High-Grade Astrocytoma with Piloid Features: A Dual Institutional Review of Imaging Findings of a Novel Entity

DISCUSSION

HGAP is a new entity within circumscribed astrocytic gliomas, often resembling glioblastoma rather than PA with survival comparable with that of grade IV IDH-mutant astrocytoma.⁵ DNA methylation profiling is performed to exclude HGAP, especially in atypical cerebellar gliomas (with ATRX loss or CDKN2A/B homozygous deletion) and HGAP-resembling tumors (like high-grade gliomas in NF1).⁶ Reinhardt et al⁸ found that molecular analysis reclassified approximately one-third of initially diagnosed glioblastoma cases as HGAP.¹² In a study by Lucas et al,¹³ advanced molecular diagnosis reclassified 14 NF1-associated gliomas as HGAP, further supporting the NF1 and HGAP link. Although the exact HGAP incidence in NF1 remains uncertain, a confirmed NF1 diagnosis should alert clinicians to the potential HGAP risk.

The NF1 gene is a tumor-suppressor gene mutated in individuals with NF1. NF1 encodes the neurofibromin 1 protein, which negatively regulates the Ras/MAPK pathways via Ras inhibition. NF1 mutations are prevalent in treatment-resistant gliomas and serve as supportive diagnostic biomarkers for HGAP and rosette-forming glioneuronal tumors.^14,15 Patients with NF1 are at a higher risk of developing low-grade gliomas. NF1 also predisposes to high-grade gliomas, with their prevalence being 10–50 times higher than in the general population. Romo et al¹⁶ reported a nearly 300-fold higher prevalence of non-optic pathway gliomas in patients with NF1 (3.2%) compared with the general population (0.01%), with a median survival of 24 months. Nonoptic gliomas in adults with NF1 often have an aggressive clinical course, further underscoring the need to understand the pathobiology of NF1-associated gliomas better.¹⁶ NF1-related gliomas resembling HGAP had worse outcomes than the NF1-associated PA but fared better than sporadic IDH wild-type glioblastoma.¹³ A substantial GB-HGAP overlap necessitates early HGAP detection because these HGAP cases often have a better prognosis and access to targeted therapies.¹³ NF1-mutated tumors may respond to mammalian target of rapamycin (mTOR) and MAPK inhibitors, such as FDA-approved everolimus, temsirolimus, and emerging complex inhibitors.¹⁵ This possibility raises the question of whether brain/spine MR imaging screening is advisable for those with NF1, but the lack of large-scale studies on HGAP prevalence in NF1 prevents a conclusive answer.

One-half of HGAP lesions in our cohort involved the posterior fossa, followed by equal (25%) spinal cord and midline diencephalic/thalamic region involvement. Reinhardt et al⁴ and Bender et al⁷ identified the posterior fossa as the primary location in 74% and 66%, respectively. Spinal cord involvement varied between 7% (5/83)⁴ and 33% (2/6).⁷ We found 5 HGAPs (P1, P3, P5, P6, P8) in the midline, aligning with the observations of Romo et al¹⁶ indicating increased midline involvement in NF1-associated tumors. Among these 5 cases, GTR was performed in 2 cases (P1, P3), with one (P1) achieving a PFS of 3 months, matching the findings of Romo et al of limited GTR success in NF1-associated tumors. While tumor-imaging descriptions for HGAP in the literature are sparse, they generally exhibit variable T1-weighted hypointensity, T2-weighted hyperintensity, and heterogeneous enhancement. Diffusion restriction was generally absent. In our cohort, HGAP lesions closely imitated glioblastoma with variable heterogeneous enhancement and increased perfusion metrics. Intramedullary HGAPs were expansile with a dorsal predilection, and one (P3) demonstrated FDG-avidity. This imaging profile is similar to the findings of Bender et al⁷ in their 6-case series, including radiotracer avidity on O-(2-[¹⁸F] fluoroethyl)-l-tyrosine PET/CT in 2 cases. Focal areas of signal intensity are common in patients with NF1 and may sometimes be challenging to differentiate from tumors. Perfusion-weighted MR imaging offers the potential to detect tumor-related neoangiogenesis, aiding in the differentiation to determine whether focal areas of signal intensity represent a low-grade tumor.¹⁷ One of our patients with HGAP (P2) demonstrated ependymal and intraspinal leptomeningeal spread, indicating aggressiveness and a poor prognosis. Leptomeningeal or dural spread of HGAP is rarely reported, except for a recent study that identified leptomeningeal spread in a confirmed case of HGAP with a 7-year interval to metastasis and at the time of initial diagnosis in a possible HGAP. Adult high-grade astrocytic tumor types, including HGAP, are capable of leptomeningeal or dural spread.⁹

MGMT promoter methylation status is crucial for therapy and prognosis. Unmethylated status predicts a poor response to alkylating chemotherapy and worse prognosis.^5,18 In our cohort, all except 1 patient received alkylating chemotherapy. We detected MGMT promoter methylation in 62.5% (5/8) of patients, higher than that reported by Bender et al⁷ (33%, 2/6) and Reinhardt et al⁴ (45%, 38/83). Four of our 5 cases of NF1 gene mutation also had co-occurring ATRX and CDKN2A/B mutations. Patients with IDH wild-type astrocytic gliomas with CDKN2A/B deletions and ATRX mutations generally have a better clinical outcome than patients with IDH wild-type glioblastoma, but the significance in NF1 is unclear.⁴ Both CDKN2A/B and ATRX mutations were common as reported by Lucas et al,¹³ who also noted MAP kinase pathway mutations (NF1, FGFR-1, BRAF genes). ATRX mutations were found in 87.5% (7/8) of our cases, slightly higher than previously reported in 45%–60% of cases.^4,7 Additionally, CDKN2A/B mutation was identified in 87.5% of our cases (7/8), previously reported in 80%–100% of cases.^4,7 ATRX, a chromatin remodeling protein, preserves genomic stability and often co-occurs with IDH mutations, potentially improving survival in low-grade gliomas, especially when lacking 1p/19q codeletions.¹⁹ While ATRX alterations are characteristic of IDH-mutant astrocytoma, they are also frequently found in the IDH wild-type HGAP in conjunction with homozygous deletion of CDKN2A/B. This loss of CDKN2A/B generally indicates a higher grade and poorer prognosis, especially compared with PA.^4,5 HGAP is diagnosed solely by methylation profiling; however, the loss of ATRX on immunohistochemistry is helpful, especially with concurrent homozygous loss of CDKN2A/B to suggest this entity. In cases without ATRX mutation, TERT can be mutated (these are mutually exclusive events).⁶

In 2 cases (P7 and P8) FGFR1 fusion was detected, which transforms primary astrocytes into highly proliferating midline gliomas.^20,21 TP53 mutations were identified in 50% (3/6) of the patients in our study, a rate higher than the 5.4% reported in cases of HGAP by Cimino et al.⁶ TP53, a tumor-suppressor gene, is involved in DNA damage response, cell cycle arrest, and apoptosis, with its mutations enhancing cell survival and chemoresistance.^5,22 BRAF V600E mutations were evaluated in 50% of cases, and all findings were negative. BRAF fusions are typically linked to World Health Organization grade 1 PA pediatric posterior fossa cases, showing a favorable outcome compared with supratentorial PAs in adults.⁴ One patient with HGAP (P4) exhibited a rare genetic profile of NTRK2 fusion, loss of P16 expression, a SETD2 mutation, lack of CDKN2A/B deletion, and MGMT promoter methylation. Considering these characteristics, this case was contemplated for potential inclusion in a future trial involving NTRK-targeted therapy (entrectinib) in the event of disease progression. The significance of these mutations remains uncertain; however, they merit further investigation in future studies. SETD2 mutations have been reported in high-grade hemispheric gliomas in older children²³ and adults with cerebellar glioblastomas.²⁴ Loss-of-function mutations in SETD2 result in a deficiency of trimethylated histone H3K36, which appears specific to high-grade tumors. The nonsense mutation in SETD2 in addition to a KIAA1549-BRAF fusion, contributes to the aggressive clinical course of this tumor.^23,24

HGAP can be challenging to distinguish from PA, DMGs, and glioblastoma on imaging. PAs, typically found in children, are low-grade circumscribed gliomas with good prognosis (10-year survival rate >95%). These are strongly associated with NF1 and BRAF-KIAA1549 fusion/duplication. Most sporadic PAs arise from the cerebellum while affecting the optic pathway in patients with NF1. However, in adults, PAs, mostly supratentorial, exhibit a more aggressive clinical course, with a 50% 5-year survival rate and rare BRAF V600E alterations.²⁵ Radiologically, they show variable appearances, ranging from large cystic lesions with enhancing nodules to solid lesions with intense enhancement (∼95%), calcification (20%), and hemorrhage.²⁶ The transformation of PA into a high-grade glioma is uncommon, especially in cases involving BRAF fusion, while about 20% of HGAPs with BRAF fusion suggest a possible link to earlier PAs or shared biologic traits.^8,20 Diffuse midline glioma (H3K27-altered) often occurs in pediatric patients without significant sex variation. It is characterized as an expansile, diffusely infiltrative lesion primarily in the thalamus, brainstem, and spinal cord, presenting as slightly T1 hypointense and T2 hyperintense on MR imaging with lesser enhancement, higher diffusion restriction, and perfusion. H3K27-altered DMGs have a poor prognosis, despite their histopathologic grade, mandating H3K27M detection and genetic analysis for definitive identification.^27,28 HGAP, prevalent in adults, associated with NF1, often displays high T2 signal and variable peripheral enhancement, and lacks diffusion restriction.⁷ Both tumors share a dismal prognosis and are managed with surgery, radiation, and chemotherapy.^7,29 Both HGAP and glioblastoma, IDH wild-type high-grade CNS tumors, share similar imaging features and poor prognosis, requiring DNA methylation profiling for accurate differentiation. Glioblastoma, the most frequent primary brain tumor, typically manifests as a heterogeneous mass in the cerebral hemisphere, displaying irregular peripheral enhancement, central necrosis, diffusion restriction, extensive peritumoral edema, and glioblastoma histologic criteria (microvascular proliferation and necrosis).⁷

Our study did not reveal significant imaging differences between HGAPs with and without NF1 mutations, except that spinal HGAPs were linked to NF1. Patient outcomes may be influenced by a complex interplay of factors, including NF1 mutations, genetic modifiers, extent of tumor involvement, resectability, and treatment response. Neither our current study nor existing literature indicate any notable prognostic differences between HGAPs with and without NF1 mutations; however, individual cases may differ due to unique genetic and molecular traits. Our study is constrained by a small patient sample, retrospective design, and varying next-generation sequencing data availability across institutions with differing gene region coverage. Nevertheless, this imaging-focused review adds valuable insight into the clinical, genetic diversity, and imaging characteristics of HGAP, particularly in NF1-associated HGAPs.^4,8 Targeted therapy may be pivotal due to frequent MAP kinase pathway gene alterations.⁶