Introduction
Malignant glial neoplasms comprise 8–10 % of primary pediatric central nervous system (CNS) tumors [
1], and are the most common cause of solid tumor-related death in children [
2]. These lesions are classified as grade III or IV by the World Health Organization (WHO), and histology is most commonly anaplastic astrocytoma (AA; grade III) and glioblastoma (GBM; grade IV) [
3]. The cornerstones of therapy remain surgical resection and adjuvant chemoradiation, but tumor recurrence is inevitable and long-term response to treatment remains extremely poor. Development of novel therapies has been restricted by limited understanding of tumor biology and dependence upon preclinical models based on adult disease.
Genomic studies of pediatric high-grade gliomas (HGG) have demonstrated that the most frequent copy number aberration (CNA) is focal amplification of the platelet-derived growth factor receptor (
PDGFRA) locus at chromosome 4q12, encoding the α isoform of platelet-derived growth factor receptor (PDGFRα) [
4‐
7]. Several exome sequencing profiles of high-grade pediatric supratentorial and brainstem gliomas revealed a frequent, novel mutation of
H3F3A, encoding histone variant H3.3, that is associated with poorer overall survival [
8,
9]; among brainstem lesions, amplification of
PDGFRA occurred exclusively in
H3F3A-mutant tumors [
10]. PDGFRα protein expression corresponds to gene amplification, and is also seen in some tumors that do not contain
PDGFRA amplification [
11]. Expression of PDGFR correlates with high-grade histology in pediatric gliomas [
12], and associated activation of downstream Ras- and AKT-pathway signaling mediators has been correlated with poor survival [
13,
14].
The second-generation receptor tyrosine kinase inhibitor nilotinib (Tasigna; Novartis Pharmaceuticals) binds its targets with similar affinity to the prototype molecule imatinib (Gleevec; Novartis Pharmaceuticals) [
15‐
19]. Unlike imatinib, it is not dependent upon the OCT-1 transporter for cell influx [
20], and although it may be subject to transporter-dependent efflux [
21,
22], it is detectable within the brains of mice with an intact blood–brain barrier [
23]. Currently in use for treating imatinib-resistant and newly diagnosed chronic myeloid leukemia, nilotinib has a well-established safety and toxicity profile [
24,
25]. In this study, we examine the effects of nilotinib on pediatric GBM-derived cell lines, in order to understand the biochemical and biological impact of inhibiting PDGFR signaling and to evaluate its suitability as a therapeutic target.
Materials and methods
Cell lines and culture
Pediatric GBM cell lines SJ-G2 and SF-188 [
26,
27] were a gift from Dr. Nada Jabado (McGill University, Montreal, Quebec), and normal human astrocytes (NHA) immortalized with E7 and hTERT were a gift from Dr. Russell Pieper (University of California, San Francisco, CA, USA). Adult GBM cell lines U-118, U-87 and U-251 were obtained from the American Type Culture Collection (ATCC). All cell lines were cultured as adherent monolayers in minimal media supplemented with 10 % fetal bovine serum (FBS), at 37 °C in 5 % CO2.
Evaluation of exogenous ligand stimulation and inhibition
Methods and results are presented below for SJ-G2; parallel experiments carried out in SF-188 are presented in supplementary sections (see detailed methods in Supplement). Culture stimulation was performed using dimerized B-isoform of recombinant human platelet-derived growth factor (PDGF-BB; Cell Signaling Technology), which has affinity for both α and β isoforms of PDGFR. Cell cultures at 70–80 % confluence were incubated overnight in growth media containing 0.5 % FBS, then washed with warm phosphate-buffered saline (PBS), and exposed to fresh minimal media containing PDGF-BB at working concentration. Stimulation was terminated by placing cultures on ice and washing with cold PBS, then scraping and suspending in cell lysis buffer.
Western immunoblot assay
Cell cultures were scraped and lysed on ice in 1X Cell Lysis Buffer (Cell Signaling) containing protease (Roche Diagnostics) and phosphatase inhibitors (Calbiochem). Protein was quantified using bicinchoninic acid (BCA) assay (Thermo Scientific), and lysate containing 25–40 μg protein was loaded onto 7–12.5 % SDS-PAGE gels for electrophoresis. A detailed list of antibodies used is presented in Supplementary methods.
Cell viability, proliferation and colony formation assays
Cell viability was assessed by MTS assay using CellTiter 96
® Aqueous One Solution reagent (Promega) according to the manufacturer’s instructions. Direct count of viable cells using the trypan blue dye exclusion method was performed with the Vi-CELL Cell Viability Analyzer (Beckman Coulter). For cell proliferation assay, chemiluminescent Cell Proliferation ELISA (Roche) using 5-bromo-2′-deoxyuridine (BrdU) labeling was carried out according to the manufacturer’s instructions, using a 12-h labeling period. Clonogenic assay and soft-agar colony-forming assay were performed to evaluate in vitro growth potential [
28]. Colonies were fixed in 10 % formalin, then stained with 0.05 % crystal violet. Each condition was carried out in triplicate, in three independent experiments. Detailed methods are described in Supplementary methods.
Subcutaneous xenograft
SJ-G2 cells (1 × 106) were injected into the right flank of 6–8-week-old NOD–SCID mice (Jackson Laboratory). On detection of palpable tumor, mice were given nilotinib or vehicle by oral gavage (10 mg/kg daily), and sacrificed upon signs of sickness or at the end of the 10-week study. Excised tumor dimensions were measured with calipers, and volume estimated by the formula 4πabc/3, where a, b and c represent tumor radius in three planes. Animal use protocols were approved by the animal care committed of the Hospital for Sick Children (Toronto, Canada) under AUP#2656.
Discussion
Demonstration of PDGFRA amplification as the most common CNA among pediatric HGG, and the association of histone H3.3 mutation in malignant pediatric brain-stem glioma with PDGFRA amplification, identified PDGFRα and its signaling pathways as likely integral to the malignant behavior of some tumors. It is therefore identified as a potential therapeutic target. Recognizing that understanding of this disease and advancement of therapy has historically been hindered by reliance upon studies of adult gliomas, we have carried out investigations of PDGFRα signaling and inhibition in cell lines derived from pediatric glioblastoma tumors.
Pediatric high-grade tumors exhibit variability in
PDGFRA copy number, as well as in expression of PDGFRα and in platelet-derived growth factor (PDGF) itself [
11,
29‐
31]. Although the SJ-G2 cell line does not contain amplification of
PDGFRA, its expression of high levels of PDGFRα, the activation of which is dependent on dosage of ligand (PDGF-BB), suggests that it represents tumors that use PDGFRα signaling to support their tumorigenic behavior. To our knowledge, this is the first demonstration in a pediatric GBM cell line that activation of PDGFR is dependent upon ligand binding. This stands in contrast, for instance, to constitutively-activated mutant forms of epidermal growth factor receptor that are frequently found in adult GBM [
32,
33]. Activation of PDGFRα occurred rapidly, with peak phosphorylation seen within a minute of exposure to PDGF-BB, and decayed, suggesting that normal mechanisms of receptor internalization and degradation remain functional in these cells [
34].
Ligand-mediated PDGFR stimulation activates the PI3K–AKT and Raf–MEK–ERK signaling pathways [
35‐
37] and is dependent upon cell type and developmental stage [
38‐
40]. Stimulation with PDGF-BB of SJ-G2 cells resulted in phosphorylation of PDGFRα and corresponding activation of AKT and ERK1/2. Exposure to nilotinib inhibited PDGF-BB-mediated stimulation of the receptor, and markedly decreased activation of AKT and ERK1/2. Phosphorylation of tyrosine in response to PDGF-BB was generally decreased in the presence of nilotinib, suggesting that PDGFRα auto-activation was inhibited.
The effect of nilotinib inhibition was subsequently examined in the heterogeneous environment of serum-supplemented growth media. In this setting, we observed a low basal level of PDGFRα phosphorylation. Consistent with ongoing PDGFRα stimulation, we also detected basal activation of AKT, S6 ribosomal protein, MEK and ERK1/2, although activation of other cell-surface receptors may have contributed to this. Exposure to nilotinib decreased PDGFRα phosphorylation in all conditions. Phosphorylation of AKT and S6 ribosomal protein was rapidly reduced, and was maintained at a significantly decreased level during sustained inhibitor exposure, suggesting that some component of their activation is PDGFR-dependent. Interestingly, there was an early increase in MEK phosphorylation, within the first 2 h of nilotinib exposure. The rapidity of this change suggests that it results from an alteration in steady-state balance between signaling pathways.
Since inhibition with nilotinib suppressed PDGFRα-mediated signaling, we then investigated its effect on SJ-G2 cell numbers and proliferation. Our results show that cell proliferation and clonogenic potential (in vitro tumor-forming ability) were significantly reduced at 3 μM nilotinib compared to vehicle-treated control. The reduced proliferation resulted in fewer viable cells over the same time interval of growth compared to untreated controls. As patients receiving typical oral dosaging of nilotinib achieve a mean peak serum concentration of 3.6 μM [
24], the plateau in effect suggests that maximal anti-PDGFRα effect occurs at a concentration that can be achieved safely in the clinical context. The in vitro results were supported by in vivo findings of prolonged survival in nilotinib-treated mice bearing SJ-G2 tumors, and smaller tumor sizes in the treated mice.
These functional assays indicate that the result of nilotinib-mediated decrease in PDGFRα activation and signaling is a reduction in SJ-G2 proliferation, colony formation and in vivo growth. This overall decrease in tumorigenic behavior suggests that SJ-G2 may represent tumors that depend upon PDGFRα signaling and would therefore respond to nilotinib inhibition.
In contrast, SF-188 expressed very little PDGFRα and demonstrated high basal activation of AKT and ERK1/2, suggesting that this cell line is dependent upon activation of alternative growth factor receptor pathways to transduce mitogenic signals (Supplementary Figs. 2S, 3S, 4S). Consistent with this, we found that treatment with nilotinib had little effect on activation of AKT and ERK1/2. Furthermore, while treated cultures of SF-188 initially demonstrated lower viability, this difference was lost with sustained inhibition. SF-188 contains a mutation of the tumor suppressor and Ras-GAP NF1 [
41], which likely maintains activation of both AKT and ERK1/2 pathways by failing to regulate Ras activation. In contrast to SJ-G2, which contains wildtype NF1, SF-188 can be seen to represent the subset of tumors that do not depend upon PDGFRα activation.
All chemotherapeutic agents intended for use in the CNS must be able to penetrate the blood–brain barrier. Although its predecessor imatinib has poor CSF penetration [
42], regarding this question nilotinib remains under clinical investigation. Interestingly, nilotinib demonstrates in vivo activity against Abelson tyrosine kinase (Abl) within neurons of murine basal ganglia [
23].
As our understanding of the biology of pediatric HGG evolves, so does our recognition that they are a clinical and pathological entity distinct from adult tumors—and, indeed, that they encompass different subtypes. In this study, we identified the SJ-G2 cell line as representative of tumors that are dependent upon PDGFRα expression and activation to drive signaling pathways leading to tumorigenic behavior, and which are therefore most likely to be responsive to inhibition with nilotinib. Based on these data, we suggest that a subset of appropriately-identified pediatric HGG should be considered candidates for nilotinib treatment.