Abstract
The PATCHED (PTC) gene encodes a Sonic hedgehog (Shh) receptor and a tumor suppressor protein that is defective in basal cell nevus syndrome (BCNS). Functions ofPTC were investigated by inactivating the mouse gene. Mice homozygous for the ptc mutation died during embryogenesis and were found to have open and overgrown neural tubes. Two Shh target genes, ptc itself and Gli, were derepressed in the ectoderm and mesoderm but not in the endoderm. Shh targets that are, under normal conditions, transcribed ventrally were aberrantly expressed in dorsal and lateral neural tube cells. Thus Ptc appears to be essential for repression of genes that are locally activated by Shh. Mice heterozygous for the ptc mutation were larger than normal, and a subset of them developed hindlimb defects or cerebellar medulloblastomas, abnormalities also seen in BCNS patients.
The human PTC gene is a tumor suppressor and developmental regulator (1). Some patients with BCNS have germline mutations in PTC and are at increased risk for developmental defects such as spina bifida and craniofacial abnormalities, basal cell carcinoma of the skin, and brain tumors (2). PTC mutations also occur in sporadic basal cell carcinomas (1), which generally have both copies of PTC inactivated.
In the fruit fly Drosophila, Ptc is a key component of the Hedgehog (Hh) signaling pathway, which controls cell fate determination during development (3). Hh protein, secreted from localized regions, antagonizes the actions of its apparent receptor, Ptc, in nearby cells (4). In the absence of a Hh signal, Ptc represses transcription of multiple target genes, includingptc itself, wingless (a Wnt gene), and the transforming growth factor β–related genedecapentaplegic. In flies, ptc mutations cause derepression of target genes, cell fate changes, and excessive growth in some tissues (5). Hh induces a high level ofptc transcription by inhibiting the function of Ptc protein, so paradoxically an abundance of ptc transcript is an indicator of a low level of Ptc function. Vertebrate ptcexpression is also regulated by Hh proteins (6), which can bind directly to Ptc (7).
The role of the Hh-Ptc pathway in skin cancer has been established by BCNS studies and with a mouse model (8), but less is known about the brain tumors associated with BCNS. About 3% of BCNS patients develop medulloblastomas (9), cerebellar tumors that usually arise in young children and have a mortality rate of ∼50% (10). ptc mutations have been detected in sporadic medulloblastomas (11), but this tumor type is rare and there are few clear animal models (12), so much remains to be learned about its origins and biology.
To study the roles of ptc in development and in tumorigenesis, we constructed micelacking ptc function. By homologous recombination, part of ptc exon 1 (including the putative start codon) and all of exon 2 were replaced withlacZ and a neomycin resistance gene (Fig.1) (13). Protein made from any alternative ATG codon would lack the first proposed transmembrane domain, flipping the orientation of the protein in the membrane. Three independent embryonic stem cell clones were used to make chimeras that were bred to B6D2F1 animals to generate heterozygous mice on a mixed background. Interbreeding of heterozygotes produced no homozygous animals among 202 offspring examined. Analysis of embryos from timed matings suggested that ptc –/– embryos die between embryonic day (E) 9.0 and E10.5, with the first gross phenotypes appearing by E8. In ptc –/– embryos, the neural tube failed to close completely and was overgrown in the head folds, hindbrain, and spinal cord (Fig.2, A to C). Embryonic lethality may have been due to abnormal development of the heart (Fig. 2B).
Generation of the ptc mutation. (A) The ptc mutant allele was generated by homologous recombination between the KO1 targeting vector andptc. External probe A detected a 3′ Eco RV polymorphism on blots and probe B detected a 5′ Sac I polymorphism. Exons are numbered. B, Bam HI; E, Eco RI; RV, Eco RV; S, Sac I; X, Xho I; neo, neomycin resistance gene; TK, thymidine kinase gene; WT, wild type. (B) Transmission of theptc allele through the germ line was confirmed by Southern blot (upper panel) and a PCR genotyping assay (lower panel). PCR primers are indicated as arrows in (A). Because the homozygous mutant embryos were being resorbed, there was much less yolk sac DNA in the −/− lanes.
Germ layer–specific derepression of Shh target genes in ptc –/– embryos. (A andB) Lateral views of E8.25 wild-type (WT) (A) and ptc –/–(B) embryos. The headfolds are overgrown in the mutant (white arrows) and the heart is not properly formed (red arrows). (C) Lateral views of E8.75 ptc +/– andptc –/– embryos stained with X-Gal (28). (D through G) Transverse sections through E8.75 ptc +/– (D and F) andptc –/– (E and G) embryos stained with X-Gal (D and E) or hybridized with a digoxigenin-labeled Gliprobe (29) (F and G). Both lacZ andGli were derepressed in the ectoderm and mesoderm but not in the endoderm (arrows). In (A) and (B), anterior is to the left and dorsal is up. In (C), anterior is up and dorsal is to the right. In (D) to (G), dorsal is up. Magnifications: (A) and (B), ∼×16; (C), ∼×20; (D) to (G), ∼×125.
In Drosophila Ptc protein inhibits ptctranscription. By inhibiting Ptc function, Hh increases production of Ptc, which may then bind available Hh and limit the range or duration of effective Hh signal (14). Hh signaling also posttranscriptionally regulates the zinc finger protein cubitus interruptus (ci) (15). In vertebrates, Sonic hedgehog (Shh) signaling induces transcription of both ptc and aci homolog, Gli (6, 16).Derepression of ptc and Gli inptc –/– mice should therefore reveal where Ptc is normally active.
The expression of ptc and Gli was greatly increased in ptc –/– embryos. Inptc +/– mice, expression of the lacZgene fused to the first ptc exon during targeting accurately reported the pattern of ptc transcription (Fig. 2, C and D). In ptc –/– embryos, expression ofptc-lacZ was extensively derepressed starting at about E8.0 in the anterior neural tube and spreading posteriorly by E8.75 (Fig. 2, C and E). Derepression was germ layer–specific: bothptc-lacZ and Gli were expressed throughout the ectoderm and mesoderm, but not in the endoderm (Fig. 2, D to G).ptc expression may be excluded from the endoderm so that Shh can signal the endoderm to the mesoderm (17). A differential requirement for Ptc may distinguish the germ layers.
As revealed by ptc mutants, an early site of Ptc activity is the neural tube, where Shh and Ptc act antagonistically to determine cell fates. Shh induces the floor plate and motor neurons in the ventral neural tube (18). These cell types fail to form inShh mutants (19). Large amounts of Shh produced by the notochord may induce floor plate by completely inactivating Ptc (18). If so, elimination of ptc function might cause floor plate differentiation throughout the neural tube. Prospective floor plate cells transcribe the forkhead transcription factor HNF3β first and then Shh itself (18). In E8.5 ptc mutants, transcription ofHNF3β and Shh was expanded dorsally (Fig.3, A to C). Ectopic Shhexpression was most extensive in the anterior, where transcripts could be detected throughout the neurepithelium (Fig. 3, B and C). Cells in this region were in a single layer with basal nuclei, like floor plate cells that are normally restricted to the ventral midline (Fig. 3, D and E). Expression of the lateral neural tube marker Pax6(20) was completely absent from ptc mutant embryos, suggesting that only ventral, and not ventrolateral, cell fates are specified (Fig. 3, F and G).
Ventralization of the neural tube inptc –/– embryos. (A) Lateral view of E8.5 ptc +/– andptc –/–embryos hybridized with aHNF3β probe. Expression is expanded dorsally in the mutant. (B and C) Transverse sections through the hindbrain of E8.5 wild-type (B) andptc –/– (C) embryos hybridized with35S-labeled Shh probe (8).Shh is expressed in the floor plate (fp) and notochord (nc) of the wild-type embryo and is greatly expanded in the ptcmutant. g, gut. (D and E) Hematoxylin and eosin stained transverse sections through the hindbrain of wild-type (D) andptc –/– (E) E8.5 embryos. Bottle-shaped cells with basal nuclei are indicated by arrows. (F andG) Transverse sections through E8.5 wild type (F) and ptc –/– (G) embryos hybridized with aPax6 probe show the absence of expression in theptc +/− mutant. (H) Dorsal view of E8.25 to E8.5 embryos hybridized with a Pax3 probe. Because of the kinking in the neural tube, the ptc –/–embryo is curled on itself. Weak Pax3 expression is seen in the posterior dorsal neural tube of theptc –/– embryo (bottom, arrow). (Iand J) Transverse sections through E8.5 wild-type (I) andptc –/–(J) embryos hybridized with a Pax3 probe. Pax3 is expressed in the dorsal neural tube (nt) and dermamyotome (dm) in the wild-type but is present only in a small dorsal domain of the mutant neural tube; s, somite. (K and L) Lateral views of E9 wild-type (K) and E8.5 ptc −/− (L) embryos hybridized with an erb-b3 probe. Staining is seen in migrating neural crest in the head and somites of wild type but not mutant embryos (red arrows). Weak staining in the head, heart, and gut (black arrows) is background or non–neural crest related. (M) Lateral view of wild-type (top) andptc –/– (bottom) embryos hybridized with anNkx2.1 probe. The body of the mutant is twisted.Nkx2.1 expression is limited to the anterior but is expanded dorsally in the mutant. (N) Lateral view of E8.5ptc +/– and ptc –/–embryos hybridized with a hoxb1 probe. Loss of expression in rhombomere four is indicated by the asterisks. In all transverse sections, dorsal is up. In (A), (K), (L), and (N), anterior is up and dorsal is to the right. In (H) and (M), anterior is to the left. Magnifications: (A), (H), (K), and (L), ∼×16; (B) and (C), ∼×100; (D) and (E), ∼×364; (F), (G), (I), and (J), ∼×160; (M) and (N), ∼×11.
In principle, dorsalizing signals from the surface ectoderm (21) could confer dorsal cell fates even in the absence ofptc function. In E8-E9 ptc homozygotes, the dorsal neural tube marker Pax3 was not expressed in the anterior neural tube but was transcribed in a very small region at the dorsalmost edge of the posterior neural tube (Fig. 3, H to J). In addition, erb-b3 transcription, which marks migratory neural crest cells (Fig. 3K) (22), was not detected in the somites of ptc mutants (Fig. 3L). We conclude that only limited dorsal fate determination occurs in the absence of ptc. Bone morphogenetic protein (BMP) signals appear to maintain dorsal gene expression (21) so either ptc is required for BMPs to work or BMP signaling is ineffective in most cells expressing Shh targets.
Ventralization of the neural tube in ptc mutants occurred without affecting cell identity along the rostrocaudal axis. In ptc –/– embryos, cells in the anterior neural tube expressed the forebrain marker Nkx2.1(23), and cells in the spinal cord transcribedhoxb1 (24) (Fig. 3, M and N). hoxb1was not transcribed in the fourth rhombomere of ptc mutants (Fig. 3N). This may reflect a transformation of hindbrain cells to floor plate, because hoxb1 is excluded from the midline of wild-type embryos. Conversely, in the anterior, Nkx2.1expression was expanded dorsally in mutants compared with wild-type embryos (Fig. 3M).
The ptc +/− mice had features in common with BCNS patients: the mice were larger than their wild-type littermates [30.72 ± 3.83 g (average ± SD;n = 29) compared with 26.54 ± 2.51 g (n = 39) at 2 to 3 months; P = 0.000001 by t test], a small fraction (3 of 389 mice examined) had hindlimb defects such as extra digits or syndactyly (Fig.4A) or obvious soft tissue tumors (1 of 243), and many developed brain tumors.
Skeletal abnormalities and medulloblastomas in ptc +/– mice. (A) Alcian blue and Alizarin red stained hindlimb from aptc +/– mouse (30). The preaxial digit is duplicated (arrows). (B and C) Dorsal views of brains from wild-type (B) and ptc +/–(C) mice. Anterior is up. In the posterior wild-type brain, the colliculi (col) are present as distinct bumps between the cortex (cor) and cerebellum (ce). In the ptc +/– mouse, a massive medulloblastoma (mb, outlined in red) grew over the colliculi and normal cerebellum, which can no longer be seen. The olfactory bulbs were removed. (D and E) Hematoxylin- and eosin-stained section through human (D) and mouse (E) medulloblastomas. The tumor cells are small with dark, carrot-shaped nuclei (arrows) and form nodules with no apparent orientation. (F) Synaptophysin immunoreactivity in a mouse medulloblastoma (26). Synaptophysin staining (brown) is seen in some processes (arrows). Nuclei are purple. Magnifications: (D) and (E), ∼×300; (F), ∼×500.
Of 243 ptc +/– mice that were between the ages of 2 and 9 months and were not killed for other studies, 18 died or were killed because of sickness. No wild-type littermates died. Ten of the affected heterozygotes were autopsied, and eight were found to have large growths in the cerebellum that resembled medulloblastomas (Fig. 4, B and C). Human medulloblastomas are believed to arise from a “primitive neurectodermal” cell type (25). They are most common in children, can be metastatic or nonmetastatic, and can have glial and neuronal properties. The histology of tumors fromptc +/− mice was similar to that of human medulloblastoma: tumor cells were small, with dark carrot-shaped nuclei and little cytoplasm (Fig. 4, D and E), and although a subset expressed neurofilament protein and synaptophysin (Fig. 4F) (26), the majority of cells appeared undifferentiated. Of the two autopsied animals without apparent medulloblastomas, one had a large tumor growing out of its rib muscle and the other died for unknown reasons. Medulloblastomas and soft tissue tumors were also observed in ptc +/– mice maintained on an inbred 129SV background: 6 of 27 had obvious medulloblastomas, 2 of 27 had soft tissue tumors, and 3 of 27 died but were not examined.
The ptc and Gli genes were strongly transcribed in the brain tumors but not in surrounding tissue (Fig.5, A and B; n = 3 of three tumors examined). There was no detectable increase inShh expression (Fig. 5C). To assess the incidence of medulloblastomas, brains from 47 asymptomaticptc +/– mice were randomly collected and stained with X-Gal. Nine brains contained medulloblastomas that were easily recognized by their disorganized morphology and intenseptc-lacZ expression (Fig. 5D). Medulloblastomas were observed in 1 of 12 (8.3%) ptc +/− mice at 5 weeks of age, 1 of 12 (8.3%) mice at 9 to 10 weeks, and 7 of 23 (30.4%) mice at 12 to 25 weeks. Tumors can therefore arise as early as 5 weeks after birth but increase in severity and frequency as the animal ages.
Derepression of ptc andGli expression in medulloblastomas fromptc +/– mice. (A to C) Semi-adjacent sections through a tumor in the cerebellum of aptc +/– mouse hybridized with 35S- labeled probes to ptc (A), Gli (B), andShh (C). ptc and Gli transcripts are abundant in the tumors (asterisks) compared with nearby cerebellar tissue (arrows). No Shh was detected in the tumor. (D) ptc +/– cerebellum (ce) and tumor (mb) stained with X-Gal (28). Anterior is to the left. Derepression of ptc expression in the medulloblastoma is reflected in the large amount of X-Gal staining. (E) Surface staining in regions (arrows) of ptc +/–cerebellum contrast with absence of β-galactosidase activity in most folia (asterisk). (F) Sagittal section through the cerebellum in (E). X-Gal staining nuclei (arrow) accumulated superficial to the molecular layer (ml), where stained nuclei are not normally seen. In unaffected regions of the cerebellum, X-Gal staining was seen in scattered cells of the molecular layer, strongly in the Purkinje cell layer (pcl), and weakly in the granule cell layer (gl). (G) ptc expression was examined in total RNA (15 μg) from wild-type (WT) andptc +/– cerebellums with a probe (M2-2) (6) that detects exons downstream of the lacZand neo insertions. Actin mRNA was used as an RNA loading control. The ptc +/– mice had ∼50% decrease in ptc transcripts. Magnifications: (A) to (C), ∼×25; (D) and (E), ∼×5; (F), ∼×90.
We looked for changes in ptc-lacZ expression that might reflect early stages of tumorigenesis. At all stages examined, about half of the animals [50% at 5 to 10 weeks (n = 24), 56.5% at 12 to 25 weeks (n = 23)] exhibited regions of increased X-Gal staining on the surface of the cerebellum (Fig. 5E). These regions were usually lateral and often extended down into the fissures separating the folia (Fig. 5, E and F). The mouse medulloblastomas may arise from these cells, which are superficial to the molecular layer of the cerebellum (Fig. 5F). During fetal development, prospective cerebellar granule cells proliferate in the external granule layer (EGL), the outermost layer of the cerebellum. Granule cells then leave and migrate past the Purkinje cells to form the internal granule cell layer of the adult animal, gradually depleting the EGL. The remnants of the fetal EGL have been proposed to be a source of human medulloblastoma progenitors, a hypothesis consistent with the higher frequency of these tumors in children (27).
The abundance of cerebellar ptc transcripts was reduced by about 50% in the ptc +/– mice compared with wild-type littermates (Fig. 5G). This reduction could lead to ectopic expression of Shh target genes and to uncontrolled cell proliferation. Brain tumors might arise from Ptc haploinsufficiency alone, from additional mutations in the second ptc allele, or from a combination of ptc mutations with mutations in other tumor suppressor loci. We have not observed basal cell carcinomas in ptc +/– mice, perhaps because somatic inactivation of the second ptc gene is required as it is in human basal cell carcinomas.
Our analysis has revealed that Ptc controls growth and pattern formation in early neural development and in the adult cerebellum. Autoregulation of ptc occurs in vertebrates as it does inDrosophila, and the balance between Hh and Ptc activities appears critical for normal development. The importance of Ptc dosage is emphasized by the phenotype of the ptc +/–mice, which develop a tumor type observed in the corresponding human cancer predisposition syndrome. Medulloblastoma is a common childhood brain tumor and the prognosis remains grim. The Hh-Ptc pathway may provide new diagnostic tools and new insights into tumorigenesis that can be directed toward potential therapies.
↵* To whom correspondence should be addressed. E-mail: scott{at}cmgm.stanford.edu
