Durable Response to Larotrectinib in a Child With Histologic Diagnosis of Recurrent Disseminated Ependymoma Discovered to Harbor an NTRK2 Fusion: The Impact of Integrated Genomic Profiling

INTRODUCTION

Ependymomas comprise approximately 10 percent of childhood CNS malignancies and are associated with dismal outcomes if metastatic, incompletely resected, or recurrent.¹ Standard therapy consists of maximal safe surgical resection followed by focal irradiation.² Recurrent fusion genes involving C11orf95-RELA or YAP1 have been identified in supratentorial ependymomas, but targetable molecular lesions are rare.^3,4 Here, we report a sustained response to larotrectinib in a pediatric patient with a histologic diagnosis of recurrent metastatic anaplastic ependymoma, whose molecular testing unexpectedly revealed a KANK1-NTRK2 fusion and methylation profile most consistent with pleomorphic xanthoastrocytoma (PXA). This case highlights the potentially profound clinical implications of integrating genomic profiling into the diagnostic evaluation of pediatric CNS tumors.

CASE REPORT

A previously healthy 6-year-old boy presented with 2 weeks of abnormal gait, progressive headaches, vomiting, and somnolence, eventually culminating in generalized tonic-clonic seizure. On examination, he was lethargic and disoriented, with anisocoria, nuchal rigidity, and generalized weakness. Magnetic resonance imaging (MRI) revealed a 3.7 × 3.1 × 3.7-cm hypervascular, diffusion-restricted, right frontal horn mass obstructing the foramen of Monro, as well as diffuse leptomeningeal dissemination (Fig 1A).

FIG 1.

Magnetic resonance imaging of brain and spine. (A) At time of diagnosis, axial postcontrast image of the brain (left) demonstrated a large enhancing mass in the right frontal horn (white arrow) and leptomeningeal dissemination (black arrows). A sagittal postcontrast image of the cervical and thoracic spine (right) revealed extensive enhancement along the anterior and posterior aspects of the cord (white arrowheads) related to disseminated leptomeningeal disease. (B) At the time of recurrence (7 months postradiation therapy), axial postcontrast image of the brain (left) showed enhancing nodules along the margins of the ventricles related to tumor dissemination (white arrows). Sagittal postcontrast image of the cervical and thoracic spine (right) demonstrated nodular enhancement along the anterior surface of the cord (white arrowheads) and thicker diffuse leptomeningeal dissemination along the dorsal aspect of the cord (black arrowheads). (C) After 10 months of treatment with larotrectinib, axial postcontrast image of the brain (left) demonstrated marked and sustained improvement of tumor nodules along the posterior lateral ventricles (white arrows). A sagittal postcontrast image of the cervical and thoracic spine (right) also confirmed marked improvement of nodular tumor dissemination anterior to the cord (white arrowheads) as well as marked decrease in the thick leptomeningeal dissemination along the dorsal aspect of the cord (black arrowhead).

Following resection of the primary tumor, histopathologic examination demonstrated a poorly differentiated tumor with regional papillary features and immunohistochemical staining positive for glial fibrillary acidic protein, epithelial membrane antigen, integrase interactor 1, and synaptophysin, negative for pan-cytokeratin, and regional strong nuclear staining for p53. A diagnosis of WHO grade III anaplastic supratentorial ependymoma with papillary features was made (Figs 2A-2F), and the patient received proton craniospinal irradiation to a dose of 39.6 Gy (relative biologic dose effectiveness) followed by primary tumor bed boost to a total dose of 54 Gy (relative biologic dose effectiveness). He recovered with moderate improvement in cognitive speed, language processing, and overall gross and fine motor function. However, 7 months after completion of radiation therapy, diffuse tumor recurrence was identified on surveillance MRI, including leptomeningeal carcinomatosis lining the ventricular walls and anterior and dorsal surfaces of the spinal cord (Fig 1B).

FIG 2.

Histopathologic tumor features. Hematoxylin and eosin staining (A) demonstrated ependymoma with papillary configuration and foci of vascular proliferation (arrow). (B and D) Areas with hyperchromatic nuclei with frequent mitotic figures (arrowheads), and (C) region with clear cell features are shown. (E) Immunostain for epithelial membrane antigen with membranous and dot-like positivity, and (F) immunostain for glial fibrillary acidic protein. Magnification: (A) × 40, (B) × 400, (C) × 100, (D) × 200, and (E and F) × 200.

Tumor molecular profiling using the Texas Children's Hospital (TCH) Comprehensive Solid Tumor Panel^5,6 was performed at the time of diagnosis. DNA analysis revealed no clinically relevant mutations, whereas RNA analysis unexpectedly identified reads representing fusion transcripts spanning the junction between KANK1 exon 3 (chr9:713464) and NTRK2 exon 16 (chr9:87482158) predicted to produce an in-frame KANK1-NTRK2 chimeric protein retaining the tyrosine kinase domain of NTRK2 (Fig 3). More extensive clinical genomic tumor profiling (exome sequencing, transcriptome sequencing, and copy-number array) performed as part of the institutional review board-approved Texas KidsCanSeq study, in addition to confirming the KANK1-NTRK2 fusion, revealed a homozygous 2.3-Mb deletion on 9p21.3 harboring the CDKN2A and CDKN2B genes (Appendix Fig A1) with copy neutral loss of heterozygosity of the rest of chromosome 9 including both the KANK1 (chr 9p24.3) and NTRK2 gene loci (chr 9q21.3), suggesting a possible chromosomal duplication following rearrangement. Other structural variants included heterozygous deletion of chromosome 22 and gain of chromosome 8 locus. Tumor mutation burden was low (0.15 mut per Mb). There were no pathogenic or likely pathogenic germline findings from panel and exome analysis of a blood sample.

FIG 3.

Tumor transcriptome sequencing identified an intrachromosomal rearrangement on chromosome 9, leading to an in-frame fusion between exons 1-3 of the KN motif and ankyrin domains 1 gene (KANK1) and exons 16-21 of the neurotrophic receptor tyrosine kinase 2 (NTRK2), predicted to retain the full-length kinase domain of NTRK2.

Therapy was initiated with the US Food and Drug Administration–approved tropomyosin-related kinase (TRK)-inhibitor larotrectinib. Within 2 weeks of treatment initiation, the patient demonstrated improved mentation speed, gait steadiness, and fluidity of movement. After 4 months, he conversed at near-baseline cognitive speed, exhibited intact short- and long-term memory, and no evidence of ataxia. Follow-up MRI of the brain and spine after 2 months of therapy revealed significant tumor response with marked decrease in diffuse leptomeningeal carcinomatosis. Both clinical and radiographic responses have been sustained through 10 months of therapy without medication-related toxicity (Fig 1C).

Tumor DNA methylation profiling was conducted using the Illumina DNA Methylation EPIC array, visualized with t-Stochastic neighbor embedding analysis, and classified using a Random Forest prediction model assembled from a published pediatric brain tumor reference cohort.⁷ The patient's sample clustered most closely with PXA (methylation classifier score of 0.64; Fig 4). No clustering with ependymal tumor subtypes or other CNS tumor types was observed (all other classifier scores ≤ 0.0527).

FIG 4.

Tumor methylation profiling. Unsupervised clustering of the tumor with a reference cohort (N = 2,801 brain tumor cases) using t-SNE dimensionality reduction showed our patient's tumor (black, indicated by arrow) clustered most closely with the PXA methylation class and appeared to be distinct from ependymoma YAP- and ependymoma RELA-fusion methylation classes. The reference cohort consists of 91 methylation classes (82 CNS methylation classes and nine control tissue methylation classes), represented in a color-coded fashion. PXA, pleomorphic xanthoastrocytoma; t-SNE, t-stochastic neighbor embedding.

METHODS

Consent was obtained to the Baylor College of Medicine (BCM) Institutional Review Board–approved Texas KidsCanSeq Study, a National Human Genome Research Institute and National Cancer Institute–funded Clinical Sequencing Evidence-Generating Research Consortium project⁸ comparing the clinical utility of targeted gene panel tumor testing versus more comprehensive germline and tumor genomic profiling.

Primary tumor tissue was used for standard diagnostic histopathologic evaluation and molecular testing, and a blood sample was later obtained at study enrollment. Clinical sequencing of DNA or RNA extracted from a formalin-fixed paraffin-embedded tumor sample and blood DNA was performed in College of American Pathologists– and Clinical Laboratory Improvement Amendments–certified laboratories at TCH and Baylor College of Medicine. Targeted tumor panel testing with the TCH Comprehensive Solid Tumor Panel consisted of a custom-designed next-generation sequencing assay using targeted DNA and RNA sequencing to detect mutations in coding and splicing regions of 2,247 exons in 124 genes as well as TERT promoter and gene fusions in 81 genes.^5,6

Tumor exome sequencing, transcriptome sequencing, and copy-number array were also performed, as well as blood exome and targeted cancer gene testing using the same TCH DNA panel. Exome sequencing—including library construction and exome capture by VCRome version 2.1 (Roche NimbleGen) supplemented with two probe sets to enhance the representation of clinically relevant genes—was performed from 250 ng of DNA followed by paired-end sequencing on Illumina HiSeq 2500 to a mean coverage of 209× and 143× (tumor and blood, respectively). Data analysis and somatic variant detection was performed as described previously.⁹ Capture-exome transcriptome sequencing was performed using 100-ng total RNA to prepare strand-specific RNA-seq libraries enriched for the coding exome, also using the supplemented VCRome version 2.1, and sequenced on Illumina HiSeq 2500. Genome-wide copy-number variation and loss of heterozygosity profiling was performed with genomic tumor DNA using the Affymetrix OncoScan FFPE Assay Kit.

DISCUSSION

Comprehensive tumor molecular profiling yielded unexpected results for a histologic diagnosis of ependymoma, significantly altering our patient's therapeutic course and providing insight into the diagnosis of an unusual pediatric CNS tumor. This case illustrates the ongoing challenge (and opportunity) to effectively integrate molecular results with histologic and morphologic findings, particularly when they appear potentially discordant. The unanticipated identification of a KANK1-NTRK2 fusion in this tumor provided a biologic rationale for targeted TRK inhibition of a recurrent tumor with poor curative potential. Consistent with early reports of the clinical efficacy and tolerability of larotrectinib in pediatric patients with primarily non-CNS TRK-fusion–positive cancers,^10-12 our patient has experienced an excellent clinical and radiologic response to date. However, agents with US Food and Drug Administration–approved pediatric indications remain uncommon. Clinical trials investigating molecularly targeted agents as upfront therapy are ongoing and data are still emerging to guide such treatment, including optimal duration of therapy.

Fusions involving the NTRK family of genes (NTRK1, NTRK2, and NTRK3) are oncogenic drivers in multiple cancer types via constitutive activation of kinase signaling pathways (MAPK, RAS/ERK, and PI3K/AKT).^13-15 NTRK fusions have been reported in < 2% of primary CNS glial or neuroepithelial tumors^16-18 and up to 4-10 percent of pediatric high-grade gliomas,^19-21 including a pediatric PXA harboring a TPM3-NTRK1 fusion,²² as well as KANK1-NTRK2 fusions in a pilocytic astrocytoma with anaplasia¹⁸ and a glioblastoma.¹⁶ However, such fusions are not characteristic of either pediatric or adult ependymomas. A large series of NTRK-fusion–positive gliomas included the case of a 42-year-old adult with a tumor harboring a MYO5A-NTRK3 as well as CDKN2A and CDKN2B (9p) and RB1 (13q) loss, despite histology and an immunohistochemical profile resembling anaplastic ependymoma. Additionally, two pediatric patients with NTRK2-positive tumors were identified with a methylation classifier score closest with PXA including one case with a KANK1-NTRK2 fusion,²⁰ as observed with our patient.

In our patient's case, upfront therapy for either anaplastic ependymoma (as diagnosed by histology or morphology) or PXA (per molecular or methylation profile) would have been similar, in consideration of his symptoms, age, and widespread metastatic disease at presentation. Both diagnoses warrant irradiation to the entire neuroaxis with boost to the primary tumor site, although different tumor bed doses (eg, 45-50.4 Gy for WHO grade II tumors like PXA v 54-59.4 Gy for higher-grade tumors including anaplastic ependymoma) may be indicated. Additionally, recurrence of either distinct diagnosis would affect potential treatment options and clinical trial eligibility.

In contrast with clinical molecular tests performed in College of American Pathologists– and Clinical Laboratory Improvement Amendments–certified laboratories, methylation profiling was conducted on a research basis. As demonstrated by this case, incorporation of methylation profiling into clinical genomic analysis has the potential for significant diagnostic utility, in particular for tumor entities that lack defining molecular alterations. This is expected to become increasingly relevant as data continue to emerge regarding clinical characteristics of molecularly defined tumor subtypes and refined molecular diagnoses become increasingly critical for proper interpretation of future clinical trials. Prospective studies of larger cohorts of patients with pediatric cancer such as with the Texas KidsCanSeq study will provide insight into the frequency at which clinically relevant genetic alterations can be detected using different genomic tests. Our case provides an example of the profound impact such testing can have for both diagnosis and effective treatment of children with high-risk pediatric cancers.

Appendix

FIG A1.

Copy-number array demonstrating homozygous deletion of a 2.3-Mb interval on chromosome 9p21.3-p21.3 harboring the CDKN2A and CDKN2B genes.

AUTHOR CONTRIBUTIONS

Conception and design: Ross Mangum, Marcia K. Kukreja, Stephen C. Mack, Sharon E. Plon, Angshumoy Roy, D. Williams Parsons, Frank Y. Lin

Financial support: Richard A. Gibbs, D. Williams Parsons, Frank Y. Lin

Administrative support: Frank Y. Lin

Provision of study materials or patients: Ross Mangum, Arnold C. Paulino, Murali M. Chintagumpala, Angshumoy Roy, Frank Y. Lin

Collection and assembly of data: Ross Mangum, Jacquelyn Reuther, Kelsey C. Bertrand, Marcia K. Kukreja, Donna Muzny, Jianhong Hu, Richard A. Gibbs, Fatema Malbari, Stephen C. Mack, Sharon E. Plon, Angshumoy Roy, D. Williams Parsons, Frank Y. Lin

Data analysis and interpretation: Ross Mangum, Jacquelyn Reuther, Kelsey C. Bertrand, Raghu Chandramohan, Marcia K. Kukreja, Arnold C. Paulino, Daniel J. Curry, Murali M. Chintagumpala, Adekunle M. Adesina, Kevin E. Fisher, Stephen C. Mack, Angshumoy Roy, D. Williams Parsons, Frank Y. Lin

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by the authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO’s conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).