Non-canonical Hedgehog/AMPK-mediated control of polyamine metabolism supports neuronal and medulloblastoma cell growth

SUMMARY

Developmental Hedgehog signaling controls proliferation of cerebellar granule cell precursors (GCPs) and its aberrant activation is a leading cause of medulloblastoma. We show here that Hedgehog promotes polyamine biosynthesis in GCPs by engaging a non-canonical axis leading to the translation of ornithine decarboxylase (ODC). This process is governed by AMPK, which phosphorylates threonine 173 of the zinc finger protein CNBP in response to Hedgehog activation. Phosphorylated CNBP increases its association with Sufu, followed by CNBP stabilization, ODC translation and polyamine biosynthesis. Notably, CNBP, ODC and polyamines are elevated in Hedgehog-dependent medulloblastoma and genetic or pharmacological inhibition of this axis efficiently blocks Hedgehog-dependent proliferation of medulloblastoma cells in vitro and in vivo. Together, these data illustrate an auxiliary mechanism of metabolic control by a morphogenic pathway with relevant implications in development and cancer.

Graphical Abstract

INTRODUCTION

Sonic Hedgehog (Shh) pathway is a critical regulator of embryonic patterning and postnatal stem or progenitor cells (Northcott et al., 2012a). In the cerebellum, Shh promotes the postnatal mitotic expansion of cerebellar granule progenitors (GCPs) (Wechsler-Reya and Scott, 1999) and the aberrant activation of the signaling in these cells is a leading cause of Medulloblastoma (MB) (Schuller et al., 2008; Yang et al., 2008), the most frequent brain malignancy of the childhood.

Shh ligand binds to the membrane inhibitory receptor Patched (Ptch), thus alleviating its inhibitory activity upon the transmembrane transducer Smoothened (Smo). These events occur at the primary cilium and trigger a cascade of intracellular processes that involve the dynamic association between the Gli transcription factors (Gli1, Gli2 and Gli3) and Sufu, which in turn regulates their activity, processing and cellular localization (Ryan and Chiang, 2012).

The aberrant activation of Shh pathway observed in MB can been attributed to mutations or amplifications of genes encoding pathway components, such as Ptch, Smo, Sufu and Gli2 or other mechanisms (Schroeder and Gururangan, 2014).

Different small molecule inhibitors of the Shh signaling have been generated and tested, thus providing hope to MB patients. However, the majority of these compounds inhibits Smo activity, and is thus inactive in case of mutations occurring at downstream level. Additionally, trials in patients and animals with tumors driven by mutations of Ptch or Smo have shown that, despite a good initial response, they quickly develop resistance. For these reasons it is now believed that alternative approaches, preferably targeting downstream components of the signaling, are required (Di Magno et al., 2015).

Polyamines are small intracellular polycations that control key aspects of cell biology, such as cell replication, translation, cell growth, differentiation and survival and can be found upregulated in cancer (Casero and Marton, 2007). The metabolism of polyamines starts from the decarboxylation of ornithine to putrescine, a rate-limiting step catalyzed by ornithine decarboxylase (ODC), the crucial gatekeeper of polyamine metabolism. Putrescine is converted to spermine, which is in turn transformed to spermidine (Tavladoraki et al., 2012).

Given their critical role, the intracellular concentration of polyamines is kept under tight control by various mechanisms affecting their biosynthesis, catabolism and transport. The conversion of ornithine to putrescine represents the most critical regulated step. Indeed, the intracellular levels of ODC are promptly adjusted to the cellular needs thanks to different mechanisms affecting its protein stability, transcription and translation (Shantz and Levin, 2007). Overexpression of ODC and polyamines are found in a number of malignancies and targeting ODC with the irreversible inhibitor difluoromethylornitine (DFMO) is highly efficient in limiting some malignancies, such as lymphomas or neuroblastomas (Nilsson et al., 2005; Rounbehler et al., 2009).

With the aim of identifying downstream inhibitors of Hh signaling, we have found a mechanism whereby this pathway controls polyamine metabolism via an AMPK-dependent control of ODC biosynthesis. This mechanism involves the small zinc finger protein CNBP, which promotes translation of ODC, in association with the cytoplasmic Hedgehog (Hh) transducer Sufu. Finally, we show that CNBP and polyamine metabolism are necessary for Hh-dependent growth and that their inhibition can be exploited for treatment of Hh-dependent MB.

RESULTS

Hh-dependent proliferation is accompanied by an increase of polyamine biosynthesis and is inhibited by DFMO

To determine whether Hh activation is associated to changes in polyamine metabolism, we studied cerebellar granule cell progenitors (GCPs), the cells of origin of Hh-dependent MB (Schuller et al., 2008). Exposure of GCPs to Shh induced a significant elevation of the three polyamines, putrescine, spermine and spermidine, accompanied to a robust increase of cell proliferation (Figures 1A, B). The ODC inhibitor DFMO counteracted the increase of proliferation (Figure 1B), thus indicating that polyamine biosynthesis is required for GCPs proliferation. The anti-proliferative effect of DFMO could not be attributed to an inhibition of canonical Hh-dependent transcription, as the mRNA levels of the Hh targets Gli1 and Ptch were not changed by this drug (Figure S1A). To investigate if the Hh-induced elevation of polyamines was caused by upregulation of ODC, we measured its mRNAs and protein levels. Interestingly, ODC protein content was increased after treatment of GCPs with Shh, while ODC mRNA levels were unchanged, indicating an Hh mediated post-trascriptional regulation of ODC (Figure 1C and S1B). The increase of ODC protein and polyamine levels was also observed in mouse embryonic fibroblasts (MEFs) treated with the Smoothened agonist SAG (Figures S1C and S1D). To understand at what level of the Hh transduction pathway this regulation occurs, we used MEF cells deleted of genes encoding two key Hh inhibitors: Ptch and the cytoplasmic transducer Sufu (Svard et al., 2006). Surprisingly, while DFMO significantly inhibited Ptch^−/− MEF proliferation (Figure 1D, left), this drug failed to inhibit Sufu^−/− cell growth (Figure 1D, right). Consistently, addition of putrescine, the downstream metabolite derived from the ODC-dependent decarboxylation of ornithine, increased the proliferation rate only in Sufu^−/− cells, further indicating that the absence of Sufu impairs this metabolic step. Supporting this finding, we observed that the protein levels of ODC (Figure 1E, bottom and S1E) and of the three polyamines (Figure 1F) were significantly higher in Ptch^−/− than in Sufu^−/− cells, while ODC mRNA levels were not different in the two cell lines (Figure 1E, top). Gli1 protein levels were also comparable in both cell lines, indicating a similar degree of activation of the Hh-dependent transcriptional output (Figure 1E). Knockdown of ODC in Ptch^−/−MEF cells caused a significant decrease of cell proliferation and polyamine content (Figure S1F), confirming the importance of this enzyme for cell proliferation in this context.

Figure 1. Hedgehog (Hh) signaling promotes polyamine biosynthesis.

(A) Polyamine levels in P7 granule cell progenitors (GCPs) treated with Shh or BSA as control for 48 hours. *p<0.01 Shh vs BSA. The polyamine content was measured by GC-MS and normalized to the protein concentration in the extracts (nmol/mg protein). Results are expressed as fold change, relative to untreated cells for each polyamine and represent the average +/− SD of three independent experiments, each performed in triplicate. Put, putrescine; Spm, spermine; Spd, spermidine.

(B) BrdU incorporation assay on GCPs treated with DFMO (5 mM) for 72 hours and with Shh or BSA as control for the last 48 hours. *p<0.01 Shh vs BSA; **p<0.05 Shh DFMO vs Shh NT. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

(C) ODC mRNA levels normalized by GAPDH levels (top) and ODC protein levels (bottom) in P7 GCPs treated with Shh or BSA as control for 24 hours; actin, loading control. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

(D) Cell proliferation assay in Ptch^−/− (left) and Sufu^−/− (right) MEF cells treated with DFMO (5mM), putrescine (Put, 10μM) alone or in combination for 48 hours. *p<0.01 DFMO vs NT 48h; **p<0.05 DFMO+Put vs DFMO 48h; ***p<0.05 Put vs NT 48h. Results are shown as the average +/− SD of four independent experiments, each performed in triplicate.

(E) ODC mRNA normalized by GAPDH mRNA levels (top) and ODC protein level (bottom) from Ptch^−/− and Sufu^−/− cells; actin, loading control. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

(F) Polyamine levels in Ptch^−/− and Sufu^{−/ −} MEF cells. *p<0.05 Sufu^−/− vs Ptch^−/−. Results are expressed as fold change of polyamine content in Sufu^−/− compared to Ptch^−/− cells and represent the average +/− SD of three independent experiments, each performed in triplicate.

See also Figure S1.

Therefore, these data indicate that ODC protein is upregulated via a Sufu-mediated mechanism acting at post-transcriptional level. Since Sufu did not affect ODC stability (Figure S1G), this suggests that Sufu regulates ODC biosynthesis, likely affecting the function of some unknown modulator.

Sufu binds and stabilizes CNBP to regulate ODC translation

To shed light on this, we expressed Flag-tagged Sufu in HEK293 cells and performed mass spectrometry analysis on Flag immunoaffinity purified Sufu-bound proteins. Among the hits with the highest coverage (Table S1), we found CNBP (CCHC type Nucleic acid Binding Protein), a zinc finger protein that binds single stranded RNA and DNA and functions as a nucleic acid chaperone (Calcaterra et al., 2010). Notably, CNBP was shown to regulate ODC translation through an IRES-mediated mechanism (Sammons et al., 2010). We generated and validated a CNBP antibody (Figure S2A) and confirmed the ability of CNBP to bind Sufu at endogenous level and in vitro (Figures 2A and S2B). We also ruled out the involvement of CNBP in the Sufu-mediated regulation of Gli transcriptional activity, stability and processing and any competition with the Gli/Sufu complex (Figures S2C–F).

Figure 2. Identification of the CNBP-Sufu-ODC axis.

(A) Co-immunoprecipitation of FLAG-tagged (top) or endogenous (bottom) Sufu with endogenous CNBP in MEF cells; In, 2.5% Input.

(B) Top, IRES translation assay on MEF cells transfected with ODC-Luc plasmid and either two distinct CNBP (shCNBP1, shCNBP2) or a scrambled shRNA vectors (shCtr). The results are expressed as fold change compared to the shCtr transfected cells. Luciferase values indicate the IRES-mediated translational activity of the ODC 5′UTR sequence and are normalized by the Renilla values, representing Cap-dependent translation of the same transcript. Results are shown as the average +/− SD of five independent experiments, each performed in triplicate *p<0.01. Middle, protein levels of ODC, CNBP and actin (loading control). Bottom, schematic representation of the ODC-Luc vector. See text for details.

(C) In vivo translation of monocistronic 7mGpppG-capped (7mGpppG-Luc), and ApppG-capped polyA⁺ Luciferase mRNA either without (ApppG-Hairpin-Luc, −IRES) or with human ODC IRES (ApppG-Hairpin-ODC-Luc, +IRES). After in vitro translation and modifications (see methods), mRNA were transfected in NIH3T3. Luciferase values were normalized by quantitative PCR analysis of luciferase mRNA levels. *p<0.05. Results represent the average +/− SD of four independent experiments, each performed in triplicate. Bottom, schematic representation of the monocistronic vector.

(D) In vivo translation of ApppG-capped polyA⁺ monocistronic Hairpin-Empty-Luc (−IRES) or Hairpin-ODC-Luc (+IRES) vectors in NIH3T3 cells transfected with siCNBP, siCtr (left) and with either CNBP or empty expression plasmids (right). Values indicate luciferase activity normalized to luciferase mRNA levels assessed by qPCR. *p<0.05 siCtr (+IRES) vs siCtr (−IRES); **p<0.05 siCNBP (+IRES) vs siCtr (+IRES); ***p<0.01 CNBP (+IRES) vs Empty (+IRES). Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

(E) Polyamine levels (left) in MEF Ptch^−/− cells transfected with either CNBP (shCNBP) or scrambled (shCtr) shRNAs. *p<0.05 shCNBP vs shCtr. Right, cell proliferation assay from cells transfected as above *p<0.05 shCNBP vs shCtr 48 hrs; **p<0.01 shCNBP+Put vs shCNBP 48hrs. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate. Put, putrescine.

(F) Analysis of CNBP and Sufu association with polysomes in subconfluent WT MEF cells. Lysates were separated on a 15%–50% sucrose gradient (−EDTA, left). To assess the specificity of protein co-fractionation, an aliquot of the cell lysate was treated with 35mM EDTA, pH 7.4 and separated on a 15%–50% gradient containing 10mM EDTA (+EDTA, right). The presence of CNBP and Sufu protein in each fraction was analyzed by Western blotting. Distribution of ribosomal proteins and purity of the fractions along the gradient was controlled with rpS19 and vinculin staining, respectively.

(G) RNA-IP of ODC mRNA. NIH3T3 cells were cross-linked and lysed. Lysates were immunoprecipitated with CNBP, Sufu or control IgG antisera. Eluted mRNAs were reverse transcribed and quantified by qPCR with primers encompassing actin and ODC 5′UTR. Results are indicated as fold difference, relative to IgG control. *p<0.05. Results are shown as the average +/− SD of four independent experiments, each performed in triplicate.

(H) Top, IRES-translation assay on MEF cells transfected with ODC-Luc and either Sufu (siSufu1, siSufu2) or scrambled siRNAs (siCtr). *p<0.01. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate. Bottom, Sufu and actin protein levels.

(I) CNBP and Sufu protein (top) and mRNA levels (bottom) normalized by HSP70 and GAPDH levels, respectively, in WT and Sufu^−/− MEF cells. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

See also Figure S2.

Knockdown of CNBP with different shRNAs caused a reduction of ODC protein levels (Figure 2B, middle). To confirm that this reduction was related to the ability of CNBP to regulate ODC IRES dependent translation, we used the bicistronic ODC-Luc vector (Sammons et al., 2010) (Figure 2B bottom). This construct contains: i) the renilla CDS, ii) the luciferase CDS and iii) the intercistronic ODC 5′UTR. This cassette is expressed as a single transcript, under the control of CMV promoter. Translation of the renilla gene (the most 5′ of the two ORFs) is Cap-dependent and terminates with the TAA stop codon. Translation of the luciferase gene occurs through the ODC internal initiation site (IRES), located in the intercistronic region. Therefore, the luciferase values represent the ODC IRES activity of the intercistronic region and are normalized for the constitutive renilla activity (Sammons et al., 2010). Knockdown of CNBP reduced the luciferase, but not the renilla activity, confirming the ability of this protein to regulate ODC IRES-dependent translation (Figure 2B, top).

Northern blot analysis of RNA extracted from control or CNBP-deficient cells transfected with ODC-Luc did not show the presence of accessory bands, demonstrating that the effect of CNBP was not related to RNA splicing processes (Figure S2G) of the bicistronic transcript.

To rule out potential plasmid artifacts and to verify that this effect was a genuine IRES dependent process, we transfected NIH3T3 cells with in vitro transcribed, monocistronic IRES^ODC-Luciferase mRNA, conjugated to the non-physiological cap analog ApppG, and containing a stable stem loop structure at the 5′ end to inhibit scanning (Gilbert et al., 2007) (Figure 2C, bottom). Compared to 7mGpppG-capped mRNA, the in vivo translational efficiency of the ApppG-capped Hairpin-containing mRNA lacking the IRES sequence (−IRES), was strongly reduced by 30 fold (Figure 2C, left). Insertion of the ODC 5′ UTR IRES sequence (+IRES) significantly increased the translational efficiency of ApppG-Hairpin mRNA (Figure 2C, right). This confirmed the presence of genuine IRES activity in the ODC 5′UTR (Pyronnet et al., 2000). CNBP depletion significantly reduced this IRES-dependent activity (Figure 2D, left), whereas exogenous CNBP induced translation of ApppG +IRES mRNA, but did not change the translational activity of control −IRES (Figure 2D, right). Thus, CNBP is a bona fide regulator of ODC IRES-dependent translation.

Ablation of CNBP decreased polyamines levels (Figure 2E, left) and the proliferation rate of MEF Ptch ^−/− cells and this effect was rescued by addition of putrescine (Figure 2E, right), supporting that loss of CNBP impairs the same metabolic step of Sufu.

To determine whether Sufu and CNBP are both associated to translational complexes, we performed polysomal fractionation. As shown in Figures 2F and S2H, CNBP and Sufu copurified in the polyribosomes fractions and EDTA treatment, which displaces the binding between the ribosomal subunits, disrupted this association, demonstrating that both proteins are associated to actively translating complexes. CNBP and Sufu were also bound to ODC mRNA, as revealed by RNA-IP (Figure 2G) further indicating their involvement in ODC translation. Consistently, the absence of Sufu caused a reduction of ODC-Luc activity (Figure 2H), indicating that Sufu is involved in the regulation of ODC translation. Interestingly, we observed a reduction of CNBP protein, but not mRNA levels in Sufu-deficient cells (Figure 2I, bottom), suggesting that Sufu regulates CNBP at post-transcriptional level.

Based on previous data illustrating the ability of Sufu to bind and stabilize Gli proteins (Chen et al., 2009; Humke et al., 2010; Wang et al., 2010), we hypothesized that Sufu could play a similar role in this context, by preventing CNBP degradation. Toward this end, we incubated WT and Sufu^−/− cells with the protein synthesis inhibitor cycloheximide (CHX) for different time points and observed that CNBP half-life was significantly shorter in Sufu-deficient cells (Figure 3A). The specific involvement of Sufu was demonstrated by its retroviral-mediated reconstitution in Sufu^−/− cells, which restored CNBP levels (Figure 3B). Incubation of Sufu^−/− cells with the proteasome inhibitor MG132 rescued the expression of CNBP and of Gli3, used as control (Chen et al., 2009), indicating the involvement of proteasomal degradation in this process (Figure 3C). In agreement with this observation, we found that both ectopic and endogenous CNBP were efficiently polyubiquitinated, a process that was increased by Sufu knockdown and decreased by its overexpression (Figures 3D,E, S3A,B). Thus, Sufu binds CNBP and prevents its ubiquitination and proteosomal degradation, thereby increasing its stability.

Figure 3. Sufu prevents CNBP degradation.

(A) Left, CNBP and actin levels in WT or Sufu^−/− MEF cells incubated with CHX (100μg/mL) for the indicated times. Right, densitometric analysis. The value 1 was assigned to time 0. Results represent the average +/− SD of three independent experiments. *p<0.05

(B) CNBP protein levels in Sufu^−/− MEF cells transduced with empty (Ctr) or Sufu (rSufu) retroviruses. Actin, loading control.

(C) Effect of 6 hours treatment with 50 μM MG132 on CNBP levels in Sufu^−/− MEF cells. Full Length Gli3 (Gli3FL) levels shown as control of MG132 efficacy.

(D) Ubiquitination assay in MEF cells transfected with FLAG-tagged CNBP, HA-Ub and siSufu (left) or Myc-Sufu (right). Filters were probed with HA antibody to detect ubiquitination and with FLAG, MYC and actin antibodies for the other proteins.

(E) Endogenous ubiquitination assay in MEF cells transfected with HA-tagged Ub and Myc-tagged Sufu vectors. Filters were probed as indicated.

See also Figure S3.

Hh activation increases ODC biosynthesis via AMPK/CNBP

Since we observed increased expression of ODC protein and polyamine biosynthesis in Hh-activated cells (Figures 1A, 1C, S1B, S1C, S1D) we next wondered whether and how activation of Hh pathway perturbs this Sufu/CNBP-mediated mechanism.

In co-immunoprecipitation studies with equal amounts of CNBP we observed that upon exposure of cells to SAG, the binding to endogenous Sufu was promptly increased (Figure 4A and S4A), leading to accumulation of CNBP protein (Figure 4B and S4B). Similarly, CNBP levels were higher in Ptch^−/− cells, compared to control cells and in GCPs treated with Shh peptide (Figures 4C and S4C).

Figure 4. CNBP controls ODC translation in response to Hh.

(A) Endogenous association of Sufu to CNBP in MEF cells treated with SAG or DMSO for 4 hours. Cells were lysed and immunoprecipitated with saturating amounts of CNBP or control (IgG) antisera, as indicated.

(B, C) CNBP levels in MEF cells treated with SAG or DMSO for 4 hours (B) and in WT and Ptch^−/− MEF cells (C); actin, loading control.

(D) Polysomal recruitment of CNBP in SAG-treated NIH3T3 cells. Cells were grown in low serum (0,5% BS) medium overnight to allow full Hh response and then treated with either DMSO or SAG for 4 hours. Cells were lysed and loaded onto a 15%–50% sucrose gradient in the presence or absence of EDTA, as described in Figure 2D. CNBP protein levels in each fraction were analyzed by Western blotting.

(E) RNA-IP of ODC mRNA. NIH3T3 cells, treated with SAG or DMSO for 4 hours, were cross-linked and lysed. Lysates were immunoprecipitated with CNBP or control IgG antisera. Eluted ODC mRNAs were analyzed as in Figure 2E. Results are indicated as fold difference, relative to IgG control. *p<0.05 CNBP DMSO vs IgG; **p<0.05 SAG CNBP vs DMSO CNBP. Results are shown as the average +/− SD of four independent experiments, each performed in triplicate.

(F) ODC and CNBP protein expression in WT and Ptch^−/− MEF cells, transfected for 72 hours with siCtr or siCNBP. Actin, loading control.

(G) WT and Ptch^−/− MEF cells pretreated with 50 μM MG132 for 2 hours and cultured for 30 minutes with [35S]-methionine, to label newly synthetized proteins. Cell lysates were immunoprecipitated with ODC antisera and the immunocomplexes were revealed by autoradiography (top). A fraction of the IP was analyzed by Western blot and showed equivalent amounts of ODC levels. Controls include analysis of whole cell extracts (WCE), where comparable amounts of [35S]-methionine were incorporated into WT and Ptch^−/− MEF cells (bottom).

(H) Activity of the ODC-Luc vector in NIH3T3 cells treated with SAG or DMSO for 6 hours after 72 hours transfections with siCNBP or siCtr. *p<0.01 siCtr SAG vs siCtr DMSO; **p<0.05 shCNBP SAG vs shCtr SAG. Results represent the average +/− SD of three independent experiments, each performed in triplicate.

See also Figure S4.

To determine whether this upregulation is associated to an increased translational activity of CNBP, we performed polysomal fractionation analysis in control and Hh-activated cells.

Following SAG activation, CNBP was loaded on polysomal fractions and EDTA treatment disrupted this association (Figure 4D and S4D, E). Consistently, treatment with SAG increased the recruitment of CNBP to ODC 5′UTR in RNA-IP experiments (Figure 4E). Paralleling CNBP, ODC protein levels were upregulated in Ptch^−/− MEF cells, compared to WT MEF cells, and knockdown of CNBP abolished this effect (Figure 4F). The increase of ODC protein was not dependent on differences in mRNA synthesis and/or stabilization, since ODC mRNA levels were unchanged (Figure S4F), neither could be attributed to variations of protein stability, as there were no differences in ODC degradation rate between WT and Ptch^−/− cells (Figure S4G).

The addition of the protein synthesis inhibitor cycloheximide (CHX) prevented the Hh-induced upregulation of ODC (Figure S4H), thus indicating that the increase was dependent on translation. To directly demonstrate an increase in ODC protein synthesis, we performed in vivo metabolic [³⁵S]-methionine labeling followed by ODC IP. Compared to WT cells, Ptch^−/− cells showed a significant increase of metabolically labeled, newly synthesized ODC protein in the presence of the proteasome inhibitor MG132 (Figure 4G, S4I). Consistently, treatment of cells with SAG strongly increased the ODC-Luc IRES-dependent activity, and this effect was disrupted by CNBP knockdown (Figure 4H). Therefore, Hedgehog induces ODC translation through CNBP.

We next sought to understand the signaling pathway connecting Smo to CNBP. Toward this end, we analyzed the CNBP sequence to identify putative residues that could be subjected to post-translational modifications. We noticed a potential AMPK-consensus site (LARECTIEAT), evolutionary conserved (Figure 5A), located in the C-terminal region that diverged from the optimal AMPK site for having the T in position +4 instead of a hydrophobic residue. A similar variation was previously identified in one of the AMPK sites of FOXO3 (Greer et al., 2007).

Figure 5. AMPK promotes ODC translation in response to Hh agonists by phosphorylating CNBP at the conserved threonine 173.

(A) Protein alignment of the amino acid sequence of CNBP surrounding T173. Optimal AMPK consensus motif shown.

(B) ODC, CNBP, total and phosphorylated AMPK protein levels in WT and AMPK^−/− MEF cells treated with DMSO or SAG for 6 hours. Actin, loading control.

(C) Activity of the bicistronic ODC-Luc vector transfected for 48 hours in WT and AMPK^−/− MEF cells and treated with DMSO or SAG for 6 hours. *p<0.01 WT SAG vs WT DMSO; **p<0.01 AMPK^−/− SAG vs WT SAG. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

(D–F) CNBP, ODC, total and phosphorylated AMPK protein levels in GPCs stimulated with Shh for 6 hours in the absence or presence of Compound C (CC, 20 μM) (D) or with the AMPK agonist A769662 (25 μM) (E) or treated with KAAD-cyclopamine (KAAD, 0.1 μM) for the indicated times (F). Tubulin, loading control.

(G) In vitro AMPK-phosphorylation assay of recombinant GST-CNBP WT or T173A mutant. Incorporation of ³²P was determined by autoradiography and the protein levels were detected by Coomassie blue staining.

(H) WT and AMPK^−/− MEF cells were transfected with a plasmid encoding FLAG-CNBP. Cell extracts were immunoprecipitated with anti-FLAG antibody and phosphorylation was revealed with an anti-phospho AMPK substrate (pAMPKsub) antibody. Equivalent amounts of FLAG-CNBP shown after FLAG immunoblot.

(I) Endogenous CNBP phosphorylation in MEF WT cells treated with DMSO or SAG for 2 hours. Cell extracts were immunoprecipitated with anti-CNBP antibody or control (IgG) antisera, as indicated. Filters were stained with pAMPKsub and reprobed with CNBP antibodies. Input=5%.

(J) MEF WT cells expressing HA-tagged WT or T173A mutant CNBP were treated with DMSO or SAG for 6 hours. Cell extracts were immunoprecipitated with anti-HA antibody and equal levels of CNBP were loaded. Phosphorylation was revealed with anti phospho-AMPK substrate antibody. Filters were reprobed with HA antibody to confirm comparable WT and mutant protein levels.

(K) Activity of the bicistronic ODC-Luc vector in CNBP-depleted MEF WT, reconstituted with WT and T173A HA-CNBP. Cell were transfected with a control siRNA (siCtr) or a siRNA targeting CNBP 3′UTR region for 72 hours and with the indicated plasmids for the last 48 hours. *p<0.01 CNBPi 3′UTR vs siCtr; **p<0.01 CNBPi 3′UTR + WT vs CNBPi 3′UTR; ***p<0.05 CNBPi 3′UTR + T173A vs CNBPi 3′UTR + WT. Results are shown as the average +/− SD of three independent experiments, each performed in triplicate.

(L) Co-immunoprecipitation of endogenous Sufu with WT, T173A, T173D HA-tagged CNBP in WT MEF cells. Cell extracts were immunoprecipitated with anti-HA antibody and the filters were stained with Sufu, HA and pAMPKsub antibodies. Input samples (5%) were probed with Sufu and HA antibodies.

See also Figure S5.

Since it was recently shown that Hh activation directly activates AMPK via non-canonical Ca²⁺-dependent activation of CAMKK2 (Teperino et al., 2012), we tested if AMPK is involved in the Hh-regulated CNBP activation.

SAG-dependent increase of CNBP, ODC protein levels and the activity of ODC-Luc reporter were disrupted in AMPK^−/− compared to WT cells (Figures 5B and 5C). Similarly, the AMPK inhibitor Compound C (CC) prevented the increase of ODC and polyamines in SAG-treated MEF cells, in Ptch^−/− MEF cells and Shh-treated GCPs (Figures S5A, B and 5D). Elevation of CNBP and ODC proteins, polyamines and ODC-Luc activity could also be observed upon exposure of cells to the direct AMPK activator A769662 (Figure 5E and S5D) and to KAAD cyclopamine (Figures 5F, S5E, F), which binds Smo and acts as an inhibitor of the canonical pathway but as a partial agonist of the non-canonical branch (Teperino et al., 2012). To determine if cyclopamine activates the non-canonical branch also in tumor cells driven by constitutive Hh activation, we tested its effect in primary medulloblastoma cells generated in conditional Math1-Cre/Ptc^C/C mice (Yang et al., 2008). We will refer to these cells as “primary medulloblastoma cells”, within text and figures. Exposure of these cells to cyclopamine did not induce AMPK activation and polyamine production, while it robustly inhibited the expression of the Hh target gene Gli1 (Figure S5G).

Consistent with previous observations (Teperino et al., 2012), activation of this non-canonical mechanism is dependent on the integrity of the primary cilium, since SAG failed to induce CNBP, AMPK phosphorylation and putrescine content in cilia-deficient Itf88^−/− MEF cells (Figure S5H). Thus, activation of the non-canonical Hh-AMPK pathway promotes upregulation of ODC and polyamines in normal cells but not in tumors sustained by loss of Ptch1.

To determine if CNBP is phosphorylated by AMPK at the putative threonine 173, we performed in vitro kinase assay with recombinant CNBP or its T173A mutant. WT CNBP efficiently incorporated ³²P in the presence of AMPK, whereas the T173A mutant did not (Figure 5G), thus demonstrating that AMPK phosphorylates CNBP only at this residue. We then immunoprecipitated ectopically expressed CNBP from WT and AMPK^−/− cells and performed western blotting with an antibody recognizing AMPK-phosphorylated substrates. CNBP was phosphorylated in WT but not AMPK^−/− cells (Figure 5H). Activation of Hh with SAG increased the AMPK-mediated phosphorylation of both endogenous and ectopic CNBP (Figure 5I,J), and this modification was completely abrogated by mutation of T173 residue to alanine (Figure 5J). To study the activity of the T173A mutant and to rule out the possibility of overexpression artifacts, we knocked down endogenous CNBP with shRNA targeting the 3′UTR region and reintroduced WT and T173A CNBP (Figure S5I). Reconstitution of CNBP-deficient cells with CNBP WT rescued ODC-Luc activity and putrescine levels, whereas the T173A mutant had a significantly reduced activity (Figures 5K and S5I). CNBP T173A mutant was no longer upregulated by SAG (Figure S5J) and showed a significantly reduced half-life compared to the WT protein (Figure S5K).

Since the stability of CNBP depends on its association with Sufu, we tested the binding of Sufu to CNBP WT, T173A and the phosphomimetic T173D mutant. Compared to WT CNBP, T173A mutant bound Sufu with reduced affinity, while the T173D showed an increased affinity (Figure 5L and S5L) indicating that phosphorylation of T173 favors the formation of the complex.

In keeping with this finding, the phospho mimetic T173D mutant promoted significantly higher ODC-Luc activity and polyamines production (Figure S5M).

Collectively, these data indicate that T173 phosphorylation represents the critical regulated checkpoint of this Hh/AMPK-mediated regulation of CNBP.

ODC and CNBP are elevated in MB and their targeting prevents MB growth

To explore the pathophysiological relevance of this mechanism, we analyzed CNBP and ODC levels in Hh-dependent MB developed in conditional Math1-Cre/Ptc^C/C mice, where Hh signaling is constitutively hyperactive (Figure S6A). Compared to normal adult cerebellum, CNBP, ODC, Sufu and phosphorylated AMPK were elevated in MB samples (Figure 6A, S6B) and this was accompanied by a strong upregulation of the three polyamines (Figure 6B).

Figure 6. CNBP and ODC proteins are highly expressed in mouse and human MB.

(A) Western blot analysis of wild-type cerebella and Ptch^−/− medulloblastoma (MB) samples. Staining for CNBP, ODC, pAMPK, AMPK, Sufu and actin as loading control are shown.

(B) Polyamine levels in normal cerebellum and Ptch^−/− MB tissue samples. Results are the average +/− SD of three different samples for each condition. *P<0.05 MB vs Cerebellum.

(C) Representative images of immunohistochemistry (IHC) staining for CNBP and ODC in human adult normal cerebellum and human MB, subdivided according to their molecular subgroup [WNT (wingless), SHH (sonic hedgehog), Group 3, and Group 4]. The nuclei were counterstained with hematoxylin. 40× original magnification at light microscopy; scale bars = 100μM.

(D) Immunohistochemistry of human adult normal cerebellum and human Large Cell Anaplastic (LCA), Classic and Desmoplastic medulloblastoma of the SHH subgroup, stained with anti-CNBP and anti-ODC antibodies. Nuclei were counterstained with hematoxylin, scale bars = 100μM.

See also Figure S6.

We next analyzed CNBP and ODC expression in a cohort of 42 human MBs, divided according to the molecular subgroup [5 WNT (wingless), 17 SHH (sonic hedgehog), 10 Group 3, and 10 Group 4] as previously described (Mastronuzzi et al., 2014; Miele et al., 2015).

CNBP and ODC staining were strongly increased in the SHH molecular subgroup compared to normal adult cerebellum, which showed a negative staining for ODC and a weak positivity for CNBP. A positive staining of CNBP and ODC was also observed in the Group 3, while the WNT and Group 4 molecular subgroups displayed a weaker staining (Figure 6C). Negative controls were also performed (Figure S6C).

We next analyzed the cohort of 17 human SHH MBs belonging to the three histological subgroups: desmoplastic, classic and large cell anaplastic - LCA (Kool et al., 2012; Northcott et al., 2012b). The staining for both CNBP and ODC was strongly positive in all the MB analyzed independently of their histological classification (Figure 6D).

The strong increase of CNBP, ODC and polyamine levels in Hh-dependent tumors prompted us to evaluate the consequence of turning off this mechanism in tumor cells.

We first tested different shRNA expressing lentiviruses targeting CNBP for their efficacy in knocking down CNBP and consequent ODC downregulation in primary SHH MB (Figure 7A). We selected the shRNA #58 and used it in all the subsequent experiments.

Figure 7. Targeting CNBP/ODC axis impairs MB growth in vitro and in vivo.

(A) Primary medulloblastoma cells infected with different shCNBP-containing lentiviral particles (clones 58, 59, 66) or a control lentivirus (shCtr) for 96 hours. Top, qPCR analysis of CNBP, ODC and Gli1 mRNA levels, normalized by GAPDH. *p<0.01. Results represent the average +/− SD of three independent experiments, each performed in triplicate.

(B) Polyamine levels in primary medulloblastoma cells infected with shCNBP clone 58 or control (shCtr) lentivirus for for 96 hours. *p<0.05. Results represent the average +/− SD of three independent experiments, each performed in triplicate.

(C) BrdU incorporation assay in primary medulloblastoma cells infected with shCNBP clone 58 or control lentivirus (shCtr) for 96 hours and stained by BrdU (red) and nuclear fluorescent stain Hoechst3342 (blue). BrdU incorporation was carried out for the last 24 hours. The % of BrdU positive cells was measured and the results expressed as fold change, relative to cells infected with scrambled. *p<0.05. Results are the average +/− SD of three independent experiments, each performed in triplicate. Bottom, representative pictures. Scale bar = 100 μM

(D) Cell proliferation assay on primary medulloblastoma cells, treated with DFMO (5mM, blue) or vehicle (H₂O, black) for the indicated times. *p<0.01. Results are shown as the average +/− SD of four independent experiments, each performed in triplicate.

(E) Corresponding polyamine levels in the same cells treated as above for 72 h. *p<0.05.

(F) CNBP, Sufu and Gli1 levels in Ptch^−/− MEF and Ewing Sarcoma TC-71 cell lines; tubulin, loading control.

(G) Cell proliferation assay in TC-71 cells treated with 5 mM DFMO, 5 μM ATO or vehicle (NT) for 72 hours. Results show a reduced cell proliferation by blocking Gli activity (ATO), but not by inhibiting ODC (DFMO). *p<0.01 ATO vs NT. Results are shown as the average +/− SD of four independent experiments, each performed in triplicate.

(H) Effect of Gli1 depletion in TC-71 cells. Cells were infected with shGli1-expressing or control lentiviruses (shCtr) for 72 hours. Cells were counted (left) and analyzed for polyamine content (middle). Gli1 RNAi efficiency was assessed by qPCR (right). * p<0.01 (0.05) vs shCtr. Results are shown as the average +/− SD of three independent experiment, each performed in triplicate.

(I) Primary medulloblastoma cells were transduced with shCNBP or control lentiviruses for 96h. Nude mice were allografted with 2×10⁶ viable infected cells and treated with 0.5% DFMO in the drinking water (DFMO= green, n=6; shCNBP DFMO= violet, n=6) or H₂O as control (shCtr H₂0= blue, n=6; shCNBP H₂0= red, n=6); tumors were measured using a caliper when they reached a volume of at least 100mm³, at the indicated days. *p<0.01. Results are shown as the average +/− SD.

(L) Polyamine levels from the above samples. Polyamine content was measured by GC-MS and normalized to tissue weight (nmol/mg tissue). Results are expressed as fold change relative to untreated mice (H₂O) and are shown as the average +/− SD. *p<0.01.

(M) Representative images of allografted nude mice at day 10 (top), of excised tumors (middle, scale bar = 1cm) and of immunohistochemical staining of CNBP in sections from the above samples (bottom, scale bar = 100 μM).

Ablation of CNBP caused a significant reduction of polyamine content (Figure 7B) and cell proliferation in primary SHH MB cells and in Shh-treated GCPs, as assessed by BrdU incorporation assays (Figures 7B, C and S7A).

Exposure of primary SHH MB cells to DFMO resulted in a robust inhibition of tumor cell proliferation (Figure 7D), without affecting Gli1 mRNA levels (Figure S7B). Accordingly, the levels of the three polyamines putrescine, spermine and spermidine in MB cells were downregulated by this drug (Figure 7E).

To determine whether DFMO could inhibit proliferation of tumors sustained by upregulation of Gli, independently of upstream activation of Smo, we used the TC-71 tumor cells. Gli1 drives the proliferation of these cells as a consequence of downstream events causing its aberrant transcription (Beauchamp et al., 2009; Beauchamp et al., 2011). Indeed, the direct Gli1 inhibitor ATO strongly inhibits proliferation of these cells, while they do not respond to the treatment with cyclopamine (Beauchamp et al., 2011).

Compared to the Ptch^−/− MEF cells, TC-71 cells showed comparable levels of Gli1 (Figure 7F). Treatment of these cells with DFMO did not affect their proliferation, while ATO had a strong inhibitory effect (Figure 7G). Knockdown of Gli1 caused a significant decrease of their proliferation, without affecting polyamine levels (Figure 7H). Thus, Gli1-driven proliferation does not require changes of polyamine content.

To study the effect of targeting the CNBP/ODC axis in vivo, we allografted tumor cells explanted from Ptch^−/− MB and stably transduced with shCNBP or control lentiviruses into nude mice. When the tumor reached the volume of 100 mm³, half of the mice from each group were treated with 0.5% DFMO in the drinking water until the end of the experiment, monitoring the growth of the tumor every other day. As shown in Figures 7I, L and M, MB growth and polyamine content were strongly reduced in CNBP-deficient tumors, compared to controls. Inhibition of ODC with DFMO also had a similar inhibitory effect on tumor growth and polyamine content. Notably, DFMO did not cause a further decrease of the growth and polyamine levels, neither a reduction of Gli1 mRNA (Figure S7C), in tumors depleted of CNBP, demonstrating that this drug requires CNBP to exert its effect.

Taken together, these data show that CNBP and ODC are elevated in Hh-dependent MB and that genetic or pharmacological targeting of the Hh-ODC axis is effective in counteracting the growth of MB both in vitro and in vivo.

DISCUSSION

In this work we have demonstrated the ability of Hh signaling to control a druggable metabolic process, which is required for normal and MB growth.

The non-canonical Smo/AMPK-dependent signaling route was originally identified in adipocytes, muscle cells and fibroblasts, where it promotes a rapid metabolic rewiring toward glycolysis and increase of glucose uptake (Teperino et al., 2012). Here we provide evidence that the same transduction axis is essential to control the proliferation of cerebellar GCPs and their tumor counterpart by regulating ODC, the key gatekeeper of polyamines metabolism.

The levels of ODC are controlled by CNBP, a small conserved protein known to regulate neuronal cell proliferation and forebrain development (Chen et al., 2003), in response to unknown signals. We show here that following Hh activation, CNBP is phosphorylated at T173 by AMPK and promotes translation of ODC, thereby controlling cell proliferation. Thus, in contrast to the classical mode of action as a tumor suppressing kinase, here AMPK is engaged in a process that supports, rather than counteracts, tumor cell growth. These observations are in line with the emerging view that AMPK may act as tumor suppressor or a contextual oncogene, depending on degree or duration of its activation (Liang and Mills, 2013). It is likely that Hh signaling activates AMPK to a level where it prevails its tumor promoting effect. Interestingly, we also show a role for AMPK in promoting, rather than inhibiting a translational process. Indeed, our findings illustrate that, by activating CNBP, AMPK promotes IRES-dependent translation, an emergency mechanism utilized by cells to cope with different adverse conditions, including metabolic, hypoxic or oncogenic stress (Spriggs et al., 2008).

In a previous work it was proposed that in cardiac myoblasts AMPK limits ODC activity, particularly when cells are exposed to isoproterenol, which was found to phosphorylate AMPK and to increase polyamine content (Passariello et al., 2012). This suggests a context-dependent role of this kinase in regulating polyamine metabolism, possibly related to its involvement in different intracellular networks.

All these aspects add a further level of complexity to the role of AMPK in cancer and underscore the importance to carefully evaluate the use of AMPK agonists as anticancer drugs, taking into account the possibility of the dual effect on tumor growth and stress adaptation.

A similar bifunctional mode of action is observed here with Sufu, which supports ODC translation and polyamine metabolism by increasing the stability of CNBP.

This observation is seemengly paradoxical, since Sufu is generally described as a tumor suppressor and its monoallelic deletion appears to be sufficient to confer a Hh gain-of-function phenotype in mice (Cooper et al., 2005; Svard et al., 2006). However, MB does not develop in Sufu^+/− mice, unless they carry a p53^−/− background (Lee et al., 2007) and conditional cerebellar Sufu^−/− mice do not display signs of early MB or pre-tumoral lesions, such as thickening of the external granule layer (Kim et al., 2011; Yang et al., 2008), typically observed in other Hh gain of function models, such as the cerebellar ptch deletion or smo activating mutations (Wu et al., 2011). Thus, it is possible that the lack of MB development in Sufu deficient animals may reflect the ability of this protein to mediate both tumor suppressing and promoting functions of Hedgehog signaling, like polyamine biosynthesis.

Mono and biallelic mutations of Sufu have been also found in human MB (Brugieres et al., 2012; Pugh et al., 2012; Robinson et al., 2012) and are believed to cause activation of the canonical pathway. Whether these mutations also affect the AMPK-dependent route remains to be determined.

In addition to SHH subgroup, CNBP and ODC protein levels are elevated in Group 3 MBs. While the mechanism underlying the increase of CNBP has to be established, it is possible that the elevation of ODC could be attributed, at least in part, to overexpression of cMyc, a typical alteration found in this molecular subgroup (DeSouza et al., 2014). Indeed, ODC is a transcriptional target of cMyc and is tipically overexpressed in cMyc-driven malignacies (Bello-Fernandez et al., 1993; Nilsson et al., 2005).

A key finding of this work is that targeting of this Hh/CNBP/ODC metabolic axis efficiently prevents the growth of MB cells in vitro and in vivo, by reducing intracellular polyamine content. Increased polyamine levels had been observed in MB in an old study (Scalabrino et al., 1982) but their role and the underlying mechanisms were not known, as well as the existence of different molecular MB subgroups. Here we demonstrate that a non-canonical, Smo-dependent and Gli-independent mechanism is responsible of the elevations of the polyamine content observed in SHH MB.

This evidence appears to be relevant since a growing body of evidence is documenting the occurrence of resistance to Smo antagonists in the treatment of MB and other Hh driven tumors. Explanations to this phenomenon have been provided with the occurrence of novel Smo mutations and/or with the identification of oncogenic signaling pathways involving ERK, AKT and mTOR/S6K1, which promote Smo-independent Gli activation or non canonical Gli-independent cascades (Di Magno et al., 2015). These observations all indicate that full inhibition of Hh pathway requires alternative approaches, inhibiting the signaling at multiple levels rather than targeting the receptor apparatus alone. Interestingly, the canonical Smo antagonists cyclopamine and GDC-0449 appear to function as selective partial agonists of this non-canonical Hh/AMPK branch in metabolic cells (Teperino et al., 2012) and in GCPs. Cyclopamine inhibits canonical Hh signaling and counteracts cell growth only in tumors with constitutive activation of the receptor apparatus (i.e. in Ptch1 loss of function). In these tumors, the activity of Smo is constitutively increased on both canonical and non-canonical branches. In addition, cyclopamine efficiently represses Gli-dependent transcription but does not result in any further activation of AMPK-polyamine axis in cultured Ptch^−/− SHH-MB cells. Therefore, the effect of cyclopamine differs between normal and tumor cells: in normal cells cyclopamine acts as a partial agonist of the non-canonical Hh signaling, but inhibits the canonical route. In contrast, in Hh-dependent tumors, cyclopamine inhibits the canonical pathway but is unable to induce any further activation of the non-canonical branch.

To address the relevance of the non-canonical AMPK-ODC pathway in SHH MB, we have targeted both CNBP (shRNA) and/or ODC (DFMO) and observed a significant inhibition of the growth of MB cells from conditional Ptch KO mice, where Smo is constitutively activated. Thus, turning off the non-canonical branch limits MB growth, even if the canonical signaling is still hyperactive, indicating that both pathways are required for growth.

Further supporting our findings, a previous report showed that treatment of Ptch^+/− mice with DFMO prevented the formation of UV-induced basal cell carcinoma, although the underlying mechanism was not clarified (Tang et al., 2004).

Compared to other chemotherapeutic compounds, DFMO is a well-tolerated drug with a modest toxicity, consisting in thrombocytopaenia, gastrintestinal effects and reverible hearing loss. Recent clinical trials have demonstrated its significant effect as a chemopreventive agent in tumors like neuroblastoma, skin and colon cancers (Jeter and Alberts, 2012). Our preclinical data support the use of DFMO also as a pharmacological agent for MB. Whether this approach, alone or in combination with other strategies, will be succesfull in human MB patients thus represents a relevant open question rised in this work.

EXPERIMENTAL PROCEDURES

Drug treatments

The ODC inhibitor DFMO (Sigma, 5mM) and putrescine (Sigma, 10μM) were added to the medium for the indicated times.

For SAG treatment, NIH3T3 cells were incubated in low serum (0.5% BS) overnight. MEF WT and Itf88^−/− were incubated in 1% BSA overnight, and then exposed to SAG (200nM, Enzo Life Sciences) or KAAD (0.1 μM, Calbiochem) for the indicated times. MEF Sufu^−/− and HEK293T cells were treated with MG132 (50μM, Sigma) for 6 hours. MEFs cells were treated for the indicated times with 100 μg/mL protein-synthesis inhibitor cycloheximide (CHX) (Sigma). GCPs were treated with recombinant Shh-N-terminal peptide (R&D Systems, 3 μg/ml in BSA), with KAAD (0.1μM, Calbiochem), A769662 (25μM, Santa Cruz Biotechnology), for the indicated times.

MEF cells and GCPs were pretreated with Compound C (20μM, Millipore) for 20 minutes and then stimulated with SAG or Shh, respectively, for the indicated times. For polyamine quantitation, MEF Ptch^−/− and GCPs were treated with 10μM CC for 24 hours (Teperino et al., 2012). Primary medulloblastoma cells were treated with KAAD (0.5 μM) for the indicated times.

TC-71 cells were treated with 5mM DFMO (Sigma) or 5 μM arsenic trioxide (ATO, Sigma) for the indicated times.

Polyamine analysis

Polyamine content was determined by Gas Chromatography-Mass Spectrometry (GC-MS). Cells or tissues were resuspended in 0,2M HClO₄ and processed. Values were normalized by the protein concentration.

In vivo translation assay

m7GpppG- and ApppG-capped polyA⁺ transcripts were obtained by in vitro transcription of the monocistronic luciferase pSP64-Hairpin-ODC–Luc (+IRES) or of pSP64-Hairpin-Empty-Luc (IRES) vector, using the MEGAscript SP6 kit (Ambion), in reactions containing physiological m7GpppG or nonphysiological ApppG caps. mRNA transfections were performed with Lipofectamine 2000 reagent (Invitrogen). After 6 hours of transfection, luciferase assays were performed and the values were normalized by QPCR-analysis of luciferase mRNA levels.

RNA immunoprecipitation (RNA IP)

Cells were treated, formaldehyde crosslinked, lysed and sonicated. Lysates were immunuoprecipitated with the specified antibodies and the immunocomplexes washed extensively. Eluted mRNAs were analyzed by real-time quantitative RT-PCR.

Medulloblastoma Allograft Models

A parental murine MB model was derived from Math1-Cre/Ptch^C/C mice (Yang et al., 2008). Nude female mice were housed under standard conditions. Briefly, freshly isolated MB were mechanically dissociated into single-cell suspensions in HBSS (GIBCO) supplemented with Pen/Strep. 5×10⁶ viable cells were harvested in duplicate and, after 24h hours, infected with control shRNA or shCNBP58 lentiviral particles. After 96 hours, 2×10⁶ viable cells were suspended in equal volumes of PBS and Matrigel (BD Pharmingen) and implanted subcutaneously in both flanks in adult athymic nude mice (Charles River Laboratories). Animals from the two subgroups were randomized and divided in two further subgroups. Treatment initiated when tumor volume reached 100 mm³. Mice (control: n=5; shCNBP: n=5) were treated with 0.5% DFMO in the drinking water or only water (control: n=5; shCNBP: n=5) and tumors size were measured every day using a caliper. Volume was calculated as V = (L × W²)/2 (Kim et al., 2013). Growth patterns were summarized graphically by plotting the mean and SD for each treatment group over time. Study events were recorded and analyzed using GraphPad Prism software (v6.05).

CNBP knockdown was confirmed by qPCR in all tumor samples at the end of the experiment.

Statistical analysis

Statistical analysis was performed using StatView 4.1 software (Abacus Concepts, Berkeley, CA). Results are expressed as mean +/− SD from at least three independent experiments, each performed in triplicate. Statistical differences were analyzed with the Mann-Whitney U test for non-parametric values and a p<0.05 was considered significant.

Full details of the Experimental Procedures and description of standard techniques are provided in Supplemental Experimental Procedures

The Institutional Review Board approved the studies involving human samples and informed consent was obtained from all subjects.

HIGHLIGHTS.

• Hh controls polyamine metabolism and medulloblastoma growth via ODC biosynthesis.

• Sufu binds and stabilizes CNBP to regulate ODC translation.

• AMPK phosphorylates and activates CNBP in response to Hh stimulation.

• Targeting this non-canonical axis inhibits medulloblastoma growth.