A Profound Basic Characterization of eIFs in Gliomas: Identifying eIF3I and 4H as Potential Novel Target Candidates in Glioma Therapy

Simple Summary Gliomas are brain tumors with currently limited therapy options. Glioma growth and proliferation is regulated by the mTOR pathway together with eukaryotic initiation factors (eIFs). In this work we show a profound basic characterization of eIFs in human gliomas and demonstrate increased mRNA and protein expressions of several eIFs in gliomas compared to healthy control brain tissue. Moreover, increased eIF3I and eIF4H levels seem to have a negative influence on the survival of patients. Our work suggests eIF3I and eIF4H as potential targets for future glioma therapy. Abstract Glioblastoma (GBM) is an utterly devastating cerebral neoplasm and current therapies only marginally improve patients’ overall survival (OS). The PI3K/AKT/mTOR pathway participates in gliomagenesis through regulation of cell growth and proliferation. Since it is an upstream regulator of the rate-limiting translation initiation step of protein synthesis, controlled by eukaryotic initiation factors (eIFs), we aimed for a profound basic characterization of 17 eIFs to identify potential novel therapeutic targets for gliomas. Therefore, we retrospectively analyzed expressions of mTOR-related proteins and eIFs in human astrocytoma samples (WHO grades I–IV) and compared them to non-neoplastic cortical control brain tissue (CCBT) using immunoblot analyses and immunohistochemistry. We examined mRNA expression using qRT-PCR and additionally performed in silico analyses to observe the influence of eIFs on patients’ survival. Protein and mRNA expressions of eIF3B, eIF3I, eIF4A1, eIF4H, eIF5 and eIF6 were significantly increased in high grade gliomas compared to CCBT and partially in low grade gliomas. However, short OS was only associated with high eIF3I gene expression in low grade gliomas, but not in GBM. In GBM, high eIF4H gene expression significantly correlated with shorter patient survival. In conclusion, we identified eIF3I and eIF4H as the most promising targets for future therapy for glioma patients.


Introduction
Gliomas are the most common primary brain tumors originating from glial cells. According to the new 2016 WHO classification, gliomas are subdivided into astrocytic tumors, including pilocytic astrocytomas (WHO grade I), diffuse astrocytic tumors, including diffuse astrocytomas (WHO grade II), anaplastic astrocytomas (WHO grade III) and glioblastoma (GBM) (WHO grade IV) [1]. Current treatment strategies for highly malignant gliomas include surgical resection followed by adjuvant radio-chemotherapy [2]. However, with a median overall survival (OS) of 12-15 months, the outcome is still dismal [3,4]. Local, or less commonly distant, recurrence is observed in most patients following initial resection and chemo-/radiotherapy, exemplifying the tumor's infiltrative nature and heterogeneity [5].
The PI3K/AKT/mechanistic target of rapamycin (mTOR) signaling pathway is a major survival pathway and has already been extensively studied in GBM. Deregulation of PI3K/AKT/mTOR signaling seems to be one of the key players driving gliomagenesis [6]. Common mutations detected in GBM also induce a constitutive activation of this pathway [7]. Amongst others, the PI3K/AKT/mTOR pathway regulates the rate-limiting step of protein synthesis [8].
Translation from mRNA into protein can be divided into four steps, denoted as initiation, elongation, termination and ribosome recycling. Translation is mainly regulated at the initiation step and is governed by eukaryotic initiation factors (eIFs). The initiation process starts with the formation of the 43S preinitiation complex (PIC) under the assistance of the eIF3 complex, eIF1 and eIF1A [9,10]. In the PI3K/AKT/mTOR-dependent eIF4F complex activation, phosphorylation of eukaryotic translation initiation factor 4E-binding protein 1 (4E-BP1) leads to the formation of the eIF4F complex consisting of eIF4A, eIF4E and eIF4G [11]. The eIF4F complex recognizes the 7-methyl-GTP cap structure at the 5 mRNA end and joins the PIC. Afterwards, the scanning for the AUG start codon and formation of the mature 80S ribosome takes place [9,10]. eIFs may serve as tumor suppressors or promote carcinogenesis as well as tumor progression in different types of cancer [12]. Alterations in the eIF signaling cascade have also been reported in previous glioma studies. In vitro silencing or inhibition of eIF3B [13], eIF3C [14], eIF3D [15] and eIF3E [16] reduced cell proliferation and increased apoptosis in GBM cell lines. Overexpression of eIF3C [14], eIF3D [15] and eIF5A [17] was described in human glioma tissues, predominantly in WHO grades III and IV. Phospho (p)-eIF4E and p-4E-BP1 levels were increased during malignant progression in astrocytoma [18,19]. Treatment of U87MG cells and U87MG-derived xenografts with the eIF4F-complex formation inhibitor 4EGI-1 decreased cell growth and induced apoptosis and mitochondrial dysfunction [20]. Recently, in silico analyses from The Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA) revealed a significant impact of various eIF3 subunits in astrocytomas (WHO grades I-IV). In particular, the prognostic value of eIF3I gene expression in IDH1/2 mutant lower grade gliomas was shown [21].
The majority of eIF subunits have not been investigated so far, or only at the protein or mRNA level. The aim of the present study was to bridge this gap and to characterize a broad panel of eIFs in terms of protein and mRNA levels in astrocytomas (WHO grades I-IV), followed by their correlation with OS in two different publicly available databases.

Impact of eIF Gene Expression on Patients' Overall Survival
To evaluate the impact of eIF gene expression on the OS of patients with glioma, the publicly available gene expression datasets from the TCGA database were analyzed. Two TCGA datasets were studied: LGGs (WHO grades I-II) and GBM (WHO grade IV). Kaplan-Meier curves were generated for eIF3B, eIF3I, eIF4A1, eIF4H, eIF4E, eIF5 and eIF6 (Figure 4 and Figure S2). Figure 4A) but did not affect the survival of GBM patients ( Figure 4B). eIF4H (p < 0.001; Figure 4D) was negatively correlated with the OS of GBM patients, but not of patients with LGGs ( Figure 4C). eIF3B ( Figure S2A) and eIF4E ( Figure S2E) showed a tendency to be associated with a reduced OS for LGG patients and eIF3B ( Figure S2F) and eIF4A1 ( Figure S2G) for GBM patients. Gene expression of eIF5 ( Figure S2D,I) and eIF6 ( Figure  S2E,J) was not associated with decreased OS.
An additional analysis with two independent datasets (TCGA and CGGA) comparing survival in tumors with high or low gene expression revealed a significant survival difference for eIF3I in TCGA data but not CGGA data (Table S6). Survival differences for eIF4H were significant in both TCGA and CGGA data (Table S6). Notably, eIF6 showed significant differences in patient survival in CGGA but not TCGA data (Table  S6). Additional correlation coefficients for eIF3I, eIF4H and other relevant genes in gliomas are provided in the supplementary material (Tables S4 and S5).  Only eIF3I and eIF4H were significantly associated with patient outcome. Higher eIF3I expression was significantly associated with reduced OS in LGG patients (p < 0.05; Figure 4A) but did not affect the survival of GBM patients ( Figure 4B). eIF4H (p < 0.001; Figure 4D) was negatively correlated with the OS of GBM patients, but not of patients with LGGs ( Figure 4C). eIF3B ( Figure S2A) and eIF4E ( Figure S2E) showed a tendency to be associated with a reduced OS for LGG patients and eIF3B ( Figure S2F) and eIF4A1 ( Figure S2G) for GBM patients. Gene expression of eIF5 ( Figure S2D,I) and eIF6 ( Figure S2E,J) was not associated with decreased OS.
An additional analysis with two independent datasets (TCGA and CGGA) comparing survival in tumors with high or low gene expression revealed a significant survival difference for eIF3I in TCGA data but not CGGA data (Table S6). Survival differences for eIF4H were significant in both TCGA and CGGA data (Table S6). Notably, eIF6 showed significant differences in patient survival in CGGA but not TCGA data (Table S6). Additional correlation coefficients for eIF3I, eIF4H and other relevant genes in gliomas are provided in the supplementary material (Tables S4 and S5).

Immunohistochemical Evaluation Confirmed Increased eIF3I and eIF4H Levels
Cell type-specific protein expression of eIF3I and eIF4H was evaluated using IHC ( Figure 4E). Scoring of positive cells was exclusively performed for tumor tissue; necrotic areas and infiltration zones were not scored. eIF3I was significantly increased in WHO grades II-IV tumors compared to CCBT (II: p < 0.001, III: p < 0.01, IV: p < 0.01; Figure 4F). eIF4H-positive cells were only detected in astrocytoma samples and were almost absent in CCBT (I: p < 0.01, II: p < 0.001, III: p < 0.01, IV: p < 0.01; Figure 4G). The IDH1 (R132H) mutation had no significant impact on expression status of eIF3I or eIF4H positive tumor cells ( Figure S3).Error bars are partly missing because of the low number of samples or similar tissue intensity scores in the immunohistochemical evaluation.
Finally, immunohistochemical results for eIF3I and eIF4H were correlated with patients' relapse-free survival and OS in LGGs (WHO grades I-II) and HGGs (WHO grades II-IV, Figure S4). As group numbers were too low to determine hazard ratios, only proportion scores were generated (Table 1).

Discussion
Gliomas are very heterogeneous tumors with varying clinical outcomes. Although much effort has gone into the development of novel therapies over the last few years, treatment options are still very limited and many clinical trials have failed [22]. New biomarkers and therapeutic approaches are urgently required. Thus, the purpose of our study was to identify eIFs that can serve as novel prognostic biomarkers or therapeutic targets in gliomas.
Amongst others, the regulation of cell proliferation and cell survival is one major task of the PI3K/AKT/mTOR signaling pathway. Dysregulation of these functions belongs to the disease characteristics of cancer and the PI3K/AKT/mTOR pathway has been identified as a key player driving carcinogenesis in various tumor entities such as breast cancer and renal cell carcinoma [23]. Within this study, the protein expression of PI3K/AKT/mTOR pathway members was analyzed and activation of this pathway in astrocytomas could be confirmed [24][25][26]. We found a significantly increased protein expression of p-AKT and p-PTEN in gliomas compared to CCBT. Increased expression levels of mTOR pathway members have been already extensively reported previously [24][25][26].
As consequence of the deregulation of the PI3K/AKT/mTOR signaling in gliomas, the eIF signaling cascade might also play an important role during gliomagenesis ( Figure 5A). 4E-BP1, the link between those two pathways, has been shown to correlate with the astrocytoma grade, as 4E-BP1 was significantly higher expressed in HGG compared to LGG [18], which was confirmed by our data. Tumor type-specific eIF4H splicing variant expression was described previously for colon cancer cell lines [30]. However, this has to be clarified in further analyses. Additionally, high eIF4H gene expression was negatively associated with the OS of GBM patients. eIF4H is an RNA binding protein that interacts with eIF4A, thereby stimulating its helicase activity [31,32]. As eIF4H and its interaction partner eIF4A1 seem to be relevant in gliomas, inhibiting the interaction between those two factors might be a novel approach in glioma therapy.    Several eIFs have already been identified as biomarkers for various cancer entities including gliomas [12]. However, research has been mainly focused on the investigation of eIF3 subunits. We wanted to close the gap and analyze not only eIF3 subunits, but also the eIF4F complex, and other eIF subunits. Additionally, we wanted to investigate protein as well as mRNA levels to get a detailed overview on eIF expression patterns in gliomas. For the basic characterization, we compared eIF expression levels of astrocytomas to CCBT. CCBT reveals a non-uniform distribution of cells and neurons and does not only contain astrocytoma cells of origin. CCBT has already served as control in previous studies [14,19].

(1) Immunoblot Analysis (2) qRT-PCR Analysis
In the course of the basic characterization, previous alterations in the eIF signaling cascade were confirmed and additional eIFs altered in gliomas were identified. Upregulation of eIF3B [13] has been described previously in glioma patient samples. eIF3C [14], eIF3D [15] and eIF4E [18] even correlated with tumor grade as their expression was significantly higher in HGG than in LGG. In silico, increased gene expression over all four tumor grades was determined for eIF3B, eIF3I, eIF3K and eIF3M [21]. Additionally, those four factors have been associated with poorer OS in glioma patients [21]. We were able to confirm a significantly increased protein expression of eIF3B and eIF3D. For eIF3C and eIF4E an increase in the protein and mRNA levels was found; however, this was not statistically significant. Besides the already known eIFs, we demonstrated significantly elevated protein levels of eIF3D, eIF3H, eIF3I, eIF3M, p-eIF4G, eIF4H, eIF5 and eIF6 in astrocytoma tissue compared to CCBT. Notably some of these findings were not consistent in Western blot and immunohistochemistry. Differing results in protein expression might partly be explained by methodological differences. However, these findings have to be interpreted carefully and their biological significance needs to be in the focus of future research.
Based on significant differences between astrocytomas (WHO grades I-IV) and CCBT, we started a selection process for promising eIFs that could serve as novel therapeutic targets or prognostic biomarkers ( Figure 5B). After immunoblot analysis, eIF3B, eIF3I, eIF4A1, eIF4H, eIF5 and eIF6 were selected for further detailed analyses. All six subunits were also significantly upregulated in mRNA level. However, eIF3I and eIF4H were the only analyzed eIFs that were significantly associated with survival differences of glioma patients. Therefore, eIF3I and eIF4H were chosen as most promising candidates for more additional immunohistochemical analysis.
eIF3I is a member of the eIF3 complex and was at first identified as TGF-β receptor II interacting protein 1 (Trip1) [27,28] and was shown to promote cell proliferation and angiogenesis in zebrafish embryos [29]. eIF3I was previously identified as a prognostic biomarker in IDH1/2 mutant LGG based on TCGA and CGGA database analyses [21]. We were able to confirm its impact on the OS of LGG patients in the TCGA database. Additionally, we also showed for the first time a stepwise increase of eIF3I expression over all four tumor grades in protein level.
Expression of eIF4H was significantly increased for mRNA level and protein level using immunohistochemistry. Interestingly, immunoblot analyses revealed an additional band at approximately 17 kDa, only present in astrocytoma samples. The product at 17 kDa might have been a degraded form of eIF4H or an astrocytoma-specific splicing variant. Tumor type-specific eIF4H splicing variant expression was described previously for colon cancer cell lines [30]. However, this has to be clarified in further analyses. Additionally, high eIF4H gene expression was negatively associated with the OS of GBM patients. eIF4H is an RNA binding protein that interacts with eIF4A, thereby stimulating its helicase activity [31,32]. As eIF4H and its interaction partner eIF4A1 seem to be relevant in gliomas, inhibiting the interaction between those two factors might be a novel approach in glioma therapy.

Patient Samples
The study was reviewed and approved by the institutional ethics committee of the Medical University of Graz (MUG) according to Austrian and European law (24-402 ex 11/12). As controls for biochemical and immunohistochemical (IHC) analysis, non-neoplastic ("healthy") cortical control brain tissue was collected post-mortem at the Department of Pathology of the MUG (biochemical analyses: n = 12, immunohistochemistry: n = 15). Astrocytoma samples for biochemical (pilocytic astrocytoma WHO grade I: n = 10, diffuse astrocytoma WHO grade II: n = 13, anaplastic astrocytoma WHO grade III: n = 8, GBM WHO grade IV: n = 13) and immunohistochemical (pilocytic astrocytoma WHO grade I: n = 18, diffuse astrocytoma WHO grade II: n = 26, anaplastic astrocytoma WHO grade III: n = 22, GBM WHO grade IV: n = 22) analysis were obtained retrospectively (before 2016) from the Biobank of the MUG and the Brain Biobank of the Division of Neuropathology, Neuromed Campus, Kepler University Linz, Austria. As all samples were collected retrospectively before 2016, they were neuropathologically classified by board-certified neuropathologists (J.H. and S.W.) according to the 2007 WHO classification [33]. Information about age, gender and IDH mutation status was collected for immunohistochemical analysis (Table S1).

Tissue Processing for Biochemical Analyses
Human tissue samples were collected during frozen section, immediately frozen in liquid nitrogen and stored until further biochemical processing. Frozen tissue samples were homogenized in NP-40 lysis buffer (0.05 M Tris-HCl, 5 mM NaCl, 0.5% NP-40, 0.1 mM Pefabloc ® , 1 mM DTT) supplemented with cOmplete™ Protease Inhibitor Cocktail and PhosSTOP™ phosphatase inhibitor using the MagNA Lyser homogenizer (all Roche Diagnostics, Risch-Rotkreuz, Switzerland). Protein concentration was determined using the Bradford protein assay (Bio-Rad Laboratories GmbH, Munich, Germany). NP-40 lysates were used for protein and mRNA analyses.

Immunoblot Analysis
Immunoblot analysis was performed as described previously [34]. Briefly, total protein lysates (30 µg/sample) were used for SDS gel electrophoresis. Primary antibodies were incubated in 5% bovine serum albumin (BSA) in TBS-T (antibody dilutions listed in Table S3) over night at 4 • C. Incubation with a horseradish peroxidase conjugated secondary antibody (dilutions: anti-mouse 1:3000, anti-rabbit 1:5000; GE Healthcare Life Sciences, Buckinghamshire, UK) was carried out for 1 h at room temperature. For membrane stripping, immunoblots were incubated in Restore PLUS Western blot stripping buffer (Thermo Fisher Scientific, Waltham, MA, USA) for 50 min at room temperature under shaking conditions. Semi-quantitative evaluation of immunoblots was performed using ImageJ software for densitometric analyses [37]. Relative densities were calculated by normalizing density values for each protein to the GAPDH loading control.

Patient Overall Survival Analysis
eIF gene expression and patient data from the The Cancer Genome Atlas database were downloaded from the Genomic Data Commons (GDC) (https://gdc.cancer.gov/ (accessed on 5 June 2020)). Upper quartile normalized gene expression data from "Brain Lower Grade Glioma (LGG)" (516 cases; last updated 2 May 2016) and "GBM" (528 cases; last updated 27 May 2016) were used for the analyses. Samples were categorized into high and low gene expression with expression values above or below the median, respectively. Kaplan-Meier survival curves were generated for every gene using the survival package [39] and survminer package [40] in R 3.3.0 [41]. The survival curves of patients with high and low gene expression were compared by the log-rank test. The significance level was set at p < 0.05.
Additional datasets from TCGA (GBMLGG) and from CGGA (693) were used for validation. Oligodrendroglioma data was removed from the TCGA datasets. Groups based on gene expression were dichotomized. A mutation for a specific gene was considered as a high expression for TCGA data. For CGGA data the expression for a gene was weighted by the clinical information if an IDH, respectively a 1p19q co-deletion, occurred. The OS of the group with the higher expression of EIF3I, respectively EIF4H, was compared to the group with the higher expression of ATRX, CDKN2A, EGFR, EIF4A1, EIF4H, EIF5, EIF6, IDH1, IDH2, MAP2, MGMT, NF1, PDGFRA, PIK3R1, PTEN and TP53. EIF3I and EIF4H were combined and the OS of the highest expressions were used, independent of the data taken from EIF3I or EIF4H. The OS from the group with the upper half expression was compared to the OS of the group with the lower half and p-values from the group comparisons were calculated. C# scripts were used to process the data. Using survcorr from the R package SurvCorr, the rho correlation coefficient between the survival curves was calculated.

Statistical Analysis
Statistical analyses and graphs were generated using Graph Pad Prism 4.03 (GraphPad Software Inc., San Diego, CA, USA). All data were tested for normal distribution using the Kolmogorov-Smirnoff test. For evaluation of biochemical and imm unohistochemical data, one-way analysis of variance (ANOVA) was used followed by Bonferroni's multiple comparison test (normally distributed data), or the Kruskal-Wallis test followed by Dunn's multiple comparison test (not normally distributed data), to analyze the eIF expression in astrocytomas of WHO grades I-V compared to control tissue. Kaplan-Meier curves of patients with high and low gene expression were compared with the log-rank test. Significance levels were set at p < 0.05. Proportion scores were calculated to analyze the impact of eIF protein expression on patients' OS.

Conclusions
In conclusion, our study provides a profound basic characterization of various eIFs in astrocytomas of WHO grades I-IV. We showed that the expression of eIF3B, eIF3I, eIF4A1, eIF4H and eIF6 was significantly increased in astrocytomas, in particular in GBM, for protein and mRNA levels. However, the OS of glioma patients only correlated for eIF3I with LGG and for eIF4H with GBM. Therefore, targeting of eIF3I and eIF4H might represent a prospective approach towards improvement of glioma therapy. However, to finally establish eIF3I and eIF4H as therapeutic targets, gain-and loss-of-function experiments are required to assess detailed pathophysiological consequences.
Supplementary Materials: The following are available online at https://www.mdpi.com/2072-6 694/13/6/1482/s1: Figure S1: Densitometric analysis of eIF immunoblots in astrocytomas (WHO grades I-IV) compared to CCBT. Figure S2: Influence of eIF gene expression on patients' OS in LGG (WHO grades I-II) and GBM (WHO grade IV). Figure S3: Immunohistochemical evaluation of eIF3I and eIF4H protein expression in astrocytomas (WHO grades I-IV) compared to CCBT in consideration of the IDH1 (R132H) mutation status. Figure S4: Influence of eIF protein expression on patients' OS in LGG (WHO grades I-II) and HGG (WHO grades III-IV). Table S1: Descriptive statistics of clinicopathological characteristics and immunohistochemical evaluation. Table S2: Oligonucleotides used for qRT-PCR analyses. Table S3: Primary antibodies used for immunoblot and immunohistochemical analyses. Antibody dilutions for immunohistochemical analyses are listed in parentheses. Table S4: Correlation coefficients comparing EIF3I and EIF4H with glioma-relevant genes. Table S5: Correlation coefficients of high and low expression of glioma-relevant genes. Table S6