Pharmacological WNT Inhibition Reduces Proliferation, Survival and Clonogenicity of Glioblastoma Cells

Abstract

Wingless (WNT) signaling has been shown to be an important pathway in gliomagenesis and in the growth of stem-like glioma cells. Using immunohistochemistry to assess translocation of β-catenin protein, we identified intranuclear staining, which suggest WNT pathway activation, in 8 of 43 (19%) adult and 9 of 30 (30%) pediatric glioblastoma patient surgical samples. WNT activity, evidenced by nuclear β-catenin in our cohort and high expression of its target AXIN2 in published glioma datasets, was associated with shorter patient survival, although this was not statistically significant. We determined the effects of the porcupine inhibitor LGK974 in 3 glioblastoma cell lines with elevated AXIN2 and found that it reduced WNT pathway activity by 50% or more as assessed by T cell factor-luciferase reporters. WNT inhibition led to suppression of growth and proliferation in the cultures and a modest induction of cell death. LGK974 reduced NANOG mRNA levels and the fraction of cells expressing the stem cell marker CD133 in neurosphere cultures, induced glial differentiation, and suppressed clonogenicity. These data indicate that LGK974 represents a promising new agent that can inhibit the canonical WNT pathway in vitro, slow tumor growth and deplete stem-like clonogenic cells, thereby providing further support for targeting WNT in patients with glioblastoma.

INTRODUCTION

Patients with glioblastoma (GBM) rarely survive more than 2 years after the initial diagnosis, making this malignant glioma one of the most lethal tumors overall (1). Glioma stem cells (GSCs) are thought to play a key role in long-term tumor growth and resistance to standard therapies. This subpopulation of cells is defined by its capacity for indefinite self-renewal and ability to initiate orthotopic xenograft formation (2, 3). GSCs also show multilineage differentiation potential (3–5), express markers found on non-neoplastic neural stem cells (6–9), and are relatively resistant to radiation and chemotherapy as compared to glioma cells lacking stem cell properties (10–12). Therefore, many research groups, including our own, have sought to identify molecular regulators required for the survival and proliferation of GSCs and to evaluate them as new therapeutic targets. One pathway implicated in GSCs and glioma pathobiology is wingless or WNT (13–17).

The WNT pathway has been shown to regulate a range of cellular interactions in normal development and diseases (18–20). It is a major pathway in stem and progenitor cells of the developing fetus (21, 22), as well as in adults (20, 23). For example, Clevers et al modulated WNT activity in the intestine of transgenic mice and showed that the pathway is required for stem cell homeostasis (24). Of note, WNT is also required for the maintenance of neural stem cells, and loss of signaling induces neural, glial and oligodendroglial differentiation (25–27).

WNTs are a large family of highly conserved protein ligands that are lipid- and palmitate-modified before being secreted as paracrine factors (28) and binding to cell surface receptors of the Frizzled-family and their LRP co-receptors (29).

Ligand/receptor binding recruits the downstream mediator Disheveled (DVL1) to the receptor site (30). Cytoplasmic levels of the key factor β-catenin (CTNNB1) are regulated though a specific degradation complex composed of the scaffolding protein AXIN, the product of the tumor suppressor gene adenomatous polyposis coli, casein kinase 1 and glycogen synthase kinase 3, which promote its phosphorylation and constitutive proteolytic degradation (18). WNT binding and recruitment of DVL1 disrupt this inactivation complex and lead to accumulation of free CTNNB1 in the cytoplasm, which then translocates into the nucleus, binds to transcriptional coactivators of the T cell factor/lymphoid enhancer factor (TCF/LEF) family, and promotes expression of genes involved in a variety of cellular processes important in tumorigenesis including growth (31–33), invasion (34–36), and therapeutic resistance (37, 38). One well-characterized WNT target is axis inhibitor protein 2 (AXIN2) (39–41). AXIN2 expression has previously been shown to directly correlate with WNT activity and aggressive behavior in GBM model systems (42–44).

In gliomas, WNT is generally activated at the level of ligand interaction rather than mutations (45). For example, the gene EVI, which is responsible for the secretion of the WNT morphogens, is frequently overexpressed in GBMs (46). Moreover, increased expression of pathway receptors Frizzled 2 and 9 as well as WNT6 by zinc finger protein PLAGL2 promotes tumor growth by impeding glioma differentiation (15). WNT5a has been shown to promote invasion by inducing the extracellular matrix metalloproteinase MMP2 (47) and proliferation of GBM cells (48). WNT3a increases nuclear translocation of CTNNB1 through induction of cell cycle regulator FoxM1 and thereby enhances gliomagenesis (38, 44). In addition, high levels of WNT receptor Frizzled 4 augments glioma invasion and therapeutic drug resistance through induction of EMT and reduced susceptibly to induce Caspase3-dependent apoptosis (49).

A number of prior studies have used modulation of upstream activators or genetic methods to inhibit WNT in gliomas. In general, these have shown inhibitory effects on GSCs and overall tumor growth (14, 36, 49, 50). However, direct pharmacological suppression of WNT activity in cancers has been challenging due to the lack of effective and specific WNT inhibitors. Recently, potent inhibitors of the WNT-specific acyltransferase porcupine (PORCN) that lead to disruption of ligand-driven activation of the pathway have been developed; these hold considerable promise as potential treatments. One such agent, LGK974, has shown therapeutic potential in experimental studies in head and neck, breast and pancreas cancers (51–53). An open-label Phase 1 clinical trial for various tumor types with documented genetic alterations upstream in the WNT pathway (https://clinicaltrials.gov/ct2/show/NCT01351103) is currently further investigating the clinical effects of this compound.

In this study, we demonstrate the ability of LGK974 to inhibit canonical WNT signaling in several in vitro GBM models with levels of AXIN2 expression similar to those seen in primary tumor specimens. GBM cells treated with the WNT inhibitor also showed significant reductions in overall cell growth, decreased proliferation and clonogenicity, lower expression of the stem cell marker CD133 and induction of glial differentiation. Taken together, these results suggest that targeting the WNT-pathway in GBM using the PORCN inhibitor LGK974 may represent a novel treatment strategy for malignant gliomas.

MATERIALS AND METHODS

Primary Tissue Samples, Cell Culture Models and Pharmacological Treatment

Snap-frozen samples from adult brain gliomas (GBMs: p349, p635, p636, p696, p770 and low-grade glioma: p824) were retrieved from the Johns Hopkins Neuropathology brain tumor tissue bank. All tissue collection and analyses were approved by the Johns Hopkins Institutional Review Board. Neurosphere cell lines (GBM1, JHH520, U87NS, GBM10) were maintained in serum free Dulbecco's modified Eagles Medium (DMEM)/F12 (Life Technologies, Carlsbad, CA), supplemented with B27 (Life Technologies), bovine fibroblast growth factor and human epidermal growth factor (both from Peprotech, Rocky Hill, NJ), and heparin (Sigma- Aldrich, St. Louis, MO), as previously described (54). The adherent glioma cell line LN229 was cultured in DMEM (Life Technologies) containing 10% fetal calf serum (Life Technologies) and 1x penicillin/streptomycin (Life Technologies). GBM1 was generously provided by A. Vescovi, Milan, Italy; AQH612 was provided by A. Quinones-Hinojosa, Department of Neurosurgery, Johns Hopkins Hospital, Baltimore, MD; JHH520 and JHH136 were provided by G. Riggins (Johns Hopkins Hospital). LN229 and U87 were purchased from American Tissue Culture Collection (Manassas, VA). GBM10 (54) and GBM14 (55) are neurosphere lines generated in our laboratory from intra-operative specimens obtained from the Department of Neurosurgery, Johns Hopkins Hospital. RNA was extracted from cell lines and tumor samples using standard techniques. RNA from pediatric brain tumor cell lines SF188, SU-DIPG, BT35, BT40, Res186, Res259, D283 and D425 were generously provided by E. Raabe, Department of Pediatric Oncology, Johns Hopkins Hospital.

The identities of the cell cultures were confirmed by analysis of 9 tandem repeats plus a gender-determining marker, Amelogenin using the StemElite kit (Promega, Madison, WI) in the John Hopkins Core Facility for DNA fragment analyses (http://grcf.med.jhu.edu/) as part of their standard STR fingerprinting service for cell lines (Supporting Information File S1). Cells were passaged before the porcupine-inhibitor (LKG974, #M60106-2S, Xcess Biosciences, San Diego, CA) was applied in the indicated concentration dissolved in cell line-specific cell culture media. Cell cultures were supplied with LGK974 every 48 hours in fresh media.

A tissue microarray containing multiple 0.6 mm cores from 35 pediatric and 45 adult glioblastoma samples was constructed as previously described (56). This tissue array as well as GBM1 cell pellets fixed in formalin then processed and sectioned in the same fashion as clinical specimens were stained for CTNNB1/β-catenin (Transduction Laboratories #610154, Lexington KY, 1:1,000 dilution) in the Johns Hopkins Hospital clinical pathology laboratory. Tumors were scored by a neuropathologist (Charles G. Eberhart) who was blinded to clinical and pathological findings as having no, weak, moderate or strong cytoplasmic/surface and nuclear immunoreactivity for CTNNB1 in at least 5% of cells.

Reporter Assay for Measurement of WNT/CTNNB1 Activity

Canonical WNT-pathway activity in the in vitro material was assessed using bioluminescence-based quantification with luciferase reporter construct (firefly luciferase cassette under the control of seven TCF binding sites, 7-TFP), as previously described (57). This reporter, which measures occupied CTNNB1 TCF/LEF-binding sites, was stably integrated into cells. Infectious lentiviral particles carrying the reporter were generated using the third generation lentiviral packaging system, as described (58, 59); stable integration was selected using puromycin (Sigma- Aldrich) at a concentration of 2 μg/ml. Cells overexpressing WNT due to the introduction of point-mutated CTNNB1/β-catenin served as positive controls and were generated in our lab, as previously described (34).

For each measurement, cells were harvested, washed in 1x phosphate-buffered saline (PBS) and lysed according to manufacturer’s descriptions using the Dual-Light luciferase & β-galactosidase reporter gene assay system (#T1003, Life Technologies). Luminescence readout was performed at 490 nm emission wave-length on an Infinite M1000Pro plate reader (Tecan, Morrisville, NC) and normalized to β-galactosidase activity.

Analysis of Gene and Protein Expression

The abundance of mRNA transcripts was assessed by conversion into complementary DNA and subsequent relative quantification using SYBR green-based fluorescence (Bio-Rad, Hercules, CA). Real-time PCR normalized to the housekeeping gene β-actin was performed with the △△Ct-method. Primer sequences can be found in Supporting Information File S2. Western blotting was performed as described before (60), and antibodies were used as per the manufacturer’s instructions (Supporting Information File S2).

For TCGA analyses, AXIN2 transcription levels acquired using the Agilent 244k microarray in 401 GBM specimens collected in The Cancer Genome Atlas ([TCGA], http://cancergenome.nih.gov) were retrieved through cBioPortal (http://www.cbioportal.org [61]) from the provisional Glioblastoma dataset in May 2015. AXIN2 expression values were correlated with available overall survival data provided by TCGA, with a cutoff of 1.75 standard deviations above the median (z-score) considered as AXIN2 high-expressing tumors.

Fluorescence Activated Cell Separation and Analyses

For CD133 cell surface immunostaining, 5 ×10⁵ cells were retrieved from a single-cell suspension and labeled with anti-CD133/1-phycoerythrin antibody (AC133, #130-080-801, Miltenyi Biotec, Cologne, Germany), according to the manufacturer’s instructions. The antibody solution consisted of anti-CD133/1-phycoerythrin (1:11) plus FcR blocking reagent (1:11, #130-059-901, Miltenyi Biotec) in 1x PBS (Life Technologies). All CD133-positive fractions were gated using respective control (AC133-pure, #130-090-422, Miltenyi Biotec). Fluorescence-activated cell analyses were performed on an Accuri C6 (BD Biosciences, Franklin Lakes, NJ). Fluorescence-activated cell sorting was performed on a SH800 (Sony Biosciences, Cambridge, MA) and FlowJo V10 software (Tree Star, Inc., Ashland, OR) was applied to perform data post-processing.

Cell Growth, Proliferation and Apoptosis Assays

Cultures were dissociated to single cell suspension and viable cells were quantified using the MUSE Count & Viability Assay Kit (#MCH100102, Merck KGaA, Darmstadt, Germany); then, 2000 cells per well (96 well plates) were plated in 100 μl triplicates. For this assay, GBM1 and JHH520 were grown as adherent cells on laminin (#L2020, Sigma-Aldrich)-coated plates (minimum of 3 hours prior cell plating with 20 μg/ml laminin solution) in stem cell culture conditions in order to replace with fresh media with consistent drug concentration (5 μM) every 48 hours. Relative cell numbers were measured at 1, 2, 3 and 4 days using the fluorescence viable cell mass assay TiterBlue according to manufacturer’s description (#G8081, Promega) on the Infinite M1000Pro plate reader (Tecan). Cell Titer Blue reagent was added directly to the cells (20 μl/well), incubated for 2 hours at 37°C and fluorescence intensity was measured at 560ex/590em nm.

To assess the effect of drug treatment on cell proliferation, we performed Ki67 quantification on day 2 and day 3 after initiation of pharmacological treatment using the Muse Ki67 Proliferation Kit (#MCH100114, Merck KGaA) on a Muse Cell Analyzer (#0500-3115, Merck KGaA). Reduction of proliferation was normalized to DMSO treated control and Ki67-stain was gated to respective IgG-stain control. For each sample, 50,000 cells were stained for 45 minutes.

Apoptotic cells were quantified using the Annexin V & Dead Cell Kit (#MCH100105, Merck KGaA) on the Muse Cell Analyzer according to manufacturer’s protocol. A minimum of 2,000 gated events were acquired.

Colony Formation Assay in Soft Agarose

Six-well plates were coated with a bottom agar/media layer made from a 1:1 mixture of a prepared 2x concentration of neurosphere media and 1% melted agarose (Life Technologies) in water. A single-cell triturated suspension was placed into a top agarose/ media mixture (0.7%) and immediately plated into the 6-well plates at a density of 3,500 cells/well (GBM1) and 5,000 cells/well (JHH520) in 1.5 ml of agarose; 1.5 ml of media (supplemented with drug or vehicle) was then placed into each well. For the effect of LGK974 on colony formation, 500 ml of fresh neurosphere media (with and without drug) was added to the treatment group every 48 hours. The experiment was stopped by viable cell visualization with nitroblue tetrazolium (Sigma-Aldrich) overnight at 37°C at day 19 and quantified using MCID Elite software (Cambridge, UK).

Statistical Evaluation

Kaplan-Meier analysis using log-ranked test compared the overall survival and was performed in Prism v4 (GraphPad Software Inc., La Jolla, CA). Statistical analyses of in vitro experiments, which each included a minimum of three biological replicates, were performed using two-tailed Student t-test and Statistica software (Statsoft, Tulsa, OK) and presented as mean values ± SD; p values ≤ 0.05 were considered significant.

RESULTS

Levels of WNT Activity in Surgical GBM Specimens and Glioma Cell Lines

We first used immunohistochemistry to examine the expression and localization of CTNNB1/β-catenin in surgical GBM specimens and in cell pellets from the GBM1 neurosphere line that were fixed and processed in parallel with the clinical samples. Of the primary tumors on our tissue microarray, 30 pediatric and 43 adult glioblastoma had sufficient material on the stained slides to score. In these, the degree of cytoplasmic and cell membrane CTNNB1 protein expression varied widely, with most cases showing no or weak expression; however, approximately one quarter showed moderate or strong expression (Fig. 1A, B). We did not identify the type of strong nuclear protein we and others have reported in tumors such as medulloblastoma, in which the pathway is activated by CTNNB1 mutation (62, 63). However, in 8/43 (19%) of adult GBM (Fig. 1A) and 9/30 (30%) pediatric GBM (Fig. 1B) we detected weak immunoreactivity in a subset of nuclei, which could potentially represent pathway activity. The presence of weak nuclear staining was seen in cases with a range of cytoplasmic expression and the 2 did not appear to correlate. The GBM1 neurosphere cells showed weak cytoplasmic staining levels similar to many GBM, but no evidence of nuclear CTNNB1 (Fig. 1C).

Figure 1.

CTNNB1/β-catenin expression in surgical adult and pediatric glioblastoma specimens. (A) Glioblastoma in an adult with moderate cytoplasmic and weak nuclear β-catenin immunoreactivity in a subset of cells (inset, arrow). (B) A pediatric glioblastoma with weak β-catenin expression in cytoplasm and in scattered nuclei (inset, arrow). (C) A weak cytoplasmic staining pattern, but no nuclear protein, was detected in formalin-fixed pellets from the GBM1 neurosphere cell line. (D) Patients whose glioblastoma contain nuclear β-catenin have shorter overall survival than those without signs of WNT activity (median overall survival: 17 vs. 20 months, p = 0.8)

The relationship between protein expression and clinical outcome was also evaluated. Patients with GBM showing nuclear CTNNB1 in their tumors had a median survival of 17 months as compared to 20 months for those without intranuclear staining. Log rank analysis of Kaplan-Meier survival curves revealed that this difference was not significant (Fig. 1D). Examining the prognostic impact of nuclear CTNNB1 in adult and pediatric cases individually revealed equal survival in adults (20 vs. 20 months) but shorter survival in patients younger than 18 years of age with nuclear protein (14 vs. 20 months), although even in these pediatric patients, the difference was not significant. Analyses of adults and pediatric GBMs with and without cytoplasmic CTNNB1 protein did not reveal any survival differences between patient groups. We also did not identify any correlation between cytoplasmic or nuclear CTNNB1 and expression of mutant IDH1 as detected by immunohistochemistry. These findings suggest that oncogenic WNT signaling is active in a subset of GBM but a possible association with worse clinical outcomes is not clear. Because immunohistochemical analysis was difficult due to the weak nuclear CTNNB1 staining, we sought to use more quantitative and sensitive methods to assess WNT signaling status.

Expression of AXIN2, an established target of canonical WNT signaling (34, 40, 44, 49, 64) has been shown to be associated with WNT activity and glioma stemness (42–44, 65). Therefore, we measured AXIN2 to determine if the brain tumor cell lines used in our laboratory had levels of pathway activity similar to those found in snap frozen patient specimens. As shown in Figure 2A, AXIN2 mRNA levels in the 6 adult tumor specimens examined (adult GBM: p349, p635, p636, p696, p770 and low-grade glioma/LGG p824) varied more than 2-fold between tumors. AXIN2 levels were even more heterogeneous in the in vitro models, including cell lines derived from 9 adult GBMs (GBM1, GBM10, GBM14, JHH136, JHH520, AQH612, U87, U87NS and LN229) and 9 pediatric brain tumors including 1 diffuse intrinsic pontine glioma (SU-DIPG, [66]), 1 anaplastic astrocytoma (BT35, [67]), 1 malignant atypical teratoid rhabdoid tumor (BT40, [67]), 2 low-grade (Res186 and Res259, [68]) and 2 high-grade glioma (KNS42, [59], SF188 [69]), as well as 2 medulloblastoma (D283 and D425, [70]).

Figure 2.

Quantification of the WNT target AXIN2 and clinical prognosis. (A) AXIN2 mRNA levels in snap frozen adult brain tumors (glioblastoma [GBM]: p349, p635, p636, p696, p770 and one LGG: p824), and brain tumor cell lines derived from adult GBMs (GBM1, GBM10, GBM14, JHH136, JHH520, AQH612, U87, U87NS and LN229) and pediatric brain tumors including diffuse intrinsic pontine glioma (SU-DIPG), anaplastic astrocytoma (AA: BT35), atypical teratoid malignant rhabdoid tumor (ATRT: BT40), low-grade glioma (LGG: Res186 and Res259), high-grade gliomas (pHGG: SF188 and KNS42) and medulloblastoma (MB: D283 and D425). (B) GBM patients with high AXIN2 have significantly shorter survival compared to patients with low AXIN2 (median overall survival: 208 days vs. 448 days) (z-score = +1.75, p = 0.06, TCGA dataset May 2015).

No clear correlation was noted between AXIN2 expression levels in our cell lines and patient age or tumor type. However, a number of the GBM lines showed levels of AXIN2 similar to those seen in primary tumors. We selected 2 neurosphere lines (GBM1 and JHH520) and 1 adherent line (LN229) that showed relatively high AXIN2 expression within the physiologically relevant range for further functional inhibition studies.

We also examined AXIN2 mRNA levels as a potential prognostic marker in 396 GBM specimens from the TCGA database with follow up data available. A number of potential thresholds for AXIN2 overexpression were tested, but while shorter survival was associated with higher levels of this WNT pathway target, the differences were not statistically significant. The most prominent effects were noted using the quite stringent thresholds for AXIN2 overexpression. In Figure 2B, we show that patients with high AXIN2 (defined as expression 1.75-fold or more SD above the median) survived 208 days vs. 448 days for the rest of the TCGA group (p = 0.06).

WNT-pathway Luciferase Reporter Assays in Cell Line Models

Luciferase reporter systems driven by CTNNB1/β-catenin binding to multimerized TCF/LEF promoter sites are frequently used to measure canonical WNT activity in cell lines. We introduced this reporter into our GBM lines using lentivirus and selected for stable integration using puromycin. Interestingly, TCF/LEF reporter signals correlated with AXIN2 mRNA levels, with moderate luciferase signals in GBM1 (0.4), higher in LN229 (0.7), and highest in JHH520 (1.1) (Fig. 3A), supporting AXIN2 as a marker of WNT activity in these tumors. When GBM1 and JHH520 reporter cultures were transduced with constitutively active (S33Y mutant) CTNNB1/β-catenin (34), we observed a more than 100-fold (GBM1) or 4-fold (JHH520) induction of luciferase activity (Fig. 3B), confirming the responsiveness of the reporter.

Figure 3.

LGK974 suppresses canonical WNT activity. (A) WNT pathway activity assessed by T cell factor (TCF) luciferase reporters. (B) Introduction of mutant active CTNNB1 (S33Y) increased the TCF reporter signals in GBM1 and JHH520 cells. (C) LGK974 inhibits WNT activity in a dose-dependent manner. (D) Treatment with 5 μM LGK974 for 48 hours effectively inhibited WNT signaling in all glioma cell lines tested (p ≤ 0.05).

LGK974 Treatment Reduces WNT Transcriptional Activity and Cell Growth in GBM Cells

We next tested the effect of LGK974 administration on WNT transcriptional activity in our culture models. In GBM1 cells, we identified 5 μM LGK974 as sufficient to suppress canonical WNT-signaling; a representative experiment is shown in Figure 3C. Additional experiments using this 5 μM LGK974 dose in GBM1 and other lines confirmed 40% to 60% WNT pathway suppression, which was significant in 2 lines (p ≤ 0.05, Fig. 3D). We chose to use this dose for functional studies to avoid potential non-specific cytotoxicity.

Glioma cells with suppressed WNT signaling showed reduced overall growth compared to their vehicle (i.e. DMSO) treated counterparts, as assessed with TiterBlue assay. Statistical significance was reached at 48 hours of treatment for GBM1 and LN229, and 96 hours for JHH520 (Fig. 4A). We also tested U87 and SF188, 2 cell lines with low AXIN2 baseline level as shown in Figure 2, but did not see significant growth inhibition after LGK974 treatment (data not shown), suggesting that WNT activity might be predictive of susceptibility to pathway inhibition.

Figure 4.

Decreased growth following WNT pathway blockade. (A-C) WNT inhibition significantly reduced cell growth as assessed with Titer Blue assay (A), cell proliferation assessed by fluorescence-based Ki67 quantification (B), and cell survival as assessed with Annexin V/Propidium iodide-based apoptosis and cell death quantification (C). *p ≤ 0.05 for all panels. 48h, 48 hours; 72h, 72 hours.

LGK974-sensitive lines showed a reduction in the percentage of proliferating cells as assessed with fluorescence-based quantification of Ki67 expression (Fig. 4B). An increase in apoptotic cell death, as detected by fluorescence-based quantification of Annexin-V/Propidium iodide-positive cells in was also seen in GBM1 and JHH520 after 48 and 72 hours of LGK974 treatment; however, no induction of apoptosis was observed in LN229 (Fig. 4C).

LGK974 Treatment Reduces In Vitro Clonogenicity and Induces Glial and Neural Differentiation

Canonical WNT signaling has been implicated as a regulator of the GSC marker CD133, and has been shown to affect glioma cell differentiation (13, 34). GBM cultures expressing the cell surface marker CD133 showed a reduction in the percentage of positive cells after treatment with LGK974 (GBM1:45% to 8%, U87NS: 17% to 9%, GBM10 50% to 43%) (Fig. 5A). The JHH520 and LN229 lines had very low baseline levels of CD133-positive cells (approximately 1%-2%), and this did not change significantly after LGK974 exposure (data not shown). When we sorted GBM1 cultures by CD133 expression using fluorescence activated cell sorting, the positive stem-like fraction had somewhat higher levels of WNT signaling, as assessed by the TCF-luciferase reporter, although the pathway was clearly also active in the better differentiated cells (Fig. 5B). LGK974 also reduced mRNA levels of the neural stem cell marker NANOG in all tested lines, reaching significance for GBM1 (Fig. 5C).

Figure 5.

LGK974 promotes differentiation. (A) LGK974 treatment reduced the fraction of cells expressing the glioma stem cell marker CD133. (B) CD133-positive GBM1 cells have higher canonical WNT activity as compared to their CD133-negative counterparts. (C) Reduced expression of NANOG after LGK974 administration (p = 0.05). (D) Increased glial fibrillary acidic protein (GFAP) (all tested cell lines) and microtubule-associated protein 2 (MAP2) (LN229) 72 hours after administration of LGK974. GAPDH, glyceraldehyde 3-phosphate dehydrogenase.

Because induction of differentiation can be an effective anti-cancer stem cell therapeutic strategy for malignant gliomas (71), we also assessed whether WNT blockade could promote differentiation. Western blot analysis revealed induction of the glial differentiation marker glial fibrillary acidic protein (GFAP) in LN229 and JH520, and a very minor increase in GBM1 (Fig. 5C). The neuronal marker microtubule-associated protein 2 was only induced in LN229 after WNT inhibition (Fig. 5D).

Given the reductions in the stem-like cell fraction, we evaluated the effects of LGK974 treatment on in vitro clonogenicity. Soft agar colony formation assays showed a significant reduction in the total number of spheres formed and average sphere diameter after drug exposure in both GBM1 and JHH520 neurosphere lines (p ≤ 0.001). However, the magnitude of the effect was much more prominent in GBM1, with over 80% average sphere number reduction and a mean colony size decrease from 147 μm to 86 μm (Fig. 6).

Figure 6.

LGK974 treatment significantly inhibited in vitro clonogenicity and reduced the average sphere diameter from (GBM1:147 μm to 87 μm, JHH520: 186 μm to 168 μm) compared to DMSO-treated control cells (p ≤ 0.001 for both parameters in both cell lines).

DISCUSSION

Recent studies have identified molecular changes affecting the WNT pathway in malignant gliomas (33, 72–74). These and other research suggest that ligand-driven upregulation of WNT is involved in glioma pathogenesis. For example, WNT3a can control tumorigenicity through regulation of the cell cycle regulator FoxM1, which promotes intranuclear accumulation of CTNNB1/β-catenin (44). In addition, ectopic expression of soluble frizzle-related proteins, secreted factors that interact with and control WNT ligands (75), inhibits glioma cell motility (76). The WNT-specific secretory protein EVI/WNTless, which controls the secretion of pathway ligands in canonical and non-canonical contexts, is overexpressed in high-grade brain tumors and promotes GBM malignancy (46). The expression of WNT inhibitory factor 1 (WIF1), a soluble inhibitor of the WNT morphogens, is downregulated in GBMs and thereby activates the WNT network, promoting tumor invasion (36). Given these data indicating that WNT signaling in gliomas can be activated at the level of ligands, we investigated the effects of LGK974 (51), an inhibitor of palmitoylation and extracellular secretion of WNT ligands (77), on adult glioma cells in vitro.

Using cell lines grown adherently or in serum-free media as neurospheres, we found that LGK974 could suppress canonical WNT activity in a dose-dependent fashion as measured by a highly sensitive luciferase-based reporter of transcriptional activity for the pathway. WNT pathway suppression was associated with significant reductions in the proliferation index and overall growth of all 3 lines tested. We also noted increased glial differentiation as evidenced by stronger expression of GFAP, and a reduction in NANOG levels and the percentage of cells expressing the stem cell marker CD133. Consistent with the notion that a stem-like fraction was being depleted, WNT inhibition caused a decrease in the size and number of colonies formed in soft agar.

In order to assess the extent of WNT activity in primary gliomas, we examined expression and localization of CTNNB1/β-catenin protein in 73 pediatric and adult surgical GBM specimens because nuclear translocation of this protein is associated with canonical signaling (62). Nuclear CTNNB1 was identified in 30% of pediatric and 19% of adult GBM samples, but expression was weak and present in relatively small numbers of cells. This assay did not show nuclear CTNNB1 in GBM1 cultures, suggesting that the immunohistochemical assay is less sensitive than TCF luciferase reporters, which identified biologically significant WNT activity in these cells. The weak nuclear staining may also reflect more modest ligand-driven activation of WNT signaling in gliomas as opposed to the adenomatous polyposis coli loss or CTNNB1 mutations, which activate the pathway in medulloblastoma (78). Nevertheless, together with prior reports our findings support the concept that WNT is active in a significant number of GBMs, and may be particularly frequent in pediatric tumors.

Several earlier studies found that increased cytoplasmic CTNNB1/β-catenin protein levels and nuclear translocation could be prognostically significant in patients with gliomas. Liu et al reported that 28% of the 43 glioblastoma they examined had cytoplasmic and nuclear protein, and that this was associated with significantly shorter survival (79). Other groups showed that increased cytoplasmic or nuclear CTNNB1/β-catenin was associated with both higher glioma grade and poor outcomes (80, 81). In contrast, Zhang et al found in a series of 63 astrocytomas that only 4 cases showed nuclear β-catenin immunoreactivity, and that while increased overall protein levels were not associated with tumor grade, they did correlate with shorter survival (82). Finally, the presence of nuclear Y333 phosphorylated β-catenin in GBM cells has been linked to shorter survival in a series of 84 patients (14).

Whereas we found a shortened overall survival in GBM patients at our institution whose tumors had nuclear CTNNB1 (17 vs. 20 months), this difference was not significant. When we analyzed only patients younger than 18 years of age the survival disadvantage of nuclear CTNNB1 was more pronounced (14 months vs. 20 months), but still not significant. Potential causes of the differences between our clinical correlation data and those from other groups include variations in staining protocols, scoring of combined nuclear and cytoplasmic protein in some studies, and the small size and mixed nature of many previously published clinical cohorts. We also examined the prognostic impact of the WNT target gene AXIN2. As was seen for nuclear CTNNB1, increased WNT activity as defined by high AXIN2 levels was associated with shorter survival, but this was not significant.

Despite the proven oncogenic role of WNT in gliomas (14, 15, 44, 83) and some progress in the development of pathway inhibitors (84, 85), little has been published on small molecule compounds targeting WNT, which are ready for clinical use in gliomas. One recent study tested the AXIN stabilizer SEN461, which showed promising therapeutic effects and inhibited WNT activity in GBM cells in vitro and in vivo (86). Lan et al used aspirin to suppress both WNT signaling and the invasion and survival of GBM cells, although this the anti-inflammatory agent can have a variety of effects (87). Another group investigated the anti-psychotic drug risperidone, which decreased CTNNB1 in gliomas and reduced the stem-like cell fraction (88). FH535, a small molecule WNT inhibitor, suppressed WNT activity in U87 and LN229 cells and reduced cellular invasion and proliferation (81). Finally, the use of genetic constructs to modulate the epigenome regulating WNT (i.e. miR-34a [89], miR-92b [90], miR-96 [74], miR-218 [35], miR-328 [91], and miR-603 [33]) is also possible, but efficient in vivo delivery will be challenging.

In summary, our data support the concept that WNT signaling is active in at least a subset of malignant gliomas, and that inhibition of this pathway can slow tumor growth, reduce the stem-like cellular fraction, and block clonogenicity in some GBM neurosphere lines. Our findings also suggest that the porcupine inhibitor LGK974, which is currently in trials for pancreatic and colorectal cancers, can effectively suppress WNT signaling in GBM in vitro. However, preclinical in vivo studies must be performed to assess the therapeutic potential of LGK974 in the treatment of brain tumor patients.