An 8‐Year‐Old Girl with Posterior Fossa Mass

Clinical History and Imaging

In 2004 as an 8‐year‐old girl presented with headaches, nausea, and vomiting. MRI showed a marginally enhancing mass, completely filling the fourth ventricle. Following gross total resection and diagnosis the patient was treated with radiation and chemotherapy. A year later, there was enhancement in the right cerebellum at the bottom of the fourth ventricle, compatible with progredient residual tumor. This was resected and evaluated. Further cycles of chemotherapy followed, and the patient recovered well. However, 13 years later, the patient had headaches and eye movement difficulties. MRI showed a new enhancing mass in the posterior fossa. Again, the tumor was resected and referred to histological examination. Shortly thereafter, a myelon‐compressing intradural T8‐T9 mass was detected and irradiation of the posterior fossa and the thoracic spine was started, when the patient presented with an acute transverse spinal cord syndrome. Resection of this spinal tumor was subsequently performed on an emergency basis. However, based on the rapidly progressive, especially meningeal tumor spread observed on MRI, supportive and palliative care was finally initiated.

Microscopic Pathology and Molecular Genetics Study

The tissue from the first operation is shown in Figure 1A. Apoptotic and mitotic cells were seen, but necrosis and vascular proliferation were not. The tumor showed focal GFAP staining, but diffuse staining for synaptophysin. The Ki‐67 proliferation was 90 %. The second tumor is shown in Figure 1B. Mitoses and necrosis were absent. GFAP and S‐100 were positive but synaptophysin was negative. The Ki‐67 proliferation was 3%. The third tumor is shown in Figure 1C. Geographic necrosis and endothelial proliferation were present, but there are no perivascular rosettes. GFAP staining was moderate, but both Olig2 and MAP2 stained strongly. Many cells showed pronounced p53 nuclear staining. Mutated H3 (K27M) was not detected, but trimethylation of Histone H3 at position K27 was lost. Ki67 proliferation was 60%. Sections of the fourth tumor resected from the spine is shown in Figure 1D. A fraction of the tumor cells strongly expressed GFAP and MAP2, but were negative for Olig2. Trimethylation of Histone H3 at position K27 was lost. Ki67 proliferation was 20 %. All four tumors were evaluated using Global Methylation analyses with r450K or 850K bead chip arrays. The first (1), second (2), and fourth (4) samples clustered into a diagnostic category (Figure 1E). In contrast, tissue from the third (3) operation was distinctly different (Figure 1E). Copy number variation profiles (CNV) of the first tumor had only a few chromosomal aberrations but that of the second resection was nearly balanced (Figure 1Fa, 1Fb). However, the third and fourth samples exhibited a complex pattern of chromosomal aberrations (Figure 1Fc, 1fd). Mutations of IDH1/2, BRAF, H3 or the TERT promoter sequence were absent in the last two resections. What are your diagnoses?

Figure 1.

 

Diagnoses

1. First, second, and fourth resection: Posterior fossa Ependymoma (WHO grade III; Methylation subgroup A)

2. Malignant glioma, corresponding to glioblastoma (WHO grade IV), most likely radiation induced

Discussion

Malignant pediatric brain tumors represent a very heterogeneous group with a great variability in the age of onset, localization, disease course and long‐term outcome as well as radiological, histological and molecular features. This makes clinical treatment complicated and still leaves patients with high mortality rates. Previously, these tumors have been classified according to distinct histomorphological features and localization. However, considerable effort has recently been made towards the molecular characterization of brain tumors, revealing that morphologically similar appearing tumors may belong to entirely different molecular entities with different biological behavior and clinical outcome (1, 2). Thus, definite diagnosis and classification of pediatric brain tumors based on histology alone is in many cases not possible and requires further molecular analyses.

An initial diagnosis of medulloblastoma seemed obvious given the embryonal features (Figure 1A). However, one year later, another resection specimen from the same localization was referred for histological examination, which exhibited none of the previously observed embryonal features, but rather formations of perivascular pseudo‐rosettes (Figure 1B) so that ependymoma (WHO grade II) was diagnosed.

Both tumors underwent 450K‐methylation analysis that identified posterior fossa ependymomas, group A (EPN_PFA) (0.67 (1) and 0.99 (2), respectively), which have aggressive behavior and bad outcomes. The CNVs of both tumors had identical changes on chromosome 22, indicate that they represent both subclones of the same tumor (Figure 1F). Therefore, the 450K‐methylation profiling strongly suggests that the initial manifestation represented like the second one an anaplastic ependymoma and not a medulloblastoma, thus prompting to consider methylation profiling for all pediatric small blue round cell tumors.

Thirteen years later, the patient developed an infratentorial GBM and a few months later she had a spinal tumor resected. While for the spinal tumor highest classifier scores were again reached for the group of EPN_PFA (0.77), methylation analysis of the infratentorial mass revealed only low classifier scores (0.5) for the methylation class family glioblastoma, IDH wildtype. The CNV profile of the infratentorial tumor was entirely different from the three other tumors (Figure 1F). Interestingly, in a t‐distributed stochastic neighbor embedding (t‐SNE) representation (3), this tumor was arranged with the group of H3 K27M mutant diffuse midline gliomas (DMG_K27) (Figure 1E), although this association was not found by the DNA methylation classifier and no mutations of H3F3A, HIST1H3B, or HIST1H3C, but loss of trimethylation of Histone H3 at position K27 were observed. However, as t‐SNE primarily indicates a high level of similarity between the tumor described here and DMG_K27M and as DNA methylation classifier and t‐SNE are fundamentally different approaches, this observation is not necessarily conflicting. As expected, the other three tumors clustered within the group of EPN_PFA.

Given the history of radiation therapy, a causal relationship between irradiation and the secondary tumor seems likely and the classic criteria for radiation induced malignancies (i.e. irradiated region, sufficient latency, distinct histology from the original tumor, and no genetic predisposition for tumor development) were met (4). Radiation induced gliomas (RIG) are a relatively rare long‐term complication of radiation therapy that occur with an average latency of 5‐10 years after irradiation (5, 6). Although they resemble their sporadically arising counterparts histologically, not much is known about their molecular characteristics. One recent study analyzed four RIGs and found that at least in these cases the classically detected genetic aberrations, such as IDH, H3, BRAF, or TERT promotor mutations were not present, indicating that different genetic alterations might play a role in the development of RIGs (7).

Still, this case emphasizes the relevance and importance of molecular analyses of pediatric brain tumors as (1) histological assessment only can be misleading, particularly in small blue round cell tumors and (2) different tumor entities in the same patient can be clearly distinguished.