CAMKV Is a Candidate Immunotherapeutic Target in MYCN Amplified Neuroblastoma

We developed a computational pipeline designed to use RNA sequencing (n = 136) and gene expression profiling (n = 250) data from neuroblastoma tumors to identify cell surface proteins predicted to be highly expressed in MYCN amplified neuroblastomas and with little or no expression in normal human tissues. We then performed ChIP-seq in the MYCN amplified cell lines KELLY, NB-1643, and NGP to identify gene promoters that are occupied by MYCN protein to define the intersection with the differentially-expressed gene list. We initially identified 116 putative immunotherapy targets with predicted transmembrane domains, with the most significant differentially-expressed of these being the calmodulin kinase-like vesicle-associated gene (CAMKV, p = 2 × 10−6). CAMKV encodes a protein that binds calmodulin in the presence of calcium, but lacks the kinase activity of other calmodulin kinase family members. We confirmed that CAMKV is selectively expressed in 7/7 MYCN amplified neuroblastoma cell lines and showed that the transcription of CAMKV is directly controlled by MYCN. From membrane fractionation and immunohistochemistry, we verified that CAMKV is membranous in MYCN amplified neuroblastoma cell lines and patient-derived xenografts. Finally, immunohistochemistry showed that CAMKV is not expressed on normal tissues outside of the central nervous system. Together, these data demonstrate that CAMKV is a differentially-expressed cell surface protein that is transcriptionally regulated by MYCN, making it a candidate for targeting with antibodies or antibody-drug conjugates that do not cross the blood brain barrier.


INTRODUCTION
Neuroblastoma is an embryonal tumor that accounts for 12% of childhood cancer-related mortality (1,2). Prognosis for low-and intermediate-risk neuroblastoma is outstanding with the majority of patients cured, but there has been minimal progress for patients with high-risk disease. About half of all high-risk neuroblastoma patients die despite intensive multimodal chemoradiotherapy (2). Recently, clinical trials of GD2-targeting monoclonal antibody therapy have demonstrated significantly prolonged relapse free survival when used after achieving a first response to standard therapy, leading to regulatory approval in the United States and Europe (3). However, patients suffer significant pain during dinutuximab infusions due to GD2 expression on pain fibers, limiting dosing and in part explaining minimal improvements in overall survival with this immunotherapy. Our understanding of how to develop successful immunotherapies has advanced significantly in the last few years, but the limiting factor remains identification of cell surface molecules uniquely and uniformly expressed on tumor cells.
Focal amplification of MYCN occurs in roughly 40-50% of high-risk neuroblastoma cases (4)(5)(6) and is associated with an aggressive phenotype and poor prognosis (2,7). MYCN encodes a basic helix-loop-helix transcription factor that functions in transcription activation when heterodimerized with MAX, or transcriptional repression when heterodimerized with MNT, MXI, MAD, or other negative co-factors by binding to Eboxes within gene promoters (8,9). Gene-expression profiling has revealed a large cohort of genes involved in cell cycle, proliferation, signaling, adhesion, differentiation, and migration to be regulated by MYCN (10)(11)(12). However, while MYC family genes are known to transcriptionally regulate a very large number of genes via enhancer invasion (13), surprisingly little is known about direct MYCN target genes.
While MYCN amplification is prevalent in high-risk neuroblastoma and some other pediatric cancers, and is an important biomarker for patient outcomes, it remains an elusive drug target. While direct targeting of the MYCN transcription factor is not yet possible, several indirect methods have been proposed such as depleting MYCN protein levels with BET or AURKA inhibitors (14)(15)(16)(17), but these appear to be with limited anti-tumor efficacy. Here, we pursue another indirect strategy, identification of direct MYCN transcriptional targets that are located in the plasma membrane and thus amendable to new immunotherapeutic strategies.

Cell Lines and Chemicals
Cell lines were grown and STR validated as described (18)(19)(20). Cell lines were tested for mycoplasma when thawed and only grown for 20 passages following thaw. SHEP-2 MYCN-ER, and SK-N-AS MYCN-ER cells were obtained from the laboratory of Dr. Michael Hogarty at the Children's Hospital of Philadelphia. Cells were treated with 1 uM tamoxifen (Sigma h7904) to induce MYCN-ER nuclear translocation.

Lentiviral Preparation and Transduction
Lentiviral preparation was carried out as described (21). Briefly, using the clone TRCN0000020695 to deplete MYCN (Sigma), plasmids encoding shRNA along with the envelope encoding plasmid pMD2.G and packaging plasmid psPAX2 were transfected into 293T cells with Fugene 6 (Roche). Supernatant was collected 48 and 72 h later, filtered and added to IMR-05 cells in the presence of 8 ug/ml polybrene (Sigma). Puromycin (Sigma) was used to select for infected cells.

qRT-PCR
Total RNA was isolated from neuroblastoma cells utilizing RNeasy mini spin kits (Qiagen) and mRNAs were converted to cDNA using SuperScript II First Strand Synthesis kits (Life Technologies). CAMKV expression was detected using a Taqman probe (Hs01062060_g1, ThermoFisher) and MYCN was detected using (Hs00232074_m1, ThermoFisher), according to the methods previously described (19,21).

ChIP-Seq
Chromatin Immunoprecipitation-Discovery Cohort Chromatin immunoprecipitation was performed on the neuroblastoma cell lines Kelly, NB-1643 and NGPs as described (19,23,24). Briefly, using anti-MYCN (Santa Cruz B8.4B, sc-53993). Cells were grown to 80% confluence on 150 mm tissue culture plates in 20 mL of medium. Four hundred and fifteen microliters of 37% formaldehyde (final concentration 0.75% w/v) were added to the medium and the plate was rocked for 10 min to fix cells.

Chromatin Immunoprecipitation-Analysis
Phred sequencing scores (25) were calculated for each sample using a perl script (https://raw.githubusercontent.com/ douglasgscofield/bioinfo/master/scripts/phredDetector.pl) for automated input into the alignment algorithm. Quality was assessed using FastQC (http://www.bioinformatics.babraham.ac. uk/projects/fastqc/) and sequences were adapter-and qualitytrimmed using Trim Galore (26,27). Following QC, bwa (28) was used to align the reads to hg 19 reference genome and Picard tools (29) was used to remove duplicates. Fragment sizes were estimated using MaSC (30) and these were used as input into MACS2 (31) for peak calling. Peaks were called significant using a q-value (minimum False Discovery Rate) cut off of 0.05. Results were returned in units of signal per million reads to get normalized peak values. Finally, repetitive centromeric, telomeric, and satellite regions known to have low sequencing confidence were removed using the blacklisted regions defined by UCSC.

Chromatin Immunoprecipitation-Data Availability
Kelly and NGP MYCN ChIP-Seq have been published under GEO accession number GSE94782 and the remaining MYCN and MYC ChIP-Seq data are available under GEO accession number GSE138295.

Cell Plugs
10-20 × 10 6 live neuroblastoma cells were suspended in 50 uL of medium. A 1% w/v agarose in PBS solution was boiled and cooled to 50 • C in a water bath. The cell suspension was mixed with 0.5 mL of the agarose and transferred into an inverted capped Eppendorf tube with the conical bottom cut off, on ice. Once solidified, the plug was removed, cut in half, and fixed in 10% v/v buffered formalin (ThermoFisher Scientific).

Immunocytochemistry
CAMKV staining was performed using the CAMKV S-17 antibody (Santa Cruz Biotechnology, Inc.). Each sample was scored by the same pathologist and was designated as "1" when <10% of cells stained positively, "2" when 10-90% of tumors stained positively and "3" when >90% of cells stained positively. Only cells of targeted tissue type were evaluated for NCAM staining.

Membrane Isolation
Membrane fractionation was performed using the Proteoextract Native Membrane Protein Extraction Kit (Millipore, 444810) according to the manufacturer's instructions.

RNA-seq Expression Analysis
RNA sequencing data from 150 primary neuroblastoma tumors (136 high risk) was generated by the Therapeutically Applicable Research to Generate Effective Treatments project (TARGET data matrix, http://ocg.programs/target/data-matrix). RNA sequencing data were obtained from The Cancer Genome Atlas Program (TCGA), the Children's Brain Tumor Tissue Consortium (CBTTC), the Pacific Pediatric Neuro-Oncology Consortium (PNOC), the GEO database (GSE60052), and the Genotype-Tissue Expression (GTEx) project. Gene level data was generated using STAR alignment and RSEM normalization using hg 38 as a reference genome and GENCODE V23 gene annotation (33). The voom procedure was used to normalize the RSEM generated expected counts followed by differential expression testing using the R package limma to obtain adjusted p-values and Log-fold changes (LogFCs). Tumors were stratified by MYCN expression and sequencing data was queried for genes that had a LogFC >1 in the MYCN amplified tumors compared to non-amplified tumors with an adjusted p-value of 0.05 by the Benhanini-Hochberg procedure.

Gene-Expression Array Analysis
Gene expression data generated using Human Exon arrays (Affymetrix) from 250 primary tumors were obtained from the TARGET data matrix. These data were processed with Robust Mutlichip Average (RMA) normalization analysis implemented in the Affymetrix Power Tools (Affymetrix, Inc.). Samples were stratified by MYCN expression and the top and bottom 15% from this stratification were then used to query for the most differentially overexpressed genes in the MYCN-high subset of tumors.

Plasma Membrane Designation
The COMPARTMENTS database (http://compartments. jensenlab.org) was used to determine association with the plasma membrane (34). We designated only those genes to be potentially associated with the plasma membrane where the max confidence score across all GO categories was associated with Plasma Membrane or Cell Surface and was ≥3.

Identification of Plasma Membrane Proteins Specifically Over-expressed in MYCN Amplified Neuroblastomas
To determine the subset of genes which were differentially overexpressed in MYCN amplified neuroblastoma, we first used RNA-sequencing data from 150 primary tumors (136 of which were high-risk) and stratified them by MYCN expression ( Figure 1A). We then filtered for genes that had a log fold change over 1 in the MYCN amplified tumors compared to nonamplified tumors with an adjusted p-value of 0.05. We similarly analyzed microarray gene expression data from 250 primary neuroblastomas stratified by MYCN expression (Figure 1B). The top and bottom 15% from this stratification were then used to query for the most differentially overexpressed genes in the MYCN-high subset. We intersected these data with a list of genes predicted to encode proteins located in the plasma membrane by the COMPARTMENTS database (34) and found 14 to be consistently differentially overexpressed in the MYCN amplified tumors ( Figure 1C, Table 1, Table S1).
To prioritize these 14 candidate immunotherapeutic targets, we first surveyed for transcription start site (TSS)-proximal Ebox motifs, and showed that these were present in 11 ( Table 1).   8 at transcription start site-proximal E-boxes ( Figure 1D, Figure S1, Table 1). We next looked at the normal expression of these genes and eliminated any with widespread normal tissue expression. For this analysis, we included genes that had normal tissue expression limited to the central nervous system, considering larger biologics such as antibodies would not cross the blood-brain barrier (35), reducing our gene list to 3 ( Table 1).
To ensure that the protein products of these genes were not only expressed at the membrane, but were membrane bound and able to be targeted with a therapeutic, we eliminated genes with secreted products. Of the remaining two genes, NME1 has been described for its prognostic value in neuroblastoma and mutations in NME1 correlate with an aggressive phenotype (36,37). While NME1 is somewhat differentially overexpressed in MYCN amplified neuroblastoma (logFC = 1, adjusted pvalue = 1.17 × 10 −6 , Table S1) the protein product of NME1, nm23-H1, has diffuse sub-cellular localization that does not always include the plasma membrane (38) and mostly consists of the cytosol and endoplasmic reticulum (39), suggesting it would not be an ideal candidate for targeted immunotherapy. Calmodulin kinase-like vesicle associated (CAMKV) is more significantly differentially overexpressed in MYCN amplified tumors than NME1 (logFC = 1.73, adjusted p-value = 2.02 × 10 −6 ). CAMKV has limited normal tissue expression in the brain ( Figure S2) and has a TSS-proximal E-box that is occupied by MYCN protein. While we had intended to nominate candidate immunotherapy targets for MYCN amplified neuroblastoma that lacked normal tissue expression, the analysis did not yield any targets. Thus, we expanded our analysis to include normal expression on CNS tissues with the goal of using an antibody-based therapeutic that wouldn't cross the blood-brain barrier. From this analysis, we nominated the unstudied protein CAMKV as a putative immunotherapeutic target for MYCN amplified neuroblastoma.
To further explore the association of CAMKV and MYCN coexpression, we stratified RNA-sequencing and gene expression data from primary tumors by MYCN amplification status. We performed correlation analysis in primary tumors between MYCN and CAMKV and found them to be significantly correlated (Figures 2A,C). Finally, we examined if CAMKV protein expression is selectively expressed in MYCN amplified neuroblastoma cell lines. Whole cell lysate from 7 MYCN amplified neuroblastoma cell lines was compared to two nonamplified neuroblastoma cell lines, immortalized retinal pigment epithelial cells, and three non neuroblastoma, non-neural crest origin cancer cell lines and found CAMKV protein to be exclusively expressed in the MYCN-amplified neuroblastoma cell lines (Figure 2B). Together, these data suggest that MYCN amplification is a predictive biomarker of high CAMKV mRNA and protein expression.
A limited number of non-amplified tumors appeared to have high CAMKV expression (Figures 2A,C). A recently defined subset of high-risk neuroblastoma tumors that lack MYCN amplification overexpress the MYC protein and have a poor prognosis comparable to those with amplified MYCN (40)(41)(42).
While the majority of CAMKV-expressing cell lines are MYCN amplified (Figure 3A), the highest CAMKV-expressing cell line is NB-69 which does not harbor MYCN amplification, but has the highest MYC expression (Figure 3B) Figure 3C). The SK-N-SH cell line has the second highest MYC expression after NB-69, the second-highest CAMKV expression among non-amplified cell lines, and also has MYC occupancy at the CAMKV TSS-proximal E-box ( Figure 3C). Together, these data suggest that in some high-risk neuroblastomas that lack MYCN amplification, MYC can drive CAMKV transcription.

MYCN Protein Directly Controls the Transcription of CAMKV
To investigate whether MYCN protein regulates CAMKV expression, we transiently depleted MYCN from MYCN amplified IMR-05 cells using shRNA and saw a significant reduction in CAMKV mRNA levels ( Figure 4A, p < 0.005), suggesting that MYCN is required for CAMKV transcription in MYCN amplified neuroblastoma cells. Next, to determine if MYCN protein could drive CAMKV transcription, we used SHEP and SK-N-AS cells with a tamoxifen-inducible construct to translocate MYCN to the nucleus in cells that do not normally express MYCN. Upon MYCN translocation, we observed a significant increase in CAMKV mRNA expression in these cells, indicating that MYCN is also sufficient to drive CAMKV transcription (Figure 4B, p < 0.01). The MYCN ChIP-seq data revealed that the CAMKV TSS has a MYCN-occupied E-box motif. To confirm this occupancy, we designed primers to amplify the locus containing the E-box motif and a 5 ′ locus that did not have an E-box motif or an occupancy peak in the ChIP-seq data sets. We performed ChIP-qPCR in three MYCN amplified cell lines (KELLY, NGP, and IMR-05) and the non-amplified cell line, SK-N-AS. These data confirmed that the MYCN and MAX proteins occupy the E-box motif at the TSS in MYCN amplified cell lines and not in SK-N-AS cells and do not have significant binding to the 5 ′ locus (Figure 4C).

CAMKV Is a Plasma Membrane-Bound Protein in Neuroblastoma Cells
CAMKV is known to localize to neurite outgrowths in the brain and thus is predicted to be a membrane-bound protein. Furthermore, CAMKV is predicted to have a transmembrane domain (43). To confirm transmembrane localization in neuroblastoma, we first performed a membrane extraction in four MYCN amplified neuroblastoma cell lines [KELLY, CHP-134, LA-N-5, and SK-N-BE (2)c], three non-amplified neuroblastoma cell lines (NB-Ebc1, SK-N-AS, SK-N-FI) and RPE-1 cells. We used ADAM12 as a marker for the membrane fraction (44) and BCL-2 as a marker for the soluble fraction of cells (45). CAMKV protein was present in the membrane fraction of the four MYCN amplified neuroblastoma cell lines and NB-EBc1 cells [which have known low basal MYCN protein expression without locus amplification (18)] ( Figure 5A). Next, we created cell plugs from both MYCN amplified and non-amplified neuroblastoma cell lines and performed immunocytochemistry for CAMKV. This revealed CAMKV protein to be expressed and membrane-bound in MYCN amplified SK-N-BE (2)c cells but not expressed in nonamplified SK-N-AS cells ( Figure 5B). Finally, the membrane localization of CAMKV was confirmed in MYCN amplified NGP neuroblastoma cells by immunofluorescence ( Figure 5C). We used NCAM1 (CD56), an established cell surface molecule expressed in neuroblastoma as a marker for the plasma membrane (32,46) and found NCAM1 and CAMKV proteins colocalize on the plasma membrane of NGP cells. Together, these data show that CAMKV is a membrane-bound protein in MYCN amplified neuroblastoma with the potential to be targeted using a CAMKV-specific biologic.

DISCUSSION
Here we present a computational pipeline to discover cell-surface immunotherapeutic targets in MYCN amplified neuroblastoma.
MYCN amplification status is routinely tested for in all newly diagnosed neuroblastoma patients (1), and we sought to discover therapeutic targets that could utilize MYCN amplification as a biomarker for efficacy. The intersection of the most differentially overexpressed genes in tumors with the highest MYCN expression and genes predicted to be located in the plasma membrane resulted in 14 potential cell-surface targets in MYCN amplified neuroblastoma tumors, 11 of which have TSS-proximal E-boxes. Chromatin immunoprecipitation with sequencing (ChIP-seq) in MYCN amplified cell lines revealed that 4 of these TSS-proximal E-boxes were bound by MYCN protein. Two of these genes had low or absent expression in normal tissues outside of the brain, and CAMKV had the highest differential expression in MYCN amplified tumors and correlated with MYCN expression across all tumors. Perhaps not surprisingly due to the neural lineage specificity of MYCN expression, all candidate immunotherapeutic targets identified here showed expression in several brain tissues, obviating cellular immunotherapeutic or small molecule targeting approaches.
CAMKV encodes a pseudokinase that has sequence similarity to the calmodulin kinase family of proteins but lacks the essential residues required for kinase activity. CAMKV is present postnatally in the mammalian forebrain and is associated with vesicles in axons and dendrites. Recombinant CAMKV was found to bind calmodulin in the presence of calcium but lacks kinase activity with a sample substrate (47). CAMKV appears to be a substrate of CDK5 in the developing brain and functions in regulating dendritic spine maintenance (48). Remarkably, CAMKV is one of a very limited number of MYCN target genes whose protein product is localized in the plasma membrane and shows significant differential expression compared to normal non-central nervous system tissues. We previously reported that MYCN mediates GPC2 expression in neuroblastoma and that GPC2 was overexpressed in neuroblastomas that have 7q gains, where the GPC2 locus is. Due to high GPC2 expression in neuroblastomas with 7q gain and diploid MYCN, GPC2 was not one of the genes to be differentially overexpressed when comparing MYCN amplified to non-amplified tumors. Here, we nominate CAMKV as an immunotherapeutic target for MYCN amplified neuroblastoma, and suggest pursuing the identification of specific human scFv binders to this protein for the creation of antibody-drug conjugate therapeutics that should not cross the blood-brain barrier.

DATA AVAILABILITY STATEMENT
The datasets generated for this study can be found in the NCBI Gene Expression Omnibus (GEO) database (GSE138295 and GSE94782).

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2020.00302/full#supplementary-material  Table S1 | Gene lists from microarray or RNAseq analysis for MYCN high tumors and transmembrane domain-containing gene products.