Acquired genetic alterations in tumor cells dictate the development of high-risk neuroblastoma and clinical outcomes

Human neuroblastoma (SH-SY5Y) cells with mixed neuroblast-like and epithelial-like
cells develop spontaneous high-risk aggressive disease in vivo

The subcutaneous administration of human SH-SY5Y cells resulted in the development
of ~200 mm
³
xenografts in ~70 % of the animals within 30 days, as described previously 26], 31], while the other 30 % of the mice were presented with multiple clinically-mimicking
aggressive metastatic tumors in the mediastinum and retroperitoneal, pelvic, abdominal,
and chest cavities as shown previously (27).

Fig. 1. Tumorosphere formation capacity of MSDAC. Representative time-lapse photomicrographs
of high-content imaging of parental SH-SY5Y and aggressive MSDACs. Cells were stained
with DiI and imaged in real-time every 20 min for 18 h with Operetta. Parental cells
(upper panel) showed monolayer spreading, MSDACs (lower panel) showed aggregation
and tumorosphere formation

Aggressive CSC-like MSDACs prompt tumorigenicity and reproduce high-risk disease

To better characterize the established high-risk aggressive disease model and to underscore
the enrichment of select clones from the parental line or ongoing acquisition of genetic
rearrangements, MSDAC clones were discretely characterized by karyotyping, whole genome
array CGH analysis, and tumorosphere-forming capacity. MSDACs are relatively small
and spherical with thin neurites. More importantly, every investigated clone of MSDACs
exhibited intrinsic CSC characteristics per the ability to readily grow ex vivo in serum-free medium and form large organized tumorospheres (28). This process is
presumed to simulate the events of tissue regeneration and maintenance from cells
that survive suspension conditions. In this process, an initial phase of symmetric
expansion of the seeding stem cells precedes a phase of asymmetric division, which
gives rise to the differentiated progeny that comprise the sphere bulk. Real-time
high-content observation of MSDACs under controlled conditions showed an aggressive
aggregation and tumorosphere formation within 18 h (Fig. 1; Additional file 1: video 1 and Additional file 2: video 2). Though parental SH-SY5Y cells and cells derived from non-metastatic xenografts
(Fig. 1) survived in serum-free medium, they exhibited monolayer cell spreading without tumorosphere
formation. In vivo, subcutaneous administration of MSDACs produced relatively large (500 mm
³
) xenografts as reported earlier 27]. The mice that received MSDACs presented with multiple metastatic tumors in the retroperitoneal,
pelvic, abdominal, and chest cavities, demonstrating the reproducibility of the high-risk
aggressive disease. Conversely, the mice that received parental cells did not exhibit
any distant metastasis, and hence served as the non-metastatic xenograft controls.

Fig. 2. Karyotyping in parental SH-SY5Y and MSDACs. Representative microphotographs showing
karyotyping patterns in parental SH-SY5Y and MSDACs by (a) G-banding analysis and (b) array CGH analysis. G-banding identical 47,XX, add(1)(q32), +del(7)?(q32), add(8)(p23), add(9)(q34), add(15)(q24), add(22)(q13)20] karyotyping in SH-SY5Y and MSDACs

G-banding certified that MSDACs from metastatic mouse tumors are derived from human
SH-SY5Y cells

Cancer cells are typically characterized by intricate karyotypes, including both structural
and numerical changes. To determine and illustrate that the aggressive tumors developing
in multiple metastatic sites were derived from the parental human SH-SY5Y cells, we
karyotyped MSDACs, with and without characterized CD133
⁺
, and compared these with the parental cells. All karyotyping was performed in double
blinded fashion. We investigated at least 20 cells per clone. SH-SY5Y cells exhibited
the 47,XX, add(1)(q32), +del(7)?(q32), add(8)(p23), add(9)(q34), add(15)(q24), add(22)(q13)20] karyotype, and served as the positive controls (Fig. 2ai ). All investigated clones of MSDACs exhibited an exact match of the parental SH-SY5Y
cells. We observed a unique marker composed of a chromosome 1 with a complex insertion
of an additional copy of a 1q segment into the long arm, resulting in trisomy of 1q.
Karyotyping also revealed six novel non-random chromosomal rearrangements on 1q32,
8p23, 9q34, 15q24, 22q13 (additions), and 7q32 (deletion; Fig. 2aii ). Consistently, array CGH analysis corroborated the karyotyping in the clones of
parental cells and MSDACs (Fig. 2b) and demonstrated that the developed aggressive metastatic tumors in mice are indeed
derived and disseminated from the parental SH-SY5Y cells.

Fig. 3. Genome wide copy number variations in parental SH-SY5Y and MSDACs. a Array CGH analysis showing digitized copy number variations (CNVs) across the genome
plotted for SH-SY5Y cells and MSDACs. b Table showing common copy number gain and/or loss across the clones of MSDACs. Chromosome
numbers, regions, and magnitude of CNV variation and corresponding genes are shown

Acquired genetic rearrangements in neuroblastoma cells drive aggressive disease

To determine any acquired genetic rearrangements and to underscore their impact on
disease progression, we utilized high-throughput whole genome array CGH analysis (Fig. 3a) coupled with quantitative transcriptional expression (QPCR). High resolution array
CGH analysis showed unique yet extensive copy-number variations (CNVs), including
insertions, deletions, and more complex changes that involve gain (duplication) or
loss (deletion) at the same locus in MSDAC clones (Fig. 3a, Fig. 4). However, in order to characterize the association of acquired genetic rearrangements
with disease progression, we considered only the common genetic variations across
the investigated clones of MSDACs. Forty-five common CNVs were observed with gain
in 30 (Chr.1,7; Chr.2, 3; Chr.4, 1; Chr.6, 1; Chr.7, 6; Chr.8, 8; Chr.11,2; Chr.17,2)
regions and loss in 15 (Chr.4,1; Chr.8,1; Chr.14,1; Chr.22,12) regions (Fig. 3b, Fig. 4). Interestingly, these CNVs correspond to the gain in the coding regions of CD1C, CFHR3, FOXP2, MDFIC, ADAM5, RALYL, CSMD3, SAMD12-AS1, MAL2, OR52N5, LOC400927, APOBEC3B, RPL3, MGAT3, SLC25A17, EP300, L3MBTL2, SERHL, POLDIP3, A4GALT, and TTLL1 genes. (Fig. 3b, Fig. 4). Unlike the healthy genome, in which changes in gene expression are carefully controlled
through transcription factors, the cancer genome adapts through the duplication of
CD1C, CFHR3, FOXP2, MDFIC, RALYL, CSMD3, SAMD12-AS1, MAL2, and OR52N5, and loss in the coding regions of ADAM5, LOC400927, APOBEC3B, RPL3, MGAT3, SLC25A17, EP300, L3MBTL2, SERHL, POLDIP3,
A4GALT, and TTLL1 genes. QPCR analysis revealed a CNV gain with a corresponding increase in transcriptional
expression of CD1C, FOXP2, RALYL, and MAL2 in MSDACs, but not in SH-SY5Y cells (Fig. 5a). Likewise, we observed a transcriptional repression of ADAM5, A4GALT, ABPOBEC3B, EP300, L3MBTL2, SERHL, SLC25A17, and POLDIP3, consistent with the CNV loss in MSDACs (Fig. 5a). Moreover, immunoblotting analysis revealed a profound increase in RALYL and FOXP2
translation in aggressive MSDAC clones as opposed to the parental SH-SY5Y cells (Fig. 5b). Like-wise we observed a robust increase in RALYL and FOXP2 expression in metastatic
tumors compared to the non-metastatic primary xenograft (Fig. 5b). Quantity one densitometry analysis revealed consistent increase in RALYL and FOXP2
expression both in ex vivo and in vivo settings (Fig. 5b side panel). Together, the definite genetic changes (CNV loss/gain) in the coding
regions of specific genes and their subsequent transcriptional/translational modulations
across MSDACs highlight the acquired genetic rearrangements in neuroblastoma progression.

Fig. 4. Copy number variations in parental SH-SY5Y and MSDACs. Representative copy number
variation charts showing gain in Chr.1, 158.35–160.00 MB; Chr.7, 114.084–114.115 MB;
Chr.8, 39.25–39.40 MB, and in Chr.8, 84.50–85.75 MB, corresponding to the coding regions
of CD1C, FOXP2, ADAM5, and RALYL, respectively, in MSDACs compared with SH-SY5Y cells

Fig. 5. Transcriptional and translational validation array CGH outcomes. a Histograms of QPCR analysis showing transcriptional amplification of CD1C, FOXP2, RALYL, NBPF20, and MAL2, and suppression of APOBEC3B, SLC25A17, EP300, L3MBTL2, SERHL, A4GALT, POLDIP3, and ADAM5 in clones of MSDACs compared with SH-SY5Y cells. b Representative immunoblots showing the expression level of RALYL and FOXP2 (both
showed gain in Array CGH analysis) in two different clones of metastatic site derived
aggressive cells (MSDAC) in comparison with the parental SH-SY5Y cells and in the
metastatic tumors derived from three different animals bearing high-risk aggressive
neuroblastoma (NB-MT-AD) in comparison with the non-metastatic primary xenograft (NB-NM-PX).
Side Panel: Histograms of Quantity one densitometry analysis showing robust increase
in RALYL and FOXP2 expression in MSDACs as well as in metastatic tumors in vivo

Fig. 6. Tumor grade associated expression of RALYL in human neuroblastoma. a Thumbnail and constructed images (20×) of human neuroblastoma tissue microarray coupled
with automated IHC showing RALYL expression levels in human neuroblastoma samples
(n = 25). b Aperio image analysis of the TMA and RALYL positivity quantification and subsequent
correlation of RALYL expression with neuroblastoma tumor grading

Acquired alterations associates with poor prognosis

To further substantiate our findings in clinical settings, we examined whether gain/loss
in the expression of such candidates correlates with high-risk neuroblastoma utilizing
a commercially available human neuroblastoma TMA. The tissues are derived from sites
including the retroperitoneal, abdominal, and pelvic cavities, the mediastinum, and
the adrenal glands. RALYL-IHC analysis revealed a significant distinction in RALYL
staining between patient samples (Fig. 6a). RALYL IHC revealed nuclear positivity with variable levels the human neuroblastoma
tissue cores analyzed. Positive RALYL staining appeared in brown and was selectively
localized in the nucleus (see 40× panel, Fig. 6a). Correlating the RALYL positivity to the tumor grade clearly identified the directly
proportional tumor-grade???RALYL expression association (Fig. 6b). RALYL positivity was relatively low in Grade 1, while its expression increased
per increased tumor invasive potential, with maximal gain in highly invasive tumors
(Fig. 6b).

Acquired genetic alterations are associated with tumor progression and poor clinical
outcomes

To underscore the importance of the observed genetic rearrangements in aggressive
disease, we first clarified their biological functions, network and communal molecular
orchestrations, and their documented role in any tumor progression systems. IPA “pathway
interaction analysis” revealed a complex yet well-organized signal transduction network
of MAL2, A4GALT, POLDIP3, RPL3, EP300, CD1C, CFHR3, APOBEC3B, RALYL, NBPF20, FOXP2, MDFIC,
TTL1, and MGAT3 (Additional file 3: Figure S1). Evidently, genes with genetic rearrangements in coding regions play
concomitant roles in multiple tumor systems, such as chronic myeloid leukemia, melanoma,
small cell carcinoma, lung carcinoma, mammary tumor, prostate cancer, pancreatic cancer,
colon adenocarcinoma, squamous cell carcinoma, and non-small cell lung adenocarcinoma.
Moreover, “IPA-Core-Analysis” revealed that this small subset of tightly inter-regulated
molecular targets showed influential participation in many canonical signaling pathways
and demonstrated defined roles in multifarious biological functions. IPA-data mining
considering only relationships where confidence?=?experimentally observed, these molecules
exhibited their role in at least 67 different canonical pathways exerting 150 biological
functions. Interestingly, in the light of tumor progression and dissemination, we
observed a significant association of these molecules in key pathways of cancer progression
viz., ATM Signaling, cAMP-mediated signaling, Cell Cycle:Checkpoint Regulation, CREB
Signaling in Neurons, Dendritic Cell Maturation, EIF2 Signaling, ERK/MAPK Signaling,
ERK5 Signaling, Estrogen Receptor Signaling, FGF Signaling, FLT3 Signaling in Progenitor
Cells, G-Protein Coupled Receptor Signaling, Granzyme A Signaling, HIF1a Signaling,
ILK Signaling, Neurotrophin/TRK Signaling, NFkB Signaling, p38 MAPK Signaling, p53
Signaling, Phospholipase C Signaling, PPAR Signaling, PPARa/RXRa Activation, Protein
Kinase A Signaling, RAR Activation, Pyrimidine Deoxyribonucleotides, TGF-b Signaling,
VDR/RXR Activation, Wnt/Ca?+?pathway, Wnt/b-catenin Signaling etc., (Additional file 4: Figure S2A). In addition to their role in molecular signaling events, these molecules
also exercise their defined (P??0.05) roles in cancer progression related bio-functions including Cancer Cell Morphology,
Progression of tumor, Cell Cycle-replicative senescence, Cellular Assembly DNA Replication,
Cell Cycle arrest, Cell Death and Survival, Cellular Function and Maintenance, Post-Translational
Modification, Cell-To-Cell Signaling, Cellular Assembly/Organization, Cellular Growth
and Proliferation, Cellular Movement, Cellular Response to Therapeutics etc., (Additional file 4: Figure S2B). To that note, all-encompassing overview of these molecules including
information on their symbol, name, subcellular location, protein functions, binding,
regulating, regulated by, targeted by miRNA, role in cell, molecular function, biological
process, cellular component, disease, role in tumor progression and metastasis etc., are provided in Additional file 5: Table S1.

To demonstrate the relevance of these genetic rearrangements to high-risk neuroblastoma
and poor clinical outcomes, we examined the correlation of individual gene expression
with overall (OS) and relapse-free survival in patients with neuroblastoma. We utilized
a web-based microarray analysis and visualization platform (http://r2.amc.nl) that correlates a select gene expression profile with clinical outcomes for samples
from multiple cohorts of patients with neuroblastoma. Kaplan-Meier plots showed a
significant association between increased expression of CFHR3, MDFIC, CSMD3, FOXP2, or RALYL (genes with gains in coding regions) and poor OS in patients with neuroblastoma (Additional
file 6: Figure S3A). This inverse association of CFHR3-, MDFIC-, CSMD3-, FOXP2-, or RALYL-gain also reflects poor relapse-free survival in these patients (Additional file 6: Figure S3A). Interestingly, SLC25A17, POLDIP3, SERHL, LOC400927, MGAT3, or TTLL1 (genes with CNV-loss in coding regions) demonstrated a definite association with
their loss and poor OS (Additional file 6: Figure S3B). The loss in any of these genes individually results in poor relapse-free
survival in children with neuroblastoma (Additional file 6: Figure S3B). Clinical outcome association analysis also revealed a strong correlation
between the expressional variations of both groups of genes listed above and stage
progression, favorable???unfavorable disease and alive???died-of-disease (data not
shown). It is pertinent to mention that gains in CD1C, NBPF20, and MAL2, and losses in ADAM5, RPL3, L3MBTL2, A4GALT, EP300, and APOBEC3B were not associated with poor clinical outcomes (Additional file 7: Figure S4). Together, these data demonstrate the direct, definite influence of genetic
rearrangements in aggressive disease on poor clinical outcomes in children with neuroblastoma.