The Sonic Hedgehog-Gli pathway regulates dorsal brain growth and tumorigenesis

The regulation of normal brain growth, on one hand, and of tumor formation, on the other, are two tightly related processes that remain poorly understood. Several secreted signals have been previously shown to participate in brain development (Kilpatrick et al., 1995; Li et al., 1998; Mabie et al., 1999). The secreted protein Sonic hedgehog (SHH) is involved in different aspects of the development of the early CNS, controlling cell differentiation as well as cell proliferation. For example, SHH is required for the differentiation of floor plate cells and ventral neurons in the early neural tube (Echelard et al., 1993; Roelink et al., 1994; Ruiz i Altaba et al., 1995), and it later regulates granule cell precursor proliferation in the cerebellum (Dahmane and Ruiz i Altaba, 1999; Wallace, 1999; Weschler-Reya and Scott, 1999).

Gli zinc-finger transcription factors participate in Hedgehog (HH) signaling and Gli1 is consistently induced in cells that receive a HH signal [reviewed by Ruiz i Altaba (Ruiz i Altaba, 1999)]. In vivo, we have shown that Gli1 is sufficient to induce basal cell carcinoma (BCC)-like skin tumors in the epidermis of frog embryos (Dahmane et al., 1997), a result later reproduced in mice (Grachtchouk et al., 2000; Nilsson et al., 2000) that further validates our approach (Wallingford et al., 1997). The expression of GLI1 in proliferating cells of BCCs, medulloblastomas (MBs) and rhabdomyosarcomas (Dahmane et al., 1997; Goodrich et al., 1997; Hahn et al., 1998; Xie et al., 1997; Pietsch et al., 1997; Vorechovsky et al., 1997; Reifenberger et al., 1998), suggests a wider role of the SHH-GLI pathway in tumorigenesis than previously suspected. However, even though GLI1 was originally identified in a glioma line (Kinzler et al., 1987; Ruppert, 1991), its proposed role in glial brain tumors has not been supported (Salgaller et al., 1991; Xiao et al., 1994).

We have investigated a role for the SHH-Gli pathway in normal and abnormal precursor proliferation in the dorsal brain. Specifically, we have investigated whether the cerebral cortex and the tectum (colliculi), two major, dorsal, layered and late developing structures of the brain, share common mechanisms for growth regulation with the cerebellum. We show that the SHH-Gli pathway modulates the growth of the dorsal brain and that it is deregulated in brain tumors, possibly being a cause for their initiation from precursor cells, as Gli1 is sufficient to induce CNS hyperproliferation. The ability of cyclopamine to inhibit the proliferation of brain tumor cells suggests that the SHH pathway is also be required for tumor maintenance.

Swiss-Webster mouse embryos were staged counting the morning after conception as embryonic day (E) 0.5. Shh homozygote mutant mice were a kind gift from C. Chiang (Chiang et al., 1996). Human tumor samples were derived from the operating room or from the NYU tumor banks. Brain tumor cell lines were obtained from ATCC and grown according to its specifications. GL261was a kind gift of Dr D. Zagzag. Frog (Xenopus laevis) embryos were obtained, reared and staged by standard methods. Tadpoles were ∼2 days old. All statistical analyses were carried out using the Student’s t test and deviations are shown as s.e.m.

Neocortical or tectal explants from embryonic day (E) 17.5 to postnatal-day (P) 3 mice were taken from the parietal region, or adjacent to the dorsal midline in the prospective superior and inferior colliculi, respectively. After removal of the meninges, the explants were grown on floating filters in serum-free media (Nothias et al., 1998; Dahmane and Ruiz i Altaba 1999). After 12 hours in culture, SHH was added and incubation continued for a further 48 hours. Explants for RNA preparation were directly collected in Trizol (Gibco-BRL). For dissociated cells, parietal cortical pieces of P3 brains were pooled (∼10 explants per experiment) and treated with trypsin for 10 minutes (0.25 mg/ml at room temperature). Tissue was triturated manually in DNase (0.5 mg/ml), and cells were centrifuged, resuspended in supplemented serum-free media and plated at a density of ∼400 cells/mm² in poly-L-lysine-coated 16- or 8-well slides. After 12 hours the media were replaced and SHH protein added if required. Cells were cultured for a further 48 hours and processed for immunocytochemistry after fixation in 4% paraformaldehyde for 1 minute. Primary gliomas from the operating room were dissociated with papain and plated in U118 media containing 10% fetal calf serum (FCS) or in DMEM:F12 serum-free media supplemented with BIT-9500 (Stem Cell Technologies) and 20 ng/ml of each of FGF2, EGF and PDGF. After two to three passages, the cells had a homogenous appearance and were then tested. Recombinant N-SHH was a kind gift from Ontogeny and was used at 5 nM. For blocking experiments, anti-SHH mAb 5E1 (Ericson et al., 1996) was used at 20 μg/ml (obtained from the University of Iowa Hybridoma Bank). Cyclopamine (a kind gift from the Poisonous Plant Laboratory or purchased from Toronto Research Chemicals) was used at 0.5-5 μM for 48 hours before assaying. Cell lines were plated at 60% confluency the night before cyclopamine treatment. FK and ddFK (Sigma) were used at 50 μM.

Injection of synthetic RNAs into frog embryos was performed into one cell at the two-cell stage, targeting the future CNS and epidermis. Frog or human Gli1 RNAs (Lee et al., 1997) were injected at 2 ng/10 nl/embryo. The N-terminal Myc-epitope tag in the frog Gli1 and Gli2 proteins (Lee et al., 1997) was used to monitor protein distribution. lacZ RNA was co-injected at 0.2 ng/10 nl/embryo and was used as a lineage tracer through X-gal staining, yielding an insoluble blue precipitate. Morpholino antisense oligonucleotides were purchased from Gene Tools and used at 0.5 mM. These were frog Gli1, 5′CGGGCGGACACTGGCGGGACGC3′; frog Gli2, 5′GCACAGAACGCAGGTAATGCTCCAT3′; and frog Shh, 5′GAGATTCGAGTTCGCAACCAGCATC3′. In all cases, the oligonucleotides were designed to be complementary to regions near the initiation ATG codon and are predicted to inhibit translation (Heasman et al., 2000).

In situ hybridization on serial ∼20 μm cryostat sections with digoxigenin-labeled antisense RNA probes and full-length frog, mouse or human Gli1 cDNA clones and histology were as previously described (Dahmane et al., 1997; Lee et al., 1997; Park et al., 2000). Visualization of the low levels of GLI1 and PTCH1 expression in CNS tumors and mouse brains older than E17 required long (2 days at room temperature) chromogenic development of the in situ hybridization reactions. Specificity was confirmed using sense RNA probes. RT-PCR of human tumors or cell lines was performed for 27, 32 and 37 cycles to determine the linear amplification range. PCR primers and specific reaction conditions are available upon request. A probe for DDR1 was made as described (Weiner et al., 2000). A 0.6 kb RT-PCR clone was used as a template for probe production for mouse Pdgfrα (platelet-derived growth factor receptor α).

BrdU incorporation in explants and dissociated cells was for 2 hours at 6 μg/ml. Primary tumor cultures were labelled with BrdU for 14 hours. Pregnant mice were injected intraperitoneally with a single dose of 50 μl of 10 mg/ml BrdU and embryos dissected 2 hours afterwards. For tadpoles, one 20 nl injection of 10 mg/ml BrdU into the lumen of the CNS and one into the endoderm were performed 1 hour before fixation. Sections of embryos or explants (14-20 μm) were prepared using a cryostat. Immunocytochemistry with monoclonal anti-BrdU antibody (Becton-Dickinson), monoclonal anti-vimentin antibody (Santa Cruz), monoclonal anti-neuronal tubulin Tuj1 antibody (Babco), rat monoclonal anti-Nestin antibody (University of Iowa Hybridoma Bank) or O4 monoclonal antibody (Chemicon, also a kind gift of Bob Miller) was performed on frozen sections and cells using fluorescein-conjugated secondary antibodies (Boehringer Mannheim).

First, we sought to clarify the patterns of expression of the Shh and Gli genes (Echelard et al., 1993; Roelink et al., 1994) in the late embryonic and postnatal mouse brain. By E15.5, the three Gli genes are expressed in the ventricular zone (vz) of the future cerebral cortex and tectum (Fig. 1A,D,G and not shown), where they are detected as early as E13.5 (Hui et al., 1994). However, Gli1 was expressed at higher levels in the deeper cortical vz/subventricular zone (svz), Gli2 mostly in vz cells and Gli3 in these two populations (Fig. 1A,D,G). Gli1 and Gli3, but not Gli2, also show a transient graded distribution along the anteroposterior axis (Fig. 1A,D,G). At E17.5, they remain co-expressed in the vz of the cerebral cortex (Fig. 1B,E,H), tectum and tegmentum (Fig. 1C,F,I). Gli expression in the tectum is localized within the vz, suggesting heterogeneity in precursor cells. At this stage, Shh is expressed in the tegmentum, but is not detectable by in situ hybridization in the tectum or cerebral cortex. However, more sensitive RT-PCR analyses indicated that Shh is indeed found in these regions from E14.5 to P3 (Fig. 1J). The neocortical meninges expressed Gli1 but not Shh (Fig. 2A).

Perinatally, the Gli and Shh genes display largely mutually exclusive expression domains. Shh RNA is found in the Purkinje layer in the cerebellum, and in the superficial layers of the tectum (Fig. 1M) and cerebral cortex (probably layer V; Fig. 1K,L) (Traiffort et al., 1999). Lack of Shh expression in dorsal progenitors would be consistent with its reported absence from late embryonic and postnatal neurospheres (Zhu et al., 1999). Shh and Gli1 were also detected in the dentate gyrus of the hippocampus (Fig. 1K,N).

Gli1, but not Gli2 or Gli3, is also expressed at low levels by single cells scattered throughout the brain (Fig. 1B,O,Q). Given that there is massive gliogenesis in the perinatal cortex, if Gli1+ scattering cells are glial progenitors that respond or have responded to SHH, the pattern of Gli1 expression should be similar to that of its target Ptch1 and possibly to that of Pdgfrα, a marker of oligodendrocyte progenitors (Goodrich et al., 1996; Pringle and Richardson, 1993). The expression of Gli1 was found in a scattered pattern similar to that of Ptch1 and Pdgfrα in serial sections (Fig. 1Q-S), although the low levels of their expression, together with the lack of specific antibodies, precluded double-labeling analyses.

SHH secreted perinatally from differentiated cells in the cortical plate could affect precursor cells in the vz/svz. To test this possibility, we cultured isolated explants of late embryonic (E17.5) and postnatal (P3) mouse cerebral cortex in vitro (Fig. 2C). SHH treatment dramatically increased Gli1 expression over that seen in untreated sibling explants (Fig. 2A,B). Upregulation was higher at P3 (Fig. 2B). In situ hybridization analyses confirmed this effect (Fig. 2D,E) and localized Gli1 expression mostly to vz cells (Fig. 2E). In response to SHH, the expression of Ptch1 showed a small but consistent upregulation, that of Gli2 and Gli3 showed a little variation, and that of endogenous Shh remained unchanged, whereas the expression of Pdgfrα in control and SHH-treated samples was uninformative, as its expression was found at high levels in both cases (Fig. 2A,B and not shown).

To test whether SHH could act as a mitogen for precursor cells, we cultured P3 neocortical explants for 48 hours and added BrdU to the media 2 hours before fixation. SHH treatment led to an approx. twofold increase in the number of BrdU-positive cells when compared with control untreated explants (Fig. 2F,G,K: control, 105±20 cells/explant section; SHH treated, 226±14 cells/explant section; P<0.005). This increase is similar to that seen in cerebellar precursors after SHH treatment (Dahmane and Ruiz i Altaba, 1999). As with Gli1 expression, the great majority of BrdU-positive cells in SHH-treated explants are in the vz (Fig. 2G).

Treatment of neocortical explants with a blocking anti-SHH antibody, extensively used previously to block SHH function specifically (mAb 5E1) (Ericson et al., 1996; Dahmane and Ruiz i Altaba, 1999; Weschler-Reya and Scott, 1999) produced a twofold decrease in BrdU incorporation after 48 hours in culture when compared with untreated controls (Fig. 2H,F,K: control, 105±20 cells/explant section; mAb 5E1-treated, 45±8 cells/explant section; P<0.005). To corroborate this result, we also attempted to block SHH signaling with cyclopamine and forskolin (FK) and with the inactive derivative 1,9-dideoxyforskolin (ddFK) as control. Cyclopamine treatment inhibits the response to SHH signaling (Incardona et al., 1998; Cooper et al., 1998), by acting on the Patched-Smoothened membrane receptor complex (Taipale et al., 2000). Cyclopamine treatment resulted in a twofold decrease in cell proliferation as measured by BrdU incorporation (Fig. 2K: untreated, 72±5.6 cells/explant section; cyclopamine treated, 38±3.4 cells/explant section, P<0.001). FK treatment leads to an upregulation of PKA activity and this inhibits the SHH pathway intracellularly (Fan et al., 1995; Dahmane and Ruiz i Altaba, 1999). FK-treated explants showed a twofold reduction in BrdU incorporation after 48 hours in culture when compared with the inactive derivative 1,9-dideoxyforskolin (ddFK)-treated sibling explants (Fig. 2I-K: ddFK-treated, 73±10 cells/explant section; FK treated, 35±5 cells/explant section; P<0.005).

To test if the results in the cortex could be extended to the tectum, we cultured explants of prospective superior and inferior colliculi. Both responded identically and we do not differentiate between them here. SHH upregulated expression of Gli1 and Ptch1 and increased the proliferation of cells (Fig. 3A-E). Expression of Gli2 or Gli3 was not upregulated (Fig. 3B). SHH treatment of tectal explants induced an approx. twofold increase in the number of BrdU-positive cells (Fig. 3C,E: control, 20.1± 3.6 cells/explant section; SHH treated, 40± 5 cells/explant section; P=0.005). As in the cortex, the increase in the number of BrdU-positive cells was seen near the vz (Fig. 3C,D), and FK, but not ddFK, was also able to inhibit BrdU incorporation by approx. two-fold (Fig. 3E: ddFK treated, 12.8±5.9 cells/explant section; FK treated, 7.3±2.5 cells/explant section; P<0.01).

To investigate the identity of BrdU-positive cells seen after SHH treatment, we tested for the mitogenic effects of SHH on dissociated P3 cortical cells. Addition of BrdU for 2 hours after 48 hours in culture showed an approx. twofold increase in the percentage of BrdU-positive cells over total cell number per field in SHH-treated versus control untreated samples (Fig. 3E, first panel: control, 7.1±0.8%; SHH treated, 14.8±0.4%; P<0.005). Double-labeling analyses showed that BrdU-positive cells with nuclear labeling were also nestin-positive cells with cytoplasmic labeling (Fig. 4A,B,E; 88±2.5% BrdU-positive cells per field were also nestin positive in control samples, and 93±3.2% in SHH-treated samples). Because the intermediate filament nestin identities precursors (Lendahl et al., 1990), including radial glia (Hockfield and McKay, 1985), this result indicates that SHH acts on precursor cells to increase their proliferation, a result consistent with the proliferation of E13.5 neurospheres after SHH treatment (Kalyani et al., 1998). By contrast, only very rarely did BrdU-positive cells express the astrocyte marker GFAP or the oligodendrocyte marker O4 (Fig. 4C-E and not shown). The identity of BrdU-positive cells as Nestin-positive precursors is in accordance with the localization of BrdU-positive cells in the vz/svz of explants (Fig. 2F,G).

Our experimental in vitro data predicts that lack of SHH in vivo will result in a downregulation of proliferation in the vz/svz of the neocortex. Shh-null mice die at birth and display severe holoprosencephaly (Fig. 5A), resulting in part from the complete lack of ventral CNS structures and the malformation of the face (Chiang et al., 1996). However, dorsal CNS structures are present, as demonstrated by the maintained expression of the Emx1, Otx2 and En2 in the mutant neocortex, diencephalon and tectum, respectively (Chiang et al., 1996). It is thus feasible to ask whether the dorsal-only brain of Shh null mice (Fig. 5L,M) displays normal proliferation prenatally.

Both RT-PCR analyses and in situ hybridization of E18.5 Shh homozygote null embryos revealed that the presumptive cortex expresses the neocortical marker Tbr1 and is present adjacent to the proboscis, which does not express Tbr1 (not shown). To analyze cell proliferation, a pulse of excess BrdU was given to pregnant mothers 2 hours before embryo collection. Within the cortex, cell proliferation was decreased nearly fourfold in mutant versus control embryos (Fig. 5B,C,F,G; 11.2±0.6% BrdU-positive cells in the vz/svz of wild-type embryos versus 3±0.7% BrdU-positive cells in mutant embryos, P<0.001; counting the vz/svz as a width of 10 cell diameters from the ventricle). Nevertheless, nestin-positive vz/svz cells (Fig. 5D,E), Tuj1-positive neurons (Fig. 5F,G) and GFAP-positive astrocytes (not shown) could be observed in both wild-type and mutant cortices. More posterior brain regions also showed a marked decrease in proliferation. For example, in the midbrain areas, the malformed tectum showed one-fifth to a half the number of BrdU-positive cells, and E18 mutant mice lacked a recognizable cerebellum, even though the adjacent choroid plexus was clearly present (Fig. 5H,I). Indeed, by E15.5, the cerebellum was already drastically reduced morphologically (Fig. 5L arrow), in line with previous work on the indispensable role of SHH in cerebellar development (Dahmane and Ruiz i Altaba, 1999; Wallace, 1999; Weschler-Reya and Scott, 1999).

In contrast to the decrease in the proliferation of the vz/svz of the brain, other parts of the mutant embryo appeared to proliferate normally or even to overproliferate. For example, basal epidermal cells appeared to have the same amount of BrdU-positive cells in mutants and wild-type littermates (Fig. 5J,K; 34.3±1.3 BrdU-positive cells per field in the basal layer of the epidermis in mutant embryos versus 36.2±3.1 BrdU-positive cells per field in the basal layer of the interfollicular epidermis of wild-type littermates, P>0.5), even though the mutant skin lacked hair follicles (Fig. 5K) (St-Jacques et al., 1998). The liver, by contrast, appeared to overproliferate, as judged by the number of BrdU- and HNF-3β-labelled cells (not shown).

Our findings in normal development raise the possibility that inappropriate activation or maintenance of the SHH-GLI pathway could lead to hyperproliferation, the basis of tumorigenesis. To test this idea, we first analyzed sporadic human brain tumors for the consistent expression of the GLI genes. We tested by RT-PCR seven glial tumors and primitive neuroectodermal tumors (PNETs), including those from the cerebellum (medulloblastomas), because the latter have been shown previously to harbor PTCH1 mutations, suggesting the activation of the SHH-GLI pathway (Wolter et al., 1997; Raffel et al., 1997). We found that all the tumors tested expressed GLI1, although at different levels (Fig. 6A). Additional analyses for a total of 22 tumors (Fig. 6B) showed that all samples contained the three GLI transcripts. Expression of PTCH1 followed the expression of GLI1/2, while that of SHH was not consistently detected (Fig. 6B).

In situ hybridization analyses of 22 independent brain tumors showed that GLI1 was expressed in neuronal and in glial tumors, including glioblastoma multiforme and low grade glioma (Fig. 6C,E). GLI1 and PTCH1 mRNAs were detected in the regions containing tumor cells and not in surrounding normal tissues (Fig. 6E,G,F,H).

Analyses of human brain tumor cell lines, including seven glioblastoma (U87MG, U118MG, U138MG, A172, T98G, M059K, M059J), two glioma (Hs683, mouse GL261), one neuroglioma (H4), three astrocytoma (CCF-STTG1, SW1088, SW1783), three medulloblastoma (Daoy, D283, D341) and two neuroblastoma (SK-N-AS, IMR32) lines, showed that all brain tumor cell lines co-expressed GLI1, GLI2 and PTCH1 (Fig. 6I). GLI3 was expressed by all but one (D341) and only a subset (U87MG, U138MG, Daoy, M059K, SW1783) expressed SHH (not shown).

As a control, we also tested a panel of unrelated sporadic human tumors by RT-PCR and found that GLI1 was expressed consistently in prostate carcinomas (9/11 cases) but not in those from the breast (1/7), suggesting that prostate cancer may also result from deregulated SHH-GLI signaling.

Expression of the GLI and PTCH1 by glioma cells raised the possibility that these harbor mutations that activate the pathway at different levels. Indeed, we expected that only a fraction of these possible mutations would affect the PTCH-SMO receptor complex. To address this possibility we have tested the effects of cyclopamine, a drug that inhibits the function of oncogenic Smoothened forms (Taipale et al., 2000). The glioblastoma/glioma lines U87, U118, U138, M059K, Hs683, C6, GL261, astrocytoma lines, SW1088 and SW1783, and the medulloblastoma line Daoy were tested and four responded to cyclopamine treatment by decreasing BrdU incorporation by ∼25-50% (Fig. 6J). These are the glioma lines U87, U118 and U138, and the medulloblastoma line Daoy: untreated U87, 22.5±1.4% BrdU-positive cells/field; 0.5 μM cyclopamine treated, 16.5±0.9% BrdU-positive cells/field, P<0.005; and 5 μM cyclopamine-treated, 13.2±1.2% BrdU-positive cells/field, P<0.001; untreated U118, 12.9±1.3% BrdU-positive cells/field; 0.5 μM cyclopamine treated, 6.7±0.8% BrdU-positive cells/field, P=0.001; and 5 μM cyclopamine treated, 7.1±1.1% BrdU-positive cells/field, P<0.005; untreated U138, 13.3±0.7% BrdU-positive cells/field; 0.5 μM cyclopamine treated, 9.5±0.6% BrdU-positive cells/field, P<0.005; and 5 μM cyclopamine-treated, 7.5±1.2% BrdU-positive cells/field, P=0.001; untreated Daoy, 37.9±2% BrdU-positive cells/field; 0.5 μM cyclopamine treated, 27.7±1.5% BrdU-positive cells/field, P=0.001; and 5 μM cyclopamine-treated, 27.1±1.7% BrdU-positive cells/field, P=0.001. While it is unclear why these four lines respond differently, these results show that their proliferation is modulated by cyclopamine-sensitive targets. Non-responsive cells could have mutations that affect the activation of the pathway downstream of the receptor complex.

Cyclopamine was also tested in three primary cortical gliomas that were dissociated and cultured in vitro. Dividing cells from all three tumors expressed vimentin (not shown), which marks neural precursors in culture (Palmer et al., 1999) among other cell types. These cells also expressed GLI1 and GLI2 but not Shh. Treatment with 5 μM cyclopamine resulted in the inhibition of BrdU incorporation in one of them by ∼60% (untreated tumor 3, 11.4±1.5% BrdU-positive cells/field; cyclopamine treated, 3.6±0.5% BrdU-positive cells/field, P<0.001), while the other two were unresponsive (untreated tumor 4, 4.5±0.5% BrdU-positive cells/field; treated, 4.6±0.5% BrdU-positive cells/field, P>0.9; and untreated tumor 5, 2.4±0.7% BrdU-positive cells/field; treated, 3.2 ±0.3% BrdU-positive cells/field, P=0.3).

The results with brain tumors and cell lines suggest that the deregulated SHH-GLI pathway may be involved in abnormal proliferation. To directly test this idea, we have misexpressed Gli1 in the CNS of the developing frog embryo. Tadpoles expressing Gli1 after unilateral injections developed ipsilateral neural tube hyperplasias (Fig. 7A; 24/36 embryos), first detected at tailbud stages, that expressed the β-gal lineage tracer (Fig. 7C; 15/15 embryos). Most hyperplasias appeared in the hindbrain and spinal cord, consistent with the more frequent distribution of the injected materials in these areas, and showed an increase in the number of BrdU-positive cells (more than fivefold: 19±1 BrdU-positive cells on average per control neural tube side in three sections counted of independent control embryos and 91±5.6 BrdU-positive cells per injected neural tube side in three sections of independent Gli1-injected tadpoles; P<0.005) when compared with the normal, uninjected contralateral side where BrdU-positive cells were confined to the vz zone (Fig. 7D). As expected, abnormal tissue contained HNF-3β-positive floor plate cells and neurons (Fig. 7B and not shown) (Lee et al., 1997; Ruiz i Altaba, 1998) but a large proportion of the hyperplastic masses had an undifferentiated appearance. In cases where the injected Gli1 RNA localized to the epidermis, the resulting skin hyperplasias or BCC-like tumors (Dahmane et al., 1997) also showed a marked increase in the number of BrdU-positive cells (not shown) over that seen in the normal, contralateral epidermis. Gli2 or Gli3 did not have these effects because they induced ectopic mesoderm earlier (Brewster et al., 2000).

The expression of Gli1 in glial tumors, together with the involvement of the SHH-GLI pathway in the development of glial lineages (Pringle et al., 1996; Poncet et al., 1996; Orentas et al., 1999; Nery et al., 2001), raised the possibility that Gli1-induced hyperplasias could contain glial precursors. Pdgfrα labels oligodendrocyte and other precursors (Pringle and Richardson, 1993) and is a marker of gliomas (Smits and Funa, 1998). In tadpoles, Pdgfrα is first expressed within the CNS in a small group of bilateral cells of the central vz of the diencephalon (Fig. 7E), overlapping Gli1 expression (not shown). Gli1-injected tadpoles showed ectopic expression of Pdgfrα within the hyperplasias (Fig. 7F; 20/45). Control uninjected tadpoles (0/60) failed to show ectopic expression of Pdgfrα.

To test if these hyperplasias include precursor cells with vz character, we have used the expression of endogenous Gli1 as a marker, as this gene is only transiently expressed in vz cells in normal development (Lee et al., 1997). Sectioning human GLI1 RNA-injected embryos after in situ hybridization with a frog Gli1 probe (which does not cross-hybridize with the injected, exogenous human RNA under the conditions used) (Dahmane et al., 1997) showed that the majority (12/16 embryos) had unilateral hyperplasias ectopically expressing endogenous Gli1 (Fig. 7G,H).

The expression of endogenous GLI1 in tumors and in induced tadpole hyperplasias (Fig. 6, Fig. 7G,H) raised the possibility that its function is required in tumor formation. This could be consistent with the observation that abnormal growths in the CNS or epidermis were first detected towards the late neurula-early tailbud stages, when most of the injected material had already been degraded. To test if endogenous Gli1 is required for tumor development, we injected human GLI1 RNA along with a morpholino antisense oligonucleotide specific for the endogenous frog Gli1 mRNA that does not recognize the injected human GLI1 RNA. Injection of human GLI1 and lacZ RNAs resulted in the development of both CNS and skin hyperplasias (Fig. 7I; 35/36). By contrast, co-injection of these same RNAs plus morpholino anti-frog Gli1 resulted in normal development without tumor formation (Fig. 7J; n=1/83). This effect is specific, as morpholino anti-frog Gli2 (n=47/52), morpholino anti-frog Shh (n=48/50) or a control unrelated morpholino (n=49/49) did not prevent tumor development by co-injected human GLI1 (not shown).

The results presented here demonstrate that the SHH-GLI pathway has a general role in controlling progenitor cell number in the developing dorsal brain, and contrast with its earlier role in ventral neural tube patterning. The approx. twofold changes in proliferation we detect in neocortical and tectal precursor after manipulating SHH signaling are similar to those we detected in the cerebellum (Dahmane and Ruiz i Altaba, 1999), indicating that SHH signaling modulates the normal growth of the three major dorsal brain regions that are layered and evolutionarily plastic. Additionally, our findings demonstrate that deregulation of SHH-GLI signaling in the CNS leads to hyperproliferation of precursor cells and suggests its involvement in the initiation and maintenance of brain tumorigenesis.

Shh is the only Hh family member reported to be normally expressed in the mammalian CNS (Echelard et al., 1993; Traiffort et al., 1999), raising the question of the localization of its sources that affect dorsal brain development. SHH is abundantly expressed in ventral brain regions throughout embryogenesis and is required for ventral development (Chiang et al., 1996). By contrast, early dorsal neural tube development requires the absence of SHH as dorsal cells, including those in the prospective cerebral cortex and tectum, can be ventralized by SHH (Roelink et al., 1994; Ericson et al., 1995; Kohtz et al., 1998; Watanabe and Nakamura, 2000; Agarwala et al., 2001). However, forebrain competence for ventralization is lost by ∼E11.5 in mice: SHH does not repress the cortical markers Emx1 or Tbr1 in E17.5 explants (not shown) as it does in similar ∼E10-11.5 explants (Khotz et al., 1998). After this early period, there is a change in the response to SHH. We show that Shh is expressed dorsally and that it is an endogenous late embryonic and postnatal mitogen modulating dorsal brain growth. From E14.5 to E17.5, we cannot localize the expression of Shh by in situ hydridization in the dorsal brain. As Shh may be expressed by layer V cortical neurons (Fig. 1) (Traiffort et al., 1999) and these are born at ∼E13-E14, it is possible that their precursors already express Shh in the vz, where they may affect neighboring cells. It is also possible, however, that as in the cerebellum (Dahmane and Ruiz i Altaba, 1999), Shh is expressed transiently by precursor cells, which later become dependent on SHH secreted from mature neurons at a distance.

SHH could also have survival functions, although TUNEL assays in our cortical explants did not show obvious differences between treated and untreated samples (not shown). Moreover, this effect is first detected at fivefold higher concentrations than those used here (Miao et al., 1997; Oppenheim et al., 1999).

Our results differ from those obtained in postnatal mice after SHH misexpression in the early forebrain using recombinant viruses, in which infected cells only become oligodendrocytes (Nery et al., 2001). This difference, however, may result from the fact that retroviral infections were performed at ∼E9.5, when SHH induces oligodendrocyte differentiation in early forebrain cells (Zhu et al., 1999) but does not yet act as a dorsal mitogen. Later on, SHH continues to induce oligodendrocytes in the ventral forebrain (Tekki-Kessaris et al., 2001). In addition to being a mitogen for precursors, our results therefore raise the possibility that endogenous SHH is involved in late embryonic and postnatal dorsal oligodendrogenesis.

Our results also differ from those obtained after misexpression of SHH in the dorsal spinal cord, in which the SHH-induced an approx. twofold increase in the proliferation of early precursors ceases before E18 (Rowitch et al., 1999). Thus, the differential growth of the dorsal regions of the brain and spinal cord could be related to the inability of the latter to proliferate in response to SHH.

The SHH-GLI pathway may not only modulate growth, and thus size, but also the shape of the brain. This idea derives from the proposal that differential use/localization of SHH would be responsible for the size and foliation patterns of the cerebellar cortex (Dahmane and Ruiz i Altaba, 1999) and from results in the ventral midbrain (Agarwala et al., 2001). Our experimental findings and the differential expression of Gli genes in the neocortex and tectum now allow us to extend this proposal to the two other major dorsal brain structures. It remains possible that SHH also regulates hippocampal cell proliferation. Indeed, changes in the spatial and/or temporal regulation of the SHH-GLI pathway within the dorsal brain could underlie the differential growth of the neocortex, tectum and cerebellum during evolution, as the sizes and shapes of these three structures vary enormously in phylogeny. A doubling of cortical cell proliferation could account for the development of the large primate neocortex (Rakic, 2000).

The late layer-specific expression of Shh in the dorsal brain (Fig. 1) suggests that it regulates Gli1⁺ precursor proliferation in germinative zones. There appears to be, therefore, a common mechanism by which SHH secreted from early differentiated neurons in the neocortex, tectum and cerebellum regulates precursor proliferation and, thus, the number, of later-born cells. This system would allow the independent growth of each dorsal structure during evolution by changing the region-specific action of SHH or its response. In this sense, the action of additional signals, such as BMPs and PACAP (Li et al., 1998; Zhu et al., 1999; Suh et al., 2001), could affect growth by antagonizing the proliferative effects of SHH.

The robust induction of Gli1 by SHH may pinpoint its primacy in the mediation of SHH signals (Lee et al., 1997; Hynes et al., 1997). However, because Gli1 null mice appear normal (Park et al., 2000), Gli2/3 could compensate for the loss of Gli1. Consistent with this, all Gli proteins have neurogenic activity (Brewster et al., 1998), and Gli3 mutant mice have smaller and disorganized cortices (Franz, 1994; Theil et al., 1999; Toole et al., 2000). Nevertheless, in the dorsal brain, as in the early embryonic neural tube (Ruiz i Altaba, 1998; Litingtung and Chiang, 2000), an antagonism between Gli3 and SHH/Gli1 may be crucial for normal development.

In addition to sporadic BCCs and PNETs (Dahmane et al., 1997; Raffel, 1997; Xie et al., 1998; Reifenberger et al., 1998), we have found that two new types of sporadic human tumors, glial brain tumors and prostate carcinomas, consistently express the GLI genes. This expression may reflect their site of origin or the types of cells affected. For example, in the GLI1-positive follicle or basal layer of the skin, maintained or ectopic GLI function could give rise to a BCC; in the GLI1-positive prostate (Podlasek et al., 1999), to carcinoma development; in the GLI1-positive cerebellum, to a medulloblastoma; in the GLI1-positive svz of the perinatal neocortex or striatum; and in adult regions where GLI1-positive precursors reside (such as the adult striatal svz (N. D., D. Lim, A. Álvarez-Buylla and A. R. A., unpublished) (Traiffort et al., 1999; Doetsch et al., 1999)), to a glioma.

Beyond being a marker of the origin of tumors, GLI expression is likely to reflect a participation of deregulated SHH-GLI signaling in tumorigenesis: GLI proteins may control precursor proliferation in many organs including the prostate and brain and their deregulation could lead to tumor formation. Support for this idea in brain tumors derives from the ability of cyclopamine to inhibit the proliferation of several human glioma cells, suggesting that these harbor mutations in the SMO-PTCH receptor (Taipale et al., 2000), that provoke constitutive signaling. Further support derives from our finding that the targeted, transient, somatic misexpression of GLI1 is sufficient to initiate a hyperplastic program in tadpoles, which could also occur in transgenic mice (Hynes et al., 1997; Park et al., 2000), that is dependent on endogenous Gli1 function. Moreover, mutations in PTCH1 and SMO have been detected in non-medulloblastoma PNETs (Vorechovsky et al., 1997; Wolter et al., 1997; Reifenberger et al., 1998) but not yet in gliomas. Nevertheless, why individuals with Gorlin syndrome heterozygous for PTCH1 develop some but not all tumor types associated with SHH-GLI signaling remains unclear. Interestingly, the consistent expression of the GLI genes in nearly all human primary brain tumors tested, as well as in all tested tumor cell lines, together with the action of cyclopamine, raises the possibility that SHH-GLI signaling also involved in tumor maintenance. If so, the viability of many human tumors in the brain and other organs could be based on persistent GLI function, thus providing an avenue for rational therapies.

We thank Moses Chao, Arturo Álvarez-Buylla, Sonya Baik, José Mullor and Gord Fishell for comments on the manuscript, the NYU tumor banks and D. Zagzag for specimens and for GL261 cells, M. Mercola for the probe for frog Pdgfrα, and B.Miller for O4 monoclonal antibody. A. R. A. and H. W. are members of the Kaplan Cancer Center. P. S. and V. P. were recipients of Fundación Ramón Areces and Pew Latin American Fellowship postdoctoral grants, respectively. H. W. was a recipient of awards from the American Brain Tumor Association/Emily Dorfman Foundation for Children and the Children’s Brain Tumor Foundation. This work was supported by grants from the NIH-NINDS, NCI, the March of Dimes and The Pew Scholars Program to A. R. A.