GOPC-ROS1 Fusion Due to Microdeletion at 6Q22 Is an Oncogenic Driver in a Subset of Pediatric Gliomas And

J Neuropathol Exp Neurol Vol. 78, No. 12, December 2019, pp. 1089–1099 doi: 10.1093/jnen/nlz093

ORIGINAL ARTICLE

GOPC-ROS1 Fusion Due to Microdeletion at 6q22 Is an Oncogenic Driver in a Subset of Pediatric Gliomas and

Glioneuronal Tumors Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020

Timothy E. Richardson, DO, PhD, Karen Tang, MD, Varshini Vasudevaraja, MS, Jonathan Serrano, BS, Christopher M. William, MD, PhD, Kanish Mirchia, MD, Christopher R. Pierson, MD, PhD, Jeffrey R. Leonard, MD, Mohamed S. AbdelBaki, MD, Kathleen M. Schieffer, PhD, Catherine E. Cottrell, PhD, Zulma Tovar-Spinoza, MD, Melanie A. Comito, MD, Daniel R. Boue, MD, PhD, George Jour, MD, and Matija Snuderl, MD

features; the third tumor aligned best with glioblastoma and Abstract showed no evidence of neuronal differentiation. Copy number ROS1 is a transmembrane receptor tyrosine kinase proto- profiling revealed chromosome 6q22 microdeletions correspond- oncogene that has been shown to have rearrangements with sev- ingtotheGOPC-ROS1 fusion in all 3 cases and methylation pro- eral genes in glioblastoma and other neoplasms, including intra- filing showed that the tumors did not cluster together as a single chromosomal fusion with GOPC due to microdeletions at 6q22.1. entity or within known methylation classes by t-Distributed Sto- ROS1 fusion events are important findings in these tumors, as chastic Neighbor Embedding. they are potentially targetable alterations with newer tyrosine kinase inhibitors; however, whether these tumors represent a dis- Key Words: 6q22, Astrocytoma, Brain tumor, GOPC, Pediatric gli- tinct entity remains unknown. In this report, we identify 3 cases oma, ROS1. of unusual pediatric glioma with GOPC-ROS1 fusion. We reviewed the clinical history, radiologic and histologic features, performed methylation analysis, whole genome copy number profiling, and next generation sequencing analysis for the detec- INTRODUCTION tion of oncogenic mutation and fusion events to fully characterize ROS1, located at 6q22.1, encodes a transmembrane re- the genetic and epigenetic alterations present in these tumors. ceptor tyrosine kinase that under normal circumstances binds Two of 3 tumors showed pilocytic features with focal expression to growth factors and undergoes dimerization and phosphory- of synaptophysin staining and variable high-grade histologic lation with transmission of growth signals downstream through intracellular second messenger systems (1). Onco- genic ROS1 fusion was originally discovered in a glioblas- From the Department of Pathology, State University of New York, Upstate toma (GBM) cell line (U118MG) in the late 1980s (2–4). Medical University, Syracuse, New York (TER, KM); Department of Pe- Subsequent work demonstrated this original alteration to be diatrics (KT); Department of Pathology (VV, JS, CMW, GJ, MS), New due to an intrachromosomal microdeletion of 240–250 kilo- York University Langone Health, New York, New York; Department of Pathology & Laboratory Medicine, Nationwide Children’s Hospital and bases (kb), resulting in fusion of GOPC and ROS1 in this The Ohio State University (CRP, CEC, DRB); Department of Neurologi- cell line (5). cal Surgery, Nationwide Children’s Hospital (JRL); Department of He- This del(6)(q22;q22) and the resulting rearrangement matology, Oncology, and Bone Marrow Transplant, Nationwide produces a constitutively active tyrosine kinase fusion prod- Children’s Hospital and The Ohio State University (MSA); Nationwide uct that has been demonstrated to be sufficient to produce Children’s Hospital, The Institute for Genomic Medicine (KMS, CEC), Columbus, Ohio; Department of Neurosurgery (ZT-S); Department of neoplastic transformation and tumor development in both Pediatrics (MAC), State University of New York, Upstate Medical Uni- cell culture and murine models (4–7). Since its discovery, versity, Syracuse, New York; and Waters Center for Children’s Cancer several studies have identified this particular alteration as a and Blood Disorders, State University of New York, Upstate Cancer relatively uncommon tumor driver in a small subset of GBM Center, Syracuse, New York (MAC). Send correspondence to: Matija Snuderl, MD, Department of Pathology, and low-grade gliomas, including very rare pediatric tumors New York University Langone Health 660 First Avenue, 2nd Floor, New (6, 8, 9). GOPC-ROS1 fusion has also been identified in sev- York, NY 10016; Email: [email protected] eral other cancers, including lung adenocarcinoma (10–12) Methylation profiling at NYU is supported in part by grants from the Fried- and cholangiocarcinoma (13), and ROS1 fusions with other berg Charitable Foundation and the Making Headway Foundation (to M.S.). A portion of this project was supported by the Nationwide Foun- partners have been discovered in many other tumor types dation Pediatric Innovation Fund. (14). Another pathogenic fusion, ZCCHC8-ROS1, was found The authors have no duality or conflicts of interest to declare. to be the underlying oncogenic driver in single a case of

VC 2019 American Association of Neuropathologists, Inc. All rights reserved. 1089 Richardson et al J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020

FIGURE 1. (A) Coronal and (B) axial T1 þ gadolinium contrast enhancing images of initial tumor presentation in Case 1 demonstrating a heterogeneous mass centered in the left thalamus with focal enhancement and extension into the left lateral ventricle and surrounding structures. Axial T1 þ gadolinium contrast enhancing imaging (C) and axial T2 FLAIR imaging (D) in the same patient demonstrating recurrence. Coronal (E) and axial (F) T1 þ gadolinium contrast enhancing images of Case 2 showing a heterogeneously enhancing intraventricular mass. Axial T1 þ gadolinium contrast enhancing images of case 3 showing a heterogeneously enhancing mass centered in the right temporal lobe (arrowheads) (G) and coronal T2 FLAIR imaging of case 3 showing multinodular tumor in the right temporal lobe (arrowheads) and a noncontiguous nodular component in the right frontal lobe (arrow) (H). congenital GBM (15). Analysis of The Cancer Genome At- MATERIALS AND METHODS las (TCGA) dataset also suggested that GOPC-ROS1 fusion may result in shorter overall survival in adult patients with Clinical Histories GBM (6). While this gene fusion product may be difficult to Case 1 readily detect due to the size of the microdeletion at 6q22 and the inability of break-apart FISH probes to identify this A 4-year-old previously healthy female patient present- particular rearrangement, commercially available immuno- ing with seizure and rapid decline of mental status was found histochemical stains for ROS1 are often positive in cases to have a 3.1 2.8 2.5 cm tumor in the left thalamus with fo- with ROS1 rearrangement with any number of fusion part- cal contrast enhancement extending into the left basal ganglia, ners (6, 11, 12). left temporal lobe, and left ventricle, resulting in hydrocepha- In this study, we identified 3 pediatric patients with lus (Fig. 1A, B). She underwent ventriculoperitoneal shunt clinically similar supratentorial brain tumors, diagnosed as placement and resection of the tumor which showed pilocytic- pilocytic astrocytoma, glioneuronal tumor, and/or GBM af- like morphology with low proliferative activity. The patient ter initial resection. Two of the 3 cases had strikingly similar was treated with carboplatin and vincristine. Over the next histology, including mixed astrocytic and oligodendroglial- 3 years, there was slow, progressive increase in tumor size like features, with diffuse glial fibrillary acidic protein (Fig. 1C, D) requiring tumor debulking and repeat administra- (GFAP) staining and focal islands of synaptophysin- tion of carboplatin and vincristine with additional vinblastine. reactive, NeuN-negative cells. Case 1 appeared to be low- The results of the second surgery showed similar histology but grade on histologic exam of the initial resection specimen; the tumor then had higher-grade features, including increased however, it recurred with features suggestive of anaplasia. proliferative activity and focal necrosis. Two years later, after Case 2 had focal high-grade features on the initial resection additional tumor growth she received left-sided laser ablation specimen. Case 3 had high-grade astrocytic features with of the thalamic tumor and repeat biopsy for additional molecu- diffusely infiltrating components. Methylation profiling lar characterization, followed by 4 cycles of Temodar and failed to classify these tumors to known tumor entities pre- Avastin, complicated by myelosuppression, with radiologic sent in the current version of the methylation classifier with evidence of further tumor progression (Table). a high degree of certainty. Subsequent whole-genome copy number profiling demonstrated microdeletions at 6q22, and Case 2 DNA and RNA sequencing using next generation sequencing (NGS) methods demonstrated GOPC-ROS1 fusion in all A 5-year-old previously healthy male patient presented 3 cases. with 3–4 weeks of somnolence, emesis, and headaches was 1090 erpto x Neurol Exp Neuropathol J TABLE. Summary of Radiologic, Histologic, Immunohistochemical, and Molecular Data in Cases 1–3 Age Gender Imaging Findings Histology Immunohistochemistry NGS Results Copy Number Pro- Diagnosis filing Results Case #1 Surgery 1 4 Female Focally enhancing Mildly hypercellular GFAPþ, OLIG2þ, focal syn- N/A N/A Low-grade glioma, most 3.1 2.8 2.5 cm astrocytic morphol- aptophysinþ, focal weak consistent with pilo- left thalamic/basal ogy with oligoden- EMAþ, IDH1 R132H-, cytic astrocytoma ganglia/medial droglial-like areas BRAF V600E-, p53-, temporal lobe/ven- and microcystic ATRX-retained,

tricular mass with areas, rare EGBs H3K27M-, Ki-67 0%–2% 2019 December 12, Number 78, Volume • hydrocephalus identified, no ganglion cells; 0 mits/ 10 HPF Surgery 2 7 Residual/recurrent Hypercellular astro- GFAPþ, OLIG2þ, focal syn- KIAA1549-BRAF N/A Glioma with increased nodular enhancing cytic morphology aptophysinþ, focal weak fusion negative proliferation, most con- 2.4 1.8 1.7 cm with microvascular EMAþ, IDH1 R132H-, sistent with pilocytic bilateral thalamic proliferation and BRAF V600E-, p53-, astrocytoma mass with exten- necrosis, rare EGBs ATRX-retained, sion into the left identified, 4 mits/ H3K27M-, Ki-67 15% lateral and third 10 HPF ventricle Surgery 3 9 Residual/recurrent Hypercellular astro- GFAPþ, OLIG2þ, focal syn- GOPC-ROS1 fusion; Loss of chromosomes Glioneuronal Tumor with nodular enhancing cytic morphology aptophysinþ, focal weak CHEK2 (I157T) 3, 8, 11, 17, 18, and anaplastic features and 3.8 2.3 2.2 cm with microcystic EMAþ, IDH1 R132H-, mutation 22q; amplification GOPC-ROS1 fusion bilateral thalamic areas, microvascu- BRAF V600E-, p53-, p16- of 7; deletion of and periatrial mass lar proliferation, retained, ATRX-retained, C19MC; microde- and focal necrosis, H3K27M-, Ki-67 20% letion at 6q22 no Rosenthal fibers or EGBs identified, 3 mits/10 HPF Case #2 Surgery 1 5 Male Heterogeneously en- Hypercellular astro- GFAPþ, focal syn- GOPC-ROS1 fusion Loss of chromosomes Glioneuronal Tumor with hancing cytic morphology aptophysinþ, faint EMAþ, 1 and 5; deletion of anaplastic features and OCRS uini eiti Gliomas Pediatric in Fusion GOPC-ROS1 4.7 4.3 4.0 cm with oligodendrog- neurofilament-in tumor C19MC, microde- GOPC-ROS1 fusion mass predomi- lial-like areas and cells but present in back- letion at 6q22 nantly involving myxoid areas, infil- ground axons, Ki-67 the right anterior trative features, 3 20%–30% lateral ventricle mits/10 HPF (continued)

1091 Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020 August 25 on user University York New by https://academic.oup.com/jnen/article/78/12/1089/5593615 from Downloaded Richardson et al J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019

found to have a 4.7 4.3 4.0 cm heterogeneously enhancing intraventricular mass involving the right anterior lateral ven-

fusion tricle with an inferior portion projecting to the foramen of Monro, resulting in obstructive hydrocephalus (Fig. 1E, F).

ROS1 He was obtunded with a dilated right pupil and required - Diagnosis emergency ventricular drain placement. The patient underwent gross total resection of the right frontal intraventricular lial tumor with features aligning best with infiltrating astrocytoma/ glioblastoma and GOPC mass. Pathology was reported to be a glioneuronal tumor with focal high-grade features. His postoperative course was

complicated by a pseudomeningocele, which was subse- Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020 quently treated by oversewing his incision and placement of a lumbar drain. Since his resection, he has had no evidence of recurrence on MRI imaging and has not required further

filing Results treatment thus far.

Case 3

A 2-year-old previously healthy male patient presented fusion Microdeletion at 6q22 High-grade neuroepithe- with new-onset seizures and was found to have a

ROS1 5.7 5.3 4.9 cm multinodular and diffuse, heterogeneously - enhancing signal abnormality present throughout portions of the right temporal lobe, right frontal lobe, and involving the GOPC right subinsular white matter as well as the right basal ganglia and posterior limb internal capsule, with its epicenter at anterior inferior right temporal lobe, and without evidence of met- astatic disease in the spine (Fig. 1G, H). He underwent partial %

, OLIG2 resection and the pathology demonstrated features consistent þ –40 with high-grade neuroepithelial tumor, aligning best with glio- , synaptophysin-, %

þ blastoma. He was started on chemotherapy as per Baby POG 1 (2 cycles of cyclophosphamide and vincristine followed by a

patchy EMA-, IDH1 R132H-, p53-, ATRX-retained, INI1- retained, NeuN-, H3K27M-, neurofilament- in tumor cells but present in background axons, Ki-67 30 third cycle of cisplatin and etoposide) for 18 months with ra-

GFAP patchy diographic improvement in his tumor, followed by one cycle of autologous hematopoietic stem cell transplant with carboplatin, thiotepa, and etoposide. He still has evidence of stable residual disease on imaging without any evidence of progression or subsequent seizure activity and remains completely in- tact neurologically at 5 years of age. cytic morphology with nuclear pleo- morphism, extensive infiltration, focal microvascular proliferation, discrete nodular areas; no oligodendroglial-like areas or evidence of neoplastic neuronal differentiation, or microcystic/ myxoid areas, 1–4 mits/ 10 HPF Hypercellular astro- Histology and Immunohistochemistry Hematoxylin and eosin (H&E)-stained slides were pre- 4.9 cm

pared from 4-lm-thick sections of formalin-fixed, paraffin-

5.3 embedded (FFPE) tissue using standard protocols. Immuno- histochemistry was performed on 4-lm-thick paraffin sections 5.7 multinodular and diffuse mass, with right temporal lobe epicenter, patchy leptomeningeal infiltration, and noncontiguous nodules in the right frontal lobe, deep white matter, and right basal ganglia; no evidence of hydrocephalus following heat-induced epitope retrieval using CC1 (Ventana, Tucson, AZ), then staining with GFAP (Leica Biosystems, Richmond, VA and Thermo Fisher Scientific, Waltham, MA), Olig2 (Cell Marque, Rocklin, CA), IDH1 R132H (Dianova, Hamburg, Germany), ATRX (Sigma-Aldrich, St. Louis, MO), p53 (Ventana and Leica Biosystems), synaptophysin (Leica Age Gender Imaging Findings Histology Immunohistochemistry NGS Results Copy Number Pro- Biosystems), neurofilament (Covance, Princeton, NJ and Leica Biosystems), epithelial membrane antigen (EMA) (Cell Marque), Histone H3 K27M (Sigma-Aldrich), INI1 (Cell Mar- Surgery 1 2 Male Variably enhancing

Continued que), Neu-N (Sigma-Aldrich), and Ki-67 (Dako, Carpinteria, CA and Cell Marque) on either a Ventana Benchmark XT or Ventana Benchmark Ultra automated stainer, using Ventana Case #3 TABLE. UltraView Universal DAB Detection kits. 1092 J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 GOPC-ROS1 Fusion in Pediatric Gliomas

Methylation Studies to 4/10 HPF) in the second specimen (Fig. 2A, B). The tumor DNA extraction was carried out using the automated was positive for GFAP (Fig. 2C), had islands of Maxwell system (Promega, Madison, WI). DNA methylation synaptophysin-positive cells (Fig. 2D), diffuse Olig2 reactiv- was analyzed by the Illumina EPIC Human Methylation array, ity (Fig. 2E), and was negative for IDH1 R132H, BRAF which assesses 850 000 CpG sites, according to the manufac- V600E, and p53, and had retained nuclear ATRX expression. turer’s instructions at the NYU Molecular Pathology laboratory Staining for neurofilament revealed no axons coursing through as described previously (16). Molecular subclassification and t- the tumor in any area, suggesting a mostly noninfiltrative pro- Distributed Stochastic Neighbor Embedding (t-SNE) visualiza- cess. Ki-67 proliferation index was 15%–20% (Fig. 2F). All tion was performed utilizing the cloud-based DNA methylation specimens were negative for H3K27M and none showed aber- classifier as described previously (17). In addition, the array data rant CD34 staining to suggest diffuse midline glioma or poly- Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020 were used to calculate a low-resolution copy number profile, as morphous low-grade neuroepithelial tumor of the young previously described (18–23). “Gain” or “amplification” in the (PLNTY). copy number profiles was determined by log2 0.3.

Next-Generation Sequencing Case 2 Targeted genome sequencing was performed on DNA isolated from FFPE tissue using NGS panels to evaluate 324 The resection specimen demonstrated similar histol- genes and gene rearrangements, microsatellite instability, and ogy to case 1 with cords of astrocytic cells, myxoid change, overall tumor mutation burden in case 1 (Foundation Medi- and oligodendroglial-like areas (Fig. 2G, H). The tumor was cine, Cambridge, MA). RNA was extracted and sequenced us- diffusely positive for GFAP (Fig. 2I) with focal synaptophy- ing a customized clinically validated and NY State approved sin (Fig. 2J) and weak cytoplasmic EMA reactivity. There RNAseq panel targeting 86 cancer related genes (NYU Fusion were areas with neurofilament-positive axons traversing the SEQer) using Anchored Multiplex PCR (ArcherDX, Boulder, tumor, suggesting a focally infiltrative process (Fig. 2K). CO) in case 2. For case 3, whole genome sequencing was Ki-67 proliferation index was relatively low in the majority performed on DNA isolated from FFPE tissue and normal pe- of the tumor, but focally elevated up to 20%–30% in a few ripheral blood. Libraries were prepared using NEBNext Ultra foci (Fig. 2L). The majority of the tumor demonstrated low- II DNA reagents (New England Biolands, Ipswich, MA). grade fibrillar and neurocytic morphology with variability in Paired-end 151-bp reads were generated on the Illumina the myxoid stroma and focal calcification. On H&E sec- HiSeq 4000 and reads were aligned to the human genome ref- tions, some areas of the tumor were noted to have pleomor- erence sequence (build GRCh37). Copy number variation was phism, hypercellularity, mitotic activity, and microvascular evaluated using VarScan (http://varscan.sourceforge.net/; last proliferation. accessed August 1, 2019) and structural variation was analyzed using Lumpy (https://github.com/arq5x/lumpy-sv; last accessed August 1, 2019). In parallel, tumor RNA derived Case 3 from FFPE underwent whole transcriptome sequencing. Paired-end 151-bp reads were generated on the Illumina Hi- The resection specimen aligned best with GBM, and Seq 4000 and reads were aligned to the human genome refer- demonstrated multiple hypercellular discrete-appearing nodu- ence sequence (build GRCh38). Gene fusions were evaluated lar components, as well as surrounding areas of extensive in- by an ensemble approach of 7 fusion callers. filtration, with pleomorphic round, ovoid, and bipolar elongated cells (Fig. 2M). The tumor cells formed secondary RESULTS structures, with areas of prominent perineuronal, perivascular, leptomeningeal, and subpial infiltration. No obvious Tumor Histology oligodendroglial-like or myxoid/microcystic areas were iden- Case 1 tified. There was focal microvascular proliferation but no defi- nite necrosis was seen (Fig. 2N). The tumor cells were The first resection specimen showed features most con- partially positive for GFAP (Fig. 2L) within a background of sistent with pilocytic astrocytoma: There were atypical astro- synaptophysin-reactive neuropil and entrapped neurons cytic areas, oligodendroglial-like areas, myxoid areas with (Fig. 2P). The tumor cells were also partially positive for microcystic changes, calcifications, and rare eosinophilic Olig2 (Fig. 2Q) and negative for EMA. IDH1 R132H and granular bodies, without any definitive ganglion cells. The tu- H3K27M were both negative, p53 showed patchy weak posi- mor was diffusely positive for GFAP with focal islands of tivity in scattered tumor cells, and ATRX and INI-1 were both synaptophysin-positive, NeuN-negative cells and focal cyto- retained, suggesting wildtype status of these 5 genes. Neurofi- plasmic EMA positivity. The tumor cells were negative for lament was focally positive with background axons and NeuN IDH1 R132H, BRAF V600E, and p53 and had retained ATRX was positive in scattered entrapped neurons, further supporting expression. Ki-67 proliferation index was 0%–2%. Subsequent an infiltrative process, and no evidence of neuronal differenti- resection specimens from tumor recurrence demonstrated a ation by tumor cells was detected. Scattered mitotic figures similar overall morphology except there was microvascular were present (1–4/10 HPF), and Ki-67 proliferation index was proliferation, focal necrosis, and increased mitotic figures (up elevated, multifocally as high as 30%–40% (Fig. 2R). 1093 Richardson et al J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020

FIGURE 2. Histologic findings in the second resection specimen of Case 1 showing H&E sections with varying histologic features (A, B), diffuse GFAP staining (C), focal islands of synaptophysin staining (D), diffuse Olig2 reactivity (E), and brisk Ki-67 proliferation rate (F). Histologic findings in Case 2 showing H&E sections with varying histologic features (G–H), diffuse GFAP staining (I), islands of synaptophysin staining (J), neurofilament-positive axons (K), and Ki-67 proliferation rate (L). Histologic findings in Case 3 showing a hypercellular glial/astrocytic proliferation with focal endothelial proliferation (M, N), diffuse-patchy GFAP staining (O), background synaptophysin staining (P), partial Olig2 nuclear staining (Q), and elevated Ki-67 proliferation index (R). (G) is taken at a total magnification of 100, (L) is taken at a total magnification of 400, all other images are taken at a total magnification of 200. All scale bars: 200 mm.

1094 J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 GOPC-ROS1 Fusion in Pediatric Gliomas

Methylation Analysis and t-SNE Panel was unable to determine if the CHEK2 I157T mutation Specimens from tumor recurrence in case 1 and initial was somatic or germline, and the patient’s family was given resection in cases 2 and 3 each underwent whole genome genetic counseling with regard to the CHEK2 mutation, which DNA methylation profiling, classification, and t-SNE cluster has been associated with a variety of other cancers, including analysis (www.molecularneuropathology.org); however, none GBM, breast cancer, colorectal carcinoma, and uterine carci- of the cases matched well with any known entity included in noma, among others (24–29). Patient 2 has not received any the classifier (V11b4, optimal score >0.9). Case 1 classified targeted therapy to date and is being followed with serial MRI best with anaplastic pilocytic astrocytoma, albeit poorly (0.49 imaging without evidence of tumor progression. Patient 3 has calibrated score) and had unmethylated MGMT promotor. not received any targeted therapy to date and is being followed Case 2 classified poorly with pilocytic astrocytoma (0.4 cali- with serial MRI imaging without evidence of progression. Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020 brated score) and also had unmethylated MGMT promoter. Case 3 did not match any known methylation class and had unmethylated MGMT promoter. t-SNE plots demonstrated DISCUSSION that these 3 cases did not cluster together as a distinct new en- Herein, we describe 3 cases of pediatric glioma with un- tity but instead clustered with various other methylation clas- usual histology, all harboring GOPC-ROS1 fusions with corre- ses including different methylation subclasses of low-grade sponding microdeletions at 6q22. Tumors with this genetic glioma and pleomorphic xanthoastrocytoma (PXA; Fig. 3). rearrangement typically fall into the category of adult astrocytoma, although rare pediatric cases with similar findings have been previously identified (6, 9). All 3 tumors were supraten- Copy Number Profiling torial masses, and cases 1 and 2 showed prominent extension Each case had large scale gains and losses as well as into the lateral ventricles. The histologic findings overlapped more focal alterations. Case 1 demonstrated heterozygous loss with pilocytic astrocytoma on H&E in cases 1 and 2, showing of chromosomes 3, 8, 11, 17, 18, and 22q, as well as gain of astrocytic regions with admixed areas of oligodendroglial-like chromosome 7, 9, 14q, and 20, deletion of the genomic region areas, myxoid change, microcystic areas, and calcifications. around C19MC, and microdeletion at 6q22 (the region encom- Immunohistochemically, they were both diffusely positive passing both GOPC and ROS1)(Fig. 4A). Case 2 demon- for GFAP with unusual islands of synaptophysin-positive, strated homozygous loss of chromosome 1 and heterozygous NeuN-negative cells, suggesting a biphasic tumor type. Both loss of chromosome 5 and with deletion of C19MC, along had relatively low overall proliferation indices and mitotic with microdeletion at 6q22 (Fig. 4B). Case 3 demonstrated counts on the initial resection specimens; however, the subse- microdeletion at 6q22 (Fig. 4C). quent recurrence in case 1 had higher-grade features including microvascular proliferation, focal necrosis, increased mitotic rates, and increased Ki-67 index. There was evidence of at Whole Genome and Whole Exome NGS Panel least focal tumor infiltration and focal increased mitotic activ- Targeted NGS panel at Foundation Medicine identified ity in case 2. Case 3 had somewhat different histologic appear- a CHEK2 I157T variant as well as GOPC-ROS1 fusion ances, with a rather unusual (for diffuse astrocytoma) (GOPC [NM_020399] exon 8; ROS1 [NM_002944] exon 35) discretely multinodular architecture in some areas, while also in case 1, confirming the results of the copy number profiling demonstrating significant extensively infiltrating components analysis. The tumor tissue was found to maintain microsatel- away from the more cellular nodules, and a different immuno- lite stability and a low mutational burden of 1 mutation/meg- histochemical profile (compared with cases 1 and 2), without abase. Targeted NGS panel at ArcherDX demonstrated any morphologic or immunohistochemical evidence of neuro- GOPC-ROS1 fusion (NM_020399) exon 4; ROS1 nal differentiation detected, but with an increased mitotic in- (NM_002944) exon 36) in case 2 (Fig. 5). In case 3, GOPC- dex and focal microvascular proliferation. ROS1 fusion (GOPC [NM_020399] exon 4; ROS1 On the basis of their variable methylation profiles that [NM_002944] exon 36) was identified by RNA sequencing, do not match closely to any known entity, t-SNE plotting was confirming the WGS copy number and structural variation performed to determine whether these 3 cases form a distinct analysis. No other clearly significant genetic variation was methylation entity as some neuroepithelial tumors with dis- detected. tinct fusions do (30, 31); however, these 3 cases did not cluster as a distinct entity (Fig. 3). It is important to note, however, that many of the entities included in the initial cluster profiles Final Pathologic Diagnosis have wider spread clusters that could account for the differ- Case 1: Glioneuronal tumor with anaplastic features and ence in these cases and a larger number of cases may be GOPC-ROS1 fusion; Case 2: Glioneuronal tumor with ana- needed to establish the signature for classification. plastic features and GOPC-ROS1 fusion. Case 3: Glioblas- The molecular mechanism underlying ROS1 fusion toma with GOPC-ROS1 fusion (Table). Methylation studies proteins driving oncogenesis has been well studied over the were performed did not demonstrate definitive clustering with past 30 years. GOPC, originally named FIG (Fused in any clinically validated tumor entity. Glioblastoma), is a constitutively expressed housekeeping All patients are currently alive as of this report. Patient 1 gene coding for a protein associated with the Golgi apparatus has been started on an ROS1 inhibitor (Lorlatinib) without evi- with a possible role in vesicular transport (32). ROS1 is a dence of further tumor progression. The Foundation Medicine transmembrane receptor tyrosine kinase proto-oncogene that 1095 Richardson et al J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020

FIGURE 3. t-Distributed Stochastic Neighbor Embedding (t-SNE) plot showing clustering of the 3 cases included in this report (arrows) in relation to other tumor entities, including diffuse leptomeningeal glioneuronal tumor (DLGNT), pilocytic astrocytoma/ganglioglioma (LGG PA/GG ST), midline pilocytic astrocytoma (LGG PA MID), rosette-forming glioneuronal tumor (LGG RGNT), ganglioglioma (LGG GG), and pleomorphic xanthoastrocytoma (PXA). Notably, despite a common fusion driver, these 3 tumors did not form a distinct cluster. functions to mediate extracellular growth factor ligand signals identified only a single case with GOPC-ROS1 fusion and an- via phosphorylation of second messenger systems (5). These 2 other with CEP85L-ROS1 fusion (8). Because of prior diffi- genes are in close proximity to one another on chromosome culty in identifying this rearrangement by FISH and other 6q22.1, such that microdeletions of 240–250 kb in this region methods (6, 36, 37), it is possible that this rearrangement is can result in chromosomal truncation and fusion of the 2 more common in both adult and pediatric glioma cohorts than genes, leading to the regulatory domain of GOPC being fused has been recognized to date, including in large TCGA studies to the kinase domain of ROS1 (5). and others (36, 38, 39). Although the process underlying 6q22 microdeletions These pediatric tumors, all with unusual morphology, resulting in GOPC-ROS1 fusion is not fully understood, the immunohistochemical profiles, and uncertain methylation pro- resulting protein has been shown to be sufficient to initiate files, are very difficult to classify and have difficult to predict neoplastic transformation in astrocyte cell culture and cause clinical behavior. The similar H&E morphology in cases 1 malignant neoplasms in murine models (5–7, 33). Fusions of and 2, similar biphasic immunohistochemical staining pat- ROS1 with GOPC and a myriad of other fusion partners has terns, and identical molecular aberrations on both copy num- been demonstrated in a number of different cancers, including ber profiling analysis and next-generation sequencing meningioma, ependymoma, low-grade glioma, GBM, and panels suggest that they may represent similar or perhaps re- several peripheral cancers; however, these are thought to be lated entities, however that was not confirmed by t-SNE more rare in pediatric gliomas (5, 6, 8, 11, 12, 34, 35). analysis. Given the small number of cases in this report, it is We have not encountered additional cases with this par- not currently certain if these tumors represent a previously ticular fusion in our database of 4000 profiled tumors, sug- unrecognized tumor entity or possibly a distinct subtype of gesting that these tumors are relatively rare in pediatrics, anaplastic pilocytic astrocytoma or glioneuronal tumors. It consistent with a previous study of 282 pediatric gliomas that is also possible that GOPC-ROS1 fusion acts as a driver in

1096 J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 GOPC-ROS1 Fusion in Pediatric Gliomas Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020

FIGURE 4. Copy number plots showing large scale chromosomal gains and losses and focal chromosome 6q22 deletion (arrows) in Case 1 (A), Case 2 (B), and Case 3 (C). histologically diverse brain tumors. Our results suggest, at oncogenic drivers in diverse tumor types. In vitro and ex vivo least, that this different genetic profile warrants further in- models of GBM have shown increased response to the more vestigation, given the potential for better therapeutic modal- conventional GBM therapy temozolomide when used in com- ities. Therefore, RNAseq studies are warranted in tumors bination with crizotinib, a targeted ROS1-inhibitor (37). Clini- that exhibit described histology and do not match with cal trials for crizotinib and another ROS1-inhibitor, known entities. entrectinib, in lung cancer patients with ROS1 fusion have With increasing emphasis on precision medicine, many also shown promising results (40, 41). More recent studies recent studies have evaluated the potential for targeting with lorlatinib, a newer tyrosine kinase inhibitor with

1097 Richardson et al J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020

FIGURE 5. Summary of GOPC-ROS1 gene fusion in Case 2. (A) A schematic representation of the 6q22 region with the GOPC and ROS1 loci. The red arrows indicate the boundaries of the 240 kb deletion leading to GOPC and ROS1 fusion (both genes are transcribed in the same direction). (B) Anchored multiplex PCR (AMP) enabled next generation sequencing (NGS) reveals both fusion partners GOPC (exon 4) and ROS1 (exon 36) and the fusion breakpoint. (C) Representation of supporting reads and coverage spanning the fusion breakpoint. better-blood brain barrier permeability, improved outcomes in 5. Charest A, Lane K, McMahon K, et al. Fusion of FIG to the receptor tyro- patients with lung cancers harboring ROS1 rearrangements sine kinase ROS in a glioblastoma with an interstitial del(6)(q21q21). Genes Chromosom Cancer 2003;37:58–71 who had brain metastases (42–44) and in mouse models of 6. Davare MA, Henderson JJ, Agarwal A, et al. Rare but recurrent ROS1 GOPC-ROS1 fusion-positive GBM (6). It is also notable that a fusions resulting from chromosome 6q22 microdeletions are targetable patient with a very similar tumor was treated, apparently suc- oncogenes in glioma. Clin Cancer Res 2018;24:6471–82 cessfully, with high-dose chemotherapy and autologous hema- 7. Charest A, Wilker EW, McLaughlin ME, et al. ROS fusion tyrosine kinase activates a SH2 domain-containing phosphatase-2/phosphatidylino- topoietic stem cell transplant (9). These advancements in sitol 3-kinase/mammalian target of rapamycin signaling axis to form molecular diagnosis and therapeutic options underscore the glioblastoma in mice. Cancer Res 2006;66:7473–81 importance of recognizing unusual oncogenic drivers in pedi- 8. Johnson A, Severson E, Gay L, et al. Comprehensive genomic profiling atric brain tumors such as those with GOPC-ROS1. of 282 pediatric low- and high-grade gliomas reveals genomic drivers, tumor mutational burden, and hypermutation signatures. Oncologist 2017; 22:1478–90 REFERENCES 9. Kiehna EN, Arnush MR, Tamrazi B, et al. Novel GOPC(FIG)-ROS1 fu- 1. Acquaviva J, Wong R, Charest A. The multifaceted roles of the receptor sion in a pediatric high-grade glioma survivor. J Neurosurg Pediatr 2017; tyrosine kinase ROS in development and cancer. Biochim Biophys Acta 20:51–5 2009;1795:37–52 10. Rikova K, Guo A, Zeng Q, et al. Global survey of phosphotyrosine sig- 2. Birchmeier C, Sharma S, Wigler M. Expression and rearrangement of the naling identifies oncogenic kinases in lung cancer. Cell 2007;131: ROS1 gene in human glioblastoma cells. Proc Natl Acad Sci USA 1987; 1190–203 84:9270–4 11. Suehara Y, Arcila M, Wang L, et al. Identification of KIF5B-RET and 3. Sharma S, Birchmeier C, Nikawa J, et al. Characterization of the ros1- GOPC-ROS1 fusions in lung adenocarcinomas through a comprehensive gene products expressed in human glioblastoma cell lines. Oncogene Res mRNA-based screen for tyrosine kinase fusions. Clin Cancer Res 2012; 1989;5:91–100 18:6599–608 4. Birchmeier C, O’Neill K, Riggs M, et al. Characterization of ROS1 12. Rimkunas VM, Crosby KE, Li D, et al. Analysis of receptor tyrosine ki- cDNA from a human glioblastoma cell line. Proc Natl Acad Sci USA nase ROS1-positive tumors in non-small cell lung cancer: Identification 1990;87:4799–803 of a FIG-ROS1 fusion. Clin Cancer Res 2012;18:4449–57

1098 J Neuropathol Exp Neurol • Volume 78, Number 12, December 2019 GOPC-ROS1 Fusion in Pediatric Gliomas

13. Gu TL, Deng X, Huang F, et al. Survey of tyrosine kinase signaling 29. Southey MC, Goldgar DE, Winqvist R, et al. PALB2, CHEK2 and ATM reveals ROS kinase fusions in human cholangiocarcinoma. PLoS One rare variants and cancer risk: Data from COGS. J Med Genet 2016;53: 2011;6:e15640 800–11 14. Uguen A, De Braekeleer M. ROS1 fusions in cancer: A review. Future 30. Sturm D, Orr BA, Toprak UH, et al. New brain tumor entities emerge Oncol 2016;12:1911–28 from molecular classification of CNS-PNETs. Cell 2016;164: 15. Cocce MC, Mardin BR, Bens S, et al. Identification of ZCCHC8 as fusion 1060–72 partner of ROS1 in a case of congenital glioblastoma multiforme with a 31. Torre M, Meredith DM, Dubuc A, et al. Recurrent EP300-BCOR fusions t(6; 12)(q21; q24.3). Genes Chromosomes Cancer 2016;55:677–87 in pediatric gliomas with distinct clinicopathologic features. J Neuropa- 16. Serrano J, Snuderl M. Whole genome DNA methylation analysis of hu- thol Exp Neurol 2019;78:305–14 man glioblastoma using illumina BeadArrays. Methods Mol Biol 2018; 32. Charest A, Lane K, McMahon K, et al. Association of a novel PDZ 1741:31–51 domain-containing peripheral Golgi protein with the Q-SNARE (Q-solu-

17. Capper D, Jones DTW, Sill M, et al. DNA methylation-based classifica- ble N-ethylmaleimide-sensitive fusion protein (NSF) attachment protein Downloaded from https://academic.oup.com/jnen/article/78/12/1089/5593615 by New York University user on 25 August 2020 tion of central nervous system tumours. Nature 2018;555:469–74 receptor) protein syntaxin 6. J Biol Chem 2001;276:29456–65 18. Sturm D, Witt H, Hovestadt V, et al. Hotspot mutations in H3F3A and 33. Charest A, Kheifets V, Park J, et al. Oncogenic targeting of an activated IDH1 define distinct epigenetic and biological subgroups of glioblas- tyrosine kinase to the Golgi apparatus in a glioblastoma. Proc Natl Acad toma. Cancer Cell 2012;22:425–37 Sci USA 2003;100:916–21 19. Wiestler B, Capper D, Hovestadt V, et al. Assessing CpG island methyla- 34. Zhao JF, Sharma S. Expression of the ROS1 oncogene for tyrosine receptor phenotype, 1p/19q codeletion, and MGMT promoter methylation tor kinase in adult human meningiomas. Cancer Genet Cytogenet 1995; from epigenome-wide data in the biomarker cohort of the NOA-04 trial. 83:148–54 Neuro Oncol 2014;16:1630–8 35. Davies KD, Doebele RC. Molecular pathways: ROS1 fusion proteins in 20. Wiestler B, Capper D, Sill M, et al. Integrated DNA methylation and cancer. Clin Cancer Res 2013;19:4040–5 copy-number profiling identify three clinically and biologically relevant 36. Lim SM, Choi J, Chang JH, et al. Lack of ROS1 gene rearrangement in groups of anaplastic glioma. Acta Neuropathol 2014;128:561–71 glioblastoma multiforme. PLoS One 2015;10:e0137678 21. Orillac C, Thomas C, Dastagirzada Y, et al. Pilocytic astrocytoma and 37. Das A, Cheng RR, Hilbert ML, et al. Synergistic effects of crizotinib and glioneuronal tumor with histone H3 K27M mutation. Acta Neuropathol temozolomide in experimental FIG-ROS1 fusion-positive glioblastoma. Commun 2016;4:84 Cancer Growth Metastasis 2015;8:51–60 22. Huse JT, Snuderl M, Jones DT, et al. Polymorphous low-grade neuroepi- 38. Brennan CW, Verhaak RG, McKenna A, et al. The somatic genomic thelial tumor of the young (PLNTY): An epileptogenic neoplasm with landscape of glioblastoma. Cell 2013;155:462–77 oligodendroglioma-like components, aberrant CD34 expression, and ge- 39. Singh D, Chan JM, Zoppoli P, et al. Transforming fusions of FGFR and netic alterations involving the MAP kinase pathway. Acta Neuropathol TACC genes in human glioblastoma. Science 2012;337:1231–5 2017;133:417–29 40. Shaw AT, Ou SH, Bang YJ, et al. Crizotinib in ROS1-rearranged non- 23. Richardson TE, Snuderl M, Serrano J, et al. Rapid progression to glio- small-cell lung cancer. N Engl J Med 2014;371:1963–71 blastoma in a subset of IDH-mutated astrocytomas: A genome-wide anal- 41. Drilon A, Siena S, Ou SI, et al. Safety and antitumor activity of the multi- ysis. J Neurooncol 2017;133:183–92 targeted pan-TRK, ROS1, and ALK inhibitor entrectinib: Combined 24. Pennington KP, Walsh T, Lee M, et al. BRCA1, TP53, and CHEK2 results from two phase i trials (ALKA-372-001 and STARTRK-1). Can- germline mutations in uterine serous carcinoma. Cancer 2013;119:332–8 cer Discov 2017;7:400–9 25. Walsh T, King MC. Ten genes for inherited breast cancer. Cancer Cell 42. Shaw AT, Solomon BJ, Besse B, et al. ALK resistance mutations and ef- 2007;11:103–5 ficacy of lorlatinib in advanced anaplastic lymphoma kinase-positive 26. Ma X, Zhang B, Zheng W. Genetic variants associated with colorectal non-small-cell lung cancer. J Clin Oncol 2019;37:1370–9 cancer risk: Comprehensive research synopsis, meta-analysis, and epide- 43. Solomon BJ, Besse B, Bauer TM, et al. Lorlatinib in patients with ALK- miological evidence. Gut 2014;63:326–36 positive non-small-cell lung cancer: Results from a global phase 2 study. 27. Havranek O, Kleiblova P, Hojny J, et al. Association of germline Lancet Oncol 2018;19:1654–67 CHEK2 gene variants with risk and prognosis of non-Hodgkin lym- 44. Shaw AT, Felip E, Bauer TM, et al. Lorlatinib in non-small-cell lung phoma. PLoS One 2015;10:e0140819 cancer with ALK or ROS1 rearrangement: An international, multicentre, 28. Siołek M, Cybulski C, Ga˛sior-Perczak D, et al. CHEK2 mutations and open-label, single-arm first-in-man phase 1 trial. Lancet Oncol 2017;18: the risk of papillary thyroid cancer. Int J Cancer 2015;137:548–52 1590–9

1099