Integrated analyses reveal two molecularly and clinically distinct subtypes of H3 K27M-mutant diffuse midline gliomas with prognostic significance.

H3 K27M-altered diffuse midline gliomas (DMGs) are highly malignant tumours that arise in the midline structures of the CNS. Most DMGs carry an H3 K27M-mutation in one of the genes encoding for histone H3. Recent studies suggested that epigenetic subgroups of DMGs can be distinguished based on alterations in the MAPK-signalling pathway, tumour localisation, mutant H3-gene, or overall survival (OS). However, as these parameters were studied individually, it is unclear how they collectively influence survival. Hence, we analysed dependencies between different parameters, to define novel epigenetic, clinically meaningful subgroups of DMGs. We collected a multifaceted cohort of 149 H3 K27M-mutant DMGs, also incorporating data of published cases. DMGs were included in the study if they could be clearly allocated to the spinal cord (n = 31; one patient with an additional sellar tumour), medulla (n = 20), pons (n = 64) or thalamus (n = 33), irrespective of further known characteristics. We then performed global genome-wide DNA methylation profiling and, for a subset, DNA sequencing and survival analyses. Unsupervised hierarchical clustering of DNA methylation data indicated two clusters of DMGs, i.e. subtypes DMG-A and DMG-B. These subtypes differed in mutational spectrum, tumour localisation, age at diagnosis and overall survival. DMG-A was enriched for DMGs with MAPK-mutations, medullary localisation and adult age. 13% of DMG-A had a methylated MGMT promoter. Contrarily, DMG-B was enriched for cases with TP53-mutations, PDGFRA-amplifications, pontine localisation and paediatric patients. In univariate analyses, the features enriched in DMG-B were associated with a poorer survival. However, all significant parameters tested were dependent on the cluster attribution, which had the largest effect on survival: DMG-A had a significantly better survival compared to DMG-B (p < 0.001). Hence, the subtype attribution based on two methylation clusters can be used to predict survival as it integrates different molecular and clinical parameters.

Transcriptomics-based characterization of the immuno-stromal microenvironment in pediatric low-grade glioma.

Immunologic treatment options are uncommon in low-grade gliomas, although such therapies might be beneficial for inoperable and aggressive cases. Knowledge of the immune and stromal cells in low-grade gliomas is highly relevant for such approaches but still needs to be improved. Published gene-expression data from 400 low-grade gliomas and 193 high-grade gliomas were gathered to quantify 10 microenvironment cell populations with a deconvolution method designed explicitly for brain tumors. First, we investigated general differences in the microenvironment of low- and high-grade gliomas. Lower-grade and high-grade tumors cluster together, respectively, and show a general similarity within and distinct differences between these groups, the main difference being a higher infiltration of fibroblasts and T cells in high-grade gliomas. Among the analyzed entities, gangliogliomas and pleomorphic xanthoastrocytomas presented the highest overall immune cell infiltration. Further analyses of the low-grade gliomas presented three distinct microenvironmental signatures of immune cell infiltration, which can be divided into T-cell/dendritic/natural killer cell-, neutrophilic/B lineage/natural killer cell-, and monocytic/vascular/stromal-cell-dominated immune clusters. These clusters correlated with tumor location, age, and histological diagnosis but not with sex or progression-free survival. A survival analysis showed that the prognosis can be predicted from gene expression, clinical data, and a combination of both with a support vector machine and revealed the negative prognostic relevance of vascular markers. Overall, our work shows that low- and high-grade gliomas can be characterized and differentiated by their immune cell infiltration. Low-grade gliomas cluster into three distinct immunologic tumor microenvironments, which may be of further interest for upcoming immunotherapeutic research.

Distinct relapse pattern across molecular ependymoma types.

BACKGROUND: Ependymoma (EPN) is not a uniform disease but represents different disease types with biological and clinical heterogeneity. However, the pattern of when and where different types of EPN relapse is not yet comprehensively described. METHODS: We assembled 269 relapsed intracranial EPN from pediatric (n=233) and adult (n=36) patients from European and Northern American cohorts and correlated DNA methylation patterns and copy-number alterations with clinical information. RESULTS: The cohort comprised the following molecular EPN types: PF-EPN-A (n=177), ST-EPN-ZFTA (n=45), PF-EPN-B (n=31), PF-EPN-SE (n=12), and ST-EPN-YAP (n=4). First relapses of PF-EPN-B (PF: posterior-fossa) and PF-EPN-SE (SE: subependymoma) occurred later than of PF-EPN-A, ST-EPN-YAP (ST: supratentorial), or ST-EPN-ZFTA (median time to relapse: 4.3 and 6.0 years vs. 1.9/1.0/2.4 years; p<0.01). Metastatic or combined recurrences in PF-EPN-B and -A more often involved the spinal cord than in ST-EPN-ZFTA (72.7% and 40.0 vs. 12.5%; p<0.01). No distant relapses were observed in ST-EPN-YAP (n=4) or PF-EPN-SE (n=12). Post-relapse survival (PRS) was poor for PF-EPN-A and ST-EPN-ZFTA (5-year PRS: 44.5±4.4/47.8±9.1%), whereas PF-EPN-B and PF-EPN-SE displayed a 5-year PRS of 89.5±7.1/90.0±9.5% (p=0.03). However, 10-year PRS for PF-EPN-B dropped to 45.8±17.3%. Neither between radiation field and relapse pattern nor between radiation field and spinal involvement at relapse an impact was identified. Notably, all patients with relapsed ST-EPN-YAP did not receive upfront radiotherapy, but were successfully salvaged using irradiation at relapse. CONCLUSIONS: Relapse patterns of specific EPN types are different. Future clinical trials, treatment adaptions, duration of surveillance and diagnostics should be planned incorporating entity-specific relapse information.

Applications of Nanopore sequencing in precision cancer medicine.

Oxford Nanopore Technologies sequencing, also referred to as Nanopore sequencing, stands at the forefront of a revolution in clinical genetics, offering the potential for rapid, long read, and real-time DNA and RNA sequencing. This technology is currently making sequencing more accessible and affordable. In this comprehensive review, we explore its potential regarding precision cancer diagnostics and treatment. We encompass a critical analysis of clinical cases where Nanopore sequencing was successfully applied to identify point mutations, splice variants, gene fusions, epigenetic modifications, non-coding RNAs, and other pivotal biomarkers that defined subsequent treatment strategies. Additionally, we address the challenges of clinical applications of Nanopore sequencing and discuss the current efforts to overcome them.

Multiomic profiling of medulloblastoma reveals subtype-specific targetable alterations at the proteome and N-glycan level.

Medulloblastomas (MBs) are malignant pediatric brain tumors that are molecularly and clinically heterogenous. The application of omics technologies-mainly studying nucleic acids-has significantly improved MB classification and stratification, but treatment options are still unsatisfactory. The proteome and their N-glycans hold the potential to discover clinically relevant phenotypes and targetable pathways. We compile a harmonized proteome dataset of 167 MBs and integrate findings with DNA methylome, transcriptome and N-glycome data. We show six proteome MB subtypes, that can be assigned to two main molecular programs: transcription/translation (pSHHt, pWNT and pG3myc), and synapses/immunological processes (pSHHs, pG3 and pG4). Multiomic analysis reveals different conservation levels of proteome features across MB subtypes at the DNA methylome level. Aggressive pGroup3myc MBs and favorable pWNT MBs are most similar in cluster hierarchies concerning overall proteome patterns but show different protein abundances of the vincristine resistance-associated multiprotein complex TriC/CCT and of N-glycan turnover-associated factors. The N-glycome reflects proteome subtypes and complex-bisecting N-glycans characterize pGroup3myc tumors. Our results shed light on targetable alterations in MB and set a foundation for potential immunotherapies targeting glycan structures.

Molecular and epigenetic ex vivo profiling of testis cancer-associated fibroblasts and their interaction with germ cell tumor cells and macrophages.

Germ cell tumors (GCT) are the most common solid tumors in young men of age 15 - 40. In previous studies, we profiled the interaction of GCT cells with cells of the tumor microenvironment (TM), which showed that especially the 3D interaction of fibroblasts (FB) or macrophages with GCT cells influenced the growth behavior and cisplatin response as well as the transcriptome and secretome of the tumor cells, suggesting that the crosstalk of these cells with GCT cells is crucial for tumor progression and therapy outcome. In this study, we shed light on the mechanisms of activation of cancer-associated fibroblasts (CAF) in the GCT setting and their effects on GCT cells lines and the monocyte cell line THP-1. Ex vivo cultures of GCT-derived CAF were established and characterized molecularly and epigenetically by performing DNA methylation arrays, RNA sequencing, and mass spectrometry-based secretome analysis. We demonstrated that the activation state of CAF is influenced by their former prevailing tumor environment in which they have resided. Hereby, we postulate that seminoma (SE) and embryonal carcinoma (EC) activate CAF, while teratoma (TER) play only a minor role in CAF formation. In turn, CAF influence proliferation and the expression of cisplatin sensitivity-related factors in GCT cells lines as well as polarization of in vitro-induced macrophages by the identified effector molecules IGFBP1, LGALS3BP, LYVE1, and PTX3. Our data suggests that the vital interaction of CAF with GCT cells and with macrophages has a huge influence on shaping the extracellular matrix as well as on recruitment of immune cells to the TM. In conclusion, therapeutically interfering with CAF and / or macrophages in addition to the standard therapy might slow-down progression of GCT and re-shaping of the TM to a tumor-promoting environment. Significance: The interaction of CAF with GCT and macrophages considerably influences the microenvironment. Thus, therapeutically interfering with CAF might slow-down progression of GCT and re-shaping of the microenvironment to a tumor-promoting environment.

Medulloblastoma in children with Fanconi anemia: Association with FA-D1/FA-N, SHH type and poor survival independent of treatment strategies.

BACKGROUND: Outcome of children with medulloblastoma (MB) and Fanconi Anemia (FA), an inherited DNA repair deficiency, has not been described systematically. Treatment is complicated by high vulnerability to treatment-associated side effects, yet structured data are lacking. This study aims at giving a comprehensive overview about clinical and molecular characteristics of pediatric FA MB patients. METHODS: Clinical data including detailed information on treatment and toxicities of six previously unreported FA MB patients were supplemented with data of 16 published cases. RESULTS: We identified 22 cases of children with FA and MB with clinical data available. All MBs with subgroup reporting were SHH-activated (n=9), confirmed by methylation profiling in five patients. FA MB patients exclusively belonged to complementation groups FA-D1 (n=16) or FA-N (n=3). Patients were treated with postoperative chemotherapy only (50%) or radiotherapy (RT)±chemotherapy (27%). 23% did not receive adjuvant therapy. Excessive treatment-related toxicities were frequent. Severe hematological toxicity occurred in 91% of patients treated with alkylating chemotherapy, while non-alkylating agents and RT were less toxic. Median overall survival (OS) was 1 year (95%CI 0.3-1.8). 1-year-progression-free-survival (PFS) was 26.3±10.1% and 1-year-OS was 42.1±11.3%. Adjuvant therapy prolonged survival (1y-OS/1y-PFS 0%/0% without adjuvant therapy vs. 53.3±12.9%/33.3±12.2% with adjuvant therapy, p=0.006/p=0.086). CONCLUSIONS: MB in FA patients is strongly associated with SHH activation and FA-D1/FA-N. Despite the dismal prognosis, adjuvant therapy may prolong survival. Non-alkylating chemotherapy and RT are feasible in selected patients with careful monitoring of toxicities and dose adjustments. Curative therapy for FA MB-SHH remains an unmet medical need.

A prognostic neural epigenetic signature in high-grade glioma.

Neural-tumor interactions drive glioma growth as evidenced in preclinical models, but clinical validation is limited. We present an epigenetically defined neural signature of glioblastoma that independently predicts patients' survival. We use reference signatures of neural cells to deconvolve tumor DNA and classify samples into low- or high-neural tumors. High-neural glioblastomas exhibit hypomethylated CpG sites and upregulation of genes associated with synaptic integration. Single-cell transcriptomic analysis reveals a high abundance of malignant stemcell-like cells in high-neural glioblastoma, primarily of the neural lineage. These cells are further classified as neural-progenitor-cell-like, astrocyte-like and oligodendrocyte-progenitor-like, alongside oligodendrocytes and excitatory neurons. In line with these findings, high-neural glioblastoma cells engender neuron-to-glioma synapse formation in vitro and in vivo and show an unfavorable survival after xenografting. In patients, a high-neural signature is associated with decreased overall and progression-free survival. High-neural tumors also exhibit increased functional connectivity in magnetencephalography and resting-state magnet resonance imaging and can be detected via DNA analytes and brain-derived neurotrophic factor in patients' plasma. The prognostic importance of the neural signature was further validated in patients diagnosed with diffuse midline glioma. Our study presents an epigenetically defined malignant neural signature in high-grade gliomas that is prognostically relevant. High-neural gliomas likely require a maximized surgical resection approach for improved outcomes.

EpiDiP/NanoDiP: a versatile unsupervised machine learning edge computing platform for epigenomic tumour diagnostics.

DNA methylation analysis based on supervised machine learning algorithms with static reference data, allowing diagnostic tumour typing with unprecedented precision, has quickly become a new standard of care. Whereas genome-wide diagnostic methylation profiling is mostly performed on microarrays, an increasing number of institutions additionally employ nanopore sequencing as a faster alternative. In addition, methylation-specific parallel sequencing can generate methylation and genomic copy number data. Given these diverse approaches to methylation profiling, to date, there is no single tool that allows (1) classification and interpretation of microarray, nanopore and parallel sequencing data, (2) direct control of nanopore sequencers, and (3) the integration of microarray-based methylation reference data. Furthermore, no software capable of entirely running in routine diagnostic laboratory environments lacking high-performance computing and network infrastructure exists. To overcome these shortcomings, we present EpiDiP/NanoDiP as an open-source DNA methylation and copy number profiling suite, which has been benchmarked against an established supervised machine learning approach using in-house routine diagnostics data obtained between 2019 and 2021. Running locally on portable, cost- and energy-saving system-on-chip as well as gpGPU-augmented edge computing devices, NanoDiP works in offline mode, ensuring data privacy. It does not require the rigid training data annotation of supervised approaches. Furthermore, NanoDiP is the core of our public, free-of-charge EpiDiP web service which enables comparative methylation data analysis against an extensive reference data collection. We envision this versatile platform as a useful resource not only for neuropathologists and surgical pathologists but also for the tumour epigenetics research community. In daily diagnostic routine, analysis of native, unfixed biopsies by NanoDiP delivers molecular tumour classification in an intraoperative time frame.

Circulating extracellular vesicles as biomarker for diagnosis, prognosis, and monitoring in glioblastoma patients.

BACKGROUND: Extracellular vesicles (EVs) obtained by noninvasive liquid biopsy from patient blood can serve as biomarkers. Here, we investigated the potential of circulating plasma EVs to serve as an indicator in the diagnosis, prognosis, and treatment response of glioblastoma patients. METHODS: Plasma samples were collected from glioblastoma patients at multiple timepoints before and after surgery. EV concentrations were measured by nanoparticle tracking analysis and imaging flow cytometry. Tumor burden and edema were quantified by 3D reconstruction. EVs and tumors were further monitored in glioma-bearing mice. RESULTS: Glioblastoma patients displayed a 5.5-fold increase in circulating EVs compared to healthy donors (P < .0001). Patients with higher EV levels had significantly shorter overall survival and progression-free survival than patients with lower levels, and the plasma EV concentration was an independent prognostic parameter for overall survival. EV levels correlated with the extent of peritumoral fluid-attenuated inversion recovery hyperintensity but not with the size of the contrast-enhancing tumor, and similar findings were obtained in mice. Postoperatively, EV concentrations decreased rapidly back to normal levels, and the magnitude of the decline was associated with the extent of tumor resection. EV levels remained low during stable disease, but increased again upon tumor recurrence. In some patients, EV resurgence preceded the magnetic resonance imaging detectability of tumor relapse. CONCLUSIONS: Our findings suggest that leakiness of the blood-brain barrier may primarily be responsible for the high circulating EV concentrations in glioblastoma patients. Elevated EVs reflect tumor presence, and their quantification may thus be valuable in assessing disease activity.

Treatment response as surrogate to predict risk for disease progression in pediatric medulloblastoma with persistent magnetic resonance imaging lesions after first-line treatment.

BACKGROUND: This study aims at clarifying the impact of persistent residual lesions following first-line treatment for pediatric medulloblastoma. METHODS: Data on 84 pediatric patients with medulloblastoma and persistent residual lesions on centrally reviewed magnetic resonance imaging (MRI) at the end of first-line therapy were analyzed. RESULTS: Twenty patients (23.8%) had residual lesions in the tumor bed (R+/M0), 51 (60.7%) had distant lesions (R0/M+) and 13 (15.5%) had both (R+/M+). Overall response to first-line therapy was minor or partial (≥ 25% reduction, minor response [MR]/PR) for 64 (76.2%) and stable disease (SD) for 20 patients (23.8%). Five-year post-primary-treatment progression-free (pptPFS) and overall survival (pptOS) were superior after MR/PR (pptPFS: 62.5 ±â€…7.0%[MR/PR] vs. 35.9 ±â€…12.8%[SD], P = .03; pptOS: 79.7 ±â€…5.9[MR/PR] vs. 55.5 ±â€…13.9[SD], P = .04). Furthermore, R+/M + was associated with a higher risk for progression (5-year pptPFS: 22.9 ±â€…17.9%[R+, M+] vs. 72.4 ±â€…12.0%[R+, M0]; P = .03). Watch-and-wait was pursued in 58 patients, while n = 26 received additional treatments (chemotherapy only, n = 19; surgery only, n = 2; combined, n = 3; valproic acid, n = 2), and their outcomes were not superior to watch-and-wait (5-year pptPFS: 58.5 ±â€…7.7% vs. 51.6 ±â€…10.7% P = .71; 5-year pptOS: 76.3 ±â€…6.9% vs. 69.8 ±â€…9.7%, P = .74). For the whole cohort, 5-year pptPFS by molecular subgroup (58 cases) were WNT: 100%, SHH: 50.0 ±â€…35.4%, group-4, 52.5 ±â€…10.5, group-3 54.2 ±â€…13.8%; (P = .08). CONCLUSIONS: Overall response and extent of lesions can function as surrogate parameters to predict outcomes in pediatric MB patients with persistent lesions after first-line therapy. Especially in the case of solitary persistent medulloblastoma MRI lesions, additional therapy was not beneficial. Therefore, treatment response, extent/kind of residual lesions and further diagnostic information need consideration for indication of additional treatments for persisting lesions.

Corrigendum: Grade 3 Meningioma Survival and Recurrence Outcomes in an International Multicenter Cohort.

[This corrects the article DOI: 10.1055/s-0044-1779888.].

Insights from a multicenter study on adult H3 K27M-mutated glioma: Surgical resection's limited influence on overall survival, ATRX as molecular prognosticator.

BACKGROUND: H3 K27M-mutated gliomas were first described as a new grade 4 entity in the 2016 World Health Organization classification. Current studies have focused on its typical appearance in children and young adults, increasing the need to better understand the prognostic factors and impact of surgery on adults. Here, we report a multicentric study of this entity in adults. METHODS: We included molecularly confirmed H3 K27M-mutated glioma cases in patients ≥ 18 years diagnosed between 2016 and 2022. Clinical, radiological, and surgical features were analyzed. Univariate and multivariate analyses were performed to identify prognostic factors. RESULTS: Among 70 patients with a mean age of 36.1 years, the median overall survival (OS) was 13.6 ±â€…14 months. Gross-total resection was achieved in 14.3% of patients, whereas 30% had a subtotal resection and 54.3% a biopsy. Tumors located in telencephalon/diencephalon/myelencephalon were associated with a poorer OS, while a location in the mesencephalon/metencephalon showed a significantly longer OS (8.7 vs. 25.0 months, P = .007). Preoperative Karnofsky-Performance Score (KPS) ≤ 80 showed a reduced OS (4.2 vs. 18 months, P = .02). Furthermore, ATRX loss, found in 25.7%, was independently associated with an increased OS (31 vs. 8.3 months, P = .0029). Notably, patients undergoing resection showed no survival benefit over biopsy (12 vs. 11 months, P = .4006). CONCLUSIONS: The present study describes surgical features of H3 K27M-mutated glioma in adulthood in a large multicentric study. Our data reveal that ATRX status, location and KPS significantly impact OS in H3 K27M-mutated glioma. Importantly, our dataset indicates that resection does not offer a survival advantage over biopsy.

SMARCA4 Loss and Mutated ß-Catenin Induce Proliferative Lesions in the Murine Embryonic Cerebellum.

Almost all medulloblastomas (MB) of the Wingless/Int-1 (WNT) type are characterized by hotspot mutations in CTNNB1, and mouse models have convincingly demonstrated the tumor-initiating role of these mutations. Additional alterations in SMARCA4 are detected in ∼20% of WNT MB, but their functional role is mostly unknown. We, therefore, amended previously described brain lipid binding protein (Blbp)-cre::Ctnnb1(ex3)fl/wt mice by the introduction of floxed Smarca4 alleles. Unexpectedly, mutated and thereby stabilized ß-catenin on its own induced severe developmental phenotypes in male and female Blbp-cre::Ctnnb1(ex3)fl/wt mice in our hands, including a thinned cerebral cortex, hydrocephalus, missing cerebellar layering, and cell accumulations in the brainstem and cerebellum. An additional loss of SMARCA4 even resulted in prenatal death for most mice. Respective Blbp-cre::Ctnnb1(ex3)fl/wt::Smarca4fl/rec mutants (male and female) developed large proliferative lesions in the cerebellum evolving from E13.5 to E16.5. Histological and molecular analysis of these lesions by DNA methylation profiling and single-cell RNA sequencing suggested an origin in early undifferentiated SOX2-positive cerebellar progenitors. Furthermore, upregulated WNT signaling, altered actin/cytoskeleton organization, and reduced neuronal differentiation were evident in mutant cells. In vitro, cells harboring alterations in both Ctnnb1 and Smarca4 were negatively selected and did not show tumorigenic potential after transplantation in adult female recipient mice. However, in cerebellar explant cultures, mutant cells displayed significantly increased proliferation, suggesting an important role of the embryonic microenvironment in the development of lesions. Altogether, these results represent an important first step toward the unraveling of tumorigenic mechanisms induced by aberrant WNT signaling and SMARCA4 deficiency.

Assessing the risk to develop a growing teratoma syndrome based on molecular and epigenetic subtyping as well as novel secreted biomarkers.

In germ cell tumors (GCT), a growing teratoma during chemotherapy with decreasing tumor markers was defined as 'growing teratoma syndrome' (GTS) by Logothetis et al. in 1982. So far, its pathogenesis and specific treatment options remain elusive. We aimed at updating the GTS definition based on molecular and epigenetic features as well as identifying circulating biomarkers. We selected 50 GTS patients for clinical characterization and subsequently 12 samples were molecularly analyzed. We further included 7 longitudinal samples of 2 GTS patients. Teratomas (TER) showing no features of GTS served as controls. GTS were stratified based on growth rates into a slow (<0.5 cm/month), medium (0.5-1.5) and rapid (>1.5) group. By analyzing DNA methylation, microRNA expression and the secretome, we identified putative epigenetic and secreted biomarkers for the GTS subgroups. We found that proteins enriched in the GTS groups compared to TER were involved in proliferation, DNA replication and the cell cycle, while proteins interacting with the immune system were depleted. Additionally, GTSrapid seem to interact more strongly with the surrounding microenvironment than GTSslow. Expression of pluripotency- and yolk-sac tumor-associated genes in GTS and formation of a yolk-sac tumor or somatic-type malignancy in the longitudinal GTS samples, pointed at an additional occult non-seminomatous component after chemotherapy. Thus, updating the Logothetis GTS definition is necessary, which we propose as follows: The GTS describes a continuously growing teratoma that might harbor occult non-seminomatous components considerably reduced during therapy but outgrowing over time again.

Morphology-based molecular classification of spinal cord ependymomas using deep neural networks.

Based on DNA-methylation, ependymomas growing in the spinal cord comprise two major molecular types termed spinal (SP-EPN) and myxopapillary ependymomas (MPE(-A/B)), which differ with respect to their clinical features and prognosis. Due to the existing discrepancy between histomorphogical diagnoses and classification using methylation data, we asked whether deep neural networks can predict the DNA methylation class of spinal cord ependymomas from hematoxylin and eosin stained whole-slide images. Using explainable AI, we further aimed to prospectively improve the consistency of histology-based diagnoses with DNA methylation profiling by identifying and quantifying distinct morphological patterns of these molecular ependymoma types. We assembled a case series of 139 molecularly characterized spinal cord ependymomas (nMPE = 84, nSP-EPN = 55). Self-supervised and weakly-supervised neural networks were used for classification. We employed attention analysis and supervised machine-learning methods for the discovery and quantification of morphological features and their correlation to the diagnoses of experienced neuropathologists. Our best performing model predicted the DNA methylation class with 98% test accuracy and used self-supervised learning to outperform pretrained encoder-networks (86% test accuracy). In contrast, the diagnoses of neuropathologists matched the DNA methylation class in only 83% of cases. Domain-adaptation techniques improved model generalization to an external validation cohort by up to 22%. Statistically significant morphological features were identified per molecular type and quantitatively correlated to human diagnoses. The approach was extended to recently defined subtypes of myxopapillary ependymomas (MPE-(A/B), 80% test accuracy). In summary, we demonstrated the accurate prediction of the DNA methylation class of spinal cord ependymomas (SP-EPN, MPE(-A/B)) using hematoxylin and eosin stained whole-slide images. Our approach may prospectively serve as a supplementary resource for integrated diagnostics and may even help to establish a standardized, high-quality level of histology-based diagnostics across institutions-in particular in low-income countries, where expensive DNA-methylation analyses may not be readily available.