LABORATORY INVESTIGATION J Neurosurg 128:911–922, 2018

Genetic landscape of sporadic vestibular schwannoma

*Aril Løge Håvik,1–3 Ove Bruland, MSc, PhD,2 Erling Myrseth, MD, PhD,4 Hrvoje Miletic, MD, PhD,5–7 Mads Aarhus, MD, PhD,8 Per-Morten Knappskog, MSc, PhD,2,3 and Morten Lund-Johansen, MD, PhD1,4,6

Departments of 1Clinical Medicine, 3Clinical Science, and 7Biomedicine, and 6K.G. Jebsen Brain Tumor Research Center, University of Bergen; 2Center for Medical Genetics and Molecular Medicine, and Departments of 4Neurosurgery and 5Pathology, Haukeland University Hospital, Bergen; and 8Department of Neurosurgery, Oslo University Hospitals, Ullevål Sykehus, Oslo, Norway

OBJECTIVE Vestibular schwannoma (VS) is a benign tumor with associated morbidities and reduced quality of life. Except for in NF2, the genetic landscape of VS remains to be elucidated. Little is known about the effect of Gamma Knife radiosurgery (GKRS) on the VS genome. The aim of this study was to characterize mutations occurring in this tumor to identify new genes and signaling pathways important for the development of VS. In addition, the authors sought to evaluate whether GKRS resulted in an increase in the number of mutations. METHODS Forty-six sporadic VSs, including 8 GKRS-treated tumors and corresponding blood samples, were sub- jected to whole-exome sequencing and tumor-specific DNA variants were called. Pathway analysis was performed using the Ingenuity Pathway Analysis software. In addition, multiplex ligation-dependent probe amplification was performed to characterize copy number variations in the NF2 gene, and microsatellite instability testing was done to investigate for DNA replication error. RESULTS With the exception of a single sample with an aggressive phenotype that harbored a large number of muta- tions, most samples showed a relatively low number of mutations. A median of 14 tumor-specific mutations in each sample were identified. The GKRS-treated tumors harbored no more mutations than the rest of the group. A clustering of mutations in the cancer-related axonal guidance pathway was identified (25 patients), as well as mutations in the CDC27 (5 patients) and USP8 (3 patients) genes. Thirty-five tumors harbored mutations in NF2 and 16 tumors had 2 mutational hits. The samples without detectable NF2 mutations harbored mutations in genes that could be linked to NF2 or to NF2- related functions. None of the tumors showed microsatellite instability. CONCLUSIONS The genetic landscape of VS seems to be quite heterogeneous; however, most samples had mutations in NF2 or in genes that could be linked to NF2. The results of this study do not link GKRS to an increased number of mutations. https://thejns.org/doi/abs/10.3171/2016.10.JNS161384 KEY WORDS vestibular schwannoma; NF2; axonal guidance; next-generation sequencing; molecular biology; USP8; CDC27

estibular schwannomas (VSs) make up 8% of in- ing VS pathobiology is the linkage of both sporadic VS tracranial tumors, with an annual incidence rate (sVS) and familial VS to a loss of the tumor suppressor ranging from 11 to 22 per million.10,58 They lead to merlin, which is encoded by the NF2 gene.49,62 aV set of well-characterized complaints. Affected individu- Merlin mediates contact inhibition and suppresses tumor- als report reduced quality of life, but the life expectancy of igenesis (both at the and in the nucleus) patients with VSs is similar to that of the average popula- through its interaction with integrins and tyrosine tion.8,37,56 , among other .28,41,69 The proportion of To date, the most important contribution to understand- sVSs harboring NF2 mutations ranges from 15% to 84%.13

ABBREVIATIONS GKRS = Gamma Knife radiosurgery; IPA = Ingenuity Pathway Analysis; LOH = loss of heterozygosity; MLPA = multiplex ligation-dependent probe amplification; MSI = microsatellite instability; PCR = polymerase chain reaction; SNV = single-nucleotide variant; sVS = sporadic VS; VS = vestibular schwannoma; WES = whole-exome sequencing. SUBMITTED May 30, 2016. ACCEPTED October 31, 2016. INCLUDE WHEN CITING Published online April 14, 2017; DOI: 10.3171/2016.10.JNS161384. * Drs. Knappskog and Lund-Johansen contributed equally to this work.

©AANS 2018, except where prohibited by US copyright law J Neurosurg Volume 128 • March 2018 911

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC A. L. Håvik et al.

This might indicate that alternative mechanisms for de- Mini Kit (QIAGEN). DNA from blood was extracted regulating NF2-related functions in schwannoma cells ex- using QIAsymphony (QIAGEN). The DNA quality and ist, leading us to search for mutations in genes linked to quantity were evaluated with 1% SeaKem gel electropho- merlin. resis and NanoDrop spectroscopy (Thermo Fisher Scien- In addition, we directed our attention toward another tific), respectively. candidate, CAV1, to see whether mutations in this gene or in genes controlling its expres- Whole-Exome Sequencing sion can provide an explanation for its universal down- 1 WES was performed as a custom service (HudsonAl- regulation in sVSs. Except for the hereditary disease pha Institute for Biotechnology). The capture was Nim- associated with VS, Type 2, little is bleGen SeqCap EZ Exome Library v3.0, and sequencing known about the development of VS. Numerous studies was paired-end sequencing (2 × 100 bp) with approxi- have been conducted in the search for an etiology, but with × 15,43,51,53 mately 85 coverage on Illumina HiSeq. The resulting inconclusive findings. Recently, Agnihotri et al. se- reads were aligned to the GRCh37/hg19 reference genome quenced the exomes of a limited number of VSs and spi- 27 3 with Burrows-Wheeler transform, and the resulting bam- nal schwannomas. To gain a better understanding of the files were delivered to us. underlying pathogenesis of sVS formation, we performed whole-exome sequencing (WES) of sVSs and normal cells Bioinformatic Analyses to identify tumor-specific mutations. We applied the Genome Analysis Toolkit (version The increasing use of radiosurgery in treatment of VS 33 has raised concerns in the neurosurgical community about 3.2) base quality score recalibration, indel realignment, radiation-induced mutational events leading to malignant duplicate removal, and we performed single-nucleotide transformation.17,31,35,39,48,54 So far, a few cases of postir- polymorphism and INDEL calling across all 92 samples radiation malignant VSs have been described. Our study simultaneously using variant quality score recalibration therefore aimed not only to study sVS, but also in particu- according to Genome Analysis Toolkit Best Practices lar to study the mutational status of cases of VS where the recommendations.14,63 For detection of somatic point tumor had undergone radiosurgery. substitutions with low allelic fraction, we used MuTect.12 ANNOVAR64 was used for functional annotation of the variants called from the 46 tumor-normal pairs. Methods Candidate variants were filtered according to the fol- Patients and Specimen Collection lowing criteria: 1) allelic frequency < 0.05 in 1000 g, Samples of VS tissue and EDTA-treated blood were esp6500 and exac03; 2) repeated alleles < 5, evaluated obtained from 46 patients who underwent first-time sub- manually in borderline cases; 3) allelic frequency in tu- occipital resection of a unilateral VS at the Department mor > 0.05, allelic frequency in germline < 0.05, sequenc- of Neurosurgery, Haukeland University Hospital, between ing depth > 7, alternative allele > 1, evaluated manually in August 2003 and October 2014 (Table 1). Five pieces of borderline cases. The allelic frequency threshold was set viable tumor tissue with diameters of 3–5 mm were col- to exclude normal variants unlikely to be pathogenic. We lected from the intracapsular part of the tumor. This was excluded areas prone to segmental duplication by setting a done in 1 session, and the tissue was collected from 1 tu- cutoff value for mapping quality to 30 (phred score). mor area only. Routine histology was done in each case. To increase the likelihood of detecting mutations im- Eight patients had been treated with Gamma Knife radio- portant to tumorigenesis, we chose to evaluate nonsyn- surgery (GKRS) for the same VS. GKRS was performed onymous single-nucleotide variants (SNVs); frameshift according to a standardized protocol of 12 Gy to the tumor indels; and stop-gain, stop-loss, and splice-site mutations periphery at minimum 95% coverage. Six patients exhib- as predicted by ANNOVAR. The missense SNVs also had ited cystic VS. Dexamethasone 4 mg, administered orally to be predicted as deleterious/probably damaging/disease 4 times for 1 day preoperatively, was routinely given, start- causing/functional (positive prediction) by one of the fol- ing the day before surgery. Written informed consent was lowing prediction algorithms used in ANNOVAR: SIFT, received from all patients before tissue harvesting, and the PolyPhen 2 HDIV and HVAR, LRT, Mutation Taster, study was approved by the regional ethical committee for Mutation Assessor, FATHMM, MetaSVM, and MetaLR. medical research in Western Norway. The samples were GERP++, PhyloP, and SiPhy annotations were used to as- either immediately snap-frozen or stored at -80° and then sess how conserved the mutated sites were across different transferred to liquid nitrogen in the Bergen Neurosurgical species. Tissue Bank at Haukeland University Hospital. One tu- After filtering of variants, the remaining candidates mor-normal pair was found to be an outlier, accounting for were validated visually using the Integrative Genomics a large number of called variants, and was hence investi- Viewer.45,60 The outlier was subjected to stricter filtering gated with microsatellite-based polymerase chain reaction criteria to reduce the number of possible passenger mu- (PCR) profiling to confirm that the samples were matched. tations: ≥ 2 of the aforementioned prediction algorithms for nonsynonymous SNVs had to be positive. The variants DNA Extraction passing these filtering criteria were used for subsequent For DNA extraction, tumor tissue was first disrupted analyses. For pathway analyses, we used QIAGEN’s Inge- using the TissueLyser (QIAGEN) and treated with prote- nuity Pathway Analysis (IPA). The samples were pooled ase. Next, DNA was extracted using the QIAamp DNA to detect clustering of mutations in relevant signaling path-

912 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC Whole-exome sequencing of sporadic vestibular schwannoma

TABLE 1. Patient demographic data and clinical characteristics » CONTINUED FROM PREVIOUS COLUMN

Time No. VS48 VS M 54 40 0 17 Since Tumor of GKRS Age Diameter NF2 * Largest tumor diameter. ID Diagnosis (mos) Sex (yrs) (mm)* Hits† Mutations‡ † Number of tumor-specific mutations, indels, or loss of alleles in NF2 as detected by exome sequencing and MLPA. VS1 VS F 58 29 2 11 ‡ Number of genes with tumor-specific mutations as detected by exome VS2 VS M 61 25 1 31 sequencing. § Preoperative GKRS. VS3 VS F 50 18 1 28 ¶ Possible clinical variant of NF2. VS4 VS M 68 31 0 9 ** Cystic appearance of tumor. VS6 VS-GK§ 33 M 61 35 1 10 VS7 VS F 67 19 0 57 VS8 VS M 58 28 2 22 ways and functions, as well as investigated on an individu- VS9 VS F 64 35 0 17 al basis. Grouping was also done according to mutational VS10 VS F 57 35 2 14 status, NF2 status, GKRS, or cystic appearance to detect differences associated with these characteristics. Identifi- VS11 VS-GK§ 114 M 58 30 2 15 cation of conserved and functional regions was done using VS12 VS F 62 27 2 17 the following reference sequences: NF2 (NM_000268); VS13 VS NF2¶ F 54 22 0 15 USP8 (NM_005154); and CDC27 (NM_001114091). VS14 VS-GK§ 35 M 66 30 1 12 To describe the status and number of mutational hits VS15 VS F 39 32 2 17 in the NF2 gene, all variants called were examined in the VS16 VS F 40 30 0 7 Integrated Genomics Viewer regardless of whether they passed the filtering criteria. A region of interest of ± VS17 VS M 59 28 2 18 2 bp was defined to increase the likelihood of mutations VS18 VS-GK§ 71 F 28 16 2 10 detected as being pathogenic. In-frame deletions were VS20 VS M 30 39 0 12 considered hits. However, synonymous SNVs and intron VS21 VS M 33 23 1 10 SNVs outside of splice donor/acceptor sites were not con- VS22 VS F 45 27 1 8 sidered hits, because it is hard to predict the outcome of VS23 VS M 48 24 1 16 these. VS24 VS-GK§ 35 M 64 35 1 17 Multiplex Ligation-Dependent Probe Amplification VS25 VS F 18 36 1 9 For detecting copy number variation in the NF2 gene, VS26 VS-GK 18 M 66 38 1 17 we performed multiplex ligation-dependent probe ampli- cystic§ fication (MLPA) according to the manufacturer’s protocol VS27 VS cystic** M 63 23 1 11 with the SALSA MLPA probe mix P044-B2 NF2.52 Data VS28 VS M 33 30 1 10 analysis was performed using the MLPA analysis tool Cof- VS29 VS-GK§ 18 F 53 26 1 16 falyser (MRC-Holland). Normal control material included VS30 VS M 55 30 2 11 chorionic villi tissue as normal for the tumor tissue and VS31 VS M 42 32 2 5 leukocyte DNA from blood donors as control for patient leukocyte DNA. Borderline results were reexamined us- VS32 VS M 57 30 1 17 ing SNVs from the WES data to determine whether there VS33 VS M 63 30 1 31 was loss of heterozygosity (LOH) in the allele in question. VS34 VS-GK 77 F 69 32 1 17 cystic§ Microsatellite Instability PCR VS35 VS cystic F 42 40 2 24 Microsatellite instability (MSI) was investigated using VS36 VS M 47 40 2 17 standard procedures with PCR amplification and capil- VS37 VS F 75 31 0 231 lary electrophoresis of the mononucleotide markers NR21, 9 VS38 VS M 66 30 2 20 NR24, NR27, BAT25, and BAT26. VS39 VS cystic F 45 40 1 9 Statistical Analysis VS40 VS M 42 27 0 4 Average and median values were calculated, with max- VS41 VS F 26 30 1 14 imum and minimum values shown in parentheses. The VS42 VS F 25 32 0 8 correlation coefficient was also calculated. To determine VS43 VS F 58 32 1 14 strength of association between a gene set and NF2, mean VS44 VS F 36 37 2 13 values for number of linking pathways and genes were VS45 VS F 58 27 2 15 calculated with standard deviations and 95% CIs. To aid VS46 VS cystic F 60 56 2 11 in determining copy number variants in NF2, tumor/nor- VS47 VS F 37 37 0 9 mal coverage ratios were calculated and normalized using mean exonic coverage for the complete reference exome. CONTINUED IN NEXT COLUMN » For significance analysis regarding linkage to specific

J Neurosurg Volume 128 • March 2018 913

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC A. L. Håvik et al. pathways or functions, p values were acquired using IPA’s genes), we discovered no significant correlations with giv- incorporated calculations. A p value < 0.05 was consid- en parameters or alterations of pathways. ered significant. With the outlier excluded, 3 genes were mutated in > 2 tumors: NF2 (n = 35, 78%), CDC27 (n = 5, 11%), and USP8 Results (n = 3, 7%). Furthermore, 16 genes were mutated in 2 tu- mors (4%): CHD4, CTAGE6, CTNNA2, EIF5B, HS6ST1, Patient Cohort KALRN, LGR5, LGSN, NAV3, OR2T3, PKD1, PLEC, Forty-six patients presenting with sVSs were included POTEJ, RAD54B, TENM2, and TTN. The mutations in in the study (Table 1). All patients had unilateral VS, and CDC27 (cell division cycle 27 gene) were clustered in none had a family history or clinical signs of neurofibro- cDNA position 754–796, corresponding to amino acids matosis Type 2. The patients in this study were among 252–266 in the predominant isoform, a location between 160 patients operated on during the period. The 8 patients the conserved domains tetratricopeptide repeat 4 and tet- who had previous radiosurgery were consecutive. The tu- ratricopeptide repeat 5 that is important for protein-pro- mor sizes and demographic data of the study patients were tein interactions. One recurrent mutation (p.G265D) was within range of all operated cases. The mean age at the found in 2 patients. Three tumors (7%) harbored missense time of surgery was 51.3 years, and the mean tumor di- mutations in the deubiquitinase gene USP8. The mutations ameter was 31 mm. Twenty-two patients were men and 24 affected a region between amino acids 764 and 798 (corre- were women. sponding to the ubiquitin-specific protease conserved site The extent of resection was total in 20, near total in 24, region) and were all predicted to be deleterious mutations and subtotal in 2 patients. Eight patients had poor postre- by several of the prediction algorithms used. None of the section facial nerve function. All patients had complete tumors harbored mutations in CAV1, but pathway analy- hearing loss on the operated side after surgery. The pa- sis revealed several associations between the gene set and tients receiving subtotal resection got early treatment with regulation of CAV1 expression. salvage radiosurgery. Among the remaining 44 patients, When pooling the 692 mutated genes from the 45 sVSs, retreatment was given in 1 case due to progression. Fol- we discovered a significant clustering of mutations in the low-up ranged from 2 to 10 years. We succeeded in col- axonal guidance pathway (Fig. 1). Twenty-five of 45 pa- lecting tissue from all patients who received prior GKRS; tients harbored a mutation in a gene implied in this path- these were rare cases and thus were important for the way, with a predominance of missense mutations. None study. The mean time between radiation and microsurgery of the genes were mutated in > 2 samples. The next top- was 50 months (range 18–114 months). None of the tumors ranked associations between the gene set and canonical demonstrated MSI. pathways included nerve (NGF) signaling, Whole-Exome Sequencing protein A signaling, breast cancer regulation by 1, and CDK5 signaling. The 5 top-ranked cel- To determine the mutational landscape of sVS, tumor lular functions associated with the gene set were cellular DNA and matched normal DNA samples were subjected growth and proliferation, cellular assembly and organiza- to WES. The 92 samples (paired tumor and blood samples tion, cellular function and maintenance, cell morphology, from 46 patients) had average exome coverage of 91 reads and cellular development (Table 2). per base; 97% of the reference exome had a coverage of 10× or more, and 95% had a coverage of 20× or more NF2 Status (Supplemental Table 1). When 1 outlier accounting for 231 mutations was excluded, a total of 716 mutations af- To determine the somatic status of the NF2 gene in our fecting 692 genes were identified. A median of 14 (4–57) sVS batch, we investigated for copy number variation by genes was mutated in each sample (Table 1, Supplemental MLPA and mutations from the WES data. In the case of Fig. 1). borderline results from MLPA, normalized ratios of cov- Among the 716 mutations, 602 were missense, 51 were erage on NF2 from tumor/normal pairs were used to aid in stop-gain, 31 were frameshift deletions, 22 were splice- determining whether an LOH was present or not. In total, site SNVs, 2 were splice-site indels, 1 was stop-loss, and 35 of the 46 (76%) sVSs harbored ≥ 1 somatic mutations 7 were frameshift insertions (detailed data on all muta- in NF2 and 16 of the 46 (35%) harbored 2 mutational hits tions available upon request). Among 676 SNVs, the most (Table 3). In total, 30 SNVs, 15 small indels, and 11 larger frequent mutations were G>A (n = 172), C>T (n = 160), deletions were present, with a predominance of nonsense and G>T (n = 125). Twelve tumors harbored mutations in SNVs, frameshift deletions, and LOH. Three mutations 13 genes coding for kinases, transcription regulators, or were recurrent throughout the batch (p.W41X, p.R57X, proteases with known inhibitors (Supplemental Table 2). and p.Y132X). We found no significant correlation between number of In total, 11 sVSs had no mutations in NF2 (non-NF2). mutations and parameters such as diameter of tumor, age, These were examined thoroughly for variants, using filter- sex, prior GKRS, cystic tumor, or NF2 status. No specific ing criteria that were not as strict. Except for a synony- mutation correlated with extent of resection, postresec- mous SNV detected in 1 sample, no other variants were tion facial nerve function, or progression-free survival. found. Excluding the single outlier without an NF2 muta- The number of mutational hits in the NF2 gene did not tion, the 10 remaining sVSs were investigated in silico for correlate with the total number of tumor-specific muta- linkage between mutated genes in the sample and NF2. tions. When stratifying the samples in groups according One tumor linked directly to NF2 from the transcrip- to number of mutated genes (0–9, 10–19, and ≥ 20 mutated tion factor GATA3, 6 tumors linked indirectly through 1

914 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC Whole-exome sequencing of sporadic vestibular schwannoma

TABLE 2. Top canonical pathways and cellular functions related to the 692 mutated genes No. of Top Pathways & Functions* p Value† Genes‡ Canonical pathways Axonal guidance signaling 5.36E−05 29 NGF signaling 8.71E−05 12 signaling 2.78E−04 25 Breast cancer regulation by 6.63E−04 15 stathmin1 CDK5 signaling 7.71E−04 10 Molecular & cellular functions Cellular growth & proliferation 5.17E−03–1.76E−09 153 Cellular assembly & organization 5.17E−03–3.91E−07 127 Cellular function & maintenance 5.17E−03–3.91E−07 186 Cell morphology 6.12E−03–4.08E−07 164 Cellular development 5.42E−03–5.15E−07 179 NGF = nerve growth factor. * Pathways and functions as defined by IPA. † Calculated with IPA’s incorporated Fisher’s exact test (right sided), which is a measure of the probability that the selected genes are associated with a pathway by chance alone. ‡ Number of genes in the VS gene set that is represented in the given path- ways and functions.

node, whereas 3 tumors linked through 2 nodes. We fur- ther divided the NF2-mutated group into 2 subgroups: 1 NF2 mutation (1hit-NF2, including 19 samples) and 2 NF2 mutations (2hit-NF2, including 16 samples). Overall, 2hit- NF2 had a stronger correlation with molecules upstream of NF2 than 1hit-NF2, although this difference was not significant. The mean fraction of the genes in each sample in non- NF2 linking to NF2 was 9.6%, and this was comparable with 1hit-NF2 and 2hit-NF2. The recurrent nodes linking genes in the samples in non-NF2 to NF2 were (n = 3), RAC1 (n = 2), ESR1 (n = 2), ubiquitin (n = 2), and GATA3 (n = 2). Although there were some discrepan- cies between the non–NF2- and the NF2-mutated groups, the recurrent nodes were mostly the same (Fig. 2). Oth- er recurrent nodes present throughout the 45 sVSs were CDC42 (n = 12), SPP1 (n = 8), CD44 (n = 6), and GRB2 (n = 4). When the mutated genes from the non-NF2 group were pooled, a significant linkage to Schwann cell prolif- eration (p = 1.02E-03) and NF2 downstream molecules (p = 2.14E-03) was observed. The genes linked to Schwann cell proliferation were spread across 3 samples and includ- ed LAMC1, MTOR, CHD4, and ATM. Mutations affecting genes downstream of NF2 were spread across 4 samples and included MTOR, MAPK3, SOS1, PRKAR1B, INADL, and ATM. The NF2-mutated group also exhibited signifi- cant linkage to Schwann cell proliferation (p = 3.24E- 03) but not to molecules downstream of NF2 (2.62E-01). None of the tumors in the non-NF2 group had a cystic FIG. The axonal 1. guidance pathway is altered The in molecules sVS. shown are implicated in the axonal guidance Knowledge pathway as defined Base. by IPA’s Molecules shaded in gray are ones in which the gene was mutated in a sample from batch. our Copyright sVS QIAGEN, Inc. (Redwood CA; City, www.qiagen.com/ingenuity). Published with color in only. online permission. available is Figure appearance. For further characterization of the non-NF2 group and comparison with the NF2-mutated group, we identified several canonical pathways (as defined by the IPA Knowl-

J Neurosurg Volume 128 • March 2018 915

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC A. L. Håvik et al.

TABLE 3. NF2 status ID Mutation AD* DP† VAF (%)‡ mRNA§ Protein¶ gPos** Exon No. of NF2 Hits†† VS1 fs del 7 18 39 c.1376_1383del p.Q459fs 30070860 13 2 VS1 nfs del 8 20 40 c.1386_1409del p.462_470del 30070870 13 VS1 LOH NF2 VS2 stop-gain 4 51 8 c.C1165T p.Q389ter 30069300 12 1 VS3 LOH NF2 1 VS4 neg 0 VS6 fs del 12 88 14 c.1122delG p.L374fs 30067937 11 1 VS7 neg 0 VS8 fs del 13 55 24 c.993delA p.R331fs 30064429 10 2 VS8 missense 25 106 24 c.A1532G p.D511G 30074270 14 VS9 neg 0 VS10 stop-gain 5 27 19 c.C396G p.Y132ter 30038223 4 2 VS10 stop-gain 22 99 22 c.G949T p.E317ter 30064385 10 VS11 splicing SNV 6 50 12 c.364-1G>A 30038190 4 2 VS11 stop-gain 15 65 23 c.C1084T p.Q362ter 30067899 11 VS12 stop-gain 29 113 26 c.C169T p.R57ter 30032794 2 2 VS12 fs del 13 63 21 c.866_867del p.K289fs 30061034 9 VS13 neg 0 VS14 LOH 1 VS15 fs del 12 51 24 c.348delT p.H116fs 30035186 3 2 VS15 missense 25 76 33 c.T623G p.L208R 30054201 7 VS15 stop-gain 30 86 35 c.C634T p.Q212ter 30054212 7 VS16 neg 0 VS17 stop-gain 10 16 63 c.C396A p.Y132ter 30038223 4 2 VS17 LOH NF2 VS18 stop-gain 18 98 18 c.C459G p.Y153ter 30050657 5 2 VS18 splicing SNV 24 121 20 c.810+1G>T 30057329 8 VS20 neg 0 VS21 stop-gain 37 80 46 c.G1639T p.E547ter 30077492 15 1 VS22 fs del 6 48 13 c.1382delC p.A461fs 30070866 13 1 VS23 fs del 13 40 33 30050657 5 1 VS24 stop-gain 6 56 11 c.G551A p.W184ter 30051617 6 1 VS25 splicing SNV 7 37 19 c.1340+2T>G 30069477 12 1 VS26 splicing SNV 17 63 27 c.810+2T>A 30057330 8 1 VS27 syn SNV 55 80 69 c.G810A p.E270E 30057328 8 1 VS27 LOH NF2 VS28 fs ins 15 39 38 c.460_461insAACA p.D154fs 30050658 5 1 VS29 splicing SNV 11 54 20 c.363+1G>T 30035202 3 1 VS30 stop-gain 7 49 14 c.C169T p.R57ter 30032794 2 2 VS30 stop-gain 19 54 35 c.G616T p.E206ter 30054194 7 VS31 stop-gain 17 84 20 c.G122A p.W41ter 30032747 2 2 VS31 splicing SNV 7 28 25 c.1341-2A>C 30070823 13 VS32 stop-gain 3 26 12 c.C331T p.Q111ter 30035169 3 1 VS33 fs del 37 71 52 c.790delA p.I264fs 30057308 8 1 VS34 fs ins 13 68 19 c.478dupC p.K159fs 30050675 5 1 VS35 nfs del 38 58 66 c.285_287del p.95_96del 30035123 3 2 VS35 LOH NF2 VS36 stop-gain 4 13 31 c.T380A p.L127ter 30038207 4 2 VS36 LOH NF2 VS37 neg 26 54 48 0

CONTINUED ON PAGE 917 » 916 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC Whole-exome sequencing of sporadic vestibular schwannoma

» CONTINUED FROM PAGE 916 TABLE 3. NF2 status ID Mutation AD* DP† VAF (%)‡ mRNA§ Protein¶ gPos** Exon No. of NF2 Hits†† VS38 missense 10 15 67 c.A367C p.K123Q 30038194 4 2 VS38 fs del 8 14 57 c.377delT p.I126fs 30038204 4 VS38 LOH NF2 VS39 stop-gain 17 83 20 c.G122A p.W41ter 30032747 2 1 VS40 neg 0 VS41 stop-gain 6 42 14 c.C1021T p.R341ter 30067836 11 1 VS42 neg 0 VS43 fs del 12 66 18 c.1607delA p.Q536fs 30077460 15 1 VS44 LOH NF2 2 VS44 del Exon 2 VS45 missense 11 41 27 c.G114C p.E38D 30000101 1 2 fs del 14 54 26 c.1009delC p.Q337fs 30067824 11 VS46 stop-gain 18 34 53 c.G1663T p.E555ter 30077516 15 2 VS46 LOH NF2 VS47 neg 0 VS48 neg 0 * Variant allelic depth. † Depth of coverage at this position. ‡ Variant allelic fraction. § Neurofibromin 2 NF2( ), transcript variant 1, NM_000268, mRNA. ¶ position according to merlin isoform 1, NP_000259.1. ** Genomic position according to GRCh37/hg19 contig. †† Number of somatic mutational hits in NF2. Two mutational hits on the same read is considered 1 hit. Synonymous and intron SNVs are not considered hits. edge Base) that were differentially associated with the 2 of a study where previous radiosurgery was not associ- groups (Supplemental Table 3). Of note, pathways signifi- ated with an increased risk of intracranial malignancy.48 cantly associated with non-NF2 and not with the NF2-mu- It should be noted, however, that WES would not have tated group included insulin receptor signaling, CNTF sig- detected chromosomal rearrangement due to radiation- naling, EGF signaling, FAK signaling, JAK/Stat signaling, induced double-strand breakage of DNA. Gamma Knife and VEGF signaling, among others. Some of the genes in treatment of VS is highly effective. All patients who re- the non-NF2 group were highly interconnected and linked ceived salvage surgery following GKRS over the study to several tumorigenesis-related pathways; these included period were included. The low number of patients previ- MAPK3, ATM, SOS1, PRKAR1B, ADCY3, MTOR, and ously treated with GKRS does not allow for any profound WNT6. When dividing the NF2-mutated group into 1hit- statistical comparisons. NF2 and 2hit-NF2 subgroups, no notable pathway corre- In recent years, several transcriptome studies using mi- lated more than expected with 1hit-NF2. croarray techniques have also added to the understanding of this disease.1,2,11,61,66 Of note, Aarhus et al. revealed al- tered expression of the tumor suppressor gene CAV1 and Discussion established the ERK pathway as the central core linking To our knowledge, we present here the largest WES the deregulated genes. Torres-Martin et al. revealed dereg- study performed on schwannomas. By comparing tumor- ulation of SPP1/MET signaling and the androgen receptor normal pairs of 46 cases of sVSs, we have elucidated the pathway. Agnihotri et al. identified overexpression of the genetic landscape of this disease. Most intriguingly, the PI3K/AKT/mTOR pathway and also validated this finding 8 patients with prior GKRS did not demonstrate a higher functionally in a schwannoma cell line. Although several number of tumor-specific mutations than the rest of the studies have investigated the mutational status of NF2, few group (Fig. 3). This is surprising, given the fact that, first, studies have been undertaken to shed light on the genetic their tumors were radioresistant, which is in itself an un- landscape of VS. The ones conducted have investigated for common feature of VS, where radiosurgery is successful copy number changes in VSs, finding loss on 22q (har- in approximately 90% of tumors, and that, second, they boring the NF2 gene) as the recurrent event and a smaller had been irradiated. number of tumors harboring aberrations on 9q34, 17q, and In addition, the types of genes mutated, and the path- 19, among others.6,19,24,65 ways and cellular functions that they were associated with, The predominant view regarding sVS pathobiology is were largely the same. These results support the findings that loss of function of merlin, encoded by the NF2 gene,

J Neurosurg Volume 128 • March 2018 917

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC A. L. Håvik et al.

FIG. 2. Recurrent nodes link the sVS gene set with NF2. The bar graph shows the recurrent nodes in 1-node pathways between sVSs and NF2 discovered using IPA. If one assumed no difference between the non-NF2 and NF2-mutated sVSs, the expected number of non-NF2 genes linking through a specific molecule would be 24% of the total. The graph shows that some of the nodes differ from that assumption, indicating differences between the groups. Figure is available in color online only. is the main driver of tumorigenesis. However, controver- Group 1 PAKs, as a potent inhibitor of tumorigenesis in sies exist because several reports indicate that a proportion neurofibromatosis Type 2–associated schwannomas, fur- of tumors have an intact NF2 gene and normal mRNA ther demonstrating the importance of these molecules.29 and protein levels.1,13,30 A few studies have demonstrated Another recurrent molecule linking our gene set with NF2 loss of the protein merlin even in the absence of a mu- is SPP1, which was found to initiate Akt-mediated phos- tated gene.49,57 Here we report, in agreement with previous phorylation of merlin at Th230 and S315, leading to its studies, that a significant number of the patients had NF2- proteasomal degradation.36 The finding of SPP1 upregula- mutated tumors. NF2 was also established as the main ge- tion in a transcriptome study of VS gives further proof of netic event in schwannoma in the study by Agnihotri et its involvement in merlin depletion in this disease.61 al.3 However, a subgroup presented without any mutations Our study provides several clues to missing merlin in this gene. function even in the absence of NF2 mutations. These data By means of pathway analysis, we have provided sev- thus support the paradigm regarding sVS biology—that eral clues to alternative mechanisms by which tumors that NF2 is the main driver of this disease even in the absence do not harbor NF2 mutations may still have impaired mer- of NF2 mutations. The predominance of truncating mu- lin function or impairment of merlin-related functions. In tations before missense mutations also lends support to total, 21 tumors linked through either Rac1 and/or Cdc42, the notion of this gene as a tumor suppressor. However, proteins belonging to the Rho family of small GTPases functional characterization of the mutated forms of these that regulate cell growth, organize the , and genes and how they affect NF2-related pathways should activate downstream protein kinases. Constitutively ac- be investigated. tive, Rac1 and Cdc42 phosphorylate Pak, which in turn In our study, we conducted a comprehensive character- phosphorylates merlin at S518, impairing its ability to bind ization of the mutational status of the NF2 gene and inves- its target proteins and thereby enabling it to function as a tigated for differential association with distinct pathways tumor suppressor.22,46,68 Manchanda et al. also showed that and/or defined sets of genes (e.g., molecules downstream activated Rac1 seemed to suppress merlin protein abun- of NF2). Although our data suggest that we can stratify the dance rather than alter its function.32 tumors into non-NF2, 1hit-NF2, and 2hit-NF2 subgroups, Licciulli et al. identified FRAX597, an inhibitor of we cannot say with certainty whether 1 or both alleles

918 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC Whole-exome sequencing of sporadic vestibular schwannoma

tumorigenesis because it has been proven to be important in several other tumors; our results warrant further inves- tigations on the subject.16 The correlation between the gene set and regulation of NF2 did not differ between the non-NF2 and NF2-mutated groups. However, we observed a more significant corre- lation between the non-NF2 group and molecules down- stream of NF2 as well as several tumorigenesis-related pathways, further suggesting alternative alterations in a group of sVSs with normal NF2. One of these pathways, insulin receptor signaling, has been implicated in VS tu- morigenesis.2,5 Our results imply that this pathway may be more important in VSs lacking mutational hits in the NF2 gene. Another pathway implicated in VS tumorigenesis, ERK/MAPK signaling, was significantly associated with non-NF2, but not with the NF2-mutated group.1,4 This sug- gests that the non-NF2 group carries some mutations that might explain why these patients still develop VS despite not having NF2 mutations. When extracting 1hit-NF2, no notable pathways were identified, suggesting that these cases harbor a second hit in NF2 not identified in the WES or MLPA data. For example, MLPA might not be sensitive enough to identify low allelic fraction deletions, and WES might miss deep intronic variants that are important for FIG. 3. GKRS did not result in an increased number of tumor-specific mutations. The scatter plot demonstrates no correlation between prior RNA splicing. GKRS and number of mutational events in the tumor as discovered by With the exception of NF2, all genes are mutated at a WES. Values on the x-axis: 1 = prior GKRS; 2 = no prior GKRS. The low percentage, indicating that deregulation of key path- outlier is excluded. ways, rather than mutations of single genes, could be im- portant. Our results demonstrate clustering of mutations in the axonal guidance pathway, suggesting that it may be a have mutational hits based on WES data because the 2 driver of sVS tumorigenesis. Torres-Martin et al. made a hits might occur on the same allele. However, the fact that similar discovery using transcriptome data from 28 sVSs 61 these mutations are tumor specific suggests that they give and 3 familial VSs. This pathway, traditionally described the cells a selective advantage. Therefore, it is not unlikely in embryonic development, has recently been implicated that 2 different mutations in the same patient affect differ- in tumorigenesis and thus warrants further functional ent alleles. characterization in VS.7,34 We also demonstrated mutations Consistent with the classic Knudson theory of tumor in genes coding for proteins with known pharmacological suppressor genes, 7 of the tumors harbored LOH in addi- inhibitors in 12 patients. However, to make use of these tion to a small mutational event, indicating hits in both al- data in a translational manner, the mutations need to be leles. For the rest of the cases, one might speculate on other functionally characterized. These mutations are present causes of impairment of normal merlin function, including in only 1 patient each, making a large-scale cost-efficient epigenetic silencing, post-translational modifications, im- screening method for functional characterization of muta- pairment of RNA splicing, or hits in related pathways and tions necessary. molecules, although epigenetic silencing of NF2 does not Interestingly, we identified 2 novel genes that were mu- seem to be a prevalent event.25,26 tated in > 2 of the 45 nonoutlier sVSs. One of them, mutat- Another issue to address is the allelic frequency by ed in 5 samples, is CDC27, a component of the anaphase- which the mutations in NF2 occur. If we use VS10 as an promoting complex. The protein product from this gene is illustrative example, one can identify a mutation in both suggested to be a tumor suppressor, making it a possible 4 and 10 with allelic fractions of 19% and 22%, re- driver in a subset of sVSs.40 Missense mutations in the spectively. If one assumes that these 2 mutations occurred deubiquitinase gene USP8 were identified in 7% of the tu- on 2 different in the same cell (thereby mors in our study. The mutations were clustered in or near cohering with the Knudson theory), only 19% of the se- the ubiquitin-specific protease-conserved site region and quenced DNA from the sample has knocked out NF2. Be- were all predicted as deleterious by the algorithms used. cause these samples are harvested from the subcapsular The protein product of this gene is required to enter the S part and examined by a neuropathologist, except for minor phase of the , positively regulates the hedgehog infiltration of blood cells and fibroblasts, these samples signaling pathway (which is implicated in tumorigenesis), consist mostly of tumor cells. Therefore, cellular heteroge- and is thought to inhibit degrada- neity within the specimens is unlikely to explain the low tion.20,38,67 allelic fraction by which NF2 is mutated. It is tempting to Row et al. reported that knockdown of this product in- suggest that clonal evolution might also play a role in VS hibited the degradation of EGFR, whereas Reincke et al.

J Neurosurg Volume 128 • March 2018 919

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC A. L. Håvik et al. reported that a missense mutation in the 14-3-3 binding tions in this gene have previously been implicated in tu- motif (amino acids 713–720) enhanced EGFR recycling morigenesis and induction of genome instability.21,50 and activity in corticotroph pituitary adenomas.44,47 These On the basis of this finding, we decided to investigate results indicate that mutations in this gene are likely to our 46 sVSs for MSI, a functional assay demonstrating have an effect on functions relevant to tumorigenesis. The replication error. A previous study found MSI in 2 of 11 ERK pathway, a pathway regulated by EGFR among oth- VSs; however, they found it to be associated with spinal ers, has previously been implicated in sVS pathobiology.1 schwannomas and not intracranial ones.55 None of the One possible explanation is that a subset of sVSs harbors tumors in our study demonstrated MSI, establishing this USP8 mutations that impair its normal function and there- phenomenon as being rare in sVS. However, the outlier by augment ERK signaling and proliferation. might have still harbored a large number of variants due Agnihotri et al. reported that ARID1A and ARID1B to impairment of DNA repair mechanisms not resulting mutations were recurrent events, although there was some in MSI. discrepancy between the frequency of mutated samples in We speculate that the correlation with the clinical be- the WES experiment and in Sanger sequencing.3 As the havior might be more than a coincidence. We assume that authors suggested, the relatively high number of mutations this tumor harbored a large proportion of passenger muta- detected in the validation cohort might reflect novel sin- tions as opposed to driver mutations. This could have pre- gle-nucleotide polymorphisms, because they did not have cluded the discovery of clustering of mutations in signal- normal controls to call tumor-specific variants. In compar- ing pathways and other relevant functions. Our study thus ison, 1 sample in our study harbored an ARID1A mutation, highlights the importance of taking mutations in DNA re- with a variant allele frequency of 4.5%. pair genes into account when performing pathway analy- Another interesting candidate reported was the DDR1 ses on gene sets from tumor-normal pairs for the discovery gene, mutated in 11% of the tumors.3 However, only 1 tu- of tumorigenesis-related genes. mor in our study harbored a DDR1 mutation. Similarly, In this study, we identified a large number of tumor- the candidate gene CDC27 from our study was only mu- specific mutations. However, it is believed that for a clone tated in 1 of 26 exome-sequenced samples, and USP8 was of cells to develop growth advantages that are sufficient not mutated in any. Still, the low frequency of candidate to make a tumor, relatively few products relevant for the genes reported in the current study and by Agnihotri et hallmarks of cancer need to be disrupted.18,23 Thus, dis- al. might represent important tumorigenic factors in sub- tinguishing the true driver mutations responsible for tu- sets of schwannomas. The authors also identified enrich- morigenesis from the passenger mutations remains a great ment of mutations in both known and novel VS pathways, challenge.42,59 In particular, we suspect that the outlier har- although the axonal guidance pathway, reported in our bored a disproportionate amount of passenger mutations. study, was not among them. This was also excluded from some of the analysis regard- The transcript of the CAV1 gene has shown it to be uni- 1,61 ing identification of clustering of mutations in relevant sig- versally downregulated in sVSs, but none of the 46 VSs naling pathways, because the large number of passenger investigated in the present study harbored mutations in the mutations would have diluted the driver mutations. Using CAV1 gene. We therefore sought to investigate whether the a combination of frequency- and functional-based ap- mutational hits found in this study could explain the ab- proaches, we identified some novel genes and lent support sence of CAV1 mRNA in sVSs. Using an in silico–based to the involvement of previously implicated pathways as approach, we discovered many plausible links between our well as novel pathways. In particular, the axonal guidance gene set and molecules upstream of CAV1. Of note, NF2 pathway seems to be involved in sVS tumorigenesis. and several nodes linking to the ERK signaling pathway might regulate CAV1 expression. Other mechanisms, such as epigenetic silencing (not investigated in this study), Conclusions might also govern the expression of this gene. We believe that our work represents the largest WES The extent of resection was total in 20, near total in 24, study on schwannomas. In conclusion, this study supports and subtotal in 2 patients. Six of 8 patients with poor post- the view that NF2 is the most central factor in VS tumori- operative facial function had total resection. We therefore genesis. In addition, the axonal guidance pathway, CDC27, assume that resection rate and facial function were pri- and USP8 are suggested as novel tumorigenic factors. In- marily determined by the surgeon’s decision to remove the terestingly, GKRS did not induce hypermutation of the entire tumor. We did not find any specific mutation cor- tumor. relating with the extent of resection, postresection facial nerve function, or progression-free survival. The outlier specimen was from the oldest patient in the study (75 years of age). The tumor was resected in 2013, Acknowledgments needed retreatment the same year with GKRS because of The National Center for Vestibular Schwannoma funded the rapid regrowth, and needed retreatment with microsurgery study. We dedicate this article to our colleague Per Møller, MD, PhD (1943–2016). Dr. Møller was Professor of Otosurgery at Ber- in 2016, exhibiting an aggressive phenotype. This tumor gen University Hospital. We thank Ms. Guri Matre at the Center harbored a missense mutation (c.A62G, p.D21G) in the for Medical Genetics and Molecular Medicine for technical assis- RAD54 L gene. The protein encoded by this gene belongs tance and Ms. Monica Katrine Finnkirk at the National Center for to the DEAD-like helicase superfamily and is involved in Vestibular Schwannoma and the Department of Neurosurgery for the homologous recombination and repair of DNA. Muta- administrative work.

920 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC Whole-exome sequencing of sporadic vestibular schwannoma

References Comparative genomic hybridization and mutation analyses of sporadic schwannomas. J Neurooncol 72:225–230, 2005 1. Aarhus M, Bruland O, Sætran HA, Mork SJ, Lund-Johansen 20. Jiang J, Hui CC: Hedgehog signaling in development and M, Knappskog PM: Global profiling and cancer. Dev Cell 15:801–812, 2008 tissue microarray reveal novel candidate genes and down- 21. Khanna KK, Jackson SP: DNA double-strand breaks: signal- regulation of the tumor suppressor gene CAV1 in sporadic ing, repair and the cancer connection. Nat Genet 27:247– vestibular schwannomas. Neurosurgery 67:998–1019, 2010 254, 2001 2. Agnihotri S, Gugel I, Remke M, Bornemann A, Pantazis G, 22. Kissil JL, Johnson KC, Eckman MS, Jacks T: Merlin phos- Mack SC, et al: Gene-expression profiling elucidates molecu- phorylation by -activated kinase 2 and effects of phos- lar signaling networks that can be therapeutically targeted in phorylation on merlin localization. J Biol Chem 277:10394– vestibular schwannoma. J Neurosurg 121:1434–1445, 2014 10399, 2002 3. Agnihotri S, Jalali S, Wilson MR, Danesh A, Li M, Klirono- 23. Knudson AG: Two genetic hits (more or less) to cancer. Nat mos G, et al: The genomic landscape of schwannoma. Nat Rev Cancer 1:157–162, 2001 Genet 48: 1339–1348, 2016 24. Koutsimpelas D, Felmeden U, Mann WJ, Brieger J: Analysis 4. Ammoun S, Cunliffe CH, Allen JC, Chiriboga L, Giancotti of cytogenetic aberrations in sporadic vestibular schwan- FG, Zagzag D, et al: ErbB/HER receptor activation and noma by comparative genomic hybridization. J Neurooncol preclinical efficacy of lapatinib in vestibular schwannoma. 103:437–443, 2011 Neuro Oncol 12:834–843, 2010 25. Koutsimpelas D, Ruerup G, Mann WJ, Brieger J: Lack of 5. Ammoun S, Schmid MC, Ristic N, Zhou L, Hilton D, Ercol- neurofibromatosis type 2 gene promoter methylation in spo- ano E, et al: The role of insulin-like growth factors signaling radic vestibular schwannomas. ORL J Otorhinolaryngol in merlin-deficient schwannomas. Glia 60:1721–1733, Relat Spec 74:33–37, 2012 2012 6. Antinheimo J, Sallinen SL, Sallinen P, Haapasalo H, Helin 26. Lee JD, Kwon TJ, Kim UK, Lee WS: Genetic and epigenetic H, Horelli-Kuitunen N, et al: Genetic aberrations in sporadic alterations of the NF2 gene in sporadic vestibular schwanno- and neurofibromatosis 2 (NF2)-associated schwannomas mas. PLoS One 7:e30418, 2012 studied by comparative genomic hybridization (CGH). Acta 27. Li H, Durbin R: Fast and accurate short read alignment with Neurochir (Wien) 142:1099–1105, 2000 Burrows-Wheeler transform. Bioinformatics 25:1754–1760, 7. Biankin AV, Waddell N, Kassahn KS, Gingras MC, Muth- 2009 uswamy LB, Johns AL, et al: Pancreatic cancer genomes 28. Li W, You L, Cooper J, Schiavon G, Pepe-Caprio A, Zhou L, et al: Merlin/NF2 suppresses tumorigenesis by inhibiting the reveal aberrations in axon guidance pathway genes. Nature DCAF1 491:399–405, 2012 E3 CRL4 in the nucleus. Cell 140:477– 8. Breivik CN, Varughese JK, Wentzel-Larsen T, Vassbotn F, 490, 2010 Lund-Johansen M: Conservative management of vestibular 29. Licciulli S, Maksimoska J, Zhou C, Troutman S, Kota S, Liu schwannoma—a prospective cohort study: treatment, symp- Q, et al: FRAX597, a small molecule inhibitor of the p21- toms, and quality of life. Neurosurgery 70:1072–1080, 2012 activated kinases, inhibits tumorigenesis of neurofibroma- 9. Buhard O, Suraweera N, Lectard A, Duval A, Hamelin R: tosis type 2 (NF2)–associated Schwannomas. J Biol Chem Quasimonomorphic mononucleotide repeats for high-level 288:29105–29114, 2013 microsatellite instability analysis. Dis Markers 20:251–257, 30. Lü J, Zou J, Wu H, Cai L: Compensative shuttling of merlin 2004 to on 518 in vestibular schwannoma. 10. Carlson ML, Habermann EB, Wagie AE, Driscoll CL, Van Laryngoscope 118:169–174, 2008 Gompel JJ, Jacob JT, et al: The changing landscape of ves- 31. Maducdoc MM, Ghavami Y, Linskey ME, Djalilian HR: tibular schwannoma management in the United States—a Evaluation of reported malignant transformation of vestibular shift toward conservatism. Otolaryngol Head Neck Surg schwannoma: de novo and after stereotactic radiosurgery or 153:440–446, 2015 surgery. Otol Neurotol 36:1301–1308, 2015 11. Cayé-Thomasen P, Borup R, Stangerup SE, Thomsen J, 32. Manchanda PK, Jones GN, Lee AA, Pringle DR, Zhang M, Nielsen FC: Deregulated genes in sporadic vestibular Yu L, et al: Rac1 is required for Prkar1a-mediated Nf2 sup- schwannomas. Otol Neurotol 31:256–266, 2010 pression in Schwann cell tumors. 32:3491–3499, 12. Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe 2013 D, Sougnez C, et al: Sensitive detection of somatic point 33. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, mutations in impure and heterogeneous cancer samples. Nat Kernytsky A, et al: The Genome Analysis Toolkit: a MapRe- Biotechnol 31:213–219, 2013 duce framework for analyzing next-generation DNA sequenc- 13. de Vries M, van der Mey AG, Hogendoorn PC: Tumor biol- ing data. Genome Res 20:1297–1303, 2010 ogy of vestibular schwannoma: a review of experimental data 34. Mehlen P, Delloye-Bourgeois C, Chédotal A: Novel roles for on the determinants of tumor genesis and growth characteris- Slits and netrins: axon guidance cues as anticancer targets? tics. Otol Neurotol 36:1128–1136, 2015 Nat Rev Cancer 11:188–197, 2011 14. DePristo MA, Banks E, Poplin R, Garimella KV, Maguire 35. Morgenstern PF, Shah K, Dunkel IJ, Reiner AS, Khakoo Y, JR, Hartl C, et al: A framework for variation discovery and Rosenblum MK, et al: Meningioma after radiotherapy for genotyping using next-generation DNA sequencing data. Nat malignancy. J Clin Neurosci 30:93–97, 2016 Genet 43:491–498, 2011 36. Morrow KA, Das S, Metge BJ, Ye K, Mulekar MS, Tucker 15. Fisher JL, Pettersson D, Palmisano S, Schwartzbaum JA, JA, et al: Loss of tumor suppressor Merlin in advanced breast Edwards CG, Mathiesen T, et al: Loud noise exposure and cancer is due to post-translational regulation. J Biol Chem acoustic neuroma. Am J Epidemiol 180:58–67, 2014 286:40376–40385, 2011 16. Greaves M, Maley CC: Clonal evolution in cancer. Nature 37. Myrseth E, Pedersen PH, Møller P, Lund-Johansen M: Treat- 481:306–313, 2012 ment of vestibular schwannomas. Why, when and how? Acta 17. Hanabusa K, Morikawa A, Murata T, Taki W: Acoustic neu- Neurochir (Wien) 149:647–660, 2007 roma with malignant transformation. Case report. J Neuro- 38. Naviglio S, Mattecucci C, Matoskova B, Nagase T, Nomura surg 95:518–521, 2001 N, Di Fiore PP, et al: UBPY: a growth-regulated human ubiq- 18. Hanahan D, Weinberg RA: Hallmarks of cancer: the next uitin isopeptidase. EMBO J 17:3241–3250, 1998 generation. Cell 144:646–674, 2011 39. Patel TR, Chiang VL: Secondary neoplasms after stereotac- 19. Ikeda T, Hashimoto S, Fukushige S, Ohmori H, Horii A: tic radiosurgery. World Neurosurg 81:594–599, 2014

J Neurosurg Volume 128 • March 2018 921

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC A. L. Håvik et al.

40. Pawar SA, Sarkar TR, Balamurugan K, Sharan S, Wang J, 60. Thorvaldsdóttir H, Robinson JT, Mesirov JP: Integrative Zhang Y, et al: C/EBPd targets cyclin D1 for proteasome-me- Genomics Viewer (IGV): high-performance genomics data diated degradation via induction of CDC27/APC3 expression. visualization and exploration. Brief Bioinform 14:178–192, Proc Natl Acad Sci U S A 107:9210–9215, 2010 2013 41. Petrilli AM, Fernández-Valle C: Role of Merlin/NF2 inacti- 61. Torres-Martin M, Lassaletta L, San-Roman-Montero J, De vation in tumor biology. Oncogene 35:537–548, 2016 Campos JM, Isla A, Gavilan J, et al: Microarray analysis of 42. Pon JR, Marra MA: Driver and passenger mutations in can- gene expression in vestibular schwannomas reveals SPP1/ cer. Annu Rev Pathol 10:25–50, 2015 MET signaling pathway and androgen receptor deregulation. 43. Prochazka M, Feychting M, Ahlbom A, Edwards CG, Nise Int J Oncol 42:848–862, 2013 G, Plato N, et al: Occupational exposures and risk of acoustic 62. Trofatter JA, MacCollin MM, Rutter JL, Murrell JR, Duyao neuroma. Occup Environ Med 67:766–771, 2010 MP, Parry DM, et al: A novel -, -, -like 44. Reincke M, Sbiera S, Hayakawa A, Theodoropoulou M, Os- gene is a candidate for the neurofibromatosis 2 tumor sup- swald A, Beuschlein F, et al: Mutations in the deubiquitinase pressor. Cell 72:791–800, 1993 gene USP8 cause Cushing’s disease. Nat Genet 47:31–38, 63. Van der Auwera GA, Carneiro MO, Hartl C, Poplin R, Del 2015 Angel G, Levy-Moonshine A, et al: From FastQ data to high 45. Robinson JT, Thorvaldsdóttir H, Winckler W, Guttman M, confidence variant calls: the Genome Analysis Toolkit best Lander ES, Getz G, et al: Integrative genomics viewer. Nat practices pipeline. Curr Protoc Bioinformatics 43:11.10.1– Biotechnol 29:24–26, 2011 11.10.33, 2013 46. Rong R, Surace EI, Haipek CA, Gutmann DH, Ye K: Serine 64. Wang K, Li M, Hakonarson H: ANNOVAR: functional an- 518 phosphorylation modulates merlin intramolecular as- notation of genetic variants from high-throughput sequencing sociation and binding to critical effectors important for NF2 data. Nucleic Acids Res 38:e164, 2010 growth suppression. Oncogene 23:8447–8454, 2004 65. Warren C, James LA, Ramsden RT, Wallace A, Baser ME, 47. Row PE, Prior IA, McCullough J, Clague MJ, Urbé S: The Varley JM, et al: Identification of recurrent regions of chro- ubiquitin isopeptidase UBPY regulates endosomal ubiquitin mosome loss and gain in vestibular schwannomas using com- dynamics and is essential for receptor down-regulation. J parative genomic hybridisation. J Med Genet 40:802–806, Biol Chem 281:12618–12624, 2006 2003 48. Rowe J, Grainger A, Walton L, Silcocks P, Radatz M, Ke- 66. Welling DB, Lasak JM, Akhmametyeva E, Ghaheri B, Chang meny A: Risk of malignancy after gamma knife stereotactic LS: cDNA microarray analysis of vestibular schwannomas. radiosurgery. Neurosurgery 60:60–66, 2007 Otol Neurotol 23:736–748, 2002 49. Sainz J, Huynh DP, Figueroa K, Ragge NK, Baser ME, Pulst 67. Xia R, Jia H, Fan J, Liu Y, Jia J: USP8 promotes smoothened SM: Mutations of the neurofibromatosis type 2 gene and lack signaling by preventing its ubiquitination and changing its of the gene product in vestibular schwannomas. Hum Mol subcellular localization. PLoS Biol 10:e1001238, 2012 Genet 3:885–891, 1994 68. Xiao GH, Beeser A, Chernoff J, Testa JR: p21-activated 50. Schmuckli-Maurer J, Rolfsmeier M, Nguyen H, Heyer WD: kinase links Rac/Cdc42 signaling to merlin. J Biol Chem Genome instability in rad54 mutants of Saccharomyces cere- 277:883–886, 2002 visiae. Nucleic Acids Res 31:1013–1023, 2003 69. Zhou L, Hanemann CO: Merlin, a multi-suppressor from cell 51. Schneider AB, Ron E, Lubin J, Stovall M, Shore-Freedman membrane to the nucleus. FEBS Lett 586:1403–1408, 2012 E, Tolentino J, et al: Acoustic neuromas following childhood radiation treatment for benign conditions of the head and neck. Neuro Oncol 10:73–78, 2008 52. Schouten JP, McElgunn CJ, Waaijer R, Zwijnenburg D, Disclosures Diepvens F, Pals G: Relative quantification of 40 nucleic acid The authors report no conflict of interest concerning the materi- sequences by multiplex ligation-dependent probe amplifica- als or methods used in this study or the findings specified in this tion. Nucleic Acids Res 30:e57, 2002 paper. 53. Schüz J, Steding-Jessen M, Hansen S, Stangerup SE, Cayé- Thomasen P, Poulsen AH, et al: Long-term mobile phone use Author Contributions and the risk of vestibular schwannoma: a Danish nationwide Conception and design: Lund-Johansen, Håvik, Bruland, cohort study. Am J Epidemiol 174:416–422, 2011 Knappskog. Acquisition of data: all authors. Analysis and inter- 54. Shin M, Ueki K, Kurita H, Kirino T: Malignant transforma- pretation of data: Lund-Johansen, Håvik, Bruland, Miletic, tion of a vestibular schwannoma after Gamma Knife radio- Knappskog. Drafting the article: Lund-Johansen, Håvik, Knappsk- surgery. Lancet 360:309–310, 2002 og. Critically revising the article: all authors. Statistical analysis: 55. Sobrido MJ, Pereira CR, Barros F, Forteza J, Carracedo A, Håvik, Bruland. Administrative/technical/material support: Lund- Lema M: Low frequency of replication errors in primary Johansen, Myrseth, Miletic, Aarhus, Knappskog. Study supervi- nervous system tumours. J Neurol Neurosurg Psychiatry sion: Lund-Johansen, Knappskog. 69:369–375, 2000 56. Stangerup SE, Caye-Thomasen P: Epidemiology and natural Supplemental Information history of vestibular schwannomas. Otolaryngol Clin North Online-Only Content Am 45:257–268, vii, 2012 57. Stemmer-Rachamimov AO, Xu L, Gonzalez-Agosti C, Bur- Supplemental material is available with the online version of the wick JA, Pinney D, Beauchamp R, et al: Universal absence article. of merlin, but not other ERM family members, in schwanno- Supplemental Tables and Figure. https://thejns.org/doi/suppl/​ mas. Am J Pathol 151:1649–1654, 1997 10.3171/2016.10.JNS161384. 58. Stepanidis K, Kessel M, Caye-Thomasen P, Stangerup SE: Socio-demographic distribution of vestibular schwannomas Correspondence in Denmark. Acta Otolaryngol 134:551–556, 2014 Morten Lund-Johansen, Department of Neurosurgery, Hauke- 59. Stratton MR, Campbell PJ, Futreal PA: The cancer genome. land University Hospital, Bergen 5009, Norway. email: mljo@ Nature 458:719–724, 2009 helse-bergen.no.

922 J Neurosurg Volume 128 • March 2018

Unauthenticated | Downloaded 09/29/21 09:50 AM UTC