Alpha-secretase inhibition reduces human glioblastoma stem cell growth in vitro and in vivo by inhibiting Notch

Abstract

The Notch pathway is dysregulated and a potential target in glioblastoma multiforme (GBM). Currently available Notch inhibitors block γ-secretase, which is necessary for Notch processing. However, Notch is first cleaved by α-secretase outside the plasma membrane, via a disintegrin and metalloproteinase–10 and –17. In this work, we used a potent α-secretase inhibitor (ASI) to test inhibition of glioblastoma growth and inhibition of Notch and of both novel and known Notch targets. Featured in this study are luciferase reporter assays and immunoblot, microarray analysis, chromatin immunoprecipitation (ChIP), quantitative real-time PCR, cell number assay, bromodeoxyuridine incorporation, plasmid rescue, orthotopic xenograft model, and local delivery of treatment with convection-enhanced delivery using nanoparticles, as well as survival, MRI, and ex vivo luciferase assay. A CBF1-luciferase reporter assay as well as an immunoblot of endogenous Notch revealed Notch inhibition by the ASI. Microarray analysis, quantitative real-time PCR, and ChIP of ASI and γ-secretase inhibitor (GSI) treatment of GBM cells identified known Notch pathway targets, as well as novel Notch targets, including YKL-40 and leukemia inhibitory factor. Finally, we found that local nanoparticle delivery of ASIs but not GSIs increased survival time significantly in a GBM stem cell xenograft treatment model, and ASI treatment resulted in decreased tumor size and Notch activity. This work indicates α-secretase as an alternative to γ-secretase for inhibition of Notch in GBM and possibly other cancers as well, and it identifies novel Notch targets with biologic relevance and potential as biomarkers.

Glioblastoma multiforme (GBM) is the most common variant of glioma, a devastating brain tumor with an average survival time from diagnosis of 12–14 months. Approximately 22 000 malignant primary brain tumors are diagnosed in the United States each year, and of these, 80% are GBM.¹ The current standard of care is surgical resection followed by radiation and treatment with temozolomide, a DNA-alkylating agent.² It is hypothesized that GBM initiation and recurrence is dependent upon a group of tumor stem cells that are resistant to current standard treatments.³ Therefore, new strategies to eliminate the GBM stem cell population are needed.

The Notch pathway has emerged as a potential therapeutic target for GBM and several other cancers, owing to its role in promoting stem cell and cancer cell survival and proliferation.^4–6 Notch is critical in brain development; its expression and activation are carefully timed to prevent cells from differentiating inappropriately and to maintain stemness.⁷ Cell–cell contact leads to the binding and activation of Notch by Delta-like or Jagged ligands, inducing cleavage of the Notch extracellular domain by metalloproteinases of the ADAM (a disintegrin and metalloproteinase) family (α-secretases),^8–10 followed by internalization and a second cleavage step that is carried out by the active γ-secretase protein complex.^10,11 The Notch intracellular domain (NICD) then translocates to the nucleus, where it activates a host of genes, including Hairy Enhancer of Split (HES), among others.¹²

We and others have demonstrated the importance of the Notch pathway in the proliferation and survival of GBM stem cells, suggesting it as a therapeutic target for this type of tumor.^6,13 Gamma-secretase inhibitors (GSIs) of Notch activity are currently being tested in clinical trials.¹⁴ However, GSI effectiveness may be hampered by the expression of efflux pumps in GBM.¹⁵ Efflux pumps have been implicated in chemotherapy resistance, because they actively eject drugs from cells. Other experimental ways to directly target Notch include use of antibodies for activated Notch isoforms on the cell surface, of ligand-binding inhibitors, and of endosomal acidification inhibitors.^16,17 We hypothesized that the use of an α-secretase inhibitor (ASI) as an alternative might be as or more effective than the use of GSIs for Notch, since ASIs act outside the plasma membrane and would not be affected by efflux pumps.¹⁸ ASIs in combination with other agents have already been tested against breast and non–small cell lung cancer lines and were shown to be effective in vitro through a mechanism of action unrelated to Notch inhibition.^19,20 The ASI compound INCB3619 (methyl(6S,7S)-7-[(hydroxyamino)carbonyl]-6-[(4-phenyl-3,6-dihydropyridin-1(2H)-yl)carbonyl]-5-azaspiro[2.5]octane-5-carboxylate) has been shown to prevent ADAM-10 and -17 from causing shedding and activation of ligands such as epidermal growth factor,¹⁹ and a nonspecific ADAM inhibitor has been shown to decrease Notch activity and inhibit myeloma growth.^19–21

We show here that human-derived adherent GBM cell lines as well as GBM stem cell lines are inhibited by treatment by INCB3619, in large part through Notch inhibition. Gene expression analysis indicated the repression of known Notch pathway genes downstream of ASIs and GSIs, but we also identified new Notch targets, such as YKL-40 and leukemia inhibitory factor (LIF). Furthermore, the treatment of human-derived GBM stem cells with ASI induced transcriptional activation of p53 and its downstream target p21. Finally, we show that local delivery of ASI via a novel nanoparticle platform significantly prolonged survival in a human GBM stem cell xenograft model in mice. These results suggest the development of INCB3619 as a therapeutic agent in GBM and other tumors with dysregulated Notch.

Materials and Methods

Cell Lines and Patient Samples

Adherent GBM cell lines U87MG, U251MG, T98G, U373MG, and A172 were acquired from the American Type Culture Collection. Human astrocytes were obtained from Lonza. GBM stem cell lines 0308 and 0822, typically maintained in neurobasal media with supplements, were derived and validated as described previously.²² All cell lines were grown under previously described conditions.

Pharmacologic Reagents

The GSI N-[(3,5-difluorophenyl)acetyl]-l-alanyl-2-phenyl]glycine-1,1-dimethylethyl ester (DAPT) was obtained from Sigma-Aldrich. INCB3619 was a gift from Incyte.

Luciferase Assay

Luciferase reporter assays were performed as previously described on a Promega Glomax 20/20 luminometer.⁶ For the measurement of Notch activity, 0308 and 0822 GBM stem cells and U251 GBM cells stably expressing CBF1-luciferase were used. CBF1-luciferase activity was normalized by dividing each well with the respective protein concentration.

Immunoblots

Immunoblots were performed as previously described.⁶ Primary antibodies included anti–Notch-1 (mN1A), anti-p21 (F-5), anti-p53 (Bp53-12) (all from Santa Cruz Biotechnology), and anti–α-tubulin (11H10) (Cell Signaling Technology). We used horseradish peroxidase–conjugated secondary antibodies to rabbit or mouse immunoglobulin G at 1:10 000 (Jackson Immunology Labs).

Microarray

The 0308 GBM tumor stem cell line was treated with dimethyl sulfoxide (DMSO, volume:volume [v:v]) as a vehicle and with 25 μM INCB3619 or 25 μM DAPT for 3 days, and then the frozen and pelleted cells were shipped to Cogenics. RNA was extracted from frozen cell pellets using the Qiagen miRNeasy Mini Kit (#217004), essentially as described by the manufacturer. RNA samples (100 ng) were labeled using the short hairpin (si)RNA Complete Labeling and Hyb Kit (Agilent Technologies #5190-0408) as described by the manufacturer. Labeled cRNAs were hybridized to Agilent human (V2) siRNA microarrays (#G4470B) and scanned with an Agilent DNA microarray scanner. Scanned image files were visually inspected and converted into data files using Agilent feature extraction software. Two replicates were performed with each experimental condition, and 2 arrays were performed for each of those replicates. The gene expression data from the biologic replicates were combined into an error-weighted average using the Rosetta Resolver Gene Expression Data Analysis System. Comparisons were made for expression of each gene between INCB3619 and DMSO and between DAPT and DMSO, generating metrics including fold-change and P value. Gene selection for Figure 2 was based on 2 criteria: fold-change of >1.5 and P < .05. Raw microarray data are provided in the Supplementary Materials. The microarray data are represented using a heatmap generated by Java Treeview 1.1.6 (freeware).

Fig. 2.

Microarray heatmap of mRNA expression level of predicted and novel ASI and GSI targets in response to 25 μM reagent or v:v DMSO in 0308 cells. Significantly different fold-changes in targets analyzed in the INCB3619 vs DMSO group and DAPT vs DMSO are shown using a microarray heatmap. Targets changed >1.5-fold and with P < .05 are shown; decreased targets are represented in green, and increased targets are represented in red. Black boxes indicate a statistically insignificant change for the target and treatment in question.

Quantitative Real-time PCR

Cells were lysed using Qiazol and then transferred to QIAshredder columns and centrifuged at 13 000 g for 3 minutes, and RNA was isolated using the miRNeasy Mini Kit according to the manufacturer's instructions (all Qiagen). Real time (RT)-PCR was used on 500 ng of RNA using the miScript Reverse Transcription Kit (Qiagen) to generate cDNA. From 100 ng of cDNA template, quantitative (q)RT-PCR analyses for HES1, HEY1, chitinase-3–like 1 (CHI3L1)/YKL-40, LIF, and α-tubulin were performed using their specific forward primers and reverse primers according to the protocol of the manufacturer (Qiagen). Primer sequences were as follows: HES1F: GGCTGGAGAGGCGGCTAA, HES1R: GAGAGGTGGGTTGGGGAGTT, 266 bp amplicon; HEY1F: ACGAGACCGGATCAATAACA, HEY1R: ATCCCAAACTCCGATAGTCC, 197 bp amplicon; CHI3L1/YKL-40F: CACACTCAAGAACAGGAACC, CHI3L1/YKL-40R: TGGCGGTACTGACTTGAT, 133 bp amplicon; LIFF: ACTTGTCTCTCCTCCCCAT LIFR: TGTACCATCTGCAGGGAAAT, 171 bp amplicon (all Real Time Primers). Sequences for α-tubulin were F: AGATCATTGACCTCGTGTTGGA, R: ACCAGTTCCCCCACCAAAG, 101 bp amplicon (National Institutes of Health). Alpha-tubulin was used as a control to normalize the levels of the respective genes. The StepOnePlus (Applied Biosystems) RT-PCR system was used to carry out qPCR, using a hot start at 95^oC for 15 minutes, then denaturation at 95°C for 15 seconds, with annealing at 60°C (30 seconds), extended at 72°C (30 seconds) for 40 cycles, followed by a melt curve analysis. Data analysis for the differences in gene expression between control and treated cells was carried out using Microsoft Excel: housekeeping gene primer threshold cycle (Ct) values were subtracted from test primer values to find the ΔΔCt, then the ΔΔCt was found by subtracting the average ΔCt of the vehicle-treated sample from itself and the drug-treated samples. Fold-change = 2^-ΔΔCt. Chromatin immunoprecipitation (ChIP) assay was performed. Nucleofection of 5 × 10⁶ 0308 GBM stem cells was performed with either an NICD-Myc plasmid or a TOPO-TA vector in duplicate using the Lonza Amaxa Mouse Neural Stem Cell Nucleofector Kit and then plated in large flasks for 3 days. Cells were processed with a ChIP assay kit (USB) according to the standard protocol. Per condition, 30 × 10⁶ cells were harvested and sonicated, with an average chromatin fragment size of approximately 800 bp. Sonicated chromatin representing approximately3 × 10⁶ cells per condition was incubated in aliquots. All aliquots were incubated with preblocked protein-A agarose beads, and then each sample supernatant was exposed to 4 μg of anti-mN1a antibody or mock bovine serum albumin. The mock incubation supernatant was saved as the input DNA positive control for the starting genomic DNA. The antibody-incubated, mock, and input DNAs were all purified by phenol extraction. The purified DNA was resuspended in 200 μL of RNase-DNase free ddH20 and assayed via qPCR with the primers listed in Supplemental Table 1. Analysis of qPCR was carried out by normalizing the input Ct according to the input dilution factor and then calculating the Ct as percent input [%input = 100*(2^{(normalized input Ct–experimental Ct)})].

Cell Counting

Cell number was evaluated to determine the effect of drug treatment. Cells were counted 4 days after treatment. Ten microliters were analyzed using a hemocytometer. For flow cytometry analysis of cell G-, S-, and M-phase cycles, cells were treated with either DMSO vehicle (v:v), INCB3619, or DAPT and stained with propidium iodide (Sigma). Cells were analyzed with the flow cytometry FACSVantage SE machine (Becton Dickinson). Assays were performed as previously described.^6,23,24

Bromodeoxyuridine Incorporation Assay

A cell-proliferation enzyme-linked immunosorbent assay (ELISA) for bromodeoxyuridine (Brdu) was used for the analysis of the number of GBM stem cells in the S-phase of the cell cycle after treatment with DMSO vehicle (equal v:v), INCB3619, or DAPT. Briefly, GBM stem cells were plated at 2000 cells per well in a 96-well plate and then treated for 48 hours. BrdU was then added and further incubated for another 24 hours. The incorporation of BrdU was determined using a colorimetric immunoassay kit according to the protocol of the manufacturer (Roche Diagnostics).

In Vivo Treatment Model

Mouse studies were approved by the Animal Care and Use Committee at the University of Virginia. Eight-week-old male Strain Code ID/NCr Balb/c mice (from the National Cancer Institute), 8–10 per group, were stereotactically implanted with 7500 0308 cells or 50 000 0308 cells stably expressing CBF1-luciferase GBM in 5–10 μL of Dulbecco's modified Eagle's medium. The surgical procedure was as previously described.¹³ For ex vivo luciferase activity and tumor volume analysis, convection-enhanced delivery (CED) of ASI-containing liposomes or control liposomes was done at 7 days post-implantation (10 mice per group). For nonmagnetic particles, 2 sequential treatments were performed at 7 and 35 days post–tumor injection. MRI of these mice was performed at 24 days post–tumor injection. Tumor volume was analyzed using 3 sequential images per mouse, by calculating pixel density in tumor-containing areas compared with mean nontumor areas plus 2 SDs. Average tumor volume per mouse was then calculated and compared. The mice were humanely killed 45 days post–tumor injection, and luciferase activity was analyzed. For ex vivo CBF1-luciferase activity, 3 mice per group were sacrificed; an intracranial tumor was isolated and exposed to luciferin and normalized by protein content as described above.

CED of magnetic liposomes (MLs) for a survival experiment was done 7 days after implantation. Animals were randomly divided into 3 groups for the survival experiment: a control group receiving empty MLs, an ASI treatment group receiving ASI-loaded MLs, and a GSI treatment group receiving GSI-loaded MLs (7 mice per group). We performed CED using the same coordinates as we used for tumor implantation. The CED volume was 20 μL, and speed was 300 nL/min; the solution also contained 7.5% mannitol to promote spread of the infusate. Distribution of MLs in brain parenchyma was visualized by MRI after completion of CED (data not shown). General appearance, neurologic status, and body weight were monitored daily, and mice were euthanized when they demonstrated signs of illness, pain, or 20% weight loss.

Synthesis and Loading of PalD2 Magnetic Nanoparticles

Core shell −type iron oxide PalD2-Mag1magnetic nanoparticles (MNPs) were synthesized as described elsewhere²⁵ by precipitation of Fe(II)/Fe(III) hydroxide from aqueous solution of the mixture of Fe(II) and Fe(III) salts, followed by transformation into magnetite in an oxygen-free atmosphere with immediately spontaneous adsorption of shell components. Technical details of MNP surface modification, gamma sterilization, and characterization were described previously.^26–29 The particles had a mean magnetite crystallite size of 4 nm and a mean hydrated particle diameter of 55 ± 11 nm. Iron content was 0.51 g Fe/g dry weight. PalD2-Mag1 suspension containing 5 mg Fe in 560 µL water was mixed with 0.5 mg ASI in 50 µL DMSO, and the resulting suspension was treated with ultrasound in a Bandelin Sonorex Digitec DT 102H water bath for 5 minutes. The resulting suspension of the ASI–PalD2-Mag1 conjugate with loading of 10% ASI based on iron weight of PalD2-Mag1 nanoparticles was sufficiently stable, with a mean hydrated particles/assemblies diameter of 216 ± 10 nm and electrokinetic potential of the particles of − 13 ± 1.3 mV when diluted in water (10 µL of the conjugate suspension/1 mL water) according to the measurements by photon correlation spectroscopy using a Malvern 3000 HS Zetasizer. A similar preparation of MLPs was also prepared, and the final concentration of ASIs in the lipoparticles was 0.5 mg/mL.

Statistics

In vitro and in vivo experimental results of 2 groups were analyzed by a 2-tailed Student's t test and plotted in Microsoft Excel. In vitro experimental results of more than 2 groups were analyzed using a 1-way analysis of variance for column plots and a 2-way analysis of variance for grouped analyses, with Bonferroni post tests to compare all columns. P values are column comparisons. The in vivo experimental results were analyzed using the Kaplan–Meier function in GraphPad Prism 5. Refutation of the null hypothesis was accepted for P <.05.

Results

ASI Treatment Decreases Notch Activity in a Similar Manner to GSI

First, we tested the effectiveness of INCB3619 (the ASI) on the inhibition of Notch activity compared with the widely used and potent DAPT (the GSI).³⁰ Treatment of an adherent GBM cell line stably expressing CBF1-luciferase (a well-established reporter of Notch activity) with INCB3619 resulted in a dose-dependent decrease in luciferase output (Fig. 1A). There was a >60% decrease in Notch activity with INCB3619 [25 μM] (Fig. 1A). We next treated 2 previously characterized human-derived GBM stem cell lines (0308 and 0822) bearing CBF1-luciferase with INCB3619 or DAPT, alone or in combination, and measured Notch activity. There was a significant decrease in Notch activity in both 0308 and 0822 cells treated with INCB3619, DAPT, and a combination of the two drugs at a concentration per drug of 12.5 μM (Fig. 1B–D). We chose the 25-μM concentration for all subsequent investigations.

Fig. 1.

CBF1 activity decreases in response to treatment of GBM cells with ASI or GSI. CBF1-luciferase reporter activity in response to increasing doses of ASI (INCB3619) or vehicle (v:v DMSO) in U251 cells (A). CBF1-luciferase reporter activity in response to 25 μM ASI or GSI (DAPT) or 12.5 μM of both ASI and GSI, in 0308 cells (B) or 0822 cells (C). Endogenous NICD/full-length Notch protein levels in response to ASI, compared with alpha-tubulin (D). CBF1-luciferase reporter activity in response to 25 μM ASI or DAPT in comparison with DMSO (v:v) in U251 cells (E). *P < .05, **P < .01, ***P < .001, ****P < .0001. Black bars represent DMSO treatment; gray bars represent drug treatment.

To further confirm the effect of INCB3619 on Notch inhibition, 0308 cells were treated, and endogenous Notch activation was assayed by immunoblot. There was a significant inhibition of the conversion of Notch-1 from the inactive full-length form to the active NICD, with an accompanying increase in full-length Notch-1 (Fig. 1D). This indicates that INCB3619 inhibits Notch activation and activity. Interestingly, we found a significant decrease in Notch activity by the ASI but not the GSI in U251 cells (Fig. 1E). This suggests that ASIs such as INCB3619 can be better inhibitors of the Notch pathway than GSIs such as DAPT in some GBM cells or under certain conditions.

Altered Expression of Notch Targets in GBM Stem Cells Treated with an ASI or GSI: New Targets of ASIs and GSIs Are Suggested Through Microarray Analysis

We used microarrays to assess global changes in gene expression downstream of an ASI and GSI. GBM 0308 stem cells were treated with 25 μM INCB3619, DAPT, or DMSO vehicle (v:v); samples were prepared and the mRNA fold-change was calculated as described in the Methods (see Supplementary Materials for raw microarray data). Data on selected mRNAs that statistically changed in fold copy number >1.5 in response to INCB3619 or DAPT vs DMSO are given in Figure 2. Most targets of either drug demonstrated decreased mRNA transcript levels, as shown in the microarray analysis. Among the mRNAs that decreased the most compared with DMSO were well-established Notch targets such as HES1 (INCB3619: –1.62, P < .00001), HEY1 (INCB3619: − 3.89, P < .00001, DAPT: − 5.61, P < .00001), and HES5 (INCB3 619: − 7.32, P < .00001, DAPT: − 35, P < .00001). Potential new secretase targets were also identified, such as CHI3L1/YKL-40 (INCB3619: − 1.62, P < .01, DAPT: − 1.99, P < .00001), a secreted protein critical for the inflammatory response in numerous diseases and an indicator of poor cancer prognosis; and LIF (INCB3619: − 1.63, P < .00001, DAPT: − 2.30, P < .00001). Other targets negatively affected by both INCB3619 and DAPT included S100A8 (INCB3619: − 2.32, P < .00001, DAPT: − 4.69, P < .00001), S100 A9 (INCB3619: − 2.45, P < .00001, DAPT: − 2.37, P < .00001), IL-33 (INCB3619: − 2.98, P < .00001, DAPT: − 2.36, P < .00001), and CABLES1 (INCB3619: − 3.25, P < .00001, DAPT: − 2.49, P < .00001). Several mRNAs increased in both INCB3619- and DAPT-treated cells in comparison with DMSO treatment: ASCL1, known to be suppressed by Notch via HES1 (INCB3619: 1.69, P = .002, DAPT: 2.84, P < .00001); DLL1, a Notch ligand (INCB3619: 1.95, P = .0005; DAPT: 2.73, P = .00003); BMF (INCB3619: 1.88, P = 0.0001, DAPT: 1.84, P < .00001); and GAD1 (INCB3619: 1.84, P = .0001, DAPT: 2.56, P < .00001). Given the diversity of putative secretase targets identified in this screen, it is uncertain how many of them are directly downstream of Notch activation, but it is likely that several of the newly identified targets are either directly downstream of Notch or are downregulated via secondary mechanisms associated with Notch inhibition.

ASI and GSI Targets Are Confirmed in Treated GBM Stem Cells: Notch Binds to the YKL-40 and LIF Promoters

Given that microarray data are not quantitatively reliable, we sought to confirm the gene expression changes seen with secretase inhibitor treatment using qRT-PCR. Adhered 0308 or 0822 GBM stem cells were treated once with INCB3619 [25 μM], DAPT [25 μM], or DMSO vehicle (v:v). RNA was harvested 48 hours later and processed as described above. INCB3619 and DAPT treatment decreased HES1 and HEY1 expression in both 0308 and 0822 cells (Fig. 3A and B). 0308 cells treated with DAPT or INCB3619 also showed downregulated expression of the proposed secretase target YKL-40/CHI3L1. YKL-40 was not detected in 0822 cells by any method (Fig. 3D). Another potential target of ASI/GSI, LIF, was downregulated in both 0308 and 0822 INCB3619 and DAPT treatment groups compared with DMSO vehicle (Fig. 3C and D).

Fig. 3.

qPCR of 0308 and 0822 cells treated with 25 μM ASI or GSI; ChIP of NICD binding to YKL-40 and LIF promoters. Predicted targets of ASI/GSI treatment HES1 (black bars) and HEY1 (gray bars) were quantified in response to 25 μM ASI or GSI, and log-scale fold expression changes in comparison with v:v DMSO are shown (A and B). Novel targets of ASI/GSI treatment YKL-40 and LIF were quantified in response to 25 μM ASI or GSI, and log-scale fold expression changes in comparison with v:v DMSO are shown (C and D). Treatment of 0822 cells did not yield an appreciable difference in YKL-40 transcript levels because the transcript was not detectable in DMSO-treated cells (D). DNA sequences pulled down with alpha-mN1a antibody in 0308 and 0822 cells were amplified via qPCR. Antibody-positive (black bars) and mock (gray bars) samples were normalized and quantified using the percent input calculation as described in the Methods. Putative NICD binding sites in the YKL-40 promoter (E) and the LIF promoter (F and G) were compared with control sites within the promoter as well as with RNase P or 18S rRNA amplification.

We identified several putative NICD1 binding domain sites (tgggaa or tggggg) in the promoter regions of both YKL-40 and LIF (see Supplementary Table 1). While these are not the most well-known canonical NICD binding sequences (t/c-gtg-g/a-gaa-a/c),^31,32 they have been shown to have the potential to bind NICD.^33,34 To test whether YKL-40 and LIF can be directly regulated by activated Notch, we performed ChIP analysis of both overexpressed NICD1-Myc (data not shown) and endogenous NICD1(Fig. 3E and F) and analyzed the potential binding domain by qPCR. The YKL-40 promoter region was pulled down by anti-Notch mN1a antibody in the samples, but only 1 putative NICD binding site was strongly positive (the site bounded by the YKL-40-3 primer set) (Fig. 3E). These results indicate that YKL-40 is bound by NICD, but the interaction may be transient or involve only a single site. In contrast, the same pull-down in both 0308 and 0822 cells revealed that LIF is likely very strongly bound by NICD at multiple sites in the promoter. The internal LIF–control promoter control primers were not amplified by the NICD pull-down sample, though they were amplified with high efficiency in the input genomic DNA, indicating specificity of the NICD binding-site interaction (Fig. 3F and G).

ASIs and GSIs Decreases GBM Cell Growth, Which Can Be Partially Rescued by Restoration of Notch Activity

Given the evidence that INCB3619 could inhibit Notch activation, and the previously shown sensitivity of GBM cells to Notch inhibition,^5,23 we hypothesized that INCB3619 treatment of GBM stem cells would inhibit cellular growth in culture. To test this hypothesis, adherent GBM stem cells were treated with DMSO (v:v), INCB3619 [25 μM], DAPT [25 μM], or both INCB3619 and DAPT [12.5 μM:12.5 μM] for a 4- or 6-day period and then analyzed as shown. Total number of 0308 cells was decreased in all treatment groups compared with DMSO vehicle, and 0822 cell numbers were also decreased by both INCB3619 and DAPT treatment (Fig. 4A and B). Flow cytometry analysis of treated vs vehicle groups in 0308 cells revealed a statistically significant accumulation of cells in G1- and a decrease in S-phase cells (Fig. 4C).

Fig. 4.

Cell-growth analysis of 0308 and 0822 cells treated with 25 μM ASI (INCB3619) or GSI (DAPT). 0308 or 0822 cells were treated with DMSO (v:v), ASI, or GSI, and cell numbers were evaluated after 6 days (A). 0308 cells were grown for 6 days in medium containing DMSO, ASI, GSI, or a combination of the drugs, and cell-cycle analysis was carried out using flow cytometry. Black bars indicate G1-phase cells, gray bars indicate G2-phase cells, and white bars indicate S-phase cells (B). 0308 and 0822 cells were treated with DMSO (v:v) or 25 μM ASI or GSI, and a BrdU ELISA was used to assess cell proliferation, with a quantitative readout at absorbance 390 nM (C). 0308 cells transfected with NICD1 and 2 plasmids or TOPO control vector were treated with DMSO (v:v) (black bars) or ASI 25 μM (gray bars), and cell metabolism was evaluated with alamar blue reagent (ex. 544 nM, em. 590 nM) (D). *P < .05, **P < .01, ***P < .001, ****P < .0001.

A BrdU-incorporation ELISA confirmed that cell proliferation was decreased in treatment groups compared with DMSO vehicle in both 0308 and 0822 cells, but INCB3619 treatment was more effective than DAPT in this regard (Fig. 4D and E). To show that Notch inhibition was an important mediator of INCB3619 decreases in cell growth, we performed a rescue experiment by transfecting cells with activated NICDs. NICD1- and NICD2-transfected 0308 cells (vs empty vector) were treated with either INCB3619 or DMSO. This experiment revealed that NICD1 and 2 transfection could substantially rescue the decrease in 0308 cell viability seen with INCB3619 (as seen in an alamar blue assay, similar to an assay by MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide)), supporting the hypothesis that decreases in cell growth observed with INCB3619 are to a large degree dependent upon Notch inhibition (Fig. 4F).

ASIs but not GSIs Increase Expression of the Cell-Cycle Modulators p21 and p53

We show in Figure 4 that treatment of GBM stem cells with INCB3619 significantly reduced their progression from G1 to S phase. It is well-known that p53 and its target p21 are cell-cycle inhibitors³⁵—we therefore tested the effects of INCB3619 on p53 and p21 expression. Treatment of both 0822 and 0308 cell lines with INCB3619 induced a strong expression of both p21 and p53 proteins compared with DMSO-treated cells (Fig. 5A and B). Notably, DAPT treatment had little effect on both p21 and p53 protein expression (Fig. 5A and B).

Fig. 5.

p21 and p53 levels in response to ASI and GSI treatment of 0308 and 0822 cells. 0822 cells were treated with INCB3619, DAPT (25 μM), or vehicle (DMSO v:v). Immunoblots of p53 (left) and p21 (right) are shown in comparison with α-tubulin loading controls (A). 0308 cells were treated with 25 μM (left) or 100 μM (right) ASI, GSI, or DMSO v:v, and p21 and p53 immunoblots are shown in comparison with α-tubulin loading controls (B).

ASI Treatment Prolongs Survival of GBM-Bearing Mice

To determine the potential of an ASI for therapy of GBM in vivo, we used an orthotopic mouse model with a human GBM stem cell line. We stereotactically injected 50 000 0308-CBF1-luciferase GBM stem cells into the brains of immunocompromised mice and waited for 1 week to allow tumors to become established. We then performed local delivery of drug-loaded MLPs containing the ASI INCB3619 or DMSO vehicle. CED was employed to increase delivery volume, and mannitol was included in the infusate to further boost this volume. MRI of established tumors 4 weeks after implantation and 3 weeks after treatment revealed statistically less tumor volume in INCB3619-treated mice compared with controls. Ball-like masses are not initially formed by 0308 tumors, as is common with adherent GBM lines, but 0308 tumors rather invade throughout the brain and resorb the calvarial bone at the site of tumor injection (Fig. 6A and B). Isolated intracranial tumors from INCB3619-treated mice also demonstrated lower Notch-driven CBF1-luciferase activity (Fig. 6C). A separate set of mice were injected with seventy-five hundred 0308 GBM stem cells, and then INCB3619, DAPT, or vehicle containing ferromagnetic nanoparticles were delivered locally using CED. Notably, a single injection of ASI-loaded nanoparticles significantly prolonged survival vs GSI- or vehicle-loaded nanoparticles (Fig. 6D).

Fig. 6.

Local nanoparticle delivery of ASI significantly prolongs SCID/NCr mouse survival in a GBM stem cell line orthotopic model. We stereotactically infused INCB3619 or vehicle liposomal particles into the region of the caudate/putamen under CED conditions into 10 mice per group with established 0308-CBF1- luciferase tumors (50 000 cells). MRI revealed less tumor volume in INCB3619-treated mice (A and B). Isolated intracranial tumors from INCB3619-treated mice also displayed lower CBF1-luciferase activity (normalized by protein content) (C). The same conditions were employed for a single infusion of drug-loaded ferromagnetic nanoparticles 1 week after 7500 GBM 0308 stem cells were injected into SCID/NCr Balb/C mice (7 mice per group). A Kaplan–Meier plot is shown for survival with the ASI-loaded, GSI-loaded, and control nanoparticles (B).

Discussion

We show here by multiple methods that the ASI INCB3619 directly inhibits the Notch pathway and its targets in GBM stem cells. Furthermore, the inhibition of Notch and its targets decreases GBM stem cell growth in culture and in vivo, and ASI treatment prolongs survival of GBM-bearing mice. While another group has recently shown that ASI treatment of lung carcinoma cell lines in vitro and in nude mice results in decreased tumor growth via Notch-1 inhibition,²¹ this is the first description of an ASI inhibiting growth of tumor stem cells. Notch and Notch ligands were identified by The Cancer Genome Atlas study of over 200 human GBMs as markers of the classical and proneural subtypes of human GBM.³⁶ Given that INCB3619 targets Notch activity and that Notch is a marker of multiple GBM subtypes, the fact that INCB3619 decreases growth of several GBM stem cells as well as adherent GBM lines in response to decreased Notch activity is likely to translate well into the clinic.

This work also suggests several new targets of both ASIs and GSIs in GBM stem cells that are important for tumor growth, including CHI3L1/YKL-40 and LIF. CHI3L1/YKL-40 expression is an indicator of poor prognosis in many cancers and inflammatory diseases. Based on our findings, serum YKL-40 levels could represent a potential pharmacodynamic biomarker for secretase inhibitors.^37–42 Previous work has determined that adherent GBM lines treated in vitro with etoposide or serum depletion exhibit increased YKL-40 expression, whereas YKL-40 expression is inhibited by fibroblast growth factor and tumor necrosis factor–α.⁴⁰ This is the first description of a small molecular pharmacologic reagent that decreases YKL-40 expression. YKL-40 has been shown to be attenuated in the presence of p53, so the increase in p53 caused by ASI treatment could be a possible indirect mechanism for the observed decrease in YKL-40.⁴⁰ However, since we also show that NICD binds the YKL-40 promoter, a direct relationship between ASI Notch inhibition and YKL-40 expression is likely. YKL-40 expression was identified by The Cancer Genome Atlas to be a marker for the mesenchymal subtype of GBM, and it is therefore significant that we identified Notch as a driver of YKL-40.³⁶ LIF is a member of the interleukin-6 family of cytokines; it is highly expressed in some GBMs.⁴³ LIF induces GBM stem cell self-renewal and tumorigenicity.⁴⁴ In addition, loss of function of LIF decreases GBM growth.⁴⁵ There is an available inhibitor that binds to the LIF receptor, but this is the first description of a reagent that decreases LIF expression.⁴⁶ This is also the first description of LIF as a direct target of Notch. The binding of the LIF promoter by NICD at 2 putative sites is likely to be physiologically relevant. Regulation of LIF mRNA expression by Notch also has implications for cell fate determination in the nervous system, as both are known to drive glial cell fate, and our data suggest cooperation between LIF and Notch.^47–49

There were several advantages of the ASI INCB3619 over the GSI DAPT. Notably, INCB3619 inhibits GBM cell growth more consistently than does DAPT in more GBM types. This is likely due to the fact that in addition to Notch inhibition, INCB3619 treatment increased p21 and p53 expression relative to DAPT treatment of GBM stem cells, and p21 and p53 are important cell-cycle inhibitors.⁵⁰ Mutations of p53 are common in GBM,⁵¹ but tumors that express wild-type p53 may be more susceptible to treatment with INCB3619 than are p53-mutant tumors. These results suggest that an ASI such as INCB3619 may be a more effective reagent than a GSI in GBM therapy.

A previous study demonstrated the in vivo efficacy of systemic INCB3619 in a subcutaneous lung cancer model.²⁰ Although this is encouraging for clinical application of this ASI, the systemic applicability of this agent for brain tumor treatment is untested. We elected to use local CED to bypass potential problems with systemic delivery. Additionally, we overcame solubility issues by incorporating INCB3619 into lipoparticles or MLs. Using this approach, we were able to achieve high local concentrations of the drug, prevent drug deposit formation (a common problem for local injection of poorly soluble drugs), and allow for MRI visualization of tumor volume. It is notable that the ASI-loaded lipoparticles decreased tumor growth, and the ASI-loaded ferromagnetic nanoparticles yielded a significant survival benefit after a single application in a challenging xenograft model. We speculate that the functionalized drug remains pharmacologically active for a prolonged period of time and that nanoparticle-immobilized ASIs may have favorable bio- and neurocompatibility. We are hopeful that future studies will demonstrate the utility of systemic ASIs for brain tumors, but it is also possible that local delivery of ASI-loaded nanoparticles may be used for targeted therapy of residual tumor. Local delivery also allows for the avoidance of potential side effects of systemic ASIs. Further studies of the ASI-loaded nanoparticles and their delivery are needed to explore their safety and therapeutic potential.

Our study of INCB3619 reveals novel targets of an important reagent that may be applicable for GBM therapy in a clinical setting. INCB3619 has already been shown to decrease Michigan Cancer Foundation–7 breast cancer cell growth in combination with lapatinib.¹⁹ This combination therapy was shown to target known breast cancer pathways, but we demonstrate here that INCB3619 also targets the Notch pathway and increases p21 and p53 expression. Evaluation of specific tumors for Notch activity in combination with wild-type p53 expression may lead to recommendations for INCB3619 treatment either alone or in combination.

Supplementary Material

Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).

Conflict of interest statement. None declared.

Funding

National Institutes of Health (R01CA134768 to BP).