Abstract
Activation of the Sonic hedgehog (Shh) signal transduction pathway is essential for normal pattern formation and cellular differentiation in the developing CNS. However, it is also thought to be etiological in primitive neuroectodermal tumors. We adaptedGAL4/UAS methodology to ectopically express full-lengthShh in the dorsal neural tube of transgenic mouse embryos commencing at 10 d postcoitum (dpc), beyond the period of primary dorsal–ventral pattern formation and floorplate induction. Expression of Shh was maintained until birth, permitting us to investigate effects of ongoing exposure to Shh on CNS precursors in vivo. Proliferative rates of spinal cord precursors were twice that of wild-type littermates at 12.5 dpc. In contrast, at late fetal stages (18.5 dpc), cells that were Shh-responsive but postmitotic were present in persistent structures reminiscent of the ventricular zone germinal matrix. This tissue remained blocked in an undifferentiated state. These results indicate that cellular competence restricts the proliferative response to Shhin vivo and provide evidence that proliferation and differentiation can be regulated separately in precursor cells of the spinal cord. Thus, Hedgehog signaling may contribute to CNS tumorigenesis by directly enhancing proliferation and preventing neural differentiation in selected precursor cells.
- Sonic hedgehog
- tumorigenesis
- GAL4
- central nervous system
- proliferation
- differentiation
- Wnt-1
- medulloblastoma
- transgenic mice
Sonic hedgehog (Shh)encodes a secreted glycoprotein that is initially expressed in mesodermal tissues underlying the ventral midline of the murine CNS (Echelard et al., 1993; Marti et al., 1995a). Shh is essential for maintenance of notochord and prechordal mesoderm (Chiang et al., 1996) as well as the induction of floorplate and ventral neuronal populations that form at different positions along the anterior–posterior (AP) axis of the neural tube (Echelard et al., 1993; Roelink et al., 1994, 1995; Marti et al., 1995b; Chiang et al., 1996; Ericson et al., 1997; Jessell and Lumsden, 1997). In addition, there is evidence that Shh signaling may specify oligodendrocyte precursors (Poncet et al., 1996; Pringle et al., 1996). Shh signal transduction is complex (Tabin and McMahon, 1997). The active Shh signal, which is produced by autoprocessing and cholesterol modification (Porter et al., 1996), binds to a receptor complex composed of at least two transmembrane proteins, Patched and Smoothened (Marigo et al., 1996; Stone et al., 1996). Shh binding to Patched is thought to relieve Patched-mediated inhibition of Smoothened activity, resulting in the activation of transcriptional targets by members of the Gli family (Ingham et al., 1991; for review, see Ingham, 1998). Although Patched and Gli-1 appear to be general transcriptional targets in vertebrates, other factors are specific to neural precursor cells, including HNF3β andNkx-2.2 (Dale et al., 1997; Ericson et al., 1997).
In contrast to its roles in neural patterning and differentiation, recent studies have implicated the Hedgehog signaling pathway in proliferation and tumorigenesis. Loss-of-function mutations in humanPATCHED are associated with activation of the Hedgehog signal transduction pathway and promotion of a neoplastic state characterized by proliferating, undifferentiated cell populations (Hahn et al., 1996; Johnson et al., 1996). Of the children with Gorlin’s Syndrome, which is caused by inherited mutations of PATCHED, 3–5% develop medulloblastoma (Vorechovsky et al., 1997). Inactivating mutations of PATCHED have also been found in sporadically occurring medulloblastoma (Raffel et al., 1997) and basal cell carcinoma, and mice heterozygous for targeted mutations ofPatched, in which Shh targets are potentially upregulated, develop cerebellar tumors (Goodrich et al., 1997). Recently,Wechsler-Reya and Scott (1999) provided evidence that Shh is required for granule cell precursor proliferation during cerebellar development, raising the possibility that similar mechanisms are involved during development and tumorigenesis.
Mitogenic effects of Shh have been observed in a number of tissues during development (Fan and Tessier-Lavigne, 1994; Forbes et al., 1996;Huang and Kunes, 1996; Bellusci et al., 1997; Jensen and Wallace, 1997;Oro et al., 1997; Duprez et al., 1998), and misexpression of chickenShh (Echelard et al., 1993), Gli-1 (Hynes et al., 1997), or a dominant-negative form of protein kinase A (dn-PKA), which activates Shh targets (Epstein et al., 1996), all resulted in embryonic CNS hyperplasia. The mechanisms underlying such proliferative effects, however, are poorly understood. To gain insight into this process, we have focused on the mitogenic role of Shh in the developing CNS. Our results show that ectopic activation of Hedgehog signal transduction causes enhanced proliferation, but only at embryonic stages. Thus, factors regulating maturation and cellular competence of CNS precursor cells temporally restrict the proliferative response to Shh in vivo.
MATERIALS AND METHODS
DNA constructs. The plasmid pGaTB and pUAST, encoding full-length GAL4 and a pentamer array of its cognate DNA binding sequence, the upstream activating sequence (UAS), were kindly provided by Drs. A. Brand and N. Perrimon (Harvard Medical School). To generate the transgene pWEXP-GAL4, plasmid pGaTB was digested with HindIII and FspI to release a DNA fragment encoding GAL4, which was cloned intoNruI-digested Wnt-1 expression vector pWEXP-2 (Echelard et al., 1993) (see Fig. 1A). The transgene was purified from vector sequences by digestion with AatII. To generate the reporter transgene pUAS-lacZ, the plasmid XB3 (Echelard et al., 1994) was digested with NotI. The pentamer array of UAS sequences from plasmid pUAST were amplified by PCR primers that incorporated NotI and EagI recognition sequences. Once digested, the PCR products were cloned into the XB3 vector to create pUAS-lacZ (see Fig.1B). The transgene was purified from vector sequences by digestion with SalI before pronuclear injection. To generate the mouse Sonic hedgehog misexpression transgene pWEXP3C-Shh, the full-length cDNA was digested from plasmid p8.1 (Echelard et al., 1993) and cloned into the Wnt-1expression vector pWEXP-3C (Danielian and McMahon, 1996). The transgene was purified from vector sequences by digestion with the restriction endonuclease SalI before microinjection. To generate the transgene pUAS-Shh, a shuttle vector, pUAS-Shuttle, was constructed as follows. The KpnI and BglII fragments of plasmid XB3 were replaced with an oligonucleotide containing an XhoI site. This construct was digested withNotI and KpnI, and anNotI–KpnI upstream fragment of pUAS-lacZ, comprising five copies of UAS, was added, generating plasmid pUAS-Shuttle. Finally, pUAS-Shuttle was digested with XhoI and BglII, and anSalI–BglII fragment of pWEXP-3C was cloned into the vector, creating plasmid pUAS-Shh (see Fig.1C). The transgene was purified from vector sequences by digestion with SalI and BglII before microinjection.
DNA sequencing of the constructs listed above was carried out using both ABI dye terminator and dideoxy chain termination methodologies. The orientation and identity of GAL4 andmShh in constructs pWEXP2-GAL4 and pWEXP3C-Shh, respectively, were confirmed by DNA sequencing using oligonucleotide 882 (5′-TAA GAG GCC TAT AAG AGG CGG-3′), which primes ∼60 bp upstream of the Wnt-1 translational initiation site.
Production and genotyping of transgenic mice.Transgenic mice were generated by microinjection of linear DNA fragments, separated from plasmid vector sequences, into pronuclei of B6CBAF1/J (C57BL/6J × CBA/J) zygotes as described (Echelard et al., 1994). The transgenic line Wnt-1/GAL4/cre-11 resulted from coinjection of the transgenes pWEXP2-GAL4 and pWEXP3C-cre (construction and characterization of the pWEXP3C-cre transgene are described elsewhere).
Founder (G0) transgenic mice were identified by Southern blot of EcoRI-digested genomic DNA and probes forGAL4 (line WEXP2-GAL4) or lacZ(lines UAS-lacZ and UAS-Shh). Subsequent genotyping of UAS-lacZ transgenic embryos or mice by PCR was carried out as described in Echelard et al. (1994). Genotyping of WEXP2-GAL4 and UAS-Shh transgenic embryos or mice used an upstream primer from exon 1 untranslated sequence ofWnt-1 (882-TAAGAGGCCATAAGAGGCGG) and a downstream primer from within GAL4 (1061-ATCAGTCTCCACTGAAGC; product size ∼600 bp) or mouse Shh (930-CTCATAGTGTAGAGACTCCTC; product size ∼600 bp) coding sequences, and the following PCR conditions: 30 sec, 93°C; 30 sec, 55°C, 1 min, 72°C for 40 cycles; then 5 min, 72°C.
Whole-mount histochemistry and skeletal preparation.Analysis of embryos for β-galactosidase activity was carried out as described by Whiting et al. (1991). For analysis of skeletal elements, 18.5 dpc bigenic fetuses were processed as described (Wallin et al., 1994). Photography of processed embryos or fetuses was performed in 80% glycerol/PBS on an Olympus SZH10 microscope using Kodak 64T film. Live embryos or fetuses were photographed in PBS using a 35 mm Nikon camera and daylight film, respectively.
Histological analysis, proliferation studies, and in situ hybridization. For histological analysis, embryos were harvested between 9.5 and 18.5 dpc, dissected in PBS, fixed overnight for 24 hr in Bouin’s fixative, embedded in paraffin, and sectioned (6–8 μm) at forelimb levels of the thoracic spine before staining with hematoxylin–eosin. To analyze proliferation, 50 μg/g of bromodeoxyuridine (BrDU) (Sigma, St. Louis, MO) was injected intraperitoneally into pregnant mothers 3 hr before they were killed at 12.5 and 18.5 dpc. Subsequently, four bigenic embryos and wild-type littermates were fixed in paraformaldehyde and sectioned as above. Dividing cells that had incorporated BrDU were identified using monoclonal IgG (Becton Dickinson, San Jose, CA) and immunoperoxidase staining (Vector Laboratories, Burlingame, CA) using diaminobenzidine (Sigma) or FITC-tyramide (DuPont NEN, Wilmington, DE). Terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) procedure was performed as described (Gavrieli et al., 1992). Both TdT and biotinylated-16-dUTP were from Boehringer Mannheim (Indianapolis, IN).
In situ hybridization on paraffin sections with radiolabeled antisense RNA probes was performed on either paraformaldehyde or Bouin’s fixed tissues according to lab protocols [after Wilkinson (1992); available on request]. Dark-field photomicrographs were collected on a Leitz Orthoplan or Nikon E600 compound microscope using a 35 mm camera and Fuji Velvia film or a SPOT I digital camera.In situ hybridization on frozen sections of paraformaldehyde-fixed tissues with digoxigenin-labeled antisense probes was performed essentially as described in Ma et al. (1997), and photomicrographs were collected on a Nikon E600 compound microscope using a SPOT I digital camera (Diagnostic Imaging). We thank the following investigators for kindly supplying the in situhybridization probes used: M. Scott (Ptc-1) (Goodrich et al., 1996); A. Joyner (Gli-1) (Hui et al., 1994); B. Hogan (HNF3β) (Sasaki and Hogan, 1993); P. Gruss (Pax-6) (Walther and Gruss, 1991) (Pax-3) (Goulding et al., 1993); J. Rubinstein (Nkx-2.2) (Price et al., 1992); R. Johnson (Lmx-1b) (Chen et al., 1998); D. Lachman (Brn-3a) (Theil et al., 1994); L. Roberston (BMP-7) (Lyons et al., 1995); R. Kageyama (HES-1) (Sasai et al., 1992); Genetics Institute, Cambridge, MA (GDF-7) (Storm et al., 1994).
Immunohistochemistry. For immunohistochemistry, embryos were either fresh frozen or fixed between 6 and 24 hr in fresh 4% paraformaldehyde before freezing and cryostat sectioning (15 μm). Antibody against a glutathione S-transferase–Hamster Lmx fusion protein or Sonic hedgehog was generated in rabbits. Rabbit antisera for Isl-1/2, Nkx-2.2, and Lim-3 were the generous gift of T. Jessell (Columbia University, New York). GalC, O4, and PDGFαR monoclonal antibodies were from Boehringer Mannheim. These and mouse monoclonal antibodies against NeuN (Chemicon, Temecula, CA), TuJ1 (BAbCo, Berkley, CA), Pax-7 (Developmental Studies Hybridoma Bank), and O4 IgM were labeled with anti-rabbit Cy3 or anti-mouse IgG or IgM-conjugated Cy2 (Jackson ImmunoResearch Labs, West Grove, PA) before visualization by fluorescence microscopy. Photomicrographs were collected on a Nikon E600 compound microscope and SPOT I digital camera (Diagnostic Instruments).
RESULTS
Gal4/UAS-targeted gene expression in theWnt-1 domain
The Wnt-1 enhancer is well suited for directing gene expression in the roofplate of the spinal cord and was used to overexpress mouse Shh in transgenic mice (D. Rowitch, B. St.-Jacques, and A. McMahon, unpublished observations). However, this resulted in a lethal CNS malformation and precluded maintenance of stable lines. Therefore, on the basis of the work of and Brand and Perrimon (1993) in Drosophila and Ornitz et al. (1991) in mice, we adapted the GAL4/UAS bigenic system for controlled gene expression in the developing murine CNS. Six lines of transgenic mice were generated in which GAL4 was expressed under control of Wnt-1 regulatory sequences (Wnt-1/GAL4) (Fig. 1A) (see Materials and Methods), as judged by whole-mount in situ hybridization (data not shown). These were subsequently crossed with a reporter line in which expression of lacZ was governed by theGAL4 cognate DNA-binding motif, “upstream activating sequence” (Fig. 1B, UAS). When mated toWnt-1/GAL4 hemizygotes, 25% of progeny embryos from three of five lines showed β-galactosidase staining in the Wnt-1pattern (Fig.2A,B), and one of these lines was selected for further study (designated UAS-lacZ). All six Wnt-1/GAL4 founders were then screened against the UAS-lacZ line and one of these was selected for further study because of its relatively high levels of activity (designated Wnt-1/GAL4). Expression oflacZ in double-hemizygous (bigenic) embryos was studied from 8.5 to 18.5 dpc. β-galactosidase staining was first detected at ∼9.0 dpc in a region of the ventral midbrain. This represented a delay of ∼24 hr in the onset of expression compared with previous observations of lacZ under direct control ofWnt-1 regulatory sequences (Echelard et al., 1994). Expression comprising the full Wnt-1 pattern was seen by 10.5 dpc (Fig. 2, compare A and B). Maintenance of β-galactosidase staining was observed at 12.5 dpc (Fig.2C) and 18.5 dpc, at which point roofplate cells could be clearly identified (Fig. 2D).
Schematic illustration of transgenes WEXP-GAL4, UAS-lacZ, and UAS-Shh used in bigenic system for misexpression in the mouse embryonic CNS. A, Plasmid pWEXP-GAL4 comprises full-length GAL4(Brand and Perrimon, 1993) cloned into the WEXP2 expression vector under control of Wnt-1 regulatory sequences (Echelard et al., 1993). B, The reporter transgenic construct pUAS-lacZ used the Wnt-1 minimal promoter (Echelard et al., 1994) and five copies of the UAS (Brand and Perrimon, 1993). C, In plasmid pUAS-Shh, full-length mouse Shh cDNA was cloned into expression vector WEXP3C (Danielian and McMahon, 1996). Wnt-1regulatory sequences were then replaced by five copies of the UAS. Binding sites for oligonucleotide primers used in genotyping the various transgenic lines are indicated (arrows).
The GAL4/UAS system for gene expression in the developing CNS. A–D, Whole-mount histochemical analysis of β-galactosidase activity in transgenic mouse embryos. A, Lateral view showing pattern oflacZ expression under the control ofWnt-1 regulatory sequences (Echelard et al., 1994).B, C, Lateral views of 10.5 and 12.5 dpc bigenic Wnt-1-Gal4 X UAS-lacZ embryos showing expression pattern of lacZ (arrow indicates roofplate expression in the spinal cord). D, Transverse (top) and bisected (bottom) views oflacZ expression in the rostral spinal cord of bigenic fetus at 18.5 dpc. Note staining in roofplate oligodendrocytes that project to the ventricular zone (arrows).E–J, Morphological analysis of wild-type (left) and Wnt-1-GAL4 X UAS-Shh bigenic (right) littermates at 10.5 dpc (E,arrow indicates anterior neural tube defect) and 12.5 dpc (F). G–J, Analysis at 18.5 dpc. Lateral views of wild-type (G) and bigenic fetuses (H). Note tissue mass protruding from midbrain that covers cerebral hemispheres (arrow).I, Dorsal view of bigenic fetus showing hyperplastic spinal cord that protrudes from the back covered by a thin epithelial membrane. Note prominent vasculature and hemorrhage (arrow). J, Dorsal view of skeletal prep of wild-type (left) and bigenic (right) fetuses at 18.5 dpc. Note absence of the membranous skull and dorsal neural arches as well as the splayed open configuration of the vertebral bodies (arrows).
Conditional expression of Shh in the developing spinal cord
To determine whether this bigenic system could also be used to control expression of Shh, the transgenic line UAS-Shh (Fig. 1C) was generated. Of eight founder lines that carried the UAS-Shh transgene, six survived and three transmitted the allele through the germline. One of these promoted expression of Shh in response to GAL4 in the CNS. The UAS-Shh line was viable; hemizygous progeny did not express ectopic Shh. The phenotype of bigenic embryos (hereafter termed Shh-Tg), comprising ventralization of the midbrain and neural hyperplasia, could first be distinguished at 9.5 dpc (data not shown) and was followed extensively from 10.5 to 18.5 dpc (Fig. 2E–J). In contrast toWnt-1-Shh transgenic founder embryos, which display a spectrum of phenotypic severity (Echelard et al., 1993; D. Rowitch, B. St.-Jacques, A. McMahon, unpublished observations), the phenotype of the Shh-Tg embryos was highly reproducible. TheShh-Tg phenotype was judged as moderate compared with the most severe Wnt-1-Shh founders, which failed to develop craniofacial mesenchyme and died at 14–16 dpc. Gross morphological analysis of Shh-Tg fetuses at 18.5 dpc revealed a tissue mass that emanated from the midbrain (Fig. 2H,arrow) and spinal cord tissues with a folded appearance protruding from the back of the animals (Fig. 2I). The membranous skull, the neural arches of the vertebrae, and epiaxial muscle, which normally overlay the brain and spinal cord, were absent (Figs. 2H–J). In addition, there was pronounced blood supply to the dorsal spinal cord (Fig.2I). Whether such effects on mesodermal derivatives are caused by Shh itself or are a consequence of neural hyperplasia remains to be determined.
Increased levels of proliferation in the spinal cord ofShh-Tg mice at embryonic but not fetal stages of development
Histological analysis of 12.5–18.5 dpc Shh-Tg bigenic embryos demonstrated hyperplasia of the dorsal spinal cord and expansion of the ventricular zone (VZ) (hydromyelia) (Fig.3, compare A and Ewith I and M). In principle, this could result from increased levels of proliferation and/or inhibition of programmed cell death. To assess proliferation in wild-type andShh-Tg embryos, mitotically active cells were labeled with BrDU at 12.5, 14.5, 16.5, and 18.5 dpc and identified in sections taken at the forelimb level (Fig.3B,F,J,N; and data not shown). To quantify levels of proliferation at 12.5 dpc, we derived a relative mitotic index (ratio of BrDU-labeled cells in the alar vs basal ventricular zone) in Shh-Tg mice (mean = 9.3, SE = 1.45; n = 4) and wild-type littermates (mean = 3.8, SE = 0.26; n = 4). This indicated that levels of proliferation in 12.5 dpcShh-Tg embryos were approximately twice that of wild-type, reflecting significant (p < 0.001, Student’st test) elevation in the dorsal compartment whereShh was ectopically expressed. At 14.5 dpc, a small population of BrDU-labeled cells were observed in the spinal cord of one of three Shh-Tg embryos, but not in wild-type (data not shown). However, by 18.5 dpc mitotic activity in Shh-Tg neural tissue (n = 3) was not above background wild-type levels (n = 3) (Fig. 3, compare Jand N).
Analysis of spinal cord morphology, rates of precursor cell proliferation, and state of Shh signal transduction in wild-type (A–D, I–L, Q) and Shh-Tg (E–H, M–P, R–T) spinal cord at 12.5 and 18.5 dpc. A, E, I,M, Histological analysis of transverse sections taken at the forelimb level. E, M, Note hyperplasia and expansion of dorsal regions of the spinal cord ofShh-Tg embryos. M, At 18.5 dpc the central canal of the spinal cord is grossly enlarged and distended (hydromyelia). B, F, J,N, BrDU incorporation in dividing cells at (B, F) 12.5 dpc and (J, N) 18.5 dpc. Note that proliferative rates are low at 18.5 dpc in both wild-type and mutant specimens. RNA in situ hybridization showing expression of Ptc-1 at 12.5 dpc (C, G) and 18.5 dpc (K, O). Note ectopic dorsal expression inShh-Tg tissues (G, O,arrows). Expression of Gli-1 at 12.5 dpc (D, H) and 18.5 dpc (L, P). Note ectopic dorsal expression in Shh-Tg spinal cord (H, P) and dorsal root ganglion (H,left arrow). Q–T, Tissue from 18.5 dpc fetuses. Q, R, Wild-type andShh-Tg cervical spinal cord tissues showing expression of Shh.S, Distribution of Shh protein demonstrated by immunostaining with anti-Shh serum (Marti et al., 1995a). T, In situ hybridization showing expression of HNF3β. Note that expression is confined to the floorplate region (arrow).
During spinal cord development, there is evidence for programmed cell death initially in neural crest precursors and the floorplate region at early stages (Homma et al., 1994) and subsequently in ventral motor neuronal populations (Lance-Jones, 1982). To examine programmed cell death in the developing spinal cord, tissue from 12.5, 14.5, and 18.5 dpc wild-type and transgenic mice was analyzed by the TUNEL procedure. No differences were observed between wild-type and Shh-Tg samples even as late as 18.5 dpc (data not shown). We conclude from these studies that hyperplasia of the dorsal spinal cord inShh-Tg mice is most likely a result of proliferation per se, rather than inhibition of apoptotic cell death. Interestingly, Shh was capable of promoting proliferation at 12.5 dpc, when precursors are normally competent to divide, but not at 18.5 dpc, when neurogenesis is complete. Thus a “clock” that normally temporally restricts the period of neural precursor proliferation in wild-type embryos also appeared to be operative in Shh-Tg mice.
Given these dramatic differences in cellular response, it was important to establish that the Hedgehog signal transduction pathway was active at both of these time points. Transcriptional targets of Shh include the transmembrane receptor Patched-1 and the zinc finger transcription factor Gli-1. Upregulation of Ptc-1and Gli-1 was observed at both 12.5 and 18.5 dpc (Fig.3G,H,O,P), confirming activation of Shh signal transduction. Additionally, high levels of Shh expression (Fig. 3R) and protein production (Fig. 3S) were maintained at 18.5 dpc. Thus, activation of Shh signal transduction was not able to promote proliferation in the spinal cord at late fetal stages.
Proliferative effects of Shh in the absence of floorplate
Induction of floorplate has been observed when notochord was ectopically grafted in the chick neural tube (Placzek et al., 1993) or when naive neural plate tissues were treated with the N-terminal fragment of Shh (N-Shh) (Marti et al., 1995b; Roelink et al., 1995), but such effects were limited to early developmental stages (Placzek et al., 1993; Ericson et al., 1996). In addition, proliferative effects of notochord and floorplate have been observed previously (van Straaten et al., 1989; Placzek et al., 1993). One possibility was that induction of floorplate in Shh-Tg mice could lead to proliferation by factors other than Shh. However, ectopic expression ofHNF3β in floorplate structures was not observed in the spinal cord at either 10.5 or 18.5 dpc (Fig. 3T), in keeping with previous observations (Echelard et al., 1993; Epstein et al., 1996). These data indicate that Shh expression in the spinal cord of Shh-Tg mice occurs beyond the phase of competence to induce floorplate (Ericson et al., 1996). At 10.5 dpc, there was broad expression of HNF3β throughout the midbrain, indicating ventralization (Echelard et al., 1993) (data not shown). This may account in part for a failure in anterior neural tube closure (Fig. 2E). Because neural tube defects can have secondary effects on proliferation, patterning, and tissue survival, we focused our studies at the forelimb and posterior cervical spinal cord levels.
Ectopic Shh expression confers mixed dorsal–ventral character to the embryonic spinal cord
Sonic hedgehog is normally expressed in organizing structures at the ventral midline from early stages of neural development. Given thatShh expression in Shh-Tg mice occurs beyond the phase of competence to form floorplate, we investigated how dorsal–ventral organization was subsequently affected in 12.5–14.5 dpc embryos. Secreted factors, such as bone morphogenetic proteins (BMPs), from the non-neural ectoderm and roofplate, are thought to act on neural plate precursors to establish dorsal identity (Liem et al., 1997; Lee et al., 1998). Dorsal neural precursors can be recognized by expression of the molecular markers Pax-3 andPax-7 at appropriate stages in wild-type mice at 12.5 dpc (Fig. 4A) (Tanabe and Jessell, 1996). In Shh-Tg embryos, Pax-3expression was maintained in the alar plate, indicating that dorsal cell types that were established before ectopic Shhexpression were maintained despite ectopic Shh activity (Fig.4B). To investigate whether precursor cells with ventral character were induced, we tested expression of the markerNkx-2.2. Nxk-2.2 was detected exclusively in the ventral neural tube of 12.5 dpc wild-type embryos (Fig. 4C) but was ectopically induced in the dorsal region of Shh-Tg embryos (Fig. 4D). Thus, populations of neural precursor cells with dorsal and ventral character occupied a similar domain in the alar plate of the spinal cord in 12.5 dpc Shh-Tg embryos. Pax-6 is expressed predominantly at 12.5 dpc in the ventral ventricular zone of the neural tube as well as in postmitotic ventral neurons (Fig. 4E). Expression ofPax-6 can be repressed by Shh in neural plate explants in culture (Ericson et al., 1997). However, we observed dramatic upregulation of Pax-6 in the alar plate of Shh-Tg mice (Fig. 4F), consistent with the findings ofMonsoro-Burq et al. (1995). Pax-6 expression appeared to be a sensitive indicator of immature ventricular zone precursors exposed to ectopic Shh.
Analysis of dorsal–ventral organization in the spinal cord of 12.5–14.5 dpc Shh-Tg embryos. In situ hybridization was performed on 12.5 dpc (A–F, O, P) and 14.5 dpc (G–N) spinal cord from wild-type (A, C, E, G, I, K, M, O) andShh-Tg (B, D, F, H, J, L, N, P) embryos. Expression of Pax-3 (A, B) is maintained in Shh-Tg tissue (arrow).C, D, Nkx-2.2 is expressed in ventral ventricular zone (C, arrow) and ectopically in Shh-Tg tissue (D,arrow). E, F, Expression of Pax-6 normally occurs in the roofplate (rp) but is strongly upregulated in theShh-Tg tissue (F, arrow).G, H, BMP-7 expression in the meninges was unaffected in Shh-Tg mice (arrows). I, J, Maintenance and expansion of roofplate cells indicated by expression ofWnt-3a in Shh-Tg tissue.K, L, Expression of GDF-7was weakly detected in wild-type mice and strongly maintained in dorsal tissues of Shh-Tg mice. M,N, Pattern of Lmx-1b expression. Note large dorsal region in which Lmx-1b expression is interrupted. fp, Floorplate. O,P, Expression of the postmitotic neuronal markerBrn-3a in dorsal spinal cord (arrows) and dorsal root ganglion (drg).
It was possible that hyperproliferating precursor cells inShh-Tg mice could be giving rise to cells with dorsal, ventral, or mixed character. BMPs function as determinants of dorsal character in the spinal cord and are expressed in the roofplate and adjacent non-neural ectoderm of the embryonic neural tube (Liem et al., 1997; Lee et al., 1998) as well as the meninges. Our results indicated that BMP-7 expression in the meninges surrounding the spinal cord was similar in wild-type and transgenic embryos (Fig.4G,H). Analysis of Wnt-3aexpression confirmed that a roofplate population was maintained and expanded in Shh-Tg mice (Fig.4I,J). Moreover, dorsal neural tissues continued to express GDF-7 (Fig.4K,L) (Lee et al., 1998). Thus, BMP and Wnt gene family members normally expressed in the roofplate were maintained in Shh-Tg mice. To further characterize development of dorsal populations, we tested expression ofLmx-1b (Chen et al., 1998), which is expressed in cells throughout a large region of the dorsal spinal cord and to a lesser extent in the floorplate (Fig. 4 m ). Interestingly, we observed that the Lmx-1b domain was interrupted in Shh-Tg mice (Fig. 4N). Taken together, these results suggested that although Shh did not suppress the commitment to early dorsal fates, hyperplastic tissue that was generated later lacked certain dorsal characteristics. Further evidence for this sequence was provided by examination ofBrn-3a expression, a marker of postmitotic dorsal neurons that can be suppressed when a source of ectopic Shh is grafted into the early chick neural tube (Fedtsova and Turner, 1997). Maintenance ofBrn-3a expression was observed in Shh-Tg embryos at 12.5 dpc, indicating that onset of Shh exposure likely followed that of early specification of Brn-3a+ neurons (Fig. 4, compareO and P). In summary, our results suggested that early dorsal patterning of the neural tube was unaffected inShh-Tg mice, consistent with the delayed onset of Shh activation. However, later populations of precursors exposed to the ectopic Shh signal gave rise to cells expressing ventral markers. Ultimately, the resulting structure comprised a hyperplastic and disorganized dorsal extension superimposed on a spinal cord in which the dorsal–ventral pattern was largely intact.
Cells of the expanded ventricular zone in Shh-Tg mice are blocked in an undifferentiated state
We next investigated the ultimate cell fates acquired in the hyperplastic dorsal spinal cord tissue of Shh-Tg mice. Because Wnt-1-GAL4 expression persists throughout the antenatal period, it is possible to characterize effects of ongoing ectopic Shh pathway activation in tissue until late fetal stages. A landmark structure in the wild-type spinal cord is the ependymal cell-lined central canal (Fig.5A,D). A striking finding in Shh-Tg mice is that the central canal is massively enlarged and often has a folded appearance in section (Fig. 5B). Of particular interest was a cell-dense, pseudostratified periventricular layer that was revealed by the nuclear stain DAPI (Fig. 5C,E). This was reminiscent of the periventricular neuroepithelial germinal zone, a structure normally found only at earlier time points in spinal cord development, which disappears as neural precursors differentiate and emigrate from this region. Consistent with this interpretation, we never observed labeling with mature (NeuN) or immature (TuJ1) neuronal markers within the persistent periventricular zones (Fig.5E,G) or immunolabeling with the astrocyte marker GFAP or the oligodendrocyte markers GalC or O4 (data not shown). Antibodies against the N-terminal fragment of Shh (Marti et al., 1995b) revealed that Shh proteins were produced exclusively within the enlarged VZ (Figs. 3R,S,5F), a region of dramatic Ptc-1 andGli-1 upregulation in Shh-Tg mice (Fig.3O,P). Further confirmation of the immature character of the VZ cells was provided by analysis of HES-1(Sasai et al., 1992) and Pax-6 expression. Overexpression ofHES-1 in the rodent neural tube causes delay or inhibition of differentiation (Ishibashi et al., 1994), whereas HES-1loss of function results in premature differentiation of neural plate precursor cells (Ishibashi et al., 1995). We observed thatHES-1 expression was upregulated in the ventricular zone ofShh-Tg mice (Fig. 5L). Our finding ofPax-6 expression in the same region (Fig.5 m ) further suggests that VZ cells at 18.5 dpc share similar properties with ventricular zone precursor cells during the period of neurogenesis (Tanabe and Jessell, 1996). Thus, markers normally associated with precursor populations and dividing cells present during embryogenesis revealed ectopic induction and striking persistence into late fetal stages in the CNS of Shh-Tg mice.
Absence of differentiation in cells lining the enlarged ventricular zone in 18.5 dpc Shh -Tg mice.A, B, Histological analysis of wild-type (A) and Shh-Tg (B) spinal cord at the forelimb level with hematoxylin–eosin. Note that the ventricular zone (VZ) is massively enlarged in Shh-Tg mice (arrows). In addition, the tissue surrounding the VZ is hyperplastic. C, Cells lining the VZ have a pseudostratified columnar appearance as revealed by the nuclear stainDAPI. D–G, Immunocytochemistry of VZ region in wild-type (D) and Shh-Tg (E–G) tissue. D,E, The mature neuronal marker NeuN (green) counterstained with DAPI (blue). Compare the sizes of ventricular zone (arrows) and the absence of NeuN labeling.F, Shh expression (red) was confined to VZ cells and did not overlap NeuN+ cells (green) in the surrounding hyperplastic tissue (arrows).G, Labeling with β-tubulin III (TuJ1,green) is excluded from the VZ (arrows).H–M, In situ hybridization of wild-type (H–J) and Shh-Tg (K–M) fetuses. H, K,Bright-field images are shown for orientation. I, L,Expression of HES-1 was upregulated inShh-Tg (arrow) in cells lining the VZ.J, M,Pax-6 was maintained in a similar pattern of expression (arrow).
Induction of neuronal and oligodendroglial lineages in hyperplastic tissue adjacent to Shh-producing ventricular zone cells
The preceding results demonstrated that regions of highest Hedgehog pathway activation formed large ventricular zone structures comprising cells maintained in an undifferentiated state. In contrast, neuronal differentiation clearly occurred in the hyperplastic tissue adjacent to the persistent VZ structures in Shh-Tg mice, as demonstrated by immunolabeling with the mature neuronal marker Neu-N (Fig. 5E). Several lines of evidence indicate that Shh is necessary and sufficient for ventral neuron (in particular motor neuron) and oligodendrocyte induction (Marti et al., 1995b; Roelink et al., 1995; Chiang et al., 1996; Ericson et al., 1996; Poncet et al., 1996; Pringle et al., 1996; Orentas et al., 1999). Thus, we determined whether ventral motor neurons or oligodendrocytes were induced inShh-Tg mice.
We initially tested expression of molecular markers associated with postmitotic ventral neuronal populations including c-ret,Isl-1, and En-1. Although ventrally located motor and interneuronal populations were clearly identified, ectopic expression of these markers was not observed (data not shown). Some dorsally located neurons normally express Isl-1 (Liem et al., 1997). However, additional numbers of Isl-1/2 neurons were only observed in 1 out of 5 Shh-Tg animals analyzed (Fig.6A). One possibility was that the dose of Shh may have been either too low or too high for efficient Isl-1 motor neuron induction, because concentration dependence of motor neuron induction has been demonstrated in chick neural explant culture (Roelink et al., 1995; Ericson et al., 1997). We therefore analyzed tissue for the presence of Lim-3 and Nxk-2.2 neurons. These markers indicate cell populations lying immediately dorsal and ventral to the region normally giving rise to motor neurons in the ventricular zone (Tanabe et al., 1998; Briscoe et al., 1999), and they can be induced with lower and higher concentrations of N-Shh, respectively (Ericson et al., 1997). As shown (Fig.6B–D), we readily detected ectopic induction of Nkx-2.2+ and Lim-3+ cells in regions adjacent to the enlarged ventricular zone. Thus, it is unlikely that the failure to form Isl-1+ neurons is related to the dosage of Shh inShh-Tg mice. A second possibility was that Shh actually inhibited the differentiation of Isl-1 motor neurons, as suggested fromin vitro studies (Kalyani et al., 1998). We therefore analyzed Shh-Tg dorsal spinal cord cells in dispersed explant cultures for generation of Isl-1 neurons. Although numerous Lim-3+ and Nkx-2.2+ neurons were detected, we failed to detect Isl-1+ neurons after 5–7 d in culture (data not shown). Thus, the most likely explanation for the lack of Isl-1 motor neuron induction is that temporal restrictions on competence are exceeded by the time of Shh production in Shh-Tg mice.
Analysis of neuronal and oligodendroglial cell fate in the dorsal spinal cord of Shh-Tg mice. Immunocytochemistry of wild-type (E, F) andShh-Tg (A–D, G–I) 18.5 dpc fetuses. A, Isl-1/2 (red) labeled cells of the dorsal root ganglion (drg, white arrow) and a few cells in the dorsal spinal cord in one of five specimens analyzed (hollow arrow). Note absence of counterstain with the mature neuronal marker NeuN (green) in dorsal cells. B, Numerous Nkx-2.2+ cells were detected in regions surrounding the VZ (diagonal arrow) but not the drg. C, The area boxed in B at higher power. Not all Nkx-2.2+ cells (e.g., solid arrow) counterstained with NeuN (hollow arrow), which suggests relative immaturity.D, Region adjacent to VZ analyzed with Lim-3 demonstrating numerous positive cells (red,arrows). E, F, Analysis of Lmx-1b populations (red) in the substantia gelatinosa region (boxed in E, and F) and roofplate (F, hollow arrow) counterstained with NeuN (green). Contrast the organization of Lmx-1b+ cells in the wild-type (G) with those of Shh-Tg spinal cord tissue. H, Labeling of cells with the dorsal marker Pax-7 (green, hollow arrows) compared with the ventral neuronal marker Nkx-2.2 (red,solid arrows). Cells expressing both markers were not detected. I, Oligodendrocyte precursors are induced in regions adjacent to the Shh-expressing VZ, as indicated by the marker PDGFαR (green). Counterstain with DAPI isblue.
We next investigated the dorsal nature of tissue in Shh-Tg mice at 18.5 dpc with the marker Lmx-1b. Although Lmx-1b is expressed in the floorplate and dorsal spinal cord at embryonic stages (Fig. 4J), Lmx-1b antisera only labels dorsal cells at both embryonic and fetal stages (data not shown). In the wild-type fetus, Lmx-1b is detected broadly in neurons of the substantia gelatinosa (orange/yellow) and the roofplate (red) (Fig.6F). In contrast, the distribution of Lmx-1b cells is disorganized in Shh-Tg mice (Fig. 6G). We conclude from this that Lmx-1b cells persist in the dorsal spinal cord and are interspersed with cells of ventral character (e.g., Nkx-2.2+ neurons). These results do not rule out the possibility that cells with mixed dorsal–ventral character were elaborated in Shh-Tg mice. To assess this, we immunolabeled dorsal spinal cord tissue with antibodies against the ventral neuronal marker Nkx-2.2 and the dorsally restricted marker Pax-7 (Tanabe and Jessell, 1996). As shown (Fig.6H), we did not detect cells that expressed both markers.
Oligodendrocyte precursors arise from a similar region of the neural tube that gives rise to motor neurons (Sun et al., 1998) and can be induced at an identical concentration of N-Shh in neural explant culture (Pringle et al., 1996). To determine whether oligodendrocyte precursors were induced in Shh-Tg mice, we performed immunolabeling with PDGFαR (Pringle and Richardson, 1993). As shown in Figure 6I, we observed numerous PDGFαR+ cells in regions adjacent to the VZ, the source of ectopic Shh. Induction of O4+ cells was observed in a similar distribution; however, GalC+ oligodendrocytes were only detected in the axons of the dorsal funiculus (data not shown). These results confirmed that the oligodendrocyte lineage was induced in the hyperplastic tissue surrounding the VZ in Shh-Tg mice.
DISCUSSION
Ectopic Hedgehog signaling has been implicated in the etiology of CNS tumors (Hahn et al., 1996; Johnson et al., 1996); however, mechanisms underlying Hedgehog-mediated tumorigenesis are poorly understood. We have used a GAL4/UAS bigenic system (Ornitz et al., 1991; Brand and Perrimon, 1993; Wang et al., 1997), which allows for the production of stable transgenic lines to produce large numbers of embryos that express a lethal transgene, to explore the effects of maintaining ectopic Shh activity in the dorsal neural tube as a model of deregulated Hedgehog signaling in the developing CNS. Analysis of bigenic embryos revealed dramatic neural hyperplasia and enhanced proliferative levels at 12.5 dpc. However, at 18.5 dpc, neural tissue was postmitotic, despite the fact that cells were exposed to Shh and still responsive, as demonstrated by upregulation of two general transcriptional targets, Ptc-1 and Gli-1.
Shh proliferative effects in the developing spinal cord
Several studies using primary CNS precursor cell cultures have demonstrated proliferative effects of the biologically active N-Shh protein after treatment for 36–48 hr (Jensen and Wallace, 1997;Kalyani et al., 1998). Whether Shh functions as a direct mitogenin vitro, however, has not been established (Jensen and Wallace, 1997). In a transgenic gain-of-function model, resolving whether proliferative effects of Shh are direct is difficult, because Shh could lead to induction of other mitogens (e.g., Wnts/BMPs) (Dickinson et al., 1994). Unfortunately, it is not feasible to remove all Wnt/BMP function from this model to determine whether the phenotype also depends on these activities. However, the expansion and patterning abnormalities clearly require Shh, and the induction of ventral cell types (e.g., Nkx-2.2) is most consistent with direct Shh signaling. Moreover, our observations are entirely consistent with a number ofin vitro studies of Shh proliferative effects on CNS precursors (Jensen and Wallace, 1997; Kalyani et al., 1998;Wechsler-Reya and Scott, 1999) and the finding that Patchedmutations in mice result in highly proliferative cerebellar tumors (Goodrich et al., 1997). An obvious question remains whether Shh has a role in regulating proliferation during normal spinal cord development, as has been reported recently for cerebellar granule cells (Wechsler-Reya and Scott, 1999).
Our findings from analysis of proliferation in vivo indicate that CNS precursors are competent to proliferate in response to activation of the Shh signal transduction pathway only at selected periods during embryogenesis. Overexpression of the Hedgehog transcriptional target Gli-1 resulted in increased levels of proliferation in the developing mouse brain (Hynes et al., 1997) andXenopus ectoderm (Dahmane et al., 1997); moreover,GLI upregulation has been associated with brain tumors and basal cell carcinoma in humans (Dahmane et al., 1997). We have usedGli-1 and Patched-1 to confirm activation of Shh signal transduction at both 12.5 and 18.5 dpc in the CNS ofShh-Tg mice. However, given that levels of proliferation were significantly elevated only at embryonic stages, it is clear thatGli-1 overexpression is itself insufficient for proliferation in neural tissues at 18.5 dpc. Rather, it is possible that Gli-1 acts in concert with other determinants of cell cycle regulation to effect a proliferative state, as has been suggested previously (Ruppert et al., 1991). In preliminary analysis, we have observed that explant cultures of dorsal spinal cord tissue from 17.5 dpc Shh-Tg fetuses resume proliferation after 3 d in serum-free media. Further work will be required to determine whether dispersal of such tissue liberates environmental (i.e., secreted or matrix-associated) signals that antagonize Shh proliferative effects.
Shh signaling prevents differentiation of neural precursors
Although our results are consistent with a mitogenic role for Shh in the neural tube, a second possibility is that proliferative effects are an indirect consequence of preventing or delaying differentiation of neural precursors. A ventricular zone germinal matrix-like structure comprising primitive undifferentiated yet nondividing cells persisted in the dorsal spinal cord of Shh-Tg mice. At 18.5 dpc, cells in these regions expressed markers indicative of mitotically active neural precursors such as Pax-6, HES-1, andDbx-1. Whether these cells represent true multipotential precursors or are restricted in their potential to form neural cell types is under study.
Our results suggest possible mechanisms downstream of Shh signaling that could function to inhibit neuronal differentiation. InShh-Tg mice, superimposition of endogenous dorsalizing signals (e.g., GDF-7, BMP-7) with Shh resulted in a broad overlap of the ventral marker Nkx-2.2 with dorsally expressedPax-3 at 12.5 dpc. We evaluated whether the mixed signals in the dorsal compartment might have prevented differentiation of cells along a coherent pathway. However, we did not detect any cells coexpressing the ventral and dorsal markers Nkx-2.2 and Pax-7, making such a mechanism unlikely. Another possibility is that persistent expression of Pax-6 or other factors associated with neural precursors could institute a block to terminal differentiation. Upregulation of HES-1, in particular, suggests that such a mechanism may be functioning in Shh-Tg mice. Interestingly,Kalyani et al. (1998) and Wechsler-Reya and Scott (1999) recently reported that Shh can directly inhibit differentiation of neuronally restricted precursor cells in vitro.
Absence of ectopic floorplate in Shh-Tg mice
We determined that Shh effects on proliferation and differentiation were not mediated by ectopic floorplate. Conversion of the entire spinal cord to floorplate has been observed inPatched-deficient mice (Goodrich et al., 1997), establishing the competence of the lateral (future dorsal) neural plate to respond to Shh signaling at early stages. The kinetics of Wnt-1/GAL4-X UAS-Shh expression initiates Shh expression at ∼9.5–10 dpc in the spinal cord, when dorsalizing signals (e.g., from roofplate and non-neural ectoderm) (Liem et al., 1997; Lee et al., 1998) have already commenced. In the face of non-naive tissues, Shh is inadequate to convert the dorsal spinal cord to floorplate (Placzek et al., 1993; Ericson et al., 1996). Indeed, maintenance ofWnt-3a and GDF-7 expression in Shh-Tg mice indicates that important dorsal organizing properties of the roofplate cannot be suppressed by Shh beyond an early naive phase.
Developmental neuropathology of CNS tumors
Given that CNS tumors can arise in tissues well after primary patterning events have taken place, it is relevant to consider the temporal role of the Hedgehog signaling pathway. Our results indicate that activation of Hedgehog signaling at 10.5 dpc in neural tissue that has already acquired dorsal character can result in mixed and complex morphology. The persistent and massively enlarged ventricular zone inShh-Tg mice was surrounded by hyperplastic and largely nestin-positive tissues comprising both dorsal and ventral neuronal cell types. For example, we observed patches of cells expressing the dorsal marker Lmx-1b adjacent to tissue containing ectopic Lim-3+ and Nkx-2.2+ neurons. Although Shh is capable of ventral motor neuron induction at early developmental stages, we did not detect induction of Isl-1+ motor neurons, most likely because Shh is produced at 10.5 dpc in Shh-Tg mice, beyond the period when neural tube is capable of forming ectopic floorplate or motor neurons. In addition, foci of both astrocytes and oligodendrocyte were also observed.
Whether these findings are relevant as an indication of effects of active Hedgehog signal transduction contributing to a tumorigenic state in humans requires further analysis. Although inactivating mutations ofPATCHED can result in medulloblastoma, a tumor of cerebellar granule cells, the spinal cord is not affected in such patients. Nevertheless, Shh causes proliferation in spinal cord precursor cellsin vitro (Kalyani et al., 1998) and in vivo(present study). Our results suggest that severe temporal restrictions on cellular competence could limit a putative “tumorigenic window” to only a few days of embryogenesis in the spinal cord. In this regard it is interesting to note that cerebellar granule cells are the latest population of CNS precursors to undergo terminal differentiation. Thus it is possible that the relatively long period of granule cell competence may facilitate stochastic events required for tumorigenic transformation in response to Hedgehog pathway activation in humans and mice.
Footnotes
We thank Drs. Makoto Ishibashi, Ryoichiro Kageyama, Qiufu Ma, Chuck Stiles, David Fischer, Yann Echelard, Rob Wechsler-Reya, and Matt Scott for comments and stimulating discussions; Yann Echelard for preparation of the Wnt-1-GAL4 transgene; and Dong-In Yuk, Bianca Klumpar, and Wendy Liu for technical assistance. D.H.R. was supported by a Physician Postdoctoral Fellowship from the Howard Hughes Medical Institute and Grant HD01182 from National Institutes of Health. B.S.J. was supported by a postdoctoral fellowship of the Medical Research Council of Canada. These studies were funded by National Institutes of Health Grants NS32691 and HD30249 (A.P.M.) and the Charles H. Hood Foundation and Brain Tumor Society (D.H.R.). D.H.R. is a recipient of a Basil O’Conor Starter Scholar Award from the March of Dimes Foundation.
Correspondence should be addressed to Dr. David H. Rowitch at his present address: Department of Pediatric Oncology, Dana-Farber Cancer Institute, Boston, MA 02115, or to Dr. Andy McMahon, Department of Molecular and Cellular Biology, Harvard University, The Biolabs, 16 Divinity Avenue, Cambridge, MA 02138.
Dr. St.-Jacques’s present address: Ontogeny, Inc., Cambridge, MA 02138.
