Abstract
Background
Ependymal tumors are glial tumors that commonly manifest in children and young adults. Their classification has remained entirely morphological until recently, and surgery and radiotherapy are the main treatment options, especially in adults. Here we sought to correlate DNA methylation profiles with clinical and pathological characteristics in the prospective cohort of the German Glioma Network.
Methods
Tumors from 122 adult patients with myxopapillary ependymoma, ependymoma, anaplastic ependymoma, subependymoma, or RELA fusion-positive ependymoma classified according to the World Health Organization (WHO) 2016 were subjected to DNA methylation profiling using the Illumina HumanMethylation450 BeadChip platform. Molecular data were correlated with histologic features and clinical characteristics.
Results
At a median follow-up of 86.7 months, only 22 patients experienced progression (18.0%) and 13 patients (10.7%) died. Each tumor could be assigned to one of the previously defined molecular ependymoma subgroups. All histologic subependymomas corresponded to subependymoma (SE) DNA methylation subgroups, but the reverse was not true: 19 histologic ependymomas (WHO grade II) were allocated to molecular SE groups. Similarly, all histological myxopapillary ependymomas were assigned to the molecularly defined spinal myxopapillary ependymoma (SP-MPE) class, but this molecular subgroup additionally included 15 WHO grade II ependymomas by histology. Overall, WHO grade II ependymomas distributed into 7 molecular subgroups.
Conclusion
Most adult patients with ependymoma show a favorable prognosis. Molecular classification may provide diagnostic and prognostic information beyond histology and facilitate patient stratification in future clinical trials. The prognostic significance of a subependymoma or myxopapillary ependymoma DNA methylation phenotype without corresponding histology requires further study.
Key Points
1. Ependymoma diagnosed in adult patients most often shows a good prognosis.
2. Molecular classification can support diagnostic and prognostic information beyond histology.
Keywords: adult, ependymal tumors, intracranial, molecular subgroups, spinal
Importance of the Study
This study used DNA methylation profiling to molecularly classify 122 adult ependymal tumors from patients enrolled in the German Glioma Network. Each tumor was assigned to one of 8 previously defined molecular subgroups of ependymal tumors. Histologically diagnosed SEs, SP-MPEs, and supratentorial RELA fusion-positive ependymomas (ST-EPN-RELA) were unambiguously assigned to the respective molecular subgroups. Only one anaplastic ependymoma (WHO grade III) was classified into the subgroup spinal subependymoma (SP-SE), the remaining 6 anaplastic ependymomas were assigned to the molecular subgroup posterior fossa group B (PF-EPN-B). WHO grade II ependymomas distributed into 7 molecular subgroups. Future studies need to determine whether molecular reclassification allows for treatment de-escalation (eg, delay of radiotherapy) in subgroups of patients. DNA methylation profiling may provide diagnostic and prognostic information beyond histology and thus may facilitate patient stratification in future clinical trials.
The 2016 World Health Organization (WHO) classification of central nervous system (CNS) tumors recognizes 5 distinct entities of ependymal tumors: subependymoma (WHO grade I), myxopapillary ependymoma (WHO grade I), ependymoma (WHO grade II), anaplastic ependymoma (WHO grade III), and RELA fusion-positive ependymoma (WHO grade II or III).1 Recently 9 molecular subgroups of ependymomas with key genetic and epigenetic characteristics across all age groups have been described.2 These were based on major CNS compartments and histopathologic and molecular findings: spine (SP), posterior fossa (PF), and supratentorial (ST) localization; subependymoma (SE), myxopapillary ependymoma (MPE), ependymoma (EPN), anaplastic ependymoma (EPN-A or -B: balanced genome or chromosomal instability), and ST-EPN-RELA, defined by RELA fusion transcript expression, as well as ST-EPN-YAP1, defined by the presence of YAP1 fusion transcripts.
The current histopathologic assessment has shortcomings, reliability and clinical significance of WHO grade II versus III have remained controversial, and histologic subtyping provides only limited guidance for clinical decision making. It might thus be worthwhile to supplement the current WHO classification by assessment of additional molecular markers or large-scale molecular profiling approaches to tailor management strategies and to avoid undertreatment as well as overtreatment of individual patients. Accordingly, we molecularly classified ependymomas of adults based on DNA methylation patterns determined by 450k DNA methylation arrays and related these data to age, tumor location, and histology. Longer follow-up of our cohort will be required to derive robust conclusions on the prognostic value of this new classifier in adult ependymoma patients.
Patients and Methods
Molecular Classification Using DNA Methylation Profiling
The present study evaluated clinical features and histopathologic findings in the ependymoma cohort of the German Glioma Network. The German Glioma Network is a prospective cohort study that enrolled adult patients with gliomas at 9 clinical centers in Germany in 2004–2012 (http://www.gliomnetzwerk.de). All patients gave written informed consent according to the research proposals approved by the institutional review boards of the participating institutions. We included all patients with newly diagnosed tumors histologically diagnosed as subependymoma, myxopapillary ependymoma, ependymoma, or anaplastic ependymoma. The histologic diagnosis was verified by central neuropathology review in all 122 patients based on the 2016 WHO classification of tumors of the CNS.1
The Illumina Infinium HumanMethylation450 (450k) array was used to obtain genome-wide DNA for tumor samples and normal control tissues. Data were generated at the Genomics and Proteomics Core Facility of the German Cancer Research Center. DNA methylation data were generated from either fresh-frozen (n = 88) or formalin-fixed paraffin-embedded (FFPE) tissue samples (n = 34). Tumor tissue originated from primary tumors (n = 114) and from relapse tumors (n = 8). For most fresh-frozen samples, >500 ng of DNA was used as input material; 250 ng DNA was used for most FFPE tissues. On-chip quality metrics of all samples were carefully controlled. Samples were also checked for unexpected genotype matches by pairwise comparison of the 65 genotyping probes included on the 450k array.
All computational analyses were performed in R version 3.2.0.3 Raw signal intensities were obtained from IDAT files using the minfi Bioconductor package version 1.14.0.4 Each sample was individually normalized by performing a background correction (shifting of the 5% percentile of negative control probe intensities to 0) by and a dye-bias correction (scaling of the mean of normalization control probe intensities to 10000) for both color channels. Subsequently, a correction for the type of material tissue (FFPE/frozen) was performed by fitting univariate, linear models to the log2-transformed intensity values (removeBatchEffect function, limma package version 3.24.15). The methylated and unmethylated signals were corrected individually. Beta-values were calculated from the retransformed intensities using an offset of 100 (as recommended by Illumina).
The following filtering criteria were applied: removal of probes targeting the X and Y chromosomes (n = 11551), removal of probes containing a single nucleotide polymorphism (dbSNP132 Common) within 5 base pairs of and including the targeted cytosine-phosphate-guanine site (n = 7998), probes not mapping uniquely to the human reference genome (hg19) allowing for one mismatch (n = 3965), and probes not included on the new Illumina EPIC array (n = 32260). In total, 428799 probes were kept for the analysis.
Copy number variation (CNV) analysis from 450k methylation array data was performed using the conumee Bioconductor package version 1.3.0. Two sets of 50 control samples displaying a balanced copy number profile from both male and female donors were used. For CNV analysis, no previous normalization steps were performed. To predict molecular subgroups, a Random Forest classifier was applied using 428799 probes. This classifier was trained on a reference set of 2801 methylation profiles of brain tumors that were previously assigned to 91 molecular subgroups which cover almost all entities listed in the 2016 WHO classification of CNS tumors. For subgrouping of the 122 samples of this study the classification model including 91 molecular subgroups of brain tumors was used. The overall prediction performance of this classifier was validated by a 3-fold cross validation indicating a very high classification accuracy with a misclassification error rate of 4.28% and a multiclass area under the curve of 0.9998.5,6
Statistical Analyses
Progression-free survival (PFS) and overall survival (OS) curves were estimated by the Kaplan–Meier method. PFS was calculated from the date of surgery to the date of progression. Patients without documented progression were censored at the last follow-up visit for PFS. OS was measured from the date of surgery to the date of death. Patients without confirmed death were censored for OS at last follow-up. Survival-related analyses were calculated using the log-rank test. The association between molecular subgroups and clinical characteristics was analyzed by Chi-square test and Fisher’s exact test and differences in age by the Mann–Whitney U-test. All statistical tests were two-tailed, and a P-value of 0.05 was set as statistically significant. All statistical analyses were performed using IBM SPSS Statistics v24. Chi-square tests were used to calculate the independence of distinct chromosomal aberrations across molecular subgroups.2 Only gains and losses of whole chromosomes and chromosome arms were included in this analysis. Independences of CNV between molecular subgroups were calculated by Chi-square tests, P-values were computed by 100000 Monte Carlo simulations.
Results
Patient Characteristics
Table 1 summarizes relevant patient characteristics. Individual patient data are provided in Supplementary Table 1. Median age was 46 years (range, 18–80 y). The distribution of ages at diagnosis was relatively uniform across 10-year age groups between 30 and 70 years (15.6–18.9%), with a peak incidence in the 41–50 year age group (28.7%) and fewer diagnoses in patients older than 70 years (4.9%). There was a male predominance (67.2%). Histologic diagnosis was WHO grade II ependymoma in the majority of cases (60.7%). WHO grade III (anaplastic) ependymomas were rarely diagnosed. The most commonly affected compartment was the spine (53.3%) followed by infratentorial (31.1%) and supratentorial (14.8%) locations. Metastatic dissemination at diagnosis was observed only once. Patients who received therapy beyond surgery at first-line treatment had myxopapillary ependymoma (n = 1), ependymoma grade II (n = 7), anaplastic ependymoma grade III (n = 5), or RELA fusion-positive ependymoma (n = 2). Each of these patients was treated with postoperative radiotherapy (RT) (Table 2, Supplementary Table 1).
Table 1.
Patient characteristics
| N (%) | |
|---|---|
| All Patients | 122 |
| Age, y | |
| Median | 46 |
| Range | 18–80 |
| Age classes | |
| <31 y | 20 (16.4) |
| 31–40 y | 23 (18.9) |
| 41–50 y | 35 (28.7) |
| 51–60 y | 19 (15.6) |
| 61–70 y | 19 (15.6) |
| >70 y | 6 (4.9) |
| Sex | |
| Male | 82 (67.2) |
| Female | 40 (32.8) |
| KPS at enrollment | |
| <70 | 1 (1.0) |
| 70–80 | 41 (40.6) |
| 90–100 | 59 (58.4) |
| No data | 21 |
| Histologic diagnosis (ICD-O) | 23 (18.9) |
| Subependymoma (9383/1) | 14 (11.5) |
| Myxopapillary ependymoma (9394/1) | 74 (60.7) |
| Ependymoma (9391/3; 9393/3) | 7 (5.7) |
| Anaplastic ependymoma (9392/3) RELA fusion-positive (9396/3) | 4 (3.3) |
| WHO 2016 tumor grade | |
| I | 37 (30.3) |
| II | 74 (60.7) |
| III | 11 (9.0) |
| Tumor localization at diagnosis | |
| Intracranial—supratentorial | 18 (14.8) |
| Intracranial—infratentorial | 38 (31.1) |
| Spinal | 65 (53.3) |
| Disseminated | 1 (0.8) |
| Follow-up | |
| Median follow-up (mo) | 86.7 |
| Progression | 22 (18.0) |
| Dead | 13 (10.7) |
| Due to tumor progression | 6 (4.9) |
| Other reason | 6 (4.9) |
| Unknown | 1 (0.8) |
KPS: Karnofsky performance score; ICD-O: International Classification of Diseases for Oncology.
Table 2 .
Treatment regimens
| N (%) | |
|---|---|
| All Patients | 122 |
| Initial tumor resection | |
| Partial | 1 (0.8) |
| Subtotal | 23 (18.9) |
| Gross total | 98 (80.3) |
| Number of surgeries | |
| 1 | 105 (86.1) |
| >1 | 17 (13.9) |
| First-line therapy beyond surgery | |
| RT alone | 15 (12.3) |
| No therapy | 107 (87.7) |
| Therapy at first progression | 22 |
| No therapy | 4 (18.0) |
| Re-resection | 8 (36.4) |
| Re-resection plus CCNU | 1 (4.5) |
| Re-resection plus RT | 5 (22.7) |
| Re-resection plus temozolomide/RT | 1 (4.5) |
| RT | 2 (13.6) |
Median follow-up was 86.7 months; 22 patients experienced disease progression (18.0%), and 13 patients (10.7%) died: 4 patients with spinal, 8 patients with intracranial, and 1 patient with disseminated disease. Progression was seen in 1 of 23 patients (4.3%) with subependymoma, 5 of 14 patients (35.7%) with myxopapillary ependymoma, 9 of 74 patients (12.2%) with ependymoma WHO grade II, 4 of 7 patients (57.1%) with anaplastic ependymoma grade III, and 3 of 4 patients with RELA fusion-positive tumors. At recurrence, 15 of 22 patients (68.2%) had a second resection, followed by RT (n = 5), CCNU chemotherapy (n = 1), or combined radiochemotherapy with temozolomide (n = 1). Two of 22 patients (9.1%) received RT alone and 4 patients received no further treatment (Table 2, Supplementary Table 1). PFS rate was 96.4% at 1 year (95% CI: 93.0–100), 90.4% at 3 years (95% CI: 84.6–96.1), and 84.6% at 5 years (95% CI: 77.3–81.8) for the entire patient cohort. OS rate was 99.1% at 1 year (95% CI: 97.5–100), 97.1% at 3 years (95% CI: 93.9–100), and 95.9% at 5 years (95% CI: 91.9–100). Deceased patients who did not die from ependymoma had other tumors (n = 2) or cardiovascular or respiratory diseases (n = 4).
Molecular Subgrouping
Ependymal tumors of all 122 patients could be assigned to one of 8 subgroups defined by distinct DNA methylation profiles: SP-SE (n = 5; 4.1%), SP-MPE (n = 29; 23.8%), SP-EPN (n = 32; 26.2%), PF-SE (n = 24; 19.7%), PF-EPN-A (n = 1; 0.8%), PF-EPN-B (n = 13; 10.7%), ST-SE (n = 14; 11.5%), and ST-EPN-RELA (n = 4; 3.3%) (Table 3, Supplementary Figures 1, 2). No tumor was assigned to another non-ependymal tumor entity and no tumor was removed from the analysis. Patients with PF-EPN-B and ST-EPN-RELA as well as with SP-MPE and SP-EPN tumors were younger (median age, 37–43 y), whereas patients in the subgroups SP-SE, PF-SE, and ST-SE tended to be older (median age, 49–59 y) (Supplementary Figure 3).
Table 3.
Patient characteristics based on ependymal tumor subgroups
| SP-SE | SP-MPE | SP-EPN | PF-SE | PF-EPN-A | PF-EPN-B | ST-SE | ST-EPN-RELA | |
|---|---|---|---|---|---|---|---|---|
| Number n (%) | 5 (4.1) | 29 (23.8) | 32 (26.2) | 24 (19.7) | 1 (0.8) | 13 (10.7) | 14 (11.5) | 4 (3.3) |
| Age, y (median range) | 59 (46–73) | 43 (18–80) | 42 (24–69) | 54 (18–79) | 42 (42–42) | 37 (19–63) | 49 (26–73) | 37 (26–62) |
| Sex | ||||||||
| Male | 2 (40.0) | 20 (69.0) | 17 (53.1) | 23 (95.8) | 1 (100.0) | 6 (46.2) | 11 (78.6) | 2 (50.0) |
| Female | 3 (60.0) | 9 (31.0) | 15 (46.9) | 1 (4.2) | 0 | 7 (53.8) | 3 (21.4) | 2 (50.0) |
| Karnofsky performance score | ||||||||
| <70 | 0 | 0 | 0 | 1 (4.2) | 0 | 0 | 0 | 0 |
| 70–80 | 2 (66.7) | 11 (52.4) | 10 (41.7) | 8 (33.3) | 1 (100.0) | 3 (25.0) | 6 (46.2) | 0 |
| 90–100 | 1 (33.3) | 10 (47.6) | 14 (58.3) | 15 (62.5) | 0 | 9 (75.0) | 7 (53.8) | 3 (100.0) |
| No data | 2 | 8 | 8 | 1 | 1 | 1 | ||
| Surgery | ||||||||
| Total | 4 (80.0) | 25 (86.2) | 28 (87.5) | 17 (70.8) | 0 | 9 (69.2) | 12 (85.7) | 3 (75.0) |
| Subtotal | 1 (20.0) | 4 (13.8) | 3 (9.4) | 7 (29.2) | 1 (100.0) | 4 (30.8) | 2 (14.3) | 1 (25.0) |
| Partial | 0 | 0 | 1 (3.1) | 0 | 0 | 0 | 0 | 0 |
| Histology | ||||||||
| Subependymoma | 3 (60.0) | 0 | 0 | 7 (29.2) | 0 | 0 | 13 (92.9) | 0 |
| Myxopapillary ependymoma | 0 | 14 (48.3) | 0 | 0 | 0 | 0 | 0 | 0 |
| Ependymoma | 1 (20.0) | 15 (51.7) | 32 (100.0) | 17 (70.8) | 1 (100.0) | 7 (53.8) | 1 (7.1) | 0 |
| Anaplastic ependymoma | 1 (20.0) | 0 | 0 | 0 | 0 | 6 (46.2) | 0 | 4 (100.0) |
| RELA fusion-positive | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 4 (100.0) |
| WHO 2016 grade | ||||||||
| I | 3 (60.0) | 14 (48.3) | 0 | 7 (29.2) | 0 | 0 | 13 (92.9) | 0 |
| II | 1 (20.0) | 15 (51.7) | 32 (100.0) | 17 (70.8) | 1 (100.0) | 7 (53.8) | 1 (7.1) | 0 |
| III | 1 (20.0) | 0 | 0 | 0 | 0 | 6 (46.2) | 0 | 4 (100.0) |
| Localization | ||||||||
| Intracranial—supratentorial | 0 | 0 | 0 | 0 | 0 | 0 | 14 (100.0) | 4 (100.0) |
| Intracranial—infratentorial | 1 (20.0) | 0 | 0 | 24 (100.0) | 1 (100.0) | 12 (92.3) | 0 | 0 |
| Spinal | 4 (80.0) | 29 (100.0) | 32 (100.0) | 0 | 0 | 0 | 0 | 0 |
| Disseminated | 0 | 0 | 0 | 0 | 0 | 1 (7.7) | 0 | 0 |
| First-line therapy beyond surgery | ||||||||
| RT alone (including cyber knife) | 0 | 2 (6.9) | 1 (3.1) | 2 (8.3) | 0 | 8 (61.5) | 0 | 2 (50.0) |
| None | 5 (100.0) | 27 (93.1) | 31 (96.9) | 22 (91.7) | 1 (100.0%) | 5 (38.5) | 14 (100.0) | 2 (50.0) |
| Progression | ||||||||
| Events, n (%) | 0 | 6 (20.7) | 2 (6.3) | 3 (12.5) | 0 | 7 (53.8) | 1 (7.1) | 3 (75.0) |
| Death | ||||||||
| Events, n (%) | 0 | 2 (6.9) | 1 (3.1) | 2 (8.3) | 0 | 3 (23.1) | 2 (14.3) | 3 (75.0) |
Fig. 1 and Supplementary Table 2 illustrate how WHO 2016 diagnoses distribute into molecular subgroups and localization. All histologic subependymomas were assigned to the molecular SE groups. However, tumors assigned to the molecular subependymoma subgroups SP-SE, PF-SE, and ST-SE also included 19 tumors histologically classified as ependymoma grade II and 1 anaplastic ependymoma. All patients diagnosed with myxopapillary ependymoma were assigned to the molecularly defined SP-MPE subgroup. This molecular subgroup, however, also included 15 grade II ependymomas by histology. All patients assigned to SP-EPN, PF-EPN-A, and PF-EPN-B corresponded to grade II or III ependymomas (Figure 1, Supplementary Table 2).
Fig. 1.
Reassignment of WHO 2016–based ependymoma diagnoses to molecular ependymoma subgroups based on DNA methylation profiles in the German Glioma Network cohort of adult patients with ependymomas stratified according to tumor location. (A) Supratentorial ependymoma (ST). (B) Posterior fossa ependymoma (PF). (C) Spinal ependymoma (SP).
The potential clinical implications become apparent when considering group assignments by compartment. This yielded few apparent changes in the supratentorial compartment (Fig. 1A). Among the posterior fossa tumors, subependymomas and anaplastic ependymomas showed a perfect match into the PF-SE and PF-EPN-B groups, whereas grade II ependymomas were heterogeneous by DNA methylation profiling and distributed to the prognostically different groups of PF-SE and PF-EPN-B (Fig. 1B). The most extensive reassignments by molecular profiling occurred in the spinal compartment, where a third of grade II ependymomas were assigned to the SP-MPE molecular subgroup (Fig. 1C). Representative histologic patterns of tumors, which were assigned to distinct molecular subgroups such as subependymoma, ependymoma grade II, myxopapillary ependymoma, anaplastic ependymoma, and RELA fusion-positive ependymoma, are shown in Supplementary Figure 4.
Chromosomal Alterations Show Distinct Patterns within Molecular Subgroups
CNVs were calculated by analyzing combined intensity values of the methylation probes.7 Whole genome copy number profiles showed distinct chromosomal aberrations in terms of frequencies and specificity across the molecular ependymal subgroups (Fig. 2, Supplementary Table 3). The single PF-EPN-PFA tumor was excluded here. Overall, molecular SE tumors within spinal, posterior fossa, and supratentorial compartments showed infrequent CNVs. Surprisingly, SE commonly showed a loss of chromosome 19, most frequently within PF-SE (79%), but also in ST-SE (50%) and SP-SE (40%), whereas genomic profiles were otherwise relatively flat. Another CNV frequently observed in SP-SE and PF-SE was partial chromosome 6 loss.
Fig. 2.
Copy number variations across molecular subgroups of adult ependymal tumors. Summary of chromosomal imbalances showing distinct alterations within 7 molecular subgroups of ependymal tumors. Copy number plots were generated based on DNA methylation, color code for losses (red), gains (green), and balanced chromosomal profiles (gray). Results were plotted as frequencies at which these aberrations occurred within each molecular subgroup; significance is illustrated by P-values, which indicate distinct distributions of alterations across the molecular subgroups (chi-square test).
Chromosome 6 loss was also frequently seen in PF-EPN-B tumors (61%). Additionally, PF-EPN-B tumors showed gains of chromosomes 15 (54%), 18 (54%), and 20 (54%), and losses of chromosome 17 (38%), which were the most frequent aberrations.2,8,9 Gain of chromosome 1q was seen in PF-EPN-B (23%) and in one ST-EPN-RELA.
In SP-EPN and SP-MPE, most chromosomal gains and losses comprised large regions, including whole chromosomes or chromosomal arms. SP-EPN showed deletions of chromosome arm 22q in more than 80% of tumors.10 Several CNVs were present in both SP-MPE and SP-EPN, the most common being loss of 22q in SP-EPN (90%) and SP-MPE tumors (47%).
Characteristic and significant CNVs for SP-SE were losses of chromosomes 18 (20%) and 19 (40%), for SP-EPN gain of chromosome 12 (56%), loss of 13q (31%), and loss of 14q (31%). SP-MPE harbored gains of chromosome 16 (25%) and losses of chromosome 10 (25%). Exclusive chromosomal changes of ST-SE were losses of chromosomes 8 (29%) and 19 (50%). Characteristic and most frequent CNVs of ST-EPN-RELA tumors were losses of chromosome 3 (75%), 9 (100%), and 11 (75%), as well as focal losses of chromosome 11q (75%), where the fusion-partner genes C11orf95 and RELA are localized.11,12
Molecular Subgroups and Survival
Survival analyses showed differences in PFS and OS, although the low number of events allowed no conclusive statistical analysis between the 8 molecular subgroups. Molecular subgroups associated with the poorest outcome were PF-EPN-B and ST-EPN-RELA, with 5-year PFS probability of 65.8% for PF-EPN-B and 25% for ST-EPN-RELA patients (Supplementary Figures 5, 6).
The 8 PF-EPN-B patients who had RT after initial surgery had a 5-year PFS rate of 72.9% (95% CI: 40.6–100%) after a median follow-up of 7.3 years, as opposed to 53.3% (95% CI: 4.7–100%) after a median follow-up of 8.5 years for patients who did not. Four of 8 patients in the RT group relapsed compared with 3 of 5 patients without further therapy. All patients in the latter group initially had a gross total resection. In the RT group only 4 of 8 patients had been gross totally resected. The age was roughly equal (median 34.5 y for patients with RT vs 37 y for patients without treatment). PF-EPN-B patients had an OS rate at 5 years of around 100%, but survival probability dropped down to 37.5% at around 10 years (Supplementary Figure 5). No significant differences in PFS (P = 0.468; data not shown) and OS (P = 0.083; data not shown) between WHO grade II and III tumors were observed in this subgroup. Notably, 3 of 6 PF-EPN-B patients with WHO grade III tumors died, but none of 7 patients with WHO grade II.
The number of ST-EPN-RELA tumors was small: 3 of 4 patients relapsed and died (Supplementary Figure 6). In contrast, SE tumors from the supratentorial and infratentorial compartments showed very good outcome. Importantly, 2 patients with ST-SE and 1 patient with PF-SE tumors died from other diseases, but not from their ependymomas (Supplementary Table 1).
Patients with spinal molecular subgroups SP-EPN, SP-MPE, and SP-SE had an excellent OS rate of 100% after 5 years. SP-MPE may relapse earlier and more often than the other spinal molecular subtypes, but longer follow-up is needed (Supplementary Figure 7).
We investigated the molecular subgroup SP-MPE in more detail and compared patients with grade I myxopapillary ependymoma (n = 14) and grade II ependymoma (n = 15), the latter representing patients whose tumors were molecularly reclassified as SP-MPE (Supplementary Table 4, Supplementary Figure 8). A comparison of clinical features revealed that patients with grade I myxopapillary ependymomas were younger than patients with grade II ependymomas (median 36 y vs 50 y, P = 0.023). We observed a trend of SP-MPE–assigned grade II ependymomas to show tumor localization of the conus medullaris or filum terminale (Supplementary Table 1). None of the patients with SP-MPE–assigned myxopapillary ependymomas died, but 2 patients with SP-MPE–assigned grade II ependymomas did, one patient for unknown reason, while death of the other patient was unrelated to the spinal tumor (Supplementary Table 4).
For SP-MPE myxopapillary ependymoma patients, 5 progressions were observed compared with only 1 progression in patients with SP-MPE–assigned grade II ependymoma (Supplementary Figure 8). The 2 cohorts of SP-MPE patients showed a trend for better PFS of patients with myxopapillary ependymomas (P = 0.071, n.s.).
The outcome comparison of the SP-EPN tumors (n = 32), all of which had grade II ependymoma histology, versus tumors of the same histology that were assigned to the epigenetic SP-MPE subgroup (n = 15) is presented in Supplementary Figure 9 and Supplementary Table 5. No relevant differences emerged. SP-EPN patients were younger by trend (median 42 y vs 50 y, P = 0.123). Three patients died (n = 1 SP-EPN and n = 2 SP-MPE).
Discussion
Ependymomas are rare brain tumors in adults. Half of these tumors occur in the spinal cord. Surgery and radiotherapy are the main treatment modalities.13 Gross total resection as safely feasible is recommended for all ependymal tumors, whereas decisions against or for RT depend on residual tumor and histology based on the 2016 WHO classification.14
The present study is the first effort to characterize adult ependymomas beyond the WHO classification of 2016, using a classifier based on DNA methylation profiling recently proposed for a largely pediatric population.2,6 The major strength of the present study is good clinical annotation regarding age, tumor location, treatment, and follow-up. Male patients were overrepresented in our cohort relative to data from larger registries.15 The low number of PFS and OS events, despite long follow-up, reflects the improved contemporary outcome for adult ependymoma patients. However, it precluded an in-depth analysis of the potential prognostic value of the new molecular classifier. Similarly, no firm conclusions on the role of therapeutic measures beyond surgery can be derived.
The vast majority of ependymomas in adults could readily be assigned to 7 of 9 recently defined molecular subgroups (Table 3). ST-EPN-YAP1 tumors were not detected. ST-EPN-RELA, ST-EPN-YAP1, and PF-EPN-A are more common in or even restricted to children. That none of the tumors included in this study were assigned a non-ependymoma diagnosis by methylation profiling may be explained by the central pathology review within the German Glioma Network that took place prior to inclusion.
The poor outcome of ST-EPN-RELA tumors observed in children may also be observed in adults (Table 3). There was a weak trend in the small group of PF-EPN-B patients, who might show a better outcome when treated with RT compared with non-irradiated patients. However, due to the low number of patients, a recommendation of upfront RT cannot be concluded. Whether PF-EPN-B patients with incompletely resected tumors will benefit from upfront RT has to be evaluated in a randomized clinical trial in the future.
The most important potential clinical implications of our study can be deduced from Fig. 1. First, although all histologically defined subependymomas of WHO grade I were assigned to the molecular subependymoma subgroups SP-SE, PF-SE, and ST-SE, these molecular subgroups also included histologically defined grade II ependymomas (n = 19) and even a single case of grade III anaplastic ependymoma. It is tempting to speculate, and is supported by the preliminary data of this study, that among the histologically defined grade II ependymomas, the molecular SE subgroup tumors have the best prognosis, and patients with these tumors may be at risk of being overtreated with RT. The newly described chromosome 19 loss in subependymomas, especially in PF-SE, could potentially support the diagnostic process in the separation of PF-SE and PF-EPN-B, where this aberration was detected in less than 4% of cases (n = 2) (Fig. 2).
Second, while all patients with histologically defined grade I myxopapillary ependymoma were assigned to the molecularly defined SP-MPE subgroup, this molecular subgroup additionally included 15 spinal grade II ependymomas by histology. The prognostic significance of this change in group allocation from histology to molecular subgroup is less clear, although some institutions adopt different post-resection strategies of wait-and-see versus RT in patients with myxopapillary ependymoma versus grade II ependymoma.14 Interestingly, gene expression profiling had previously indicated that grade II ependymoma and myxopapillary ependymoma are biologically distinct entities, although a subset of cases showed transcriptional profiles that could not be firmly assigned to either group.10 The present study also suggests that a subset of spinal grade II ependymomas is molecularly related to myxopapillary ependymomas by DNA methylation profiling. Interestingly, a previous study reported on the presence of both classic and myxopapillary histologic features in about 10% of lumbosacral ependymal tumors.16 Thus, histology may be ambiguous or not fully representative in a subset of spinal ependymal tumors, and additional DNA methylation profiling may thus be helpful in refining the classification of these cases. In this context, DNA methylation patterns of brain tumors are usually stable from the cell of origin until the development of a brain tumor and even at relapse.2,17
Third, we hypothesize that once targeted treatments for ependymal tumors become available, a molecular classifier will be superior to histology to enrich for clinical trial populations of patients with tumors sharing a similar cell of origin and biology.
Limitations of this study include the limited sample size per subgroup and the low number of PFS and OS events. Long-term follow-up and independent cohort studies are required to confirm our assumption that molecular subgroup assignment of adult patients with ependymal tumors provides superior outcome description and enrichment for future clinical trials assessing therapeutic interventions.
Funding
This work and the German Glioma Network were supported by German Cancer Aid (Deutsche Krebshilfe 70-3163-Wi 3). This study was supported by grants from the Stiftung Sibylle Assmus to H.W. This work was supported in part by the Cooperation Program in Cancer Research of the Deutsches Krebsforschungszentrum (DKFZ) and Israel’s Ministry of Science and Technology (MOST) to H.W.
Conflict of interest statement
Felix Sahm: Speakers’ bureau: Agilent, Illumina; research funding: Agilent, Illumina. Travel, accommodations, expenses: Agilent, Illumina, Roche.
Ulrich Herrlinger: Consulting or advisory role: Roche, BMS, Norocure, Noxxon Pharma; speakers’ bureau: Roche, Medac; research funding: Roche. Travel, accommodations, expenses: Roche, BMS, Medac.
Torsten Pietsch: Research funding: Affymetrix.
Guido Reifenberger: Honoraria: Amgen, Celldex, Medac; consulting or advisory role: Celldex; research funding: Roche, Merck.
Michael Weller: Honoraria: Merck Serono, Roche, Lilly, MSD, ImmunoCellular Therapeutics; consulting or advisory role: Bristol-Myers Squibb, Roche, Merck Serono, Magforce, Celldex, Lilly, Pfizer, Teva, Abbvie; research funding: Bayer, Roche, Merck Serono, Piqur, Actelion, Acceleron Pharma, Novocure, OGD2, Merck Sharp & Dohme.
Authorship statement
Conception and design: HW, DG, BH, TP, GR, SMP, JCT, MW
Collection and assembly of data: HW, DG, BH, JF, FS, DC, JCT, TP, GR, MW
Data analysis and interpretation: HW, DG, BH, MS, SMP, TP, GR, MW
Manuscript writing, final approval of manuscript, accountability for all aspects of the work: All authors.
Supplementary Material
References
- 1. Louis DN, Perry A, Reifenberger G, et al. The 2016 World Health Organization classification of tumors of the central nervous system: a summary. Acta Neuropathol. 2016;131(6):803–820. [DOI] [PubMed] [Google Scholar]
- 2. Pajtler KW, Witt H, Sill M, et al. Molecular classification of ependymal tumors across all CNS compartments, histopathological grades, and age groups. Cancer Cell. 2015;27(5):728–743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing;2015. [Google Scholar]
- 4. Aryee MJ, Jaffe AE, Corrada-Bravo H, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics. 2014;30(10):1363–1369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Breiman L. Random forests. Machine Learning 2001; 45(1):5–32. [Google Scholar]
- 6. Capper D, Jones DTW, Sill M, et al. DNA methylation-based classification of central nervous system tumours. Nature. 2018;555(7697):469–474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Hovestadt V, Remke M, Kool M, et al. Robust molecular subgrouping and copy-number profiling of medulloblastoma from small amounts of archival tumour material using high-density DNA methylation arrays. Acta Neuropathol. 2013;125(6):913–916. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Witt H, Mack SC, Ryzhova M, et al. Delineation of two clinically and molecularly distinct subgroups of posterior fossa ependymoma. Cancer Cell. 2011;20(2):143–157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Korshunov A, Witt H, Hielscher T, et al. Molecular staging of intracranial ependymoma in children and adults. J Clin Oncol. 2010;28(19):3182–3190. [DOI] [PubMed] [Google Scholar]
- 10. Mack SC, Agnihotri S, Bertrand KC, et al. Spinal myxopapillary ependymomas demonstrate a warburg phenotype. Clin Cancer Res. 2015;21(16):3750–3758. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Parker M, Mohankumar KM, Punchihewa C, et al. C11orf95-RELA fusions drive oncogenic NF-κB signalling in ependymoma. Nature. 2014;506(7489):451–455. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Pietsch T, Wohlers I, Goschzik T, et al. Supratentorial ependymomas of childhood carry C11orf95-RELA fusions leading to pathological activation of the NF-κB signaling pathway. Acta Neuropathol. 2014;127(4):609–611. [DOI] [PubMed] [Google Scholar]
- 13. Gilbert MR, Ruda R, Soffietti R. Ependymomas in adults. Curr Neurol Neurosci Rep. 2010;10(3):240–247. [DOI] [PubMed] [Google Scholar]
- 14. Rudà R, Reifenberger G, Frappaz D, et al. EANO guidelines for the diagnosis and treatment of ependymal tumors. Neuro Oncol. 2018;20(4):445–456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Vera-Bolanos E, Aldape K, Yuan Y, et al. ; CERN Foundation Clinical course and progression-free survival of adult intracranial and spinal ependymoma patients. Neuro Oncol. 2015;17(3):440–447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Jeibmann A, Egensperger R, Kuchelmeister K, et al. Extent of surgical resection but not myxopapillary versus classical histopathological subtype affects prognosis in lumbo-sacral ependymomas. Histopathology. 2009;54(2):260–262. [DOI] [PubMed] [Google Scholar]
- 17. Hovestadt V, Jones DT, Picelli S, et al. Decoding the regulatory landscape of medulloblastoma using DNA methylation sequencing. Nature. 2014;510(7506):537–541. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.