Teratogen-Mediated Inhibition of Target Tissue Response to Shh Signaling

See allHide authors and affiliations

Science  05 Jun 1998:
Vol. 280, Issue 5369, pp. 1603-1607
DOI: 10.1126/science.280.5369.1603


Veratrum alkaloids and distal inhibitors of cholesterol biosynthesis have been studied for more than 30 years as potent teratogens capable of inducing cyclopia and other birth defects. Here, it is shown that these compounds specifically block the Sonic hedgehog (Shh) signaling pathway. These teratogens did not prevent the sterol modification of Shh during autoprocessing but rather inhibited the response of target tissues to Shh, possibly acting through the sterol sensing domain within the Patched protein regulator of Shh response.

A striking aspect ofShh function is its role in developmental patterning of the head and brain, as revealed in Shh −/− mouse embryos by the occurrence of severe holoprosencephaly (HPE) (1). HPE is characterized by development of the prosencephalic derivatives as a single undivided vesicle that consists of the fused remnants of the dorsal telencephalic lobes, with an undivided eye field and an absence of ventral forebrain structures such as the optic stalks, the optic chiasm, and the pituitary (1,2). Externally, severe HPE is characterized by an absence of midline facial structures and development of a proboscis consisting of fused nasal chambers at a location overlying a cyclopic eye. Loss-of-function mutations at the human Shh locus are associated with a milder and more variable form of HPE that is inherited in autosomal dominant fashion, indicative of haploinsufficiency at the human Shh locus (3).

Hedgehog (Hh) proteins undergo an intramolecular autoprocessing reaction that entails internal cleavage and covalent addition of cholesterol to generate the mature signaling molecule (4-6). Given this critical role for cholesterol modification in the biogenesis of Hh proteins, it is noteworthy that certain perturbations of cholesterol homeostasis cause HPE. For example, HPE is induced in rat pups exposed during gestation to distal inhibitors of cholesterol biosynthesis such as triparanol, AY9944, or BM15.766 (7-9). Milder forms of HPE are observed in 5% of patients with Smith-Lemli-Opitz Syndrome, which is thought to be caused by a defect late in the cholesterol biosynthetic pathway (10). HPE is also observed in mouse embryos deficient in megalin, a member of the low-density lipoprotein (LDL) receptor family that is expressed in embryonic neuroectoderm and binds and internalizes LDL (11). Finally, HPE is induced in lambs born to pregnant ewes that consume Veratrum californicum, and the teratogenic effects of this plant have been traced to the alkaloids cyclopamine and jervine (12). These two closely related compounds resemble cholesterol in structure, and jervine acts as a distal inhibitor of cholesterol biosynthesis (13).

As seen in Fig. 1, B to E, exposure of chick embryos to jervine at the intermediate to definitive streak stage (14) induced external malformations characteristic of HPE, with a variable extent of loss of midline structures and consequent approximation and fusion of paired lateral structures such as the mandibular and maxillary processes as well as the optic vesicles and olfactory processes. We circumvented the inherent variability of these in ovo treatments by using an explant assay, which facilitates a more uniform application of these hydrophobic compounds (15). Medial neural plate with notochord attached was dissected from a region just rostral to Hensen's node (Fig. 1F), a level where the notochord expresses Shh (16) but the neural plate does not yet express floor plate cell (HNF3β) or motor neuron (Isl1) markers (17). Induction of these cell types depends on Shh signaling both in vivo and in vitro (1, 18), and, as seen in Fig. 1G, both HNF3β and Isl1 were induced within the explanted neural plate tissue after a 40-hour incubation. Induction of HNF3β and Isl1 was inhibited by 4 μM jervine (Fig. 1I). At 0.5 μM, jervine still blocked expression of HNF3β, but expression of Isl1 was maintained or enhanced (Fig. 1H; see also Fig. 3). Partial and complete inhibition of Shh signaling was also obtained with increasing concentrations of AY9944, triparanol, and cyclopamine (19). In contrast, the structurally related but not teratogenic alkaloid tomatidine (20) did not block induction of HNF3β and Isl1, even at concentrations an order of magnitude higher than the inhibitory concentration of jervine (Fig. 1J).

Figure 1

Jervine induces holoprosencephaly and blocks endogenous Shh signaling. (A) Scanning electron micrograph of external facial features of an untreated embryo. (B to E) Embryos were exposed to 10 μM jervine (14) with variable loss of midline tissue and resulting fusion of the paired, lateral olfactory processes (Olf), optic vesicles (Opt), and maxillary (Mx) and mandibular (Mn) processes. Complete fusion of the optic vesicles and lenses (L) results in true cyclopia (E). (F) Midline tissue was removed from stage 9 to 10 chick embryos at a level just rostral to Hensen's node (white dashed line) and further dissected (black dashed lines) to yield an explant containing an endogenous source of Shh signal (notochord) and a responsive tissue (neural plate ectoderm) (15). (G) After 2 days of culture in a collagen gel matrix, expression of floor plate cell [HNF3β, rhodamine (red)] and motor neuron [Isl1, fluorescein isothiocyanate (FITC) (green)] markers is induced in untreated neural ectoderm or (J) in explants treated with the nonteratogenic alkaloid tomatidine (TOM) (50 μM). (H) Intermediate doses of jervine (JER) (0.5 μM) block induction of HNF3β,while permitting induction of Isl1 (see text). (I) Higher doses of jervine (4.0 μM) fully inhibit HNF3β and Isl1 induction.

To examine the potential effects of these teratogens on Shh processing, we used HK293 cells carrying a stably integrated construct for expression of Shh under ecdysone-inducible control (21). The Shh protein was efficiently processed (Fig.2A, lanes 2 and 3), and addition of jervine, cyclopamine, tomatidine, AY9944, or triparanol during the 24-hour induction period did not diminish production of Shh-Np, the processed NH2-terminal product, or induce accumulation of unprocessed precursor [relative molecular mass (M r) of 45 kD], even at doses 6- to 50-fold higher than those required to completely inhibit Shh signaling (Fig.2A, lanes 4 to 7 and 10 to 15). All of the NH2-terminal cleavage product generated in the presence of these compounds was detected in cell lysates, not in the culture medium, and had the same electrophoretic mobility as cholesterol-modified Shh-Np(compare with lanes 8 and 9, Fig. 2A), consistent with the presence of a sterol adduct in the NH2-terminal cleavage product. In addition, chick embryos treated with jervine after floor plate induction displayed normal apical localization of Shh protein within floor plate cells (22), indicative of normal secretion and sorting of intracellular Shh-Np.

Figure 2

Teratogens do not inhibit Shh autoprocessing. (A) Stably transfected HK293 cells containing an ecdysone-inducible Shh expression construct (lane 1, uninduced) were treated by addition of muristerone A (lanes 2 to 7 and 10 to 15) either alone (lanes 2 and 3) or in combination with jervine (lanes 4 and 5), cyclopamine (lanes 6 and 7), tomatidine (lanes 10 and 11), AY9944 (lanes 12 and 13), or triparanol (lanes 14 and 15) (21). Shh from control (lane 3) or drug-treated (lanes 4 to 7 and 10 to 15) cell lysates is efficiently processed with no detectable accumulation of precursor protein (M r = 45 kD). The processed NH2-terminal product (Shh-Np) is cell associated and migrates faster than unprocessed Shh-N protein (lane 8) from the medium of cultured cells transfected with a construct carrying an open reading frame truncated after Gly198(both Shh-Np and Shh-N are loaded in lanes 9 and 16), indicating that Shh-Np from treated cells likely carries a sterol adduct. The slower migrating species resulting from tomatidine treatment is ∼1.9 kD larger, suggestive of a minor inhibition of signal sequence cleavage (see asterisk; lanes 10 and 11). Immunoblotted actin for each lane is shown as a loading control. (B) Coomassie blue–stained SDS-polyacrylamide gel showing in vitro autocleavage reactions of the bacterially expressed His6Hh-C protein (∼29 kD) incubated for 3 hours at 30°C in the absence of sterols (lane 1), with 50 mM dithiothreitol (DTT) (lane 2), 12 μM cholesterol (lane 3), 12 μM 7-dehydrocholesterol (lane 4), 12 μM desmosterol (lane 5), 12 μM lathosterol (lane 6), 12 and 350 μM lanosterol (lanes 7 and 8, respectively), and 12 and 350 μM muristerone (lanes 9 and 10, respectively). The 27-carbon cholesterol precursors (lanes 4 to 6) stimulate His6Hh-C autoprocessing as efficiently as cholesterol (lane 3). Lanosterol (lanes 7 and 8) and muristerone (lanes 9 and 10) do not stimulate autoprocessing above background (lane 1). The NH2-terminal product migrates as a ∼7-kD species (lane 2) when generated in the presence of 50 mM DTT and as a ∼5-kD species (lanes 3 to 6) with a sterol adduct.

Because the plant alkaloids resemble cholesterol in structure, including the presence of a 3β hydroxyl, their effects were tested in a cholesterol-dependent in vitro autoprocessing reaction (5). None of these compounds could replace cholesterol or inhibit its stimulatory effect (23). In contrast, cholesterol could be replaced efficiently in the in vitro reaction by desmosterol and 7-dehydrocholesterol (Fig. 2B, lanes 4 and 5), the major precursors that accumulate in cells treated with triparanol and AY9944 (9). Other 27-carbon cholesterol precursors, including lathosterol (but not lanosterol, a 30-carbon cholesterol precursor), could participate in the reaction (Fig. 2B, lanes 6 to 8). These and other observations (24) suggest that many, possibly all, 27-carbon sterol intermediates in the biosynthetic pathway are potential adducts in the autoprocessing reaction, and this may account for the unimpaired efficiency of processing in the presence of distal synthesis inhibitors.

To examine the possibility that teratogens affect the response of target tissues to Shh signaling, we used an intermediate neural plate explant (15) (Fig. 3A) that responds to recombinant Shh-N protein in a concentration-dependent manner (18). The dorsal marker Pax7 was repressed at low concentrations (2 nM, Fig. 3, B and C) (25), and the ventral markers Isl1 and HNF3β were induced at progressively higher concentrations (6.3 nM, Fig. 3D, and 25 nM, Fig. 3E) (18). These teratogens completely blocked the repression of Pax7 (at 2 nM Shh-N, Fig. 3, F to I) and the induction of Isl1 and HNF3β (at 25 nM Shh-N, Fig. 3, P to S). A complete inhibition of the response to 25 nM Shh-N required doses of teratogenic compounds twofold to fourfold higher than those required to completely block the 2 nM response. In addition, at a drug concentration half of that required for complete inhibition of 25 nM Shh-N treatment, Isl1 expression was retained or expanded (Fig. 3, K to N). Inhibition of the response to higher concentrations of Shh-N thus requires higher drug concentrations, and, at a fixed concentration of Shh-N, distinct degrees of pathway activation can be produced by distinct inhibitor concentrations. Tomatidine, in contrast, did not inhibit Pax7 repression (Fig. 3J) and only partially inhibited HNF3β and Isl1 induction (Fig. 3, O and T), even at concentrations 100- to 200-fold higher than those required for complete inhibition by jervine and cyclopamine.

Figure 3

Teratogens inhibit response of neural ectoderm to recombinant Shh-N protein. (A) Intermediate neural plate ectoderm, free of notochord and other tissues, was dissected as shown (dashed lines) from stage 9 to 10 chick embryos at a level just rostral to Hensen's node (see Fig. 1F). (B) Explanted intermediate neural plate tissue cultured in a collagen gel matrix for 20 hours expresses the dorsal marker Pax7 (FITC) but not the floor plate marker HNF3β (rhodamine). (C) Addition of recombinant, purified Shh-N at 2 nM suppresses Pax7 expression. (D) Markers of motor neuron (Isl1, FITC) and floor plate cell (HNF3β,rhodamine) fates are induced upon explant culture for 40 hours in the presence of 6.3 nM Shh-N. (E) At 25 nM Shh-N, HNF3β expression expands at the expense of Isl1 expression. The repression of Pax7 expression by 2 nM Shh-N is inhibited by (F) 0.5 μM AY9944 (AY), (G) 0.25 μM triparanol (TRI), (H) 0.13 μM jervine, and (I) 0.063 μM cyclopamine (CYC) but not by (J) 50 μM tomatidine. (K to N) Induction of HNF3β is reduced, whereas induction of Isl1 at 25 nM Shh-N is maintained or expanded at intermediate concentrations of AY9944 (1.0 μM) (K), triparanol (0.25 μM) (L), jervine (0.25 μM) (M), and cyclopamine (0.13 μM) (N). (O) Tomatidine at 25 nM displays a slight inhibitory effect with a decrease in HNF3β expression and an increase in the number of Isl1-expressing cells. (P to S) HNF3β and Isl1 induction is completely blocked at inhibitory doses twofold higher than those in (K) to (N). (T) Tomatidine at 50 μM markedly reduces HNF3β induction and enhances Isl1 induction.

In a test of the specificity of these compounds, explants from the ventral neural plate (Fig. 4A) responded to BMP7 protein by formation of neural crestlike migratory cells that express the HNK-1 surface antigen (26) (compare Fig. 4, B and C), even in the presence of jervine at 10 μM (Fig. 4D). This concentration is 20-fold higher than that required for a complete block of Shh-N signaling. Similar results were obtained with tomatidine and cyclopamine (27).

Figure 4

Jervine does not inhibit the response of neural ectoderm to BMP7. (A) Ventral neural plate ectoderm was dissected as shown (dashed lines) from stage 9 to 10 chick embryos at a level just rostral to Hensen's node (see Fig. 1F). (B) Ventral neural plate explants cultured for 24 hours in a collagen gel matrix do not give rise to HNK-1–positive migratory cells unless BMP7 (100 ng/ml) is added (C). (D) Induction of migratory HNK-1–positive cells by BMP7 (100 ng/ml) is not inhibited by the presence of 10 μM jervine (C and D, explant borders outlined by white dashed lines) nor by the addition of the other plant-derived compounds (10 μM cyclopamine and 50 μM tomatidine) (37).

Given that these teratogens inhibit the response of target tissues to Shh signaling and that some of them affect distal cholesterol biosynthesis (9, 13), it is noteworthy that the Patched (Ptc) protein, which controls the response to Shh signaling within responding tissues (28), contains a sterol sensing domain (SSD). The SSDs of two other proteins, hydroxymethylglutaryl coenzyme A (HMG CoA) reductase and SCAP (SREBP cleavage-activating protein), confer differential responses to high and low concentrations of intracellular sterols; a third SSD-containing protein, NPC1 (Niemann-Pick C1), is proposed to function in intracellular transport. The possibility thus emerges that these teratogens inhibit Shh signaling through the Ptc SSD, either through effects on Ptc protein stability, Ptc-dependent activation of downstream catalytic events, or Ptc-dependent effects on transport, as suggested by analogy to apparent SSD functions in HMG CoA reductase, SCAP, and NPC1, respectively (29).

The teratogenic effects of these compounds cannot simply be due to a reduction of cholesterol synthesis or due to the accumulation of an inhibitory sterol precursor because none of the explant responses to Shh-N were either blocked or restored by a potent proximal inhibitor of cholesterol biosynthesis (30). It is thus important to note that, in addition to its effects on distal cholesterol biosynthesis, AY9944 inhibits cholesterol esterification (31). This property is shared by a group of compounds termed class 2 transport inhibitors (32), which appear to act by reducing the flux of cholesterol and its sterol precursors from the plasma membrane (PM) to the endoplasmic reticulum (ER), thus preventing action of acyl-CoA cholesterol acyltransferase on exogenously delivered cholesterol and causing accumulation of cholesterol biosynthetic precursors. Class 2 sterol transport inhibitors appear to increase the activity of HMG CoA reductase and to stimulate SCAP activity (33). Given the ER localization of these two SSD-containing proteins, it is possible that disruption of sterol transport from PM to ER by class 2 compounds decreases sterol concentrations within intracellular compartments, despite normal or increased concentrations of cellular sterols overall. We have found that the other teratogens studied here also inhibit cholesterol esterification (34) and that other structurally dissimilar class 2 compounds inhibit Shh signaling in our explant assays (35). Because the Ptc protein is found at intracellular locations (36), a teratogen-induced defect in sterol transport could conceivably perturb Ptc function through its SSD. Further studies with these teratogenic compounds may help elucidate the mechanistic roles of Ptc and intracellular transport in the Shh signaling pathway.

  • * Present address: Ontogeny Incorporated, 45 Moulton Street, Cambridge, MA 02138.


Stay Connected to Science

Navigate This Article