DNA methylation profiling in central nervous system tumors: Where do we draw the line of clinical utility?

Abstract

DNA methylation profiling has emerged as a transformative tool in the epigenetic classification and diagnosis of central nervous system (CNS) tumors, offering unprecedented resolution in distinguishing morphologically overlapping entities. The advent of methylation-based classifiers has refined the diagnostic landscape, enabling improved tumor stratification, prognostication, and even therapeutic decision-making. In CNS malignancies, DNA methylation profiling offers tumor-specific epigenetic “fingerprints” that improve diagnostic accuracy and repeatability. It improves the classification of glioneuronal tumors, ependymomas, and juvenile gliomas, eliminates 9%-25% of histopathological discrepancies, improves grading in up to 18% of instances, and achieves > 70% concordance with integrated histo-molecular diagnosis. However, its application in routine neuropathology practice raises critical questions regarding necessity, accessibility, and cost-effectiveness. While methylation signatures have proven indispensable in diagnostically ambiguous or rare cases, their universal application to all CNS tumors remains debatable, especially in settings where robust histopathology, immunohistochemistry, and targeted molecular assays already provide reliable diagnostic accuracy. This editorial explores the strengths and limitations of DNA methylation profiling, examining its role as a complementary vs mandatory tool in routine practice. We argue for a balanced approach, where methylation is deployed selectively, prioritized for diagnostically challenging, morphologically ambiguous tumors or in the context of clinical trials rather than applied indiscriminately. Such an evidence-driven framework may optimize resource utilization while ensuring diagnostic precision in CNS tumor pathology.

Key Words: DNA methylation profiling; Central nervous system tumors; Epigenetic classification; Prognostification; Diagnostic precision

Core Tip: DNA methylation profiling has revolutionized central nervous system tumor classification by providing objective, reproducible, and prognostically relevant molecular signatures. However, its integration into routine neuropathology practice requires judicious application. This editorial highlights the need to balance diagnostic precision with practical limitations such as cost, accessibility, and turnaround time. We emphasize that methylation profiling should complement, not replace, conventional histopathology, immunohistochemistry, and targeted molecular assays, reserved for diagnostically ambiguous or high-impact cases where it can meaningfully influence patient management.

INTRODUCTION

An emerging area in neuro-oncology, using objective biomarkers to connect structure, function, and result, is highlighted in the new paper by Parija et al[1]. The field of molecular neuropathology, where DNA methylation profiling has revised tumor categorization throughout the central nervous system (CNS), exhibits a similar translational zeal. However, despite its diagnostic accuracy, there remains disagreement over whether methylation profiling is necessary in every case or if it is going beyond what is clinically required.

The diagnosis and classification of CNS disorders have undergone a remarkable transformation with the advent of DNA methylation profiling[2]. With the introduction of newly identified molecularly defined entities exhibiting novel characteristics, the subclassification of known entities based on prognostic value, and the resolution of ambiguity in challenging cases, methylation profiling has made significant progress, providing critical insights into the classification of CNS tumors[3]. However, its clinical deployment raises questions about boundaries, challenges, and actual value in routine neuro-oncologic practice. This editorial critically evaluates DNA methylation profiling in CNS tumors, tracing its development, highlighting its utility and limitations, and debating where clinicians should draw the line for its optimal use.

This editorial draws on published evidence from major validation studies and guidelines to evaluate the role of methylation profiling.

PROMISE OF DNA METHYLATION PROFILING

DNA methylation is currently at the forefront of molecular diagnostics in CNS tumor classification, which has shifted the concept of purely morphological diagnosis to an epigenetic signature (methylation pattern) of the tumor[3]. As a genome-wide, high-throughput technique, DNA methylation can distinguish between different types of tumors, define novel subtypes, resolve diagnostic ambiguity, and aid in the prognostication of tumors[4,5]. The development of the Deutsches Krebsforschungszentrum (DKFZ) classifier by the German Cancer Research Centre, in collaboration with Heidelberg University Hospital, has enabled widespread acceptance globally for refining diagnoses and updating them with the latest research in the CNS, as outlined in the World Health Organization (WHO) 5^th edition[2,6].

CLINICAL UTILITY: WHAT DOES THE EVIDENCE SAY?

Methylation profiling provides a tumor type-specific epigenetic signature, lending objectivity and reproducibility to CNS tumor classification. The discordance in CNS tumor diagnosis at the academic centers is 12% and at the community setting, it is even higher, which amounts to 26%[4]. DNA methylation identified a diagnostic mismatch in 14% of cases, with histologically recognized instances described in the WHO 2021 guidelines. In 5% of cases, no match was found, and 17% of cases were given a descriptive diagnosis[4]. In previous retrospective series, 12%-17% cases misdiagnosed were correctly identified by DNA methylation[4].

This diagnostic inaccuracy can be improved by DNA methylation studies, which provide tumor-specific “fingerprints” that are essentially the epigenetic signature[7]. Extensive studies have shown that the concordance rate of DNA methylation with histopathology is 74% and the discordance rate is 14%[4]. In 14% of cases, the resolution was achieved through DNA methylation, thereby averting the histopathological misdiagnosis. Prognostic refinement was done in 7% of cases. Among the cases where no histopathological consensus was reached, DNA methylation resolved 86% of those[4]. Grade was modified in 18% of the cases (upgrade or downgrade). Among the tumor-specific diagnoses, astrocytoma isocitrate dehydrogenase (IDH)-mutant were among the most misclassified tumors (31%) followed by low-grade glial/glioneuronal tumors (27%) on histology, followed by ependymomas (25%), and glioblastoma (13%)[4]. Other studies have also shown that DNA methylation changes can alter histopathological diagnosis in 9.8%-25% of cases[8]. This percentage is significant, and a layered approach that combines histopathology, immunohistochemistry, targeted molecular study, and methylation is necessary in “difficult-to-classify” cases.

As previously mentioned, the diagnostic concordance of CNS tumors with histopathology, immunohistochemistry, fluorescence in situ hybridization, and next generation sequencing, along with methylation profiling, is above 70%[4,9]. The discrepancy seen may vary from 9% to 25%[4,8]. The main reason for the discrepancy lies in overlapping histological features. These discrepancies were apparent in pediatric CNS tumors. The most frequent discrepancies seen in histologically low-grade glioneuroal tumors, which, on methylation, may be grouped under specific glioneuronal tumors with distinct molecular features, or glioblastoma, IDH wild-type, etc[5,10]. Ependymomas are given a particular class on methylation like supratentorial ependymoma, zinc finger translocation associated-fusion positive; posterior fossa ependymoma group A/B; spinal ependymoma, v-myc avian myelocytomatosis viral related oncogene, neuroblastoma derived amplified, etc. However, sometimes the supratentorial ependymomas are classified as posterior fossa ependymomas. Many IDH wild-type glioblastoma are misclassified as ganglioglioma on the methylation classifier[5,10].

The Galbraith et al[4] cohort consisted of 79% internal cases and 21% consultation cases, whereas Wu et al[8] had all consultation cases. This might inflate the discordance rate to 25%. Galbraith et al[4] reported a cohort with 28% pediatric cases, where the descriptive diagnosis was solved in 91%-92% of cases, which is very significant. Wu et al[8] found that pediatric tumors like medulloblastoma and ependymoma cases were refined by methylation rather than changing the complete diagnosis. Tumor purity is the most significant determinant of methylation score, as low purity may cause no match[8]. Galbraith et al[4] had used a methylation score > 0.9 for “high confidence” while Wu et al[8] accepted > 0.84 for “high confidence”, which had apparently increased the success rate of diagnosis (66%).

The authors have argued that methylation avoids the delay of running multiple sequential IHC and NGS tests. Conversely, the risk of delay from methylation itself (e.g., waiting for a “no match” result) is acknowledged as a workflow challenge but not quantified. Wu et al[8] have shown that a significant number of low-scoring cases (41.6%) required additional methods to reach a final diagnosis, which may impair the fast-track diagnosis process and cause delay. They argue that this should be kept in an integrated diagnosis workflow.

DISCOVERY OF NEW ENTITIES AND SUBTYPES

A profound impact of DNA methylation profiling is the identification of novel subtypes, completely changing the approach to histologically rare or ambiguous cases. This refined methylation classification of CNS tumors yields numerous novel entities, which are included in the CNS WHO 5^th edition, that cannot be confidently diagnosed solely based on histopathology[11]. Among the methylation classes, novel entities include: Diffuse pediatric type high-grade glioma, H3-wild type, IDH-wild type; CNS Neuroblastoma, FOXR2-activated; CNS embryonal tumor with BCL6 corepressor (BCOR) internal tandem duplication; high-grade astrocytoma with piloid features; diffuse glioneuronal tumor with oligodendroglioma-like features and nuclear clusters, etc. Without requiring a time-consuming rollout of new assays, the sector may swiftly adjust to scientific advancements by retraining classifiers on fresh data[12-17].

PRACTICAL LIMITATIONS: ACCESSIBILITY, COST, AND TURNAROUND TIME

Despite its diagnostic value, the DNA methylation study faces logistical hurdles. Some of the necessary elements for the smooth running of the DNA methylation set include highly specialized equipment, expensive chemicals for testing, skilled bioinformatics staff, and turnaround times of 7-8 working days per batch (typically eight samples)[10,18]. For a basic methylation research setup, a cost of around $75000 to $150000 is necessary[19]. The main cost lies in the scanner/platform (e.g., refurbished Illumina iScan system with additional computing cost.

For advanced, fully automated, and premium facilities, the price may exceed 1 million United State Dollar[19]. New iScan systems are more expensive than refurbished ones. Automation requitred $50000-$100000. Computing and software require $50000-$200000+. For accreditation and validation, an additional $12000-$50000 is needed for consumables.

Accessibility is primarily limited to tertiary institutions and cooperative networks, and the cost per sample remains high. These obstacles may prevent broad adoption in environments with limited resources or postpone important treatment choices, especially for high-grade malignancies that require prompt attention.

UNCLASSIFIABLE CASES AND LOW-SCORE RESULTS

Not every CNS tumor produces conclusive matches for methylation classifiers. Studies suggest that 6%-17% of samples remain unclassifiable, most notably in situations with low tumor purity, inadequate DNA input, or odd age groups (young children/adolescents)[10]. An ideal classifier score is greater than 0.9; however, sensitivity and specificity are balanced by a classifier score threshold (e.g., 0.84 and above). Values below 0.3 indicate “no match”, which frequently delays adjuvant therapy and prolongs the time to diagnosis[20]. These unclassifiable tumors may be the result of biologically unique entities not included in existing databases or technical limitations, and they may be associated with a lower survival rate. These groups of unclassified cases with low scores may be best classified based on histopathology, supplemented by FISH, immunohistochemistry, or NGS as necessary.

INTEGRATING METHYLATION DATA INTO CLINICAL WORKFLOW

DNA methylation results should never be interpreted in isolation for the best outcomes. The integration of methylation profiles with the complete clinical context, including histology, immunoprofiles, targeted sequencing, and clinical data, is strongly encouraged by the consortium recommendations (cIMPACT-NOW which stands for the consortium to inform molecular and practical approaches to CNS tumor taxonomy - now)[10]. By anchoring diagnoses in the established clinical workflow, this comprehensive “layered” diagnosis allows doctors to leverage epigenetic data meaningfully. Additionally, instead of being used routinely for all malignancies, it enables the careful selection of cases for methylation study, concentrating on those with genuine diagnostic, prognostic, or therapeutic ambiguity. Similarly, while methylation classes have been defined for meningiomas and pituitary adenomas, their clinical testing is generally unnecessary for standard diagnosis, though it may provide improved risk stratification in histologically ambiguous or high-risk cases[10].

For patients above 55 years of age with older age-related characteristics, high-grade glioma histology, endothelial proliferation, and necrosis, and an IDH R132H mutation that is wild-type, methylation is not required[21]. IDH mutant astrocytomas generally do not need methylation unless prognostication is required[22]. IDH mutant oligodendrogliomas also typically do not require methylation studies; however, extensive research has revealed false-positive results of FISH for 1p/19q co-deletion, leading to the reclassification of this subtype[4]. Not all pediatric high-grade gliomas require methylation studies. Some surrogate markers, such as H3K27M and H3G34R, can diagnose specific types of pediatric-type high-grade glioma without requiring methylation[23]. Pleomorphic xanthoastrocytoma and diffuse pediatric type high-grade glioma, H3-wild type, IDH-wild type, sometimes show overlapping features, and it is crucial for methylation studies of these cases[13]. Low-grade glioneuronal tumors, especially in adults, are the strongest candidates for methylation studies. Ependymomas are moderate contenders for methylation classification. Among the embryonal tumors, medulloblastomas usually do not need methylation. New entities, such as CNS neuroblastoma, FOXR2-activated & CNS embryonal tumor with BCOR internal tandem duplication, are classified using a methylation classifier. Newly diagnosed embryonal tumors with BRD4: LEUTX fusion or pleomorphic adenoma gene 1 amplification require methylation or NGS studies at a minimum[22-26]. Pineal tumors may require methylation, but not always[12,27]. Mesenchymal tumors, such as solitary fibrous tumor, Ewing sarcoma, and capicua transcriptional repressor-rearranged sarcoma, can be classified based on immunohistochemistry. Undifferentiated/unclassified sarcomas may require methylation or NGS study[11,27]. Since methylation diagnosis can take time, it is to be ensured that standard patient-of-care is not delayed to this. Treatment should be started based on initial (H&E + IHC + molecular) diagnosis. Wu et al[8] have shown that 90% of glioblastoma, IDH-wild type cases were confirmed as it was.

A “reflex testing pathway” should always be pre-planned for the methylation “intermediate score” (0.3-0.84 or 0.3-0.9) cases so that time is not wasted to think what to do next. t-SNE/UMAP analysis or Targeted sequencing are the safe methods considering morphology of the cases.

Galbraith et al[4] argue that while methylation has a cost, it is “significantly lower” than large NGS panels and decreases “tissue waste” associated with sequential testing. A cost-effectiveness analysis would compare the single cost of methylation against the cumulative cost of multiple IHC stains + FISH + NGS panels. Since methylation diagnosis can take time, it is essential to ensure that the standard patient care is not delayed as a result. Treatment should be initiated based on the initial (H&E, IHC, and molecular) diagnosis. Wu et al[8] have shown that 90% of Glioblastoma IDH-wild-type cases were confirmed as they were.

TECHNICAL ADVANCEMENTS AND FUTURE DIRECTIONS

The classifier is undergoing refinements in classifier algorithms, with expanding reference sets, and enhancements in specificity for rare and pediatric entities[28]. Classifier V12.8, for example, aims to capture the biological diversity of CNS malignancies better and reduce unclassifiable rates. Global reach is being expanded by adaptations for tiny biopsies, fornalin-fixed-paraffin-embedded material, and rapid-turnaround techniques (including commercial services like MedGenome in India).

WHERE SHOULD WE DRAW THE CLINICAL LINE?

The growing use of DNA methylation profiling in CNS tumors is promising; however, careful consideration of the benefits and risks should accompany its application. It offers better diagnostic, prognostic, and predictive capabilities as a strong molecular platform, but its use should be prudent rather than careless. The clinical line should be drawn with consideration for several factors.

Diagnostic uncertainty

DNA methylation profiling should be reserved for cases where histopathology, immunohistochemistry, FISH, and NGS studies do not reveal a clear answer, or where diagnosis directly impacts management[7,10]. One of the most critical factors is grading, where there is a chance of upgrading or downgrading of tumors[4].

Resource allocation

The urgency of the treatment should be weighed against cost and turnaround time. Most tumors can be diagnosed with histology and immunohistochemistry much earlier than with a methylation classifier.

Tumor purity and sampling

Be aware of the limitations in situations where tumor density is low or DNA yield is insufficient; the results may not be accurate or informative in these situations[6,10].

Clinical impact

Use DNA methylation results to refine therapy only when clear evidence demonstrates a change in risk categorization, treatment, or surveillance.

CLINICAL PARADIGM: METHYLATION AS A LAYER, NOT A REPLACEMENT

DNA methylation profiling has become a crucial component of the diagnostic paradigm for CNS tumors, particularly for cases with diagnostic ambiguity. It enhances rather than replaces focused molecular testing, immunohistochemistry, and traditional histology. It can resolve ambiguity, uncover new subtypes, and advance precision medicine, but its use requires strong clinical and financial judgment.

ETHICAL AND PRACTICAL IMPLICATIONS

As the use of methylation platforms grows, ethical issues about equality, patient access, and the fallout from unclassifiable diagnoses need to be taken into account[29]. Essential elements of safe use include multidisciplinary case reviews, transparent classifier ratings, standardized reporting, and patient counseling.

By combining molecular and histological data, the integration of methylation classifiers with machine learning improves the accuracy of CNS tumor detection. Artificial intelligence (AI) developments are enhancing classifier accuracy and interpretability, promoting more equitable diagnosis internationally, despite ethical discrepancies arising from restricted access in low-resource settings[30].

PATH FORWARD

In the field of CNS tumor diagnoses, DNA methylation profiling is a game-changer, blurring the boundaries between pathology, genetics, and clinical neuro-oncology. Clinicians must, however, carefully consider when and how to use it, saving its advantages for situations in which it clearly enhances patient outcomes. The “line” of clinical use will change as technology advances. However, the fundamental idea remains the same: Methylation profiling is a powerful tool whose usefulness is enhanced when integrated into a multi-layered, patient-centered, and context-rich diagnostic framework.

The future also promises to integrate a DNA methylation classifier with machine learning data to accurately classify CNS tumors for diagnostic, prognostic, and therapeutic purposes. Recent studies have shown that leveraging deep learning to connect digital pathology images with a DNA methylation classifier is a promising tool for improving diagnostic precision[31,32]. The newly introduced deep learning models, such as DEPLOY, can predict DNA methylation states directly from histopathology glass slides with excellent accuracy and precision[31]. These AI models can reduce turnaround time and reliance on costly and capital-intensive molecular assays.

When combined with pathologist consensus, methylation classifiers like SNUH-MC (Seoul National University Hospital Methylation Classifier) and DKFZ-MC (DKFZ Methylation Classifier) have demonstrated enhanced concordance, benefiting from cross-validation using imaging, molecular, and clinical characteristics[33].

The argument: If AI models (like DEPLOY) can predict methylation classes from cheap H&E slides, they can act as a “poor man’s methylation”.

This allows low-resource settings to use AI as a screen. If the AI is highly confident, they might skip the expensive physical methylation test. If the AI is unsure, then they cross the “line” and pay for the expensive test.

CONCLUSION

DNA methylation profiling is a significant advancement in the diagnosis of CNS tumors; however, its application should be driven more by clinical necessity than by technological zeal. It should be used as a supplementary, case-selective modality, reserved for cancers that are difficult to diagnose, have overlapping histology, or have an unknown prognosis, as it can have a direct impact on treatment choices. Without contextual support, over-reliance runs the risk of increasing expenses and growing access gaps. The future of neuropathology practice will depend on a balanced approach to CNS tumors, a hierarchical structure, and multidisciplinary tumor board meetings with neurosurgeons, radiation oncologists, and medical oncologists, ensuring that every test carried out actually improves patient care rather than merely increasing molecular data.

ACKNOWLEDGEMENTS

I extend my sincere gratitude to Swati Singh, PhD Scholar in Neuropathology at AIIMS, New Delhi, for her invaluable support in enhancing the quality of this manuscript.