Estrogen-related receptor β activation and isoform shifting by cdc2-like kinase inhibition restricts migration and intracranial tumor growth in glioblastoma

Abstract

Glioblastoma (GBM; grade 4 glioma) is a highly aggressive and incurable tumor. GBM has recently been characterized as highly dependent on alternative splicing, a critical driver of tumor heterogeneity and plasticity. Estrogen-related receptor β (ERR-β) is an orphan nuclear receptor expressed in the brain, where alternative splicing of the 3′ end of the pre-mRNA leads to the production of 3 validated ERR-β protein products: ERR-β short form (ERR-βsf), ERR-β2, and ERR-β exon 10 deleted. Our prior studies have shown the ERR-β2 isoform to play a role in G₂/M cell cycle arrest and induction of apoptosis, in contrast to the function of the shorter ERR-βsf isoform in senescence and G₁ cell cycle arrest. In this study, we sought to better define the role of the proapoptotic ERR-β2 isoform in GBM. We show that the ERR-β2 isoform is located not only in the nucleus but also in the cytoplasm. ERR-β2 suppresses GBM cell migration and interacts with the actin nucleation-promoting factor cortactin, and an ERR-β agonist is able to remodel the actin cytoskeleton and similarly suppress GBM cell migration. We further show that inhibition of the splicing regulatory cdc2-like kinases in combination with an ERR-β agonist shifts isoform expression in favor of ERR-β2 and potentiates inhibition of growth and migration in GBM cells and intracranial tumors.—Tiek, D. M., Khatib, S. A., Trepicchio, C. J., Heckler, M. M., Divekar, S. D., Sarkaria, J. N., Glasgow, E., Riggins, R. B. Estrogen-related receptor β activation and isoform shifting by cdc2-like kinase inhibition restricts migration and intracranial tumor growth in glioblastoma.

Glioblastoma (GBM; grade 4 glioma) is a highly aggressive tumor, incurable, and markedly resistant to most systemic chemotherapies. Surgery, radiation, and adjuvant treatment with the DNA alkylating/methylating agent temozolomide (TMZ) are the current first-line standard of care. However, development of TMZ-resistance is rapid, which is in part due to re-expression of O6-methylguanine methyltransferase (MGMT), the DNA repair enzyme that removes TMZ-induced O6-methylguanine DNA adducts. Median survival is only ∼14–16 mo with TMZ, and after resistance has developed, there is no established second-line regimen (1, 2). This is due in part to our limited understanding of the molecular drivers of GBM and the inherent challenge of developing effective therapeutic strategies that penetrate the blood–brain barrier (BBB). GBM is also a heterogeneous and invasive tumor that disseminates into healthy brain tissue. Even when the bulk of the tumor is surgically resected, chemotherapeutic options to target residual disease are limited by the BBB (3), which has severely limited repurposing of drugs from other cancers for GBM.

Dysregulation of alternative mRNA splicing is one potential mechanistic driver of GBM (4), given that the brain contains the most alternatively spliced transcripts of any organ and the expression of many splicing factors is up-regulated in GBM vs. normal brain (5, 6). Serine/arginine rich (SR) proteins are a prominent group of splicing regulatory factors that are phosphorylated and thereby regulated by the cdc2-like kinases (CLKs) (7), some of which have been mechanistically implicated in GBM (8). Although CLK inhibitors have not yet entered clinical trials, preclinical studies of TG-003 (a pan-CLK inhibitor) show that this agent can cross the BBB in mouse models of autism (9). Ongoing clinical trials are testing first-generation splicing regulatory drugs, such as H3B 8800 for myelodysplastic syndromes, acute myeloid leukemia, and chronic myelomonocytic leukemia (10).

Given that improved therapeutic options are an urgent clinical need for GBM, the nuclear receptor superfamily (members of which are highly successful targets in breast and prostate cancers) provides another novel target strategy. Estrogen-related receptor β (ERR-β) [ERR-β gene (ESRRB), also known as nuclear receptor subfamily 3, group B, member 2] is the founding orphan member of the nuclear receptor superfamily (11). By definition, orphan nuclear receptors lack known endogenous ligands, though their function can be modified by coregulatory proteins or synthetic ligands that increase or decrease their transcription factor activity (12, 13). DY131 is one synthetic agonist that has been shown to enhance ERR-β transcription factor activity in the murine arcuate nucleus (14, 15), which is strongly suggestive of BBB penetrance. ERR-β is alternatively spliced at the 3′ end, leading to the production of ERR-β short form (ERR-βsf), ERR-β2, and ERR-β exon 10 deleted (ERR-β-∆10) [Fig. 1A (16)]. These 3′ splicing events are unique to primates, with all lower vertebrate organisms containing genomic sequences for only the ERR-βsf isoform (16). Inclusion of additional 3′ exons in ERR-β2 and ERR-β-Δ10 produces 67- and 75-aa carboxyl-terminal extensions, or F domains, which can modify transcription factor function and recruit distinct coregulatory proteins (17).

Figure 1.

A) Expression of ESRRB correlates with GBM survival. Normalized RNAseq data (fragments per kilobase of transcript per million mapped reads upper quartile) from the GBM The Cancer Genome Atlas (TCGA) dataset were downloaded from XenaBrowser. Log-rank survival analysis was performed for upper vs. lower quartile of ESRRB expression. P = 0.0332. B) Differential splicing at the 3′ end of the ESRRB pre-mRNA leads to the production of 3 known ERR-β transcripts and protein products: the short form (ERR-βsf) and 2 longer forms, ERR-β2 and ERR-β-∆10. The ERR-βsf isoform is conserved in zebrafish and mice, with percent identity for each ortholog compared with the human sequence. AF-1, activation function-1. C) ERR-β2 and ERR-βsf proteins are expressed in primary NHAs as well as in multiple TMZ-sensitive (8MGBA, 42MGBA) and -resistant (8MGBA-TMZres, 42MGBA-TMZres, T98G) GBM cell lines. MGMT, a known marker of therapy resistance in GBM, is heavily expressed in the resistant cell lines. D) DY131, a small molecule agonist of ERR-β, inhibits cellular proliferation in TMZ-sensitive and -resistant cell lines but not in NHAs or normal human oligodendrocytes. Data in the growth curves are presented as means (point) ± sd (error bars), and data in the inset are presented as means (line) ± minimum/maximum values (box) for 3 (NHA), 6 (oligodendrocytes, T98G), or 8 (all other cell lines) technical replicates, which were analyzed by 2-way ANOVA with post hoc Dunnett’s multiple comparisons test. Data are representative of at least 2 independent biologic replicates. **P < 0.01, ***P < 0.0001, vs. DMSO control. E) Expression of ERR-β2 and ERR-βsf vary across multiple PDXs of GBM, as does wild-type (wt) and vIII mutant EGFR, the DNA damage marker PARP, and the cyclin-dependent kinase inhibitor p21. Lanes labeled β2 and βsf denote lysate from cells transfected with the indicated cDNA construct.

We previously identified the ERR-β2 isoform as a cytoplasmic and centrosome-adjacent protein (18) and showed that activation of this isoform can delay mitosis and partially repress the transcription factor activity of the ERR-βsf isoform (19). The goal of this study was to better define the function and interacting partners of the proapoptotic ERR-β2 isoform in GBM and test the ability of splicing kinase inhibitors to shift isoform balance toward the ERR-β2 isoform. We found that the cytoplasmic ERR-β2 isoform suppresses GBM cell migration and interacts with the actin nucleation-promoting factor cortactin, and that an ERR-β agonist remodels the actin cytoskeleton and also suppresses migration. CLK inhibition with TG-003 in combination with the ERR-β agonist DY131 shifts isoform expression in favor of ERR-β2 and leads to suppression of growth and migration in TMZ-resistant GBM cells. Finally, we use a novel zebrafish model to show that the combination of TG-003 and DY131 has antitumor activity in vivo in a setting in which the BBB is intact.

MATERIALS AND METHODS

Cell lines and culturing conditions

Primary normal human astrocytes (NHAs) were purchased from Lonza (CC-2565; Basel, Switzerland). Immortalized human oligodendrocyte MO3.13 cells were a kind gift from Dr. Alexandra Taraboletti [Lombardi Comprehensive Cancer Center (LCCC)]. TMZ-sensitive 42MGBA and 8MGBA cell lines were provided by Dr. Jeffrey Toretsky (LCCC), and the de novo TMZ-resistant T98G cell line was provided by Dr. Todd Waldman (LCCC). Acquired TMZ-resistant 42MGBA-TMZres and 8MGBA-TMZres cell line variants were developed by our laboratory and previously described (20). All cells tested negative for Mycoplasma contamination and were maintained in a humidified incubator with 95% air and 5% carbon dioxide. All cell lines were fingerprinted by the LCCC Tissue Culture Shared Resource to verify their authenticity using the standard 9 short tandem repeat loci and Y-specific amelogenin. Both the 42MGBA-TMZres and 8MGBA-TMZres variants are documented to be of the same origin as their respective parental cell lines. NHAs were used within 1 passage and maintained in astrocyte growth medium (CC-3187; Lonza) supplemented with l-glutamine, gentamicin sulfate, ascorbic acid, human epidermal growth factor, insulin, and 3% fetal bovine serum (FBS) (CC-4123; Lonza). MO3.13, 42MGBA, 8MGBA, 42MGBA-TMZres, and T98G cells were grown in DMEM (high glucose, 11965092; Thermo Fisher Scientific, Waltham, MA, USA) with 10% FBS. The 8MGBA-TMZres cells were grown in DMEM with 10% FBS and 100 µM TMZ. TMZ (S1237; Selleckchem, Houston, TX, USA) was dissolved in DMSO (D8418; Millipore-Sigma, Burlington, MA, USA) to 130 mM and used at the concentrations indicated. DY131 (2266; Tocris Bioscience, Bristol, United Kingdom) was dissolved in DMSO to 10 mM and used at the concentrations indicated.

Western blotting

Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (4906837001; Roche, Basel, Switzerland) for protein extractions and separated by PAGE using 4–12% gradient gels (NP0321BOX; Novex by Thermo Fisher Scientific) as previously described (19). They were then transferred onto nitrocellulose membranes (IB23001; Thermo Fisher Scientific) with the iBlot2 (IB21001; Thermo Fisher Scientific) and probed with the following antibodies: ERR-β2 (1:500, PP-H6707-00; R&D Systems, Minneapolis, MN, USA), ERR-βsf (1:1000, PP-H6705-00; R&D Systems), poly ADP ribose polymerase (PARP) (1:1000, 9542L; Cell Signaling Technology, Danvers, MA, USA), cyclin-dependent kinase inhibitor 1A/p21 (1:300, sc-756; Santa Cruz Biotechnology, Dallas, TX, USA), epidermal growth factor receptor (EGFR) (1:1000, 2232S; Cell Signaling Technology), MGMT (1:1000, 2739S; Cell Signaling Technology), cortactin (1:1000, 05-180; Upstate, Lake Placid, NY, USA), phosphorylated (p)SR proteins (clone 1H4, 1:500, MABE50; MilliporeSigma), total SR proteins (1:1000, MABE126; MilliporeSigma), phosphorylated histone H3 serine 10 (1:1000, 3377S; Cell Signaling Technology), and total histone H3 (1:1000, 9715S; Cell Signaling Technology). β-Tubulin (1:5000, T7816; MilliporeSigma) and β-actin (1:5000, A5316; MilliporeSigma) were used as loading controls. Proteins were detected with horseradish peroxidase–conjugated secondary antibodies [1:5000, NA931-1ML (Mouse) or NA934-1ML (Rabbit); GE Healthcare, Waukesha, WI, USA] and enhanced chemiluminescent detection HyGlo Quick Spray Chemiluminescent (E2400; Denville Scientific, Holliston, MA, USA) using film (E3212; Denville Scientific).

Immunofluorescence

Cells were seeded at a density of 40,000–50,000 cells onto 18-mm-diameter #1.5 round coverslips (101413-518; VWR, Radnor, PA, USA) in 12-well dishes. On the following day, the medium was removed, and cells were fixed and permeabilized in 3.2% paraformaldehyde with 0.2% Triton X-100 in PBS for 5 min at room temperature. Three washes were performed with PBS in the 12-well plate, and then coverslips were inverted onto 120 μl of primary antibody in the antibody block (0.1% gelatin with 10% normal donkey serum in PBS) on strips of parafilm and incubated for 1 h. Coverslips were first incubated with either ERR-β2 (1:150) or ERR-βsf (1:200) for 1 h. After incubation with primary antibodies, coverslips were washed 3 times with PBS. Then, coverslips were inverted onto 100 µl of antibody block with secondary antibodies (Alexa Fluor 488 anti-mouse, 1:200, A11029; Thermo Fisher Scientific) and DAPI (DNA, 1:500 dilution) for 20 min in the dark. Coverslips were again washed 3 times with PBS and then gently dipped 4 times into molecular biology-grade water before inversion onto 1 drop of Fluoro-Gel (with N-[tris(hydroxymethyl)methyl]-2-aminoethanesulfonic acid buffer, 17985-30; Electron Microscopy Sciences, Hatfield, PA, USA) and allowed to air-dry in the dark for at least 10 min. Slides were stored at 4°C until image collection on the LCCC Microscopy and Imaging Shared Resource’s Leica SP8 microscope (Leica Microsystems, Buffalo Grove, IL, USA) with the ×63 oil objective.

Cell growth assays

Cells were seeded in 96-well plastic tissue culture plates at 1000 cells/well 1 d prior to treatment with the indicated concentrations of DY131. Cells were treated for a total of 8–10 d, with medium changed and drug replenished on d 4 or 5. Staining with crystal violet, resolubilization, and analysis of staining intensity as a proxy for cell number was performed as previously described (19).

Immunohistochemistry on human GBM tumor samples

Immunohistochemical staining of GBM human tumor samples was performed for ERR-β2 or ERR-βsf. Sections (5 μm) from formalin-fixed paraffin-embedded tissues were deparaffinized with xylenes and rehydrated through a graded alcohol series.

For ERR-β2, heat-induced epitope retrieval was performed by immersing the tissue sections at 98°C for 20 min in 10 mM citrate buffer (pH 6.0) with 0.05% Tween. Immunohistochemical staining was performed using a horseradish peroxidase–labeled polymer from Agilent Technologies (K4001, K4003; Santa Clara, CA, USA) according to the manufacturer’s instructions. Briefly, slides were treated with 3% hydrogen peroxide and 10% normal goat serum for 10 min each and exposed to primary antibody at a 1:125 dilution overnight at 4°C. Slides were exposed to the appropriate horseradish peroxidase–labeled polymer for 30 min and 3,3′-diaminobenzidine chromagen (Agilent Technologies) for 5 min.

For ERR-βsf, heat-induced epitope retrieval was performed by immersing the tissue sections at 98°C for 20 min in 110 mM Tris, 1 mM EDTA pH 9.0 buffer (Genemed, San Francisco, CA, USA). Immunohistochemical staining was performed using the VectaStain Kit from Vector Laboratories (Burlingame, CA, USA) according to the manufacturer’s instructions. Briefly, slides were treated with 3% hydrogen peroxide, avidin/biotin blocking, and 10% normal goat serum and exposed to primary antibody at a 1:200 dilution overnight at 4°C. Slides were exposed to appropriate biotin-conjugated secondary antibodies (Vector Laboratories), Vectastain avidin–biotin complex reagent, and 3,3′-diaminobenzidine chromagen (Agilent Technologies).

For both primary antibodies, slides were counterstained with hematoxylin (Harris Modified Hematoxylin, Thermo Fisher Scientific), blued in 1% ammonium hydroxide, dehydrated, and mounted with Acrymount. Consecutive sections with the primary antibody omitted were used as negative controls.

ERR-β knockdown

Short hairpin RNA (shRNA)-mediated silencing of ERR-β2 and ERR-βsf in T98G GBM cells was previously reported (19). shRNA sequences targeting ESRRB are as follows: shESRRB-1: 5′-TGAGGACTACATCATGGAT-3′; shESRRB-2: 5′-TGCAGCACTTCTATAGCGT-3′.

Migration assays

For scratch-wound assays, cells were plated at 150,000–200,000 cells/well depending on the cell line and allowed 48 h to create a monolayer. After monolayer formation, a P200 tip was used to make a scratch, and images were taken at 0, 24, 48, and 72 h time points. Analysis was done in ImageJ (National Instiutes of Health, Bethesda, MD, USA) (21) to determine percent closed, with 0% being at 0 h. For transwell migration assays, cells were cultured in low-serum conditions (0.5% FBS) overnight. The following day, 100,000 cells in 200 µl of 0.5% DMEM were loaded into the top of a modified Boyden chamber (24-well, 8-µm cell culture insert, 353097; Becton Dickinson, San Diego, CA, USA), which was then placed in a 24-well plate containing 500 µl of DMEM supplemented with full serum (10% FBS). Cells were allowed to migrate for 4 and 24 h at 37°C, at which time the nonmigratory cells were removed from the top of the membrane with cotton swabs, and the bottom of the membrane was stained with crystal violet as described for cell growth assays. After staining, the membranes were excised from the insert, mounted on glass slides using Fluoro-Gel, and allowed to air-dry. Cells per field at ×20 magnification were counted by light microscopy.

Immunoprecipitation

On d 0, cells were seeded at 200,000 cells/well in 6-well dishes. The next morning, on d 1 and 2, separate tubes were made and then combined for the transfection. Tube #1 contained 2.5 µg plasmid DNA (psg5, ERR-β2, ERR-βsf, or ERR-β-∆10), 2.5 µl Plus reagent (from Lipofectamine LTX, 15338100; Thermo Fisher Scientific), and Opti-MEM (31985070; Thermo Fisher Scientific), which brings the total volume to 50 µl. Tube #2 contained 500 µl Opti-MEM and 6.25 µl Lipofectamine LTX. Tube #1 was then added to tube #2 and pipetted up and down a few times to mix thoroughly. The tube #2 with all reagents was left untouched for 25–30 min. Before transfecting, medium was changed to serum-free DMEM. Then, the transfection reagent was slowly taken up in a P1000 tip and added drop-wise to the cells. At 4–6 h later, medium was changed back to 10% DMEM overnight. The next morning (d 2), medium was changed again with fresh 10% DMEM. In the afternoon of d 2, cells were lysed in RIPA for protein extraction. Protein was quantified, and amounts were normalized between all samples to be used in the immunoprecipitation (IP). Total volume for each IP was brought up to 500 µl of RIPA with inhibitors. An aliquot was also set aside for input for each sample. In the evening of d 2, 1 µl of the cortactin antibody (ab81208, 1:500 dilution; Abcam, Cambridge, MA, USA) was added to each sample and allowed to rotate overnight at 4°C. The next morning, protein A/G beads (20421; Pierce, Rockford, IL, USA) were vortexed, and 30 µl of beads were added to an Eppendorf tube per sample. Beads were then washed in cold RIPA buffer with inhibitors and centrifuged at 4°C for 5 min. Supernatant was aspirated, and cold RIPA buffer with inhibitors was added to bring up the volume. Then, 30 µl of washed beads were added to each IP tube and allowed to rotate again at 4°C for 1 h. Tubes were then centrifuged for 5 min at 10,000 rpm at 4°C. Supernatant was aspirated, and 500 µl of cold RIPA with inhibitors was used to wash IPs. Tubes were then centrifuged again for 5 min at 10,000 rpm at 4°C. IPs were washed 2 times more in Tris-saline buffer (50 mM Tris base, pH 7.5, 150 mM NaCl in distilled H₂O). After the last wash, supernatant was aspirated and 30 µl of 2× loading buffer was added. Samples were then vortexed and boiled for 9 min. Samples were then spun down in a centrifuge at room temperature for 1–2 min. Finally, both IP and input samples were loaded onto a 4–12% gradient gel, and the Western blotting protocol previously described was followed.

F-actin flow cytometry

Cells were seeded at 100,000–200,000 cells/well in 6-well plastic tissue culture dishes and then treated with 5 µM DY131 the next day. After 24 h, cells were collected by trypsinization, combined with nonadherent cells in the culture medium, and then washed with PBS prior to fixation in 4% paraformaldehyde for 5 min. Fixed cells were permeabilized using 0.2% Triton X-100 in PBS for 5 min and then blocked in 0.1% gelatin with 10% normal donkey serum in PBS for 30 min at room temperature. Cells were stained in suspension with a 1:300 dilution of Anti-Stain 488 Fluorescent Phalloidin (PHDG1; Cytoskeleton, Denver, CO, USA), gently mixed, and incubated in the dark at room temperature for 30 min. Stained cells were washed twice with PBS, and then 30,000 cells were acquired by flow cytometry on a Becton Dickinson Fortessa. Data were analyzed using FCSExpress 6 (DeNovo Software, Glendale, CA, USA).

Cell cycle analysis

On d 0, cells were seeded at 200,000 cells/well in 6-well plastic tissue culture dishes 1 d prior to treatment with the indicated concentrations of drug. For experiments with TG-003 ± DY131, treatment time was 24 h. After 24 h, cells were collected, washed with PBS, ethanol fixed, stained with propidium iodide, and analyzed for cell subG₁ (fragmented/apoptotic) DNA content and cell cycle profile. In all, 30,000 cells were acquired by flow cytometry on a Becton Dickinson Fortessa. Files were modeled using ModFit software (Verity Software, Topsham, ME, USA) to determine subG₁, G₁, S, and G₂/M cell cycle stage.

Zebrafish intracranial xenografts

The zebrafish model used is a transgenic line with green blood vessels in a double mutant, transparent background with the genotype Tg(kdrl:GRCFP)zn1; mitfab692/b692; ednrb1b140/b140. This line is propagated in house by raising equal numbers of fry from 5 independent group matings, each consisting of 3 females and 3 males, in order to maintain background genetic diversity. Correct genotype is determined by visual inspection; the fish lack all pigment, except for the eyes and the blood vessels, which fluoresce bright green. A new generation is raised biyearly, and the oldest generation is discontinued when they reach 2 yr of age. All procedures were performed in accordance with National Institutes of Health guidelines on the care and use of animals and were approved by the Georgetown University Institutional Animal Care and Use Committee (2017-0078).

Tumor cells were labeled in suspension with a 1:100 dilution of Vybrant CM-1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) cell-labeling solution (V22885; Thermo Fisher Scientific) at a density of 5 × 10⁶ cells/ml for 20 min at 37°C. The labeled cells were washed 5 times in PBS, suspended in PBS at 5 × 10⁷ cells/ml, and then back-loaded into a pulled borosilicate microinjection needle (David Kopf Instruments, Tujunga, CA, USA). The needles were placed in a vertical position with the point facing down for 10 min to allow the cells to settle toward the tip. At 36 h postfertilization embryos were anesthetized in 160 μg/ml buffered tricaine solution (MilliporeSigma). The embryos were loaded onto an injection plate at 37°C and covered with 1.5% low-melting-point agarose (Thermo Fisher Scientific), where they were positioned for injection. The plate was allowed to cool to room temperature and then was covered in fish water (0.3 g/L sea salt) containing 160 mg/ml tricaine. The embryos were injected with 1–2 nl of cell suspension (50–100 cells) into the area of the midbrain-hindbrain border. Following injection, the embryos were manually freed from the agarose and allowed to recover at 28°C for 1 h, followed by incubation at 33°C. At 5 d postfertilization (dpf), tricaine-anesthetized embryos were mounted in 3% methyl cellulose and imaged on an Olympus XI-71 Inverted epifluoresce microscope. Following imaging, the fry were transferred to 24-well plates in 1 ml fish water/well. At 10 dpf, fry were anesthetized with 112 μg/ml tricaine, mounted in 3% methyl cellulose, and imaged on an Olympus XI-71 Inverted epifluorescence microscope or mounted in 1.5% low-melting-point agarose for imaging on a Leica SP8 confocal microscope.

Treatment groups were assigned to achieve similar numbers of larvae containing similar distributions of tumor sizes. The larvae were treated starting at 6 dpf. Final drug concentrations were made using 1:1000 dilution of stock in DMSO. Control larvae were treated with 0.1% DMSO. The drugs were refreshed daily by removing most of the solution, leaving ∼100 μl to keep the fry wet, followed by the addition of 1 ml of fresh drug solution. For BBB imaging experiments, Cascade Blue dextran, 10,000 MW (D1976; Thermo Fisher Scientific) at 25 mg/ml was back-loaded into a pulled borosilicate microinjection needle. At 10 dpf, larvae were anesthetized in 112 μg/ml buffered tricaine solution, loaded onto an injection plate at 37°C, and covered with 1.5% low-melting-point agarose, where they were positioned for injection. Approximately 2 nl of Cascade Blue dextran solution was injected into the common cardinal vein. The injected larvae were then mounted in 1.5% low-melting-point agarose and imaged on a Leica SP8 confocal microscope at 1 h postinjection.

Computational, image, and statistical analysis

Images and figures were compiled using either Photoshop or Illustrator (Adobe Systems, San Jose, CA, USA). Densitometry of the ERR-β2:ERR-βsf ratio was calculated using ImageJ (21). Calculation of zebrafish intracranial xenograft tumor size was carried out in Illustrator as follows: zebrafish images were imported into Illustrator. The red channel was selected, which highlighted the GBM cells as DiI fluoresces red. The magic wand tool was chosen to select the red fluorescent tumor, in which after selection, area was calculated by Illustrator’s measure area function. These measurements were exported, and pre- and posttreatment tumor areas were compared. Motif analysis of ERR-β2 F domain sequences was performed using ScanSite3.0 (22). Statistical analysis and graphing were performed using Prism 8.0 (GraphPad Software, La Jolla, CA, USA), with the following exceptions. Supplemental Figure S1A was generated using RNA sequencing (RNAseq) data from European Molecular Biology Laboratory—European Bioinformatics Institute [dataset E-MTAB-4840 (23, 24)]. Figure 1A was generated using RNAseq data from XenaBrowser (25). Supplemental Figure S3A was generated using RNAseq data from the Chinese Glioma Genome Atlas through the International Cancer Genome Consortium (https://icgc.org), analyzed in R using GGally and ggplot2 packages. Supplemental Figure S3B was generated and analyzed in Prism 8.0 from RNAseq data obtained through the Ivy Glioblastoma Atlas Project (26). SR splicing factor (SRSF) 6 consensus binding sites (27) in ESRRB exon 10 were predicted using RBPmap (28). Data are presented as means ± sd unless otherwise indicated. The number of biologic and technical replicates and specific statistical tests performed are reported in the figure legend for each figure. Statistical significance is defined as α ≤0.05.

RESULTS

The conserved ERR-βsf and primate-specific ERR-β2 isoforms of ERR-β have distinct functions in cell cycle regulation, with activation of ERR-β2 by the synthetic agonist ligand DY131 causing mitotic arrest and apoptosis (19). These mechanistic discoveries demonstrated the ERR-β2 isoform to play a more antitumorigenic role than ERR-βsf. Therefore, the goal of this study was to better define the function and interacting partners of the proapoptotic ERR-β2 isoform in GBM and test the ability of splicing kinase inhibitors to shift isoform balance toward the ERR-β2 isoform.

ERR-β isoforms are expressed in GBM

Nuclear receptors are attractive drug targets, and successful inhibitors to nuclear receptors like estrogen receptor α have fundamentally changed cancer treatment outcomes for patients with breast cancer. However, targeting estrogen receptor α with tamoxifen in GBM has been unsuccessful (29), in part because its expression in the brain is quite low, in contrast to ERR-β (gene symbol ESRRB), an orphan nuclear receptor expressed in the brain (Supplemental Fig. S1A). In The Cancer Genome Atlas, high (upper quartile) ESRRB mRNA expression is significantly associated with longer overall survival (Fig. 1A). Two ERR-β protein isoforms (ERR-β2 and ERR-βsf) are expressed across multiple TMZ-sensitive and -resistant GBM cell lines (20) and primary NHAs (Fig. 1B, C), as well as immortalized human oligodendrocytes (MO3.13, Supplemental Fig. S1B). DY131 is a small molecule agonist of ERR-β that has antiproliferative activity in several preclinical cancer models (18, 30). DY131 is growth inhibitory in GBM cell lines but not NHAs or immortalized oligodendrocytes, in which it shows a slight increase in growth (Fig. 1D). A limitation of conventional GBM cell lines is that they do not always recapitulate essential GBM molecular features or pathobiology (31). Patient-derived xenografts (PDXs) from primary tumors have been established to address this limitation, and we show that ERR-β2 and ERR-βsf proteins are broadly expressed across whole-cell lysates from flank-grown GBM PDX specimens representative of multiple molecular subtypes and clinical/pathologic features (Fig. 1E and Supplemental Table S1). EGFR wild-type and vIII mutant–amplified tumors are more frequently categorized in the classic molecular subtype (32), whereas PARP1 expression is enriched in classic and proneural subtypes (33), and cyclin-dependent kinase inihbitor 1A/p21 expression is indicative of primary vs. secondary GBM (34).

Studies in CV-1 (simian) in origin, and carrying the SV40 genetic material cells transfected with ERR-β isoform cDNA show that ERR-βsf localizes to the nucleus, whereas ERR-β2 is localized to the cytosol and nucleus, suggesting differential functions (16). This pattern of subcellular localization of endogenous ERR-βsf (nucleus) vs. ERR-β2 (cytosol and nucleus) is similarly observed in NHAs, oligodendrocytes, and GBM cells (Fig. 2A and Supplemental Fig. S1C) and in primary tumor specimens (Fig. 2B). Collectively, these data demonstrate that ERR-β isoforms are expressed, and a well-established ERR-β agonist is growth inhibitory in multiple TMZ-sensitive and -resistant GBM models.

Figure 2.

A) Immunofluorescent staining of ERR-β2 and ERR-βsf depicts nuclear localization of ERR-βsf and nuclear and cytoplasmic localization of ERR-β2 in NHAs and TMZ-sensitive and -resistant GBM cell lines. B) Immunohistochemistry of ERR-β2 and ERR-βsf also illustrates broad expression of these 2 isoforms in 3 different primary GBM tumor samples. H&E, hematoxylin and eosin.

The long ERR-β2 isoform suppresses GBM cell migration and interacts with the actin nucleation-promoting factor cortactin

Prior studies implicate ERR-βsf as a transcription factor with activity at multiple DNA response elements but suggest that ERR-β2 has little or no transcription factor activity and may partially repress βsf-mediated gene transcription (16, 18, 19). This knowledge, coupled with the localization of ERR-β2 expression adjacent to centrosomes and diffusely throughout the cytoplasm (Fig. 2A), led us to test whether ERR-β2 might have additional functions in GBM cell migration. Knockdown of ERR-β2, but not ERR-βsf, significantly enhances T98G cell migration in wound healing (Fig. 3A, B) and transwell migration assays (Supplemental Fig. S2).

Figure 3.

Knockdown of ERR-β2, but not ERR-βsf, significantly enhances T98G cell migration, as measured by scratch-wound assay. A) Cells stably transduced with the indicated shRNA are compared with parental cells or those stably transduced with a scrambled shRNA control. Lanes labeled Δ10, β2, and βsf denote lysate from cells transfected with the indicated cDNA construct. B) Data are presented as the median (line) ± minimum/maximum for 3–8 fields of view, in each of at least 2 independent biologic replicates. N.s., not significant. Data were analyzed by 1-way ANOVA with post hoc Tukey’s multiple comparisons test. ***P < 0.001, ****P < 0.0001. C) ERR-β2, but not ERR-βsf or ERR-β-Δ10, forms a complex with endogenous cortactin as analyzed by IP. –, control IP of ERR-β2-transfected cell lysates.

F domains are carboxyl-terminal extensions of the ligand binding domain of nuclear receptors, which can modify transcription factor function and recruit distinct coregulatory proteins (17). Inspection of the 67-aa extended carboxyl-terminal F domain unique to ERR-β2 shows a number of putative protein-protein interaction motifs and post-translational modification sites [Supplemental Table S2, (22)]. Of these, a proline-rich region consisting of aa 467–472 (PLPPPP) forms the core of the consensus binding motif for the Src homology 3 domain of the cytoskeletal protein and actin nucleation-promoting factor cortactin (35, 36). Cortactin mRNA expression is increased with increasing severity of gliomas, including GBM (Supplemental Fig. S3A), and is significantly enriched at the leading edge of tumors [Supplemental Fig. S3B, (26)]. Cortactin has previously been implicated in GBM cell migration (37) and is expressed across all GBM cell lines (Fig. 1B). Transfection of GBM cells with ERR-β cDNA shows that ERR-β2, but not ERR-βsf, forms a complex with endogenous cortactin (Fig. 3C). Together, these data suggest that the ERR-β2 isoform plays a role in cell migration and is uniquely capable of binding cortactin.

ERR-β activation by DY131 suppresses GBM cell migration and remodels the actin cytoskeleton

We next examined the impact of activating endogenous ERR-β on actin cytoskeletal remodeling and cell migration in GBM cells. Treatment with the ERR-β agonist DY131 has no effect on ERR-β2, ERR-βsf, cortactin, or MGMT protein expression but modestly increases cleavage of PARP in 42MBGA-TMZres and T98G cell lines, indicative of apoptosis (Fig. 4A). However, DY131 induces a marked redistribution of cortactin away from the cell membrane and toward the center of the cell body in multiple GBM cell lines (Fig. 4B, white arrowheads vs. asterisks). In 2 of 3 TMZ-resistant GBM cell lines (42MGBA-TMZres and T98G), DY131 also significantly increases F-actin polymers, as measured by fluorescence-activated cell sorting for F-actin (Fig. 4C). Loss of cortactin membrane localization coupled with increased F-actin polymers can be indicative of impaired cell movement (38, 39), and consistent with this, we show that DY131 significantly inhibits the migration of the 8MGBA and 42MGBA parental and TMZ-res variant GBM cell lines (Fig. 4D).

Figure 4.

A) Western blot analysis of PARP cleavage, cortactin, ERR-β2, ERR-βsf, and MGMT expression in designated cell lines treated with DY131 or DMSO control (denoted as -) for 24 h. B) Immunofluorescent staining of cortactin in designated cell lines treated with 5 μM (42MBGA and 42MBGA-TMZres) or 2.5 μM DY131 (T98G) or DMSO control for 24 h prior to fixation, staining, and imaging. DY131 treatment causes a redistribution of cortactin away from the membrane (DMSO, white arrowheads vs. DY131, white asterisks). C) Changes in F-actin polymerization were measured by flow cytometry analysis of cells stained with Acti-Stain 488 phalloidin treated with 5 μM DY131 or DMSO control for 24 h. Fold increase in F-actin polymerization in each DY131-treated cell line relative to its own DMSO control is presented as the mean ± sd for 3–4 independent biologic replicates. Data were analyzed by Mann-Whitney test, *P < 0.05. D) Scratch-wound analysis of 2-dimensional migration over 72 h of treatment with the indicated concentration of DY131 or DMSO control. DY, DY131. Data are presented as means ± sd for 3 independent biologic replicates. Data were analyzed by Mann-Whitney test at each time point. *P < 0.05, **P < 0.01, *** P < 0.001, **** P < 0.0001.

Inhibition of CLKs shifts ERR-β isoform expression and potentiates DY131-mediated inhibition of growth and migration in GBM cells

ERR-β2 and ERR-βsf have distinct roles in transcription factor activity and differentially suppress tumor cell growth and motility, but how splicing of ERR-β (gene = ESRRB) is regulated to produce these isoforms is unknown. The spliceosome is the primary driver of differential splicing, but heterogeneous nuclear ribonucleoprotein (hnRNP) and SRSF (SR protein) families cooperate with the spliceosome to enhance the exclusion vs. inclusion (respectively) of specific exons (Fig. 5A). SR proteins broadly promote exon inclusion by binding cis exon splicing enhancers (ESEs), whereas hnRNPs are able to antagonize SR proteins, which promotes exon skipping [e.g., Rahman et al. (40)]. The ERR-β2 transcript has a unique exon inclusion event (exon 10), and multiple high-confidence SRSF6 consensus ESEs are predicted in ESRRB exon 10 (Fig. 5B) (27, 28).

Figure 5.

A) Simplified schematic showing interplay between SR proteins and hnRNPs, which broadly promote exon inclusion or exclusion, respectively. star = phosphorylation; U1,2 = spliceosomal components. ESS, exon splicing suppressor. B) RBPmap prediction of high-confidence SRSF6 binding sites in ERR-β2–specific exon 10. C) Representative Western blot of pSRs (1H4), ERR-β2, and ERR-βsf expression, and Spearman rank-order correlation for pSRSF6 with ERR-β2, but not ERR-βsf, in GBM PDXs. Arrows indicate ERR-β2 and pSRSF6. D, E) Changes in ERR-β2 and ERR-βsf expression, PARP cleavage, and serine 10 phosphorylation of histone H3 were measured in 42MGBA-TMZres (D) and T98G (E) cells treated with 50 μM TG-003, the combination of 50 μM TG-003 and 5 μM DY131, or DMSO for 24 h. Asterisks indicate increased ERR-β2 (D) or decreased ERR-βsf expression (E), both of which serve to increase the relative abundance of ERR-β2. Bar charts depict the normalized ratio of ERR-β2 to ERR-βsf protein expression by densitometry. DY, ERRβ-agonist; pHis3, phosphorylated histone 3; tHis3, total histone 3; TG, pan-cdc2-like kinase inhibitor.

Serine phosphorylation of SR proteins modifies their ability to promote exon inclusion. Hyperphosphorylated SR proteins are recruited to sites of active transcription and pre-mRNA processing, where they can bind ESEs and strengthen splicing recognition sites (7). After exon-exon joining in the now-mature transcript, SR proteins become hypophosphorylated and are exported from the nucleus (41). Using a pan-pSR antibody, we show that there is a significant positive correlation between pSRSF6 and the ERR-β2 isoform in GBM PDX models (Fig. 5C) but no correlation between pSRSF6 and ERR-βsf or ERR-β2 and any other pSR protein. One of the kinase families responsible for catalyzing nuclear phosphorylation of SR proteins is the CLK family (41). TG-003 is a pan-CLK-1, -2, and -4 inhibitor that suppresses the phosphorylation of multiple SR proteins, leading to differential splicing of a broad range of responsive mRNAs (42, 43). We therefore tested TG-003 in 2 TMZ-resistant GBM models as a strategy to shift the balance of ERR-β2 vs. ERR-βsf isoforms (Fig. 5D, E). In 42MGBA-TMZres cells, the combination of TG-003 and DY131 markedly up-regulates ERR-β2 expression (Fig. 5D, asterisk and bars), thereby increasing the ERR-β2:ERR-βsf ratio. In T98G cells, the combination of TG-003 and DY131 also increases the ratio of ERR-β2:ERR-βsf, although here, this is achieved through decreased expression of the ERR-βsf isoform (Fig. 5E, asterisk and bars). Differential patterns of SR protein expression and phosphorylation are observed in these cell models (Supplemental Fig. S4). In both cell lines, the combination of TG-003 and DY131 enhances PARP cleavage and serine 10 phosphorylation of histone H3, which are suggestive of increased apoptosis and of mitotic arrest, respectively.

As pan-CLK inhibition shifts isoform expression in favor of ERR-β2, and ERR-β2 has a more pronounced proapoptotic, antimitotic, and antimigratory activity [(19) and Figs. 3B and 5D, E and Supplemental Fig. S2], we tested whether combination treatment with TG-003 and DY131 would more robustly induce G₂/M arrest, apoptosis, and inhibit cell migration across all of our GBM models. Multiple TMZ-sensitive and -resistant GBM cell lines exhibit significantly enhanced cell death and mitotic arrest in response to combination treatment with TG-003 and DY131 (Fig. 6A, B and Supplemental Fig. S5), with 42MGBA-TMZres and T98G cells being the most sensitive. Importantly, T98G cells treated with the combination of TG-003 and DY131 undergo equivalent G₂/M arrest to T98G cells in which ERR-βsf has been knocked down (Supplemental Fig. S5), further suggesting that this phenotype is driven predominantly by ERR-β2. By contrast, MO3.13 oligodendrocytes showed no significant induction of G₂/M arrest or cell death by either drug alone or the combination. In scratch-wound assays using dose-reduced concentrations of TG-003 and DY131 to isolate specific effects on migration vs. general cell viability, combination treatment significantly inhibits cell migration in multiple TMZ-sensitive and -resistant GBM cell lines (Fig. 6C).

Figure 6.

A, B) Flow cytometric cell cycle analysis of SubG₁ [fragmented DNA (A)] and G₂/M (B) fractions of immortalized oligodendrocyte MO3.13 and GBM cell lines treated with 5 μM DY131, 50 μM TG-003, the combination of 5 μM DY131 + 50 μM TG-003 (TGDY), or DMSO control for 24 h. For each panel, data are presented as means ± sd for 3 independent biologic replicates. Data were analyzed by 1-way ANOVA with post hoc Tukey’s multiple comparisons test. C) Scratch-wound analysis of 2-dimensional migration over 72 h of treatment with dose-reduced concentrations of DY131 and TG-003, as indicated in the legend. For each panel, data are presented as means ± sd for 3 independent biologic replicates. Data were analyzed by 1-way ANOVA at each time point with post hoc Tukey’s multiple comparisons test. DY, ERRβ-agonist; TG, pan-cdc2-like kinase inhibitor. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Combination treatment with TG-003 and DY131 inhibits the growth of GBM intracranial xenografts

A key limitation of novel treatment strategies for GBM is failure of therapeutic agents to effectively penetrate the BBB. The BBB protects the CNS from a range of endogenous and exogenous insults, and although this can be locally compromised by the primary tumor, disseminated GBM cells remaining after surgical resection are largely shielded from systemic therapies. DY131 and TG-003 are both reported to be brain penetrant in mice (9, 14, 15), so we tested these drugs individually and in combination using a zebrafish intracranial xenograft model (Fig. 7A). The zebrafish BBB forms at ∼3 dpf and is functionally similar to that of higher organisms (44, 45). We modified published intracranial xenograft procedures (Fig. 7B) (46, 47), injecting labeled 42MGBA-TMZres cells into 1.5-dpf zebrafish embryos. At 2–3 d after the BBB forms (by 6 dpf), tumors were imaged, fish were treated daily by addition of test compounds to the fish water, and tumors were imaged again at 10 dpf. Comparing the area of pre- vs. posttreatment confocal tumor images shows that 54% of tumors in control-treated fish increased in size, but 5 d of treatment with the dose-reduced combination of TG-003 (15 μM) and DY131 (1.5 μM) reduces this to 32% (Fig. 7C, Cochran-Armitage P = 0.09). In parallel, Cascade Blue–conjugated dextran imaging at 10 dpf was used to assess the integrity of the BBB. In both DMSO control- and combination-treated fish, signal is occluded from the brain (Fig. 7D and Supplemental Fig. S6). Collectively, these results support penetrance of TG-003 and DY131 with an intact BBB and provide evidence of antitumor activity of this combination in vivo against TMZ-resistant GBM.

Figure 7.

A) Brightfield (left) and fluorescent (right) images of 5-dpf zebrafish bearing xenografts of 42MGBA-TMZres cells. DiI-labeled tumor cells are red, and the vasculature is green. B) Experimental schematic for in vivo zebrafish model with representative images of DiI-labeled 42MBGA-TMZres cells pre- and posttreatment. C) Growth or shrinkage of 42MGBA-TMZres xenografts following 5 d of treatment with 1.5 μM DY131, 15 μM TG-003, the combination of 1.5 μM DY131 + 15 μM TG-003, or DMSO shows a trend toward tumor shrinkage in the combined treatment group. Data are presented as fractions of fish showing tumor growth vs. tumor shrinkage for 19–26 fish/group. Data were analyzed by χ² test for trend (Cochran-Armitage). P = 0.09. D) Fluorescent images of 10-dpf zebrafish bearing xenografts of 42MGBA-TMZres cells following 5 d of treatment with DMSO control (left) or the combination of 1.5 μM DY131 + 15 μM TG-003 (right). DiI-labeled tumor cells are red, the vasculature is green, and Cascade Blue–conjugated dextran shows integrity of the BBB. E) Summary schematic of key findings. CLK inhibition shifts isoform expression in favor of ERR-β2 and potentiates ERR-β agonist-mediated inhibition of growth and migration in GBM. DY, ERRβ-agonist; TG, pan-cdc2-like kinase inhibitor.

DISCUSSION

Here, we show that the ERR-β agonist, DY131, is growth inhibitory in a broad set of GBM cell lines. The cytoplasmic and nuclear ERR-β2 isoform is expressed in multiple GBM cell lines, PDX models, and human tumors, in which it suppresses GBM cell migration and interacts with the actin nucleation-promoting factor cortactin. Treatment with the ERR-β agonist DY131 remodels the actin cytoskeleton and suppresses migration, and we further show that broad CLK inhibition shifts isoform expression in favor of ERR-β2 and potentiates ERR-β agonist-mediated inhibition of growth and migration in GBM cells. Combined treatment with DY131 and TG-003 also produces a nonsignificant trend toward inhibition of intracranial tumor growth.

ERR-β2, unlike ERR-βsf, is specifically expressed in primates and localized to both the cytoplasm and the nucleus. Altered subcellular localization of splice variants is a common phenotype, especially in brain cancer and brain development [e.g., Lee et al. (48)] and suggests a specialized function of the cytoplasmic primate-specific ERR-β2 nuclear receptor isoform. Inclusion of a unique F domain that may allow ERR-β2 to bind cortactin, an essential regulator of the actin cytoskeleton and cell motility in GBM, gives further insight into its specific function. GBM is a highly migratory tumor in which cortactin plays a critical role (37). We show that the ERR-β agonist DY131 decreases migration, whereas shRNA-mediated silencing of ERR-β2, but not ERR-βsf, increases the migration of GBM cells in the absence of this agonist. This suggests that changing the balance between ERR-β2 and ERR-βsf could successfully limit GBM migration and local invasion. However, small interfering RNA/small hairpin RNA-dependent strategies have many limitations and few clinical successes (49). Splice-switching antisense oligonucleotides designed to modify isoform expression are emerging as useful approaches in musculoskeletal disorders (50), but to our knowledge, they have not yet been tested in GBM.

In the absence of established nucleotide-based strategies, we looked to splicing modulatory drugs, specifically splicing kinase inhibitors, to increase the ERR-β2:ERR-βsf ratio. We show that the pan-CLK inhibitor TG-003 in combination with DY131 increases the relative expression of the ERR-β2 isoform in 2 TMZ-resistant GBM cells and causes a more robust inhibition of G₂/M arrest and migration, which have previously been shown to be ERR-β2-dependent. Coupled with the knowledge that differential splicing is a key driver of phenotypic diversity and plasticity in the malignant brain (4–6), our results suggest a new paradigm in which splicing modulatory drugs could be an effective approach to inhibit GBM migration and invasion by changing the balance of pro- vs. antimigratory isoforms of ERR-β and other proteins. TG-003 is a broad spectrum CLK inhibitor with modest activity against the dual specificity tyrosine-phosphorylation–regulated kinase 1 kinase family (51), and we do not yet know which of the CLKs (or DYRKs) are most important for shifting isoform expression in favor of ERR-β2. Sakuma et al. (52) showed that the strength of uracil-rich polypyrimidine tracts, exon length, and abundance of splicing factor binding sites characterize TG-003–responsive exons in human and mouse skeletal muscle cells. Additional studies are necessary to determine how ESRRB sequence features and the repertoire of RNA binding factors beyond SR proteins dictate context-dependent pre-mRNA processing and the expression of distinct ERR-β isoforms. Nevertheless, we provide evidence that ERR-β, specifically the ERR-β2 isoform, has proapoptotic and antimigratory functions in GBM and that splicing modulatory drugs such as CLK inhibitors are a novel strategy for shifting the balance of ERR-β isoforms to potentiate ERR-β agonist-mediated inhibition of growth and migration in GBM cells and intracranial tumors (Fig. 7E).

Poor brain penetrance is also a common cause of failure for novel therapeutic agents in GBM clinical studies despite apparently robust antitumor activity in vitro (3), which is why intracranial GBM models are an essential preclinical test of candidate small molecules and treatment strategies. We show that the combination of DY131 and TG-003 numerically (but not significantly) restricts the growth of TMZ-resistant intracranial GBM xenografts in a context in which the BBB remains intact. Zebrafish are an ideal model organism for early-phase screening of GBM intracranial xenografts because of their evolutionarily conserved BBB structure, capacity for intrinsic neovascularization of implanted tumors, and in pigmentation-deficient transgenic animals, tumor development, and response to treatment can be tracked in real time in live animals (53). Whereas others (9, 14, 15) show that DY131 and TG-003 are brain penetrant, and our data demonstrate that these small molecules may have antitumor activity in combination, further optimization and drug development efforts will be necessary to improve the efficacy of these and other splicing modulatory drugs and nuclear receptor ligands.

A potential limitation of our study is that, aside from coimmunoprecipitation studies in Fig. 3C, it does not address the contribution of ERR-β-Δ10 to GBM. The field currently lacks reagents that are selective enough to study the endogenous function of the ERR-β-∆10 isoform. The extended carboxyl-terminal F domain of the ERR-β-Δ10 isoform has a proline-rich sequence that conforms more closely to the consensus binding motif for the Src homology 3 domain-containing protein amphiphysin (Supplemental Table S2) (54), yet we show that exogenously expressed ERR-β-Δ10 is unable to interact with endogenous cortactin. Previously published studies in cells transfected with cDNA for ERR-β-Δ10 show that this isoform, like ERR-βsf, localizes to the nucleus (and not the cytoplasm) and has transcription factor activity (16). A second limitation is that ERR-β agonist ligands and splicing kinase inhibitors can have activity against other members of the estrogen-related receptor family or off-target effects. DY131 is also an agonist for the closely related estrogen-related receptor γ (12, 55), and a prior report implicates this compound as an antagonist of Hedgehog signaling (56). However, we have shown previously that shRNA-mediated silencing of estrogen-related receptor γ does not abrogate DY131-mediated phenotypes, and the Hedgehog pathway inhibitors cyclopamine and vismodegib do not phenocopy DY131 in GBM cells (19). High (50) micromolar concentrations of TG-003 are the norm in studies testing CLK-dependent regulation of specific splice variants [e.g., Marcel et al. (43)], and this also carries the risk of off-target effects. A third limitation of the study is the nonsignificant trend toward inhibition of intracranial tumor growth by the DY131 + TG-003 combination. One potential explanation is that zebrafish viability is optimal at 28°C, and human GBM cell lines are adapted to grow at 37°C, so a typical compromise employed by zebrafish xenograft studies (including ours) is to maintain the fish at 33°C. This may contribute to both poor tolerance of the embryos to intracranial injection and the shrinkage shown in the DMSO-treated cells. Orthotopic injections in immunocompromised mouse models are an alternate approach and future strategy that will be considered to circumvent this limitation.

In summary, our current work builds upon prior studies that first established the more robust antitumor effects of the ERR-β2 isoform. We demonstrate that ERR-β2 is broadly expressed in multiple GBM cell lines and PDXs, plays an antimigratory role that we hypothesize is mediated by its interaction with cortactin via a unique F domain, and induces a robust G₂/M arrest and apoptosis upon treatment with the ERR-β agonist DY131. We further demonstrate the efficacy of splicing modulatory drugs to shift the ERR-β2:ERR-βsf isoform ratio in lieu of established antisense oligonucleotide–dependent approaches, or shRNA/small interfering RNA–mediated isoform targeting that has thus far been unsuccessful clinically in cancer. Lastly, we show preliminary potential for combination treatment with DY131 and TG-003 to shrink tumor growth in vivo.

Glossary

BBB

blood–brain barrier

CLK

cdc2-like kinase

dpf

day postfertilization

DY131

ERRβ-agonist

EGFR

epidermal growth factor

ESE

exon splicing enhancer

ERR-β

estrogen-related receptor β

ERR-β-∆10

ERR-β exon 10 deleted

ERR-βsf

ERR-β short form

ESRRB

ERR-β gene

FBS

fetal bovine serum

GBM

glioblastoma

hnRNP

heterogeneous nuclear ribonucleoprotein

IP

immunoprecipitation

LCCC

Lombardi Comprehensive Cancer Center

MGMT

O6-methylguanine methyltransferase

NHA

normal human astrocyte

PARP

poly ADP ribose polymerase

PDX

patient-derived xenograft

RNAseq

RNA sequencing

shRNA

short hairpin RNA

SR

serine/arginine rich

SRSF

SR splicing factor

TG-003

pan-cdc2–like kinase inhibitor

TMZ

temozolomide

TMZres

TMZ resistant

AUTHOR CONTRIBUTIONS

D. M. Tiek contributed to study design, performed experiments, analyzed data, and wrote the paper; S. A. Khatib performed experiments, analyzed data, and wrote the paper; C. J. Trepicchio performed experiments and analyzed data; M. M. Heckler performed experiments; S. D. Divekar performed experiments; J. N. Sarkaria provided patient-derived xenograft samples; E. Glasgow contributed to study design and performed experiments; R. B. Riggins contributed to study design, performed experiments, analyzed data, and wrote the paper; and all authors reviewed, edited, and approved the manuscript.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.