Analysis of netrin‐3 expression in cancer

We first attempted to identify the cancer‐associated expression of netrin‐3 by conducting a bioinformatics search in public cancer databases. We first focused on a dataset encompassing RNA‐sequencing data from 675 cancer cell lines (Klijn et al, 2015) overlapping 16 cancer cell types. Compared to netrin‐1 gene expression, which is displayed by most cancers, netrin‐3 gene expression was largely represented by two specific clusters corresponding to neuroblastoma (NB) and small cell lung cancer (SCLC) (Fig 1A). Of note, the expression of netrin‐3 and netrin‐1 seemed to be mutually exclusive (Fig 1B). While netrin‐1 was detectable (FPKM ≥ 1) in 43.3% of cell lines, netrin‐3 was only detected in 4.8%, and their common expression occurred in only 0.015% of cell lines (P = 0.037—Fisher’s exact test) (Fig 1C), arguing in favor of a selective event driving high netrin‐3 gene expression specifically in these two neuroepithelial cancer indications (Rindi et al, 2018).

Figure 1

Netrin‐3 expression in cancer cell lines

Exploring the expression by RNA sequencing of netrin‐1 and netrin‐3 among 675 cancer cell lines spanning 16 cancer types.

AExpression heat map of netrin‐3 and netrin‐1. Receptors of netrin‐1 and putative receptors of netrin‐3: DCC, UNC5‐A, UNC5‐B, UNC5‐C, UNC5‐D, Neogenin, APP, CD146 are presented. The color legend (right‐hand side) describes all tumor types included in the heat map. Asterisks indicate the two main pathologies in which netrin‐3 expression is detectable.

BAnalysis of netrin‐1 and netrin‐3 expression in tumor cell lines and ranked according to netrin‐1 expression.

CPercentage of cell lines positive for netrin‐1 and/or netrin‐3 expression. Co‐occurrence of expression was calculated using Fisher’s exact test. P‐value is indicated in the table.

Netrin‐3 as a prognostic marker for NB

We thus investigated more closely the expression of netrin‐3 gene in NB, which is the most common extracranial pediatric solid tumor, responsible for 15% of all childhood cancer‐related deaths, and arises from the sympatho‐adrenal lineage of neural crest cells (Pugh et al, 2013; Matthay et al, 2016; Mohlin et al, 2019; Chang et al, 2020). According to the INSS classification, NB patient outcome is strongly associated with tumor grade, which encompasses five stages in the case of NB, namely 1, 2, 3, 4, and 4S. We thus analyzed netrin‐3 gene expression by qRT–PCR in a panel of 181 human NB samples (Gibert et al, 2014). While netrin‐3 mRNA was barely detectable in stages‐1, ‐2, and ‐3, which are localized NB with good prognosis, its expression increased in more advanced, highly aggressive and metastatic tumors with poor outcome (Fig 2A). Indeed, stage‐4 displayed higher levels of netrin‐3 compared to stages‐1 (P = 0.0102), ‐2 (P = 0.1068), or ‐3 (P = 0.0042). Remarkably, stage 4S restricted to neonates and encompassing highly metastatic tumors that often spontaneously regress, displayed a low level of netrin‐3 (P = 0.0434; stage‐4 vs. 4S) (Fig 2A). NB patients were also stratified in two groups according to MYCN amplification, the lower risk group consisting of non‐MYCN‐amplified and localized tumors (stages‐1, ‐2, and ‐3) and metastatic forms for children under 18 months (stages‐4 and 4S) (Matthay et al, 2016). The high‐risk group includes all MYCN‐amplified NB and non‐MYCN‐amplified stage‐4 tumors, for children above 18 months (Ambros et al, 2003). Netrin‐3 gene expression was once again correlated with poor outcome as it was significantly higher in the high‐risk group (P = 0.039) (Fig 2B). Furthermore, elevated netrin‐3 gene expression levels were strongly correlated with poor overall survival (OS) in this cohort (median expression ranking), with an OS at 150 months of 72.5% for low netrin‐3‐expressing tumors and 46.6% for high netrin‐3‐expressing tumors (P = 0.029) (Fig 2C). A group of NB patients (n = 19) that displayed a twofold increase or more in netrin‐3 expression was extrapolated (=netrin‐3 very high) from this original cohort and exhibited an OS of 28.5% (P = 0.002) (Figs 2C and EV1A). Moreover, in the aggressive stage‐4 sustained netrin‐3 gene expression was correlated with poor prognosis, potentially underlining a function for netrin‐3 in NB tumor progression and aggressiveness (Fig 2D). Finally, we confirmed the data extracted from the 181 patients, using a published cohort of 498 cases of NB patients (Zhang et al, 2015), and further validated this gene expression with an RNAscope analysis on fixed frozen NB‐tissue samples (Figs 2E and EV1B–D).

Figure 2

Analysis of netrin‐3 expression on neuroblastoma (NB) prognosis

AQuantification of netrin‐3 gene expression by qRT–PCR in a panel of 181 human NB stages 1, 2, 3, 4, and 4S. Number of cases is indicated on the graph. Error bars indicate s.e.m. Statistical treatment of the data was performed using a two‐sided Student’s t‐test.

BQuantification of netrin‐3 gene expression by qRT–PCR in a panel of 181 human neuroblastoma patients, defined as low‐ and high‐risk NB. The number of cases is indicated on the graph. Error bars indicate s.e.m. Statistical treatment of the data was performed using a two‐sided Student’s t‐test.

CNetrin‐3 high expression is a marker of poor prognosis in NB. 280 months overall Kaplan–Meier survival curves in a panel of 181 patients of all NB stages. The cohort was dichotomized base on netrin‐3 median expression (in green and red). Statistical treatment of the data: Mantel–Cox; P‐value is indicated below the graph.

DNetrin‐3 high expression is a marker of poor prognosis in aggressive NB. 280 months overall Kaplan–Meier survival curves in a panel of 64 stage four patients. The cohort was dichotomized base on netrin‐3 expression, as presented in panel C. Statistical treatment of the data: Mantel–Cox; P‐value is indicated below the graph.

EConfirmation in 498 RNA‐seq NB patients that netrin‐3 high expression is a marker of poor prognosis in NB. 6,000 days overall Kaplan–Meier survival curves. The cohort was dichotomized based on netrin‐3 median expression (in green and red). Statistical treatment of the data: Mantel–Cox; P‐value is indicated below the graph.

Figure EV1

Netrin‐3 role in NB

ANetrin‐3 high expression is a marker of poor prognosis in NB. 280 months overall Kaplan–Meier survival curves in a panel of 181 patients of all NB stages. The cohort was dichotomized base on netrin‐3 median expression as Fig 2C (in green). An additional group (purple) defined as netrin‐3 very high, expressing twice the median expression. High netrin‐3 is the group composed by the patient of the high group of Fig 2C, without the very high group of patients. Statistical treatment of the data: Mantel–Cox; P‐value is indicated below the graph.

BRepresentative netrin‐3 and netrin‐1 detection, using RNAscope on NB high‐grade tumor cryosection. Negative control DAPB, positive control PPiB. Each brown dot represents a unique molecule of mRNA of each target gene. Scale bars 20 µm.

CQuantification of netrin‐3 expression by RNA sequencing in a panel of 498 human NB stages 1, 2, 3, 4, and 4S. The number of cases is indicated on the graph. Statistical treatment of the data was performed using a two‐sided Student’s t‐test with Welsh correction (bars indicate SD).

DNetrin‐3 high expression is a marker of poor prognosis in aggressive NB. 6,000 days overall Kaplan–Meier survival curves in a panel of 183 stage 4 patients (extracted from the 498 RNA‐seq cohort). The cohort was dichotomized base on netrin‐3 expression.

To refine the molecular characterization of netrin‐3‐expressing NB tumors, we performed Gene Set Enrichment Analysis (GSEA) to identify associated pathways (Subramanian et al, 2005). We observed that MYCN‐regulated pathways were also over‐activated in the netrin‐3 high group (P = 0.039; fdqr = 0.025) (Fig 3A). The E2F signature was also over‐represented among netrin‐3 high NB tumors (P ≤ 0.001; fdqr = 0.0014), the G2/M cell hallmark correlated with genomic instability was also predominant in netrin‐3 high stage four patients (Fig EV2A‐B). These two signatures are correlated with high levels of mitosis and cancer aggressiveness.

Figure 3

Netrin‐3 is a target gene of MYCN in NB

AGene Set Enrichment Analysis (GSEA) in high netrin‐3 NB patients. The cohort was dichotomized with the n = 50 lowest vs. 50 highest: amplification the MYCN pathway (P = 0.039; fdqr = 0.025).

BNetrin‐3 expression is higher in MYCN‐amplified NB patients. The cohort was dichotomized based on mycn amplification and sorted for netrin‐3 expression. Statistical treatment of the data: Welsh test; P‐value is indicated on the graph. Error bars indicate s.e.m (boxes = 25^th to 75^th percentile; central band = mean; whiskers = min to max).

CQRT–PCR analysis of netrin‐3 gene expression in IMR32 (n = 5) and IGRN91 (n = 5) NB cell lines after MYCN silencing by siRNA (U‐test). Error bars indicate s.e.m.

DChIP‐seq analysis of MYCN binding, active enhancer epigenetic marks H3K27ac, H3K4me1, and active promoter epigenetic mark H3K4me3, on netrin‐3 locus (red line). An enrichment of MYCN, associated with active enhancer marks, was detected in three different neuroblastoma cell lines NB‐1643, COGN415, and LAN5.

EQuantitative analysis showing the size of IGRN91 primary tumors implanted on CAM and silenced or not for netrin‐3 (n = 11 siScr, n = 9 siNetrin‐3; U‐test). Error bars indicate s.e.m.

FQuantitative analysis showing the size of IMR32 primary tumors silenced or not for netrin‐3 (n = 5 siScr, n = 7 siNetrin‐3; U‐test). Error bars indicate s.e.m.

Figure EV2

Netrin‐3 is a target gene of MYCN in NB

AGene Set Enrichment Analysis (GSEA) in high netrin‐3 NB patients. The total cohort was dichotomized with the n = 50 lowest vs. 50 highest: amplification of E2F pathway (P ≤ 0.001; fdqr = 0.0014).

BGene Set Enrichment Analysis (GSEA) realized in stage 4 patients high vs. low netrin‐3 NB patients (n = 183 characterized in EV‐1C). The hallmarks G2/M checkpoint is amplified, indicating a strong proliferation in the High netrin‐3 group (P = 0.025; fdqr = 0.044).

CAnalysis of netrin‐3 expression in MYCN‐amplified NB patients. The cohort was dichotomized based on MYCN amplification and sorted for netrin‐3 expression. Statistical treatment of the data: Welsh test; P‐value is indicated on the graph (n = 181, bars indicate SD).

DAnalysis of patient survival based on netrin‐3 expression on patient none MYCN‐amplified from the cohort described in Figs 2A and Fig EV2A (n = 157). Statistical treatment of the data: Mantel–Cox; P‐value is indicated below the graph.

EQRT–PCR analysis of netrin‐1 gene expression in IMR32 (n = 4) and IGRN‐91 (not shown; netrin‐1 not detectable) cell line after MYCN silencing by siRNA (U‐test, bars indicate s.e.m.).

FAnalysis of MYCN binding on netrin‐3 promoter in BE2(C) NB cell line after Chip‐seq experiment. MYCN peaks upstream of netrin‐3 are associated with a broad H3K27ac‐enriched region.

GChIP‐seq analysis of MYCN binding, active enhancer epigenetic marks H3K27ac, H3K4me1, and active promoter epigenetic mark H3K4me3, on netrin‐1 locus (pink line). An enrichment of MYCN, associated with active enhancer marks, was detected in three different neuroblastoma cell lines NB‐1643, COGN415, and LAN5.

HSchematic representation of the experimental chick chorioallantoic membrane (CAM) model. IMR32 or IGR‐N91 cells were transiently transfected with scramble, netrin‐1 and netrin‐3 siRNAs and xenografted on CAM on day 10.

INB cell lines were subjected to immunoblots.

JRepresentative picture of cleaved PARP staining. Quantification of IGR‐N91 cells positive for cleaved PARP, in tumors transfected or not by siNetrin‐3 or siControl (n = 4, U‐test, bars indicate SD, scale bars 100 µm).

As MYCN amplification is believed to be one of the most drastic events impacting patient survival (Rickman et al, 2018), we decided to further analyze the links between MYCN and netrin‐3 in NB among the 498 patients. As such, those with an amplified MYCN presented a higher level of netrin‐3 compared to non‐amplified MYCN‐bearing patients (P ≤ 0.001), suggesting a putative direct regulation by this transcription factor (Figs 3B and EV2C). Interestingly, netrin‐3 did not constitute a prognostic marker in non‐MYCN‐amplified patients (Fig EV2D). Netrin‐3 gene expression decreases when mycn was silenced using specific siRNA in two NB cell lines, whereas netrin‐1 remained unaltered (Figs 3C and EV2E). MYCN regulation of netrin‐3 gene expression was further supported, as we detected an enrichment in MYCN binding sites in the netrin‐3 promoter locus after ChIP‐sequencing analysis of NB cell lines (Robinson et al, 2011; Zeid et al, 2018; Upton et al, 2020). This was correlated with Histone H3 acetyl‐Lysine 27 (H3K27ac) and monomethylated H3K4 (H3K4me1), histone modifications that are associated with active enhancer regions (Rada‐Iglesias et al, 2011), indicating that MYCN directly contributes to increasing the expression of netrin‐3 in NB (Figs 3D and EV2F). A broad domain of the active promoter‐associated epigenetic modification, trimethylated H3K4 (H3K4me3), was also enriched at netrin‐3 transcription start site (TSS) and across the gene body (Figs 3D and EV2F), consistent with this being an actively transcribed gene potentially involved in cell identity (Benayoun et al, 2015) (Fig EV2F). This binding was undetectable for Netrin‐1 explaining, at least in part, the specificity of these expressions (Fig EV2G).

To ascertain whether netrin‐3 impacts neuroblastoma tumor formation, we conducted netrin‐3 silencing via specific siRNA assays in the IGR‐N91 cell line. The chorioallantoic membrane (CAM) of chicken embryos is a well‐described model to study tumor progression (Stupack et al, 2006; Berthenet et al, 2020). NB cells were xenografted on the CAM of ten‐day‐old chick embryos (Fig EV2H). Seventeen‐day‐old chick embryos were then analyzed for primary tumor size. As shown in Fig 3E, tumors silenced for netrin‐3 were substantially smaller than controls (U‐test; siScr vs. siNetrin‐3, P = 0.0011). Staining of the engrafted tumors resulted in an increase in the number of apoptotic cells detected in netrin‐3‐silenced tumors (Cleaved PARP, U‐test, P = 0.0162). It is interesting to note that proliferation appears as not modified in the same experiments (Fig EV2I). Similar results were obtained with IMR32 NB cell line (U‐test, P = 0.0025) (Fig 3F).

Elevated netrin‐3 expression in small cell lung cancer

We next investigated the putative role of netrin‐3 in SCLC, which is the deadliest histological subtype of lung cancers, associated with high rates of metastatic disease at diagnosis (Augert et al, 2020; Ko et al, 2021). Some drugs including immunotherapy have recently been tested in the clinic for this indication but have failed to provide benefits for a majority of patients. Netrin‐3 expression was detectable by RNA sequencing in lung cancer cell lines, particularly in SCLC (P ≤ 0.001) and more sporadically in carcinoid and mesothelioma cells which are also neuroendocrine lineages (Figs 4A and EV3A). We performed qRT–PCR analyses to select NCI‐H82 and NCI‐H69 cells model, as they expressed netrin‐3 but not netrin‐1 (Fig 4B).

Figure 4

Characterization of Netrin‐3 in SCLC

ADeep analysis of netrin‐3 gene expression in lung cancer cell lines (n = 76; RNA‐Seq).

BQRT–PCR analysis of netrin‐1 and netrin‐3 expression, in lung cancer cell Lines (Error bars indicate SD, n = 3).

CRepresentative netrin‐3 and netrin‐1 mRNA detection, using RNAscope on SCLC paraffin‐embedded tumor sections. Negative control DAPB, positive control PPiB. Each brown dot is a unique molecule of mRNA of each targeted gene (n = 10, scale bars 20 µm).

DQuantification of netrin‐3 and netrin‐1 expression in human SCLC paraffin‐embedded tumor sections (High ≥ 50% marked cells; Low ≤ 5%).

Figure EV3

Netrin‐3 function in SCLC

AQuantification of netrin‐3 expression by RNA sequencing in lung cancer cell lines. Statistical treatment of the data was performed using a two‐sided Student’s t‐test with Welsh correction (adenocarcinomas n = 36; squamous n = 4, NSCLC n = 47, SCLC n = 30, large cells n = 7, carcinoids n = 2, mesothelium n = 12; bars indicate SD).

BNetrin‐3 is not expressed in the tumor microenvironment. Representative netrin‐3 mRNA detection, using RNAscope technology on SCLC paraffin‐embedded tumor sections. T = tumoral part; NT = not tumoral. Each brown dot is a unique molecule of mRNA of netrin‐3 in the tumoral part. Scale bars 20 µm.

CRepresentative netrin‐3 mRNA detection, using RNAscope technology on SCLC paraffin‐embedded tumor sections. Each brown dot is a unique molecule of mRNA of netrin‐3 in the tumoral part. Scale bars 20 µm.

To further analyze the expression of netrin‐3 in human SCLC, we performed an RNAscope analysis on FFPE sections (Fig 4C). Netrin‐3 was expressed at a high level in nine out of 10 human SCLC samples tested, whereas netrin‐1 was detected in only 10% of the samples, confirming our previous observation on their exclusive expression pattern (Fig 4D). Interestingly, RNAscope analysis further supported the view that netrin‐3 was specifically expressed by cancer cells and not by the tumor microenvironment (Fig EV3B and C).

In an effort to unravel the mechanisms underlying the expression of netrin‐3 in SCLC, we additionally carried out an in silico screen of transcription factors described as promoters of genes associated with SCLC aggressiveness. Indeed, netrin‐3 expression is associated with the expression of two key transcription factors: the neuronal transcription factor neurogenic differentiation 1 (NeuroD1) and Achaete‐Scute Family BHLH Transcription Factor‐1 (ASCL‐1) (Borromeo et al, 2016; Rudin et al, 2019). These factors are upregulated in a variety of aggressive neural/neuroendocrine carcinomas and important for the development and function of several neural/neuroendocrine tissues. After analysis of published ChIP‐sequencing datasets, we identified that specific binding sites for both NeuroD1 and ASCL‐1 are enriched upstream of the netrin‐3 gene promoter in SCLC cell lines (Robinson et al, 2011; Borromeo et al, 2016; Huang et al, 2018; Upton et al, 2020) (Figs 5A and EV4A and B). These data show that the NeuroD1 and ASCL‐1 peaks are centered around a broad domain of H3K27ac that encompasses the entire netrin‐3 upstream intergenic region, as well as most of the gene body. This suggests that NeuroD1 and ASCL‐1 binding at this site is associated with transcriptional activation of netrin‐3 (Fig 5B). By comparison, we did not observe a similar enrichment in H3K27ac in a wide range of other cell types, indicating selectivity to SCLC cell lines expressing NeuroD1/ASCL‐1 (Fig EV4A and B). We detected no binding in the netrin‐1 gene promoter (Fig EV4C). Furthermore, netrin‐3 gene expression decreases when neuroD1 and ascl‐1 were silenced using specific siRNA in two SCLC cell lines, indicating that these transcription factors directly regulate the netrin‐3 transcript (Fig 5C).

Figure 5

Netrin‐3 is a target gene of NeuroD1 and ASCL1 in SCLC

AChIP‐seq analysis of NeuroD1 and ASCL‐1 binding on the netrin‐3 gene promoter. A major binding site was detected (red and blue peaks for NeuroD1 and ASCL1, respectively). Input controls are shown below (black).

BChIP‐seq analysis of NeuroD1 and ASCL‐1 chromatin enrichment, which was associated with broad H3K27ac‐enriched regions at the netrin‐3 locus (red line), indicating a transcriptionally active region in SCLC cell lines (rep = experimental replicates).

CQRT–PCR analysis of netrin‐3 gene expression in NCI‐H82 (n = 6) and NCI‐H69 (n = 4) SCLC cell lines after Neurod‐1 and ASCL‐1 silencing by siRNAs (U‐test). Error bars indicate s.e.m.

Figure EV4

Netrin‐3 is a target gene of NeuroD1 and ASCL1 in SCLC

AAnalysis of NeuroD1 ChIP‐sequencing in two cell lines (light blue), and ASCL1 ChIP‐sequencing data in three cell lines (light red) along with input controls for SCLC cell lines expressing either NeuroD1 or ASCL1 (black) on netrin‐3 gene promoter (red line). H3K27ac‐enriched region is specific of SCLC cell lines.

BAnalysis of NeuroD1 or ASCL1 and H3K7ac recruitment to the netrin‐3 promoter region in two SCLC cell lines.

CAnalysis of NeuroD1 ChIP‐sequencing in two cell lines (light blue), and ASCL1 ChIP‐sequencing data in one cell line (light red) along with input controls for SCLC cell lines expressing either NeuroD1 or ASCL1 (black). On netrin‐1 gene promoter (pink line).

Novel tumor growth‐promoting activity by netrin‐3

Since netrin‐1 has been shown to promote tumor progression in various cancer indications due to its ability to block DCC and UNC5B‐induced cell death, and since our current data show that in NB and SCLC tumors, netrin‐3 is upregulated rather than netrin‐1, we first investigated whether netrin‐3 could bind to netrin‐1 receptors. Data previously reported suggested that netrin‐3 could interact with netrin‐1 receptors (Wang et al, 1999). To extend this analysis, we conducted bio‐layer interferometry assays (Fig 6A). Attempts to produce recombinant human netrin‐3 were unsuccessful due to a non‐soluble production of h‐netrin‐3 and prompted us to use chicken netrin‐2L, which is phylogenetically considered as the ortholog of mammalian netrin‐3. As shown in Fig 6A and B, netrin‐2‐like (NTN2L) was able to bind to UNC5‐B and UNC5‐C with a Kd similar to netrin‐1, whereas it interacted with lower affinity with DCC, UNC5A, and Neogenin.

Figure 6

Tumor promoting role of netrin‐3

ABio‐layer interferometry analysis of netrin‐2L/receptor interactions. Other members of the netrin family were also tested: netrin‐1, netrin‐4 and netrin‐G2 (red: netrin‐2L; blue: netrin‐1; green netrin‐4).

BKd calculation after bio‐layer interferometry analysis, *= no interaction detected.

CNCI‐H82 CRISPR/Cas9 were subjected to immunoblots. sGx, designates the number of guide RNAs, control cells (sGc).

DNetrin‐3 silencing delays or inhibits SCLC engraftment in vivo. NMRI nude mice were subcutaneously engrafted with either NCI‐H82‐CRIPSR/cas9‐NTN3 with three different guides RNAs (sG1 P = 0.0022; sG2 P ≤ 0.001; and sG3 P ≤ 0.001) or control cells (sGc) n = 6/group. Positive tumor engraftment was reported when tumors reached 50 mm³ (Mantel–Cox).

ENCI‐H69‐CRISPR/Cas9 were subjected to immunoblots. sGx designates the number of guide RNAs, control cells (sGc).

FNCI‐H69‐CRISPR/Cas9 cells presented in E. were subcutaneously engrafted with either NCI‐H69‐ CRIPSR/cas9‐NTN3 with three different guide RNAs (sG1 P = 0.017; sG2 P = 0.028; and sG3 P = 0.001) or control cells (sGc) n = 6/group. Positive tumor engraftment was reported when tumors reached 20 mm³ (Mantel–Cox).

Source data are available online for this figure.

In an effort to analyze the function of netrin‐3 in cancer cells, we first silenced netrin‐3 and CRISPR/Cas‐9 in NCI‐H82 SCLC cells, using three different guide RNAs and isolated three polyclonal cell lines devoid of netrin‐3 expression (Fig 6C). In this cell line and in the NCI‐H82 KO polyclones, DCC was not detected. Interestingly, an increase in the level of UNC5B receptor was detected. This may imply that UNC5B is internalized when bound to netrin‐3 as formerly described for DCC when bound to netrin‐1 (Bin et al, 2015). In vitro, these netrin‐3 KO cells did not behave differently than control cells in terms of proliferation or cell death similarly to results obtained with siRNAs (Fig EV5A). To determine whether netrin‐3 had a function in tumor formation, we engrafted these cell lines in immunodeficient mice. As shown in Fig 6D, cell populations silenced for netrin‐3 expression showed a strong decrease in tumor onset, supporting a role for netrin‐3 in tumor development. Similar data were observed with the NCI‐H69 cell line (Fig 6E and F).

Figure EV5

Netrin‐3 as a therapeutic target

AProliferation and cell death assay analyses H82 and H69 transfected with siNetrin‐3 and in CRISPR/Cas9 clones lacking netrin‐3 in NCI‐82 and NCI‐H69. Proliferation was determined by the MTT method; apoptosis by cleaved caspase‐3 monitoring. (Error bars indicate SD, n = 4, U‐test).

BTable presenting the percentage of conservation of the NP137 epitope across netrin family members.

CAnalysis of NP137 binding in a dose‐dependent manner on netrin‐1 and netrin‐2L by bio‐layer interferometry assays.

DNMRI nude mice were engrafted with NCI‐H82 cells by subcutaneous injection of 2 million cells. When the mean tumor volume reached approximately 80 mm³, animals were treated 3 × time/weekly by intra‐abdominal injection of PBS; NP001 (Human IgG1, isotype); NP137 (anti‐netrin1/3) for 20 days. Immunoblots of netrin‐1 and netrin‐3 were conducted at end point with the corresponding antibodies.

ENMRI nude mice were engrafted with NCI‐H2286 cells by subcutaneous injection of 5 million cells. When the mean tumor volume reached approximately 80–100 mm³, animals were treated 3 × time/weekly by intra‐abdominal injection of PBS; NP001 (Human IgG1, isotype); NP137 (anti‐netrin1/3) for 20 days. n = 10 animals/group (*P < 0.05, two‐way ANOVA, Error bars indicate s.e.m.).

FRepresentative images of cleaved PARP staining in H82 treated with NP001 (Human IgG1, isotype) or NP137 (anti‐netrin1/3). Scale bars 100 µm.

Netrin‐3 as a therapeutic target in SCLC

Having demonstrated that netrin‐3 was upregulated in two subtypes of neuroectodermal tumors, where it potentially promotes tumor development, we next sought to study its relevance as a therapeutic target. Of interest, a therapeutic anti‐netrin‐1 mAb, called NP137, was recently developed and is currently being tested in clinical phase 1 and 2 trials in patients harboring tumoral netrin‐1 expression (Cassier et al, 2019). Netrin‐1 and netrin‐3 are 75% homologous in terms of protein conservation, and this homology reaches 90.1% for the epitope region recognized by the NP137 (Figs 7A and EV5B). To investigate whether this similarity is sufficient to allow NP137 to bind to netrin‐3, we tested the binding of NP137 on recombinant chicken netrin‐2 (rchNTNT2) by bio‐layer interferometry. As shown in Fig 7B, NP137 clearly bound netrin‐2L within a similar range of affinity compared to netrin‐1 (Kd rhNTN1: 0.96 ± 0.30 nM; Kd for rchNTN2: 1.53 ± 0.96 nM) (Figs 7B and EV5C). Moreover, using the netrin‐1/UNC5B interaction as a positive control, we further showed that NP137 also inhibits the netrin‐2L/UNC5B interaction in a dose‐dependent manner (Fig 7C). Next, we assessed whether as observed for netrin‐1‐expressing cancer cells (Grandin et al, 2016), NP137 could affect netrin‐3‐expressing SCLC tumor cells. NCI‐H82 cells were thus grafted in immunodeficient mice, which were then received systemic injections of NP137 once tumors had reached 100 mm³. As shown in Fig 7D, NP137 triggers a significant tumor growth inhibition and an enhanced mouse survival (Figs 7E and EV5D).

Figure 7

Netrin‐3 is a therapeutic target in SCLC

AAnalysis of the NP137 antibody (anti‐netrin‐1) epitope across netrin family members.

BAnalysis of NP137 binding on netrin‐2L, netrin‐1; netrin‐4 or netrin‐G2 by bio‐layer interferometry assays. Kd rhNTN1: 0.96 ± 0.30 nM; Kd for rchNTN2: 1.53 ± 0.96 nM.

CNP137 inhibits NTN1/UNC5B or NTN‐2L/UNC5B interaction, detected by Elisa assays. Here the recombinant proteins are the murine forms (Control IgG = mouse IgG2A, mNP137 = murine Fc, UNCB = rat UN5H2) bars indicate SD, n = 3.

DNMRI nude mice were engrafted with NCI‐H82 cells by subcutaneous injection of 2 million cells. When the mean tumor volume reached approximately 80 mm³, animals were treated 3 × time/weekly by intra‐abdominal injection of PBS; NP001 (Human IgG1, isotype); NP137 (anti‐netrin1/3) during 20 days. n = 10 animals/group (*P < 0.05, two‐way ANOVA, error bars indicate s.e.m.).

ENP137 enhances survival of mice engrafted with NCI‐H82 cell line (see D). Analysis of Kaplan–Meier survival curves of mice treated or not with NP137. Mantel–Cox test; n = 10 animals/group.

FQuantification of cell death (cleaved PARP) in H82‐treated tumors in D. Tumors were analyzed 7 days after the first treatment (i.e., 400 mm³; n = 3/group. U‐test, bars indicate SD).

Similar data were observed with the NCI‐H2286 cell line, highlighting netrin‐3 as a potential therapeutic target (Fig EV5E). To decipher, the mechanism of action of the antibody targeting netrin‐3 we conducted an immunohistological analysis of anti‐netrin‐treated NCI‐H82 tumors, which revealed a significant increase of cleaved PARP staining (Figs 7F and EV5F). Interestingly, no significant modification of Ki67 is detected confirming chorioallantoic membrane experiments, further supporting the view that netrin‐3 interference is in vivo associated with cancer cell death.

Human tumor samples

Patients or their legal guardians signed a written informed consent agreeing on the use of tumor samples for research, according to the French regulations on the protection of persons (French Ethics Committee). Human NB samples were collected from a cohort of 181 patients staged in accordance with the International Neuroblastoma Staging System (Gibert et al, 2014). Material was collected by the Institut Gustave Roussy (Villejuif, France). Human SCLC tumors were collected by the Biobank Tissue unit, Pasteur Hospital (Nice, France). Informed consent was obtained from all subjects, and the experiments conformed to the principles set out in the WMA Declaration of Helsinki and the Department of Health and Human Services Belmont Report. Following patient agreement, primary tumors were obtained, either by biopsy or after surgery, and were directly frozen before storage.

Mouse experiments and pathology

Animals were maintained in a specific pathogen‐free animal facility (Anican, Lyon—France) and stored in sterilized filter‐topped cages. Mice were handled in agreement with the institutional recommendations and procedures approved by the animal care committee (Comité d'Evaluation Commun au Centre Léon Bérard, à l'Animalerie de transit de l'ENS, au PBES et au laboratoire P4; CECCAP). Seven‐week‐old (20–22 g body weight) female NMRI mice were obtained from Janvier Labs (Saint Berthevin—France) (Delloye‐Bourgeois et al, 2013). The animals were housed in standard cages (five animals/cage) with litter and food. Enrichment was achieved by adding cotton for nesting.

Generation of stable netrin‐3 knocked out cell line

Different small guide RNAs (sgRNA) were designed using the sgRNA designer tool (CRISPRi), (https://portals.broadinstitute.org/gpp/public/analysis‐tools/sgrna‐design‐crisprai). The sequences are as follows: sg1: 5'‐CCGCCCTCGCTGCAGCCGGG‐3'; sg2: 5'‐CGCCCACGGCCCTTCCCGGG‐3'; sg3: 5'‐TCGCTGCAGCCGGGAGGAGG‐3'.

The Esp3I linker was added to sgRNAs, and the oligonucleotides were then annealed by incubation for 3 min at 90°C and 15 min at 37°C and ligated in pLV hU6‐sgRNA hUbC‐dCas9‐KRAB‐T2a‐Puro (Addgene #71236) (Thakore et al, 2015). Insertion of sgRNA was confirmed by sequencing. The use of pLV hU6‐sgRNA hUbC‐dCas9‐KRAB‐T2a‐Puro, which is constructed around a 3^rd generation lentiviral backbone, allows for the simultaneous expression of Cas9 and gRNA and for selection for puromycin resistance. Resulting plasmids were used to produce lentiviruses to further infect cell lines of interest. Lentiviral vectors were a generous gift from Carine Maisse (INRA, UMR754, Lyon, France).

Lentiviral transduction

Production of VSV‐g pseudotyped lentivirus was made by co‐transfecting 70–80% confluent 293FT cells (gift from Dr Fabrice Lavial, CRCL) with lentiviral vectors described in 1) and 2) plus pCMV‐dR8.91 and pCMV‐VSV‐g plasmids using Lipofectamine 3000 in OptiMEM medium + GlutaMAX supplemented with 1 mM Sodium Pyruvate and 5% FBS. Transfection medium was replaced 6 h later by OptiMEM + Pyruvate + 5% FBS. Lentiviral supernatant was harvested 48 h later, filtered through a 0.45‐μm syringe filter, and used either fresh or snap‐frozen. For transduction, cell lines of interest were plated in 0.25 ml in 24‐well plates and 0.25 ml of viral supernatant was added in the presence of 10 µg/ml polybrene. 24 h later, viral supernatant was discarded and cells were cultured in their normal culture medium supplemented with puromycin for one week.

Cell culture

NCI‐H82 and NCI‐H69 were obtained from ATCC and cultured in RPMI 1640 Medium (ATCC 30‐2001) adding fetal bovine serum (FBS, ATCC 30‐2020) to a final concentration of 10% and 1% penicillin/streptomycin (P/S). NCI‐H2286 was obtained from ATCC and cultured in DMEM: F12 Medium (Catalog No.30‐2006) adding 0.005 mg/ml Insulin, 0.01 mg/ml Transferrin, 30 nM Sodium selenite (final concentration), 5% FBS and 1% P/S. HeLa cells were obtained from our lab and cultured in DMEM with 10% FBS and 1% P/S. All cells were tested for mycoplasma contamination.

Quantitative RT–PCR

Tumor cells, patient samples, or xenografts were disrupted with MagNALyser kit (Roche Applied Science). Total RNAs of either cell lines or tumors were extracted by using NucleoSpin^® RNAII Kit (Macherey Nagel) according to the manufacturer’s protocol. RT–PCR reactions were performed with iScript cDNA Synthesis Kit (BIO‐RAD). One microgram total RNA was reverse transcribed by using the following program: 25°C for 5 min, 42°C for 30 min, and 85°C for 5 min. Real‐time quantitative RT–PCR was performed with a LightCycler 96 apparatus (Roche) using LightCycler^® TaqMan^® Master kit (Roche, Basel, Switzerland). Expression of target genes was normalized to hypoxanthine‐guanine phosphoribosyltransferase (HPRT), glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH), and beta‐glucuronidase (GUSB) using the comparative CT method.

Relative expression of each gene was calculated according to the comparative 2−ΔΔCt quantification method in which ΔCt = Ct (Sample)−Ct (Normalizer) and ΔΔCt = ΔCt (Sample)−ΔCt (Calibrator).

hASCL1 F (probe 38) CGACTTCACCAACTGGTTCTG; hASCL1 R (probe 38) ATGCAGGTTGTGCGATCA; hMYCN F (probe 38) TAATATGCCCGGGGGACT; hMYCN R (probe 38) GGGCTGGAACTGGCTTTT; hNeuroD1 F (probe 30) CTGCTCAAGGACCTACTAACAACAA; hNeuroD1 R (probe 30) GTCCAGCTTGGAGGACCTT; hNTN3 F (probe 56) GCCCTGTGTTAAGACCCCTA, hNTN3 RR (probe 56) TGCAGTGCGAGTCACAGTC; hNTN1 F (probe 3) AAAAGTACTGCAAGAAGGACTATGC; hNTN1 R (probe 3) CCCTGCTTATACACGGAGATG; hHPRT F (probe 73) TGACCTTGATTTATTTTGCATACC; hHPRT R (probe 73) CGAGCAAGACGTTCAGTCCT; hGUS F (probe 57) CGCCCTGCCTATCTGTATTC; hGUS R (probe 57) TCCCCACAGGGAGTGTGTAG.

Bio‐layer interferometry

BLI experiments were performed using The OctetRED96 system. All interactions were analyzed at 30°C with constant shaking at 1,000 rpm in PBS, 0.02% Tween‐20, 0.1% BSA (binding buffer: BB). To determine the affinity between hNetrin‐1 or chNetrin‐2 with the extra‐cellular domain (ECD) of hUNC5H1, rUNC5H2, hUNC5H3, and hUNC5H4, netrin‐1 or netrin‐2‐coated‐HIS1K biosensors were incubated with an increasing concentration series of ECDs (3.12 nM, 6.25 nM, 12.5 nM, 25 nM, 50 nM, 100 nM) and binding was observed for 5 min at 30°C. Biosensors were then incubated in BB for a further 5 min to observe dissociation of the complex. To determine the affinity between h‐netrin‐1 or ch‐netrin‐2 with the extra‐cellular domain (ECD) of mNeogenin and mDCC, ECDs‐coated‐AHC biosensors were incubated with a concentration series of hNetrin‐1 or chNetrin‐2 (3.12 nM, 6.25 nM, 12.5 nM, 25 nM, 50 nM, 100 nM). In this assay, 1 µg/ml Dextran 5,000 was added to BB to prevent non‐specific binding of the netrins to the biosensors. Association and dissociation were monitored as described above. To determine the affinity of the human anti‐netrin NP137 antibody for h‐netrin‐1 and ch‐netrin2, Netrin‐1 or Netrin‐2‐coated‐HIS1K biosensors were incubated with an increasing concentration series of NP137 (1 nM, 2 nM, 5 nM, 10 nM, 20 nM, 50 nM, 100 nM) and association was observed for 5 min at 30°C. Biosensors were then incubated in BB for a further 5 min to observe dissociation of the complex. To determine the ability of the anti‐netrin antibody to antagonize the binding of h‐netrin‐1 and ch‐netrin‐2 with UNC5H2, the ECD of UNC5H2 was captured on AHC biosensors, incubated with 25 nM of netrin‐1 or netrin‐2L in presence of various concentrations of either the murine anti‐netrin antibody or an isotype control (64 pM, 320 pM, 1.6 nM, 8 nM, 40 nM). The association was observed for 5 min at 30°C to evaluate kon. Binding kinetics were evaluated with ForteBio Octet RED Evaluation software 6.1 using a 1:1 binding model to derive kon, koff, and KD values.

Western blot

Cells were harvested and lysed in SDS buffer (5% SDS, 10% Glycerol, 10 mM Tris–HCl pH 7.6, 1% Triton X‐100, 0.1 M DTT) and then sonicated using a Branson Digital sonifier 450 for 20 pulses (50% amplitude, pulse on 0.2 s, pulse off 1.0 s). Protein concentration was measured with the 660 nm Protein assay kit (Pierce Biotechnology, Rockford, IL, USA) using bovine serum albumin (BSA) as a standard curve according to manufacturer’s instructions. Protein extracts (20–50 mg per lane) were loaded onto 4%–15% SDS‐polyacrylamide gels (Bio‐Rad) and transferred onto nitrocellulose membranes using Trans‐Blot Turbo Transfer System (Bio‐Rad). Membranes were blocked with 5% skimmed milk in PBS/0.1% Tween‐20 (PBS‐T) for 1 h and then incubated overnight with primary antibody: 1:1,000 dilution for anti‐netrin‐1 (Abcam, anti‐NTN1 ab126729), 1:500 dilution for anti‐netrin‐3 Abcam (anti‐NTN3 ab185200),

After three washes with PBS‐T, membranes were incubated with the appropriate HRP‐conjugated secondary (1:5000 dilution) antibody for 1 h at room temperature. Detection was performed using West Dura Chemiluminescence System (Pierce). Membranes were imaged on the ChemiDoc Touch Imaging System (Bio‐Rad).

Human histopathology

For histological examination, tissue samples were fixed in 10% buffered formalin and embedded in paraffin (FFPE) (see Supplementary Materials section). 4‐μm‐thick FFPE tissue sections were prepared according to conventional procedures. Sections were then stained with hematoxylin/eosin and examined under a light microscope. mRNA ISH was performed using the RNAscope 2.5 VS Reagent kit—BROWN with a custom designed netrin‐3 probe (Advanced Cell Diagnostic, Hayward, CA) according to the manufacturer’s guidelines. The RNAscope procedure was performed using the Discovery XT autostainer with mRNA amplification, pretreatment & DAB PTO kit (Roche, Meylan, France). Tissue control was assessed by performing RNAscope analysis of common housekeeping genes PPIB and DAPB, as a negative control. Finally, sections were scanned with panoramic scan II (3DHistech, Budapest, Hungary) at 40× and Z‐stacked (3 sections at 0.6 µm intervals).

10‐µm‐thick cryosections were obtained from tissues and stored at −80°C. Frozen tissue slides were immersed in 4% formalin for 1 h and then rinsed in PBS. Tissues were dehydrated in increasing percentages of alcohol and air‐dried.

The mRNA ISH was performed using the RNAscope 2.5 VS reagent kit brown with custom designed netrin‐3 probes (Advanced Cell Diagnostic, Hayward, CA) according to the manufacturer’s instructions. The RNAscope procedure was performed in the Discovery XT autostainer with mRNA amplification, pretreatment and DAB PTO kit (Roche, Meylan, France), without steps of deparaffinization and HIER. Tissue control was assessed by performing RNAscope analysis of a common housekeeping gene PPiB. Finally, sections were scanned with panoramic scan II (3DHistech, Budapest, Hungary) at 40× and Zstack (3 levels spaced of 0.6 µm).

ChIP‐seq analysis

NeuroD1 ChIP‐seq peaks in NCI‐H524 and NCI‐H82 cells, ASCL1 ChIP‐seq peaks in H2107, NCI‐H889, and NCI‐H128 cells, and the corresponding input controls for NCI‐H524 and NCI‐H889 cells, were downloaded from NCBI GEO dataset GSE69398 (PMID: 27452466) and GSM1700637; GSM894066; GSM894072; GSM894100; GSM89409; GSM1700641; GSM1526706; GSM1526702 and analyzed using IGV software. NeuroD1 and ASCL1 ChIP‐seq data were downloaded from NCBI GEO dataset GSE69398 (PMID: 27452466). All H3K27ac datasets were downloaded from GSE115123 (PMID: 29945888). Data were analyzed using IGV software.

All neuroblastoma ChIP‐seq tracks were downloaded from NCBI GEO dataset GSE138315 (bioRxiv 829754; https://doi.org/10.1101/829754), GSM2113521, GSM2127461, GSM1680108. Data analyzed using IGV software.

Cell death and proliferation assays

Apoptosis was monitored by measuring caspase‐3 activity as described previously [18] using Caspase 3/CPP32 Fluorometric Assay Kit (Gentaur Biovision, Brussel, Belgium) 24 h after transfection with siRNAs.

For proliferation assays, 10⁵ cells were seeded onto a 96‐well plate. After 24 h, 20 µl of CellTiter96^®AQueous One Solution Reagent was added to each well containing 100 µl of culture medium. The plate was then incubated at 37°C for 2 h in a humidified, 5% CO₂ atmosphere. Absorbance at 490 nm was recorded using a TECAN‐infinite m1000.

CAM, mouse experiments, and pathology

The chorioallantoic membrane (CAM) of chicken embryos is a well‐described model to study tumor progression (Stupack et al, 2006; Berthenet et al, 2020). The fertilized eggs were obtained from a producer (EARL Les Bruyères, Dangers, France). They were kept in specific incubators. IGR‐N91 and IMR32 cells (15 and 10 million) were xenografted on the CAM of ten‐day‐old chick embryos with Matrigel (100 μl/tumor) at day 10. Seventeen‐day‐old chick embryos were then analyzed for primary tumor size as previously described (Gibert et al, 2014).

For the engraftment experiments, mice were implanted with either NCI‐H82 or NCI‐H69 SCLC cells; control or CRISPR‐netrin‐3 #1, #2, #3, by subcutaneous injection of 10⁶ cells in 100 µl of PBS (and 100 µl of Matrigel) into the right flank of mice. Tumor volume was calculated with the formula V = (length*width2)/2. A positive tumor catch is declared when the tumor reaches 50 mm³ for NCI‐H82 and 20 mm³ for NCI‐H69.

Mice were implanted with either NCI‐H82 or NCI‐H69 SCLC cells and NCI‐H2286 mix cell line, by subcutaneous injection of 10⁶ cells in 100 µl of PBS into the right flank. Mice were treated with either PBS, NP001 (isotype control human IgG1) or NP137 (anti‐Netrin‐1/3, human IgG1) by i.v. injection at 10 mg/kg twice a week during 3 weeks, once tumors were established at a volume close to 100 mm³ after external quantification using calipers. Tumor volume was calculated with the formula V = (length*width²)/2. At the end of the treatment, tumors were harvested and embedded in paraffin, then sectioned into 10‐µm slices. Tumor histology was studied after hematoxylin‐Phloxine B‐saffron, Ki67, cleaved caspase‐3, and cleaved PARP staining’s of tumor slides at the Pathology‐Research Platform (Lyon‐France).

Statistical analyses

Quantitative reverse‐transcriptase PCR (qRT–PCR) and RNA‐sequencing data analyses were performed using the GraphPad Prism (San Diego, California) and R software (free software foundation, University of Auckland, New Zealand) to generate heat maps. Unpaired Student’s t‐test with Welch correction was used to compare the different groups. Cohorts were separated based on median netrin‐3 expression. Survival curves were produced according to the Kaplan–Meier method on the GraphPad software. Data were analyzed with a Mantel–Cox test. n defines the total replicates. All statistical tests were two‐sided.