Sonic hedgehog medulloblastomas are dependent on Netrin-1 for survival

Abstract

Netrin-1 signaling is an essential prototypical neuronal guidance mechanism during embryonic development that also regulates tumor cell survival in a variety of adult cancer entities. In line with these data, a monoclonal netrin-1 blocking antibody (anti-netrin-1 mAb/NP137) has been preclinically developed and netrin-1 blockade has recently been investigated in phase 1 and 2 clinical trials in several adult cancers. Here, we investigate the role of netrin-1 in the most common malignant pediatric brain cancer, Medulloblastoma. Interestingly, we find that netrin-1 is upregulated in medulloblastoma subgroups associated with developmental dysregulation, in particular in medulloblastoma with Sonic Hedgehog (SHH) activation. First, we demonstrate that genetic deletion of netrin-1 or systemic treatment with the clinical-stage anti-netrin-1 blocking antibody significantly reduces tumor growth in vivo in various orthotopic models of SHH medulloblastomas. Second, in vitro and in vivo, we unexpectedly uncover that SHH medulloblastomas treated with an SHH-inhibitor targeting Smoothened (SMO) increase netrin-1 expression, paving the way for combinatorial therapy. In line with that, we next show that netrin-1 blockade potentiates the efficacy of SMO inhibitor therapy in vivo. Together, our data indicate that, netrin-1 blockade, used as monotherapy or in combination with SMO inhibitors, is a promising therapeutic strategy in SHH medulloblastomas.

Introduction

Medulloblastoma (MB) is the most common malignant pediatric brain tumor. Standard treatments for MB consist of surgical resection, cytotoxic chemotherapy, and radiation in patients over 3 years old¹. This multimodal therapy improves average survival, but 30% of patients eventually die from recurrent disease, and survivors suffer from severe long-term side effects, including neurological and cognitive deficits. Identification of more effective therapies with reduced side effects can thus improve outcomes and quality of life for MB patients.

MB comprises four main molecular subgroups: WNT (Wingless), SHH (Sonic Hedgehog), Group 3, and Group 4^2,3. These subgroups exhibit specific transcriptional, epigenetic and mutational signatures, derive from different cells of origin, and show specific clinical characteristics such as histological features, prognosis and metastatic frequencies⁴. The SHH MB subgroup represents approximately 30% of all MBs and is associated with an intermediate prognosis. Further sub-classification defined four subtypes (α, β, γ and δ) and identified SHH subtypes α and β with poor outcomes⁵. SHH MB is characterized by mutations affecting key effectors of the SHH signaling pathway, such as PTCH1 (PATCHED1) and SMO (SMOOTHENED) receptors, SUFU (SUPPRESSOR OF FUSED) protein, or GLI2 as well as MYCN amplifications, and TP53 mutations^6,7. TP53 mutations are associated with a dismal prognosis, making the development of new therapies a priority for this group of patients⁸. SHH MB is thought to arise from an aberrant proliferation of cerebellar granule neuron progenitors (GNPs) within the developing cerebellum^9,10. During cerebellar development, SHH, secreted by Purkinje cells, activates GNPs proliferation^11,12,13. Binding of the SHH ligand to the PTCH1 receptor releases inhibition of the SMO receptor, which leads to the activation of GLI transcription factors and subsequent expression of pro-proliferative genes in GNPs¹⁴.

The importance of SHH signaling in a large subset of MB has led to attempts to use pharmacological inhibition of the SHH pathway as a therapeutic strategy for patients with SHH MB. SMO inhibitors, including GDC-0449 (Vismodegib, Erivedge®) and NVP-LDE225 (Erismodegib), have been developed and extensively studied^15,16. These agents are effective and FDA-approved for basal cell carcinoma of the skin, another SHH-driven tumor, and have been tested in clinical trials for SHH MB^17,18. Unfortunately, MBs develop resistance upon SMO inhibitor therapy and recur¹⁹, making the elucidation of the underlying resistance mechanism a major concern in the field.

Netrin-1 signaling plays key roles in many other biological functions, including cell survival, inflammation, angiogenesis and tumorigenesis, and in many diseases, including diabetes, cardiovascular diseases, and cancer^20,21,22,23. Netrin-1 signaling is mediated by the interaction of netrin-1 ligand with a diverse set of several dependence receptors^24,25, including the Uncoordinated-5-Homologs (UNC5H, i.e., UNC5A, UNC5B, UNC5C and UNC5D), the Deleted in Colorectal Carcinoma (DCC), and the DCC homolog, Neogenin (NEO1). In the presence of their ligands, these receptors promote biological functions, such as cell survival, cell motility, or more recently epithelial-to-mesenchymal transition (EMT)^26,27,28. However, in the absence of their ligands, the receptors trigger a cascade of signaling events leading to apoptotic cell death^29,30,31. Netrin-1 has been shown to be up-regulated in many types of cancer, and a function-blocking anti-netrin-1 monoclonal antibody named NP137 has been pre-clinically developed³² and was reported to show anti-tumor activity in adult cancer models²⁸. Recent phase 1 data assessing NP137 in monotherapy in adult patients with advanced solid cancer demonstrated the interest of targeting netrin-1 in adult cancer³³. NP137 is currently being assessed in a series of phase Ib/II trials in gynecological cancer (https://clinicaltrials.gov/ct2/show/NCT04652076), in pancreatic cancer (https://www.clinicaltrials.gov/study/NCT05546853), in liver cancer (https://www.clinicaltrials.gov/study/NCT05546879), and in patients resistant to immunotherapy (https://www.clinicaltrials.gov/study/NCT05605496).

Of interest, a crosstalk between Netrin-1 and SHH signalings has been highlighted during embryonic development. Netrin-1 and SHH coordinate guidance of commissural neurons during embryonic development^34,35. Moreover, several studies on the developing nervous system featured netrin-1 expression downstream of the SHH signaling^36,37. Given these observations, we sought to explore the expression and the role of netrin-1 and its receptors in the context of MB. Netrin-1 was previously found to be upregulated in MB cell lines and reported as a potential candidate biomarker for MB patients, as urinary netrin-1 is detected in invasive MB tumors³⁸. We therefore investigated whether Netrin-1 and SHH signaling pathways interact in MB to influence tumor behavior.

Here, we show in human samples that the SHH MB subgroup has higher NTN1 (netrin-1 gene) expression compared to WNT, Group 3 and Group 4 tumors. Using mouse models, we uncover in vivo that genetic deletion of Ntn1 in postnatal GNPs does not prevent tumor formation induced by SHH activation. Interestingly, acute deletion of Ntn1 or administration of a function-blocking anti-netrin-1 monoclonal antibody inhibits tumor growth, in vitro and in vivo, while having no impact on GNPs survival. Surprisingly, we observe an increase of netrin-1 expression level upon SMO inhibition (using both Vismodegib or Erismodegib) in SHH MB tumors suggesting a potential role for netrin-1 as a tumor cell survival factor. Consistent with this, we demonstrate that both SMO inhibitors and anti-netrin-1 antibody synergize, preventing both tumor cell survival and tumor growth in vitro and in vivo. Taken together, we report here not only that SHH MB are dependent on netrin-1 for survival but also the rational for the development of a novel approach by combined targeting of SHH signaling and netrin-1.

Results

Netrin-1 is upregulated in mouse and human SHH MB

In order to investigate the role of netrin-1 in MB, we initially examined the expression of Ntn1 and its receptors in SHH MB mouse models. First, we looked at Ntn1 mRNA levels in spontaneous SHH MBs arising from tumor prone Ptch1^+/− heterozygous mice³⁹ compared to proliferative P7 granule neuron progenitors (GNPs), the cells of origin of SHH MB. Interestingly, we found higher expression of Ntn1 in SHH MB samples as compared to P7 proliferative GNPs (Fig. 1a). Conversely, we detected a downregulation of expression for netrin-1 receptors, Unc5A/B/C/D, Dcc and Neogenin in SHH MB.

Fig. 1: Netrin-1 is upregulated in mouse and human SHH MB.

a Boxplots summarizing the expression of murine Ntn1, Unc5A, Unc5B, Unc5C, Unc5D, Dcc and Neo1 analyzed from the Affymetrix dataset of Kutscher et al.⁵⁹ in Granule Neuron Progenitors (GNPs; n = 3) compared to SHH MB tumors (n = 6). Results are expressed as Log2FoldChange. The box part of the boxplots represents the interquartile range while the whisker bonds of the boxplots indicate the highest and smallest values within 1.5 times interquartile range above and below the 75th and 25th quantiles, respectively. (two-tailed t test. ****p < 0.0001, *p = 0.012 (Ntn1 and Neo1) = 0.037 (Unc5A) = 0.023 (Dcc), ns: not significant). b t-SNE projection of cells from WT P7 cerebella (n = 5) and SHH MB tumors (n = 5), from Math1-Cre/SmoM2 transgenic mice, based on PCA and color-coded for the expression of indicated markers. Analyses were performed using previously published single-cell RNA sequencing (scRNA-seq) data⁴⁰. c Histograms represent the quantification of cells expressing detectable transcript from WT P7 cerebella (n = 5) and SHH MB tumors (n = 5) of Math1-Cre/SmoM2 transgenic mice, shown in (b). The results are expressed as mean ± SEM. (two-tailed Mann–Whitney test. **p = 0.0079, *p = 0.0159). d Heatmap of NTN1, UNC5A, UNC5B, UNC5C, UNC5D, DCC and NEO1 genes expression analyzed from Cavalli’s transcriptomic data⁵ performed by Affymetrix array using 763 primary medulloblastoma samples of patients. Color bars indicate medulloblastoma subgroups. Genes in red color are overexpressed, and those in blue are downregulated. e Violin plots summarizing the expression of NTN1 and UNC5C in the different SHH MB subtypes from the transcriptomic dataset of Cavalli⁵ (SHH-α = 65, SHH-β = 35, SHH-γ = 47, SHH-δ = 76). Results are expressed as Log2FoldChange. (two-tailed t test. ****p < 0.0001, *p = 0.01). f Heatmap of NTN1, UNC5A, UNC5B, UNC5C, UNC5D, DCC and NEOGENIN protein expression analyzed from Archer’s proteomic dataset of 45 primary medulloblastoma samples of patients⁴¹. Color bars indicate medulloblastoma subgroups. Genes in red color are overexpressed, and those in blue are downregulated.

We next investigated the cellular patterns of netrin-1 ligand and receptor expression in the postnatal murine cerebellum and in murine SHH MBs, using previously published single-cell RNA sequencing (scRNA-seq) data⁴⁰. We focused on GNPs and differentiated granule neurons (GNs) in the WT cerebella, and on tumor cells in MBs. We isolated these populations by subjecting the publicly available gene expression matrices to Principal Component Analysis (PCA) and Louvain clustering, and then identifying and excluding clusters of cells with gene expression patterns that identified them as astrocytes, oligodendrocytes, myeloid cells, endothelial cells, or fibroblasts. We subjected the remaining cells to independent component analysis (ICA) and generated 2-dimensional UMAP plots showing the cells from WT cerebella and from Math1-Cre/SmoM2 MBs. Using cyclin expression to identify proliferative cells and GN markers to identify differentiating cells, we defined proliferative, intermediate and differentiating subsets of cells (Fig. 1b and c). We noted that the expression patterns of netrin-1 ligand and receptors correlated with differentiation state in both WT cerebella and SHH MB. Ntn1 was predominantly expressed by proliferative cells in both WT cerebella and SHH MB. Moreover, SHH MBs showed significantly higher fractions of Ntn1+ cells compared to WT cerebella. Conversely, Unc5 receptors were less expressed in SHH MB than in WT cerebella, consistent with the above observations (Fig. 1a–c).

Considering the increased Ntn1 expression in mouse MB, we then interrogated whether human SHH MB primary tumors similarly up-regulated NTN1. To this end, we analyzed a large transcriptomic MB dataset regrouping 763 primary MB samples from all subgroups⁵. Remarkably, the Affymetrix data revealed NTN1 high expression level in both SHH and WNT subgroups compared to Groups 3 and 4 (Fig. 1d). Moreover, comparison of all netrin-1 receptors across the four subgroups demonstrated that UNC5A, UNC5C, DCC and NEO1 were also enriched in SHH MB samples compared to other subgroups (Fig. 1d). Human SHH MB can be further divided in four subtypes: SHH-α, β, γ and δ with specific molecular features and patient clinical outcomes⁵. Interestingly, our data showed increased NTN1, UNC5A and DCC in SHH-δ compared to other subtypes (Figs. 1e and S1a). By contrast, SHH-δ expressed less UNC5C and NEO1 (Figs. 1e and S1a). We also looked at the pattern of expression of both netrin-1 and its receptors at the proteome level, using an independent MB cohort⁴¹. Interestingly, proteomics analyses confirmed the pattern of expression observed at the transcriptome level for netrin-1 and most of its receptors in human MB samples (Fig. 1f). Finally, we assessed NTN1 expression in validated MB patient derived xenografts (PDXs) from both SHH and Group 3 MB. Comparing two independent PDXs from both groups, we were able to confirm an upregulation of NTN1 in SHH MB PDXs (Supplementary Fig. 1b). Further molecular characterization of these SHH PDXs by WES, RNAseq, and proteome showed that both SHH PDXs were Ptch1 mutants, and that they clustered perfectly with human SHH MB at the transcriptome and proteome levels, providing evidence that these PDXs recapitulate the biology of SHH MB (Supplementary Fig. 1c and d). We next investigated intratumoral distribution of netrin-1 by immunohistochemistry (IHC) in the two SHH MB PDXs (Supplementary Fig. 1e). Interestingly, netrin-1 was detected throughout the proliferative areas within the tumors (Ki-67+) (Supplementary Fig. 1e), consistent with the scRNA-seq data (Fig. 1b). Altogether, our data indicate that SHH MB expresses key elements of Netrin-1 signaling, including ligand and receptors, in patterns that correlate with other clinicopathologic features, suggesting a role in tumor biology.

Netrin-1 deletion in postnatal cerebellar granule neuron progenitors does not impair SHH MB formation

Considering the robust netrin-1 expression in SHH MB and the role of Netrin-1 signaling in many cancer types, we interrogated whether the loss of netrin-1 affected SHH MB initiation in vivo. In order to design our experimental strategy, we first explored the pattern of expression of netrin-1 during mouse cerebellar development. First, we assessed the expression and localization of netrin-1 by in situ hybridization and immunostaining in mouse cerebellar sections (Fig. 2a). At P7, Ntn1 mRNA was mainly found within the outer external granule layer (oEGL) where Atoh1+ cells reside^42,43. Further immunohistochemical staining revealed that netrin-1 protein is mainly localized in proliferative GNPs and is nearly absent in differentiated granule cells (Fig. 2a), confirming the Ntn1 expression pattern in the scRNA-seq analysis (Fig. 1b).

Fig. 2: Netrin-1 deletion in postnatal cerebellar granule neuron progenitors does not impair SHH MB formation.

a Representative image of in situ hybridization of Ntn1 mRNA and immunostaining of netrin-1 and CaBP, in P7 mouse cerebellum. The nuclei were stained with DAPI (scale bar: 50 µm). b Genotyping obtained by PCR on Atoh1^CreERT2+; Ptch1^flox/flox mice depending on Ntn1 expression: Ntn1^wt/wt, Ntn1^flox/wt and Ntn1^flox/flox. c Kaplan-Meier survival curve of tamoxifen treated-Atoh1^CreERT2+; Ptch1^flox/flox mice depending on Ntn1 expression: Ntn1^wt/wt (n = 8) compared to Ntn1^flox/wt (n = 16) and Ntn1^flox/flox (n = 8) (ns: not significant). d Representative images of SHH MB developed in tamoxifen treated-Atoh1^CreERT2+; Ptch1^flox/flox mice depending on Ntn1 expression: Ntn1^wt/wt and Ntn1^flox/flox. Atoh1^CreERT2− mice were used as a control. Cerebellum slices were stained with Hematoxylin and Eosin (H&E), and Ki-67 immunostaining was performed (scale bar: 500 µm). e Gli1, Gli2 and Atoh1 mRNA expression analyzed by quantitative PCR in cerebellum from tamoxifen treated-Atoh1^CreERT2+; Ptch1^flox/flox mice depending on Ntn1 expression: Ntn1^wt/wt (n = 6), Ntn1^flox/wt (n = 8) and Ntn1^flox/flox (n = 8). Atoh1^CreERT2− (n = 5) mice were used as a control. The results are expressed as mean ± SEM. (two-tailed Mann-Whitney test, compared to Atoh1^CreERT2− group: Ntn1^wt/wt **p = 0.0087 (Atoh1), *p = 0.0426 (Gli1) = 0.0303 (Gli2); Ntn1^flox/wt **p = 0.0047 (Gli1) = 0.0027 (Gli2) = 0.001 (Atoh1); Ntn1^flox/flox **p = 0.0047 (Gli1) = 0.0016 (Gli2) = 0.0054 (Atoh1)).

In line with these observations, to determine the role of Netrin-1 signaling in SHH MB tumorigenesis, we conditionally deleted Ntn1 at the postnatal stage in Atoh1-positive cells in combination with SHH activation. For that purpose, we took advantage of a previously described SHH MB mouse model in which tamoxifen-inducible Cre deletes homozygous floxed Ptch1 (Ptch1^flox/flox) alleles in Atoh1 expressing cells. We interbred Ptch1^flox/flox and Atoh1^CreER mouse strains to generate Atoh1^CreER/Ptch1^flox/flox mice. Postnatal injection of tamoxifen in Atoh1^CreER/Ptch1^flox/flox animals induces MB at 100% penetrance with a median age of tumor onset of 13 weeks¹⁰. To similarly delete Ntn1 at the earliest stage of MB formation from GNPs, we crossed Atoh1^CreER/Ptch1^flox/flox animals with Ntn1^flox/flox animals or Ntn1^flox/wt44 resulting in Atoh1^CreER/Ptch1^flox/flox/Ntn1^flox/flox or Atoh1^CreER/Ptch1^flox/flox/Ntn1^flox/wt recipient counterparts. Pups were then treated at P2/P3 with tamoxifen, and depletion of both Ptch1 and Ntn1 in animals as well as in MB samples were confirmed by PCR (Fig. 2b). As expected, control animals (Atoh1^CreER/Ptch1^flox/flox/Ntn1^wt/wt) developed SHH MB within 30 weeks (Fig. 2c). Animals lacking one or both Ntn1 alleles (Atoh1^CreER/Ptch1^flox/flox/Ntn1^flox/wt or Ntn1^flox/flox) developed a tumor and did not show any statistical difference in survival when compared to wild-type condition for Ntn1 (Atoh1^CreER/Ptch1^flox/flox/Ntn1^wt/wt). Further characterization by Hematoxylin & Eosin (H&E) staining confirmed the formation of a tumor arising from all two genotypes (Fig. 2d). All tumors were highly positive for the proliferative marker Ki-67 (Fig. 2d) and qPCR for Gli1, Gli2 and Atoh1, key markers of SHH MB, confirmed their overexpression in all tumor samples (Fig. 2e). Collectively, these data suggest that Ntn1 expression is not required for SHH MB formation driven by the deletion of Ptch1 in postnatal GNPs in vivo.

Netrin-1 is required for SHH MB maintenance by promoting SHH MB tumor cells survival

While we found that netrin-1 is not required for SHH MB tumor initiation, we next sought to assess a possible role of netrin-1 in SHH MB maintenance. To this end, we acutely inactivated Ntn1 expression in SHH MB tumor cells. First, we used retroviral vector-mediated gene transfer to allow expression of shRNA-Ntn1 (shNtn1 #1 and #2) or scrambled shRNA (shCtrl), as a control, together with GFP in primary murine SHH MB cells, originated from the Ptch1^+/− mouse model. As shown in Supplementary Fig. 2a and b, both shNtn1 induced a significant reduction of netrin-1 expression of 50% and 80% at both mRNA and protein levels, respectively, compared to the control. This downregulation of netrin-1 expression by shNtn1 was also confirmed by immunoblotting of secreted netrin-1 in the cell supernatant (Supplementary Fig. 2c). We then monitored the impact of Ntn1 loss in murine SHH MB infected cells using IncuCyte live-imaging system (Fig. 3a and b). Interestingly, inactivation of Ntn1 expression strongly inhibited cellular growth of murine SHH MB cells (Fig. 3a). We also monitored cell death after incorporation of propidium iodide (PI) over time. At 72h post-infection, we detected a significant increase of PI positive cells by 2-fold and 2.8-fold for shNtn1 #1 and #2, respectively, when compared to the control condition (Fig. 3b). Furthermore, we evaluated the impact of NTN1 loss on SHH MB PDX (ICN-MB-PDX12). Similarly to our previous experiments, shRNA-based knockdown of NTN1 in SHH MB PDX (ICN-MB-PDX12) impaired cellular growth, which was consistent with the increase of cell death over time (Figs. 3c, d, and S2d). Overall, these data demonstrate that netrin-1 confers a cellular growth advantage to SHH MB by promoting cell survival in vitro.

Fig. 3: Netrin-1 plays a role in SHH MB cell survival and tumor growth.

a, b Primary murine SHH MB cells (from Ptch1^+/− mouse model) were infected with shNtn1 prior measurement of survival using the IncuCyte live-imaging system. Cell confluence (a) was monitored during 72h, and cell death (b) was assessed using propidium iodide (PI) uptake over time. (mean ± SD; n = 6; two-way ANOVA, compared to shCtrl: ****p < 0.0001). c, d SHH MB PDX cells (ICN-MB-PDX12) were infected with shNTN1 prior measurement of survival using IncuCyte. Cell number was quantified by counting at indicated time. Living cell number was expressed as a fold increase over control ((c); mean ± SEM; n = 6; two-way ANOVA. ****p < 0.0001, ***p = 0.0001, **p = 0.0012) and cell death ((d); mean ± SEM; n = 7; two-way ANOVA. ****p < 0.0001, ***p = 0.0004, **p = 0.0046) corresponds to the percentage of PI-positive cells. e Kaplan-Meier curves show the effect of loss of Ntn1 expression on survival of pups orthotopically transplanted with murine SHH MB cells (from the Ptch1^+/− mouse model) infected with shCtrl (n = 11) or two different shNtn1, #1 (n = 13) or #2 (n = 14). (Log-Rank (Mantel-Cox) test, compared to shCtrl: ***p = 0.0004, **p = 0.002). f, g Primary murine SHH MB cells (from the Ptch1^+/− mouse model), infected with shCtrl or two different shNtn1, were daily treated or not with a murine recombinant netrin-1 protein (25 ng/ml) to assess the specificity of the shNtn1. Cell confluence (f) was monitored during 96h using IncuCyte. (mean ± SEM; n = 3; two-way ANOVA. ****p < 0.0001, ***p = 0.0004, **p = 0.0033). Cell death, assessed using PI uptake over time, was represented at 72h (g). (mean ± SEM; n = 3; one-way ANOVA. shNtn1#1 *p = 0.0261; shNtn1#2 *p = 0.013). h, i SHH MB PDX cells (ICN-MB-PDX12), infected with shCtrl or a human shNTN1, were daily treated or not with a human recombinant NETRIN-1 protein (600 ng/ml) to assess the specificity of shNTN1. Cell survival was monitored during 96h using IncuCyte. Cell number was quantified by counting at indicated time. Living cell number was represented as a fold increase over control ((h); mean ± SEM; n = 5; two-way ANOVA. ****p < 0.0001; ***p = 0.0002; shCtrl vs shNTN1 **p = 0.0027; shNTN1 vs shNTN1 + recNTN1 **p = 0.009; *p = 0.012) and cell death ((i); mean ± SEM; n = 5; two-way ANOVA. ****p < 0.0001; ***p = 0.0005; shCtrl vs shNTN1 **p = 0.0031 (24h) = 0.0067 (96h); shNTN1 vs shNTN1 + recNTN1 **p = 0.0046 (48h) =0.0072 (72h), *p = 0.0319 (24h) =0.0106 (96h)) corresponds to the percentage of PI-positive cells.

Based on the in vitro studies, we hypothesized that loss of netrin-1 could impact tumor growth in vivo. To test this, we used an orthotopic allograft mouse model, in which murine SHH MB cells transduced with either shControl (shCtrl) or two independent shRNA-Ntn1 (shNtn1 #1 and #2) were directly implanted into the cerebellum of five-day postnatal immunocompetent Swiss CD-1 pups (P5). As shown in Fig. 3e, the loss of netrin-1 in tumor cells slowed down tumor growth in vivo and significantly improved animal survival compared to controls, suggesting that inhibition of netrin-1 could have a potential therapeutic advantage against SHH MB.

In order to rule out possible off-target effects, we then examined whether growth defects induced by netrin-1 knockdown could be rescued by recombinant netrin-1 expression. We thus transduced murine SHH MB cells with shNtn1 #1 or #2, or with shCtrl prior treatment with a murine recombinant netrin-1 protein during 96h. We next monitored cell survival using IncuCyte live-imaging system over time (Fig. 3f and g). Strikingly, recombinant netrin-1 protein was able to fully rescue the inhibition of both cellular growth and cell death observed through depletion of netrin-1 in murine SHH MB cells in vitro. We next confirmed this result using SHH MB PDX. Here again, recombinant NETRIN-1 was able to rescue the genetic silencing of NTN1 on both cellular growth and cell death in vitro (Figs. 3h, i, and S2e).

Overall, our results demonstrate that netrin-1 is essential for SHH MB cell survival and tumor development, and could therefore be considered as an attractive potential therapeutic target.

Anti-netrin-1 antibody inhibits human SHH MB tumor development in vitro and in vivo

Small molecule inhibitors targeting SHH pathway have emerged as an attractive SHH MB therapy, and Smoothened (SMO)-targeted drugs have moved into the clinic and shown to be beneficial to cancer patients⁴⁵. Nevertheless, the first generation of SMO-targeted drugs has shown some limitations due to acquired drug resistance¹⁹. Also, SMO inhibitors cannot be employed to poor prognosis SHH MB patients harboring an activation of the SHH pathway downstream SMO⁶. Together, this emphasizes the need to identify novel, effective treatments for SHH MB. Given the role of netrin-1 in SHH MB tumor maintenance through tumor cell survival in vitro and in vivo, we next explored pharmacological inhibition of netrin-1 as a potential therapeutic strategy in the treatment of SHH MB.

To this end, we took advantage of the blocking anti-netrin-1 antibody, NP137/Net1-H-mAb, which is currently assessed in various phase Ib/II clinical studies in adult patients with advanced cancer (https://clinicaltrials.gov/ct2/show/NCT04652076, https://www.clinicaltrials.gov/study/NCT0554685, https://www.clinicaltrials.gov/study/NCT05546879, https://www.clinicaltrials.gov/study/NCT05605496). Mechanistically, the NP137 antibody disrupts the interaction between the ligand netrin-1 and its associated UNC5 family receptors, thereby preventing netrin-1 activity without affecting netrin-1 expression levels (Supplementary Fig. 3a). NP137 binds both human and murine netrin-1 in the nanomolar range³². We first confirmed the innocuity of the isotypic control, in vitro, on both PDX and murine SHH MB cell survival using the Incucyte system over time (Supplementary Fig. 3b and c). Next, SHH MB PDX cells were treated with NP137 or the control for 96h. Cell survival was evaluated, in vitro, using the IncuCyte system, and revealed that NP137 antibody reduced significantly SHH MB PDX cellular growth by 30% and increased cell death, evaluated by PI staining, compared to the control condition (Figs. 4a–c and S3d). This increase in cell death is consistent with a decrease in Death-Associated Protein Kinase 1 (DAPK1) phosphorylation induced by NP137 treatment at 80 µg/ml (Supplementary Fig. 3e). This observation correlates with previous studies reporting that in the absence of netrin-1, UNC5 receptors trigger apoptosis via Serine/Threonine-protein phosphatase 2 A (PP2A) mediated dephosphorylation of DAPK1^46,47.

Fig. 4: Pharmacological inhibition of Netrin-1 signaling by NP137 induces cell death and inhibits tumor growth of SHH MB PDX.

a–c SHH MB PDX (ICN-MB-PDX12) cells were treated once with Control or NP137 (80 µg/ml) during 96h, under IncuCyte monitoring. Cell number was quantified by counting and was represented either as a ratio to seeded cell number at 0h ((a); n = 4; mean ± SEM; two-way ANOVA. ****p < 0.0001) or as a fold over control ((b); n = 4; mean ± SEM; two-way ANOVA. 24h **p = 0.0022; 48h ***p = 0.0003; 72h **p = 0.0012; 96h ***p = 0.0001). Cell death was assessed using PI ((c); n = 6; mean ± SEM; two-way ANOVA. 24h *p = 0.0314; 48h *p = 0.0436). d, e GNP cells (P7), isolated from mouse developing cerebellum, were treated once at the indicated NP137 concentrations for 96h. Cell confluence (d) and cell death (e) were assessed using IncuCyte. (mean ± SD; n = 3; two-way ANOVA. not significant). f–i Nude mice (n = 7/group), injected subcutaneously with SHH MB PDX (ICN-MB-PDX12) cells, were treated with NP137 (15 mg/kg, 3 times/week) or vehicle. f Graph shows tumor growth. (mean ± SEM; two-way ANOVA. ****p < 0.0001). g Kaplan-Meier curves show the overall survival rates. (Log-Rank (Mantel-Cox) test. **p = 0.005). Images ((h), magnification ×40; scale bar: 50 µm) and quantification (i) of immunostaining for Ki-67 or cleaved Caspase-3 (CC3) in tumor samples. (n = 7/group; mean ± SEM; two-tailed unpaired t test. **p = 0.0049, *p = 0.017). j The ability of NP137 (15 mg/kg) to cross the BBB was evaluated in healthy C57BL/6 mice treated during 24h (n = 3/group). Images of cerebellum immunostained for anti-human IgG1, used to detect NP137 antibody, are shown (scale bar: 50 µm). k, l NP137 (15 mg/kg) or vehicle were injected subcutaneously in symptomatic C57BL/6 Ptch1^+/− mice (n = 4/group) for 4 days, as illustrated in the diagram ((k), Created in BioRender. https://BioRender.com/s18l552). l Images (scale bar: 40 µm) and quantification of immunostaining for Ki-67 and CC3 in tumor samples. (mean ± SEM; two-tailed unpaired t test. ****p < 0.0001). m Immunocompetent C57BL/6 mice, injected orthotopically with murine SHH MB cells (Ptch1^+/−) were treated with NP137 (15 mg/kg, 3 times/week) or vehicle. Kaplan-Meier curves show the overall survival rates. (Log-Rank (Mantel-Cox) test. *p = 0.047). In (f, g, and m), the arrow indicates the start of treatment.

Cerebellar development lasts three weeks in mice and is fully completed by the end of 2 years old in humans, including, GNPs proliferation⁴⁸. Since MB could occur during this timeframe, we therefore asked whether the inhibition of netrin-1 using the antibody NP137 would also affect the proliferation of GNPs. GNPs were isolated from postnatal day 7 mouse cerebella and treated with different concentrations of NP137 for 96h. Interestingly, we observed that NP137, even at high concentration, did not affect proliferation or induce cell death of GNPs in vitro (Fig. 4d and e). Interestingly, these results contrast with the phenotype that we observed in tumor cells, highlighting not only that NP137 could be a safe approach for targeting tumor cells located within an environment under development, but also underlining a tumor specific mechanism for netrin-1 in SHH MB.

Following these in vitro studies, we wished next to evaluate the impact of the pharmacological inhibition of netrin-1 through NP137 on tumor growth in vivo. To this end, as a first approach,  Nude mice were grafted, subcutaneously, with freshly isolated SHH MB PDX cells, prior treatment with either NP137 (15 mg/kg; iv) or the control three times a week. Interestingly, as shown in Fig. 4f and g, NP137 treatment strongly reduced by 45% SHH MB PDX tumor growth and significantly improved mice survival, compared to the vehicle control group. Further analysis of the tumors revealed that NP137 treatment significantly decreased Ki-67 staining in treated tumor masses compared to the control group and increased apoptosis by 2-fold as shown by increased immunohistochemical staining for cleaved Caspase-3 (Fig. 4h and i).

Together, these results demonstrate that the pharmacological inhibition of netrin-1 efficiently reduces human SHH MB tumor development both in vitro and in vivo, suggesting the use of an anti-netrin-1 mAb as an attractive therapeutic strategy for SHH MB in clinic. In that context, we investigated whether the NP137 antibody could cross the blood-brain barrier (BBB). First, we treated subcutaneously healthy mice with NP137 and then, after 24h, collected their cerebellum for immunohistological staining of IgG1 to detect the presence of NP137. As shown in Fig. 4j, NP137 was detected in the cerebellum of healthy treated mice. On the contrary, and as expected, no IgG1 staining was observed in vehicle-treated mice. Thus, these results provide evidence that the anti-netrin-1 mAb, NP137, is able to cross the BBB. Next, we investigated whether pharmacological inhibition of netrin-1 by NP137 could inhibit cell survival in intracranial mouse SHH MB tumors. We therefore used two transgenic C57BL/6 mouse models, the Gfap-cre/Ptc^fl/fl and the Ptch1^+/− mice, which develop aggressive SHH MB at around P10 (pediatric model) and several months (adult model), respectively (Figs. 4k, l, and S3f, g). When tumors developed, these mice were treated subcutaneously few days with NP137 prior to tumor harvesting. Interestingly, in these two models, we observed a reduction of Ki-67 positive tumor cells by 80% in mice treated with NP137 compared to vehicle-treated mice (Figs. 4l and S3g). In addition, in the Ptch1^+/− mouse model, NP137 induced a significant 3-fold increase in apoptosis as shown by the increase of Cleaved Caspase-3 staining, compared to control mice (Fig. 4l). Finally, in light of these results, we investigated whether NP137 could improve animal survival. We therefore injected orthotopically primary murine SHH MB cells, purified from the Ptch1^+/− mouse model, into the cerebellum of adult C57BL/6 immunocompetent mice. Mice were treated three times a week with NP137 (15 mg/kg) or control vehicle until the end of the experiment. Consistent with the reduction in tumor cell proliferation and increase in apoptosis induced by NP137 previously observed, we noted a significant improvement of animal survival compared to the control group upon pharmacological inhibition of Netrin-1 by NP137 (Fig. 4m). All together, these results demonstrate that NP137 could be a promising therapeutic strategy for SHH MB, since NP137, by crossing the BBB, strongly inhibits tumor cell proliferation and induces apoptosis, thereby improving animal survival.

Inhibition of SMO synergizes with anti-netrin-1 in SHH MB

As previously mentioned, therapeutic strategies targeting SHH signaling and specifically the receptor SMO have been developed for SHH MB. Given the crosstalk between both SHH and netrin-1 during brain development, we wondered whether targeting both axes could be an appealing target strategy in SHH MB. For that purpose, we investigated whether the inhibition of SHH signaling through SMO inhibition could synergize with the inhibition of netrin-1.

Murine SHH MB cells, from the Ptch1^+/− mouse model, were treated with either NP137 or the SMO inhibitor GDC-0449 alone, or with the combination of both agents, as indicated in Fig. 5a and b, prior cell death assessment over time, using the IncuCyte system. While neither NP137, nor GDC-0449 induced tumor cell death when used at low concentration, combining NP137 and GDC-0449 induced a robust cell death at 72h, as shown in Fig. 5b. In addition, synergistic effects between SMO and netrin-1 inhibitions were confirmed in these murine SHH MB cells, using the highest single agent (HSA) reference model (Fig. 5c).

Fig. 5: The combination of a SMO inhibitor and an anti-netrin-1 mAb synergizes to inhibit SHH MB tumor growth.

a, b Murine SHH MB cells, from the Ptch1^+/− mouse model, were treated with either 1) NP137 (40 µg/ml), 2) GDC-0449 (10 nM), 3) the association of NP137 and GDC-0449, or 4) respective controls for 72h. The effect of these treatments was evaluated on cell survival over time, using the IncuCyte. a Cell confluence images at 0h and 72h (scale bar: 150 µm). b Cell death, corresponding to the percentage of PI positive cells, is represented at 48h and 72h. (mean ± SD; n = 3; two-way ANOVA. **p = 0.0072, *p = 0.0409). c Representation of calculated synergy maps (3D on the left and 2D on the right) based on dose response matrix (5 × 5). Murine SHH MB cells (Ptch1^+/− model) were treated with a range of NP137, GDC-0449, or a combination of both drugs for 48h, followed by a luminescence-based assay for cell viability. Synergy score was calculated using the HSA reference model. Negative scores indicate antagonism (green) and positive scores indicate synergy (red). d Nude mice (n = 5/group), injected subcutaneously with murine SHH MB (Ptch1^+/− model) cells, were treated with NP137 (15 mg/kg, 3 times/week or its control) and GDC-0449 (2.5 mg/kg, 5 times/week or its control). Graph shows tumor growth. (mean ± SEM; two-tailed Mann-Whitney test at final point, compared to “NP137 + GDC-0449” group: **p = 0.0041 (Control), *p = 0.0221 (NP137), *p = 0.0303 (GDC)). e, f Nude mice (n = 6/group), injected subcutaneously with human SHH MB PDX (Med-1712FH) cells, were treated with NP137 (10 mg/kg, 3 times/week or its control) and GDC-0449 (35 mg/kg, 5 times/week or its control). e Graph shows tumor growth. (mean ± SEM; two-way ANOVA, compared to “NP137 + GDC-0449” group: ****p < 0.0001 (Control), **p = 0.0018 (NP137), *p = 0.0231 (GDC)). f Human GLI1 and NTN1 mRNA levels were determined by quantitative RT-PCR in tumor samples. (mean ± SEM; one-way ANOVA, compared to Control group: GDC-0449 ***p = 0.0007 (GLI1), *p = 0.0467 (NTN1); Combo ***p = 0.0005 (GLI1), *p = 0.0331 (NTN1)). In (d and e), the arrow indicates the start of treatments.

Based on these in vitro studies, the impact of the combined administration of NP137 and GDC-0449 was then evaluated on SHH MB tumor growth in vivo. To conduct this investigation, primary murine SHH MB, from the Ptch1^+/− mouse model, were engrafted subcutaneously into the flank of immunocompromised recipient Nude mice. Mice were treated by either intravenous injection with NP137, or gavage with GDC-0449, or a combination of both treatments (Fig. 5d). To assess the potential effects of combined therapy, we used concentrations of NP137 (15 mg/kg) and GDC-0449 (2.5 mg/kg) that do not induce a strong tumor response in vivo. While no significant differences were observed neither with NP137 nor GDC-0449 treatments alone, by contrast, mice treated with a combination of both molecules showed a significant impact on tumor growth (Fig. 5d). This potentiating effect was further confirmed in a human SHH MB PDX model (Med-1712FH). In Nude mice engrafted subcutaneously with SHH MB PDX cells (Med-1712FH), while neither NP137 (10 mg/kg), nor GDC-0449 (35 mg/kg) alone, reduced significantly tumor growth, when used at low concentrations, the combination of both treatments induced a significant inhibition of tumor growth by half (Fig. 5e). Together, these data demonstrate that SMO inhibition synergizes with netrin-1 inhibition to reduce cell survival and tumor growth.

Accordingly, molecular analyses of tumor samples collected from SHH MB PDX model (Med-1712FH) at the end point revealed, as expected, a reduction in the expression of the SHH target gene, GLI1, in mice treated with GDC-0449. We next wondered whether the expression of NTN1 was affected. Strikingly, we found in tumors treated with GDC-0449 an increased expression of NTN1 (Fig. 5f). This latter observation strongly suggests that SMO inhibitors might activate a survival mechanism in SHH MB cells and could mechanistically explain the synergistic effect observed in vivo.

Overall, our data are consistent with the notions that (i) Smo inhibitors might sensitize cells to NP137 by increasing the expression of netrin-1, and (ii) netrin-1 being a survival factor, its upregulation through GDC-0449 treatment might explain a potential mechanism underlying GDC-0449 resistance, and therefore highlights the putative interest to combine a netrin-1 interfering drug with GDC-0449 to overcome resistance effects and treat effectively SHH MB.

SMO inhibitors activate netrin-1 expression

Given the above data, we then sought to explore how the inhibition of SMO affects netrin-1 expression. First, we confirmed, in vitro, that murine SHH MB tumor cells from Ptch1^+/− mouse model, treated with GDC-0449, exhibited, an inhibition of Gli1 expression, as expected, but also an increase in the expression of netrin-1 at both mRNA and protein levels, as well as its receptors Unc5B and Unc5C (Figs. 6a, b and S4a). In addition, we used another SMO inhibitor, named NVP-LDE225, and showed a dose-dependent increase in Ntn1 expression in murine SHH MB cells at nanomolar concentrations, for both SMO inhibitors, GDC-0449 and NVP-LDE225 (Fig. 6c and d). As expected, both SMO inhibitors, by inhibiting Gli1 expression, reduced SHH MB cell survival in a dose-dependent manner (Figs. 6c, d, and S4b–g). To validate the specificity of netrin-1 upregulation by SMO inhibition, we downregulated Smo expression using two independent shRNAs against Smo (shSmo #1 and #2) in murine SHH MB cells. shSmo decreased the expression of Smo and Gli1, as expected, and inhibited SHH MB cell survival (Figs. 6e and S4h–j). Interestingly, the genetic loss of Smo expression also significantly increased Ntn1 expression at the mRNA level (Fig. 6e). These results were further confirmed in a SHH MB PDX model. Indeed, the treatment of SHH MB PDX cells (ICN-MB-PDX12) with the SMO inhibitor, GDC-0449, for 48h, increased by 2-fold NTN1 expression, as well as UNC5B and UNC5C expression, while GLI1 expression showed a strong decrease (Fig. S4k and l).

Fig. 6: SMO inhibitors activate Netrin-1 signaling pathway.

a qPCR of Ntn1 and Gli1 mRNA levels in murine SHH MB cells upon GDC-0449 treatment (1 µM, 24h). (mean ± SEM; n = 5; two-tailed Mann-Whitney test. **p = 0.0079). b Netrin-1 and Gli1 protein levels were determined by Western Blot in murine SHH MB cells upon GDC-0449 treatments (1–2.5 µM, 48h). Results are representative of three independent experiments. c qPCR of Ntn1 and Gli1 mRNA levels in murine SHH MB cells upon GDC-0449 treatments (24h). (mean ± SEM; n = 6; Kruskal Wallis test, compared to Control: ****p < 0.0001, ***p = 0.0002, **p = 0.0019 (Ntn1/GDC 100 nM) = 0.0032 (Ntn1/GDC 500 nM), *p = 0.0147 (Gli1/GDC 100 nM) = 0.0221 (Ntn1/GDC 250 nM). d qPCR of Ntn1 and Gli1 mRNA levels in murine SHH MB cells upon NVP-LDE225 treatments (24h). (mean ± SEM; n = 6; Kruskal Wallis test, compared to Control: ****p < 0.0001, ***p = 0.0002, **p = 0.0065, *p = 0.042 (Ntn1/LDE 50 nM) = 0.0108 (Ntn1/LDE 100 nM). e qPCR of Ntn1, Gli1, and Smo mRNA levels in murine SHH MB infected cells with shCtrl or two shSmo. (mean ± SEM; n = 5; two-tailed Mann-Whitney test, compared to shCtrl: **p = 0.0079). f Boxplots of Ntn1 mRNA expression upon GDC-0449 treatment (1 µM, 8h) in murine SHH MB cell line (SMB56, from Ptch1^+/− mouse model), analyzed from the public RNAseq database of Liu et al.⁴⁹ (Series GSE130485). (n = 2, median, interquartile 25th/75th. Two-tailed Wald test. ****p < 0.0001). g qPCR of Ntn1 and Gli1 mRNA levels in murine SHH MB cells treated 24h with GDC-0449 (1 µM) or GANT61 (25 µM). (mean ± SEM; n = 4; two-tailed unpaired t test, compared to Control: ****p < 0.0001, *p = 0.0211 (Ntn1) = 0.0299 (Gli1)). h qPCR of Ntn1, Gli1 and Gli3 mRNA levels measured 24h after Gli1- or Gli3-overexpression. (mean ± SEM; n = 3; two-tailed Mann-Whitney test. ****p < 0.0001, ***p = 0.0004). i qPCR of Ntn1 and Gli1 mRNA levels in murine SHH MB cells, from two distinct Ptch1^+/−; Trp53^−/− mouse models, upon GDC-0449 treatment (1 µM, 24h). (mean ± SEM; n = 4; two-tailed Mann-Whitney test. **p = 0.0079, *p = 0.0286). j–l Analyses from public Affymetrix database from Buonamici et al.⁵² (Series GSE19657). SHH MB, from Ptch1^+/−; p53^−/− mouse model, were treated in vivo with NVP-LDE225 (20 mg/kg/day, 4h or 16h; n = 3/group). j Boxplots of Ntn1 mRNA expression. (expressed as Log2FoldChange. median, interquartile 25th/75th. two-tailed t test. ****p < 0.0001, ns: not significant). k Volcano plot of differentially expressed genes. l Enrichment pathways analysis performed with PANTHER tool.

Next, using published RNA-seq dataset of spontaneous MB cells from Ptch1^+/− MB mouse model treated with 1 µM of GDC-0449 for 8h in vivo, we confirmed the upregulation of Ntn1 by SMO inhibitors⁴⁹ (Fig. 6f). Since the increase in Ntn1 expression was already visible at the transcriptional level at 8h, we further investigated the possible molecular mechanism involved. GDC-0449 being an inhibitor of SHH signaling, we primarily explored a potent role of Gli transcription factors on Ntn1 mRNA levels. Our results indicated that while GDC-0449 increased Ntn1 expression in murine SHH MB cells, neither GANT61, a Gli1 inhibitor, nor the overexpression of Gli1 and Gli3 modulated Ntn1 gene expression, suggesting an increase of Ntn1 mRNA level independent of Gli activity (Fig. 6g and h). Moreover, previous reports showed that netrin-1 expression is regulated by TP53⁵⁰. We thus wondered whether netrin-1 could be activated through TP53 in SHH MB. To this end, we took advantage of a genetic SHH MB mouse model in which TP53 is inactivated⁵¹ by crossing Ptch1^+/− and Trp53^−/− animals. As expected, mice developed MB at 100% incidence. Murine SHH MB tumor cells were isolated from two independent Ptch1^+/−; Trp53^−/− animals and were treated with GDC-0449 during 24h. Interestingly, consistent with what we previously observed in tumor cells from SHH MB Trp53 wild-type mouse model, we found an upregulation of Ntn1 expression upon GDC-0449 treatment in the absence of Trp53 (Fig. 6i). In parallel, we analyzed an Affymetrix dataset of murine SHH MB Ptch1^+/−; Trp53^−/− tumors treated with NVP-LDE225, at 20 mg/kg during 4h and 16h⁵². NVP-LDE225 induced not only Ntn1 but also Unc5B and Unc5C expression in a time-dependent manner in murine SHH MB Ptch1^+/−; Trp53^−/− tumors (Figs. 6j and S4m). Overall, these data highlight that SMO inhibitors increased the expression of Ntn1 and its receptors Unc5B and Unc5C in SHH MB independently of p53 status.

TP53 mutations in SHH MB are deleterious and associated with poor survival. Although SHH MB associated with TP53 mutations and SMO activation are not common events, as most of the patients with TP53 inactivation have rather downstream activation of the SHH pathway, they still can be found in some cases. Thus, we sought to assess whether interference with both netrin-1 and SMO could also synergize in Trp53-mutated tumors. Interestingly, similarly to Trp53 wild-type tumors, we observed a synergistic effect between NP137 and GDC-0449 on cell survival of murine SHH MB Trp53^−/− model compared to each monotherapy over time, in vitro (Supplementary Fig. S5a and b).

Together, our results support the view that neither Gli transcription factors nor TP53 accounted for netrin-1 induction upon SMO inhibitor treatment. Hence, to clarify the molecular mechanism involved in netrin-1’s regulation by SMO inhibitor, we decided to compare the differentially expressed genes between NVP-LDE225 treated mice (20 mg/kg; 16h) and non-treated control recipient mice from a murine SHH MB model (Fig. 6k). As expected, we observed a downregulation of SHH signaling genes, such as Gli1 and Ptch1 upon NVP-LDE225. Consistent with the above results, we could also observe an increase of Ntn1 expression in treated mice. Interestingly, our analysis revealed the upregulation of several genes involved in axon guidance cues by NVP-LDE225 including Efna3 and Efna4. In line with this observation, further pathway enrichment analysis revealed that axon guidance mediated by Netrin-1 signaling was the top activated pathway, which could argue for a possible molecular mechanism controlled by SMO inhibitors (Fig. 6l).

Netrin-1 is a survival factor in proliferative cells upon anti-SMO treatment

Interestingly, SMO inhibitor therapy induced the expression of netrin-1 ligand and netrin-1 receptors. Since our scRNA-seq studies showed Ntn1 and its receptors were expressed in different subpopulations of GNPs/GNs and MB cells, consistent with paracrine signaling, we investigated the effects of SMO inhibition on Ntn1 and its receptor expression at the single cell level in tumors. We therefore analyzed the scRNA-seq data from the spontaneous Math1-Cre/SmoM2 SHH MB mouse model (M-Smo mice), treated or not with Vismodegib, published by Ocasio and colleagues⁴⁰ (Fig. 7a). These transgenic M-Smo mice harbor a mutant, constitutively active allele of Smo in the Atoh1 (aka Math1) lineage which mediates SHH hyperactivation and therefore results in only a partial response to Vismodegib due to the persistence of Vismodegib-resistant cell populations⁴⁰. For this analysis, we digitally isolated SHH MB tumor cells, performed ICA and generated UMAPs, and then divided the cells into proliferative, intermediate and differentiating subsets, as in Fig. 1b. In addition, we quantified the percentage of cells expressing specific genes of interest in these three different subsets. First, as expected, we observed that Vismodegib reduced the proliferative population and enriched the differentiated population. In addition, Vismodegib-treated tumors showed increased Ntn1 expression, particularly in the proliferative and intermediate subsets, suggesting that netrin-1 could contribute to maintaining proliferative cells during Vismodegib therapy. By contrast, Vismodegib increased expression of the Unc5 family of netrin-1 receptors specifically in cells of the differentiated node, suggesting an additional role for Netrin-1 signaling in the differentiation processes. These changes in Netrin-1 signaling elements in response to Vismodegib demonstrate a crosstalk between SHH and Netrin-1 signalings, and suggest that netrin-1 may contribute to Vismodegib resistance by supporting proliferation when SHH is inhibited. Overall, these results are consistent with the idea that Netrin-1 signaling could contribute to the persistence of so-called “Vismodegib-resistant” tumor cells.

Fig. 7: Netrin-1 is a survival factor in proliferative cells upon anti-SMO treatment.

a t-SNE projection of tumor cells from the spontaneous Math1-Cre/SmoM2 SHH MB mouse model (M-Smo mice) treated with vehicle (n = 5) or Vismodegib (n = 5), based on PCA and color-coded for the expression of indicated markers. Metaclusters are defined by hierarchical clustering of single-cell RNA clusters. The percentage of cells expressing the genes of interest are quantified in the proliferative (green), intermediate (blue) and differentiating (purple) subsets. Analyses were performed using previously published single-cell RNA sequencing (scRNA-seq) data⁴⁰. b Representation of calculated synergy maps (3D above and 2D below) based on dose response matrix. Primary murine SHH MB cells, from the hGFAP-Cre/SmoM2 (G-Smo) model, were treated with a range of NP137, GDC-0449, or a combination of both drugs in an 5 × 5 matrix as indicated, for 72h, followed by a luminescence-based assay for cell viability. Synergy score was calculated using the HSA reference model. Negative scores indicate antagonism (green) and positive scores indicate synergy (red). Results are representative of three independent experiments.

Next, to validate our hypothesis on the contribution of netrin-1 to the survival of SMO inhibitor-resistant tumor cell populations, we investigated whether the pharmacological inhibition of netrin-1 could improve the response of SHH MB-resistant tumors to SMO inhibitors. For this purpose, primary murine SHH MB from the hGFAP-Cre/SmoM2 (G-Smo) mouse model, freshly isolated, were treated in vitro with a range of NP137, GDC-0449, or a combination of both drugs prior to cell survival assessment, as shown in Fig. 7b. HSA analysis confirmed the synergy between NP137 and GDC-0449 and showed that the inhibition of netrin-1 could potentiate the effect of GDC-0449 in SMO inhibitor-resistant mouse model.

Overall, our finding provide interesting insights into the potential of combining an anti-netrin-1 strategy with an anti-SMO approach for the treatment of SHH MB.

Discussion

In the present study, we characterize the role of the protein netrin-1 in the context of MB, and especially in the SHH subgroup. We were able to show that netrin-1 represents a key moiety implicated in SHH MB cell survival, which could be further exploited for alternative treatment approaches.

More specifically, we primarily showed that genetic deletion of Ntn1 in postnatal GNPs has no impact on SHH MB tumorigenesis. This absence of phenotype observed during SHH MB initiation could be explained by the early deletion of Ntn1 which may contribute to the escape of the cells to netrin-1 dependency observed in SHH MB. Consistent with this idea, postnatal GNPs were found unaffected by the anti-netrin-1 antibody NP137 treatment. However, we cannot exclude that the lack of phenotype is due to a compensation mechanism that might take place as netrin-1 is expressed by other cells in postnatal cerebellum, such as in cerebellar interneurons^53,54. Although it is fair to say that it is now acknowledged that contrary to what was initially thought, netrin-1 is not a diffusible soluble protein but is retained closely to its site of secretion^44,55.

Interestingly, we here clearly pinpointed a netrin-1 dependent cell survival mechanism in SHH MB tumors. We specifically observed a high level of Ntn1 in Ptch1 mutant SHH MB compared to P7 wild-type GNPs, while netrin-1 receptors displayed inverse expression profiles. This observation suggests that netrin-1 dependence receptors are not primed to trigger cell death of SHH MB cells. Along with studies in SHH MB mouse models, high expression of netrin-1 was detected in SHH MB patients, as well as in human SHH patient-derived xenografts, when compared to other MB subgroups. Interestingly, acute or pharmacological inhibition of netrin-1 prevented tumor cell growth both in vitro and in vivo. Hence, we can argue that the balance of expression levels between ligand/receptors might confer a selective survival advantage of SHH MB cancer cells. Nevertheless, one intriguing question remains as to how this dependency to netrin-1 is acquired by the SHH MB cells while such dependency does not exist for normal GNPs. One mechanism could involve a temporal regulation of netrin-1 and its receptor expression levels.

Together, these data not only provide evidence for the importance of netrin-1/receptor pairs in the regulation of SHH MB tumor progression, but also suggest a specific therapeutic strategy for this subgroup of MB. Although the mechanism underlying netrin-1 overexpression is not known, and still needs further investigations, netrin-1 inhibition through anti-netrin-1 mAb showed positive outcomes. This is of great interest as anti-netrin-1 mAb is currently in several clinical trials alone or in combination with chemotherapies and immune-checkpoint inhibitors. Based on our data, it could therefore be interesting to explore further the clinical benefit of anti-netrin-1 therapy in SHH MB. Of note, the anti-netrin-1 mAb did not show any toxic effects in vitro on GNPs making the idea even more appealing. Finally, even though it remains to be further studied, the use of anti-netrin-1 mAb in SHH MB treatment appears very attractive as it seems to be able to cross the blood-brain barrier (BBB), a barrier described as intact in the SHH MB subgroup⁵⁶. In line with this, recent studies have shown that endothelial netrin-1 and its receptor Unc5B are important regulators of BBB integrity⁵⁷. Also, it was shown that an antibody blocking netrin-1 binding to Unc5B causes a transient BBB breakdown followed by neurovascular barrier resealing⁵⁸. Therefore, we could envision a scenario whereby the anti-netrin-1 mAb by blocking the interaction between netrin-1 and its receptors of the Unc5 family might also modify the integrity of the BBB, facilitating its passage through the barrier and then impacting on MB cell death. The use of the anti-netrin-1 mAb could thus have two independent effects that ultimately enhance its anti-tumoral activity in SHH MB.

We next demonstrated that the combination of anti-netrin-1 antibody with SMO inhibitors, a therapy that has already been assessed in clinic for SHH MB, led to a synergistic effect on cancer cell survival inhibition compared to each of the monotherapies alone in all models tested. So far, SMO inhibitors have had limited success due to acquired drug resistance. In this context, characterization of the mechanisms underlying such resistance or identification of alternative targetable pathways represent today more effective therapeutic strategies to block SHH MB. Interestingly, our results showed an increase of netrin-1 expression upon SMO inhibition. At the cellular level, looking at a single cell RNA-seq dataset, we found that SHH MB cells that remain proliferative after Vismodegib treatment correlated with enrichment of Ntn1 expression. Furthermore, we have shown in a resistant model that netrin-1 inhibition improved the response of SHH MB-resistant tumors to SMO inhibitors. This prompted us to envision a mechanism by which tumor cells by upregulating a survival pathway could bypass drug-induced cell death. We thus investigated the mechanism underlying netrin-1 regulation in the presence of SMO inhibitors. Our results point out a transcriptional regulation of netrin-1. Since we noted strong netrin-1 expression in the SHH MB subgroup, we would expect SHH signaling to directly control netrin-1 expression. However, it turns out that Glis transcription factors are not involved in the transcriptional regulation of netrin-1 in SHH MB. On the other hand, in the presence of SMO inhibitors, it is very likely that the activity of as yet unknown transcriptional regulators is responsible for the increase in netrin-1 transcriptional activity. Further studies will be therefore required to assess the detailed basis of netrin-1 upregulation in SHH MB upon SMO inhibition, providing a mechanism that may be important for tumor progression, and potentially blocking SMO-mediated resistance. In addition, our data demonstrate that synergy between anti-netrin-1 and SMO inhibitors impaired SHH MB tumor growth independently of p53 status. This result reinforces that combination of drugs targeting the SHH pathway with other therapeutic agents could be efficient towards promising therapies. This data also suggests that in the context of SHH MB, inhibition of both Netrin-1 and SHH signaling pathways could provide clinical benefit for patients with SHH MB regardless of their p53 status, hence opening a therapeutic window for a subset of SHH MB mutated-p53 patients.

Methods

Ethics statement

Animal studies were conducted in accordance with the French Animal Care Guidelines and approved by the Ethic Committee of the Institut Curie CEEA-IC #118 (Authorization APAFiS#26879-2020081315161665-v1 given by the National Authority) and the local Animal Ethic Evaluation Committee of Centre Leon Bérard (CECCAPP: C2EA-15 approved by the French Ministry of High School and Research -authorization n°:  APAFiS#9058). The protocol also complied with the internationally established 3R principles, in accordance with the UKCCCR guidelines.

Human samples for xenograft studies were obtained with informed consent of patients, and all experimental procedures were performed following guidelines from the Institutional Review Board at Necker Hospital.

Bioinformatics analyses

Two primary tumor datasets were used to show that netrin-1 is upregulated in human SHH subgroup (Figs. 1d–f and S1a): a normalized primary human medulloblastoma gene expression data (763 samples of patients, GEO accession GSE37382) from Affymetrix array, with sample affiliation published in Cavalli et al.⁵ and a normalized primary human medulloblastoma protein expression data (45 samples of patients, MassIVE online repository: MSV000082644) with sample affiliation published in Archer et al.⁴¹. Complex Heatmaps for the genes of interest were generated by ComplexHeatmap R package.

Netrin-1 upregulated was also displayed in mouse SHH MB (Fig. 1a) using the Affymetrix gene expression (3 GNPs vs 6 SHH MB) from Kutscher et al.⁵⁹.

The effect of SMO inhibition on Ntn1 and netrin-1 receptor expression levels was analyzed using two datasets (Fig. 6f and j): raw mouse medulloblastoma model Affymetrix gene expression (3 controls, 12 sensitive samples and Vehicle vs NVP-LDE225 treatment after 4 and 16h, GEO accession GSE19657) published in Buonamici et al.⁵² and Normalized RNAseq expression data (4 samples: 2 Controls vs 2 GDC treated samples, GEO accession GSE130485) published in Liu et al.⁴⁹.

For all the Affymetrix datasets, the raw CEL files were pre-processed using affy R package⁶⁰, with the custom chip definition file (CDF) hugene11sthsensgcdf for human and mouse4302mmentrezgcdf for mouse data, and then normalized by the rma method⁶¹. The differentially expressed genes were detected using limma R package⁶², and sorted by Foldchange and adjusted p-values (two-tailed t test). The box part of the boxplots represents the interquartile range while the whisker bonds of the boxplots indicate the highest and smallest values within 1.5 times interquartile range above and below the 75th and 25th quantiles respectively. Volcano plot (Fig. 6k) was generated by EnhancedVolcano R package. Geneset enrichment analysis was applied to the Affymetrix data using PANTHER tool (Fig. 6l). For the RNAseq dataset, differential expression analysis between Control and GDC was carried out using DESeq2 R package⁶³ (two-tailed Wald test).

PDX whole exome sequencing

Briefly, genomic DNA was extracted from PDX samples using the DNeasy Blood & Tissue Kit (Qiagen), according to the manufacturer’s protocol. Then, 50 ng of genomic DNA were used for the library preparation with the SureSelect XT-HS v2 Human All Exon v8+NCV kit (Agilent). The pool of multiplexed libraries was finally sequenced on a NovaSeq 6000 sequencer (Illumina), from Somatic Genetic Unit at Institut Curie, in the 100 bp paired-end model. Raw data were processed using VEGAN pipeline developed at the Institut Curie Bioinformatics Core Facility and available at https://gitlab.curie.fr/data-analysis/vegan. The current pipeline, VEGAN v2.1, is inspired from nf-core/sarek⁶⁴. Briefly, raw sequencing reads were first checked for quality with Fastqc and aligned on the human hg38 reference genomes using bwa-mem algorithm. Aligned BAM files were then filtered by using SAMtools. Somatic variants (SNVs and Indels) were called using MuTect2⁶⁵ with the default setting in tumor only mode applying germline resource (gnomAD VCF file). The raw variants are then filtered by FilterMutectCalls and annotated by SnpEff⁶⁶. Finally, copy number analysis was performed by FACETS⁶⁷ using unmatched control. These WES data have been deposited in the SRA public database under the accession code PRJNA1234717.

PDX RNA sequencing

Transcriptomic data were analysed from the public database of Forget et al.⁶⁸, deposited in the European Genome-Phenome Archive under the accession code EGAD00001004327. Raw data were processed using an in-house pipeline developed at the Institut Curie Bioinformatics Core Facility and available at https://github.com/bioinfo-pf-curie/RNA-seq. Briefly, raw sequencing reads were first checked for quality with Fastqc and trimmed for adapter sequences. Trimmed reads were then aligned on the human hg38 (PDXs) reference genomes using the STAR mapper, up to the generation of a raw count table per gene. The downstream analysis was then restricted to protein-coding genes, and all genes having 0 counts across all samples were filtered out before. Finally, raw counts were normalized (Variance stabilization process) using the rlog method from Bioconductor DESeq2 package⁶³. Differential gene expression analyses were performed using DESeq function using the filtered raw counts.

PDX proteome

Proteomic data were analysed from the public database of Forget et al.⁶⁸, deposited in the ProteomeXchange database (PRIDE: PXD006607). All peptide samples were separated and analysed by nanoLC-MS/MS using an UltiMate 3000 RSLCnano system coupled to an Orbitrap Fusion mass spectrometer, as described in the published work of Forget et al.⁶⁸ The peptides were quantified by Spectronaut pipeline (v17.1)⁶⁹. For peptides identification, the data were searched against the Homo Sapiens (UP000005640_9606_012021) UniProt database. For protein quantification, label-free quantification (LFQ) was carried out in MyProMS⁷⁰ using the XICs from all non-proteotypic peptides and allowing missed cleavages. Median correction and variance scale normalization were applied on the total signal to correct XICs. Proteins with at least three detected peptide ratios with no more than 34% missing values across all samples were considered. After imputation by PCA method⁷¹, the number of proteins deemed valid for downstream analyses was boosted to 6869 proteins.

The molecular subgroup of the PDXs at RNAseq and Proteome level was confirmed applying the UMAP clustering⁷² and using the same reference public primary tumor cohort of 40 MBs (RNAseq: EGAD00001004327 and Proteome: PXD006607)⁶⁸.

Single cell RNA sequencing

We analyzed Ntn1 and netrin-1 receptor expression in publicly available scRNA-Seq data⁴⁰ (GEO accession number GSE129730), using software tools as previously described⁴⁰. We re-analyzed gene expression matrices generated from cerebella from P7 WT mice and MBs from P15 Math1-Cre/SmoM2 mice. After PCA and Louvain cluster analysis, we identified and excluded cells in clusters showing stromal cell types, and thus subsetted the matrix to only GNPs, GNs and MB cells. In separate analyses for cerebella and for MBs, we identified four components using independent components analysis (ICA). From the ICA components, we generated clusters and a 2-dimensional Uniform Manifold Approximation and Projection (UMAP). Finally, we generated “metaclusters” by hierarchical clustering of cluster centroids to sort cells into three main groupings. We separately projected the WT or tumor cells into 2-D UMAPs, color-coded the expression of specific marker genes on the UMAPs, and quantified by percent of cells expressing specific genes of interest in the proliferative, quiescent and differentiating subsets.

Cell lines

HEK293T (ATCC, CRL-3216) cells, used for virus production were cultured in DMEM (Dulbecco’s Modified Eagle’s Medium) supplemented with 10% fetal bovine serum and 10,000 Unit/ml penicillin/streptomycin (Thermo Fischer Scientific), at 37°C in a humidified atmosphere containing 5% CO₂.

GNP primary culture

Cerebellar GNPs (Granule Neuron Progenitors), obtained from 6/7-day old Swiss CD-1 (RjOrl:SWISS) mouse pups, were purified and cultured as previously described⁷³. Briefly, GNPs were isolated by enzymatic digestion with trypsin/DNase (Worthington Biochemicals) followed by trituration, and then, they were separated by centrifugation on a Percoll density gradient (Sigma-Aldrich), collected and washed. GNPs were grown in Neurobasal medium with B27 supplement, 2 mM glutamine, 10,000 Unit/ml penicillin/streptomycin (all from Thermo Fischer Scientific), 0.45% D-Glucose, 1X SPITE medium supplement and 1X linoleic acid-oleic acid (all from Sigma-Aldrich). SHH-N Terminus recombinant or the Smoothened ligand (SAG) at 100 nM (Enzo Life Sciences) was also added to the culture medium. For culture experiments, GNPs were maintained on a surface coated with a poly-D-lysine solution (Sigma) and Matrigel (BD Biosciences).

SHH MB (PDX and murine) primary culture

Murine SHH MB tumor cells were obtained from spontaneous medulloblastoma arising from Ptch1^+/−;Trp53^+/+, Ptch1^+/−;Trp53^−/− and hGFAP-Cre/SmoM2 mouse models, as previously described⁷⁴.

Patient-derived xenografts (PDXs) were generated from primary human SHH MB sample (ICN-MB-PDX12) and from primary human group3 MB sample (ICN-MB-PDX3 and ICN-MB-PDX7), from the Children’s Necker Hospital in Paris, transplanted orthotopically, after cell dissociation, into the cerebellum of immunocompromised NOD/SCID mice, as previously described⁷⁵. Once established, PDX models were maintained by serial transplantations into the subscapular fat pad of Nude mice. The SHH MB PDX, Med-1712FH, was generated by the Olson lab^76,77.

SHH MB tumor cells were purified and cultured as previously described⁷⁴. Briefly, SHH MB tumor cells were dissociated by enzymatic digestion with DNaseI (Worthington Biochemicals), Collagenase P (Roche) and Collagenase/dispase (Roche), and then, they were separated by centrifugation on a Percoll density gradient (Sigma-Aldrich), collected and washed. SHH MB tumor cells were grown in Neurobasal medium with B27 supplement, 2 mM glutamine, 10,000 Unit/ml penicillin/streptomycin (all from Thermo Fischer Scientific), 0.45% D-Glucose, 0.4% BSA, 16 µg/ml N-acétyl Cystéine (all from Sigma-Aldrich). For culture experiments, SHH MB tumor cells were maintained on a surface coated with a poly-D-lysine solution (Sigma) and Matrigel (BD Biosciences).

Treatments

The anti-netrin-1 antibody (Net1-H-mAb/NP137) and the isotypic control were provided by the company NETRIS Pharma. GDC-0449 and NVP-LDE225 was purchased from Selleckchem (Vismodegib, GDC-0449, No. S1082 and Sonidegib, NVP-LDE225, No. S2151). Human and murine recombinant netrin-1 were purchased (R&D Systems, 6419-N1 and 1109-N1) and added every day in the culture medium. SHH was purchased (SHH-N Terminus recombinant, R&D Systems). Heparin was purchased from Sigma. For in vitro treatment experiments, cells were seeded and treated the following day.

shRNA constructs, virus production, and infection

Ntn1 shRNA experiments were performed using pLKO.1 vectors from Sigma (SHCLNG-NM_008744 for mouse). siRNA sequence were designed using http://biodev.extra.cea.fr/DSIR/DSIR.html and http://www.broadinstitute.org/rnai/public/seq/search. For Control shRNA, siRNA were designed against the anthozoan coelenterate Renilla luciferase. Then forward oligonucleotides were annealed with their reverse sequence (Supplementary Table 1) and cloned in AgeI/EcoRI linearized pLKO.1 vector. A GFP tag were introduced in all Ntn1 shRNA pLKO.1 vectors using In-Fusion HD cloning kit (Clontech). Briefly, vectors were linearized with BamHI enzyme and a GFP-P2A sequence amplified from a pSBTet-GH vector (AddGene) was inserted upstream of puromycin sequence in pLKO.1 by recombinational cloning.

Smo shRNAs, using pLV-EGFP, were purchased from VectorBuilder.

Retroviruses were generated, as previously described⁷⁴, by co-transfecting pLKO-GFP vectors encoding murine shRNA targeting Ntn1 or the scramble and packaging plasmids encoding an ecotropic envelope and the Gag and Pol genes, in HEK293T cells, using X-tremeGene 9 DNA transfection Reagent (Sigma-Aldrich). Retroviral particles were harvested 36–48–60h post-transfection, and then, murine primary SHH MB cells were infected with the particles. For all experiments, murine SHH MB were sorted on a FACSAria™III (BD Biosciences) cell sorter on the basis of GFP expression.

Lentiviruses pLKO-GFP vectors encoding human shRNA targeting NTN1 or the scramble were produced in HEK293T cells by co-transfecting packaging plasmids pS-PAX2 and pMD2-VSVG with pLKO-derived vectors, using X-tremeGene 9 DNA transfection Reagent (Sigma-Aldrich). Lentiviral particles were harvested 36–48–60h post-transfection, and then, human primary SHH MB cells were infected with the particles.

Western blot

Flash-frozen medulloblastoma tumor samples or cells were lysed in RIPA buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) mixed with protease inhibitor cocktail (Sigma Aldrich). Proteins were quantified by BCA protein assay (Pierce), then separated by SDS-PAGE and transferred to Immobilon-P membranes (Millipore). Membranes were immunoblotted with primary antibodies listed in Supplementary Table 2. Antibody binding was visualized with the Enhanced Chemiluminescence System (Pierce).

To detect secreted netrin-1 in the supernatant, cells were treated with heparin (200 µg/ml) for 24h prior to harvesting to solubilize netrin-1 which could remain trapped in the matrix and on cell surface. Then, supernatant was collected and directly loaded on SDS-PAGE gel for netrin-1 expression analysis by immunoblotting. To monitor cell number and heparin efficiency, cells from each corresponding well were lysed, proteins were then quantified and analyzed by western blot using netrin-1 and actin antibodies. We were not able to detect any netrin-1 expression in cell lysate upon heparin treatment.

RNA extraction, cDNA synthesis, and qPCR

Total RNA was extracted either using RNeasy Plus mini kit (Qiagen) according to manufacturer’s instructions for cells or using TRIzol reagent (Thermo Fisher Scientific) for tumor samples. DNase I (Thermo Fisher Scientific) treatment of total RNA was performed to eliminate the potential genomic contamination. cDNA was synthesized using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific) following provider’s protocols. Quantitative real-time PCR was performed with either Power SYBR Green PCR Master Mix (Applied Biosystems) or Premix Ex Taq^TM PCR Amplification kit (Takara) and TaqMan® probes (Roche). Primers for qPCR are listed in Supplementary Table 3. Data were normalized with the endogenous controls (TBP and HPRT) and expressed as the fold change with respect to the control sample value.

Cell viability and cell death assay using IncuCyte live-imaging system

For cell viability and cell death experiments, SHH MB tumor cells were plated in 96-well plates at a density of 5 × 10⁴ and 7.5 × 10⁴ cells/well for murine and PDX models, respectively, prior infection with shNtn1 or not. For pharmacological studies, the next day (day 0), cultured tumor cells were treated once with either NP137, GDC-0449, or different combinations and respective controls, at the indicated concentrations in the Figures. GNPs were also plated in 96-well plates at a density of 1 × 10⁵ cells/well and treated once with NP137 at indicated concentrations. Propidium iodide (PI at 0.3 µg/ml; Sigma) was also added in the medium at day 0. Then, the plates were scanned for phase contrast and PI staining every 3h during 72–96h, using the IncuCyte imager with a 4X objective (Essen BioScience). For GNPs and murine SHH MB cell models, proliferation was measured using quantitative kinetic processing metrics from time-lapse image acquisition and was presented as percentage or ratio of cell confluence over time, normalized to time 0h of each condition. For PI staining, which fluorescently stains the nuclear DNA of cells that have lost plasma membrane integrity, the percentage of PI positive cells, corresponding to red object confluence, was divided by the phase object confluence percentage in each well, indicating the level of dead cells in each well. However, for the SHH MB PDX cell model, the metrics of the IncuCyte system to measure cell confluence were not used as they were not suitable for this model. In this case, cell number (living and dead cells) was quantified by counting, in six random fields by condition, on images taken by the IncuCyte system, at indicated time. The mean of cell number was represented either 1) as a ratio to seeding number cells at time 0h or 2) as a fold over control, or 3) as the percentage of PI positive cells to evaluate cell death.

Synergy assessment

Cells were seeded into 96-well plates and treated with a range of NP137, GDC-0449, or the association of both drugs in a drug-combination matrix, at indicated concentrations and indicated time, for murine SHH MB. CellTiter-Glo® Luminescent Cell Viability Assay, performed according to the manufacturer’s instructions (Promega Corporation, Madison, WI, USA), was used as a readout of cell viability. Synergy score was calculated with the Synergy Finder on line software, using the highest single agent (HSA) reference model⁷⁸. Negative scores indicate antagonism, whereas positive scores indicate synergy.

Animal husbandry

Pups (males and females) at Postnatal P7 or P5 ages for, respectively, GNPs culture or orthotopic transplantations were obtained from Swiss CD-1 (RjOrl:SWISS) mice that were purchased from Janvier Labs. Adult immunodeficient female NMRI Nude mice were also obtained from Janvier Labs. Adult C57BL/6 mice (males and females) were purchased from Charles River.

All these animals, used for this study, were housed under pathogen-free conditions with sterilized food and water provided ad libitum and maintained on a 12h light and 12h dark cycle, temperatures of 19–21°C with 40–60% humidity, at the animal facility, in accordance with the recommendations of the European Community (2010/63/UE) for the care and use of laboratory animals.

Transgenic animals

Ntn1 flox C57BL/6 mice were described in Dominici et al.⁴⁴. Briefly, these mice were created (Genoway) by inserting two loxP sites flanking the coding sequences containing both the principal ATG (based on Ntn1 cDNA sequence NM_008744) and the cryptic ATG (based on Ntn1 cDNA: BC141294) and the alternative promoter described in intron 3⁷⁹. Removal of the floxed sequence by Cre recombinase generates a null allele resulting in the silencing of Ntn1 in the Cre-targeted tissue.

Ptch1 flox mice have been described previously⁸⁰. Briefly, two loxP sites flanking exon 3 at Ptch1 locus were introduced. First, inducible Atoh1^CreERT2; Ptch1^flox/flox and Atoh1^CreERT2; Ntn1^wt/flox were generated using Atoh1^CreERT2 (obtained from Jackson Laboratory, 007684)^43,81. Then, both strains were crossed to generate breeder pairs of control (Atoh1^CreERT2; Ptch1^flox/flox; Ntn1^wt/wt) and Ntn1 deficient (Atoh1^CreERT2; Ptch1^flox/flox; Ntn1^flox/flox) mice. The silencing of Ptch1 and Ntn1 Atoh1-Cre-expressing tissue was induced by tamoxifen injection into newborn animals at P2 or P3. Tamoxifen solution was dissolved in corn oil (Sigma) at a concentration of 10 mg/ml and was delivered by a unique subcutaneous injection at a dose of 75–100 μg/g body weight. The genotypes of littermate were then determined.

Spontaneous SHH MBs were obtained from either Ptch1^+/− C57BL/6 mice³⁹, Gfap-Cre/Ptch^fl/fl C57BL/6 mice¹⁰ or hGFAP-Cre/SmoM2 C57BL/6 mice (G-Smo)⁴⁰.

All experiments with transgenic mice were performed in males and females equally distributed between groups.

In vivo animal treatment studies

For in vivo experiments, 2 × 10⁶ SHH MB PDX cells (Med-1712FH and ICN-MB-PDX12) or 5 × 10⁴ primary murine SHH MB from Ptch1^+/− mouse model, purified as previously described, were injected either 1) subcutaneously in the flank of 6-week-old female Nude mice or 2) orthotopically in the cerebellum of 8-week-old female C57BL/6 mice. For subcutaneous models, when tumors were measurable (50-100 mm³), the mice were randomly distributed into the different groups (n = 5/7 per group) and treated with the different molecules, as indicated in the Figures. For monotherapy experiment, mice were treated three times a week, by intravenous (iv) injection, with either NP137 (15 mg/kg in trehalose buffer) or the vehicle, until the end of the experiment. For combination experiment, mice were treated, until the end of the experiment, either with 1) NP137 (10 mg/kg for Med-1712FH PDX model and 15 mg/kg for murine SHH MB model, 3 times a week by iv) or 2) GDC-0449 (35 mg/kg for Med-1712FH PDX model and 2.5 mg/kg for murine SHH MB model, in 2% DMSO, 30% PEG300, 5% Tween80, H20 by oral gavage, 5 times a week) alone and respective controls, or with 3) the combination of both molecules, NP137 + GDC-0449 at the same above concentrations. The control group was treated with the vehicle of these two molecules. Tumor growth was monitored and measured with a caliper over time. Tumor volume was calculated with the following formula: (length × width²)/2. We confirm that the volume of subcutaneous tumors never exceeded the maximum size of 2000 mm³ approved by our ethics committee.

For Gfap-Cre/Ptc^fl/fl and Ptch1^+/− C57BL/6 mice treatments, symptomatic mice were randomly divided into two groups (n = 3 or 4 per group, respectively) and injected subcutaneously with NP137 (15 mg/kg) or vehicle, for 24h or 4 days, respectively. Cerebellar tumor masses were analyzed by immunostaining.

Orthotopic transplantation in mouse pups

Primary murine SHH MB cells, originated from the Ptch1^+/− mouse model, were purified, infected with shControl or two shNtn1 (#1 and #2) as previously described, and then FACS-sorted, based on GFP expression. Those infected murine SHH MB cells were orthotopically injected into the cerebellum of 5-day old Swiss CD-1 (RjOrl:SWISS) mouse pups (males and females equally distributed between groups). The pups were anesthetized by hypothermia and analgesized with buprenorphine. Then, a small incision was made in the back of the skull, between the two ears, and 1 × 10⁵ murine SHH MB cells (resuspended in 2 µl of Matrigel) were injected into the cerebellum to a depth of 1.5 mm through a Hamilton syringe. Mice were checked daily for tumor growth and euthanized when they had first symptoms of cerebellar tumor. Tumor samples were collected for gene expression analyzes.

Immunohistochemistry

Cerebellum and tumor tissues were collected and fixed in 4% paraformaldehyde. Then, they were embedded either in paraffin, cut into 4 µm sections and colored with Hematoxylin/Eosin (H&E) or in PBS containing 7.5% gelatin and 10% sucrose and cut into 20 μm sections with a cryostat. Sections were then stained for Ki-67, Cleaved Caspase-3, Netrin-1, CaBP, or anti-IgG1 with antibodies listed in Supplementary Table 2. Immunodetection was performed either using DAB Substrate-Chromogen (VectorLabs) and counterstained with hematoxylin or fluorescent secondary antibodies and Hoechst. Image acquisitions were performed on a Zeiss microscope. Immunohistochemistry staining was assessed by quantification of stained positive cell areas using ImageJ software.

In situ hybridization was done as previously described⁴⁴. Briefly, the sections were fixed in 4% PFA, treated 2 min with Proteinase K (20 µg/ml, Invitrogen), and acetylated. Then, the sections were blocked 2h in blocking buffer 1 (50% formamide, 5X SCC, 1X Denhardt, 0.2 mg/ml salmon sperm, 0.25 mg/ml tRNA from yeast) and hybridized overnight at 68°C with mNetrin-1 digoxygenin labeled probe. The sections were washed in 5x SCC and 0.2x SCC at 72°C for respectively, 5 min and 2h. Then, the sections were blocked with blocking buffer 2 (100 mM TrisHCl pH 7.6, 0.15 mM NaCl, 0.1% tween20, 5% NGS) for 2h and the sections were incubated overnight at room temperature with anti-Digoxygenin-AP (Roche) at 1/5000 in blocking buffer 2. Finally, Netrin-1 mRNA is detected with NBT/BCIP (Roche).

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (version 9, La Jolla, CA, USA). Unpaired two-tailed Student’s t tests or Mann–Whitney test were used to analyze experimental data between two groups. Data with more groups were analyzed either with the One-way or Two-way ANOVA test. For the survival curves, performed with Kaplan–Meier analyses, the p-values have been calculated using a log-rank (Mantel–Cox) analysis. Results were given as mean ± SEM for in vivo and in vitro experiments using primary cell culture. For IncuCyte experiments, analyses were performed at least in triplicate and were representative of three independent experiments, and results were given as mean ± SD or SEM. Statistical significance was assumed when p-value < 0.05.

Differential gene expression (DGE) analysis was performed on scRNA data using the likelihood-ratio test for single cell gene expression⁸² implemented in Seurat V3 (satijalab.org).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.