This evaluation of our molecular data confirms the great value of glioma targeted NGS for routine brain tumor diagnostics. Our Ion Torrent based NGS panel allows simultaneous assessment of a number of mutations and CNA that are relevant for gliomas. We introduced this platform into routine diagnostics once a validation series showed an improved classification and prognostication of the NGS panel in comparison to classical histopathology, which finding has been confirmed by other series [10, 19, 22]. The improved classification of gliomas using molecular criteria resulted in to the major 2016 revision of the WHO criteria for brain tumors . The current study on a large cohort of patients underscores the diagnostic strength of the platform we developed, with a clear separation of glioma subgroups with different outcomes consistent with our findings in the earlier study on EORTC study 26,951 .
Although results of NGS in diffuse glioma have been published before, none of these other studies showed data on the diagnostic value of the NGS panel in cases in which no diagnosis could be established based on the histological findings, nor did they include survival information. Several studies found a high sensitivity of NGS panels to detect genetic alterations known to be present by conventional techniques [17, 19, 22]. Zacher et al. also used a 20-gene panel for an integrated histological and molecular diagnosis of 111 diffuse gliomas, allowing reclassification of oligoastrocytoma and glioblastoma by IDH-status and identification of tumors with H3F3A mutations . Ballester et al. used a more extensive NGS panel (46–50 genes) in 381 brain tumors and found that the most clinically relevant genes for brain tumor classification in their panel were IDH1, IDH2, TP53, PIK3CA, BRAF, EGFR, PDGFRA and FGFR1/2/3. Sahm et al.  found that information derived from their NGS protocol identified potential targets for experimental therapy (i.e. EGFR, BRAF, PTEN in 37/47 (79%) glioblastomas, 9/10 (90%) pilocytic astrocytomas, and 5/14 (36%) medulloblastomas in a prospective cohort (n = 71) . This is in line with our results, although it is fair to say that at present only BRAF V600E mutations represent a validated target for precision medicine (Table 3).
A major advantage of a glioma-targeted NGS approach like this panel is the simultaneous detection of several markers relevant for glioma diagnostics, including copy number alterations, allowing glioma diagnostics according to the revised WHO 2016 classification. These can each be individually assessed by other tests (eg, immunohistochemistry, FISH, Sanger sequencing) and then usually carried out sequentially but that makes the diagnostic process more time consuming. Although NGS may be a relatively expensive diagnostic method, it yields with one assay information that otherwise would require several tests. Moreover, the costs of NGS are rapidly decreasing making it more affordable. Also, the test can be done on very limited amounts of tissue (minimal requirement is 1 ng of DNA from approximately 150 cells consisting of at least 30% neoplastic cells), independent of the method by which the tissue has been obtained; ie. resection, biopsies or even cytology.
The routine assessment of potentially actionable mutations that may have implications for prognosis and treatment is another argument for the routine use of NGS in glioma. BRAF mutations are in particular interesting considering their potential treatment implications. Although BRAF mutations are no part of the WHO classification and not tumor specific, they are present in certain glioma subtypes with an increased rate and have potentially clinical implications since drugs are available that are active against some BRAF mutations (in particular the BRAF V600E mutation). We have been taken by surprise in several cases where despite a histological diagnosis without an increased likelihood of a BRAF mutation a BRAF V600E mutation was identified. Another major advantage is the further classification of grade II and III IDHwt astrocytomas, of which some have molecular features that allow them to be classified as glioblastoma, holding prognostic and treatment implications [2, 14]. The recently published 3rd paper of the cIMPACT-NOW committee for the integration of new information into the classification of brain tumors now recommends to classify IDHwt grade II or III astrocytoma with either high level EGFR amplification, or whole chromosome 7 gain in combination with whole chromosome 10 loss, or TERT promoter mutation as ‘diffuse astrocytic glioma, IDH-wildtype, with molecular features of glioblastoma, WHO grade IV’ . This is similar to the criteria we used for the diagnosis ‘molecular glioblastoma’. Vice versa, our data confirm the clear difference in prognosis between IDHmt and IDHwt glioblastoma. The median OS in our IDHmt glioblastoma possibly reflects a bias towards performing NGS in glioblastoma patients with an unusual long survival. There is also evidence suggesting homozygous deletion of CDKN2A/B identifies poor prognosis IDHmt astrocytoma . This finding and the 3rd cIMPACT-NOW report underscore the diagnostic importance of the routine use of a panel that simultaneously assesses glioma relevant mutations and more copy number alterations than only 1p and 19q.
The routine use of an NGS panel still requires a critical evaluation of the clinical, radiological features and histopathological findings in the case under consideration. In our series, a mistake almost made was the diagnosis of a molecular glioblastoma in a fossa posterior tumor in a young patient with histological characteristics of a medulloblastoma (7+/10-, TERT +). These alterations can however also be found in medulloblastoma. Also, independent of the technique used there is always the possibility of inconclusive findings. In a few cases indeed no molecular diagnosis was obtained. In most of these cases rare brain tumors without a typical molecular profile were diagnosed histopathologically. The opposite also happened: cases in which the histopathology remained inconclusive or failed to identify tumor whereas the NGS panel yielded a clear diagnosis. The most impressive experiences were of course the 6 cases in which the pathologist was unable to positively identify tumor but in which a very characteristic mutation and/or CNA pattern associated with glioma was identified, even from biopsy samples. This has of course a major clinical impact for patients.
Early 2017 the platform has been revised and expanded, to allow detection of mutations in de TERT promoter, in genes important for pediatric brain tumors and other adult non-glioma brain tumors and more CNA’s (including 9p, 17) that are relevant for pediatric, adolescents and young adults. Clearly, platforms like this are a moving target, and require the reconsideration of their design with new information being reported. The diagnostic specificity also depends on the specificity of the mutations and CNA’s identified within the histological context (eg. H3F3A K27 M mutations have now also been identified in cases of less aggressive circumscript fossa posterior lesions , BRAF mutations are not specific for a diagnostic category).
NGS is primarily aiming at mutations and allows simultaneous assessment of copy number alterations or of fusion genes, depending on the used technology. There is an increasing interest in the use of DNA genome wide methylation based classification of central nervous system tumors, which diagnostic sensitivity and clinical usefulness has been demonstrated in a recent series . Each of those more broad molecular diagnostic panels have the major advantage of assessing more than one molecular feature, resulting in more in depth diagnostics. Obviously, any new version of these diagnostic panels needs to be well validated before it is introduced into routine clinical diagnostics. Ideally, this requires the close collaboration of pathologists, molecular biologists and clinicians at all stages of that process.
Limitations of the study are the testing of selected patients, in part of tertiary referrals and on clinical indications (eg, screening for trials targeted trails with targeted agents, long term glioblastoma survivors). Also, germline DNA was not investigated, which is less of an issue in case of targeted NGS but still requires the distinction between DNA variants without clinical significance and tumorigenic mutations. Next, the criteria for molecular glioblastoma are not required by the WHO 2016 classification schema to call a glioblastoma, but were used by us to have positive molecular criteria for glioblastoma, and in some cases of histological glioblastoma these were not found. Of note, the c-IMPACT-NOW 3rd update proposes the exact same criteria for ‘molecular features of glioblastoma’. Also, typical molecular abnormalities of some entities are not covered by our panel (e.g., fusion genes like RELA fusion genes, relevant for supratentorial ependymoma, BRAF-KIAA fusion gene relevant for pilocytic astrocytoma, FGFR fusion genes, potentially targetable). At the period studied TERT promoter mutations could not be assessed with our panel, but testing for this mutation has been added to the 2017 revised version of the panel. Then, of some entities characteristic mutations are not yet identified, for these methylation analysis may be better suited (e.g., posterior fossa ependymoma).