Chromogranin A regulates neuroblastoma proliferation and phenotype

ABSTRACT Neuroblastoma is a commonly encountered solid tumor in early childhood with high neuroplasticity, and differentiation therapy is hypothesized to lead to tumor mass shrinkage and/or symptom relief. CgA is a tissue specific protein restricted to the diffuse neuroendocrine system, and widely expressed in neuroblastomas. Using knockdown and knockout approaches to deplete CgA levels, we demonstrated that CgA loss inhibits SH-SY5Y cell proliferation and leads to a morphological shift with increased expression of Schwann and extracellular matrix specific molecules, and suppression of chromaffin features. We further confirmed the effects of CgA in a series of neuroblastoma cells with [BE(2)-M17 and IMR-32] and without (SK-N-SH) N-Myc amplification. We demonstrated that CgA depletion reduced IGF-II and IGFBP-2 expression, increased IGFBP-3 levels, and suppresses IGF downstream signaling as evidenced by reduced AKT/ERK pathway activation. This was further supported by an increased anti-proliferative effect of the ERK inhibitor in the CgA depleted cells. In an in vivo xenograft neuroblastoma model, CgA knockdown led to increased S-phenotypic marker expression at both protein and mRNA levels. Together these results suggest that CgA maintains IGF secretion and intracellular signaling to regulate proliferation and differentiation in neuroblastomas.


INTRODUCTION
Neuroblastoma is one of the most commonly encountered early-childhood extracranial tumors, and arises from the neural crest during embryonic development. This tumor retains plasticity and can differentiate into several tissue lineages, resulting in diverse clinical manifestations in terms of lesion location, tumor composition, disease stage and progression (Ngan, 2015;Tsokos et al., 1985). Additionally, a broad range of treatment outcomes is observed, including unresponsiveness to conventional radiation and chemotherapy, treatment-induced maturation to a benign ganglioneuroma and/or ganglioneuroblastoma, and even spontaneous regression (Cooper et al., 1991).
A heterogeneous cellular composition is typically observed in neuroblastoma tumor tissues and cultured cell lines, where distinct neuroblastic (N)-, substrate-adhesive (S)-and intermediate (I)-types have been demonstrated (Ross et al., 2003). N-type cells represent immature sympathoblasts, expressing neuronal skeleton markers such as neurofilament and neurotransmitter synthesizing enzymes including tyrosine hydroxylase; S-type cells show fibroblast or epithelial-like characteristics and express smooth muscle-specific proteins, including alpha smooth muscle actin, basic calponin and desmin. The final I-type cells are considered to be stem cells that can differentiate into either N-type or S-type cells (Piacentini et al., 1996), and interconversion between N-and S-type cells has been observed (Tsokos et al., 1985). Intuitively therefore, characterization of the mechanisms underlying this plasticity and improved understanding of factors that could direct tumor differentiation toward N-or S-type could not only advance our understanding of neural crest development, but potentially provide novel therapeutic strategies for neuroblastoma disease control.
Chromogranin A (CgA, NM_001275.3) is a 456-amino acid hydrophilic acidic protein of the granin family, expressed in a variety of endocrine, neuroendocrine, peripheral and central neural tissues (Bartolomucci et al., 2011). CgA functions as a key component of dense-core secretory granules and modulates the storage and processing of neuropeptide and peptide hormones in health and disease (Helle, 2004). Circulating CgA levels are elevated in a variety of neuroendocrine tumors (NETs), including carcinoids, pancreatic NETs, pheochromocytoma, paraganglioma, and neuroblastoma (Modlin et al., 2010). Serum CgA levels in neuroblastoma patients correlate with tumor burden and can be used as a sensitive and specific diagnostic and prognostic disease marker (Hsiao et al., 1990;Pagani et al., 2002). In vitro studies have demonstrated alterations in CgA transcription during neuroblastoma differentiation induced by retinoic acid and cAMP (Gaetano et al., 1995). However, the potential role, if any, for CgA itself in regulating neuroblastoma proliferation and/or differentiation remains unclear. In the current study, we have characterized CgA effects in a series of neuroblastoma cell lines and demonstrated that CgA depletion results in reduced neuroblastoma proliferation in vitro and in vivo and changes the neuroblastoma phenotype, indicating that CgA may be a promising therapeutic target for treatment of neuroblastoma and potentially other neuroendocrine tumors.

shRNA-directed CgA depletion inhibits in vitro neuroblastoma cell proliferation
To elucidate the biological function of CgA in modulation of neuroblastoma proliferation and differentiation, we used a short hairpin RNA (shRNA)-directed knockdown approach to deplete CgA expression in neuroblastoma SH-SY5Y cells in vitro. CgA knockdown efficiency was confirmed by real-time PCR [CgA mRNA expression (fold change), nonsense versus shRNA CgA, 1.0±0.1 versus 0.3±0.01, P<0.01, Fig. 1A] and western blotting [CgA protein expression (fold change), nonsense versus shRNA CgA 1.0±0.01 versus 0.1±0.03, P<0.05, Fig. 1B]. SH-SY5Y cells comprise three distinct morphologic phenotypes, among which N-type is the major type, followed by S-and I-type (Ross et al., 2003). We first observed morphological change of shRNA CgA cells compared to nonsense controls. The SH-SY5Y nonsense control cells displayed a typical N-type morphology predominantly characterized by an extensive network of neurite-like cytoplasmic processes (Fig. 1C, top panel). In contrast, SH-SY5Y shRNA CgA transfectants exhibited enlarged, firmly attached, polygonal shaped cells with occasional short processes (Fig. 1C, bottom panel). Concomitantly, a dramatic reduction in cell proliferation rates was observed in the shRNA CgA knockdown cells compared to nonsense control neuroblastoma cells as measured by the CellTiter-Glo ® luminescent cell viability assay (Fig. 1D) and BrdU incorporation assay (nonsense versus shRNA CgA, 1.0±0.1 versus 0.6±0.05, P<0.01, Fig. 1E). The calculated doubling time of CgA knockdown neuroblastoma cells was 1.5-fold longer than the nonsense control cells [T 1/2 (days), nonsense versus shRNA CgA, 2.0 versus 3.1, P<0.005, Fig. 1D]. We also used Caspase-3 activation assay to determine the effect of CgA knockdown in cell survival, but no difference in cell death was observed (data not shown), indicating a pro-proliferative effect of CgA. In soft agar assays, fewer colonies formed in the SH-SY5Y shRNA CgA transfectants compared to nonsense control cells (number of colonies, nonsense versus shRNA CgA, 1416±254 versus 118±72, P<0.01, Fig. 1F), indicating that anchorage-independent growth was markedly impaired in the CgA knockdown cells. To further substantiate the shRNA-mediated CgA knockdown effect in neuroblastoma proliferation, we performed rescue experiment using an shRNA-resistant CgA plasmid which contained optimized CgA codon sequences to avoid recognition and degradation by the CgA shRNA [CgA mRNA expression (fold change), vector versus CgA Rescue, 1.0±0.1 versus 11.3±0.04, P<0.01, Fig. 1G, left panel]. Rescue re-expression of CgA in the shRNA CgA neuroblastoma cells resulted in increased BrdU incorporation (vector versus CgA rescue, 1.1±0.1 versus 2.1±0.05, P<0.05, Fig. 1G, right panel). These findings demonstrated that shRNA-directed CgA depletion reduced cell proliferation, inhibited anchorage-independent growth, and resulted in a striking morphology change in the human neuroblastoma SH-SY5Y cells in vitro.

Confirmation of the role of CgA in cell proliferation and differentiation in multiple neuroblastoma cell lines
To further support our findings on the actions of CgA to alter neuroblastoma phenotype and proliferation rates in vitro, we also used CRISPR-Cas9 to completely deplete CgA expression in the SH-SY5Y cells. We used a 20-bp region in CgA Exon 2 as a single guiding RNA (sgRNA) to direct Cas9-mediated insertions or deletions (indels) to knockout CgA in the SH-SY5Y cells, which was confirmed by western blotting (Fig. 3A). Complete CgA loss in SH-SY5Y cells led to marked inhibition of in vitro proliferation measured by CellTiter-Glo ® luminescent cell viability assay (Fig. 3B) and BrdU incorporation assay (control versus CgA sgRNA, 1.1±0.2 versus 0.57±0.08, P<0.05, Fig. 3C). We also observed the same altered phenotype in the CgA knockout cells with differentiation toward an S-type as evidenced by increased VIM, FN and COL4A1 mRNA expression detected by real-time PCR in the CgA knockout cells compared to control cells (relative mRNA expression, control versus CgA sgRNA, VIM 1.0±0.3 versus 4.5±0.9; FN 0.9±0.1 versus 2.2±0.6; COL4A 0.9±0.1 versus 3.2±0.5, P<0.05, Fig. 3D). These knockout studies corroborated our prior findings using shRNA and demonstrated that reducing CgA expression in neuroblastoma SH-SY5Y cells inhibits in vitro cell proliferation and promotes cell differentiation toward a Schwannian cell phenotype. To evaluate the role of CgA more broadly in neuroblastoma, we compared endogenous CgA expression in three additional cell lines with (BE(2)-M17 and IMR-32) or without (SK-N-SH) N-Myc amplification. We found that BE (2) Fig. 4F). Collectively, these results supported our initial findings in SH-SY5Y cells and confirmed the effects of CgA in cell proliferation and differentiation in neuroblastoma.
CgA knockdown suppresses IGFR, AKT and MAPK activity, and IGF2 addition rescues the proliferation defects of the knockdown cells As noted above, CgA plays a crucial role in the biogenesis of secretory granules, and is involved in sorting and processing a wide spectrum of peptide hormones, amines and ions that maintain physiological homeostasis by both endocrine-and auto/paracrine-dependent actions (Bartolomucci et al., 2011). Insulin-like growth factors (IGFs) are intricately involved in regulation of tumor growth and differentiation, and previous studies have highlighted the role of IGF-II in neuroblastoma (Dake et al., 2004;Grellier et al., 2002). IGF binding protein-2 (IGFBP-2) expression frequently positively correlates with IGF-II levels and cell proliferation (El-Badry et al., 1991;Grellier et al., 2002). Secreted IGFs are non-covalently bound to IGFBP-1 to -6 with high affinity during transportation, and IGFBPs function as a reservoir to buffer IGF bioavailability, where IGFBP-3 is the predominant circulating IGF carrier (Tanno et al., 2004;Yu and Rohan, 2000). IGFBP-3-mediated actions in cancer vary in a disease-specific manner and some studies suggest that intracellular IGFBP-3 functions may be independent of its IGF binding ability (Baxter, 2014). We observed that knockdown of CgA in neuroblastoma SH-SY5Y cells resulted in a reduction in both IGF-II and IGFBP-2 mRNA expression (relative mRNA expression, nonsense versus shRNA CgA, IGF-II 0.9±0.1 versus 0.1±0.01, P<0.01, Fig. 5A; IGFBP-2 0.9±0.2 versus 0.1±0.1, P<0.01, Fig. 5B). CgA knockdown also led to increased IGFBP-3 expression (nonsense versus shRNA CgA, 1.0±0.1 versus 8.9±1.4, P<0.01, Fig. 5B). In contrast CgA knockdown did not alter expression of IGFBP-4, 5 or 6 (Fig. 5B), and expression of both IGF-1 and IGFBP-1 were under the detection limit of the assay in both the control and shRNA CgA SH-SY5Y cells (data not shown). IGF-II actions are transduced by cognate transmembrane receptor tyrosine kinases that include the tetrameric type I insulin-like growth factor receptor (IGF1R), the insulin receptor (INSR), and the monomeric IGF2R which lacks intracellular tyrosine kinase domain (Ryan and Goss, 2008). We next quantified phosphorylated IGF1R by western blot. As shown in Fig. 5C, lower phospho-IGF1Rβ (Y1135/1136) expression was seen in shRNA CgA cells compared to nonsense control cells, indicating attenuation of IGF pathway activation following CgA depletion. To further support a role for altered autocrine IGF-II proliferative actions following CgA depletion, we performed a 'rescue' experiment by treating both shRNA CgA and nonsense control SH-SY5Y neuroblastoma cells with exogenous IGF-II. As depicted in Fig. 5D, exogenous IGF-II treatment (100 ng/ml) resulted in increased proliferation in the ShRNA CgA cells compared to control cells (IGF-II versus medium control, nonsense 1.1±0.05 versus 1.0±0.01, P>0.05; shRNA CgA 1.4±0.003 versus 1.0±0.01, P<0.05, Fig. 5D).
As IGFs signal through the PI3K/AKT/MAPK pathways, we next examined phosphorylated AKT and ERK1/2 expression in our CgA knockdown cells. As depicted in Fig. 5E, CgA knockdown led to reduced phosphorylated AKT and phosphorylated ERK1/2 protein expression compared to that observed in control cells. To extend these findings, we next used ERK inhibitor (MEK-162) and demonstrate the treatment caused greater inhibition of cell proliferation in shRNA CgA compared to nonsense control cells  Fig. 5F). An illustration of the in vitro effects we have observed following neuroblastoma CgA depletion is described in Fig. 5G with reduced expression of IGF-II and IGFBP-2, combined alteration of which may contribute to reduced growth factor signaling as evidenced by reduced p-IGF1R signaling and increased responsivity to pharmacological inhibitor.

Flank xenografts of neuroblastoma cells lacking CgA show a shift towards an S-phenotype
We next tested effects of CgA depletion in neuroblastoma tumor growth in vivo. Human neuroblastoma SH-SY5Y cells stably transfected with shRNA CgA or nonsense control were subcutaneously inoculated into five-week-old male athymic nude mice at the density of 1×10 6 cells/animal. Mice were examined for tumor presence twice a week, and tumor volume was measured following palpable tumor development 4 weeks after tumor injection. The mice inoculated with SH-SY5Y shRNA CgA developed tumors later than the mice inoculated with nonsense control neuroblastoma cells (Fig. 6A). By 6 weeks after tumor injection, the animals were euthanized as tumors in the control group became debilitating. Tumor volume in the SH-SY5Y shRNA CgA group was reduced compared to that of nonsense control group [tumor volume (cm 3 ), nonsense versus shRNA CgA, 0.9±0.4 versus 0.3±0.1, P=0.1596, Fig. 6B]. The average tumor weight in shRNA CgA group was also reduced [tumor weight (g), nonsense versus shRNA CgA, 0.3±0.1 versus 0.2±0.1, P=0.5237, Fig. 6C], although these differences did not attain statistical significance. Immunohistochemical analysis for the S-type marker VIM in frozen tumor tissues demonstrated VIM was expressed in 72.7±5.7% cells from SH-SY5Y shRNA CgA bearing animals (n=4) compared to 2.6±0.5% in SH-SY5Y nonsense control tumors (n=4, P<0.01, Fig. 6D), and VIM mRNA expression was also elevated in tumors from shRNA CgA group (n=3) compared to those from nonsense control (n=2, nonsense versus shRNA CgA 1.0±0.1 versus 13. 3±1.1, P<0.005, Fig. 6D). These results corroborated our in vitro findings that CgA loss results in a shift towards an S-phenotype.

DISCUSSION
Neuroblastoma is a commonly encountered solid tumor in early childhood with a prevalence of ∼1 case per 8000 live births, and an annual incidence of ∼10 cases/million children aged <15 years (Ratner et al., 2016). Approximately half of these neuroblastoma tumors are classified as high-risk, and despite intensive multimodal therapy, including surgery, radiation and chemotherapy, the overall survival rate of this group of patients is <40% (Ferrari-Toninelli et al., 2010;Sun et al., 2013). Thus innovative therapeutic approaches are urgently needed to successfully treat this disease (Karmakar et al., 2011). Neuroblastoma arises from tissues derived from the neural crest and it is proposed that cells that become arrested along this migratory differentiation pathway develop into neuroblastomas (Thiele, 1990). Given the apparent multipotent neuroplasticity of the disease, it has been hypothesized that therapy which promotes differentiation and commitment toward specific cell lineages could lead to tumor shrinkage and/or relief of symptoms (Ratner et al., 2016). 13-Cis-retinoic acid, which induces a neuronal cell linage differentiation, is the most widely used agent in this regard, but after an initial response, chemo-resistance and tumor relapse is often seen (Lasorella et al., 1995). Better understanding of the mechanisms involved in differentiating those tumors could ultimately foster novel treatment approaches for neuroblastoma.
CgA is a tissue specific protein restricted to the diffuse neuroendocrine system, and widely expressed in neuroblastomas. Prior studies have demonstrated reduced CgA transcription following atRA and 12-O-tetradecanoylphorbol-13-acetate (TPA)induced neuroblastoma neuronal differentiation, concomitant with growth arrest (Gaetano et al., 1995). Conversely, glucocorticoidtreatment in vitro increased CgA expression and promoted chromaffin cell differentiation accompanied by increased N-Myc expression, a well characterized indicator of a poor prognosis (Ross et al., 2002;Rozansky et al., 1994). Underpinning the clinical relevance of this finding, a primary neuroblastoma located in or near the adrenal gland is often a higher grade tumor with a two-year survival rate of less than 20% (Ross et al., 2002). N-Myc amplification is prevalent in this group (Ross et al., 2002), and it has been proposed that the high regional steroid concentrations from the adjacent adrenal cortex inhibit sympathoblast neuronal differentiation and promote chromaffin maturation, resulting in a more aggressive disease phenotype (Gestblom et al., 1997). Furthermore, the glucocorticoid receptor antagonist RU-486 has been observed to decrease neuroblastoma cell proliferation (Maggi et al., 1998;Sengupta et al., 2000). Our studies herein show that CgA depletion using knockdown and knockout approaches resulted in reduced cell proliferation and promoted differentiation toward a Schwannian cell phenotype. Schwann cells are considered part of the stromal element in neuroblastoma and stroma-rich tumors have a more favorable prognosis compared with stroma-poor tumors (Ross et al., 2003). It has also been shown that the interplay between tumor and Schwann cells dynamically shapes neuroblastoma differentiation and disease outcome (Ross et al., 2003). For example, Schwann cell proliferation is promoted by the growing axons from differentiating neuroblastoma cells through direct contact and production of soluble chemotactic or mitogenic factors (Ambros et al., 2001). In parallel, Schwann cells secrete soluble factors and extracellular matrix that can promote tumor differentiation and inhibition of growth and angiogenesis (Ambros and Ambros, 1995). Our findings demonstrate that CgA depletion in neuroblastoma cells resulted in a cell morphological shift associated with increased expression of Schwann and ECM specific molecules (PMP22, SERPINF1, FN, LAMB2 and COL4A1), and suppression of features associated with the chromaffin phenotype (reduced IGF-II).
The MAPK/ERK signaling cascade is under the control of multiple extracellular ligands, mitogens and growth factors, including the insulin like growth factor family (IGFs). In some types of mesenchymal cells, such as bone marrow mesenchymal stem cells, constitutive activation of the MAPK/ERK cascade blocks differentiation into a smooth muscular lineage (Tamama et al., 2008). Confirming prior studies in other tumor types, we observed reduced AKT/ERK activation following CgA knockdown in neuroblastoma cells (Gong et al., 2007;Khan et al., 2012;Yu et al., 2003). Additionally, we demonstrated reduced IGF-II expression in CgA depleted cells and exogenous IGF-II treatment caused greater pro-proliferation effect in ShRNA CgA cells in comparison to control cells, which leads to our proposal that loss of CgA impairs IGF-II-mediated autocrine and paracrine proliferation signaling. This in turn leads to increased sensitivity to the antiproliferative effect of chemical inhibitor that we observed in the shRNA CgA knockdown neuroblastoma cells.
In summary, we demonstrated that depletion of CgA in neuroblastoma inhibits cell proliferation and leads to Schwannian differentiation in vitro. CgA depletion also led to impaired IGF autocrine signaling with reduced AKT/ERK pathway activation, which may underlie the molecular basis for our observed findings. Although we did observe reduced tumor growth in vivo in the shRNA CgA knockdown neuroblastoma cells, this did not attain statistical significance. This may have occurred for several reasons. Firstly, the number of animals was small and our study may not have been sufficiently statistically powered. Secondly, given that CgA is an ubiquitous protein and the growth factors such as IGFs are secreted by endothelial and other cells, the tumor in vivo microenvironment may have been able to compensate for the CgA-related loss. However, given the striking alterations we observed of CgA depletion in vitro with a marked phenotypic shift, we still feel that our findings are significant and could enable novel therapeutic approaches in neuroblastoma differentiation therapy.

Cell culture and reagents
Human neuroblastoma SH-SY5Y ( purchased from ATCC, CRL2266), BE(2)-M17 (from ATCC, CRL-2267), SK-N-SH and IMR-32 cells (kindly provided by Dr M. Sue O'Dorisio, Department of Pediatrics, University of Iowa) (Sue O'Dorisio et al., 1992) were cultured as monolayers at 37°C, 5% CO 2 using 1:1 mixture of ATCC-formulated Eagle's Minimum Essential Medium (30-2003) and F12 Medium containing 10% fetal bovine serum (FBS), and penicillin/streptomycin. They were recently authenticated and tested for mycoplasma contamination. The cultures were detached with trypsin and transferred to new 75-cm 2 culture flasks (Thermo Fisher Scientific) once a week. DMEM, FBS and antibiotics were purchased from Life Technologies, Inc. For in vitro cell culture studies, all trans-retinoic acid (atRA) and MEK-162 were solved in DMSO, and IGF-II (Thermo Fisher Scientific) was solved in DMEM/F12 at concentrations of 10 mM and 0.1 mg/ml respectively and stored in −80C freezer in aliquots, and later diluted as described in experiments.

CellTiter-Glo ® proliferation assay
Cells were suspended in 100 μl DMEM supplemented with 10% FBS, and plated in 96-well plates (2×10 4 viable cells/well) and cultured overnight. Cell viability was determined using CellTiter-Glo ® Luminescent Cell Viability Assay kit (Promega) with a luminometer (Wallac 1420 Victor 2 multipliable counter system). Results are presented as proliferation rate fold change (relative luminescence signal to nonsense control or medium control as indicated) and all experiments were repeated at least three times and presented as mean±s.d.

BrdU cell proliferation assay
Cell proliferation was quantified in 96-well plate (2×10 4 cells/well) using a colorimetric BrdU Cell Proliferation ELISA Kit (Abcam, ab126556) according to manufacture instruction. Results are presented as BrdU incorporation fold change (relative absorbance at 450 nm to nonsense control as indicated) and all experiments were repeated at least three times and presented as mean±s.d.

Anchorage-independent growth assay
Anchorage-independent growth (soft agar assay) was performed as described in our previous studies (Zhang et al., 2009). Briefly, 1×10 5 cells suspended in 0.33% soft agar were seeded over a bottom layer of 0.5% agar in 10% FBS DMEM in each well of six-well plates. The plates were incubated in 5% CO 2 incubator at 37°C for 3 weeks. Colonies were inspected under a microscope and only colonies with ≥32 cells were counted and all experiments were repeated at least three times and presented as mean±s.d.

Tumor Xenograft model
The use of mice was approved by the University of California Los Angeles (UCLA) Animal Research Committee and complied with all relevant federal guidelines and institutional policies. SH-SY5Y shRNA CgA or nonsense control neuroblastoma cells (1×10 6 ) in 100 μl matrigel were injected subcutaneously into five-week-old male Nu/J (JAX) mice to generate neuroblastoma tumors (n=10 each group). Tumor presence was checked twice weekly. Tumor diameters were measured in two dimensions with Vernier calipers and volumes calculated using the equation length×width 2 ×0.5. When the tumor diameter reached 2 cm, mice were euthanized using CO 2 inhalation. Tumors were excised, weighed and stored at -80C. Tumor tissues (n=4 each for SH-SY5Y nonsense-and shRNA CgA-bearing animals) embedded in OCT compound were cut with a cryostat (5 μm) and mounted on commercially available charged slides (Thermo Fisher Scientific). A specific anti-Vimentin antibody (DAKO, M0725, 1:100 dilution) (Hijaz et al., 2016) was used for IHC staining and the percentage of Vimentin immunoreactive cells was calculated by counting cytoplasmic Vimentin positive staining versus total nuclei in ten high power fields. Total RNA from frozen tumor tissue (n=2 for SH-SY5Y nonsense-bearing mice and n=3 for shRNA CgA-bearing mice) was extracted with RNeasy kit (Qiagen), and reverse transcribed into first-strand cDNA using a cDNA synthesis kit (Invitrogen). Quantitative PCR reactions were carried out to detect tumor tissue VIM mRNA expression.

Statistics
All in vitro experiments were repeated at least three times. Results are expressed as mean±s.d. Differences were assessed by Student's t-test using GraphPad Prism 4 (GraphPad Software, La Jolla, CA). P-values less than 0.05 were considered significant.