Activation of ERBB4 in Glioblastoma Can Contribute to Increased Tumorigenicity and Influence Therapeutic Response

Glioblastoma (GBM) is often resistant to conventional and targeted therapeutics. ErbB2 Receptor Tyrosine Kinase 4 (ERBB4) is expressed throughout normal brain and is an oncogene in several pediatric brain cancers; therefore, we investigated ERBB4 as a prognostic marker and therapeutic target in GBM. Using RT-qPCR, we quantified mRNA encoding total ERBB4 and known ERBB4 variants in GBM and non-neoplastic normal brain (NNB) samples. Using immunohistochemistry, we characterized the localization of total and phosphorylated ERBB4 (p-ERBB4) and EGFR protein in archived GBM samples and assessed their association with patient survival. Furthermore, we evaluated the effect of ERBB4 phosphorylation on angiogenesis and tumorigenicity in GBM xenograft models. Total ERBB4 mRNA was significantly lower in GBM than NNB samples, with the juxtamembrane JM-a and cytoplasmic CYT-2 variants predominating. ERBB4 protein was ubiquitously expressed in GBM but was not associated with patient survival. However, high p-ERBB4 in 11% of archived GBM samples, independent of p-EGFR, was associated with shorter patient survival (12.0 ± 3.2 months) than was no p-ERBB4 (22.5 ± 9.5 months). Increased ERBB4 activation was also associated with increased proliferation, angiogenesis, tumorigenicity and reduced sensitivity to anti-EGFR treatment in xenograft models. Despite low ERBB4 mRNA in GBM, the functional effects of increased ERBB4 activation identify ERBB4 as a potential prognostic and therapeutic target.

Two investigators (A.B.L. and J.F.D.) were blinded to the samples and assessed the staining independently in accordance with a pre-established scoring chart. Sections with no staining were classified as 'negative'. Sections that were weakly stained with less than 50% of tumor cells stained were classified as 'low'. Sections that were intensely stained with more than 50% of the tumor cells stained were classified as 'high'. No sections were observed with weak staining of a large proportion of cells or intense staining of a small proportion of cells.
For microvessel counts, samples were scanned for areas of high vessel concentrations ('hotspots'), and four fields of view per sample were captured using a Leica microscope and capture unit using a 20x lens. Using ImageJ software, microvessels or proliferating cells were counted for each field of view, and the mean counts±SEM were recorded.

Derivation of U87ERBB4 and U87ERBB4 E317K Cell Lines
U87MG GBM cells were obtained from the ATCC. A full-length cDNA clone encoding ERBB4, pDNR-Her4, was obtained from the PlasmID Database, Harvard University. PCR was used to amplify the insert cDNA for subcloning into the retroviral vector pBABE-puro [1]. Site-directed mutagenesis was used to introduce the E317K amino acid substitution (to generate the constitutively active form of ERBB4 identified in melanoma [2], U87ERBB4 E317K ), using the primer pair 5′-AGATGGAAGTAGAAAAAAATGGGATTAAA-3′ and 5′-TTTAATCCCATTTTTTTCTACTTCC ATCT-3′. The Plat-A amphotropic retrovirus packaging cell line was transfected with retroviral vectors corresponding to wild-type and mutant ERBB4 by using FuGENE (Roche, North Ryde, NSW, Australia). Virus-containing supernatants were then used to infect U87MG cells. Transfected cells were selected and maintained in the presence of puromycin (2 µg/mL).

MTS Proliferation Assay
Cells were seeded in triplicate at 1,000 cells/well in 96-well plates, allowed to adhere overnight, then washed in serum-free medium and incubated in medium containing 0.5% FBS at 37 °C, 5% CO2 for 6 days. CellTiter 96 Aqueous MTS solution (20 µl/well; Promega, Madison, WI, USA) was added, and the absorbance (490 nm) was measured, after 3 h at 37 °C, with a FLUOstar Optima plate reader (BMG Labtech, Offenburg, Germany).

Bio-Plex Assay
Cells were seeded at 200,000 cells/well in six-well cell culture plates and left to attach overnight. Cells were washed with PBS and serum-free medium and then serum-starved in the presence of 10 µg/mL panitumumab overnight. Next, cells were washed and lyzed with Bio-Plex Cell Lysis buffer, and the lysates were analyzed for phospho-ERK (p-ERK) and p-AKT levels using Bio-Plex Phosphoprotein Detection Assays (Bio-Rad Laboratories, Hercules, CA, USA), according to the manufacturer's protocol, and the Bio-Rad Bio-Plex 200 System with Bio-Plex Manager 5.0 software.

xCELLigence Real-Time Analysis
The Real-Time Cell Analyzer (RTCA) SP instrument (Roche Diagnostics GmbH, North Ryde, NSW, Australia) was placed in a humidified incubator maintained at 37 °C with 5% CO2. 2,500 SF767 cells were seeded in 96-well plates (E-plate 96, Roche Diagnostics GmbH) in media. Cells were initially monitored once every 2 min for 1 h and then once every hour. After addition of ErbB4-Fc, cells were monitored once every 10 min for 3 h and thereafter once every hour.

RT-qPCR Analysis
All RNA was converted to cDNA using SuperScript III (Sigma-Aldrich). The total ERBB4 and ERBB4 variant levels were analyzed by PCR with the 7900 FAST Real-Time PCR Thermocycler (Life Technologies). Thermocycling conditions were as follows: 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s).
A generic human ERBB4 20x Probe/Primer pre-mix (Applied Biosystems; Life Technologies) was used for measuring total ERBB4 expression. ERBB4 variant primers and probes were designed based on allele-specific PCR, which relies upon the 3′ end of either the forward or reverse primer to discriminate between closely related mRNAs. The same reverse primer and probe was used for all ERBB4 variants, whereas different forward primers were used to distinguish ERBB4 variants [JM The sequences used to design primers and probes were sourced from NCBI NM_0052352 Human ERBB4 JM-a-CYT-1 Human and NM_001042599 Human JM-a-CYT-2. Efficiency of amplification for each primer set was validated using a 2-fold dilution series of cDNA. Efficiency ranged from 98.7-99.9%.
RT-qPCR was performed by amplifying triplicate samples containing 10-µL reaction mixture consisting of 2× gene expression mastermix (Applied Biosystems), 18 µM primer, 5 µM probe and 5 µg DNA. The H6PD gene was chosen as reference standard for normalization using the comparative Ct method [2-ΔΔCt] and analyzed using SDS 2.3 software (Applied Biosystems).

Xenografts
Mice were injected subcutaneously in each flank with 1×10 6 SF767 cells or U87MG cells retrovirally transduced with U87ERBB4 or U87ERBB4 E317K . Tumors were measured twice per week with digital calipers, and volumes calculated as width 2 × 0.5 length. Once the average tumor volume per group reached 100 mm 3 , 1 mg/mL panitumumab was administered intraperitoneally on alternate days for 14 days. Dacomitinib (Pfizer, West Ryde, NSW, Australia), a pan-ERBB inhibitor, was administered intraperitoneally (10 mg/kg in captisol) daily for 10 days. Once tumor volumes reached 1,000 mm 3 , mice were euthanized and tumor collected for analysis. Tumor volumes are presented as mean volume ± SEM.
Alginate plugs were generated as previously described [3]. Briefly, 500 µl of 1.5% sodium alginate (Sigma-Aldrich) dissolved in sterile saline was layered on top of a 5mL volume of 80 mM calcium carbonate (Sigma-Aldrich) followed by incubation at 37 °C for 30 minutes. Set plugs were then washed in sterile water and stored at 4 °C. Plugs were then injected with either 100 µl of PBS or serum free conditioned media collected from U87MG cells over three days (U87-CM) and concentrated 5 times using an Amicon ultrafuge-15 (Merck-Millipore, Australia). BALB/c mice were anesthetized with ketamine (10 mg/mL) and xylazine (5 mg/mL), and an alginate plug was placed subcutaneously into both flanks. Mice then received daily injections of PBS (vehicle) or dacomitinib (10 mg/kg) for 7 days. Grafts were then collected and fixed in 10% buffered formalin for processing and IHC. Table S1. Endogenous levels of ERBB4 in cancer cell lines.

Cell line Total ERBB4 JM-a JM-b JM-c JM-d CYT-1 CYT-2
GBM cell lines SF767    A and B), the area was removed from p-ERBB4 analysis. If p-ERBB4 and p-EGFR did not co-localize (C and D), the area was included in the analysis. This example is considered p-ERBB4 hi /p-EGFR neg . Asterisks identify areas of comparison. Scale bars, 100 µm.       Figure. 6. A, Samples were stained for CD31 (VECTOR Blue, blue) and smooth muscle actin (DAB, brown). Arrows indicate vessels. Scale bars, 50 µm. B, Samples were stained for CD31 (VECTOR Blue, blue), and nuclei were counterstained with Ki67 (VECTOR Red, red). Arrows indicate vessels. Scale bars, 50 µm.