Molecular and clinicopathologic features of gliomas harboring NTRK fusions

Fusions involving neurotrophic tyrosine receptor kinase (NTRK) genes are detected in ≤2% of gliomas and can promote gliomagenesis. The remarkable therapeutic efficacy of TRK inhibitors, which are among the first Food and Drug Administration-approved targeted therapies for NTRK-fused gliomas, has generated significant clinical interest in characterizing these tumors. In this multi-institutional retrospective study of 42 gliomas with NTRK fusions, next generation DNA sequencing (n = 41), next generation RNA sequencing (n = 1), RNA-sequencing fusion panel (n = 16), methylation profile analysis (n = 18), and histologic evaluation (n = 42) were performed. All infantile NTRK-fused gliomas (n = 7) had high-grade histology and, with one exception, no other significant genetic alterations. Pediatric NTRK-fused gliomas (n = 13) typically involved NTRK2, ranged from low- to high-histologic grade, and demonstrated histologic overlap with desmoplastic infantile ganglioglioma, pilocytic astrocytoma, ganglioglioma, and glioblastoma, among other entities, but they rarely matched with high confidence to known methylation class families or with each other; alterations involving ATRX, PTEN, and CDKN2A/2B were present in a subset of cases. Adult NTRK-fused gliomas (n = 22) typically involved NTRK1 and had predominantly high-grade histology; genetic alterations involving IDH1, ATRX, TP53, PTEN, TERT promoter, RB1, CDKN2A/2B, NF1, and polysomy 7 were common. Unsupervised principal component analysis of methylation profiles demonstrated no obvious grouping by histologic grade, NTRK gene involved, or age group. KEGG pathway analysis detected methylation differences in genes involved in PI3K/AKT, MAPK, and other pathways. In summary, the study highlights the clinical, histologic, and molecular heterogeneity of NTRK-fused gliomas, particularly when stratified by age group.


Introduction
The tropomyosin receptor kinase (TRK) family of tyrosine receptor kinases is comprised of TRKA, TRKB, and TRKC, which are encoded by neurotrophic tyrosine receptor kinase (NTRK) genes NTRK1, NTRK2, and NTRK3, respectively. The three TRK proteins are structurally similar, with an extracellular region containing leucine-rich repeats, cysteine-rich clusters, and immunoglobulin-like domains, a transmembrane region, and an intracellular region including a tyrosine kinase domain [2]. Binding of neurotrophin ligands to the extracellular region triggers TRK dimerization and transphosphorylation of tyrosine residues within the activation loop of the kinase domain, which ultimately results in the upregulating of multiple pathways including mitogen-activated protein kinase (MAPK), phosphatidylinositol-3-kinase / protein kinase B (PI3K/AKT), and phospholipase C-γ (PLC-γ) signaling cascades [27]. TRK receptors are highly expressed in neural tissue, where they have a physiologic role in neuronal survival, development, proliferation, and synaptic plasticity, as well as memory and cognition [2].
Fusions involving the NTRK genes can be oncogenic drivers and typically involve the 5′ end of the fusion partner and the 3′ end of NTRK preserving the tyrosine kinase domain. Reported gene fusion partners are numerous and in many cases contain structural motifs such as coiled-coil domains and zinc finger domains that promote dimerization [10]. Thus, oncogenic NTRK fusions can result in aberrant ligand-independent TRK receptor dimerization and constitutive activation of TRK signaling pathways [3], leading to upregulated proliferation and resistance to apoptosis. NTRK fusions in which the fusion partners lack dimerization domains might alternatively promote tumorigenesis through loss of extracellular TRK regulatory domains [5].
Clinical interest in NTRK-fused tumors has increased substantially due to the efficacy of Food and Drug Administration (FDA) approved TRK inhibitor therapies [4, 13, 15-17, 23, 30]. The aim of the current study is to provide insights into the clinicopathologic and molecular features of gliomas with NTRK fusions.

Cohort
The surgical, consultation, and molecular pathology archives of Brigham and Women's Hospital (BWH) (Boston, MA), Boston Children's Hospital (BCH) (Boston, MA), Children's Hospital of Philadelphia (CHOP) (Philadelphia, PA), Washington University School of Medicine (WashU) (St Louis, MO), Northwestern University (NWU) (Chicago, IL), and Foundation Medicine (FM) (Morrisville, NC) were reviewed for gliomas with NTRK rearrangements. In this retrospective multi-institutional study, a total of 42 cases were identified, composed of 8 cases from BWH, 7 cases from BCH, 5 cases from CHOP, 1 case from WashU, 1 case from NWU, and 20 cases from FM. The study contains 2 cases (cases 7 and 15) that have been previously reported in the literature [31,46]. The study was conducted under BCH IRB protocol IRB-CR00027359-1 and DFCI protocol 10-417. The cases were grouped in infantile (age less than 1 year), pediatric (age ranging from 1 year to 18 years), and adult (age over 18 years). Available routine hematoxylin and eosin stained sections and immunohistochemical stains prepared from formalin-fixed, paraffin-embedded (FFPE) tissue from the 42 identified cases underwent review by neuropathologists (SA, MT, SHR, MSa, CH), with 22 of the cases undergoing central review by SA; all tumors with material for methylation were centrally-reviewed. In general, there was agreement with the initial clinical diagnosis, and a specific World Health Organization (WHO) diagnosis was sought whenever the histology allowed it. A complete set of slides was not available for the 20 cases from FM; however, these were all reviewed by one neuropathologist (SHR). A subset of the pediatric tumors had concerning histology, with occasional mitoses and pleomorphism, but, overall, these features did not reach the threshold for WHO histologic grade 3. The histologic diagnoses in cases that posed this challenge were: glioma with anaplastic features (4 cases) and anaplastic pilocytic astrocytoma (APA) (1 case). Therefore, a specific histologic grade could not be assigned for these tumors and they are referred to as having "uncertain WHO histologic grade" in the manuscript. Patient information was abstracted from the electronic medical records or from the clinical information provided on the pathology report.

Figures
The Oncoprint figure was created using R 3.6.0, RStudio 1.2.1335, and the Oncoprint function of the Complex-Heatmap 2.2.0 package. The Circos plot was generated using the online Circos Table Viewer (http://mkweb. bcgsc.ca/tableviewer). All other figures were created using GraphPad Prism (v.8) software.

Next generation sequencing (NGS)
NTRK rearrangements were detected by either DNAbased next generation sequencing (NGS) or RNA-based fusion panel performed at the time of clinical diagnosis. Given that this is a retrospective multi-institutional study, a limitation is that the NGS panels utilized are institution-specific (albeit similar in coverage of genes of interest and scope).
Oncopanel interrogates the exons of 447 genes and 191 introns across 60 genes, and structural rearrangements are evaluated with BreaKmer analysis as previously described [20]. The Foundation Medicine NGS assay evaluates 324 genes for mutations and copy number alterations, as well as select intronic regions of a subset of genes to detect gene rearrangements. Details about the Foundation Medicine NGS assay can be found at https://www.foundationmedicine.com/genomic-testing/foundation-one-cdx. The CHOP Comprehensive Solid Tumor Panel includes sequencing and copy number analysis of 237 genes as well as targeted RNA-based anchored multiplex PCR using custom probes for over 100 genes, as previously described [49]. Caris Life Sciences performs exome sequencing on 592 genes for mutational analysis, evaluates a proportion of these genes for copy number alterations, and assesses for fusions involving targeted genes with RNA-based anchored multiplex PCR (https://www.carismolecularintelligence.com/ profiling-menu/mi-profile-usa-excluding-new-york/). GlioSeq uses amplification-based DNA and RNA sequencing to evaluate for mutations, copy number alterations, and structural rearrangements involving genes relevant to primary central nervous system (CNS) tumors. A list of genes included in the GlioSeq panel can be accessed at https://mgp.upmc.com/Home/Test/Glio-Seq_details. Copy number data was determined from DNA-based NGS and methylation profile plots.

DKFZ CNS tumor classification of NTRK gliomas
Genome-wide methylation profiling was performed on DNA extracted from FFPE tissue from 18 cases with available material using the Illumina EPIC Array 850 Bead-Chip (850 k) array to evaluate the DNA methylation status of over 850,000 CpG sites, as described previously [40]. The raw idat files were then analyzed by the Brain Tumor Classifier developed by Capper et al. [7], which is clinically validated at NYU. Each NTRK fusion case was compared against the CNS reference tumor cohort (82 methylation classes and 9 control tissues) using the Random Forest Classifier. The classifier generates Methylation classifier scores for each sample along with t-distributed stochastic neighbor embedding (tSNE) dimensionality reduction of queried samples against the reference cohort classes.

NTRK cohort genome-wide methylation profiling and analysis
To analyze the NTRK cohort in our study, the raw idats generated from iScan were processed and analyzed using Bioconductor R package Minfi. All the Illumina EPIC array probes were normalized using quintile normalization and corrected for background signal. Samples were then checked for their quality using mean detection p-values (p-value < 0.05). Unsupervised principal component analysis (PCA) was performed to check for biological variation within the cases. To identify the differentially methylated CpG probes, the samples were grouped based on NTRK gene involved, histologic grade, and age. Beta values were generated and probes with FDR cutoff (q < 0.05) were considered the most significantly variable probes. Heatmaps were generated in a semi-supervised manner, showing the hierarchical clustering pattern of the top 10,000 significant differentially methylated genes/probes by NTRK gene involved. KEGG pathway analysis with ClusterProfiler [54] was used to identify the signaling pathways enriched in the top 10,000 most variable genes/probes.
The types of molecular assays performed on each case are listed in Supplemental Table 1.
Targeted therapy with larotrectinib was administered in 3 patients: 2 pediatric patients (one of whom showed a decrease in tumor burden ( Fig. 2a-b), and one of whom has shown stable disease), and 1 infantile patient whose course of larotrectinib was terminated due to elevated liver function tests.

Methylation profiling of NTRK-fused gliomas
Methylation profiling with clustering analysis was performed on 18 cases with available material. Two tumors matched to known methylation class families with high confidence (Fig. 6, Table 3): case 3, an infantile HGG with features of PXA, matched to methylation class family PXA (calibrated score = 0.989) and case 14, a 3month-old with a histologic diagnosis of GBM, matched to infantile hemispheric glioma (IHG, calibrated score = 0.9836). In both instances, the histology was consistent with the matched methylation class family. All other cases matched with low confidence or not at all to known methylation class families (i.e. scores were lower than the recommended threshold value of ≥0.9 [7] or the less conservative threshold of ≥0.84 [8]. Seven cases had methylation classifier scores between 0.5 and 0.84 [8], with 2 (cases 9 and 21) having histology consistent with the closest methylation class family. Overall, a disproportionately high number of case either classified with calibrated score < 0.9 or did not classify with any reference cohort compared to previously published data [7] suggesting perhaps that NTRK fusions alter the DNA methylation pattern from non-NTRK driven cases of similar histology (Table 3; also please see Supplemental Table 3 for link to all methylation reports and t-sne plots). While some newly described CNS tumor entities driven by gene fusions form unique entities [45], unsupervised PCA of the methylation profiles of NTRKfused cohort showed no obvious grouping by NTRK gene involved, histologic grade, or age (Supplemental Fig. 1). KEGG pathway analysis of the top 10,000 most variably methylated genes/probes in the cohort demonstrated enrichment in pathways involving PI3K-AKT signaling (and related human papillomavirus infection signaling) and MAPK signaling, among others (Fig. 7).
One of the findings of this multi-institutional study is the considerable clinicopathologic and molecular heterogeneity of NTRK-fused gliomas. NTRK fusions do not appear to define ipso facto a single glial entity but rather are a genetic feature occurring in multiple tumor types.
The present study adds to the literature by demonstrating that the histology and histologic grade of NTRK-fused gliomas vary by patient age. NTRK-fused gliomas in all infantile and most adult patients were histologically high-grade, with the majority diagnosed as GBM. In contrast, pediatric NTRK-fused gliomas were more likely to be of low-grade (46.2%) or uncertain WHO grade (38.5%), and there was no single predominant histologic diagnosis in this cohort. These features of the pediatric NTRK-fused gliomas make their diagnosis and clinical management difficult.
Gliomas with NTRK fusions have been previously reported to possess co-occurring genetic alterations such as IDH [18,21,39,56]  In contrast, the vast majority of NTRK-fused gliomas diagnosed in pediatric patients are of low-grade histology or of uncertain WHO grade. (Bottom) Histologic diagnoses of NTRK-fused gliomas: there were 12 unique histologic diagnoses assigned to NTRK-fused glioma in the study. Slightly less than half of all cases were diagnosed as GBM. Within the infantile and adult age cohorts, the majority of cases were diagnosed as GBM. A more diverse spectrum of tumors was diagnosed in the pediatric age cohort for which there was no single predominant histologic diagnosis     [21]. Our study matches many of these molecular findings and further demonstrates that the frequency of pathologically significant mutations in NTRK-fused gliomas appears to increase with patient age. In addition, TERT promoter mutations are observed only in histologically high-grade adult tumors, PTEN alterations are almost exclusively seen in histologically high-grade tumors, and CDKN2A/2B loss is rare in histologically low-grade tumors. Notably, 22.7% of adult NTRK-fused gliomas in our cohort are IDH1 p.R132H mutated, raising questions about the oncogenic driver event in these specific tumors and whether they display oncogenic dependence on the NTRK fusion. In one case report of secondary IDH-mutant GBM [39], an NTRK fusion was detected in only a subclonal tumor population and was absent in the original AA, suggesting that the NTRK fusion was a secondary alteration. In contrast, in NTRK-fused gliomas without co-occurring pathologically significant mutations [9,21,52], typically arising in younger patients, NTRK fusions are almost certainly the oncogenic driver event. This is supported by multiple in vivo models that have demonstrated the capability of NTRK fusions to drive gliomagenesis/tumorgenesis [11,28,33,51,52].
In our series, gliomas with rearrangements involving the same NTRK gene and fusion partner do not necessarily have the same histology or methylation class, which suggests that other factors such as age, co-occurring genetic events, cell of origin and microenvironment potentially play an important role in tumor biology. We have also observed that the NTRK gene involved in the rearrangement differs in frequency by age, with pediatric gliomas having a high percentage of rearrangements involving NTRK2 (69.2%), and adult gliomas having a high percentage of rearrangements involving NTRK1 (68.2%).
Most NTRK-fused gliomas in our cohort were not a perfect match with known methylation entities. Only two cases matched with high confidence to methylation class families. Furthermore, cases that matched with low confidence generally had histology that was not characteristic of the methylation class family, mirroring the experience described in a prior case study of NTRKfused glioneuronal tumor [29]. Another study reported a proportion of NTRK-fused gliomas matching to methylation classes including IHG, diffuse midline glioma H3 K27M mutant, PXA, and GBM, IDH wildtype, subclass midline [9]. Overall, the findings from our study and prior studies with methylation data [9,29] suggest that better methylation profile classifier guidelines are needed to account for NTRK-fused. Our unsupervised PCA demonstrating no obvious grouping by NTRK gene involved, age, or histology highlights the heterogeneity within the NTRK-fused glioma methylome.
In summary, NTRK-fused gliomas are clinically, histologically, and molecularly diverse, with notable differences by age group and associated genetic alterations. Additional studies are needed to develop clinical guidelines for the diagnostic workup of potential NTRK-fused CNS tumors and to improve methylation classifier guidelines for NTRK-fused gliomas. Further mechanistic work is required to determine the role of NTRK fusions in driving Fig. 7 (a) Heatmap of the top 10,000 differentially methylated genes/probes by NTRK gene involved (blue indicates hypomethylation, and red indicates hypermethylation). (b) KEGG pathway analysis reveals several pathways enriched in the top differentially methylated gene/probes. The dot plots represent the ratio of genes (x-axis) involved in each signaling pathway (y-axis). The size of the dots shows the gene counts, and the color denotes the significance level gliomagenesis in the setting of concurrent oncogenic drivers such as IDH mutations and to demonstrate how downstream TRK signaling pathways may be mediated by different NTRK gene involved, location of NTRK fusion breakpoint, fusion partner, and cell of origin.