Molecular Alterations in Pediatric Low-Grade Gliomas That Led to Death

Abstract

Pediatric low-grade gliomas (PLGGs) have excellent long-term survival, but death can occasionally occur. We reviewed all PLGG-related deaths between 1975 and 2019 at our institution: 48 patients were identified; clinical data and histology were reviewed; targeted exome sequencing was performed on available material. The median age at diagnosis was 5.2 years (0.4–23.4 years), at death was 13.0 years (1.9–43.2 years), and the overall survival was 7.2 years (0.0–33.3 years). Tumors were located throughout CNS, but predominantly in the diencephalon. Diagnoses included low-grade glioma, not otherwise specified (n = 25), pilocytic astrocytoma (n = 15), diffuse astrocytoma (n = 3), ganglioglioma (n = 3), and pilomyxoid astrocytoma (n = 2). Recurrence occurred in 42/48 cases, whereas progression occurred in 10. The cause of death was direct tumor involvement in 31/48 cases. Recurrent drivers included KIAA1549-BRAF (n = 13), BRAF(V600E) (n = 3), NF1 mutation (n = 3), EGFR mutation (n = 3), and FGFR1-TACC1 fusion (n = 2). Single cases were identified with IDH1(R132H), FGFR1(K656E), FGFR1 ITD, FGFR3 gain, PDGFRA amplification, and mismatch repair alteration. CDKN2A/B, CDKN2C, and PTEN loss was recurrent. Patients who received only chemotherapy had worse survival compared with patients who received radiation and chemotherapy. This study demonstrates that PLGG that led to death have diverse molecular characteristics. Location and co-occurring molecular alterations with malignant potential can predict poor outcomes.

BRAF, MAPK pathway, NF, Pediatric low-grade glioma, Targeted exome sequencing

INTRODUCTION

Pediatric low-grade glioma (PLGG) is the most common brain tumor observed in children <19 years old and is associated with excellent long-term survival (1, 2). Several factors influence long-term survival in PLGG patients, including radiation therapy, age of diagnosis, histology, primary site, and degree of resection (2, 3).

Although PLGG have in common the designation of World Health Organization (WHO) grade 1 or 2, there is significant heterogeneity in this group of tumors. They occur throughout the neuroaxis, including the cerebral hemispheres, diencephalic structures (thalamus, hypothalamus, and third ventricle), brainstem, cerebellum, and spinal cord. There are also a variety of histologic entities included in this group, including pilocytic astrocytoma, ganglioglioma, diffuse astrocytoma, and pleomorphic xanthoastroctyoma (PXA) (4). Furthermore, many tumors do not fall neatly into any particular diagnostic entity and are referred to as low-grade glioma, not otherwise specified (LGG-NOS).

Most PLGG also have in common the upregulation of the mitogen-activated protein kinase (MAPK) pathway for tumorigenesis, including KIAA1549-BRAF gene fusion in the majority of pilocytic astrocytomas, BRAF V600E in ganglioglioma and the majority of PXA, and NF1 loss-of-function mutation in optic pathway glioma (5, 6). Recent publications highlight a wider molecular heterogeneity amongst PLGG, which includes activating point mutations and gene fusions involving FGFR1/2/3, MYB/MYBL1, and NTRK1/2/3 (5–9). IDH1/2 mutations are rare in young children but can be seen in diffuse gliomas in older children and adolescents (8, 10–12).

Large-scale next-generation sequencing studies reinforce the theme of molecular heterogeneity amongst PLGG and also highlight important risk stratification schemes (6). However, the molecular characteristics amongst deceased PLGG patients have not been specifically studied. Here, we analyze the molecular characteristics of tumors from deceased PLGG patients and integrate clinical, radiological, and treatment-related factors that may have influenced each patient’s course.

MATERIALS AND METHODS

With Institutional Board Approval, a total of 48 subjects were identified in the archives at Boston Children’s Hospital and Dana Farber Cancer Institute utilizing Queueview software to search clinical records for patients with a diagnosis of “LGG” and vital status of “expired.” Clinical research coordinators abstracted clinical data, including age at diagnosis, tumor location, number of surgical occurrences, treatment history, and date of death. Age at diagnosis was defined as the date of first surgical intervention, as presurgical imaging data were not available for every case. Clinical circumstances leading to death were evaluated by experienced neuro-oncologists (T.M.C. and P.B.) in order to categorize cause of death as direct tumor involvement (death due to tumor growth and/or progression), semidirect tumor involvement (death due unintended treatment- or surgery-related complications), unrelated to tumor (death due to factors not related to the patient’s tumor), or unknown (detailed clinical history surrounding patient death was unavailable).

Pathology slides were reviewed for all cases and a diagnosis of LGG on initial tissue was confirmed by 2 neuropathologists (S.A. and K.L.L.). Subsequent surgical samples and autopsy tissue was evaluated, when available, and examined for histologic evidence of tumor progression.

Targeted next-generation sequencing (Illumina HiSeq) was performed on DNA isolated from formalin-fixed paraffin-embedded tissue using a next-generation hybrid capture targeted exome sequencing assay (OncoPanel) that interrogates the exons of 447 genes and 191 introns across 60 genes for structural rearrangements (13).

Swimmer plots were generated in order to examine differences in survival and the relative occurrence of disease recurrence and/or progression based upon histologic examination of tumors from additional surgical procedures and/or autopsy-derived samples. Kaplan-Meier analysis was used to compare overall survival between groups based on location, histologic diagnosis and treatment. Log-rank test was used to determine statistical significance between groups. For single comparisons, statistical significance was considered p < 0.05. For multiple comparisons, significance was adjusted using Bonferroni correction. Statistical analyses were performed using GraphPad Prism software.

RESULTS

We searched the clinical database at our institution for pediatric patients diagnosed with LGG between the years 1975 and 2019 with a vital status of “expired.” In total, 48 patients were identified (outlined in Table 1). The male to female ratio was 27:21. Median age at diagnosis was 5.2 years (range: 0.4–23.4 years) and at death was 13.0 years (range: 1.9–43.2 years). Median overall survival was 7.2 years (range: 0.0–33.3 years). Tumors were located in the cerebral hemispheres (n = 11), diencephalon (including hypothalamus, thalamus, optic pathway, and third ventricle; n = 27), brainstem/spinal cord (n = 8), or cerebellum (n = 2).

TABLE 1.

Demographics of Expired Pediatric Low-Grade Glioma

Total	n = 48
Gender
Male	27 (56.2%)
Female	21 (43.8%)
Timepoints: median (range)
Age at diagnosis, years	5.2 (0.4-23.4)
Age at death, years	13.0 (1.9-43.2)
Overall survival, years	7.2 (0.0-33.3)
Histologic diagnosis
LGG-NOS	25 (52.1%)
Pilocytic astrocytoma, WHO I	15 (31.3%)
Diffuse astrocytoma, WHO II	3 (6.2%)
Ganglioglioma, WHO I	3 (6.2%)
Pilomyxoid astrocytoma	2 (4.2%)
Tumor location
Cerebral hemispheres	11 (22.9%)
Diencephalon	27 (56.2%)
Brainstem/spinal cord	8 (16.7%)
Cerebellum	2 (4.2%)
Treatment history
Chemotherapy only	17 (35.4%)
Radiation only	2 (4.2%)
Chemotherapy + radiation	18 (37.5%)
Neither	10 (20.8%)
Unknown	1 (2.1%)
Cause of death
Direct tumor involvement	31 (64.6%)
Semidirect tumor involvement	6 (12.5%)
Unrelated to brain tumor	5 (10.4%)
Unknown	6 (12.5%)
Autopsy
Yes	20 (41.7%)
No	28 (58.3%)

Surgical and autopsy pathology slides were centrally reviewed by 2 neuropathologists (S.A. and K.L.L.) and a diagnosis of LGG was confirmed on the first surgical specimen for each patient. Histologic diagnoses in this cohort included pilocytic astrocytoma (n = 15), diffuse astrocytoma (n = 3), ganglioglioma (n = 3), and pilomyxoid astrocytoma (n = 2). LGG-NOS was applied to tumors that did not meet criteria for any specific histologic entity (n = 25). Interestingly, there were no examples of PXA in our cohort. Pilocytic astrocytoma and LGG-NOS were observed most frequently in the diencephalon (10/15 and 13/25 cases, respectively). Diffuse astrocytoma was located once each in the diencephalon, brainstem, and cerebral hemispheres. Ganglioglioma was observed in the cerebral hemispheres in 2 cases and once in the diencephalon. Both pilomyxoid astrocytomas were observed in the diencephalon. Histologic progression to a higher-grade tumor in subsequent surgical samples or at autopsy was observed in 10 cases (Table 2).

TABLE 2.

Disease Course

Disease recurrence/progression (R/P)
At least 1 R/P	42
At least 2 R/P	17
Three or more R/P	4
Timepoints: median (range)
Time to first R/P, months	40.5 (0.2–220.0)
Time to second R/P, months	37.9 (3.3–133.2)
Time to third R/P, months	10.8 (2.4–14.9)
Histologic progression
Present	10 (20.8%)

Treatment history was abstracted from the clinical records (outlined in Tables 1 and 2) and included only chemotherapy (n = 17), radiation therapy (n = 2), both chemotherapy and radiation therapy (n = 18), and neither chemotherapy nor radiation therapy (n = 10). Treatment history was not available for 1 patient. At least 1 disease recurrence resulting in additional surgical procedure was noted in 42 cases (87.5%). Two disease recurrences were noted in 17 cases (35.4%), and 4 patients experienced 3 disease recurrences (6.3%). No patients in this cohort experienced more than 3 disease recurrences. The median time to first disease recurrence, defined as the time from first surgery to second surgery, was 40.5 months (range: 0.2–220.0 months). The median time to second disease recurrence was 37.9 months (range: 3.3–133.2 months), and the median time to third disease recurrence was 10.8 months (range: 2.4–14.9 months).

Detailed clinical notes and autopsy reports (when available) were used to classify patient death into 1 of 4 categories (Table 1): direct tumor involvement (i.e. death due to tumor growth and/or progression; n = 31), semidirect tumor involvement (death due unintended surgery/treatment-related complication or cardiopulmonary compromise in patients with brainstem/diencephalic tumors; n = 6), unrelated to tumor (death due to factors not related to the patient’s tumor; n = 5), or unknown (when detailed clinical history surrounding patient death was unavailable; n = 6).

Targeted next-generation sequencing (OncoPanel) was attempted in all 48 cases, with successful results obtained in 33 (68.8%). In 11 cases, there was insufficient tissue for sequencing. Sequencing failed in the other 4 cases. The sequencing results for the 33 successful tumors are summarized in Figures 1 and 2. KIAA1549-BRAF fusion was the most common oncogenic alteration, observed in 14 cases. A variety of other oncogenic driver events were present in the remaining cases, including BRAF V600E (n = 3), NF1 loss-of-function mutation (n = 2), FGFR1-TACC1 fusion (n = 2), IDH1 R132H (n = 1), FGFR1 N546K (n = 1), and FGFR1 internal tandem duplication (ITD, n = 1; Fig. 2A). Interestingly, there were several cases with molecular features characteristic of high-grade glioma, including activating point mutations in EGFR (n = 3), FGFR3 focal gain/amplification (n = 1), PDGFRA amplification (n = 1), mismatch repair deficiency (MMR-D, n = 1), and focal gain/amplification in FGFR1 and MDM2 in 1 case. There were no tumors with histone alterations in our cohort. Finally, 1 case harbored both KIAA1549-BRAF fusion and BRAF V600E. A definitive oncogenic driver was not found in 2 cases.

FIGURE 1.

Molecular driver alterations in pediatric LGG by intracranial site. (A) A total of 33 tumors were successfully sequenced with 1 additional case from a patient with NF1 diagnosis clinically (n = 34 total cases). KIAA1549-BRAF gene fusion representing the most common single driver alteration overall. A variety of driver alterations were observed from tumors located in the cerebral hemispheres (B), diencephalon (C), and brainstem/spinal cord (D). Only a single tumor from the cerebellum was successfully sequenced, which did not reveal a driver alteration (E).

FIGURE 2.

Molecular landscape of pediatric LGG from expired patients. Driver alterations were identified in 31 of 33 tumors successfully profiled with targeted next-generation sequencing panel (OncoPanel). Additional co-occurring molecular alterations were also identified in 14 of 33 tumors. Histologic diagnosis, tumor location, and cause of death are also included.

Not surprisingly, KIAA1549-BRAF fusion was more associated with tumors located in the midline (8/17 diencephalon tumors, 47.1%; 2/5 brainstem/spinal cord, 40%). Otherwise, there was remarkable heterogeneity of oncogenic drivers in these brain regions, including 8 different drivers in tumors from the cerebral hemispheres, 7 different drivers in diencephalic tumors, and 5 different drivers in brainstem/spinal cord tumors (Fig. 1B–D). Of the 2 cerebellar tumors in this cohort, successful sequencing results were obtained for 1 case, which demonstrated no identifiable driver alterations (Fig. 1E).

A number of cases (14/33, 42.1%) exhibited co-occurring genetic events commonly associated with negative prognosis (Fig. 2). CDKN2A/B homozygous deletion was observed in 4 instances (once each with KIAA1549-BRAF, BRAF V600E, FGFR1-TACC1, and IDH1 R132H). CDKN2C loss was observed in 4 cases (twice with KIAA1549-BRAF and once each with BRAF V600E and FGFR1-TACC1). TP53 and ATRX mutations were observed in 6 and 2 cases, respectively, with only 1 case (IDH-mutant diffuse astrocytoma) showing mutations in both TP53 and ATRX. The majority of patients with tumors with MAPK pathway alteration and coexisting TP53 or CDKN2A/2C alterations died due to tumor progression (Fig. 2). Several significant gene amplifications were observed, including EGFR (n = 2), MDM2 (n = 2), and TERT (n = 2). Two cases also showed polysomy 7. Combined polysomy 7 and monosomy 10, along with CDKN2A/B homozygous deletion and CDKN2C loss, was observed in 1 case also harboring an FGFR1-TACC1 fusion, but, overall, copy number changes were not a prominent feature in this cohort. Several cases displayed only a single genetic alteration without additional events, including 10 cases with KIAA1549-BRAF fusion, 2 with EGFR activating mutations, and single instances with BRAF V600E, NF1 loss-of-function mutation, FGFR1-TACC1, and FGFR1 N546K. The cause of death in these patients was due to direct tumor involvement (death due to tumor growth and/or progression) in 12 of these 16 cases (75%). There were no cases with 1p/19q codeletion, H3.1/3 mutation, or MYB/MYBL1 rearrangements.

Several PLGG cases harbored molecular alterations generally ascribed to high-grade/adult-type glioma, including 3 cases with activating EGFR mutations, 1 case with focal FGFR3 amplification, 1 case with focal PDGFRA amplification, and 1 case with MMR-D; 1 case had a peculiar molecular profile, with both KIAA1549-BRAF fusion and BRAF V600E mutation. Select histologic images from these cases are demonstrated in Figure 3. The case with FGFR3 gain occurred in the cerebral hemisphere of 12-year-old female and demonstrated palisading of tumor cells reminiscent of angiocentric glioma with areas of vacuolated microcysts (Fig. 3A). The MMR-D case was from a 1-year-old with a frontal lobe tumor that displayed minimal hypercellularity and glial atypia with very low Ki67 proliferation index (Fig. 3B). The case with combined KIAA1549-BRAF fusion and BRAF V600E occurred in the optic pathway of an 8-year-old and demonstrated peripheral palisading of nuclei around eosinophilic cytoplasmic processes (mimicking pseudopalisading necrosis at low power) and perivascular distribution (Fig. 3C). Finally, one example of pathogenic EGFR mutant (D770_N771insNPH) glioma in the thalamus of a 4-year-old showed moderate cellularity with mild atypia on initial biopsy; however, at autopsy, tumor progression was observed with worsened nuclear atypia and pronounced mitotic activity that was not present on initial surgery (Fig. 3D).

FIGURE 3.

Pediatric low-grade glioma (LGG) with molecular features of high-grade glioma. Representative hematoxylin-eosin (H&E)-stained sections from histologically LGGs. (A) LGG with FGFR3 amplification demonstrates palisading of tumor cells reminiscent of angiocentric glioma (top panel, ×10) with areas of vacuolated microcysts (lower panel, ×20). (B) LGG with mismatch repair deficiency shows minimally hypercellular brain tissue with mildly atypical glia cells (top panel, ×20) and low Ki67 proliferation index (1%) (lower panel, ×20). (C) Glioma with both KIAA1549-BRAF and BRAF V600E alterations shows peripheral palisading of nuclei around eosinophilic cytoplasmic processes mimicking pseudopallisading necrosis at low power (top panel, ×10) and arranged around blood vessels (lower panel, ×20). (D) EGFR-mutated LGG at initial resection demonstrated moderate cellularity with mild atypia (top panel, ×20); however, at autopsy, tumor progression was observed with worsened nuclear atypia and pronounced mitotic activity (lower panel, ×20).

Swimmer plots were created to assess for differences in overall survival based on oncogenic driver mutation (Fig. 4). When the entire cohort is organized by descending overall survival (Fig. 4A) or grouped based on oncogenic driver mutation (Fig. 4B), there were no trends or patterns observed. Indeed, there is striking heterogeneity in overall survival between different molecular groups. However, the occurrence of disease recurrence/progression that led to additional surgical procedures is notable and was observed in almost every molecular subgroup.

FIGURE 4.

Swimmer plots of pediatric LGG from expired patients. Individual horizontal bars represent individual patients and are color-coded based on oncogenic driver alteration. Black/pink lines within the patient bars represent additional surgery due to tumor recurrence (black) or tumor progression (pink). A black asterisk (*) at the end of a bar indicates autopsy was performed and pink asterisk indicates histologic progression was observed at autopsy. (A) Patients arranged in order of decreasing overall survival. (B) Patients grouped by molecular alteration and then in order of decreasing median overall survival within each group.

Finally, overall survival was assessed by Kaplan-Meier analysis to assess the effects of tumor location, histologic diagnosis, and treatment history (Fig. 5). There were no statistically significant differences in survival between tumors in supratentorial locations versus infratentorial locations. Although there was a trend towards worse overall survival in patients with diencephalic tumors (p = 0.0157), this was not statistically significant after Bonferroni correction for multiple comparisons (Fig. 5A). There were no differences in overall survival based on histologic diagnosis (Fig. 5B). When overall survival was compared between patients with radiation (both with and without chemotherapy) to those without radiation therapy, there was no difference in overall survival. However, patients who were treated with only chemotherapy had worse overall survival compared with those treated with both chemotherapy and radiation therapy, which was statistically significant after Bonferroni correction for multiple comparisons (p < 0.0001). No other differences in overall survival based on treatment history were observed (Fig. 5C).

FIGURE 5.

Kaplan-Meier analysis to assess overall survival in expired patients with pediatric LGG. (A) Tumor location, (B) histologic diagnosis, and (C) treatment history were compared with assess differences in overall survival. The only statistically significant difference was worse overall survival in patients who received only chemotherapy compared with those who received both chemotherapy and radiation therapy (p < 0.0001). Bonferroni correction was applied for multiple comparisons.

DISCUSSION

PLGGs are the most common brain tumors in childhood and are generally associated with excellent long-term survival. A small subset of patients, however, experience poor outcome. In large population-based studies, mortality rates in patients with PLGG range from 3.1% to 10.5% and are strongly associated with administration of radiation but disease recurrence/progression has also been reported as a significant contributor to death (2, 3, 14). The molecular characteristics of PLGG that led to patients’ death are insufficiently described: although large series of pediatric gliomas and their outcomes exist in the literature, there are no studies that specifically analyze the molecular profile of PLGGs that led to patients’ death (6, 15). Furthermore, some of the existing literature suggests that particular tumorigenic drivers, such as BRAF V600E mutation, particularly when associated with CDKN2A homozygous deletion, are associated with worse outcome (16, 17). Our study demonstrates molecular heterogeneity amongst these tumors. The majority of tumors in our cohort harbor alterations affecting the RAS/MAPK pathway, in keeping with other large-scale molecular studies of PLGG (5–8, 10, 11). Prior studies have demonstrated an association with worse outcome in PLGG with BRAF V600E, particularly for those that concurrently harbor CDKN2A homozygous deletion (16, 17). The most common alteration observed in our cohort was KIAA1549-BRAF fusion (n = 14/33; 42.1%), whereas BRAF V600E mutation was only observed in 4 tumors. Similarly, CDKN2A homozygous deletion was only observed in 1 BRAF V600E mutated tumor. Our results suggest that other molecular subgroups of PLGG can also have poor outcome and should be monitored closely for disease relapse and/or progression.

Additional alterations not associated with RAS/MAPK pathway upregulation were also present in our cohort, including mutations in IDH1, EGFR, PDGFRA, and MMR-D. Despite uniform low-grade histology, these molecular alterations are more characteristic of high-grade glioma (18–21). Eventual progression to high-grade histology was observed in 3 out of 6 cases, including 2 cases with EGFR mutation and 1 case with PDGFRA amplification. These data suggest that routine molecular profiling should be performed in pediatric gliomas in order to identify molecular alterations that confer additional risk for poor outcome, as histology does not always provide a perfect correlation. Particularly in small biopsies, in addition to a correlation with imaging studies and clinical presentation, determining the tumorigenic driver and, if possible, co-existing genetic events with prognostic implications (e.g. CDKN2A homozygous deletion, TP53 mutations), may guide clinical management. A combination of tests that investigate for presence of fusions and mutations in genes associated with PLGG (e.g. MAPK pathway genes), as well as ruling out drivers of high-grade gliomas, such as histone alterations, would be ideal, particularly in light of the recent cIMPACT-NOW recommendations (22, 23). Since there is not a national standard for molecular modalities, the tests employed vary widely, with most centers utilizing a combination of immunohistochemical surrogates for specific alterations (e.g. BRAF V600E, IDH1 R132H, H3K27M immunostains), fluorescence in situ hybridization (e.g. BRAF duplication, MYB alterations, NTRK genes alterations), digital droplet polymerase chain reaction (ddPCR) for single nucleotide variants (e.g. BRAF V600E, HIST1H3B, H3F3A), and, sometimes, larger DNA and/or RNA-based panels, such a targeted exome sequencing and fusion panels.

PLGG are often associated with single-molecular events, particularly rearrangement-driven PLGG (5), and do not generally harbor additional molecular events that are common in adult gliomas, such as mutations in TP53 and ATRX (24, 25). In our cohort, 5 of 16 PLGG driven by rearrangements possessed additional molecular alterations at initial diagnosis (involving CDKN2A, CDKN2C, MDM2, BRAF, EGFR, ATRX, and TP53), including 4 KIAA1549-BRAF altered tumors and 1 FGFR1-TACC1 tumor (Fig. 2). These additional alterations may portend worse behavior by these tumors. Similarly, CDKN2A/B homozygous deletion is commonly associated with worse outcome in pediatric glioma patients (26, 27), particularly those with BRAF V600E mutation (6). Overall, 14 of 33 tumors in this study contained clinically significant co-occurring molecular events. These additional molecular events could prove to have significant impact on tumor grade and treatment decisions in pediatric patients, particularly those with multiple recurrences and/or progression.

The most common cause of death overall in our study was due to primary effects of the brain tumor, including tumor growth and/or progression. This is in keeping with the results of Renzi et al. (14) who specifically examined the cause of death in PLGG (28). It is unclear what factors influence progression in PLGG as this phenomenon is overall rare (29, 30) and the majority of PLGG remains quiescent or even spontaneously regress as patients age (31, 32). Thus, identification of factors that influence disease recurrence and progression in these patients is critically important. Tumors from this cohort were most frequently located in deep seated locations in the brain, such as the diencephalon and brainstem. Complete surgical resection is difficult to achieve for tumors in these locations, making long-term management difficult. Additional molecular profiling should be considered in patients that do not respond to conventional treatment, especially if subsequent surgeries are warranted and additional tissue is rendered. For those patients who did not receive treatment, only 2 died from direct tumor involvement and 2 from semi-direct tumor involvement. Death was unrelated to tumor in 3 patients and unknown for 3 patients.

Finally, examination of tumors should not end with patient death. Indeed, important information can be gathered from autopsy-derived tissue, including histologic changes from initial diagnosis, molecular evolution over time, and response to treatment(s) (33). Autopsy was performed in 19 patients in this study, and histologic progression was confirmed in 5 of these cases (Fig. 4). Ongoing studies are currently underway to determine how the molecular characteristics of PLGG change over time and in response to different treatments. Supporting these efforts, the majority of parents/caregivers of patients with PLGG are willing to consent to research autopsy as a desire to benefit others and provide additional significance to their child’s life (34). Autopsy studies have provided important insight to pediatric gliomas in which surgical tissue is not readily accessible, such as rare entities and tumors in surgically inaccessible locations (35, 36).

In summary, we provide molecular characterization of PLGG from a large cohort of expired patients from 1 institution. We demonstrate unique heterogeneity in tumor location, histologic diagnosis, molecular driver alteration, and co-occurring molecular alterations. Importantly, occasional tumors with low-grade histology harbor molecular alterations more characteristic of high-grade tumor behavior. Many of these patients experienced multiple disease recurrences and tumor progression. These data provide important information about possible underlying genetics that could influence poor outcome in these patients and argue for broad molecular testing in patients with PLGG.

The authors have no duality or conflicts of interest to declare.

Author notes

Keith L. Ligon and Sanda Alexandrescu contributed equally to this work.