Ependymomas in infancy: underlying genetic alterations, histological features, and clinical outcome

Ependymomas in 28 children under the age of 18 months were retrieved from the archives of the Institute of Neuropathology, University of Bonn Medical Center (Bonn, Germany), and of the DGNN German Brain Tumor Reference Center (Bonn, Germany). All patients were diagnosed with anaplastic ependymoma (WHO grade III) in a period from 2002 to 2011 and enrolled into the prospective non-randomized multicenter HIT ependymoma study (ClinicalTrials.​gov NCT00303810). This included risk-adapted HIT-SKK chemotherapy (without intraventricular methotrexate) and irradiation. In infant patients, the aim was to postpone irradiation until the age of 18 months if possible. In case of residual tumor, patients were evaluated for potential second-look surgery after each therapy element. All examinations were carried out on the basis and according to the legal requirements of the revised Declaration of Helsinki of the World Medical Association in 1983 and according to the guidelines of the institutional review boards. Informed consent was given at study inclusion by the parents.

All tumors were centrally reviewed at the DGNN German Brain Tumor Reference Center (Bonn, Germany) and reclassified according to the revised WHO classification of tumors of the CNS 2016 [5]. Mitotic activity was assessed in 4μm-thick HE-stained FFPE slides by counting mitotic figures in 10 high power fields (HPF; area, 2.38 mm²). Presence of more than 10 mitoses/10 HPF was defined as high mitotic activity. Further histological features including necrosis, vascular proliferation, and clear cell morphology were assessed. Immunohistochemical staining—performed on an immunostaining system (BenchMark XT, Ventana-Roche, Mannheim, Germany)—included Ki67 (MAb MIB-1, Dako, Glostrup, Denmark), phosphohistone-H3 (Biocare, Concord, USA), p16 protein (MAb E6H4, Ventana-Roche, Darmstadt, Germany), p65-RelA (rabbit antibody D14E12, Cell Signaling, Danvers, U.S.A.), and trimethylated histone3-Lys27 (H3-K27me³; rabbit MAb C36B11, Cell Signaling). For H3-K37me³, positive nuclei of endothelial cells served as internal control.

On hematoxylin and eosin (HE)-stained FFPE sections, representative areas with at least 80% tumor cell content were identified for microdissection and DNA extraction from serial sections using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. In order to identify genome-wide copy number alterations and allelic disbalances, we used molecular inversion probe (MIP) assays (OncoScan FFPE express 330 K arrays, versions 2 and 3; Affymetrix, Santa Clara, CA, US) as previously described [21, 22]. Raw data were further analyzed with the Nexus Copy Number software, version 7.5 (BioDiscovery, El Segundo, CA, USA). DNA from FFPE cerebellar tissue served as normal control. The SNP-FASST-2 segmentation algorithm was used for data processing. All cases were reviewed individually, and if necessary, diploidy correction was performed manually.

FFPE or fresh frozen tissue (if available) of supratentorial ependymomas was further tested for presence of C11orf95-RELA and YAP1-MAMLD1 fusions. RNA was extracted, followed by RT-PCR and sequencing, described previously [15, 16, 18].

Seven patients of 28 (25%) had an ependymoma in a supratentorial location. These occurred in five girls and two boys. Subtotal resection (STR) was performed in five (71%), and in three (43%), a relapse occurred. However, all patients were alive after a median follow-up of 7.96 years (range 0.28–12 years). High mitotic activity, necrosis, and vascular proliferation each were present in all but one case. All cases had at least one of these criteria. All tumors were classified as anaplastic ependymoma (WHO grade III). Two cases showed clear cell morphology, and balanced genomes were found in four cases (57%) compared to structural (two cases) and numerical genomic group (one case). Chromosome 1q gain was not present at all, while one case showed chromosome 22 loss. Six cases could be studied for fusion transcripts; for the remaining one, material for further analysis was not available. RELA fusions were detected in four cases, which showed nuclear accumulation of p65-RelA protein as indicator of activation of NFkB signaling. Two of the RELA-fused cases displayed clear cell morphology, and three of the four relapsed. YAP1-MAMLD1 fusions were found in two other cases, not carrying RELA fusions. Concerning the latter two patients, none relapsed, although in one, gross total resection (GTR) could not be achieved. One of the two cases harboring a YAP1-MAMLD1 fusion showed monosomy of chromosome 22. CDKN2A loss/LOH or p16 protein loss was not identified in any of these 7 cases. For further information, see Table 1 and Fig. 1.

Twenty-one of 28 patients (75%) developed tumors in the posterior fossa. The male to female ratio was even (11 vs. 10). All cases were classified as anaplastic ependymoma (WHO grade III). In terms of histological criteria associated with anaplasia, 19 (91%) revealed vascular proliferation, 18 (86%) necrosis, and 15 (71%) high mitotic activity, while neither true ependymal rosettes nor clear cell morphology were encountered. Most strikingly, all cases showed a balanced genomic profile with consequently no gain of chromosome 1q and loss of the trimethylated lysine 27 of histone 3, typical of pediatric “group A” PF ependymomas [23, 24] (Table 1, Fig. 1).

Nine patients (43%) had postoperative residual tumor, and 15 (71%) suffered a relapse; none showed initial metastasis. After a median follow-up time of 5.44 years, 5-year EFS and OS were 43% and 71%, respectively.

Kaplan-Meier analysis revealed no statistically significant difference between ST and PF ependymoma regarding EFS and OS, respectively. Regarding OS, there was nevertheless a clear trend towards a favorable prognosis of ST ependymoma of infancy, since none of the patients died (Fig. 2).

Strikingly, no patient with a supratentorial ependymoma below the age of 18 months at diagnosis died during the period of follow-up although all tumors in this age group were graded as anaplastic (WHO grade III). RELA fusions—representing the most frequent alteration in pediatric supratentorial ependymomas—were present in four cases (57%), which is similar in frequency as previously reported [3, 15, 17]. All three genomic groups are represented in tumors of this location. However, in line with Upadhyaya et al., we cannot confirm the poor prognosis reported by Pajtler et al. in a retrospective case collection [3, 11]. Although some patients relapsed, they were successfully treated, either with second resection, chemotherapy, or radiotherapy. One explanation for a favorable prognosis of this specific subgroup may be the absence of CDKN2A loss/LOH in RELA-positive tumors. CDKN2A loss has been previously reported in ST ependymoma and is known to be an indicator of adverse prognosis in RELA-fused ependymomas [25] and other brain tumors such as IDH-mutated anaplastic astrocytoma [26].

In addition, in this location, tumors harboring a YAP1-MAMLD1 fusion—apparently a parameter of favorable prognosis [3, 11, 15, 17]—accounted for two patients (29%), and neither of them suffered from relapse or died.

It is generally accepted that patients with PF ependymomas face a worse prognosis, especially at young age; however, a recent publication showed no age-dependent prognostic difference for these tumors [9]. Although radiation therapy can be effective in infants, most neurooncological centers are reluctant to apply irradiation in this age group because of severe long-term secondary effects. Therefore, the treatment of infants imposes difficulties compared to older patients with PF ependymoma; hence, there is a need for identification of specific risk factors and novel treatment options in this subgroup of patients.

The rate of incomplete resection, a parameter of inferior prognosis, occurred in 43% of infratentorial compared to 71% in supratentorial ependymoma of infancy. The shorter overall survival of patients with posterior fossa ependymoma may result from re-resection being more difficult compared to the supratentorial compartment.

All PF ependymomas in this study were classified as anaplastic (WHO grade III). The prognostic impact of WHO grading was demonstrated in some studies but not found in others [9, 11].

Furthermore, genomic analysis revealed that all 21 patients in this group showed balanced genomic profiles and were confirmed to represent PF-A group/CIMP + ependymoma, immunohistochemistry demonstrating a complete loss of H3K27me³, which represents the subgroup of ependymoma with the worst prognosis [3, 27]. Interestingly, chromosome 1q gain and the associated structural genomic profile—both reported indicators of adverse prognosis [2, 5, 13, 14, 28, 29]—were not found in any of the 21 PF tumor samples in our cohort.

In summary, we were able to demonstrate that ependymomas of infancy represent distinct location- and age-associated ependymoma entities in terms of underlying biology and genetics. Even though statistical significance regarding survival could not be reached, most probably due to the small cohort sizes, ST ependymomas seem to have a superior prognosis compared to PF ependymoma in terms of OS. The EFS of patients with tumors of the two locations may not be different, but re-resection being more easily achieved in the ST-compartment may lead to the long-term survival benefits noted in the previous studies.

ST ependymomas harbor distinct genetic alterations (such as fusions) which represent possible therapeutic targets; these fusions are absent in PF ependymomas. Besides the optimal timing for radiation therapy, the field of targeted therapies remains to be explored for ependymoma in the future.

Our results show that risk stratification for pediatric ependymoma remains challenging and suggest that separate analysis—regarding risk factors and treatment regimens—is required for tumors in different locations and among different age groups.