Practical implementation of DNA methylation and copy-number-based CNS tumor diagnostics: the Heidelberg experience.
Capper D, Stichel D, Sahm F, Jones DTW, Schrimpf D, Sill M, Schmid S, Hovestadt V, Reuss DE, Koelsche C, Reinhardt A, Wefers AK, Huang K, Sievers P, Ebrahimi A, Schöler A, Teichmann D, Koch A, Hänggi D, Unterberg A, Platten M, Wick W, Witt O, Milde T, Korshunov A, Pfister SM, von Deimling A. · 2018
Read Practical implementation of DNA methylation and copy-number-based CNS tumor diagnostics: the Heidelberg experience. by Capper D, Stichel D, Sahm F, Jones DTW, Schrimpf D, Sill M, Schmid S, Hovestadt V, Reuss DE, Koelsche C, Reinhardt A, Wefers AK, Huang K, Sievers P, Ebrahimi A, Schöler A, Teichmann D, Koch A, Hänggi D, Unterberg A, Platten M, Wick W, Witt O, Milde T, Korshunov A, Pfister SM, von Deimling A. (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6060790.txt.
Introduction
Capper and co-authors describe how Heidelberg put DNA methylation testing into brain tumor diagnosis. Their report appeared in Acta Neuropathologica. A machine learning classifier sorts central nervous system tumors by their methylation pattern. The center ran the test on more than 2,000 tumors between 2015 and 2018.
About half came from local neurosurgery departments. The rest arrived from outside institutions. The authors explain how results merge with classic microscope findings. They give a catalog of copy-number changes for each tumor class. Hard cases where methods disagree get special attention. Pitfalls and practical fixes fill much of the text. Readers get a working model for molecular tumor boards.
Catalog record
| CREATOR | Capper D, Stichel D, Sahm F, Jones DTW, Schrimpf D, Sill M, Schmid S, Hovestadt V, Reuss DE, Koelsche C, Reinhardt A, Wefers AK, Huang K, Sievers P, Ebrahimi A, Schöler A, Teichmann D, Koch A, Hänggi D, Unterberg A, Platten M, Wick W, Witt O, Milde T, Korshunov A, Pfister SM, von Deimling A. |
|---|---|
| YEAR | 2018 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 111 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Practical implementation of DNA methylation and copy-number-based CNS tumor diagnostics: the Heidelberg experience By David Capper, Damian Stichel, Felix Sahm, David T. W. Jones, Daniel Schrimpf, Martin Sill, Simone Schmid, Volker Hovestadt, David E. Reuss, Christian Koelsche, Annekathrin Reinhardt, Annika K. Wefers Published in: Acta Neuropathologica DOI: 10.1007/s00401-018-1879-y ABSTRACT Recently, we described a machine learning approach for classification of central nervous system tumors based on the analysis of genome-wide DNA methylation patterns [6]. Here, we report on DNA methylation-based central nervous system (CNS) tumor diagnostics conducted in our institution between the years 2015 and 2018. In this period, more than 1000 tumors from the neurosurgical departments in Heidelberg and Mannheim and more than 1000 tumors referred from external institutions were subjected to DNA methylation analysis for diagnostic purposes. We describe our current approach to the integrated diagnosis of CNS tumors with a focus on constellations with conflicts between morphological and molecular genetic findings. We further describe the benefit of integrating DNA copy-number alterations into diagnostic considerations and provide a catalog of copy-number changes for individual DNA methylation classes. We also point to several pitfalls accompanying the diagnostic implementation of DNA methylation profiling and give practical suggestions for recurring diagnostic scenarios. Electronic supplementary material The online version of this article (10.1007/s00401-018-1879-y) contains supplementary material, which is available to authorized users. INTRODUCTION DNA methylation-based tumor classification has emerged as a promising tool to dissect tumor classes and to improve diagnostic accuracy [6, 32, 45, 46]. The underlying concept is that cellular differentiation is driven by gene expression programs dependent on transcriptional control. DNA methylation at gene promoters and enhancer regions is crucial for coordinating such programs. In cancer, the genome-wide DNA methylation pattern (the “methylome”) likely represents a combination of both the cell of origin and somatically acquired DNA methylation changes [10, 14]. Interestingly, the dynamics of the methylome crucial for cellular differentiation in non-transformed cells seem to be somewhat arrested in cancer cells and DNA methylation signatures likely remain relatively stable during the course of disease [32]. Therefore, analysis of tumoral DNA methylation patterns may represent an approximation of the DNA methylation pattern of the early transformed cell and may even be close to the patterns of the cell of origin. Analysis of DNA methylation can be performed by readily available tools. Application of this method has shown that many accepted WHO central nervous system (CNS) tumor entities can be more precisely defined by DNA methylation profiling than by morphological features: medulloblastomas are separated into four major clinically relevant sub-groups [48], ependymomas grouped by DNA methylation profiles are clinically more homogenous than what WHO classification and grading can accomplish [32], and supratentorial primitive neuroectodermal tumors (PNET) have been demonstrated to fall into very different sub-entities [45]. Consequently, the potential of methylation-based characterization has been expanded to other CNS tumor entities and the overarching concept and feasibility has just been published [6]. We introduced this technique in our routine diagnostic workup of CNS tumors in 2015. To assist in the availability of this approach for the community, we have constructed a webpage allowing the upload of methylation data and providing an analysis based on the underlying reference tumors and algorithms (www.molecularneuropathology.org). The DNA methylation-based CNS tumor classifier (the “Classifier”) does not rely on the copy-number variation (CNV) pattern of a given tumor for classification. However, constellations of losses or gains of chromosomal regions are of high diagnostic impact in some instances. For example, combined loss of 1p/19q by now is a prerequisite for diagnosing canonical oligodendroglioma [28] and the combination of Chr.7 gain, Chr.10 loss, and/or EGFR amplification is characteristic for glioblastoma [46]. We, therefore, provide an overview of CNV profiles across the CNS tumor entities currently included in the Classifier and how these profiles might be utilized for diagnostic decisions. Currently, DNA methylation-based CNS tumor classification is being introduced in several diagnostic institutions. Here we describe the experience with this approach in Heidelberg and how it is implemented within the recently introduced WHO concept of an “integrated diagnosis”. We deliberately give a focus on constellations with conflicting morphological and molecular findings or other confusing aspects. Since, for many of these constellations, a sound scientific basis of how to best proceed is not yet available, we formulate suggestions as to how we go about these situations on a working basis. We expect that many of these suggestions will be replaced by more sound recommendations in the future. MATERIALS AND METHODS DNA methylation-based tumor analysis—the Classifier and copy-number analysis All methylation data were generated using the Illumina HumanMethylation450 (450k) or MethylationEPIC (850k) array platforms. Unsupervised clustering of methylation array data from a previously established reference set of unambiguously diagnosed CNS tumors (“the reference cohort”) revealed clear formation of separate tumor groups. The “Classifier” tool was developed based on a random forest algorithm. The current version v11b4 of the Classifier is based on the analysis of 10,000 CpG sites present on both the 450k and the EPIC arrays. An in-depth description on the theoretical background and development of this tool is given elsewhere [6]. The error rate of the Classifier calculated by a cross-validation analysis of the reference cohort was estimated to be approximately 1%. However, for statistical reasons in this cross validation, the class prediction was defined as “class with highest score” and did not include the threshold of ≥ 0.9 as an additional requirement for a correct prediction (for further information on the threshold see below). Therefore, we assume that many of the cases constituting the 1% “technical” error rate would actually resolve as cases with a calibrated score below ≥ 0.9, and we have not yet identified cases with such a technical error scoring in our routine workup. The same set of data generated by employing the Illumina 450k or Illumina 850k/EPIC arrays can be used to calculate copy-number alterations using the ‘conumee’ package for R (http://bioconductor.org/packages/conumee). Patients More than 1000 specimens from patients of the University Hospitals of Heidelberg and Mannheim operated between 2015 and 2018 have been subjected to DNA methylation analysis. There was a bias for rare tumor entities, diagnostically difficult cases and pediatric patients. In addition, more than 1000 referrals from outside hospitals have undergone the same procedure. Sample preparation For methylation analysis, we aimed at a tumor cell content of 70% or more. Both fresh frozen and FFPE samples were included. Fresh frozen tissue was examined by frozen section to ensure sufficient tumor cell content. Shaves were then taken and processed according to protocols for the Invisorb Genomic DNA Kit II (Stratec Molecular, Berlin). DNA was dissolved in ddH2O to a concentration of approximately 25 ng/µl. When using FFPE tissue, suitable regions on the H&E section were marked, and for most cases, a punch of 2 mm diameter and up to 3 mm depth was taken from the corresponding position of the paraffin block. For the remaining cases, unstained sections were cut and the corresponding tumor areas were macrodissected. Following deparaffinization, DNA was extracted with the Maxwell® 16FFPE Plus LEV DNA Kit (Promega, Fitchburg, USA) and a Maxwell DNA extractor. In our experience, old age alone of the FFPE material did not preclude extraction of suitable DNA. More important appears to be the quality and composition