EGFR and C/EBP-β oncogenic signaling is bidirectional in human glioma and varies with the C/EBP-β isoform

Abstract

We investigated the intersection of epidermal growth factor receptor (EGFR) and CCAAT enhancer binding protein (C/EBP)-β signaling in glioblastoma (GBM), given that both gene products strongly influence neoplastic behavior. C/EBP-β is known to drive the mesenchymal transcriptional signature in GBM, likely through strong microenvironmental influences, whereas the genetic contributions to its up-regulation in this disease are not well described. We demonstrated that stable overexpression and activation of WT EGFR (U87MG-WT) led to elevated C/EBP-β expression, as well as enhanced nuclear translocation and DNA-binding activity, leading to up-regulation of C/EBP-β transcription and translation. Deeper investigation identified bidirectional regulation, with C/EBP-β also causing up-regulation of EGFR that was at least partially dependent on the STAT3. Based on ChIP-based studies, we also found that that the translational isoforms of C/EBP-β [liver-enriched transcription-activating protein (LAP)-1/2 and liver inhibitory protein (LIP)] have differential occupancy on STAT3 promoter and opposing roles in transcriptional regulation of STAT3 and EGFR. We further demonstrated that the shorter C/EBP-β isoform, LIP, promoted proliferation and migration of U87MG glioma cells, potentially via induction of cytokine IL-6. Our molecular dissection of EGFR and C/EBP-β pathway interactions uncovered a complex signaling network in which increased activity of either EGFR or C/EBP-β leads to the up-regulation of the other, enhancing oncogenic signaling. Disrupting the EGFR-C/EBP-β signaling axis could attenuate malignant behavior of glioblastoma.—Selagea, L., Mishra, A., Anand, M., Ross, J., Tucker-Burden, C., Kong, J., Brat, D. J. EGFR and C/EBP-β oncogenic signaling is bidirectional in human glioma and varies with the C/EBP-β isoform.

Epidermal growth factor receptor (EGFR) is a transmembrane glycoprotein receptor tyrosine kinase, and its gene, EGFR, is frequently altered in human glioblastoma (GBM). Over 60% of GBMs have EGFR amplification (1), and half of these harbor mutations, the most frequent being the gain-of-function mutation EGFRvIII, in which deletion of exons 2–7 generates an in-frame truncation in the extracellular domain that leads to constitutive kinase activity (2–4). EGFR amplification results in the activation of downstream mitogenic signaling cascades, including the AKT and MAPK pathways to support glioma growth (5).

CCAAT enhancer–binding protein (C/EBP)-β is a nuclear transcription factor belonging to the basic leucine zipper C/EBP superfamily and controls the growth properties of numerous neoplastic diseases (6–8). The C/EBP-b gene contains only 1 exon, yet generates 3 distinct isoforms with variable N-terminal lengths because of the presence of 2 internal AUG start sites. The 2 longer isoforms, LAP1 and -2, are transcriptional activators differing by only 21 residues at the N terminus, whereas the shorter isoform, liver inhibitory protein (LIP), lacks 185 amino acids from the N terminus. The LIP isoform is incapable of transactivation, yet presumably competes with LAPs for promoter binding and has distinctive oncogenic properties in specific contexts (9, 10). The expression ratio of LAP to LIP acts as a critical determinant of C/EBP-β mediated function in specific biologic events, such as liver regeneration, osteoclastogenesis, and mammary gland development (11, 12). Many human cancers express high levels of C/EBP-β, including those of skin (13), liver (14), breast (12, 15), ovary (16), colon (17), and lymphoma (18). LIP overexpression has been demonstrated in breast cancer (15).

Carro et al. (19) implicated C/EBP-β as a critical regulator of glioma biology when they demonstrated that STAT3 and C/EBP-β sit atop a master transcriptional hierarchy within the mesenchymal subclass of GBM, accounting for ∼80% of downstream transcripts. Studies from our laboratory demonstrated that C/EBP-β expression was strongly associated with the extent of necrosis in GBM samples, suggesting a strong microenvironmental influence on both the expression of C/EBP-β and the transcriptional class of GBM. We also showed that high expression was associated with shorter survivals (20). Genetic alterations associated with C/EBP-β expression and activity in GBM are not well documented. EGFR signaling has been shown to increase C/EBP-β (LAPs and LIP) expression in cultured cells, murine cancer models (21), and human disease (22, 23). Leutz et al. (8) recently showed that LIP overexpression predisposed mice to many forms of cancer and elevated LIP:LAP ratios mediated by EGFR contribute to aggressive cancerous phenotypes (24). It has also been suggested that LIP promotes cancer progression by downstream interference with TGFβ and IGF-1R signaling (25, 26). C/EBP-β occupancy on EGFR promoter has been established (27), yet direct links between LIP/LAP and the regulation of EGFR expression has not. STAT3 could be an intermediary of these events (19), at least in glioma.

MATERIALS AND METHODS

Tissues, cells, and reagents

Human GBM cell lines were grown in standard DMEM supplemented with 5% fetal bovine serum, l-glutamine, and penicillin/streptomycin, and incubated at 37°C in a humidified 5% CO₂ chamber. EGF concentration of 20 ng/ml was used for 24 h for stimulation experiments. U87MG cells with a stable, muristerone-inducible wild-type (WT) PTEN expression construct were induced to express PTEN by treating with 1 μM muristerone A for a period of 48 h. Deidentified glioblastoma and nonneoplastic brain tissue samples were obtained from the Cancer Tissue and Pathology Shared Resource at Winship Cancer Institute and protein lysates were obtained from digested tissue for Western blot analysis.

Comparison of C/EBP-b gene expression across molecular subsets and transcriptional subtypes

The Cancer Genome Atlas (TCGA) U133 microarray data were downloaded from cBioPortal for Cancer Genomics (http://www.cbioportal.org/). Data were accessed on June 17, 2016, and ANOVA and the multiple comparison analysis test were performed to determine the gene expression of C/EBP-β across TCGA transcriptional subtypes of GBM (proneural, neural, classic, and mesenchymal). After identifying samples with data for both transcriptional subtype and C/EBP-b gene expression data (z scores from U133 microarray), 383 GBM samples were used to compare C/EBP-b in 89 classic, 123 mesenchymal, 58 neural, and 113 proneural samples. We also compared C/EBP-b gene expression between IDH-mutant (n = 13) and IDH-WT (n = 370) GBMs using the same dataset.

Real-time quantitative RT-PCR

RNA was extracted using Trizol (15596-026; Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturer’s instructions. DNA was removed by using Turbo DNA-free kit (1907 M; Thermo Fisher Scientific, Waltham, MA, USA). cDNA preparation was performed according to the manufacturer’s protocol with Ready-To-Go You-Prime First-Strand Beads (27-9264-01; GE Healthcare, Pittsburgh, PA, USA) from total RNA (1–2 μg). The quantitative determination of cDNA levels by real-time quantitative RT-PCR (qRT-PCR) used SYBER Green dye chemistry. Primer sequences for genes are described in Supplemental Table 1. All experiments were run with 3 technical and 3 biologic replicates.

ChIP assay

Chromatin isolation from glioma cells was performed with a chromatin immunoprecipitation (ChIP) assay kit according to the manufacturer’s instructions (17-295; EMD Millipore, Billerica, MA, USA). Chromatin solution (1  ml) was immunoprecipitated with the following antibodies: anti-RNA polymerase II (sc-899; Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-AcH3 (2 μg, from the ChIP assay kit), anti-C/EBP-β antibody (Sc-150; Santa Cruz Biotechnology), FLAG (3165; Sigma-Aldrich), and hemagglutinin (H9658; Sigma-Aldrich). For the negative control, IgG was used in a normalized quantity. DNA recovered after immunoprecipitation was used as a template for PCR amplification. Two percent of the chromatin solution (20 μl) was used for the input DNA as a control. The primer sequences for STAT3 promoter for the analysis of the binding sites for C/EBP-β, RNA Pol II, and acetylated H3 were 5′-CTGGCTGGTCGTGGGTAG-3′ (forward) and 5′-GGGAGCATAATTTAACCTAGAAAAAG-3′ (reverse). The amplified DNA was analyzed on a 1.5% agarose gel with ethidium bromide staining. All the experiments were performed in triplicate.

Nuclear extract preparation and EMSA

Nuclear extract was prepared by NE-PER Nuclear and Cytoplasmic Extraction Reagents (78833; Thermo Fisher Scientific Life Sciences), by adding a phosphatase inhibitor cocktail (Complete Mini 11836153001; Roche, Indianapolis, IN, USA). Consensus oligonucleotides of C/EBP-β (sc2525, sc2526) and Oct-1 (sc-2506; all supplied by Santa Cruz Biotechnology) were further labeled with γ-[³²P]-ATP (NEG035C; Perkin Elmer, Waltham, MA, USA) with T4 polynucleotide kinase (U2010; Promega, Madison, WI, USA). The binding reactions were also performed to determine the specificity of DNA probes by using the heterologous cognate sequence of the Oct-1 binding site. A competition assay using excess cold C/EBP-β was also performed, to demonstrate specificity.

Transfection and Western blot analysis

Glioma cells were transfected with X-tremeGENE 9 (06-365-787-001; Roche) or Lipofectamine 2000 (11668027; Thermo Fisher Scientific), according to the manufacturer’s protocol. Whole-cell extracts were obtained by lysing cells in RIPA buffer [50 mM Tris (pH 7.4) 150 mM NaCl, 0.5 mM EDTA, 0.5% sodium deoxycholate, 0.1% SDS, 1% Triton X-100, and 1× mini-Complete anti-proteases; Roche]. Cytosolic and nuclear fractions were prepared by using an NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Fischer Scientific) as per the manufacturer’s instructions. Protein concentrations were determined by using the DC protein assay (Bio-Rad, Hercules, CA, USA) and Nanodrop (Thermo Fisher Scientific). Protein samples (15–100 μg per well) were resolved by 4–20% gradient SDS-PAGE on Criterion TGX gels (Bio-Rad), followed by transfer onto an Immobilon PVDF Membrane (EMS Millipore). Membranes were immunoblotted with antibodies against EGFR (PA-1110; Thermo Fischer Scientific); phosphatase and tensin homolog (PTEN; sc-7974) and C/EBPβ (sc-150; Santa Cruz Biotechnology); pEGFR (2234- Y1068) and STAT3 (9132; Cell Signaling Technology, Danvers, MA, USA); β-actin (sc-47778; Santa Cruz Biotechnology); topoisomerase I (5929-1; Epitomics, Burlingame, CA, USA); GAPDH (60004-1-Ig; Protein Technologies, Tucson, AZ, USA); α-tubulin (GTX628802; Gen Tex, Millwood, NY, USA); FLAG; and hemagglutinin. Immunodetection was performed with the corresponding secondary horseradish peroxidase–conjugated antibodies (Santa Cruz Biotechnology). Horseradish peroxidase activity was detected with a SuperSignal West Pico chemiluminescence kit (Thermo Fisher Scientific).

Small interfering RNA-mediated transient silencing

For small interfering RNA (siRNA) silencing experiments, transfection of DNA/siRNA was performed in 6-cm dishes at 20–30% cell confluence with Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific). EGFR siRNAs (sc-29301) and C/EBP-β siRNA (sc-44251; both Santa Cruz Biotechnology) and empty green fluorescent protein (GFP) plasmids were transfected according to study design, and the cells were harvested after 72 h, per the manufacturer’s protocol. The GFP and scramble siRNA were used as controls to verify specificity of knockdown and transfection, respectively. Nuclear, cytoplasmic and total proteins were extracted, and Western blot analysis was performed, as previously described. The commercial siRNA used was a pool of 3 target-specific 19- to 25-nt siRNAs designed to knock down gene expression in 10–40 nM range.

Immunofluorescence

Immunofluorescent staining was performed using antibodies against C/EBP-β (sc-150; Santa Cruz Biotechnology) and EGFR (PA-1110; Thermo Fisher Scientific). For intracellular staining, cells were fixed and permeabilized with the Cytofix/Cytoperm/Permeabilization Kit (554715; BD Biosciences, Franklin Lakes, NJ, USA). Images were captured via an Axio Observer microscope (Zeiss, Thornwood, NY, USA). All experiments were performed in triplicate.

Lentiviral particle production, transduction, and generation of glioma cells with stably overexpressing LIP isoform

The pBABE-C/EBP-β-LIP plasmid was constructed in Dr. Joan Massague’s laboratory (Memorial Sloan Kettering Cancer Center, New York, NY, USA) and purchased from Addgene (plasmid no. 15713; Cambridge, MA, USA). The LIP fragment was excised and subcloned into a pGIPZ lentiviral vector (Thermo Fisher Scientific). Further lentivirus-based stable transduction was performed in a GBM cell line (U87MG), with Lipofectamine 2000. The cells were incubated for 48 h initially without antibiotic and later with the antibiotic puromycin for 1 wk. Transduction was confirmed by GFP signals of Immunofluorescence and eventually Western blot for GFP and C/EBP-β were used to verify the cell lines generated.

Cell proliferation assay

Cell proliferation was measured with MTT-based CellTiter96 Aqueous One Solution Cell Proliferation Assay Kit (G3582; Promega). GFP-overexpressing, LIP-overexpressing (LIP-OE) and LIP vector control (LIP-VC) U87MG cells were plated in 96-well plates at a concentration of 1000 cells/100 μl medium/well. MTS (20 µl) solution was added to each well and incubated at 37°C for 2 h. Cell proliferation was recorded at 0, 4, and 24 h by calculating corresponding absorbance using Synergy HT microplate reader (BioTek, Winooski, VT, USA). Equal number of cells of LIP-VC and LIP-OE were seeded, and experiments were run with 3 technical and 3 biologic replicates.

Cell migration assay

GBM cells were plated in 35 mm culture dishes with a detachable culture insert (35-mm μ-Dish, 80206; Ibidi, Madison, WI, USA). Five hundred cells were suspended in 200 μl medium and plated in dishes. The culture insert was removed the next day. Images of cell migration were taken at 0, 2, 4, 8, and 24 h with an AX-10 confocal microscope (Zeiss). An equal number of LIP-VC and LIP-OE cells were seeded for experiments.

ELISA for cytokine IL-6 detection

IL-6 secretion by U87MG LIP-OE and LIP-VC cells was measured with ELISA (Quantikine ELISA kit; R&D Systems; Minneapolis, MN, USA) as per the manufacturer’s protocol. Briefly, 1 × 10⁶ cells of each type (LIP-VC and LIP-OE) were plated in 100-mm dishes. Conditioned medium was collected and concentrated using Amicon filter units (Millipore). IL-6 antibody-coated 96-well plate was treated with 100 μl of conditioned medium, and subsequent buffers and secondary antibody were added according to protocol. Absorbance was measured at 450 nm, and background correction was performed using readings obtained at 540 nm.

Statistical analysis

All data are expressed as means ± sem. Statistical analysis was performed with ANOVA, and P < 0.05 was statistically significant.

RESULTS

EGFR and C/EBP-β are variably expressed in human glioma

To determine expression levels of C/EBP-β protein in gliomas as compared to nonneoplastic brain tissue, human GBM specimens (n = 5) and normal brain (n = 2) samples were collected and analyzed (Fig. 1A). Western blots showed that total C/EBP-β (LAP1, LAP2, and LIP) expression was significantly elevated in GBM samples, as compared to nonneoplastic normal brain tissue. C/EBP-β expression was slightly higher in this small group of GBMs with EGFR amplification than in the GBMs without EGFR amplification. The expression of the shorter LIP isoform in GBM samples was variable.

Figure 1.

EGFR and C/EBP-β expression profile in human GBM. A) Western blot for C/EBP-β (LAP1/2 and LIP) in human GBM (n = 5) and nonneoplastic brain tissue samples (n = 2) showing higher expression of LAP1/2 in GBMs than in nonneoplastic samples. EGFR-amplified tissues (n = 3) had slightly more expression levels of C/EBP-β (LAP1/2 and LIP) than the EGFR-WT (nonamplified, n = 2) samples. LIP expression was variable among the 5 GBMs. B) In silico analysis of C/EBP-b expression using TCGA GBM data. Box plots show mRNA expression of GBM transcriptional subtypes as a z-score using data collected from the U133 array platform. C/EBP-b was highest in the mesenchymal subtype and lowest in proneural, whereas the classic and neural subtypes had similar, intermediate levels of CEBP-b expression. Symbols above and below box plots represent outliers. C) Western blot showing EGFR and C/EBP-β (LAP1/2 and LIP) protein expression in specific glioma cell lines. U87MG-WT and LN229-WT represent U87MG and LN229 cells stably expressing WT-EGFR, respectively. α-Tubulin served as the loading control. D) qRT-PCR showing total expression of C/EBP-b and EGFR mRNA transcripts in human glioma lines relative to GAPDH. Data are expressed as means ± sem (n = 3).

To investigate larger numbers of GBMs, we next performed an in silico analysis of CEBP-b expression among transcriptional subtypes of GBM, with the U133 microarray platform and ANOVA (n = 383; Fig. 1B). C/EBP-b expression means of the 4 transcriptional subtypes (classic, n = 89; mesenchymal, n = 123; neural, n = 58; and proneural, n = 113) were not equal (P = 2.11 × e⁻³⁰). This in silico analysis did not discriminate between CEBP-β isoforms (LAP1, LAP2, and LIP) and assessed only total CEBP-b expression. The mesenchymal subtype had the highest expression, with multiple-comparison analysis indicating higher levels than the classic, neural, and proneural subtypes (P = 3.77 × e⁻⁹, 3.81 × e⁻⁹, and 3.77 × e⁻⁹, respectively). Both classic and neural GBMs, which are enriched with EGFR amplification, had higher expression than proneural (P = 4.50 × e⁻³ and 3.10 × e⁻³, respectively), which are not. Classic and neural subtypes of GBM had similar CEBP-b expression. Using the U133 data, we also compared the expression of CEBP-b in IDH-mutant (n = 13) and IDH-WT (n = 370) GBMs and found no significant difference (P = 0.16; not shown).

We next screened a panel of 6 glioma cell lines (T98G, LN18, U118, MG251, U87MG, and LN229) by Western blot to determine relative baseline expression levels of EGFR and C/EBP-β. Three of 6 cell lines (U118, U87MG, and LN229) showed high expression of LAP1 and -2, whereas LIP was variably expressed in 5 of 6 lines (Fig. 1C). In contrast to LN229 cells, which had more LIP than LAP1/2, U87MG cells expressed higher levels of LAP isoforms. To investigate a potential correlation of C/EBP-β and EGFR expression under unstimulated conditions, we examined these proteins in U87MG and LN229 gliomas that stably express wild-type EGFR (U87MG-WT and LN229-WT, respectively). In both EGFR-overexpressing glioma lines, LIP expression was lower than in the corresponding cells without EGFR overexpression. LAP1/2 showed higher expression in U87MG with overexpression of EGFR, but LN229 did not.

Results from qRT-PCR analysis of C/EBP-b expression in glioma lines largely corresponded to C/EBP-β protein expression, but did not appear to correlate with EGFR expression in these unstimulated preparations. We did observe up-regulated C/EBP-b mRNA level in U87MG-WT as compared to control U87MG, similar to the expression of LAP protein isoforms in these cell lines (Fig. 1D).

Because C/EBP-β protein levels varied among glioma cells and did not correlate closely with EGFR expression status, we hypothesized that other genetic factors regulate C/EBP-β to cause this discrepancy. To test the effects of PTEN, a tumor suppressor that is downstream of EGFR and is mutated in a subset of GBMs, on C/EBP-β expression, we used a muristerone-inducible (MuA) construct to express WT PTEN in U87MG cells, which harbor a PTEN mutation. We demonstrated increased PTEN protein expression in the cytoplasm and nucleus in the presence of muristerone (Supplemental Fig. S1A, B) and found that C/EBP-β was strongly down-regulated after the expression of WT PTEN. Thus, the variability of C/EBP-β expression in glioma cell lines most likely reflects the effects of multiple oncogenes and tumor suppressors, including PTEN.

EGFR drives nuclear translocation of C/EBP-β in glioma cells

To examine the effect of EGFR expression and activation on the expression and nuclear translocation of C/EBP-β, we performed Western blots of nuclear and cytoplasmic extracts obtained from EGF-induced and noninduced U87MG cells (Fig. 2A). In the presence of EGF, we found that LAP1/2 expression was significantly shifted from the cytoplasmic compartment to the nuclear compartment in both U87MG and U87MG-WT, resulting in increased nuclear expression in both cell lines. In U87MG-WT cells, which stably overexpress EGFR, EGF exposure and receptor activation, as noted by phospho-EGFR (Y1068), was associated with reduced expression of EGFR protein, as reported by others (Fig. 2B). However, nuclear C/EBP-β expression remained elevated in the nucleus of cells in EGF-induced conditions, as compared to the cytoplasm. We next examined whether inhibiting EGFR kinase activity would diminish nuclear localization of C/EBP-β. We found that exposure of U87MG-WT cells to EGF led to enhanced nuclear expression of C/EBP-β by immunofluorescence, similar to that noted by Western blot (Fig. 2C). However, exposure to EGF in the presence of tyrphostin (AG1478; Sigma-Aldrich), an EGFR-specific kinase inhibitor, attenuated the EGF-induced nuclear localization of C/EBP-β. Together, these findings suggest that EGFR activation facilitates nuclear translocation of C/EBP-β in U87MG-WT cells.

Figure 2.

EGFR overexpression and activation causes nuclear translocation of C/EBP-β. A) Western blot for nuclear (N) and cytoplasmic (C) extracts showing differential expression of C/EBP-β (LAP1/2) in U87MG and U87MG-WT cells in the presence and absence of EGF. Expression of C/EBP-β is shifted from the cytoplasmic compartment to the nuclear compartment in the presence of EGF in U87MG and U87MG-WT. Topoisomerase I and actin were loading controls for nuclear and cytoplasmic lysate, respectively. B) Western blot for nuclear and cytoplasmic extracts showing the increased detection of phosphorylated (activated) EGFR (Y1068) in the presence of EGF. Activation of EGFR by EGF also leads to reduced expression of EGFR. C) Immunofluorescence of U87MG demonstrating increased nuclear expression of C/EBP-β in presence of EGF and its attenuation by AG1478 (EGF+AG1478), an inhibitor of EGFR phosphorylation. D) EMSA demonstrated increased binding of C/EBP-β on its cognate DNA (arrow) in U87MG-WT compared to U87MG and also shows the increased binding of C/EBP-β in the presence of EGF in both cell lines. Densitometry values for intensity of bands are noted at the bottom of each lane. Asterisk indicates nonspecific and free probes.

Given the increased nuclear expression of C/EBP-β in the presence of EGFR overexpression, we next examined the effects of EGFR expression and activation on DNA binding activity of C/EBP-β by EMSA. When incubated together with the [³²P]-labeled C/EBP-β consensus DNA sequence, nuclear extract prepared from U87MG-WT cells showed much higher DNA binding activity than that from U87MG cells, further confirming enhanced nuclear localization of C/EBP-β in the presence of EGFR overexpression (Fig. 2D, with quantitative densitometry values shown at bottom). C/EBP-β DNA binding was also increased in the presence of EGF in both U87MG and U87MG-WT, indicating that receptor activation enhances this effect as well. We verified specificity of binding of C/EBP-β to cognate DNA by competition assays using a mutant C/EBP-β competitor probe. The equal extent of transactivation for a heterologous probe of transcription factor Oct-1 was used to confirm that modulation of C/EBP-β binding was specific to this nuclear transcription factor (Supplemental Fig. S2), without influencing the basic transcriptional mechanism. A competition gel shift assay was also performed to examine the specificity of cognate DNA sequences for C/EBP-β, using GBM nuclear extracts (Supplemental Fig. S4). DNA binding activity of C/EBPβ was abolished in the presence of unlabeled, 100 M excess of a specific competitor (C/EBPβ), which reappeared when a nonspecific competitor (Oct-1) was used.

EGFR and C/EBP-β participate in a costimulatory loop in human gliomas

To investigate the mechanism of C/EBP-b up-regulation by EGFR, we first analyzed the mRNA expression of C/EBP-b in U87MG-WT and U87MG cells and, as expected, observed increased expression in the setting of EGFR overexpression (Fig. 3A). When EGFR was transiently knocked down using siRNA, we found that C/EBP-β protein expression was reduced in a dose-dependent manner in both U87MG and in U87MG-WT cells (Fig. 3B). These data suggest that EGFR expression is capable of increasing the expression of C/EBP-β and that at least part of this regulation is at the transcriptional level.

Figure 3.

EGFR and C/EBP-β regulate each other at mRNA level. A) qRT-PCR showing increased mRNA levels of EGFR and C/EBP-b relative to GAPDH in U87MG-WT vs. U87MG cells. All the experiments were performed with 3 technical and 3 biologic replicates. *P < 0.05, by Student’s t test. B) Western blots of EGFR and C/EBP-β (LAP1/2), after transient knockdown of EGFR by siRNA, demonstrating a dose-dependent reduction of C/EBP-β after siRNA. Exp, exposure time. C) qRT-PCR demonstrating that transient knockdown of C/EBP-b by siRNA leads to reduced C/EBP-b mRNA relative to GAPDH. In the setting of siRNA for C/EBP-b, EGFR mRNA expression is suppressed in a dose-dependent manner. siNT-C/EBP-β, scrambled siRNA control. si1-C/EBPβ (10 nM) and si-2-C/EBPβ (20 nM) represent 2 doses of siRNA. *P < 0.05; **P < 0.01, by Student’s t test. D) Western blot of EGFR and C/EBP-β after siRNA-mediated knockdown of C/EBPβ, showing a significant reduction in the expression of both proteins. Lanes 1 and 2 are controls from lysates transfected with scrambled control siRNA.

When we explored the regulation of EGFR by C/EBP-β, to determine whether the effects were bidirectional, we found that siRNA-mediated knockdown of C/EBP-b resulted in a consistent reduction of EGFR mRNA (Fig. 3C), as well as its protein product (Fig. 3D), in a dose-dependent manner. These results suggest an unexpected transcriptional regulation of EGFR by C/EBP-β (Fig. 3C) and that EGFR and C/EBP-β regulate one another, with at least a component of this regulation occurring at the transcriptional level.

Since the promoter of STAT3 has C/EBP-β binding sites (19) and EGFR can also be regulated by STAT3 (28), we hypothesized that C/EBP-β-mediated EGFR regulation is directed by STAT3. Using ChIP, we analyzed the STAT3 promoter for the differential binding of LAP1/2 and LIP, to determine potential contributions from these C/EBP-β isoforms. As expected, ChIP analysis verified that C/EBP-β (total) binds avidly to the STAT3 promoter (Fig. 4A). We found that overexpression of the shorter isoform, LIP, with its deleted transactivation domain, suppressed total C/EBP-β DNA binding of the STAT3 promoter (quantitative densitometry of 3 independent experiments is shown in Fig. 4B). The results suggest that higher expression of the LIP isoform in the background of endogenous LAP suppresses promoter binding of LAP to the STAT3 promoter, introducing a second level of regulation of EGFR expression by C/EBP-β.

Figure 4.

C/EBP-β protein isoforms (LAP1/2 and LIP) differentially regulate EGFR expression. A) ChIP assay showing occupancy of C/EBPβ on STAT3 gene promoter. B) Densitometry quantification of ChIP assays (n = 3). Data are from 3 different experiments and are expressed as means ± sem (n = 3). IgG, negative control; H3-Ac, acetylated histone3, positive control for active promoter. The y axis indicates densitometric intensity. *P < 0.05 by Student’s t test. C) Western blot showing protein expression levels of EGFR, C/EBP-β (LIP and LAP), and STAT3 from cell lines transfected/cotransfected with C/EBP-β isoforms (LAP and LIP). NT, vector control. When LIP was overexpressed, EGFR and STAT3 expression were suppressed. D) qRT-PCR showing differential effects of C/EBP-β isoforms on mRNA levels of STAT3. LAP overexpression induced STAT3, whereas only LIP suppressed STAT3 mRNA expression relative to GAPDH. Data are expressed as means ± sem (n = 3). *P < 0.05, **P < 0.01 by Student’s t test.

Consistent with our ChIP analysis, Western blot analyses (Fig. 4C) verified that overexpressing the LIP isoform alone modestly down-regulated the expression of STAT3 and EGFR, and this suppression was seen in conditions of high- and low-LAP expression. In contrast, overexpression of LAP 1/2 up-regulated STAT3 and EGFR levels, again demonstrating the differential roles of C/EBP-β isoforms in modulating STAT3 and EGFR. This finding was further confirmed by qRT-PCR, which showed that LIP suppressed STAT3, whereas LAP1/2 induced its mRNA expression (Fig. 4D).

The LIP isoform of C/EBP-β regulates glioma proliferation and migration

To understand the functional consequences of LIP overexpression on gliomas, we performed proliferation assays (Fig. 5A) using LIP-OE U87MG cell lines, which stably overexpress LIP, after lentiviral-based transduction. Overexpressing LIP led to a significant increase in in vitro growth at 24 h when compared to the vector controls. Migration assays (Fig. 5B) also showed that LIP-OE cells had greater rates of migration in scratch assays compared to LIP-VC cells, which became apparent as early as 4 h and were more obvious at 24 h. Because IL-6 has been widely implicated as a mediator of GBM behavior, we wondered whether LIP-OE cells that showed enhanced proliferative and migratory properties also showed increased IL-6 secretion. We performed ELISA (Fig. 5C) and demonstrated increased levels of IL-6 secretion by LIP-OE cells. We further infused LIP-OE cells with IL-1 and observed a strong (6-fold) up-regulation of IL-6 (from 29.2 to 172.5 ng/ml) when compared to LIP-VC cells. These results demonstrate a strong up-regulation of IL-6 in LIP-OE cells that may correspond to their enhanced biologic properties.

Figure 5.

LIP-OE U87MG cells drive glioma phenotypes. A) MTT-based cellular proliferation assays showing that LIP-OE U87MG cells with stable LIP overexpression are more proliferative over the course of 4 d than LIP-VC cells expressing empty vector. B) Migration assay showing enhanced migration of LIP-OE U87MG glioma cells at 4 and 24 h compared with LIP-VC cells with an empty vector construct. Migration was recorded at 0, 2, 4, and 24 h. C) ELISA demonstrated increased secretion of IL-6 in LIP-OE gliomas compared with LIP-VC cells under conditions of IL-1β stimulation (right) or no stimulation (left).

DISCUSSION

In the present study, we elucidated mechanisms by which EGFR and C/EBP-β signal transduction pathways intersect in malignant gliomas and demonstrated differential roles of C/EBP-β protein isoforms. These studies are significant, because EGFR and C/EBP-β are central regulators of GBM behavior. EGFR is amplified in more than 60% of GBMs, leading to enhanced signaling through Akt and MAPK pathways (1). Given its up-regulation in a large subset of GBMs and the importance of enhanced signal transduction in glioma biology, EGFR remains a primary oncogene for targeted therapies, although attempts thus far have not been successful. A prior study also indicated that C/EBP-β and STAT3 are among a small number of master transcriptional regulators that sit on top of the signal transduction hierarchy and drive the mesenchymal signature of GBMs (19). We have previously demonstrated that hypoxia and necrosis are closely associated with enhanced expression of C/EBP-β and suggested that these microenvironmental factors drive the mesenchymal signature (20). However, the contribution of genetic alterations to C/EBP-β expression and activity has not been fully explored in mesenchymal or nonmesenchymal GBMs. Because EGFR amplification is not a key component of the mesenchymal signature, we wondered how EGFR signaling interfaces, if at all, with C/EBP-β. Our studies demonstrated that oncogenic signaling by EGFR provides an additional mechanism for up-regulating C/EBP-β in the absence of hypoxia and necrosis, at least in a subset of GBMs.

In our in silico analysis of the TCGA dataset, we found that CEBP-b expression was similar in IDH WT and IDH mutant GBMs. Of note, the number of IDH mutant tumors was relatively small, and EGFR amplification was not present within this subset. We also compared CEBP-b expression across the transcriptional subtypes of GBM and showed, as previously reported (19), that its expression was highest in the mesenchymal subtype. The classic and neural subtypes, which are enriched for EFGR amplifications, showed similar intermediate levels of CEBP-b expression, whereas the proneural subset, which is not characterized by EGFR-amplified GBMs, had the lowest. This type of in silico analysis does not include information on the isoforms of C/EBP-β, but assesses only the total expression of C/EBP-b mRNA. Therefore, differential gene expression of C/EBP-b conveys only partial information on the activity of C/EBP-β, because there are often variations in the ratio of LAP and LIP in specific forms of cancer. We demonstrated that C/EBP-β expression was up-regulated in malignant gliomas, as compared to normal brain, consistent with prior findings, but our results add complexity, because specific isoforms of C/EBP-β appear to have differential effects on glioma biology, with the LIP isoform inhibiting binding of LAP to the STAT3 promoter and driving proliferation and migration. These findings are consistent with those of Leutz and colleagues (8), who demonstrated a distinctive role of LIP, separate from LAP, in numerous forms of cancer in a mouse model, but did not include brain tumors. Prior studies that used genome-averaging techniques, such as gene expression arrays, may have missed the independent contributions of C/EBP-β isoforms on transcriptional signatures and glioma biology, whereas our study emphasized that expression patterns of C/EBP-β isoforms will be a critical consideration for future studies.

We were intrigued that the LIP isoform was down-regulated by EGFR overexpression in U87MG and LN229 gliomas, because EGFR has been shown to cause LIP overexpression in breast cancer (21). Nonetheless, our studies and those on breast cancer both concluded that the ultimate consequences of LIP expression were similar, leading to increased tumor cell proliferation in both cases. It may be that LAP-mediated oncogenic signaling is fully operational in the setting of EGFR overexpression in gliomas, whereas LIP-mediated effects may be required only in the setting of low EGFR. This interpretation would also explain our findings that LIP overexpression resulted in suppression of LAP as well as EGFR in U87MG gliomas. Precise molecular mechanisms of LIP/LAP regulation are not currently known, yet CUG-BP1, an RNA CUG triplet repeat binding protein (hnRNP), appears to regulate the expression of the LIP isoform (29, 30). We have not demonstrated a direct link, but we have shown that LIP is overexpressed specifically in those glioma cell lines that show the highest expression of CUG-BP1 (T98G, U251, U87MG, and LN229; Supplemental Fig. S3), adding a layer of C/EBP-β protein regulation that may explain differences among cancer types.

In the presence of activating levels of EGF, we showed that increased phosphorylation of EGFR was associated with enhanced nuclear translocation and DNA binding of C/EBP-β in U87MG cells. Our experiments also uncovered a bidirectional regulation, in which EGFR and C/EBP-β modulate each other at the transcriptional level. At least a component of transcriptional regulation of C/EBP-β is dependent on Ras signaling, likely downstream of EGFR (27), but EGFR signaling can also modulate the DNA-binding activity of numerous critical transcription factors, such as AP-1 and CREB/ATF-1 (31), to the C/EBP-β promoter (32, 33). Conversely, our data and those of others also strongly indicate that C/EBP-β is capable of enhancing EGFR promoter activity (27). We were able to demonstrate that C/EBP-β-mediated up-regulation of EGFR was at least partially through STAT3, which may be expected, because the EGFR promoter is directly regulated by STAT3 and C/EBP-β can stimulate STAT3 activity in glioma (19). However, it is also clear that the LIP isoform may have the ability to attenuate C/EBP-β recruitment to the STAT3 promoter or function as a dominant negative to impede transcription.

Finally, we found that differential expression of C/EBP-β isoforms has biologic implications for glioma proliferation and invasion (8, 34). We found that LIP overexpression caused enhanced proliferation, consistent with reports on other cell types, including breast cancer (35). Similarly, our studies implicated LIP as a proinvasive factor in gliomas. Studies on C/EBP-β and invasion in head and neck cancers (36) and renal cell carcinoma (37) have suggested that increased invasiveness occurs by induction of target genes such as cytokines (Il-1, Il-6). Although the same number of cells were seeded, LIP-OE cells divided more rapidly and were more motile, demonstrating increased global migration in our scratch assay. Our data also indicate that IL-6 secretion is dramatically increased by LIP and correlates with invasion, yet we do not have evidence of a causal relationship.

In summary, in our study, EGFR and C/EBP-β regulated one another bidirectionally in gliomas, and STAT3 served as an important intermediary. We also observed differential roles of long and short isoforms of C/EBP-β, in which the shorter isoform, LIP, promoted glioma-specific biologic properties.

Glossary

C/EBP

CCAAT enhancer-binding protein

ChIP

chromatin immunoprecipitation

EGFR

epidermal growth factor receptor

GBM

glioblastoma

GFP

green fluorescent protein

LAP

liver-enriched transcription-activating protein

LIP

liver inhibitory protein

LIP-OE

LIP overexpressing

LIP-VC

LIP vector control

PTEN

phosphatase and tensin homolog

qRT-PCR

quantitative RT-PCR

siRNA

small interfering RNA

STAT

signal transducer and activator of transcription

TCGA

The Cancer Genome Atlas

WT

wild-type

AUTHOR CONTRIBUTIONS

L. Selagea, A. Mishra, M. Anand, and D. J. Brat conceived and designed the experiments; L. Selagea, A. Mishra, M. Anand, J. Ross, C. Tucker-Burden, and J. Kong performed the experiments and acquired data; L. Selagea, A. Mishra, M. Anand, J. Ross, C. Tucker-Burden, J. Kong, and D. J. Brat analyzed and interpreted the data; L. Selagea, A. Mishra, J. Ross, J. Kong, and D. J. Brat drafted and revised the manuscript; and all authors approved of the final manuscript.