Detection of EGFR Mutation Distribution and Transcriptional Variants in IDH-Wildtype High-Grade Gliomas Using a Next-Generation Sequencing Oncopanel

\Purpose: To detect the epidermal growth factor receptor gene (EGFR) mutation prole and transcriptional variants in high-grade gliomas (HGGs), we sequenced EGFR and evaluated the EGFR splicing prole using a next-generation sequencing (NGS) oncopanel. Methods: We analyzed 124 HGGs—10 grade (cid:0) IDH-wildtype anaplastic astrocytomas (AAs) and 114 grade (cid:0) IDH-wildtype glioblastomas (GBMs). Results: The EGFR mutations were observed in 6.0% of grade (cid:0) GBMs and in 33% of grade (cid:0) AAs. Four cases harbored missense mutations in the EGFR kinase domain (L747A, S768I, V774M, and T790M). A total of 25% of the GBMs showed EGFR amplication. Moreover, 27% of the EGFR mutations occurred in the kinase domain. EGFRv(cid:0) positivity was detected in 8.0% of EGFR-amplied GBMs. We identied two other EGFR variants in GBM cases with deletions of exons 6–7 (Δe 6-7) (one case) and exons 2–14 (Δe 2-14) (two cases). Interestingly, in one case, the initial EGFRvIII mutation transformed into an EGFR Δe 2-14 mutation during recurrence. The frequency of EGFR alterations in our cohort was lower but the frequency of EGFR mutations in the kinase domain in our cohort was higher than that in The Cancer Genome Atlas and Memorial Sloan Kettering Cancer Center cohorts. Conclusions: We suggested that the EGFR gene proles of GBM differ among cohorts and identied rare EGFR variants with longitudinal and temporal transformations of EGFRv(cid:0).


Introduction
The epidermal growth factor receptor gene (EGFR) is the most commonly ampli ed and overexpressed proto-oncogene in glioblastoma multiforme (GBM) and causes the most malignant subtype. EGFR ampli cation is detectable by next-generation sequencing (NGS) in approximately 40-50% of GBM cases [1][2][3], and 14.4-26% of GBMs harbor EGFR mutations [2,3]. Recently, the Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy (cIMPACT-NOW) proposed a diagnostic entity of grade II-III isocitrate dehydrogenase (IDH)-wildtype astrocytoma that should behave similar to GBM. The molecular features de ning this new tumor entity include telomerase reverse transcriptase promoter (TERTp) mutation, EGFR ampli cation, or a combination of whole chromosome 7 gain and whole chromosome 10 loss [4]. EGFR may serve as a diagnostic and prognostic biomarker. Indeed, EGFR has garnered interest as a drug target in GBM because of the high frequency of EGFR alterations in this disease. However, GBMs respond poorly to EGFR inhibitors [5,6]. As EGFR mutations in GBM occur in the extracellular domain rather than the intracellular kinase domain, EGFR inhibitors that target the kinase domain are less effective [6].
To evaluate the EGFR mutational and splicing pro le in IDH-wildtype high-grade gliomas (HGGs), we performed NGS analysis using a custom gene panel that we recently reported [15]. In this study, we collected a higher number of samples to clarify EGFR mutational distribution and developed a method to detect EGFR transcriptional variants.

Materials And Methods
Diffuse glioma samples One hundred and twenty-four formalin-xed para n-embedded (FFPE) tumor tissue samples were selected from the Central Nervous System Tumor Tissue Bank at Kagoshima University Hospital. The study was approved by the Institutional Review Board of Kagoshima University (approval number: 180104) and complied with the tenets of the Declaration of Helsinki. Informed consent was obtained from all patients. Resected tumors were xed with phosphate-buffered 10% formalin within 24 h of sampling and routinely processed for para n embedding, followed by sectioning for hematoxylin and eosin staining. All tumors were originally classi ed according to the World Health Organization classi cation of 2016. The tumor series consisted of 10 grade III IDH-wildtype anaplastic astrocytomas (AAs) and 114 grade IV IDH-wildtype GBMs. All tissues were histologically evaluated by board-certi ed pathologists (M.K. and A.T.) to ensure an estimated tumor cell content of 30% or more. In all patients, when analyzing copy number variations, we sequenced leukocyte DNA for comparison against matched tumor DNA.

DNA extraction and quanti cation
For DNA preparation from FFPE samples, we used the Maxwell 16 FFPE Tissue LEV DNA Puri cation kit (Promega) according to the manufacturer's instructions. Thereafter, the concentration of DNA was measured using a Qubit 3.0 Fluorometer dsDNA BR Assay kit (Life Technologies), and DNA quality was monitored using the QIAseq DNA QuantiMIZE kit (QIAGEN). The extracted DNA was diluted to a concentration of 5-10 ng/μL as a template, and polymerase chain reaction (PCR) was performed using the QIAseq DNA QuantiMIZE kit.
Next-generation sequencing NGS was performed using an amplicon-based glioma-tailored gene panel as described previously [15]. Amplicon sequences were aligned to the human reference genome GRCh37 (hg19) in the target region of the sequence. Data were analyzed using the QIAGEN Web Portal service (https://www.qiagen.com/).

Detection of EGFR variants
In all patients, to analyze rearrangement events as signi cant targets of intragenic copy number breakpoints, we sequenced leukocyte DNA for comparison against matched tumor DNA. The EGFR variants were detected by reduced read counts obtained for EGFR sequences in tumor DNA relative to those in matched leukocyte DNA. In addition, as a validation method, the EGFR variants were detected by determining the copy number (CN) of each exon in EGFR and identifying exon sites with -2SD below in CN compared with the average CN of all exons.

Complementary DNA analyses of clinical specimens
The total RNA from FFPE samples of clinical specimens was extracted using the Maxwell 16 LEV RNA FFPE Puri cation kit (Promega) and converted into cDNA using the ReverTra Ace qPCR RT Kit & Master Mix (TOYOBO Inc., Osaka, Japan). The cDNA fragments around the targeted site were ampli ed by PCR using the KOD One PCR Master Mix (TOYOBO Inc.) using the primers listed in Supplementary Table S1. The PCR products were analyzed by electrophoresis on 2% agarose gels. The PCR fragments were puri ed using the Exo-CIP Rapid PCR Cleanup kit (New England Biolabs Inc., Ipswich, MA) and analyzed by GENEWIZ Japan Corp.

Immunohistochemistry
Surgical specimens were xed within 10 min of excision in 10% neutral buffered formaldehyde for 24 h, embedded in para n, cut into 3-μm-thick sections, and mounted on glass slides coated with poly-l-lysine. Subsequently, the sections were probed with 1:25 anti-EGFR (mouse monoclonal antibody, clone EGFR.113; Leica Biosystems) and 1:200 anti-EGFRv (rabbit monoclonal antibody, cat#64952; Cell Signaling Technology). The sections were then stained with diaminobenzidine tetrahydrochloride and hematoxylin.

TCGA and MSKCC data analyses
We retrieved the molecular characteristics of the GBM cohort from previous publications. After excluding H3F3A, IDH1/2, and BRAF V600E-mutant cases, we analyzed 343 cases from The Cancer Genome Atlas (TCGA) cohort and 456 cases from the Memorial Sloan Kettering Cancer Center (MSKCC) cohort [16]. All cases were conclusively diagnosed as IDH-wildtype GBM using the cBioPortal for Cancer Genomics (https://cbioportal.org).

De nition of actionable alterations
Actionable gene alterations were predicted to confer sensitivity to either an approved targeted agent or an experimental targeted agent currently under clinical trials. The JAX Clinical Knowledgebase (JAX-CKB) (https://ckb.jax.org/) was used to identify clinically actionable variants.

Data analysis
We used OncoPrinter (cbioportal.org/oncoprinter) and MutationMapper (cbioportal.org/mutation_mapper), which are tools in the cBioPortal for Cancer Genomics, to visualize and analyze our data [17,18]. Statistical analyses were performed using JMP Pro v13 software (SAS Institute, Cary, NC). Groups were compared using the χ 2 test. Differences were considered signi cant at p < 0.05.

Longitudinal and temporal transformation of EGFRv
As an example, a 57-year-old woman with complaints of seizures underwent resection for a lesion in the left temporal lobe with high FLAIR signals observed by magnetic resonance imaging and high methionine accumulation observed by positron emission tomography. The pathological diagnosis was IDH-wildtype AA with an EGFRv (Δe 2-7) variant (Fig. 3a). Two years later, a tumor recurred in the same area, and the patient underwent another resection. The pathological diagnosis was IDH-wildtype GBM with EGFR exon 2-14 deletion (Δe 2-14) without EGFRv (Δe 2-7) (Fig. 3b). In other words, the initial EGFRv (Δe 2-7) mutation transformed into EGFR exon 2-14 deletion (Δe 2-14), with extensive deletion of exons 8-14 at the time of recurrence.

Discussion
In this study, we collected higher number of samples than in our previous study [15]. We found that the frequency of EGFR mutations was low but the frequency of EGFR mutations in the kinase domain was high in wildtype-GBMs. Moreover, we developed a method to detect EGFR transcriptional variants; we found a novel EGFR transcriptional variant and identi ed a case with a rare EGFR variant with longitudinal and temporal transformations of EGFRv . EGFR ampli cation is detectable in approximately 40-50% of all GBMs [1][2][3]. The frequency of EGFR alterations in our cohort was lower than that in TCGA and MSKCC cohorts. However, a study from Japan reported that 25.5-33.1% of GBMs have EGFR ampli cations [19,20], which is consistent with our ndings. Recent reports indicate that 14.4-26% GBMs harbor EGFR mutations [2,21]; however, the EGFR mutation frequency in GBMs in our cohort (6.0%) was considerably lower than that in the previous study cohort. These results suggest that EGFR alterations may be less frequent in Japan than in other countries. Moreover, in our study, EGFR A289D/T/V was the most common missense mutation, and two cases harbored EGFR T790M missense mutations in the EGFR kinase domain. Previous studies have shown that approximately 4% EGFR mutations in GBM have mutations in the kinase domain [2,21]; however, in our study, 27% EGFR mutations were in the kinase domain in GBMs, and this is higher than that reported previously.
Two cases (1 GBM and 1 AA) harbored EGFR T790M missense mutations in the kinase domain. EGFR T790M , which is commonly observed in lung cancer [22], is a very rare mutation in glioma, with only one case of GBM reported previously [23]. Osimertinib is an oral, third-generation tyrosine kinase inhibitor (TKI) that irreversibly inhibits EGFR and was developed speci cally to target the EGFR T790M -resistant mutation in EGFR-mutated non-small-cell lung cancer [22, 24, 25]. Makhin et al. reported a GBM case with two EGFR point mutations (C628F and A289V) that responded well to osimertinib. Thus, glioma cases with mutations in the EGFR kinase domain may bene t from EGFR TKIs [26]. In our study, a high EGFR ampli cation was identi ed in 5/10 (50%) patients with IDHwildtype AA, consistent with the ndings of a previous study [27]. In patients with IDH-wildtype AA, 33% EGFR mutations (2/6 mutation sites) had T790M and S768I mutations in the kinase domain. However, in IDH-wildtype AA, the distribution of mutations within the EGFR coding sequence has not been reported.
EGFRv is the most common EGFR splice variant. EGFRv activates multiple downstream signaling pathways and exhibits high tumorigenic potential [8, 11,12]. Recent reports indicate that 50-60% EGFR-ampli ed GBMs harbor EGFRv variants [9,10]. EGFRv positivity was detected in 2 of the 25 EGFR-ampli ed glioblastomas (8.0%) in our study, suggesting that EGFRv positivity may have been less frequent in our study than in previous studies. These discrepancies may be due to ethnic differences in patient cohorts or differences in analysis methods; we used NGS and not immunohistochemistry or RT-PCR. A comparative study found that RT-PCR was more sensitive and speci c than immunostaining using two different EGFRv -speci c antibodies [28]. Another study showed that the sensitivity of NGS-based EGFRv detection is lower than that of immunohistochemistry or RT-PCR, re ecting that EGFRv may be restricted to small subclones of glioma cells, which may not lead to a detectable reduction in exon 2-7 gene dosage [29].
In this study, we identi ed one uncharacterized EGFR variant with deletions of exons 6-7 (Δe 6-7). Deletions of exons 2-14 (Δe 2-14) constitute a very rare variant in glioma, with only one case of GBM having been reported previously [30]. In one case, the initial EGFRv (Δe 2-7) mutation transformed into a EGFR exon 2-14 deletion (Δe 2-14) at the time of recurrence. Recent reports indicate that 16-59% of GBMs that were initially EGFRv positive, lost EGFRv at recurrence [9,10,14]. Some reports suggest that the frequency of EGFRv loss at recurrence is altered by the treatment received [31], whereas others suggest that it is not dependent on the treatment received [14]. In addition, it has been reported that the expression of EGFRv is a result of epigenetic regulation [32], but the mechanism by which EGFRv is lost at recurrence is unknown. Based on our ndings, it is possible that the initial EGFRv mutation is transformed to other variants during recurrence. The functional signi cance of the novel EGFR variants needs to be analyzed in further studies.
In conclusion, we report the distribution of mutations within the EGFR coding sequence and two novel EGFR variants, one of which showed longitudinal and temporal transformation of EGFRv . In addition, we showed that the frequency of driver gene alterations in GBMs differs across cohorts. Thus, to implement personalized medicine, it is necessary to accurately assess the genetic pro le of each cohort.

Declarations Funding
The authors received no nancial support related to this study.

Con icts of interest/Competing interests
The authors declare that they have no con ict of interest.

Availability of data and material
All data used and analysed in the current study are available from the corresponding author on reasonable request.   Supplementarytable.S1.xlsx supplementary g.1.pdf