Nmyc upregulation by sonic hedgehog signaling promotes proliferation in developing cerebellar granule neuron precursors

Signaling by the hedgehog family of secreted glycoproteins has been highly conserved through evolution (Ingham and McMahon, 2001). Hedgehog activity is required for fundamental processes during vertebrate development, including determination of the embryonic left-right axis and limb polarity as well as induction and patterning during organogenesis. Additionally, sonic hedgehog (Shh) signaling regulates proliferation of the developing lung(Bellusci et al., 1997),pancreas (Kim and Melton,1998), retina (Jensen and Wallace, 1997) and central nervous system (CNS)(Dahmane and Ruiz-i-Altaba,1999; Kalyani et al.,1998; Wallace,1999; Wechsler-Reya and Scott,1999). Proliferative effects of Shh have been especially well characterized in the external granule layer (EGL) of the cerebellum, where expansion of granule cell precursors is dependent upon Shh signaling during early postnatal development (Dahmane and Ruiz-i-Altaba, 1999; Wallace,1999; Wechsler-Reya and Scott,1999).

Cerebellar granule neuron precursors (CGNPs) are generated in the rostral hindbrain during late embryonic development. They then migrate dorsally, where a second phase of postnatal proliferation takes place in the external granule layer of the cerebellum (EGL) (Altman and Bayer, 1997). Proliferating CGNPs express Math1, a bHLH transcription factor required for development(Ben-Arie et al., 1997;Helms and Johnson, 1998). Mature granule cells express other markers, including NeuN and the zinc finger transcription factor Zic (Aruga et al.,1994; Wechsler-Reya and Scott,1999). Postmitotic granule precursors migrate to their final destination in the internal granule layer (IGL), where they undergo terminal differentiation (Altman and Bayer,1997).

Shh pathway activation is implicated in the etiology of cerebellar tumors in humans as well as other types of cancer, and patched (PTCH), an inhibitory component of hedgehog signaling, has been identified as a tumor suppressor (Ruiz i Altaba,1999). Individuals with Gorlin's syndrome, resulting from mutations of PTCH, have high rates of basal cell carcinoma and medulloblastoma (Hahn et al.,1996). Medulloblastoma, which is thought to derive from cerebellar granule cell precursors, is one of the most common solid tumors in children(Provias and Becker, 1996). Mutations of PTCH have also been found in 10-20% of sporadic medulloblastomas (Pietsch et al.,1997; Xie et al.,1997) and mice heterozygous for targeted mutations ofPtch develop cerebellar tumors(Goodrich et al., 1997). However, the intracellular mechanisms that underlie Shh effects on cell cycle progression during cerebellar development and tumorigenesis are poorly understood.

Shh is thought to bind to a receptor complex composed of the transmembrane proteins Ptch and smoothened (Smo). This in turn relieves Ptch-mediated inhibition of Smo activity (Ingham and McMahon, 2001). Smo, a member of the serpentine G-protein-coupled receptor family (van den Heuvel and Ingham, 1996), is thought to activate an inhibitory G protein(DeCamp et al., 2000). Consequences of Shh signaling, in vivo and in vitro, can be inhibited experimentally by increasing cAMP levels or protein kinase A activity(Dahmane and Ruiz i Altaba,1999; Fan et al.,1995; Hammerschmidt et al.,1996; Kenney and Rowitch,2000; Wallace,1999; Wechsler-Reya and Scott,1999), or by treatment with the alkaloid, cyclopamine and related compounds (Berman et al., 2002;Dahmane et al., 2001;Incardona et al., 1998).

Smo signaling in Drosophila leads to post-translational activation of Cubitus interruptus protein, a member of the Gli family of zinc-finger transcription factors (Ruiz i Altaba,1999). In vertebrates, Gli family members appear to be general targets of hedgehog signaling. Gli1 behaves as a transcriptional activator,whereas Gli3 functions as a repressor of hedgehog signaling(Dai et al., 1999;Sasaki et al., 1999). Although Gli2 appears to have distinct activator and repressor properties(Ruiz i Altaba, 1999;Sasaki et al., 1999),Gli1 can fully compensate for Gli2 function during CNS development (Bai and Joyner,2001). Gli proteins are implicated in regulation of neural proliferation (Ruiz i Altaba,1999), and GLI1 was originally identified as a locus occasionally amplified in human gliomas(Kinzler et al., 1987). However, cerebellar proliferation is apparently normal in Gli1knockout mice (Park et al.,2000), indicating that additional factors must mediate proliferative effects of Shh.

Identification of Shh signaling intermediates that function in cell cycle regulation is key to understanding the role of this pathway in CNS proliferation during development and tumorigenesis. Shh treatment of CGNPs results in rapid upregulation of D-type cyclin genes, via a mechanism requiring protein synthesis (Kenney and Rowitch, 2000). Many mitogens signal to the cell cycle machinery by inducing expression of classical immediate-early genes (IEGs), such as members of the Fos, Jun and Myc families. Mitogen induction of IEGs occurs as a direct result of signaling pathway activation and does not require new protein synthesis (Sheng and Greenberg,1990).

We report that Shh signaling in CGNPs induces the proto-oncogene Nmyc in a protein synthesis-independent manner. Of the Myc family members tested, onlyNmyc was expressed in the proliferative zone of the cerebellum,suggesting a potential role for Nmyc in CGNP proliferation in vivo.Nmyc upregulation also occurred in proliferating neural precursor cells exposed to ectopic Shh in transgenic mice and in medulloblastomas ofPtch heterozygotes. Nmyc overexpression in vitro promoted CGNP proliferation and cyclin D1 protein upregulation in the presence of the Shh signaling inhibitor cyclopamine. Finally, antagonism of Nmyc activity significantly decreased CGNP proliferation induced by Shh. Together, our data provide compelling evidence that Nmyc functions downstream of Shh to promote cell cycle progression in cerebellar neuronal precursor cells during development and tumorigenesis.

Primary cultures of postnatal day (P) 4/5 mouse cerebellar granule cells were established precisely as described(Kenney and Rowitch, 2000),including an overnight adherence period in 10% fetal calf serum. To generate monolayer cultures enriched in CGNPs, cells were pre-plated twice for 30 minutes on 35 mm dishes coated with poly-D-lysine, then strained through a 70μm nylon cell strainer (Falcon) before final resuspension and counting(Hatten, 1998). Quantitation of GFAP immunostaining revealed that glial cells accounted for less than 10% of cells in cultures enriched in this manner. Cells were resuspended at 3×10⁶ cells/ml. One ml/well or 0.2 ml/coverslip was plated on poly-L-ornithine (Sigma)-treated glass cover-slips or six-well tissue culture plates (Falcon). For cycloheximide experiments,cells were `rested' for 1 hour after the adherence period by incubation in DMEM-F12 with antibiotics alone to allow for downregulation of serum-stimulated intracellular signaling pathways. `resting' medium was then exchanged for DMEM-F12/N2 supplement/25 mM KCl/antibiotics and Shh (3μg/ml) or Shh vehicle±cycloheximide (10 μg/ml, Sigma). Shh was obtained from Biogen (Cambridge MA).

Cells destined for retroviral infection were prepared as above, except that Shh (1.5 μg/ml) was included during the adherence period. For 24 hours before infection, CGNPs were incubated in serum-free medium containing Shh (3μg/ml) or Shh vehicle (indicated in figure legends). For infection,conditioned medium was removed and saved, and cells were exposed to freshly thawed retroviral supernatants for 2-3 hours, with intermittent rocking. Supernatants were removed and conditioned medium, with treatments added as indicated, was replaced. Alternatively, fresh medium was used with new Shh. Control cultures were treated with non-infectious, conditioned packaging cell medium, or UV-treated retroviral supernatants during the infection period. During analyses, no differences were observed between results obtained from cells cultured under these conditions. Cyclopamine, provided by William Gaffield (USDA, Albany, CA), was used at 1 μg/ml, as recommended.

All proliferation assessment experiments were repeated using pups from at least two separate litters. For immunochemistry experiments, two coverslips per treatment/infection were used, for each litter. Five fields of 100 cells per coverslip were analyzed by an unbiased observer, using DAPI staining to select homogeneously distributed cell fields. Thus, micrograph data shown are representative of a total n=20, 100-cell fields. Data are portrayed as fold change from Shh-treated GFP-infected cultures. GFP infected cultures did not differ from uninfected Shh-treated cultures in levels of proliferation(not shown). For flow cytometry (see below), experiments were performed using pups from three separate litters. With each separate litter, two wells of a six-well plate were used for each treatment/infection. Data represented in graphs are from n=6 experiments. Error bars represent standard error of the mean. Significance (P>0.01 in comparison with Shh-treated uninfected/GFP-infected CGNPs) was determined using the two-tailedt-test (Excel software). Sub-G₁-DNA quantitation was performed on two samples (two wells at 36 hours post-infection) for each experimental treatment. For northern blot and immunoblot analysis, two wells per treatment were pooled to maximize recovery of RNA and protein. Experiments for those analyses were repeated with three separate litters.

Mouse Nmyc cassette (Wood et al., 2000), and ΔMB2 cassette were prepared using standard methods for site-directed mutagenesis. Mxi-SR was a kind gift of Ron Depinho (Dana-Farber Cancer Institute) and eGFP was purchased from Clontech. All cassettes were cloned into the pWZL retroviral vector (Jay P. Morganstern, Millenium Pharmaceuticals, Cambridge, MA). pWZL IRES-GFP vector was provided by Steve Lessnick (Dana-Farber Cancer Institute). 293 EBNA (Invitrogen) packaging cells were co-transfected with retroviral constructs, gagpol, and vesicular stomatitis virus G glycoprotein plasmids, using Fugene 6 transfection reagent (Roche). Packaging cells were re-fed 12 hours after transfection. Retroviral supernatants (4 ml) were harvested every 12 hours for 72 hours and kept at 4°C until they were pooled, filtered through 0.45 μm syringe filters, aliquoted and stored at-80°C until use.

Flow cytometric analysis of propidium iodide (PI) staining was used to determine levels of proliferation and cell death as described(Darzynkiewicz et al., 1997;Kenney and Rowitch, 2000),using a FACScan (Becton Dickinson) and Cellquest software (Becton Dickinson)for data acquisition. Modfit software (Verity Software House) was used for quantifying sub-G₁ DNA levels.

For S-phase assessment, CGNPs plated on coverslips were pulsed with 25μg/ml 8-bromo-deoxyuridine (BrdU) for 2 hours prior to fixation in 4%paraformaldehyde. Cell were washed in PBS and treated for 2 minutes with 2N HCl. Cells were then processed for immunocytochemistry using standard methods. Primary antibodies included mouse anti-BrdU (Becton-Dickinson), rabbit anti-GFP (Seedorf et al.,1999), mouse anti-GFAP (Sigma) and mouse-anti-NeuN (Chemicon). Fluorochrome-conjugated secondary antibodies were Cy-3 anti-rabbit, Cy-2 anti-mouse or FITC-conjugated anti-mouse (Jackson Immunoresearch Laboratories). Cells were co-stained with DAPI to label nuclei. Staining was visualized with a Nikon Eclipse E600 microscope. Images were captured using a SPOT 1 digital camera (Diagnostic Instruments) and processed using Adobe Photoshop 5.0 software.

Total RNA was prepared by CsCl gradient centrifugation and ethanol precipitation. Ten μg of each sample was electrophoresed through a 1%agarose formaldehyde gel and transferred to a Hybond-N+ membrane(Amersham-Pharmacia). ³²P-labeled probes used for northern blotting were Gli1 (Hui et al., 1994),Jun (Tom Curran, St. Jude Children's Research Hospital, Memphis TN), Fos(Charles Stiles, DFCI, Boston MA), Mxi, Myc, Nmyc and Lmyc (all provided by Ron DePinho, DFCI, Boston, MA). Math1 cDNA was a gift of Jane Johnson(University of Texas Southwestern Medical Center, Dallas TX). Mouse GAPDH cDNA probe was PCR amplified from a mouse cDNA library.

Protein lysates were prepared and quantified as described(Kenney and Rowitch, 2000;Matsushime et al., 1994). Each sample (25 μg) was separated on 8% or 12.5% SDS-polyacrylamide gels, then transferred to Immobilon PVDF (Millipore) membranes. Primary antibodies used were anti-cyclin D1 and anti-Nmyc (sc-450 and sc-751, Santa Cruz), and anti-β-tubulin (T4026, Sigma). Peroxidase-conjugated secondary antibodies included donkey anti-mouse (Jackson Immunoresearch Laboratories) and goat anti-rabbit (Pierce). Blots were developed using enhanced chemiluminescence(ECL) (Amersham-Pharmacia), according to the manufacturer's instructions. Chemiluminescent immunoreactivity was detected using Kodak X-OMAT X-ray film.

Neonatal cerebella (PN1, PN7 and PN15) from 1% formalin/PBS perfused SWB mice and Ptch adult heterozygotes(Goodrich et al., 1997) orShh-transgenic 12.5 dpc embryos(Rowitch et al., 1999) were drop-fixed in fresh 4% paraformaldehyde at 4°C. Samples were embedded in OCT and frozen 15-18 μm parasagittal sections were prepared on Superfrost plus slides. In situ hybridization on frozen sections was performed according to standard protocols (detailed protocol available upon request). Antisense digoxigenin-labeled riboprobes were generated from Myc, Nmyc and Lmyc constructs (provided by Ron DePinho, DFCI, Boston MA), Gli1 (provided by Alex Joyner, Skirball Institute of Biomolecular Medicine, NY) or cyclin D1(Ccnd1) (from Steve Elledge, Baylor College of Medicine, Houston TX)constructs.

Upregulation of G₁ cyclin mRNA in CGNPs by Shh requires synthesis of unidentified protein intermediates(Kenney and Rowitch, 2000). Here, we asked whether known IEG transcription factors might link Shh signaling to cell cycle progression. IEGs are operationally defined as genes that are (1) induced by drugs that perturb protein synthesis and (2)`superinduced' when protein synthesis is perturbed in the presence of growth factors, cytokines and other stimuli(Greenberg et al., 1986;Kelly et al., 1983;Lau and Nathans, 1985). The molecular basis of both the induction and super-induction responses of IEGs such as Fos and Myc family members has been well characterized(Greenberg et al., 1986;Lau and Nathans, 1987). The induction of IEG mRNA in the presence of protein synthesis inhibitors,including cycloheximide (CHX), occurs because IEG mRNA degradation depends upon IEG mRNA translation (Lau and Nathans, 1987). Constitutive low-level IEG transcription results in mRNA accumulation in the presence of protein synthesis inhibitors. The super-induction response results when factor-specific induction is combined with protein synthesis inhibition-mediated induction(Greenberg et al., 1986). To determine whether Shh signals to known classic IEGs in proliferating CGNPs, we examined the mRNA accumulation and super-induction responses of selected IEGs in CGNPs treated with Shh, in the presence/absence of the protein synthesis inhibitor cycloheximide (CHX).

As shown (Fig. 1A), 3 hours of CHX treatment alone resulted in increased levels of mRNA transcripts forMyc and Fos in CGNPs regardless of treatment, in keeping with original observations made in fibroblasts(Greenberg et al., 1986;Lau and Nathans, 1987). Cycloheximide-mediated accumulation of mRNAs for these genes indicates that they are transcribed in primary granule cell cultures, but it is evident from our analysis that their transcription is not specifically affected by Shh. Shh did not induce expression of Lmyc compared with vehicle-treated CGNPs(Fig. 1A), and Jun mRNA was undetectable in CGNP cultures (data not shown). Thus, Myc, Fos, Lmycand Jun are unlikely to be direct targets of Shh in proliferating CGNPs.

Gli1 is a known transcriptional target of the Shh pathway in CGNPs(Weschler-Reya and Scott, 1999) and Gli proteins can be activated by hedgehog activity (Ruiz i Altaba,1999). Gli1 was strongly induced after 3 hours of treatment with Shh alone (Fig. 1A). CHX treatment alone did not induce Gli1 mRNA, and sharply reduced the degree of Gli1 upregulation in the presence of Shh. These results indicate that Gli1 does not share the induction/super-induction response characteristic of IEGs. Newly synthesized Gli1 may regulate its own transcription(Dai et al., 1999). This could explain the antagonistic effect of CHX on Gli1 mRNA induction by Shh.

Unlike other genes tested, Nmyc expression was strongly upregulated by Shh in the presence or absence of CHX(Fig. 1A). CHX treatment alone resulted in a low level of Nmyc mRNA elevation, in the manner of classically defined IEGs (Fig. 1A). These data suggest that Shh signaling pathway activation induces Nmyc by a mechanism that does not require new protein synthesis. Proliferation in CGNPs reaches high levels by 24 hours of Shh treatment (Kenney and Rowitch,2000; Wechsler-Reya and Scott,1999). Shh effects on the cell cycle machinery are first observed after 3 hours of treatment (Kenney and Rowitch, 2000). We did not detect significant upregulation ofNmyc mRNA until 3 hours of treatment with Shh+CHX(Fig. 1A,B).

Cerebellar homogenate cultures are an established system to characterize Shh proliferative effects on CGNPs (Kenney and Rowitch, 2000; Klein et al., 2001; Wechsler-Reya and Scott, 1999). These heterogeneous cultures comprise ∼85%CGNPs, which express the bHLH transcription factor Math1(Ben-Arie et al., 1997;Helms and Johnson, 1998;Kenney and Rowitch, 2000), as well as glia and a small percentage of other cell types (collectively defined as `non-granule cell component'). Previous work indicates that CGNPs are the only cells in these cultures that proliferate in response to Shh(Dahmane and Ruiz-i-Altaba,1999; Kenney and Rowitch,2000; Wechsler-Reya and Scott,1999). Nonetheless, to minimize any contribution of non-granule cells, we enriched for CGNPs in cerebellar homogenates (see Materials and Methods), then assayed cultures prepared in this manner for Nmycupregulation by Shh. As shown in Fig. 1B, these enriched cultures exhibited protein-synthesis independent upregulation of Nmyc in response to Shh. These results confirm that Shh-induced Nmyc upregulation in cerebellar homogenate cultures is attributable to CGNPs, and rules out any significant contribution of the non-granule cell component.

Other neuronal cell types responding to extracellular stimuli typically upregulate IEG mRNA within minutes (Morgan and Curran, 1991; Sheng and Greenberg, 1990). By contrast, we observed a delay inNmyc upregulation, which is most probably due to our culture conditions (see Materials and Methods) that include a 1 hour `rest' period in factor-free medium before Shh treatment(Kenney and Rowitch, 2000). Cerebellar homogenates were initially maintained in serum-containing media to allow them to adhere and recover from harvest (Hatten, 1998). Although serum is not mitogenic for CGNPs, it is likely to contain factors that affect signaling pathways, which could antagonize or synergize with Shh activity. The resting period was thus used to minimize any potential contributions of serum-regulated pathways and reveal effects specific to Shh signaling. The timecourse we show for induction of Nmyc coincides with that of D-type cyclins in the absence of CHX(Kenney and Rowitch, 2000;Wechsler-Reya, 1999). Similarly, Gli1 upregulation was first detectable after 3 hours of treatment in the absence of CHX(Fig. 1C), and lasted for the duration of this time course analysis. In summary, 3 hours is the earliest time point at which we have observed concerted activation of the hedgehog pathway and effects on cell cycle.

Proliferation in CGNPs reaches high levels by 24 hours of Shh treatment(Kenney and Rowitch, 2000;Wechsler-Reya and Scott,1999). Cultures treated with Shh for 24 hours showed maintenance of Math1 (Atoh1 — Mouse Genome Informatics) expression(Fig. 1C), a CGNP-specific marker in the cerebellum required for granule cell development(Ben-Arie et al., 1997). Expression of Math1 was used to confirm the presence of immature CGNPs in heterogeneous Shh-treated homogenate cultures(Kenney and Rowitch, 2000;Wechsler-Reya and Scott,1999). Like Gli1, a classical Shh pathway target(Ruiz i Altaba, 1999),Nmyc was upregulated in CGNPs treated with Shh for at least 24 hours(Fig. 1C), indicating that the induction of Nmyc is not a transient response to Shh.

In mice, granule neuron precursor expansion in the EGL occurs primarily during the first 2 weeks of life (PN1-PN14)(Altman and Bayer, 1997). However, a role for Nmyc in this process has not been described, nor hasNmyc expression been characterized during expansion of granule neurons in vivo. We used in situ hybridization to determine expression ofNmyc, cyclin D1 and Gli1, an established Shh transcriptional target, in the PN7 cerebellum. As shown inFig. 2A-C, we observed co-expression of Nmyc, cyclin D1 and Gli1 in the EGL. To assess whether Nmyc is expressed in an appropriate developmental time course in vivo, we subjected cerebella from PN1-15 mice to analysis.Nmyc expression was readily detectable at PN1 and PN7(Fig. 2D,E). At PN7,Nmyc expression was restricted to the proliferating outer zone of the EGL (EGLa) and was not detected in the inner granule layer(Fig. 2E,G), where differentiated granule cells lie. Furthermore, at PN15, when CGNP proliferation has ceased, Nmyc mRNA transcripts were no longer detected (Fig. 2F). Other myc family members tested (Myc, Lmyc) were not specifically expressed in the EGL (Fig. 2H,I) at any stage, consistent with results in Fig. 1. Thus, the temporal-spatial expression pattern of Nmyc,but not other myc family members, precisely coincides with regions of hedgehog proliferative activity in the developing cerebellum

Gene over-expression has been employed to study hedgehog-stimulated proliferation in CNS precursors in cerebellar slice preparations(Solecki et al., 2001) and transgenic mice (Epstein et al.,1996; Rowitch et al.,1999). These methods, however, cannot distinguish primary versus secondary proliferative effects on surrounding cells. By contrast, monolayer cultures of CGNPs are suitable for the study of the cell-autonomous functions expected of transcription factors. Moreover, they minimize formation of aggregates, which can stimulate granule cell proliferation(Gao et al., 1991). Retroviral infection of monolayer cultures was the method we chose to investigate the role of Nmyc in Shh-regulated CGNP proliferation. Retroviruses exclusively infect proliferating cells (Roe et al.,1993), so we optimized conditions to enhance proliferation of CGNPs in vitro and minimize proliferation of non-CGNP cell types, such as glia and fibroblasts (see Materials and Methods). In particular, we observed that cultures grown in the presence of Shh for 24 hours before infection demonstrated high rates of infection with jellyfish green fluorescent protein(GFP)-carrying retroviruses (Fig. 3B) relative to cultures pre-treated with vehicle alone(Fig. 3A). To confirm selective infection of CGNPs under our conditions, primary CGNP cultures were infected with GFP-carrying retroviruses, then incubated in serum-free medium without Shh for 48 hours to ensure that infected cells exited the cell cycle and commenced differentiation. As shown inFig. 3C,D, fluorescence immunocytochemistry indicated co-localization of GFP with NeuN, a marker of post-mitotic granule cells (Wechsler-Reya and Scott, 1999), but not cells expressing the astroglial marker,glial fibrillary acidic protein (GFAP).

For proliferation studies, CGNP cultures were infected with retroviruses expressing either GFP, wild-type mouse Nmyc or mutant Nmyclacking the Myc box 2 domain (ΔMB2). The MB2 domain is required for myc interaction with TRRAP (McMahon et al., 1998), a co-factor essential for recruitment of histone acetyltransferase to target DNA. Cyclopamine, an alkaloid derived from the veratrum lily (Keeler, 1969),antagonizes hedgehog signaling during development in vivo(Incardona et al., 1998)because it prevents activation of Smo(Taipale et al., 2000). Cyclopamine also inhibits Shh-induced expansion of chick mid-brain precursor populations (Britto et al.,2002). Effects of cyclopamine on cycling CGNPs have not been previously investigated.

We used cyclopamine to assess whether (1) Smo inhibition blocks Shh-induced CGNP proliferation, (2) Nmyc lies downstream of Smo and (3) Nmyc can functionally rescue CGNP proliferation when Smo is inactivated. Proliferation was assessed 48 hours post-infection (∼80 hours post-plating). Fresh medium with new Shh was provided after the infection period. It is known that GCNPs will proliferate at high levels with Shh treatment for up to 2 weeks and that proliferation is maximal at 3-4 days post-plating(Wechsler-Reya and Scott,1999). Thus, our assay was performed within the time window during which Shh proliferative effects have been observed. Further, we confirmed upregulation of cyclin D1 proteins and ongoing effects of Shh on the molecular apparatus controlling cell cycle progression at these times(Fig. 4A,B). As shown inFig. 4A, S-phase levels in cyclopamine/Shh-treated cells were significantly reduced in comparison with Shh-treated CGNPs. Western blot analysis confirmed that Shh-treated CGNPs expressed higher levels of Nmyc protein than vehicle-treated cells(Fig. 4A), and that cyclopamine treatment reduced levels of Nmyc protein in Shh-treated control/GFP-infected CGNPs.

Using flow cytometry, we measured levels of proliferation in CGNP cultures that were treated with cyclopamine/Shh and infected with Nmyc- orΔMB2-carrying retroviruses. Retroviral protein expression was verified by western blotting (Fig. 4A). ΔMB2-infected, cyclopamine/Shh-treated CGNPs showed similar levels of cells in S-phase as GFP-infected Shh/cyclopamine-treated CGNPs (Fig. 4A). By contrast, Nmyc-infected cyclopamine/Shh treated cells maintained levels of proliferation similar to CGNPs exposed to Shh alone. These findings indicate that Nmyc expression is sufficient for CGNP proliferation in the absence of Smo-mediated hedgehog pathway activation. To confirm results of flow cytometry by an independent means, we measured proliferation in CGNP-enriched cultures by BrdU immunochemistry. As shown inFig. 4B, this analysis yielded similar results. Cyclopamine alone had no effect on the ability of Nmyc to promote proliferation in CGNPs (Fig. 4B). CGNPs infected with Nmyc+GFP and subsequently treated with Shh vehicle alone also showed levels of proliferation similar to cells infected with GFP and treated with Shh (data not shown). Infection withNmycΔMB2 and treatment with cyclopamine or Shh vehicle resulted in proliferation and cyclin D1 protein levels and similar to GFP-infected cells treated with Shh+cyclopamine (Fig. 4B and data not shown).

Nmyc might promote synthesis and release of secondary mitogens from infected cells. To determine whether proliferative effects of Nmyc on cell division were cell-autonomous, we used a retrovirus expressing bothNmyc and GFP (see Materials and Methods) to follow infected cells in Shh/cyclopamine-treated cultures. As shown inFig. 4C, when Shh/cyclopamine-treated cultures were infected with Nmyc+GFP virus,we observed immunolabeling for BrdU exclusively in cells that co-expressed GFP. No BrdU-positive, GFP-negative cells were observed, out of twenty 100-cell fields examined. These results suggest that Nmyc acts within individual CGNPs to promote cell cycle progression.

Although our culture conditions were optimized for preferential infection of CGNPs, it was formally possible that the few glial cells present may have been infected with the Nmyc retrovirus. However, in Nmyc+GFPretrovirus-infected cultures treated with Shh/cyclopamine, we observed that both BrdU and GFP labeling consistently segregated away from cells expressing the astroglial marker, GFAP (Fig. 4C). Furthermore, the similar proliferative effects of Nmyc virus(Fig. 4B) despite the lack of glial cells in pre-plated, CGNP-enriched cultures indicates that these cells are not required to support CGNP proliferation in vitro. Together, these data rule out that CGNP proliferation in Nmyc-infected cultures reflects any significant contribution from glial cells.

A hallmark of the CGNP response to Shh is upregulation of D-type cyclin mRNA and protein (Kenney and Rowitch,2000). Immunoblot analysis showed increased levels of cyclin D1 in Shh-treated cells, compared with vehicle treated controls(Fig. 4A,B). Cyclopamine co-treatment eliminated cyclin D1 protein expression in Shh-treated control cells and those infected with GFP or ΔMB2retroviruses. By contrast, Nmyc-infected CGNPs maintained cyclin D1 protein expression despite treatment with cyclopamine. These results suggest that a downstream consequence of Nmyc activity is upregulation of cyclin D1 protein expression.

To further investigate potential roles for Nmyc, we asked whether Nmyc is required for Shh-mediated CGNP proliferation. A well-established means of blocking Myc function in vitro is by overexpression of Mad family members(Ayer and Eisenman, 1993;Lahoz et al., 1994;Lee and Ziff, 1999;Wu et al., 1996). Mad proteins, including Mad1, 3, 4 and Mxi, repress Myc target genes(Baudino and Cleveland, 2001)by binding to E-box sequences and recruiting histone deactelylase. Mxi overexpression blocks Myc- and Nmyc-mediated, but not E1a-mediated transformation of fibroblasts (Lahoz et al., 1994).

As shown in Fig. 5A,endogenous Mxi is expressed in CGNPs in vitro. We infected Shh-treated CGNPs with a retrovirus carrying Mxi-SR, the strong repressor form of Mxi (Schreiber-Agus et al., 1995), then compared proliferation levels with control/GFP-infected CGNPs. Expression of Mxi-Sr is shown inFig. 5A. Thirty-six hours after infection, proliferation in CGNPs was assayed by flow cytometry.Mxi-infected CGNPs treated with Shh showed significantly reduced proliferation compared with controls (Fig. 5A). To determine whether this decrease in proliferation reflected cell death, we quantified levels of fragmented `sub-G₁' DNA, a hallmark of cell death that can be quantified by flow cytometry(Darzynkiewicz et al., 1997). The anti-proliferative effects of Mxi were not due to increased levels of cell death, as sub-G₁ DNA levels in Mxi-infected cells (5.3%±0.5,n=2) were similar to those in GFP-infected CGNPs (5.6%±0.7,n=2). These findings indicate that Mxi overexpression negatively regulates Shh-stimulated proliferation in CGNPs. Because Mxi represses Myc targets, and Nmyc is the only Myc family member expressed in CGNPs,our observations imply that Nmyc target gene activity is crucial for Shh to achieve its full mitogenic potential in CGNPs.

Although cell growth regulation targets substantially overlap between Myc and Mad (James and Eisenman,2002), several targets for activation by Myc are not targets for Mad-mediated repression, and vice versa(O'Hagan et al., 2000). We wished to confirm our findings using a more specific means of blocking Nmyc activity. To this end, we infected CGNPs with a virus expressingΔMB2+GFP and assessed levels of proliferation in these cells,compared with cells infected with GFP-carrying retroviruses alone. Proliferation was measured by quantitation of BrdU incorporation. These data indicated that ΔMB2 overexpression results in a significant decrease in Shh-induced CGNP proliferation (Fig. 5B). This suggests that overexpression of NmycΔMB2can compete with endogenous Nmyc in regulating growth-promoting target genes,resulting in reduced levels of proliferation. Together, the findings suggest that Nmyc activity is required to achieve the full proliferative response to Shh within CGNPs.

Our findings indicate that Nmyc acts a link between hedgehog pathway activation and cell cycle progression within CGNPs. Shh has proliferative effects elsewhere in the CNS (Dahmane et al., 2001; Jensen and Wallace,1997; Kalyani et al.,1998), and ectopic expression of Shh using theWnt1 regulatory element results in increased levels of neural precursor proliferation in the dorsal spinal cord of transgenic mice(Rowitch et al., 1999). We observed Nmyc upregulation in dorsal spinal cord of such transgenic embryos (Fig. 6D), suggesting that Nmyc is similarly upregulated in other CNS regions where Shh has mitogenic effects. These data further show that, in contrast to Gli1,Nmyc is not a general Shh transcriptional target. For example,Nmyc expression is absent in the floorplate, where Shh is strongly expressed (fp) (compare Fig. 6A,C with Fig. 6B,D). Second, it is clear that Nmyc can be activated by mitogens other than Shh. Nmyc expression is observed in proliferating progenitors in the dorsal half of wild-type spinal cord at 12.5 dpc (at a long distance from a source of Shh;Fig. 6B) and in the dorsal root ganglion (drg; Fig. 6B). Together with results shown in Figs1,2, these findings indicate thatNmyc can be recruited by Shh solely within immature, proliferating CNS precursor cells.

Mutation of the tumor suppressor Ptch is associated with abnormal activation of hedgehog signaling and generation of medulloblastoma(Goodrich et al., 1997). We performed in situ hybridization on sections of medulloblastoma fromPtch adult heterozygotes. As shown inFig. 6E, we observed dramatic induction of Nmyc in the tumors, in contrast to adjacent normal cerebellum. Indeed, the borders of the tumor appeared to be well demarcated byNmyc expression, which highlighted an area of continuity between the IGL and tumor that largely occupied the space dorsal and posterior to the cerebellum (Fig. 6F). These findings suggest that Nmyc upregulation is a consequence of pathological hedgehog pathway activation as well as normal Shh signaling during development.

Sonic hedgehog signaling is required for expansion of neuronal precursor cells in the EGL of the cerebellum, and Shh is the most potent known mitogen for CGNPs in vitro. Activation of Shh signaling in cultured CGNPs rapidly results in increased levels of G₁ cyclin mRNA transcripts/proteins(Kenney and Rowitch, 2000), by a mechanism that requires new protein synthesis. We have shown that Shh activates expression of the proto-oncogene Nmyc, without requiring intermediate protein synthesis. The hedgehog signaling pathway is highly conserved and no genetic or biochemical interactions with Myc family members have yet been thoroughly explored, although recently published expression profiling studies using oligonucleotide microarrays have suggested that Nmyc may lie downstream of Shh signaling in the developing cerebellum(Zhao et al., 2002) and medulloblastoma (Pomeroy et al.,2002). These studies raise important questions regarding the role of Nmyc as a novel member of the hedgehog proliferative pathway.

Of numerous IEGs tested by northern blotting, only Nmyc expression was upregulated by Shh in the presence of cycloheximide in CGNPs, suggesting its transcription is directly regulated by the Shh signaling pathway.Nmyc overexpression was sufficient for CGNP proliferation, despite treatment with the hedgehog pathway antagonist cyclopamine, which interferes with Shh signaling by affecting the activation state of Smo(Taipale et al., 2000). Cyclopamine has been shown to interfere with the expansion of chick midbrain precursors, a Shh-mediated process (Britto et al., 2002). Interestingly, cyclopamine also reduces proliferation of certain brain tumor cells(Dahmane et al., 2001), which do not express Shh but do express Shh signaling targets. More recently, use of cyclopamine to inhibit the hedgehog signaling pathway was shown to block growth of mouse medulloblastomas in vitro and in vivo(Berman et al., 2002). Our data indicate that cyclopamine blocks Shh-induced proliferation and reduces levels of Nmyc protein in CGNPs. Nmyc overexpression sustained CGNP proliferation, despite cyclopamine treatment, indicating that Nmyc function lies downstream of Smo in the Shh signaling pathway.

It seems likely that Nmyc is not the only Smo target with a role in regulating cell cycle progression. Investigation of the proliferative targets of Shh has often focused on the Gli transcription factors, orthologs of the Drosophila hedgehog target cubitus interruptus(ci). Increased levels of Gli1 mRNA transcripts are characteristic of Shh signaling pathway activation, and can be seen in CGNPs treated with Shh (Kenney and Rowitch,2000; Wechsler-Reya and Scott,1999). Our studies have not focused on a relationship between Gli activity and Nmyc regulation per se. Recently, a microarray analysis by Yoon and colleagues identified cyclin D2 as a Gli1 target in transformation assays (Yoon et al., 2002).Nmyc was not among the Gli1 targets described in that study.

Mutation of the DNA-binding domain of Gli1 is not reported to result in cerebellar defects (Park et al.,2000), indicating the existence of Gli1-independent mechanisms for regulating cerebellar expansion. It is possible that Gli2 activity or antagonism of Gli3-mediated repression promotes Shh-stimulated Nmyc expression. The regulation of these factors is thought to involve post-translational processing (Ruiz i Altaba, 1999; Wang et al.,2000), a process that would enable target gene activation in the presence of protein synthesis inhibitors. However, this mechanism may not involve the induction/super-induction property of IEG mRNA regulation. An intriguing alternative is that Nmyc up-regulation by Shh occurs via a Gli-independent mechanism.

Nmyc is a member of the myc proto-oncogene family, which also includes Myc and Lmyc. Activity of Myc proteins is known to promote proliferation and/or transformation in many cell types(Henriksson and Luscher,1996). Complete loss of Myc activity results in severely compromised cell cycle regulation (Mateyak et al., 1997). We found that Nmyc was highly expressed in the expansion zone of the EGL, where precursor proliferation is known to depend upon Shh activity. By contrast, we did not detect Myc orLmyc expression in the developing EGL, nor were they upregulated by Shh in CGNP cultures. These findings suggest an important role for Nmyc in proliferation of CGNPs. We observed Nmyc upregulation in CGNP cultures and the dorsal embryonic spinal cord exposed to Shh, suggesting that signaling via Nmyc may be a general feature of Shh-induced proliferation in the CNS. However, in contrast to known general transcriptional targets of Shh (e.g. Gli genes), we note that Shh signaling effects on Nmyc regulation are unique to immature, proliferating precursor cells. Thus, cellular competence is evidently a critical determinant of the transcriptional response of Nmyc to Shh.

Increased levels of MYCN mRNA have been reported in certain cases of human medulloblastoma (Garson et al.,1989). More recently, MYCN upregulation has been observed in the desmoplastic form of medulloblastoma, the specific subtype of medulloblastoma associated with pathological activation of the hedgehog pathway (Pomeroy et al.,2002). By contrast, MYCN expression was not elevated in the majority of other (non-desmoplastic) medulloblastoma types analyzed. Consistent with findings in humans, we observed dramatic Nmycupregulation in medulloblastomas from Ptch heterozygotes(Goodrich et al., 1997). Together, these findings are strong evidence that similar mechanisms operate downstream of hedgehog signaling during central nervous system development and tumorigenesis. Our results further suggest that MYCN expression in desmoplastic medulloblastoma is a direct consequence of hedgehog pathway activation.

To establish a role for Nmyc in the Shh proliferative pathway in developing cerebellar precursors, it is necessary to examine cell cycle progression in the absence of Nmyc function. However, mice homozygous forNmyc null or certain hypomorphic alleles mice die in utero or at birth (Charron et al., 1992;Moens et al., 1992;Sawai et al., 1991;Stanton et al., 1992),preventing assessment of Nmyc function during the proliferative postnatal phase of CGNP development. Recently, conditional targeting of murineNmyc in the developing CNS has been reported to result in severe hypoplasia of cerebellum and defects in granule cell precursor proliferation(Knoepfler et al., 2002). These findings indicate that Nmyc function is necessary for cerebellar development. Yet, they leave unanswered the question of whether Nmyc activity is required downstream of Shh signaling during the postnatal phase of granule cell expansion.

In studies of Myc biology, overexpression of Mad family members is an established method for inhibiting Myc functions in proliferation,transformation and apoptosis (Ayer and Eisenman, 1993; Hurlin et al.,1995; Schreiber-Agus et al.,1995; Wu et al.,1996). These studies indicate that the repressive activity of Mad family members is specific, and Mad proteins are unlikely to behave as general transcriptional repressors; indeed, Mxi overexpression cannot inhibit transformation by E1a, another potent oncogene(Lahoz et al., 1994). The ability of Mxi overexpression to abrogate the mitogenic effects of Shh in vitro raises the possibility that Mad family members may regulate CGNP cell cycle exit during cerebellar development in vivo. In keeping with this,expression profiling studies indicate upregulation of several Madfamily members at PN7, which coincides with the major wave of CGNP differentiation (Q. Zhao, A. Kho, I. Kohane and D. H. R., unpublished).

Our results imply that Mxi represses Nmyc targets, whose activity is crucial for Shh signaling to achieve its full proliferative capacity in CGNPs. To show that Nmyc activity is specifically involved in the proliferative response of CGNPs to Shh, we overexpressed a mutant Nmyc that lacks the Myc box 2 domain. This also resulted in a significant reduction in CGNP proliferation. Analogous mutants of Myc have dominant negative effects on transformation (Sawyers et al.,1992), and can attenuate entry into S-phase in fibroblasts(Conzen et al., 2000). The NmycΔMB2 mutant is incapable of interacting with TRRAP, a co-factor required for the transformation activity of Myc(McMahon et al., 1998). Although the NmycΔMB2 mutant lacks transforming activity, it can partially rescue cell cycle defects in Myc-null fibroblasts(Nikiforov et al., 2002). The activity of the NmycΔMB2 mutant, as well as requirements for TRRAP in neuronal precursors, have not been investigated. The inability of NmycΔMB2 to rescue proliferation in cyclopamine-treated CGNPs and its interference in Shh-induced CGNP proliferation raise the interesting possibility that neuronal precursors possess unique requirements for TRRAP activity in regulating proliferation-associated gene activation.

Although the levels of proliferation were reduced comparably with Mxi-SR or NmycΔMB2, in neither case did we observe complete cessation of proliferation. This is not surprising, as both of these methods allow for some residual Nmyc activity. When Myc activity is reduced, but not eliminated,Myc-dependent processes can still occur(Bazarov et al., 2001). Indeed,cultured fibroblasts that lack any Myc family member remain capable of proliferation albeit with an extended cell cycle length(Mateyak et al., 1997). Thus,abrogation of Nmyc activity might not necessarily lead to cell cycle exit within the time frame of our studies. Taken together, our findings of decreased CGNP proliferation with specific Nmyc antagonists in vitro, coupled with data that demonstrate essential functions for Nmyc during cerebellar development in vivo (Knoepfler et al., 2002), strongly suggest that Nmyc functions as an integral component of a Shh regulated pathway during proliferation of granule neuron precursors.

Myc proteins are characterized by specific functional domains, including a C terminus DNA-binding domain and a helix-loop-helix/leucine zipper domain,which mediates dimerization with the obligate partner of Myc, Max [Myn in the mouse (Prendergast et al.,1991] (Henriksson and Luscher,1996). Two highly conserved amino acid regions, Myc box 1 and Myc box 2, reside in the N-terminal transactivation domain. Myc box 2 associates with TRRAP protein (McMahon et al.,1998), which functions to recruit histone acetyltransferase to the transactivation domain of Myc (Park et al., 2001). This interaction is required for the transforming potential of Myc and Nymc(McMahon et al., 1998). The Myc box 2 domain also appears to be required for the proliferative function of Nmyc in CGNPs, as overexpression of an Nmyc mutant lacking Myc box 2 was unable to rescue CGNP proliferation in the absence of Shh signal, and interfered with CGNP proliferation in the presence of Shh.

It is likely that Nmyc activates yet to be identified genes involved in cell cycle regulation. Recently, Id2 was shown to be an Nmyc target in neuroblastoma (Lasorella et al.,2002). Some Myc targets include genes such as cyclin D2(Bouchard et al., 1999) andCdc25 (Galaktionov et al.,1996), which control cell cycle progression. Metabolic regulators that enhance cell growth may also be targets of Myc(Dang, 1999). Finally, Myc has been implicated in repression of genes that promote cell cycle exit and differentiation (Claassen and Hann,1999). Given the high level of relatedness between Nmyc and Myc protein structure, and the observation that Nmyc can rescue many of the defects in Myc null mutant mice (Malynn et al., 2000), it is possible that they have similar transcriptional targets.

We observed increased levels of cyclin D1 in Nmyc-infected cells despite the presence of cyclopamine. Nmyc is a direct transcriptional target of Shh signaling, whereas cyclin D1 mRNA upregulation was inhibited by cycloheximide (Kenney and Rowitch,2000). Nmyc could be a candidate regulator of cyclin D1 expression. However, our present cyclin D1 protein measurements were taken 48 hours after infection and treatment, too long a length of time to assay for direct activation of Nmyc target genes. Nmyc may act indirectly, via activation of genes that otherwise promote cell growth, or by repressing genes that control the degradation of cyclin D1 message.

In summary, our results indicate that Nmyc is a direct downstream target of the canonical Shh signaling pathway in proliferating CGNPs.Nmyc expression was also found in medulloblastomas of Ptchmutant mice. Nmyc function is a key component of a Shh proliferative pathway essential for normal expansion of CGNP populations. These findings are the first to identify a direct connection between the Shh signaling pathway and a cell-intrinsic regulator of the cell cycle apparatus in primary neuronal cells. Proximal events in the progression from Smo activation to upregulation of Nmyc expression remain to be determined. Identification of Nmyc transcriptional targets will be crucial for understanding Shh-mediated cell cycle progression in CNS development. Moreover, because cerebellar granule cells are postulated to be the cell-of-origin for medulloblastoma, effective means to inhibit Nmyc activity might provide new approaches to control the growth of cerebellar tumors.

The authors thank Paul Knoepfler and Robert Eisenman for sharing data prior to publication. Sovann Kiang, Dongin Yuk and the DFCI core flow cytometry facility staff provided technical assistance. John Kim (Children's Hospital,Boston, MA) kindly contributed samples from Ptch mutant mice. We also appreciate the generosity of Ron DePinho, Steve Lessnick and Ronan O'Hagan in freely giving reagents and helpful advice. Roz Segal, Chuck Stiles and Hans Widlund patiently and critically reviewed our manuscript. A. M. K. has been supported by the Justin Porter postdoctoral fellowship (American Brain Tumor Association) and by a grant from the Medical Foundation. These studies were funded by grants to M. D. C. from the NCI and to D. H. R. from the NINDS (R21 NS41764-01, RO1 NS4051); and by the Edward Mallinkrodt, Jr Foundation, the National Multiple Sclerosis Society and the James S. McDonnel Foundation. D. H. R. is a Kimmel Scholar.