within the cytoplasm surrounding nuclei. In C, E, and G, cells with no cholesterol accumulation are indicated with the unfilled arrowhead, cells with moderate accumulation indicated with the arrow, and cells with high accumulation indicated with the filled arrowhead. In B, D, F, and H all cells in each field have high accumulation of cholesterol in lysosomes. Cells carrying NPC1 mutations appear identically when similarly challenged with LDL (Blanchette-Mackie et al., 1988). Cells treated with 0.12 ␮M cyclopamine or 0.12 ␮M tomatidine appeared the same as control cells (data not shown), but cells treated with 0.12 ␮M veratramine showed significant lysosomal filipin staining (E). Scale bar (G) represents 50 ␮m.

cholesterol in lysosomes of cultured cells, assayed by filipin cholesterol was produced by 1.2 ␮M cyclopamine (Fig. 5C), cytochemistry (Fig. 5 and Table 2). Cyclopamine and toma- observed in 61% of treated cells, and 1.2 ␮M tomatidine tidine were found to have very similar effects on trafficking (Fig. 5G), in 58% of treated cells. Both compounds at 12 ␮M of lysosomal LDL-cholesterol. A moderate accumulation of (Figs. 5D and 5H, respectively) produced a maximal effect comparable to 30 ␮M progesterone (Fig. 5B), with 100% of treated cells showing filipinϩ lysosomes. Veratramine was significantly more potent in this assay, producing a moder- TABLE 2 ate cholesterol accumulation at 0.12 ␮M (Fig. 5E), with 67% Quantification of the Accumulation of LDL-Cholesterol in of treated cells showing filipinϩ lysosomes. Lysosomes Induced by Steroidal Alkaloids and Progesterone All the steroidal alkaloids tested thus have roughly similar cholesterol transport blocking activity, with veratra- ϩ Concentration Filipin cells/total mine more potent in some assays. These data are consistent Drug (␮M) (percentage) with observations that most hydrophobic amines can block None N/A 2/27 (7) cholesterol transport, but clearly indicate that effects on Veratramine 0.12 18/27 (67) intracellular cholesterol trafficking do not correlate with 12 25/25 (100) their teratogenic potency in blocking the Shh response. It is Cyclopamine 1.2 17/28 (61) notable, however, that NPC1 appears to play a key role in 12 27/27 (100) the cholesterol trafficking pathway(s) measured by these Tomatidine 1.2 19/33 (58) three assays, as defects in each are observed in cells from 12 28/28 (100) NP-C patients. Because some of these drugs may affect a Progesterone 30 29/29 (100) vesicular compartment linked to NPC1 function (Higgins

Copyright © 2000 by Academic Press. All rights of reproduction in any form reserved. 448 Incardona et al. et al., 1999; Kobayashi et al., 1999; Neufeld et al., 1999; quired for induction of holoprosencephaly in whole animals Holtta-Vuori et al., 2000; Watari et al., 2000), it is possible and to block Shh signaling in vitro. In contrast, the struc- that this same compartment is linked to Ptc function and tural requirements associated with inhibition of intracellu- Shh signaling. lar cholesterol transport are different. Although the large majority of compounds with this activity are simply hydrophobic amines with very diverse structures (Lange and Ptc-1 and NPC1 Colocalize Extensively Steck, 1994), there is signiﬁcant structural speciﬁcity for in Transfected Cells the inhibition of cholesterol transport by steroids, indicated NPC1 is normally found on cytoplasmic vesicles with by the potency of progesterone relative to other natural late endosomal characteristics (Higgins et al., 1999; steroids (Butler et al., 1992). Moreover, U18666A, which is Neufeld et al., 1999; Holtta-Vuori et al., 2000; Watari et al., both a hydrophobic amine and a steroid, is one of the most

2000). In the Drosophila embryo Ptc is found on multive- potent blockers of cholesterol transport, with an IC50 rang- sicular bodies, indicating that it is present in a late endo- ing from 100 to 800 nM depending on the assay (Underwood cytic compartment (Capdevila et al., 1994). Recent evi- et al., 1996). Nevertheless, both progesterone and U18666A dence suggests that Ptc-1, the vertebrate Ptc-family are relatively weak Shh antagonists, but U18666A is sig- member expressed in neural plate cells, mediates the endo- nificantly more potent than progesterone. Although high cytosis of Shh during signaling in neural plate cells (Incar- concentrations of cyclopamine can inhibit cholesterol dona et al., submitted for publication). In transfected cells, transport, we believe that this activity reflects a side effect epitope-tagged Ptc-1 (PtcHA) is found predominantly intra- unrelated to inhibition of Shh signaling. cellularly (Fig. 6A), with a significant component present in The effects of cyclopamine on both embryos treated in endocytic vesicles (Incardona et al., submitted for publica- ovo and neural plate explants must reflect inhibition of the tion). Although PtcHA is essentially undetectable at the signal transduction cascade at a very early point. Activation cell surface, PtcHA-expressing cells rapidly internalize Shh, of Shh target genes in neural precursors occurs through which colocalizes with PtcHA in vesicles (Incardona et al., either of at least two downstream branches: HNF3␤ expres- submitted for publication). The pattern of PtcHA immuno- sion in the floor plate is controlled by Gli transcription fluorescence is virtually identical to that of NPC1 overex- factors, whereas induction of motor neuron-specific genes pressed in the same cell type (Fig. 6B and Higgins et al., is independent of Gli function (Ding et al., 1998; Matise et 1999), with immunolabeled vesicles of 0.5–1 ␮m diameter al., 1998; Park et al., 2000) and appears to involve another appearing peripherally in filopodia, as well as a prominent transcription factor (Krishnan et al., 1997a,b). Because floor perinuclear cluster of positive vesicles. In cells cotrans- plate and motor neuron inductions are both remarkably fected with PtcHA and NPC1 constructs, both proteins sensitive to inhibition by cyclopamine, the target of cyclo- colocalize extensively in peripheral vesicles and in the pamine must be proximal to the bifurcation in the pathway perinuclear region (Figs. 6C–6E), suggesting that Ptc-1 and leading to these distinct fates. Other elements of the Shh NPC1 are targeted to the same compartment(s). signaling cascade that are upstream of and influence Gli function include Fused and Suppressor of fused (Carpenter et al., 1998; Ding et al., 1999; Kogerman et al., 1999; Pearse DISCUSSION et al., 1999; Stone et al., 1999). The fly homolog of Gli, Ci, also interacts with Costal2 (Robbins et al., 1997; Sisson et It was recently proposed that defects in intracellular al., 1997), but no vertebrate Costal2 homolog has been cholesterol transport and homeostasis are the common identified. A role for any of these elements in motor neuron basis for the induction of holoprosencephaly by cyclopamine development has not been demonstrated, but it seems and the 7-DHC reductase inhibitor AY-9944 (Cooper et al., likely that cyclopamine acts upstream of their involve- 1998). This hypothesis was based on the findings that ment. The only other upstream factors known to date are compounds known to block intracellular cholesterol trans- Ptc and Smo. port also inhibited Shh signaling in vitro and that high The simplest model explaining our data is a common concentrations of both AY-9944 and cyclopamine could mechanism of inhibition by cyclopamine, U18666A, and also block cholesterol transport. Our data do not support progesterone. It is unlikely that perturbation of Shh respon- this hypothesis for the following reasons: (1) Among the siveness is a direct or indirect consequence of NPC1 an- compounds tested, only cyclopamine completely blocked tagonism because Shh signaling is intact in embryos lack- Shh signaling in the absence of cytotoxicity, and terato- ing NPC1. No malformations are associated with NPC1 genic concentrations did not block cholesterol transport in mutations in humans or mice, although defects in choles- neural plate cells. (2) Cyclopamine and steroidal alkaloids terol transport can be detected at fetal stages (de Winter et with weak Shh inhibitory activity were generally equipo- al., 1992; Vanier et al., 1992; Dumontel et al., 1993; tent in assays of cholesterol transport. (3) U18666A and Rodriguez-Lafrasse et al., 1994). Moreover, defects in intra- progesterone are weak Shh antagonists, despite markedly cellular cholesterol distribution similar to that observed in affecting cholesterol homeostasis in neural plate explants. NP-C cells were recently associated with several other The unique structure of cyclopamine is absolutely re- genetic lipid storage diseases (Puri et al., 1999), none of

FIG. 6. Immunolocalization of PtcHA and NPC1 in transfected cells. COS-1 cells were transfected and processed for immunofluorescence as described under Materials and Methods. Each panel shows a high-magnification confocal image of a single transfected cell. (A) A cell transfected with human NPC1 alone. NPC1 immunofluorescence is evident as punctate signal representing numerous vesicles within cellular processes and in a cluster near the nucleus (nu). (B) A cell transfected with PtcHA alone. PtcHA immunofluorescence is evident in numerous vesicles within cellular processes, as well as in a cluster near the nucleus (nu). The perinuclear signal in this cell also includes structures with a more tubulovesicular appearance that may indicate Golgi-associated protein. (C–E) A cell cotransfected with human NPC1 and PtcHA. PtcHA (C, green) and NPC1 (D, red) colocalize extensively in both peripheral vesicles within cellular processes as well as in the perinuclear region, as shown by yellow signal in the merged image (E). Scale bar represents 10 ␮m.

which are known to produce defects in Shh signaling. Why and the regulation of Shh signaling by Ptc. This might would Shh signaling be inhibited as a consequence of explain the disruption of Shh signaling by drugs that block pharmacologic disruption of intracellular cholesterol trans- cholesterol transport (U18666A, progesterone), as it is port, but not genetic disruption of the same pathways? The known that intracellular trafﬁcking of NPC1 is dramati- relationship of Ptc to NPC1 and the known effects of cally affected by the same agents (Higgins et al., 1999; U18666A and progesterone on NPC1 function suggest that Neufeld et al., 1999). On the contrary, loss of one passenger Ptc function is the direct or indirect target of these com- on this pathway (e.g., NPC1) by mutation would not be pounds in the Shh signaling pathway. expected to disrupt the trafﬁcking of another (Ptc). Al-

We favor the hypothesis that a common vesicular traf- though the IC50s for inhibition of both cholesterol transport ﬁcking pathway may underlie both cholesterol transport and HNF3␤ induction are similar for both U18666A and

Copyright © 2000 by Academic Press. All rights of reproduction in any form reserved. 450 Incardona et al. progesterone, the idea that these two compounds affect Shh Butler, J. D., et al. (1992). Progesterone blocks cholesterol translo- signaling via inhibition of cholesterol transport is not cation from lysosomes. J. Biol. Chem. 267, 23797–23805. supported by the genetic data. Capdevila, J., et al. (1994). Subcellular localization of the segment The extensive colocalization of PtcHA and NPC1 and the polarity protein patched suggests an interaction with the wing- potential role of endocytic trafficking of Shh in signaling less reception complex in Drosophila embryos. Development 120, 987–998. (Incardona et al., submitted for publication) suggest impor- Carpenter, D., et al. (1998). Characterization of two patched tant functional ties to the structural similarity of Ptc and receptors for the vertebrate hedgehog protein family. Proc. Natl. NPC1. It is possible that all the compounds indirectly affect Acad. Sci. USA 95, 13630–13634. both NPC1 and Ptc by affecting another protein or lipid Carstea, E. D., et al. (1997). Niemann-Pick C1 disease gene: with which they have a common functional interaction. On Homology to mediators of cholesterol homeostasis. Science 277, the other hand, the potency of cyclopamine could reflect an 228–231. activity distinct from the other hydrophobic amines and Chen, Y., and Struhl, G. (1998). In vivo evidence that Patched and steroids. Nevertheless, recent studies provided evidence for Smoothened constitute distinct binding and transducing compo- regional lipid specializations within endosomal compart- nents of a Hedgehog receptor complex. Development 125, 4943– ments (Kobayashi et al., 1998; Mayor et al., 1998; Mukher- 4948. jee et al., 1999), including a lumenal region of late endo- Chiang, C., et al. (1996). Cyclopia and defective axial patterning in somes involved in cholesterol transport that is enriched mice lacking Sonic hedgehog gene function. Nature 383, 407– 413. with a unique lipid, lysobisphosphatidic acid, and perturbed Chiang, C., et al. (1999). Essential role for Sonic hedgehog during by U18666A (Kobayashi et al., 1999). We speculate that the hair follicle morphogenesis. Dev. Biol. 205, 1–9. function of both NPC1 and Ptc may be related to such Cooper, M. K., et al. (1998). Teratogen-mediated inhibition of target intracellular lipid specializations and that the regulation of tissue response to Shh signaling. Science 280, 1603–1607. Smo activity by Ptc involves vesicular trafficking. Dale, J. K., et al. (1997). Cooperation of BMP7 and SHH in the induction of forebrain ventral midline cells by prechordal meso- derm. Cell 90, 257–269. ACKNOWLEDGMENTS de Winter, J. M., et al. (1992). Prenatal diagnosis of Niemann–Pick disease type C. Clin. Chim. Acta 208, 173–181. We thank Ed Neufeld, Shutish Patel, and Joan Blanchette- Ding, Q., et al. (1998). Diminished Sonic hedgehog signaling and Mackie for discussions about NPC1 function and sharing unpub- lack of floor plate differentiation in Gli2 mutant mice. Develop- lished data, Cliff Tabin for the PtcHA construct, and Hidemichi ment 125, 2533–2543. Watari and Jerry Strauss for the NPC1 construct. This work was Ding, Q., et al. (1999). Mouse suppressor of fused is a negative supported by Grant ES09210 from the NIEHS to H.R. and R.P.K, by regulator of sonic hedgehog signaling and alters the subcellular a McKnight Scholar Award and an Atorvastatin Research Award to distribution of Gli1. Curr. Biol. 9, 1119–1122. H.R., by Grant ES07033 from the NIEHS to the University of Dumontel, C., et al. (1993). Fetal Niemann–Pick disease type C: Washington Center for Ecogenetics and Environmental Health, and Ultrastructural and lipid findings in liver and spleen. Virchows by Grant HL28448 to Y.L. from the NIH. J.P.I. was supported by a Arch. A Pathol. Anat. Histopathol. 422, 253–259. Medical Teratology Training Grant (HD07456) from the NIH. Ericson, J., et al. (1997). Pax6 controls progenitor cell identity and neuronal fate in response to graded Shh signaling. Cell 90, 169–180. REFERENCES Fitzky, B. U., et al. (1998). Mutations in the Delta7-sterol reductase gene in patients with the Smith–Lemli–Opitz syndrome. Proc. Bae, S. H., et al. (1999). Cholesterol biosynthesis from lanosterol. Natl. Acad. Sci. USA 95, 8181–8186. Molecular cloning, tissue distribution, expression, chromosomal Gaffield, W., et al. (1986). Carbon-13 and proton nuclear magnetic localization, and regulation of rat 7-dehydrocholesterol reduc- resonance spectra of Veratrum alkaloids. J. Nat. Prod. 49, 286– tase, a Smith–Lemli–Opitz syndrome-related protein. J. Biol. 292. Chem. 274, 14624–14631. Gaffield, W., and Keeler, R. F. (1993). Implication of C-5, C-6 Blanchette-Mackie, E. J., et al. (1988). Type-C Niemann–Pick unsaturation as a key structural factor in steroidal alkaloid- disease: Low density lipoprotein uptake is associated with pre- induced mammalian teratogenesis. Experientia 49, 922–924. mature cholesterol accumulation in the Golgi complex and Gaffield, W., and Keeler, R. F. (1996a). Craniofacial malformations excessive cholesterol storage in lysosomes. Proc. Natl. Acad. Sci. induced in hamsters by steroidal alkaloids. J. Nat. Toxins 5, USA 85, 8022–8026. 25–38. Bottenstein, J. E. (1992). Environmental influences on cells in Gaffield, W., and Keeler, R. F. (1996b). Steroidal alkaloid terato- culture. In “Neuromethods: Practical Cell Culture Techniques” gens: Molecular probes for investigation of craniofacial malfor- (A. A. Boulton, G. B. Baker, and W. Walz, Eds.), Vol. 23, pp. 392. mations. J. Toxicol. Toxin Rev. 15, 303–326. Humana Press, Totowa, NJ. Giron, M. D., Havel, C. M., and Watson, J. A. (1993). Isopentenoid Briscoe, J., and Ericson, J. (1999). The specification of neuronal synthesis in eukaryotic cells. An initiating role for post- identity by graded sonic hedgehog signalling. Semin. Cell Dev. translational control of 3-hydroxy-3-methylglutaryl coenzyme A Biol. 10, 353–362. reductase. Arch. Biochem. Biophys. 302, 265–271. Brown, D., and Keeler, R. F. (1978). Structure–activity relation of Higgins, M. E., et al. (1999). Niemann–Pick C1 is a late endosome- steroid teratogens. 1. Jervine ring system. J. Agric. Food Chem. resident protein that transiently associates with lysosomes and 26, 564–566. the trans-Golgi network. Mol. Genet. Metab. 68, 1–13.

Holtta-Vuori, M., et al. (2000). Mobilization of late-endosomal Mayor, S., Sabharanjak, S., and Maxfield, F. R. (1998). Cholesterol- cholesterol is inhibited by rab guanine nucleotide dissociation dependent retention of GPI-anchored proteins in endosomes. inhibitor. Curr. Biol. 10, 95–98. EMBO J. 17, 4626–4638. Hua, X., et al. (1996). Sterol resistance in CHO cells traced to point Mukherjee, S., Soe, T. T., and Maxfield, F. R. (1999). Endocytic mutation in SREBP cleavage-activating protein. Cell 87, 415– sorting of lipid analogues differing solely in the chemistry of 426. their hydrophobic tails. J. Cell Biol. 144, 1271–1284. Incardona, J. P., et al. (1998). The teratogenic Veratrum alkaloid Murone, M., Rosenthal, A., and de Sauvage, F. J. (1999). Sonic cyclopamine inhibits Sonic hedgehog signal transduction. Devel- hedgehog signaling by the patched-smoothened receptor com- opment 125, 3553–3562. plex. Curr. Biol. 9, 76–84. Ingham, P. W. (1998). Transducing Hedgehog: The story so far. Neufeld, E. B., et al. (1999). The Niemann–Pick C1 protein resides EMBO J. 17, 3505–3511. in a vesicular compartment linked to retrograde transport of Kelley, R. L., et al. (1996). Holoprosencephaly in RSH/Smith– multiple lysosomal cargo. J. Biol. Chem. 274, 9627–9635. Neumann, C. J., et al. (1999). Transient establishment of antero- Lemli–Opitz syndrome: Does abnormal cholesterol metabolism posterior polarity in the zebrafish pectoral fin bud in the absence affect the function of Sonic hedgehog? Am. J. Med. Genet. 66, of sonic hedgehog activity. Development 126, 4817–4826. 478–484. Park, H. L., et al. (2000). Mouse Gli1 mutants are viable but have Kim, S. K., and Melton, D. A. (1998). Pancreas development is defects in SHH signaling in combination with a Gli2 mutation. promoted by cyclopamine, a hedgehog signaling inhibitor. Proc. Development 127, 1593–1605. Natl. Acad. Sci. USA 95, 13036–13041. Pearse, R. V., 2nd, et al. (1999). Vertebrate homologs of Drosophila et al. Kobayashi, T., (1998). A lipid associated with the antiphos- suppressor of fused interact with the gli family of transcriptional pholipid syndrome regulates endosome structure and function. regulators. Dev. Biol. 212, 323–336. Nature 392, 193–197. Pepinsky, R. B., et al. (1998). Identification of a palmitic acid- Kobayashi, T., et al. (1999). Late endosomal membranes rich in modified form of human Sonic hedgehog. J. Biol. Chem. 273, lysobisphosphatidic acid regulate cholesterol transport. Nat. Cell 14037–14045. Biol. 1, 113–118. Plemenitas, A., Havel, C. M., and Watson, J. A. (1990). Sterol- Kogerman, P., et al. (1999). Mammalian suppressor-of-fused modu- mediated regulation of mevalonic acid synthesis. Accumulation lates nuclear–cytoplasmic shuttling of Gli-1. Nat. Cell Biol. 1, of 4-carboxysterols as the predominant sterols synthesized in a 312–319. Chinese hamster ovary cell cholesterol auxotroph (mutant 215). Kolf-Clauw, M., et al. (1996). Inhibition of 7-dehydrocholesterol J. Biol. Chem. 265, 17012–17017. reductase by the teratogen AY9944: A rat model for Smith– Puri, V., et al. (1999). Cholesterol modulates membrane traffic Lemli–Opitz syndrome. Teratology 54, 115–125. along the endocytic pathway in sphingolipid-storage diseases. Krishnan, V., et al. (1997a). Identification of a novel sonic hedgehog Nat. Cell Biol. 1, 386–388. response element in the chicken ovalbumin upstream promoter- Robbins, D. J., et al. (1997). Hedgehog elicits signal transduction by transcription factor II promoter. Mol. Endocrinol. 11, 1458–1466. means of a large complex containing the kinesin-related protein Krishnan, V., et al. (1997b). Mediation of Sonic hedgehog-induced costal2. Cell 90, 225–234. expression of COUP-TFII by a protein phosphatase. Science 278, Roddick, J. G. (1979). Complex formation between solanaceous 1947–1950. steroidal glycoalkaloids and free sterols in vitro. Phytochemistry Kumagai, H., Chun, K. T., and Simoni, R. D. (1995). Molecular 18, 1467–1470. dissection of the role of the membrane domain in the regulated Rodriguez-Lafrasse, C., et al. (1990). Abnormal cholesterol metab- degradation of 3-hydroxy-3-methylglutaryl coenzyme A reduc- olism in imipramine-treated fibroblast cultures. Similarities tase. J. Biol. Chem. 270, 19107–19113. with Niemann–Pick type C disease. Biochim. Biophys. Acta Lange, Y., and Steck, T. L. (1994). Cholesterol homeostasis. Modu- 1043, 123–128. Rodriguez-Lafrasse, C., et al. (1994). Free sphingoid bases in tissues lation by amphiphiles. J. Biol. Chem. 269, 29371–29374. from patients with type C Niemann–Pick disease and other Lange, Y., Strebel, F., and Steck, T. L. (1993). Role of the plasma lysosomal storage disorders. Biochim. Biophys. Acta 1226, 138– membrane in cholesterol esterification in rat hepatoma cells. 144. J. Biol. Chem. 268, 13838–13843. Roelink, H., et al. (1995). Floor plate and motor neuron induction Liscum, L., and Faust, J. R. (1989). The intracellular transport of by different concentrations of the amino-terminal cleavage prod- low density lipoprotein-derived cholesterol is inhibited in Chi- uct of sonic hedgehog autoproteolysis. Cell 81, 445–455. nese hamster ovary cells cultured with 3-beta-[2-(diethyl- Roessler, E., et al. (1996). Mutations in the human Sonic hedgehog amino)ethoxy]androst-5-en-17-one. J. Biol. Chem. 264, 11796– gene cause holoprosencephaly. Nat. Genet. 14, 357–360. 11806. Roessler, E., and Muenke, M. (1998). Holoprosencephaly: A para- Liscum, L., and Munn, N. J. (1999). Intracellular cholesterol trans- digm for the complex genetics of brain development. J. Inherit. port. Biochim. Biophys. Acta 1438, 19–37. Metab. Dis. 21, 481–497. Loftus, S. K., et al. (1997). Murine model of Niemann–Pick C Roff, C. F., et al. (1991). Type C Niemann–Pick disease: Use of disease: Mutation in a cholesterol homeostasis gene. Science hydrophobic amines to study defective cholesterol transport. 277, 232–235. Dev. Neurosci. 13, 315–319. Marigo, V., et al. (1996). Biochemical evidence that patched is the Roux, C., and Aubry, M. (1966). Action teratogene chez le rat d’un Hedgehog receptor. Nature 384, 176–179. inhibiteur de la synthese du cholesterol, le AY 9944. C. R. Matise, M. P., et al. (1998). Gli2 is required for induction of floor Seances Soc. Biol. Fil. 160, 1353–1357. plate and adjacent cells, but not most ventral neurons in the Siebert, J. R., et al. (1990). “Holoprosencephaly: Overview and mouse central nervous system. Development 125, 2759–2770. Atlas of Cases.” Wiley–Liss, New York.

Sisson, J. C., et al. (1997). Costal2, a novel kinesin-related protein type C disease: Current strategy from an experience of 37 in the Hedgehog signaling pathway. Cell 90, 235–245. pregnancies at risk. Am. J. Hum. Genet. 51, 111–122. Stone, D. M., et al. (1996). The tumour-suppressor gene patched Watari, H., et al. (1999). Niemann–Pick C1 protein: Obligatory encodes a candidate receptor for Sonic hedgehog. Nature 384, roles for N-terminal domains and lysosomal targeting in choles- 129–134. terol mobilization. Proc. Natl. Acad. Sci. USA 96, 805–810. Stone, D. M., et al. (1999). Characterization of the human suppres- Watari, H., et al. (2000). NPC1-containing compartment of human sor of fused, a negative regulator of the zinc-finger transcription granulosa-lutein cells: A role in the intracellular trafficking of factor Gli. J. Cell Sci. 112, 4437–4448. cholesterol supporting steroidogenesis. Exp. Cell Res. 255, 56–66. Yamada, T., et al. (1993). Control of cell pattern in the neural tube: Tanabe, Y., and Jessell, T. M. (1996). Diversity and pattern in the Motor neuron induction by diffusible factors from notochord and developing spinal cord. Science 274, 1115–1123. floor plate. Cell 73, 673–686. Tint, G. S., et al. (1994). Defective cholesterol biosynthesis associ- Yoshikawa, H. (1991). Effects of drugs on cholesterol esterification ated with the Smith–Lemli–Opitz syndrome. N. Engl. J. Med. in normal and Niemann–Pick type C fibroblasts: AY-9944, other 330, 107–113. cationic amphiphilic drugs and DMSO. Brain Dev. 13, 115–120. Underwood, K. W., et al. (1996). Quantitative analysis of hydrophobic amine inhibition of intracellular cholesterol transport. J. Received for publication December 14, 1999 Lipid Res. 37, 1556–1568. Revised March 31, 2000 Vanier, M. T., et al. (1992). Prenatal diagnosis of Niemann–Pick Accepted April 21, 2000