EGFRvIII mutations can emerge as late and heterogenous events in glioblastoma development and promote angiogenesis through Src activation

Abstract

Background

Amplification of the epidermal growth factor receptor (EGFR) and its mutant EGFRvIII are among the most common genetic alterations in glioblastoma (GBM), the most frequent and most aggressive primary brain tumor.

Methods

In the present work, we analyzed the clonal evolution of these major EGFR aberrations in a small cohort of GBM patients using a unique surgical multisampling technique. Furthermore, we overexpressed both receptors separately and together in 2 patient-derived GBM stem cell lines (GSCs) to analyze their functions in vivo in orthotopic xenograft models.

Results

In human GBM biopsies, we identified EGFR amplification as an early event because EGFRvIII mutations emerge from intratumoral heterogeneity later in tumor development. To investigate the biological relevance of this distinct developmental pattern, we established experimental model systems. In these models, EGFR⁺ tumor cells showed activation of classical downstream signaling pathways upon EGF stimulation and displayed enhanced invasive growth without evidence of angiogenesis in vivo. In contrast, EGFRvIII⁺ tumors were driven by activation of the prototypical Src family kinase c-Src that promoted VEGF secretion leading to angiogenic tumor growth.

Conclusions

The presented work shows that sequential EGFR amplification and EGFRvIII mutations might represent concerted evolutionary events that drive the aggressive nature of GBM by promoting invasion and angiogenesis via distinct signaling pathways. In particular, c-SRC may be an attractive therapeutic target for tumors harboring EGFRvIII as we identified this protein specifically mediating angiogenic tumor growth downstream of EGFRvIII.

EGFR amplification, which is detected in 30%–40% of tumors, remains among the most frequently occurring focal genetic aberrations observed in primary GBM.¹ EGFRvIII, a mutant EGF receptor overexpressed in 50%–60% of EGFR-amplified GBMs, lacks the extracellular ligand-binding domain (exons 2–7 deletion) and is constitutively active.² EGFRvIII alone has no transforming capacity, but it can contribute to the transformation of normal cells in the context of other mutations.^3,4

Genetic heterogeneity within tumors likely contributes to the failure of molecular-targeted therapies in cancer treatment.^5–7 Surgical multisampling of GBM, in addition to recent advances in sequencing technology, has revealed profound heterogeneity within individual tumors, even at the single cell level.^8,9 For instance, a recent study using single-cell nucleus sequencing showed intratumoral heterogeneity of a rare EGFR mutant (EGFRvII), while EGFR amplification was found in all tumor cells.¹⁰ At present, it is unclear to what extent this heterogeneity accounts for the more frequently occurring EGFRvIII mutation, which we addressed in a small cohort of GBM patients in the present study.

Invasion and angiogenesis are hallmarks of GBM, and a number of studies have shown that both wtEGFR and EGFRvIII can activate these growth processes in malignant and normal cells.^11–21 However, the majority of these studies focused on only one of the receptors. Thus, there is an urgent need to study both receptors within one model system in vivo to accurately identify their functions within GBM development.

Methods

Patient Recruitment and Sample Collection

This project received approval from the UK National Research Ethics Service and the University of Cambridge National Health Service Foundation Trust Research & Development department. Details are described in Supplementary methods.

Copy Number Analyses

Assessments of EGFR copy number changes (gene amplifications) were performed as described in Supplementary methods.

Identification of EGFR Deletions

Intragenic deletions of EGFR exons 2–7 were identified by long-range PCR amplifications covering the regions of potential breakpoint (intron 1 and intron 7). PCRs were performed using the Qiagen Long Range PCR kit according to the manufacturer's instructions and using the primers described previously.²²

Screening for EGFR Point Mutations

Point mutations in the EGFR gene were assessed by PCR amplification of exons 2–7 with subsequent capillary sequencing of the resulting PCR products. Details are described in Supplementary methods.

Cell Culture

Cells from the human GBM biopsy (referred to as P3) were passaged 20 times in nude rats, as previously described.²³ NCH421k denotes cells derived from a biopsy of another patient, as previously described,²⁴ and were kindly provided by Dr. Christel Herold-Mende (University of Heidelberg, Germany). Cells were cultured in Neurobasal (NB) medium as previously described.²⁵ NB medium with bFGF is a complete medium (NBM). The human embryonic kidney cell line 293T and human glioma cell line U87 were obtained and maintained, as previously described.²⁶ Unless otherwise indicated, all cell lines were grown at 37°C in a humidified atmosphere of 5% CO₂.

Molecular Cloning

The cloning of EGFRvIII, wtEGFR, shSrc, and constitutive active Src (Src⁺) are described in Supplementary methods.

Lentiviral Vector Production and Transduction of Glioblastoma Cells

Lentiviral vectors carrying wtEGFR, EGFRvIII, shSRC, or Src⁺ were produced and titrated in NBM as previously described.²⁶ P3 and NCH421k cells were transduced with lentiviral vectors as previously described.²⁶

Flow Cytometric Analysis and Cell Sorting

Flow cytometric analyses were performed on an Accuri-6 flow cytometer (BD Biosciences). Subsequent data analysis was performed using FlowJo software (Tree Star Inc.). Details are described in Supplementary methods.

In Vitro Signaling Complex Immunocapture Mass Spectrometry

The procedures were performed as previously described.²⁷

In Vivo Experiments

Nude immunodeficient rats (rnu/rnu Rowett) were fed a standard pellet diet and provided with water ad libitum. Animal procedures were approved by the Norwegian National Animal Research Authority. Details are described in Supplementary methods.

Immunohistochemistry

Immunohistochemistry of paraffin sections was performed as previously described.²⁸ Antibody dilutions are described in Supplementary methods.

Western Blotting

Protein was extracted from tumor tissue in lysis buffer (Kinexus), and electrophoresis and blotting were performed as previously described.²³ Antibody dilutions are described in Supplementary methods.

Magnetic Resonance Imaging

MRI was performed and analyzed based on methods described previously.²⁹ Perfusion and permeability parameters were calculated as previously described³⁰ using routines custom-developed in Matlab (MathWorks, Inc.).

Statistical Analysis

Survival was analyzed by a log-rank test based on the Kaplan-Meier test using SPSS software. Differences between pairs of groups were determined with the Student t test. Statistical significance of changes between perfusion, permeability, and contrast-enhancement parameters in the MRI study was assessed by Wilcoxon rank-sum tests using Matlab (MathWorks, Inc). Individual pixels were used as observations. P values < .05 were considered significant.

Results

The Evolutionary Dynamics of EGFR Amplification and EGFRvIII Mutation in Glioblastoma

To characterize the clonal evolution of EGFR aberrations, we studied EGFR-amplified GBMs because only these tumors also carry the EGFRvIII mutation. We analyzed 3 GBMs harboring EGFR amplification using a unique surgical multisampling technique, as previously described.⁸ This approach enabled us to address intratumoral heterogeneity of EGFR aberrations (Fig. 1A). EGFR gain/amplification was present in all tumor regions in all patients, while EGFRvIII, which was detected in 2 patients, exhibited regional differences (Fig. 1A; Supplementary material, Table S1). Interestingly, EGFRvIII was found in 3 different tumor areas (T1, T2, and S) from patient sp54 (Fig. 1A and B; Supplementary material, Table S1). DNA breakpoint analysis revealed the same breakpoint for areas T1 and T2; however, a different breakpoint was observed in area S (Fig. 1B), indicating parallel evolution of different subclones in the tumor. A similar pattern was observed in a second patient (sp55) showing EGFRvIII mutations in the 2 areas (T4 and T) with 2 different DNA breakpoints (Fig. 1A and B). In the same patient, a hotspot EGFR point mutation (R222C) within the extracellular domain was found in a mutually exclusive fashion in samples T2 and S (Fig. 1A; Supplementary material, Table S1). We also performed molecular clock analysis of these 2 patients, as previously described,^8,31,32 and integrated the clonal evolution of EGFR mutations that appeared in distinct subclones over time (Fig. 1C). These results strongly suggest that EGFRvIII mutations (i) show intratumoral heterogeneity and (ii) develop from EGFR-amplified tumor clones during tumor progression.

Fig. 1.

The evolutionary dynamics of EGFR amplification and EGFRvIII mutation in patient glioblastoma (GBM). (A) MRI of GBM patients sp54 and sp55 with schematic illustration of tumor sampling. EGFR amplification is detected in all samples from both patients, while EGFRvIII mutations and the point mutation EGFR R222C are found only in distinct tumor areas. (B) Deletions within the EGFR. Any PCR products shorter than the expected amplicon size were subjected to capillary sequencing for identification of deletion breakpoints (indicated by arrowheads). Asterisks indicate that the given sequence is antisense. (C) Molecular clock analysis of tumor samples from patients sp54 and sp55 demonstrate evolution of EGFRvIII mutations and R222C point mutation over time.

Multimodal Imaging Reveals Increased Angiogenic and Aggressive Growth of EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ Tumors

To establish biological relevance of the distinct clonal evolution patterns of wtEGFR and EGFRvIII, we investigated the functions of both receptors in clinically relevant animal models. We used GSCs from 2 different EGFRvIII-negative human GBM biopsies with minimal (P3) or undetectable (NCH421k) wtEGFR expression (Supplementary material, Fig. S1A). NCH421k cultures and xenografts from P3 were previously characterized genetically.^25,26 We modified the cell lines to overexpress either wtEGFR (wtEGFR⁺), EGFRvIII (EGFRvIII⁺), or both receptors together (wtEGFR⁺EGFRvIII⁺) using lentiviral vectors. Transduced cells were sorted using fluorescence-activated cell sorting (FACs) to achieve high purity (Supplementary material, Fig. S1B).

To define the growth and angiogenic properties of the different cell populations in vivo, we injected the modified cell populations and parental control cells derived from both patients separately into the brains of nude rats. Kaplan-Meier survival analysis revealed significantly shorter survival for animals with EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors compared with controls and those with wtEGFR⁺ tumors (Fig. 2A; Supplementary material, Fig. S2A). There was no significant difference in survival between animals harboring wtEGFR⁺ and animals with control tumors for either of the 2 xenograft models (Fig. 2A; Supplementary material, Fig. S2A).

Fig. 2.

EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors show greater contrast enhancement and vascular permeability and are more aggressive than control and wtEGFR⁺ tumors. (A) Kaplan-Meier survival curves of EGFRvIII⁺, wtEGFR⁺EGFRvIII⁺, wtEGFR⁺, and control P3 xenografts show that the difference in survival between EGFRvIII⁺/wtEGFR⁺EGFRvIII⁺ and control/wtEGFR⁺ tumors is statistically significant (log-rank, P < .01/P < .01); n = 4. (B) T2- and T1-weighted w/contrast MRIs. (C) Blood flow (F), blood volume (BV), blood-to-tissue transfer coefficient (Ktrans), and interstitial space (Ve) were obtained by pharmacokinetic analysis of DCE-MRI data. (D) Quantification of all parameters revealed significantly larger values in the EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ groups compared with wtEGFR⁺ and control groups in P3 xenografts (P < .0001). Data are represented as mean ± standard errors; n ≥ 3.

We performed MRI with and without contrast agent to characterize angiogenesis. In xenografts from both patients, EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors revealed more angiogenesis as observed by increased contrast enhancement compared with control and wtEGFR⁺ tumors (Fig. 2B; Supplementary material, Fig. S2B). These results indicate that EGFRvIII, but not wtEGFR, fosters angiogenic tumor growth. We characterized additional MRI parameters by using dynamic contrast enhancing (DCE)-MRI. We assessed blood volume (Bv), blood flow (F), volume transfer coefficient (Ktrans), and interstitial space (Ve). These parameters are indicators of perfusion and vascular permeability representing the angiogenic propensities of the tumors. EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors from both xenografts showed significantly greater values for all 4 parameters compared with both wtEGFR⁺ and control tumors (Fig. 2C; Supplementary material, Fig. S2C).

Histologic and Molecular Analyses Confirm Increased Angiogenesis by EGFRvIII

To substantiate the imaging results, we performed histological, immunohistochemical (IHC), and molecular analysis of tumor specimens derived from the xenografts. Histology confirmed highly angiogenic tumors with microvascular proliferations and/or necrosis in the EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ groups from both xenograft models, while control and wtEGFR⁺ tumors were less angiogenic (Fig. 3A; Supplementary material, Fig. S3A). This was further verified and quantified by vWF immunohistochemistry (Fig. 3A and B; Supplementary material, Fig. S3A and B). To confirm the difference in angiogenesis at the molecular level, we performed Western blots of HIF1A and a panel of important proangiogenic factors. EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ P3 tumors demonstrated upregulation of HIF1A as well as the angiogenic factors VEGFA and angiopoietin 1 and 2 compared with both control and wtEGFR⁺ P3 tumors, which exhibited minimal expression of these proteins (Fig. 3C). HIF1A and VEGFA were also upregulated in EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ NCH421k tumors compared with control and wtEGFR⁺ tumors (Supplementary material, Fig. S3C).

Fig. 3.

EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors promote angiogenesis, while wtEGFR⁺ tumors do not. (A) H&E and immunostaining for vWF show typical microvascular proliferation (arrows), indicating angiogenic tumor growth in the EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ groups from P3 xenografts. wtEGFR⁺ and control tumors show normal vasculature. (B) Vessel area fractions as indicated by vWF immunostaining. Quantification was performed at 200× magnification. Values represent mean ± SD. * P < .05; ** P < .01: **** P < .001; n = 10. Scale bars 50 µm. (C) Western blot of EGFR, pEGFR, and angiogenic factors showing upregulation of HIF1A,VEGF, ANGPT1, and ANGPT2 in EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors compared with both wtEGFR⁺ and controls tumors. (D) Immunohistochemical staining of EGFRvIII⁺ GBM biopsies with antibodies against EGFRvIII, pEGFR, and vWF. Scale bars, 50 µm. (E) Area fraction of vascular elements immunostained with vWF from EGFRvIII⁺ vs EGFRvIII⁻/pEGFR⁺ areas from 3 different patients. Quantification was performed at 200× magnification. **** P < .001; n = 15. Values represent mean ± SD.

EGFRvIII Correlates With Increased Angiogenesis in Glioblastoma Patient Samples

To validate the results from our xenograft models in patients, we selected 3 specimens with robust EGFRvIII expression from several EGFR-amplified GBM biopsies. Importantly, these specimens showed activated EGFR (pEGFR⁺) in areas that did not express EGFRvIII, corresponding with activation of wtEGFR. Tumor sections were immunostained for vWF to investigate vessel size and to quantify vessel area. EGFRvIII⁺ tumor cells were frequently found around microvascular proliferations and dilated tumor vessels, while EGFRvIII⁻/pEGFR⁺ areas contained smaller vessels (Fig. 3D). Quantification of vessel area per microscopic field (HPF) revealed significantly greater values for EGFRvIII⁺ areas compared with EGFRvIII⁻/pEGFR⁺ areas, which confirmed induction of angiogenesis by EGFRvIII but not by wtEGFR (Fig. 3E).

wtEGFR, but not EGFRvIII, Promotes Invasion of Glioblastoma Cells

We have previously shown that wtEGFR promotes invasion of EGFR-amplified glioblastoma cells.²⁶ To further validate these findings in the present models, we analyzed invasion of the different P3 and NCH421k cell populations in vivo. We observed an increased capacity of wtEGFR⁺ tumor cells from both patients to invade into the brain parenchyma compared with EGFRvIII⁺, wtEGFR⁺EGFRvIII⁺, and control tumor cells (Supplementary material, Fig. S4A). EGFRvIII⁺ tumors showed more circumscribed growth and were even less invasive than control tumors. The results were verified by quantification of nestin⁺ invading cells into the cortex (Supplementary material, Fig. S4B).

c-SRC Is Upregulated and Activated in EGFRvIII⁺ Tumors

Since the results presented above point at different biological functions of EGFRvIII and wtEGFR we hypothesized that this was due to differential activation of downstream-signaling pathways. To further explore this hypothesis, we used signaling-complex immunocapture mass spectrometry²⁷ to generate signaling network analyses of wtEGFR⁺ and EGFRvIII⁺ P3 cells with and without EGF stimulation. wtEGFR showed strong interaction with adaptor proteins corresponding to classical RTK signaling, particularly Erk signaling (Fig. 4A; Supplementary material, Table S2). This result confirms observations from previous studies implicating Erk as a major contributor to cell migration upon EGF stimulation.^33,34 In contrast to wtEGFR, EGFRvIII showed a more intricate interaction scenario in the signaling network analyses, associating with a number of other receptors with oncogenic potential as well as with Src family kinases (SFKs) (Fig. 4A; Supplementary material, Table S3). Based on previous observations implicating SFKs in EGFR signaling,³⁵ we decided to explore the role of SFKs in our model.

Fig. 4.

c-SRC is upregulated and activated in EGFRvIII⁺ tumor cells. (A) Network analysis of the protein pool was identified to accumulate >2-fold in the stimulated wtEGFR versus unstimulated wtEGFR interactome (left) versus network analysis of the protein pool identified to accumulate >2-fold in the stimulated EGFRvIII vs wtEGFR interactome (right). Note that the SFKs YES, and FYN are implicated in the EGFRvIII interactome while they are not implicated in the wtEGFR interactome. (B) Immunohistochemical stainings for SFKs in P3 xenograft tissue specimens. (C) Western blot of in vivo lysates from control, wtEGFR⁺, and EGFRvIII⁺ P3 tumor specimens. (D) Immunostaining of glioblastoma (GBM) biopsies with antibodies against EGFRvIII and c-Src. (E) Double immune-fluorescence staining of GBM biopsies with antibodies against EGFRvIII (green) and c-SrcY216 (red).

The highly conserved nature of SFK amino acid sequence³⁶ may prevent accurate distinction of individual SFKs by high-throughput mass spectrometry. Thus, in order to validate our findings, we performed Western blots of the prominent SFKs c-SRC, FYN, c-YES, and LYN on lysates from P3 and NCH421k cells differentially expressing wtEGFR and EGFRvIII in vitro. Both P3 and NCH421k cells expressed c-SRC and LYN (Supplementary material, Fig. S5A), but we could not detect FYN or c-YES (data not shown). All cell groups expressed c-SRC and LYN independent of exogenous EGF stimulation in a heterogeneous fashion, and the cells appeared to upregulate these SFKs in an EGFR-dependent manner.

To evaluate the SFK distribution in vivo, we performed immunohistochemistry to detect different SFKs in P3 and NCH421k tumor xenograft specimens (Fig. 4B; Supplementary material, Fig. S5B). Both xenografts showed staining for c-SRC, LYN, and c-YES. Expression of c-SRC was profoundly upregulated in EGFRvIII⁺ tumors compared with control and wtEGFR⁺ tumors (Fig. 4B; Supplementary material, Fig. S5B). LYN was expressed in tumor cells and the vasculature, but it was expressed heterogeneously in both P3 and NCH421 across all tumor groups (Fig. 4B; Supplementary material, Fig. S5B). c-YES, meanwhile, was not expressed in tumor cells and was rather contained to the vascular compartment. Indeed, EGFRvIII⁺ tumors showed more pronounced staining of c-YES as these tumors exhibit enhanced angiogenesis (Fig. 4B; Supplementary material, Fig. S5B).

To confirm our findings by a second method, we performed Western blots on control, wtEGFR⁺, and EGFRvIII⁺ P3 tumor lysates (Fig. 4C; Supplementary material, Fig. S5C). The blots were consistent with immunostaining as c-SRC expression was substantially upregulated only in EGFRvIII⁺ tumor lysates. c-SRC activation, as determined by phosphorylation at both tyrosine 216 and 419, was also greater in lysates from EGFRvIII⁺ tumors. Furthermore, consistent with our previous findings, the angiogenic factors bFGF and VEGFA were upregulated only in lysates from EGFRvIII⁺ tumors compared with lysates from control and wtEGFR⁺ tumors.

To validate that c-Src is expressed and activated in patient tumors expressing EGFRvIII, we stained consecutive sections of 5 EGFRvIII⁺ patient biopsies with EGFRvIII and Src antibodies. As shown in Fig. 4D, c-SRC is expressed in EGFRvIII⁺ tumor areas. In addition, we performed double immunofluorescence staining with EGFRvIII and phospho c-SRC Y216 antibodies demonstrating coexpression of both proteins (Fig. 4E).

Angiogenic and Aggressive Tumor Growth Mediated by EGFRvIII Is Dependent on c-SRC

To investigate whether c-SRC is responsible for the EGFRvIII-mediated secretion of angiogenic factors, we performed in vitro angiogenesis arrays and blocked c-SRC activation in P3 EGFRvIII⁺ cells by either overexpressing a c-SRC shRNA construct (shSRC; Supplementary material, Fig. S6) or using anti-Src small molecule inhibitors. Additionally, we overexpressed a constitutively active Src (Src⁺; Supplementary material, Fig. S6) in P3 control cells. Both EGFRvIII⁺ and Src⁺ P3 cells secreted higher levels of VEGFA than P3 control cells, and suppression of c-SRC activity using shSRC, dasatinib, or saracatinib counteracted or reduced the enhanced VEGFA secretion (Supplementary material, Fig. S7). Next, we implanted P3EGFRvIII⁺ shSrc and control cells orthotopically into nude rats. Animals harboring EGFRvIII⁺ shSRC tumors survived significantly longer than those harboring control tumors (Fig. 5A). Importantly, EGFRvIII⁺ shSRC tumors did not show contrast enhancement on T1-weighted MRI, while control tumors did as expected (Fig. 5B).

Fig. 5.

Aggressive tumor growth and angiogenesis, mediated by EGFRvIII, are dependent on c-SRC. (A) Kaplan-Meier survival curve showing survival of rats with P3 EGFRvIII⁺ shSRC and EGFRvIII⁺ shRNA control tumors. The difference in survival is statistically significant (log-rank; P < .05). n = 4 (B) T2- and T1-weighted MRIs of control and treated tumors with and without contrast. (C) H&E and immunostaining for vWF (tumor center) show typical microvascular proliferation, indicating angiogenic tumor growth in EGFRvIII⁺ control tumors. EGFRvIII⁺ shSRC tumors, meanwhile, show significantly less angiogenesis. H&E and immunostaining for human nestin (tumor border) show significantly more invasive growth in the EGFRvIII⁺ shSRC tumors compared with control tumors. (D) Quantification of vessel area in vWF immunostained sections was performed at 200× magnification (n = 10); ** P < .01. (E) Quantification of nestin⁺ cells was performed at 400× magnification (n = 10); ** P < .01. (F) Western blot of EGFRvIII⁺ control and EGFRvIII⁺ shSRC tumor lysates.

To verify that angiogenesis was reduced in EGFRvIII⁺ shSRC compared with EGFRvIII⁺ shRNA control P3 tumors, we performed histology and vWF immunohistochemistry on tissue specimens. Control tumors showed significantly greater vessel-area staining than EGFRvIII⁺ shSRC tumors (Fig. 5C and D). Furthermore, we determined if the EGFRvIII⁺ shSRC tumors exhibited a more invasive behavior compared with control tumors because Src has previously been shown to be involved in invasion.^35,37 Interestingly, immunostaining for human nestin showed significantly greater invasion in EGFRvIII⁺ shSRC tumors compared with controls (Fig. 5C and E).

To further evaluate the differences in angiogenesis at the molecular level, we performed Western blots on lysates derived from P3 EGFRvIII⁺ shSRC and EGFRvIII⁺ shRNA control P3 tumors. Indeed, c-SRC expression and activation was reduced in EGFRvIII⁺ shSRC specimens compared with controls, and this correlated with the level of bFGF and VEGFA expression (Fig. 5F).

Constitutively Active Src Induces Angiogenesis in EGFRvIII-negative Tumors in Vivo

Since EGFRvIII appeared to be dependent on c-SRC in mediating angiogenesis and aggressive tumor growth, we hypothesized that P3 cells could become angiogenic without EGFRvIII as long as c-SRC was overexpressed and active. Accordingly, we sorted and prepared P3 Src⁺ and corresponding vector control cells for orthotopic injection into nude rats. P3 control animals lived significantly longer than those harboring Src⁺ tumors (Fig. 6A). Furthermore, we performed MR imaging that showed more heterogeneous tumors in the Src⁺ group compared with controls on T2-weighted images. Focal loss of the hyperintense signal within Src⁺ tumors may represent intratumoral microbleeding indicative of increased vascular permeability. In addition, Src⁺ P3 tumors showed increased contrast enhancement compared with control tumors (Fig. 6B).

Fig. 6.

Constitutively active Src induces angiogenesis in EGFRvIII-negative P3 tumors in vivo. (A) Kaplan-Meier survival curve showing survival of rats with P3 Src⁺ and control tumors. The difference in survival is statistically significant (log-rank; P < .05). n ≥ 4 (B) T2- and T1-weighted MRIs of control and treated tumors with and without contrast. (C) H&E and immunostaining for vWF (tumor center) show microvascular proliferation, indicating angiogenic tumor growth in Src⁺ tumors. Control tumors, meanwhile, show significantly less angiogenesis. H&E and immunostaining for human nestin (tumor border) show significantly less invasive growth in the Src⁺ tumors than in control tumors (P < .001). (D) Quantification of vessel area in vWF immunostained sections was performed at 200× magnification (n = 10); ** P < .01. (E) Quantification of nestin⁺ cells was performed at 400× magnification (n = 10); **** P < .001. (F) Western blot of control, Src⁺, wtEGFR⁺ and EGFRvIII⁺ tumor lysates.

We also performed histology and vWF IHC on xenograft tumors to further characterize the angiogenic and aggressive characteristics of Src⁺ P3 tumors. Src⁺ tumors showed significantly greater vessel-area staining than controls, indicating increased angiogenesis (Fig. 6C and D). Microbleeding around vessels was detected on hematoxylin and eosin (H&E) sections and verified the MRI findings (data not shown). We then investigated the invasive properties of P3 Src⁺ and control tumors. Notably, human nestin IHC revealed significantly greater invasion in control specimens compared with Src⁺ tumor specimens (Fig. 6C and E). To confirm the induction of angiogenesis by Src⁺ P3 tumors at the molecular level, we performed Western blots and found that upregulation of VEGFA was indeed dependent on c-SRC overexpression and activation (Fig. 6F).

Discussion

In the present study, we identified distinct clonal evolution patterns of EGFR amplification and EGFRvIII mutations in a small cohort of patient GBMs. DNA breakpoint analysis revealed the emergence of distinct EGFRvIII mutations in different regions within the same tumor. In contrast, EGFR amplification/gain was present in all tumor areas, which indicates that EGFR amplification is an early event in tumorigenesis, while EGFRvIII emerges later during tumor progression. Intratumoral heterogeneity was also recently observed for the less prominent EGFRvII mutation using single-cell nucleus sequencing.¹⁰ Tumor heterogeneity can also emerge within a therapeutic setting, as was recently shown for the EGFRvIII mutation: Treatment of EGFRvIII-mutated GBM cells with EGFR tyrosine kinase inhibitors resulted in drug resistance due to elimination of EGFRvIII from extrachromosomal DNA,²² underscoring the complexity of tumor heterogeneity.

To establish biological relevance for the sequenced EGFR aberrations, we investigated the functions of wtEGFR and EGFRvIII in human GBM xenografts in vivo. EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors grew faster and more angiogenic compared with control and wtEGFR⁺ tumors. A more aggressive phenotype mediated by EGFRvIII has also been observed in U87 cells.^38,39 However, we also found a substantial increase in angiogenesis in EGFRvIII⁺ and wtEGFR⁺EGFRvIII⁺ tumors in addition to the enhanced tumor growth. Indeed, EGFRvIII has been linked to angiogenesis in other animal models^16,18,20 and in a study investigating patient tumors by perfusion MRI.⁴⁰ We further confirmed the angiogenic capacity of EGFRvIII in patient samples by demonstrating that EGFRvIII⁺ tumor cells are more frequently found around microvascular proliferations/dilated vessels compared with EGFRvIII⁻/pEGFR⁺ tumor cells representing those with wtEGFR activation.

Interestingly, we detected a decrease in invasion by EGFRvIII, while wtEGFR instead enhanced invasion and concomitantly reduced angiogenic tumor growth in our xenograft models. We have previously shown that EGFR amplification mediates invasion independent of angiogenesis,²⁶ and we confirmed here in another model system that wtEGFR does not promote GBM angiogenesis but rather enhances invasion. Our results further support the notion that EGFR status has an important impact on the balance between invasive and angiogenic growth that culminates in intratumoral heterogeneity.

The distinct functions of wtEGFR and EGFRvIII also suggested a difference in downstream signaling. Compared with EGFRvIII, wtEGFR showed a stronger association with adaptor proteins related to classical RTK signaling. In contrast, we observed an enrichment of SFK signaling in EGFRvIII⁺ cells compared with wtEGFR⁺ cells. Specifically, EGFRvIII tumors upregulated c-SRC. Meanwhile, other SFKs including c-YES and FYN were not expressed by the tumor cells, and LYN was expressed—but not specifically upregulated—in EGFRvIII⁺ tumor cells. Although previous findings have demonstrated that c-SRC interacts with EGFR in GBM cells,^35,41 we showed here that this interaction is specific for EGFRvIII.

We verified that the angiogenic capacity of the EGFRvIII⁺ tumors was dependent on c-SRC by performing functional experiments. The angiogenesis-promoting effects of Src activation have indeed been shown in GBM cells in addition to other cell lines and tumor models including those of ovarian and colon cancer,^42–44 but this has not been previously linked to EGFRvIII signaling.

In contrast to other findings implicating Src activation in tumor invasion,^35,37,45 we propose a mechanism of c-SRC activation that is directly linked to oncogenic EGFRvIII signaling and promotes angiogenic and aggressive tumor growth. In fact, we found significantly reduced invasion in EGFRvIII⁻Src⁺ tumor specimens and significantly increased invasion in EGFRvIII⁺ shSrc tumor specimens. We attribute the discrepancies mainly to the notion that constitutive Src activation, as mediated by EGFRvIII, might produce a different biological response compared with nonconstitutive Src activation by other means. We did not observe high levels of SFK activation in wtEGFR⁺ cells in our model system, nor did we find substantial SFK activation in several GBM xenografts with EGFR amplification (unpublished results). SFK activation, and specifically activation of c-SRC, appears to be mediated by EGFRvIII and not wtEGFR.

In conclusion, the presented work suggests that sequential evolutionary events of EGFR amplification and EGFRvIII mutations together contribute to the progression of primary GBM from an invasive tumor to one that is highly angiogenic. Furthermore, our results reveal that c-SRC mediates the angiogenic tumor growth downstream of EGFRvIII and that targeting this distinct signaling axis may be an attractive approach for clinical intervention of GBM.

Supplementary Material

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

Funding

G.V. Røsland was supported by a PhD fellowship from the University of Bergen, Norway. E. Eskilsson was supported by a PhD fellowship from the Norwegian Cancer Society. R. Jirik was supported by Czech Science Foundation (Grant GA102/12/2380). This work was also supported by the Research Council of Norway, The Norwegian Cancer Society, Helse Vest, Haukeland University Hospital, the K.G. Jebsen Research Foundation, the Bergen Medical Research Foundation and in part by a grant to G.G. from the Italian Association for Cancer Research (AIRC 2013 IG 14042), grant GA102/12/2380 from the Czech Science Foundation, and by the Ministry of Education, Youth, and Sports of the Czech Republic (project No. LO1212).