Kinome multigenic panel identified novel druggable EPHB4‐V871I somatic variant in high‐risk neuroblastoma

Abstract Neuroblastoma (NB) is the most common extracranial neoplasm in children. The overall outcome for high‐risk NB patients is still unacceptable, therefore, it is critical to deeply understand molecular mechanisms associated with NB, which in turn can be utilized for developing drugs towards the treatment of NB. Protein kinases (TKs) play an essential role in the regulation of cell survival and proliferation. Different kinases, such as anaplastic lymphoma kinase (ALK), Aurora kinase, RET receptor tyrosine kinase, are potential therapeutic targets in various cancers, including NB. We analysed a cohort of 45 high‐risk NB patients and 9 NB cell lines by a targeted—(t)NGS custom gene panel (genes codifying for the kinase domains of 90 TKs). We identified somatic variants in four TK genes (ALK, EPHB4, LMTK3 and EPHB6) in NB patients and we functionally characterized an interesting somatic variant, V871I, in EPHB4 gene. EPHB4 plays a crucial role in cardiovascular development and regulates vascularization in cancer‐promoting angiogenesis, tumour growth and metastasis. Several EPHB4 mutations have previously been identified in solid and haematological tumour specimens but EPHB4 mutations were not described until now in NB. Interestingly, a re‐analysis of public CGH‐array showed that the EPHB4 gain is associated with advanced diseases in NB. We further demonstrated that higher EPHB4 expression is correlated to stage 4 of NB and with poor overall survival. Additionally, we also revealed that the EPHB4‐V871I accounts for increased proliferation, migration and invasion properties in two NB cell lines by acting on VEGF, c‐RAF and CDK4 target genes and by increasing the phosphorylation of ERK1‐2 pathway. The use of two EPHB4 inhibitors, JI‐101 and NVP‐BHG712, was able to rescue the phenotype driven by the variant. Our study suggested that EPHB4 is a promising therapeutic target in high‐risk NB.


| INTRODUC TI ON
Neuroblastoma (NB) is the most common extracranial neoplasm in children and contributes to about 15% of all paediatric cancer-related deaths. 1 Despite major advances in therapies over the past decade, the overall outcome for high-risk NB patients is still unacceptable. [2][3][4][5] Current therapies include chemotherapy drugs that are highly toxic for healthy cells and have significant long-term side effects. 6 Therefore, developing novel targeted therapies for high-risk NB is critical to achieve higher efficacy and to alleviate adverse effects. Future improvements in high-risk NB outcomes will require the identification of disease and patient-specific oncogenic variations that can be druggable. 7 Among high-risk patients, gene signatures can identify children with high-risk disease who would benefit from new and more aggressive therapeutic approaches. 2 MYCN amplification is a strong characteristic of highrisk NB patients and is a genetic marker of disease. 3 However, finding therapeutic strategies to directly target MYCN is a difficult task due to its protein structure. High-throughput sequencing-based studies have highlighted that recurrent mutations of single genes are infrequent in primary NB with activating mutations in ALK, inactivating mutations in ATRX and TERT rearrangements being the most frequent. 5,[8][9][10] RAS/P53 and FA and RAC pathways are among the most significantly mutated pathways in NB. [11][12][13][14][15] Kinases play a crucial role in the regulation of cell survival and proliferation. 16 Different kinases, such as anaplastic lymphoma kinase (ALK), 13,17 Aurora kinase, 14 RET receptor tyrosine kinase, 15 are potential therapeutic targets in various cancers, including NB. [18][19][20] Indeed, molecules as ALK inhibitors were found to be appropriate in patients whose tumours harbour activating ALK mutations.
Although some mutations seem to be resistant to current ALK inhibitors (ie F1174L), new drugs have already been formulated to overcome this resistance. 17 Moreover, these drugs are actively being evaluated in the New Approaches to Neuroblastoma Therapy (NANT) consortium. 19 In this study, we analysed a cohort of 45 high-risk NB patients and 9 NB cell lines by targeted (t)-NGS customized TK domains panel.
We identified a somatic variant p.V871I in EPHB4 gene. EPHB4 plays a crucial role in cardiovascular development and regulates vascularization in cancer-promoting angiogenesis, tumour growth and metastasis. 21 Several EPHB4 mutations have previously been identified in solid and haematological tumour specimens. [21][22][23] Many other EPHB4 variants have been identified in other types of tumours and cell lines and catalogued in "The Cancer Genome Atlas project." 24 However, EPHB4 mutations were not described until now in NB. We demonstrated that higher EPHB4 expression is correlated with poor overall survival. Moreover, the functional study highlighted the role of the variant by increasing proliferation, migration and invasion in NB cells. Of note, the treatment of the cells with two EPHB4 inhibitors, JI-101 and NVP-BHG712, was able to rescue the phenotype driven by the variant suggesting that EPHB4 is a promising therapeutic target in high-risk NB.

| Targeted-(t)NGS panel design
The "kinome" custom sequencing panel was designed to cover the kinase domains of TKs. KinBase (http://kinase.com/; a database of protein kinases) was queried to retrieve the coding regions of kinase domains. 20 Six hundred seventeen coding regions with a mean length of 260.66 bp (40-9408 bp) were selected, and their genomic coordinates were extended of 50 bp up-and down-stream to get a final target of about 222.5 Kb.

| High throughput sequencing
For high-risk NB samples, kinome targeted regions were captured and enriched using the SeqCap EZ Library SR (Roche NimbleGen).
Captured DNAs were subjected to massively parallel sequencing using an Illumina HiSeq 1000 obtaining 90 bp paired-end reads.
For NB cell lines, kinome targeted regions were captured and enriched using the Agilent HaloPlex target enrichment system (Agilent Technologies) according to the manufacturers' protocol. The sequencing was performed on an Illumina HiSeq 1000 yielding 90 bp paired-end reads.

| Sequencing data processing and mutation calling
Illumina paired-end reads of NB samples were mapped versus the reference genome (GRCh37/hg19 downloaded from UCSC Genome Browser) using the BWA (Burrows-Wheeler Aligner) 25  ESP 6500, ExAC, CG46); (d) without sequencing strand bias (variants with a strand bias over the 90% were set aside to reduce false-positive calls); (e) protein sequence impacting (eg missense); (f) with CADD score greater than 10 because these variants were considered as pathogenic. The set of variants obtained was manually curated and visually inspected with the IGV-Integrative Genomics Viewer. 29

| Public data sets analysis
Gene expression, survival (GEO Id: GSE45547; n = 649) and copy number analysis (GEO Id: GSE103123; n = 553) were performed by using the R2 Genomics Analysis and Visualization Platform (http:// r2.amc.nl). The overall and the event-free (relapse-free) survival probability were calculated by using the Kaplan-Meier method, and the significance of the difference between Kaplan-Meier curves was calculated by the log-rank test. We defined "high" and "low" groups containing samples with EPHB4 expression greater than, or lower than, its median value, respectively. Copy number gains for GSE103123 (a data set including 346 aCGH and 207 SNP arrays) were defined as in Depuydt et al 30 The selected DNA region, downloaded from R2 web tool, ranged from 100350145 to 100960971 on chromosome 7q. The cut-off to call copy number gains was set at Log Ratio = 0.2 and 0.15 for aCGHs and SNP arrays, respectively.
In gene expression and copy number analysis, we assessed the significance of differences between groups by t test and chi-square test, respectively. Statistical significance was set at 5%.

| Patients
A total of 45 surgical NB specimens were used for the kinome panel.
All the specimens were obtained at the time of diagnosis, before radiation therapy or chemotherapy, and were subjected to histopathological review according to the WHO criteria. DNA was obtained for genetic analysis from patients after signed informed consent, according to the Declaration of Helsinki, and as approved by local university ethical committees.

| RNA isolation, cDNA preparation and quantitative real-time PCR
Analysis total RNA was extracted from the cell lines using Trizol reagent (Invitrogen). Synthesis of cDNA from total RNA (2 mg) used SuperScript II First-Strand kits (Invitrogen). Quantitative real-time PCR (qRT-PCR) was performed using the SYBR-green method, following standard protocols with an Applied Biosystems ABI PRISM 7900HT Sequence Detection system. Relative gene expression was calculated using the 2 −ΔCt method, where ΔCt indicates the differences in the mean cycle threshold (Ct) between selected genes and the internal control. 31 Q RT-PCR primers for each gene were designed using Primer Express software, version 2.0 (Applied Biosystems).
Primer sequences are available upon request. The significance of the gene expression differences was determined using Student's t tests; statistical significance was established at P < .05.

| Vector cloning and site direct mutagenesis
cDNA encoding full-length wild-type EPHB4 (6584384) was obtained by Origene and then subcloned into pCMVtag1 vector (Agilent) using the HindIII-SalI restriction sites. The point mutation c.G2611A, p.V871I, was introduced into pCMVtag1-EPHB4 by sitedirected mutagenesis, as previously described. 32

| Western blotting
Total lysates of 50 µg were loaded and run on 12% polyacrylamide gels, which were then blotted onto polyvinylidene difluoride membranes (BioRad). These membranes were incubated with the following anti-

| Cell proliferation assay
The proliferation of SKNBE2 and SHSY5Y cells was assessed by seeding the cells stably expressing empty vector (EV), EPHB4-WT and EPHB4-V871I in medium containing 10% FBS in 96-well plates (20 000 cells/well). Cell viability was analysed after 72 hours using

| Colony formation in soft agar
SKNBE2 and SHSY5Y cells stably expressing EV, EPHB4-WT and EPHB4-V871I were plated at 2 × 10 5 cells/well in triplicates in 0.35% agarose-coated 6-well plates in the presence of medium containing 10% foetal bovine serum. After 2 weeks, the colonies were stained with crystal violet, and the numbers of colonies were counted.

| Monolayer wound-healing assay
SKNBE2 and SHSY5Y cells stably expressing EV, EPHB4-WT and EPHB4-V871I were plated in wells of a 6-well culture dish. Three parallel scratch wounds of approximately 400 mm width were made perpendicular to the marker lines with a P200 pipette tip (Corning).
The wounds were observed after 48 hours using phase-contrast microscopy as previously described. 34

| Migration assays
Migration of SKNBE2 and SHSY5Y cells stably expressing EV, EPHB4-WT and EPHB4-V871I through membranes with 8 µm pores was assessed using Transwell filter inserts assembled in 24well plates (Corning). Cells (7 × 10 5 in 200 mL serum-free medium) were then placed into the upper well of the membrane. Medium (500 mL) containing 10% FBS was added to the lower chamber as the chemoattractant. The assay plates were incubated at 37°C and 5% CO 2 for 16 hours for SHSY5Y cells and for 4 hours for SKNBE2 cells. Quantification of migration through the porous membranes was carried out by counting the stained cells (0.1% crystal violet/20% methanol) using a microgrid as previously described. 34

| Availability of data and materials
The NGS data sets generated and analysed during this study are available from the corresponding author on reasonable request. Public data and data repositories are referenced within the manuscript.

| Statistical analysis
All the data are presented as means + standard errors. Statistical significance was calculated using Student's t test. P-value < .05 was considered statistically significant.

| Kinome sequencing and identification of EPHB4-V871I mutation in NB patients
We performed targeted sequencing of TK domains on a total of 45 NB normal-primary tumour matched pairs and 9 NB cell lines. All tumour samples were high-risk patients according to the COG Risk Group Classification System (Table S1). Our NGS panel comprised the kinase domains of all 90 members of the TK gene superfamily (Table S2). After the sequence alignment, the mean read depth was about 360x and 1213x for NB samples and NB cell lines, respectively. The multigene panel showed high sensitivity and specificity.
Indeed, on average, the 97.08% and the 98.61% of analysable target bases were covered by at least 20 reads in NB samples and NB cell lines, respectively (Tables S3 and Table S4). The somatic variant calling returned a total of 796 (17.69 per sample) exonic variants.
After stringent filtering steps (see Material and Methods), we obtained five somatic mutations in four TK genes (Table 1). We found two missense mutations in ALK, one of these (F1174L) is the most frequent in NB.
Interestingly, we found two mutations in EPHB4 (V871I) and in EphB6 (A417S) genes, both involved in axon guidance pathway.
The variant V871I in the kinase domain of EPHB4 showed a high pathogenic score ( Figure 1A and Table 1). The same tumour analysed by whole-exome sequencing in our previous study 11 confirmed the variant V871I.
The screening of kinome regions in NB cell lines resulted in 11 filtered mutations in eight genes. Here, we detected four additional ALK mutations. Of these, three were F1174L changes, and the other was the R1275Q change (Table 1). We also found a mutation (N457K) in FGRF1 gene already found in other studies. 5,11,35 An analysis of public NB gene expression data set (GEO Id: GSE45547) 36 demonstrated that EPHB4 expression positively correlated with MYCN amplification. Indeed, we found higher expression in MYCN amplified NBs compared with MYCN non-amplified samples (P = 6.1 × 10 −17 ) ( Figure 1B). Survival analysis showed that high EPHB4 expression reduced both overall and event-free survival probabilities of NB patients (P = 1.0 × 10 −3 and P = 5.1 × 10 −3 , respectively) ( Figure 1C). To evaluate the independence of EPHB4 gene expression, and its effect on patient survival, from the MYCN amplification status, we restricted the survival analysis to patients without MYCN amplification. We confirmed the trend of association between high EPHB4 expression and reduced overall and event-free survival rates but with non-statistically significant values (P = 1.01 × 10 −1 and 6.2 × 10 −2 , respectively, Figure S1A,B).
Moreover, we surveyed the GSE3446 data set of 102 non-MYCN amplified NBs published, 37 and we again observed that high EPHB4 expression reduced event-free survival probabilities of NB patients (P = 6.5 × 10 −4 ; Figure S2). Moreover, we evaluated the presence of copy number variants involving EPHB4 region in a public data set of CGH and SNP arrays from a previous study (GEO Id: GSE103123). 31 We found that the 40% (206/515) of stage 4 NBs showed a copy number gain whereas only the 13.16% of the lower stages NBs (5/33) had a copy number gain event (P = 1.85 × 10 −3 ; Chi-square test) ( Figure 1D).

| Expression analysis of EPHB4 in NB cell lines
In order to select the cell lines to perform the functional charac- We selected SKNBE2 in the first group and SHSY5Y in the second one. To evaluate the expression of the EPHB4 mutant, we cloned EPHB4 wild-type (WT) and EPHB4 mutant V871I (MUT) in pCMV6-tag1. We then modelled our patient's genotype in vitro by stable clone transfection of WT and mutant EPHB4 expression plasmids into SKNBE2 and SHSY5Y cells. We found that the mutation did not impair EPHB4 expression at the mRNA and protein level ( Figure 2B,C). showed an increased number of colonies compared with WT and EV in both cell lines ( Figure 4B).

| EPHB4-V871I increases the expression of some target genes and enhance the phosphorylation of ERK1-2 pathway
We then tried to study possible targets of EPHB4. We studied three EPHB4 downstream target genes by analysing the mRNA levels of: VEGF, c-RAF and CDK4 genes by qRt-PCR. All three of these genes showed significantly higher levels of expression for EPHB4-V871I compared with EPHB4-WT for both the cell systems, SKNBE and SHSY5Y ( Figure 5A). We further assessed the phosphorylation of ERK1-2 pathway by Western blotting and we found increased phosphorylation status in EPHB4-V871I compared with EPHB4-WT for both cell systems, SKNBE and SHSY5Y ( Figure 5B).

| Treatment with TK inhibitors rescues the phenotype induced by EPHB4-V871I in NB cell lines
To test if EPHB4 could be a promising druggable target for high-risk NB, we treated our EPHB4 stable clones with two EPHB4 inhibitors,

| D ISCUSS I ON
We herein reported a novel somatic non-synonymous variant, V871I, in the EPHB4 gene occurring in our high-risk NB cohort. The Our custom kinome targeted sequencing on a total of 45 NB normal-primary tumour matched pairs, and 9 NB cell lines identified 11 filtered mutations in eight genes. According to previous sequencing screenings, 11 we observed very few somatic mutations in coding regions. Our data further confirmed ALK as most frequently mutated kinase gene in NB. Furthermore, here we report the FGFR1 mutation N457K, already found in our previous work and other sequencing studies on primary and relapsed NBs. 5,11 Despite FGFR1 mutation remains infrequent in NB, we can speculate on the presence of a mutational hot-spot in the kinase domain of FGFR1. FGFR1 mutations could be functional not only for the development but also for the selection of resistant and metastatic clones in NB.
Interestingly, we found two variants in EPHB4 (V871I) and EPHB6 (A417S), both genes are involved in axon guidance pathway.
These findings suggest that defects in genes involved in neuronal growth are an important category of tumour-driving events in NB.
Of note, we already reported another EPHB4 somatic pathogenic mutation (P257L) in our previous study comprising a total of 82 NB samples. 11   and extracellular signal-regulated kinase (ERK). 48 We found that the variant V871I enhances the expression of the mRNA levels of all the three genes in both the cell lines analysed. To confirm this, the phosphorylation of ERK1-2 was also increased in EPHB4 mutant compared to WT. This data demonstrated that downstream signalling pathway of EPHB4-V871I is enhanced compared with EPHB4-WT.

F I G U R E 5 Analysis of target genes and ERK pathway of EPHB4-V871I in NB cells. A, Quantification of VEGF, c-RAF and CDK4
Identifying genomic and genetic variants in Eph receptors' gene family may have valuable clinical implications since they are attractive targets for therapeutic applications in cancer. 49 Many strategies have been applied to evaluate the interference of tumour-promoting effects or the enhancement of tumour suppressive effects. The inhibition of the Eph-ephrin system may be particularly useful for anti-angiogenic therapies. 50 The ephrin-binding pocket in the extracellular N-terminal domain of Eph receptors and the ATP-binding pocket in the intracellular kinase domain could be potential binding sites for peptides and small molecules. 51 Consequently, multi-targeted tyrosine kinase inhibitors could also inhibit the kinase activity of EPHB4; some of these, such as EXEL-7647 and dasatinib, are in clinical trials. 50 Additionally, some new EPHB4 inhibitors have been obtained from kinase inhibitor libraries. NVP-BHG712 was reported to target EPHB4 and to also inhibit VEGF in vivo. 51 NVP-BHG712 also shows a synergistic effect with other chemotherapy drugs for solid tumours. JI-101 is an oral multi-kinase inhibitor that was demonstrated to inhibit VEGFR2, platelet-derived growth factor receptor F I G U R E 6 Rescue of the phenotype by treatment with EPHB4 tyrosine kinases inhibitors, NVP-BHG712 and JI-101, of EPHB4-V871I in NB cells. A, Proliferation assays by MTT in SKNBE2 cell line showing: EPHB4-WT stable clone treated with the vehicle-DMSO in black line; EPHB4-WT stable clone treated with NVP-BHG712 in red line; EPHB4-V871I stable clone treated with the vehicle-DMSO in black dotted line; EPHB4-V871I stable clone treated with NVP-BHG712 in red dotted line. The data shown are mean from three independent experiments, each carried out in sixfold. *P < .05, EPHB4-V871I treated with vehicle-DMSO vs EPHB4-V871I treated with NVP-BHG712, °P < .05 EPHB4-WT treated with vehicle-DMSO vs EPHB4-WT treated with NVP-BHG712. B, Proliferation assays by MTT in SKNBE2  cell line showing: EPHB4-WT stable clone treated with the vehicle-DMSO in black line; EPHB4-WT stable clone treated with JI-101 in red  line; EPHB4-V871I stable clone treated with the vehicle-DMSO in black dotted line; EPHB4-V871I stable clone treated  β (PDGFR-β), and EPHB4. 52 It was used in a phase I trial in patients with advanced solid tumours. 52 We used these two drugs to verify the possible rescue of the phenotype observed in EPHB4-V871I mutant. We assessed the proliferation rate of EPHB4-V871I after treatment with both drugs showing restoring of the proliferation rate at basal levels in both the cell lines. The rescue of the phenotype was also obtained regarding the migration properties after NVP-BHG712 and JI-101 treatment. Thus, the inhibition of EPHB4 tyrosine kinase activity rescued the malignant phenotype of EPHB4-V871I. So, we speculated on the possible use of these drugs in future therapeutic application to treat those patients carrying EPHB4 gain-of-function somatic mutations.
We demonstrated that our t-NGS panel comprising the kinase domains of TKs could be a promising screening tool to identify druggable mutations in TKs in NB. Moreover, our data suggest that genomic/genetic alterations can promote NB tumour progression by the activation of the druggable gene EPHB4.

CO N FLI C T O F I NTE R E S T S
We have nothing to disclose.