Combined Epidermal Growth Factor Receptor and Beclin1 Autophagic Protein Expression Analysis Identifies Different Clinical Presentations, Responses to Chemo- and Radiotherapy, and Prognosis in Glioblastoma

Dysregulated EGFR in glioblastoma may inactivate the key autophagy protein Beclin1. Each of high EGFR and low Beclin1 protein expression, independently, has been associated with tumor progression and poor prognosis. High (H) compared to low (L) expression of EGFR and Beclin1 is here correlated with main clinical data in 117 patients after chemo- and radiotherapy. H-EGFR correlated with low Karnofsky performance and worse neurological performance status, higher incidence of synchronous multifocality, poor radiological evidence of response, shorter progression disease-free (PDFS), and overall survival (OS). H-Beclin1 cases showed better Karnofsky performance status, higher incidence of objective response, longer PDFS, and OS. A mutual strengthening effect emerges in correlative power of stratified L-EGFR and H-Beclin1 expression with incidence of radiological response after treatment, unifocal disease, and better prognosis, thus identifying an even longer OS group (30 months median OS compared to 18 months in L-EGFR, 15 months in H-Beclin1, and 11 months in all GBs) (P = 0.0001). Combined L-EGFR + H-Beclin1 expression may represent a biomarker in identifying relatively favorable clinical presentations and prognosis, thus envisaging possible EGFR/Beclin1-targeted therapies.


Introduction
Glioblastoma (GB) is the most frequent primary brain neoplasm and one of the most lethal tumors. Standard treatment is the maximal safe surgical resection followed by adjuvant radiotherapy (RT) and chemotherapy (CHT) [1]. Despite this multimodal treatment, prognosis remains poor, with less than 5% of the patients alive beyond five years from diagnosis [2]. Extensive multiplatform genomic characterization is increasing our understanding of the molecular bases of GB and is leading to the discovery of promising novel therapeutic targets, although efficient new treatments are still not available [3][4][5].

Ethics
Approval. This study was approved by both the Institutional Review Board and Ethics Committee of the University Hospital of Siena, and all the patients had provided signed informed consent before any treatment.

Patients.
We retrospectively reviewed the medical records of patients affected by GB (Grade IV-WHO Classification [26]) submitted to the Radiation Oncology Unit of the University Hospital of Siena for postoperative adjuvant CHT-RT from February 2002 to November 2013. Patients who had undergone a full-course RT and CHT (i.e., standard RT and concurrent and sequential temozolomide (TMZ) administration) were included in the study, whereas patients enrolled in clinical trials with experimental RT/CHT with antiangiogenetic, anti-EGFR, and any other targeted therapy were excluded.
Patients were referred, treated, and followed up on a three-month interval basis after therapy in our unit. All clinical and pathological data were recorded, including extent of surgery and histological diagnosis, clinical examination, blood counts and chemistry, Karnofsky performance status (KPS), neurological performance status (NPS), chest Xray, and pre-and postoperative and follow-up magnetic resonance imaging (MRI).

Postoperative Treatment.
RT and TMZ CHT schedules were adopted as previously described, according to a protocol-driven schedule [27]. Briefly, RT consisted of a 54-60 Gy total dose administered with three-dimensional conformal irradiation in all cases with a unifocal presentation. The planning target volume (PTV) [28] included residual tumor mass and postoperative cavity with a 2-3 cm margin. Suitable patients with small lesions received a boost dosage up to 70 Gy, limited to the gross tumor volume (GTV) [28] if no obvious progression or relevant toxicity occurred during the previous irradiation course. Patients with multifocal lesions were submitted to whole brain irradiation, up to 50 Gy. Five weekly 1.8-2 Gy sessions were administered during the entire RT course, in all cases. All patients received TMZ concurrent with RT (75 mg/mq/die) up to a maximum of 7 weeks, and most of them also received sequential TMZ CHT (150-200 mg/mq for 5 days, every 28 days), unless tumor progression or relevant toxicity occurred. The patients included in this evaluation completed at least 80% of the planned treatment.

Follow-Up.
After treatment, all patients were included in a follow-up program. General and neurological examinations, with blood counts and chemistry, were performed every three months, as previously outlined.
NPS: Patients were assigned to five categories after surgery (1: no neurological impairment; 2: some neurological impairment; 3: moderate impairment; 4: major functional impairment; 5: no conscious response), according to the Medical Research Council Brain Tumor Working Party [29].
MRI disease presentation (unifocal versus synchronous multifocal disease) was assessed at the preoperative MRI examination. Multifocality consisted of at least two lesions at the gadolinium-enhanced T1 sequence, separated by a distance of at least 1 cm.
After RT and concurrent TMZ administration, MRI was repeated, in order to assess the subsequent tumor volume evolution with respect to pre-RT status, the first time at 2-3 weeks after completion, then on a 3-month basis, and in any case of suspicion of tumor progression on clinical grounds.
Radiological response (RR) was so detected and classified into complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD) at the first MRI examination. Objective response is defined as OR = CR + PR. RR was assessed using either MacDonald's criteria [30]or response assessment in neurooncology (RANO) criteria [31], respectively, before and after 2010.
2.6. Molecular Determination of the Methylation Status of the MGMT Gene Promoter. MGMT gene promoter methylation was assessed by methylation-specific polymerase chain reaction (PCR). Briefly, genomic DNA was extracted from paraffin-embedded tumor sections and treated with sodium bisulfite using the EZ DNA Methylation-Gold kit (HISS Diagnostics, GmbH, Freiburg, Germany). Primer sequences were used to detect methylated and unmethylated MGMT promoter sequences. PCR products were separated on 2% agarose gel. A glioma cell line, with a completely methylated MGMT promoter, and peripheral blood mononucleated cells served as positive and negative control samples, respectively.
The evaluation of the signal was performed by Ultra-Vision LP Large Volume Detection System HRP Polymer (Bio-Optica, Milan), with the diaminobenzidine chromogen (Dako) for 8 min. Sections were then counterstained with Meyer's hematoxylin. In all cases, negative controls were performed by repeating the procedure and omitting the primary antibody.

Assessment of Immunostaining.
Staining was independently evaluated by two of the authors (CM, MAGMB), at medium resolution (20x objective, eye piece 1.25) all throughout tumor sections. EGFR membranous and/or cytoplasmic and Beclin1 cytoplasmic immunoreactivity scored 0 if negative and from 2 to 5, if positive, on the basis of both the stain's intensity (1: weak, 2: moderate, and 3: strong) and the percentage of positive cells (1: ≤50%, 2: >50%). We considered scores 0-2 as a low (L) and scores 3-5 as a high (H) protein expression, respectively.

Double EGFR-Beclin1
Immunofluorescence Stain. In order to colocalize EGFR and Beclin1 protein in tumor cells, in representative cases of each group of low or high protein expressing GBs, a double immunofluorescence stain was performed. Briefly, 4 m thick sections were deparaffinized in xylene and rehydrated in graded ethanol solutions (100%, 95%, 80%, and 70%), 5 minutes each, and washed in dH 2 O. Then, antigen retrieval was obtained by incubation with 10 mM sodium citrate buffer (pH 6.0) at a subboiling temperature for 20 min. Sections were then cooled for 10 min, washed in phosphate-buffered saline (PBS), and incubated overnight at 4 ∘ C with the following antibodies: mouse anti-EGFR (Undiluted, Novocastra); rabbit anti-Beclin1 (diluted 1 : 200, Sigma-Aldrich). The slides were washed three times with PBS and incubated with the secondary antibody fluorochrome conjugate (goat anti-rabbit Alexa Fluor 488, goat anti-mouse Alexa Fluor 568) for 1 hour at room temperature in the dark. The nuclei were counterstained by incubating the sections for 10 min with 4 ,6-diamidino-2-phenylindole (DAPI). Slides were washed in PBS and mounted with Antifade. In each case, a negative control was generated by omitting the primary antibody. Images were acquired and analyzed with a microscope Leica AF CTR6500HS (Microsystems).
Images of the bands were digitized and the densitometry was performed using the open source image processing program ImageJ (http://imagej.nih.gov/ij/); -actin bands were used for normalization.

Statistical Analysis.
We performed a correlation analysis using Spearman's rho two-tailed correlation test between clinical parameters and L-and H-EGFR/Beclin1-expression groups. Overall and progressive disease-free survival (OS and PDFS, resp.) were calculated with the Kaplan-Meier method. The univariate survival analysis was used to identify prognostic parameters as follows: clinical factors (age, KPS, NPS, and synchronous multifocality), treatment-related factors (extent of surgery, RT dose, sequential TMZ, and radiological response from MRI scans), and biological factors (EGFR/Beclin1 protein expression). We used the log-rank test to assess the significance of survival differences for the considered parameters ( values ≤ 0.05 were considered as statistically significant). We also performed a multivariate analysis (Cox regression) to quantify the relationship between survival and potential predictors, in order to identify a subgroup of independent factors significantly related to survival. All statistical analyses were performed with the SPSS 15.0 software package for Windows. Forty-three patients (36.6%) received a RT dose of <54 Gy (due to a multifocal presentation or constraints, such as critical structures very close to the tumor), 47 (33.8%) a dose of 54-60 Gy, and 27 (23.7%) a boost up to 70 Gy, according to the aforementioned protocol. Ninety-one (77.6%) patients completed the full-course of TMZ concurrently to RT; out of them, sequential TMZ was then administered in 76 patients (64.9%), until the events of tumor progression or severe toxicity (64.9%), whereas 41 (35.1%) patients did not have this treatment scheduled, due to early tumor progression or toxicity at the end of the concurrent RT and TMZ administration.

Methylation Status of MGMT Gene Promoter.
MGMT methylation status was assessed in 83 patients: the MGMT promoter was unmethylated in 45 cases (54.2%) and methylated in 38 (45.8%).

EGFR and Beclin1
Immunohistochemistry. EGFR membranous and/or cytoplasmic positivity was observed in most GBs, while few nuclei were decorated by EGFR in a minority of cases. Some cases were completely negative for EGFR. EGFR staining was not observed in vessel endothelia or in normal/reactive glia (Figures 1(a), 2(a), and 3(a)). On the other hand, Beclin1 stained normal cells and was heterogeneous in GBs, with a higher number of cases negative or lowly expressing the protein; a pin point cytoplasmic staining and variable nuclear immunopositivity were observed ( Figures  1(b), 2(b), and 3(b)). In several cases, a heterogeneous expression of both proteins was observed in limited areas. evidenced by the double EGFR-Beclin1 immunofluorescence stain ( Figure 5).

Discussion
Glioblastoma has a poor prognosis. Patients under current therapies have a median survival of approximately one year  after diagnosis, and it rarely exceeds 15 months for patients enrolled in clinical trials [5]. An increased efficacy of the standard of care was recently achieved, in fact, with TMZ concomitant and adjuvant in respect to radiotherapy [32,33]. However, response to therapy and prognosis highly depends on both clinical and molecular determinants. Clinical trials as well as retrospective case series analyses outline the prognostic profile of subgroups of patients with a more favorable outcome, based on classical factors (i.e., young age, a good KPS and NPS, circumscribed neoplasms arising in noneloquent areas allowing for a gross total resection, and a high RT dose, i.e., 60 Gy [2,34]), which were all associated with a better prognosis also in our series, novel chemoradiotherapeutic approaches [33], and molecular biomarkers (i.e., themethylation of O6-methylguanine-DNA methyltransferase (MGMT) gene promoter [32] and mutations of isocitrate dehydrogenase (IDH1) gene [35]). We also found a significant correlation of MGMT methylation with a longer survival, although the analysis was not conducted in all the cases. The network of prognostic factors is continuously growing and large-scale genome analyses are further subgrouping molecular subtypes of GB associated with different prognosis [3,7]. However, to date, among molecular biomarkers, only the MGMT promoter methylation status predictive biomarker has an undoubtedly high impact on clinical practice, being used for stratification of RT and TMZ-CHT treatment regimes, leading to a limited but significant improvement of survival [32,34]. Therefore, further molecular prognostic biomarkers and targets for identifying patients at a higher prognostic risk are needed, for enlarging the horizon of future individualized therapies. Our study suggests that the combined evaluation of EGFR and Beclin1 autophagic protein expression in tumor tissue sections could add valuable information to the prognostic molecular profile of GB.
In our experience, in fact, multivariate analysis indicated that, combining the two variables, that is, L-EGFR and H-Beclin1 expression, a subgroup of patients (20.4%) with a more favorable prognosis could be identified. This subgroup In the latter group, in particular, in no case was there multifocality of the neoplasm at onset. EGFR protein overexpression and gene mutation/amplification are known drivers of gliomagenesis and GB aggressiveness, activating signaling cascades that trigger tumor cell proliferation and invasiveness, angiogenesis, and suppressing apoptotic cell death [20,[36][37][38][39], largely contributing to the high RT-CHT therapy resistant GB phenotype [6,39,40]. EGFR alterations are found in most GBs and characterize the most frequent molecular classic subtype [3], thus being an ideal targetable molecule. However, the use of monoclonal antibodies for therapy has not yielded promising results to date in clinical trials [41]. Novel strategies are intended to target the key phosphorylated kinases and/or altered metabolic pathways downstream EGFR [41]. Autophagy is one of the altered metabolic pathways inhibited by EGFR, which acts via mTOR [8] or by direct inhibition of Beclin1, a cytoplasmic protein that induces autophagy by binding to Vps34-Vps15 core [8,18,19,42]. It is known that this catabolic process acts as a tumor suppressor in the early phases of carcinogenesis, while, in advanced neoplasms, depending on tumor cell context and type, it may either promote or inhibit cancer progression and therapy resistance, being also able to induce an autophagy-related or type II programmed tumor cell death [11]. The authors [22], as others [23], found that Beclin1 was underexpressed in most GBs: this was associated with a decreased apoptosis and negatively impacted on prognosis [24]. The overall negative impact of activated EGFR on autophagy could partly explain the alternate, almost mutually exclusive expression of EGFR and Beclin1 that we observed in most GBs. By examining the colocalization of the two proteins with a double immunofluorescence stain, in areas showing heterogenous protein expression, cells positive for EGFR were negative for Beclin1 and vice versa. EGFR's direct inhibition of Beclin1 and/or the existence of mutated forms of Beclin1 recently observed in human nonsmall cell lung carcinoma cells [19], in which EGFR was found colocalized with Beclin1 in autophagic vacuoles, support the coexpression of the two proteins that we also observed in GB cells. The distribution of the two proteins was, in fact, heterogenous in many cases, although subgroups of GBs could be identified based on the dominant protein expression pattern. Other unknown-to-us factors may occur in cases in which both proteins were negative. Synchronous multifocality of newly diagnosed GBs and GB aggressiveness are partly linked to the invasion ability of neoplastic cells, resulting from an orchestrated activation of cell migration, in which EGFR plays a pivotal role [38,[43][44][45].
Low expression of Beclin1 was also related to distant metastasis risk in breast cancer [46] and recently we demonstrated that the combined silencing of EGFR and induction of autophagy by rapamycin has additive effects both in increasing radiation sensitivity and in inhibiting cell migration ability, in U373 and T98G GB cells [25]. Therefore, both L-EGFR and H-Beclin1 might have interacted in contributing to the minor aggressiveness, the higher MRI evaluated response to therapy, and the lack of multifocal presentation, which we observed in the GB group with a more favorable prognosis, whereas the leading EGFR ability on cellular invasiveness may partly justify why H-EGFR patients showed a significantly higher multifocal presentation, besides a more frequent progression of the disease at imaging after treatment and a poorer prognosis.
MGMT methylation was positively correlated with prognosis. However, it was not correlated with either EGFR or Beclin1 expression, further supporting their independent role as prognosis biomarkers, despite the fact that MGMT analysis was not performed in all our GB cases.

Conclusion
Our results provide a preliminary assessment of the role of EGFR and Beclin1, extrapolated from a series of GB patients treated according to a prospectively definite protocol. Some correlations, in terms of clinical features at referral, imaging data, response to RT-TMZ treatment, and survival, can be established with EGFR and Beclin1 expression. In particular, some parameters of aggressiveness in GBs, such as multifocality (probably related to tumor cell invasiveness) and the type of response to postoperative treatment, deserve further study with regard to these relationships, in our opinion. It is noteworthy that a combined L-EGFR and H-Beclin1 GB profile seems to identify, in our observations, a subgroup of long-surviving patients. This new disclosure might contribute to the other available data suitable for prognostic stratification of GB patients and also envisage future implications for targeted therapies.