Molecular GBM versus Histopathological GBM: Radiology-Pathology-Genetic Correlation and the New WHO 2021 Definition of Glioblastoma

DISCUSSION

The WHO CNS tumor classification represents the primary source of updates on diagnostic classes, grades, and criteria. The characterization and grading of CNS tumors has conventionally been on the basis of histopathology. It has now been established that such a classification scheme is not an accurate predictor of a tumor’s biologic behavior. Molecular and genetic markers provide more information with lower likelihood of undersampling, even in small biopsy specimens. Molecular information provides higher diagnostic accuracy, offering more precise prognosis and optimized treatment options, forming important elements of personalized medicine, and eventually resulting in better outcome.^2,3 Molecular information in CNS tumor classification was for the first time formally incorporated in the definition of these tumors in 2016. The 2016 WHO classification represented an update of the fourth edition of WHO classification (2007) rather than being the formal fifth edition. The 2016 CNS WHO divided GBM into 1) GBM, IDH-wild-type (approximately 90% of cases), corresponding most frequently with the clinically defined “primary” or de novo GBM (>55 years of age); 2) GBM, IDH-mutant (approximately 10% of cases), which corresponds closely to so called “secondary” GBM with a history of prior lower-grade diffuse glioma (younger age-group), and 3) GBM, NOS where IDH evaluation could not be performed. GBM was previously classified as “glioblastoma multiforme” where “multiforme” referred to the highly variable histopathological features. Given the emphasis on molecular (IDH) markers in the revised fourth edition (2016) of WHO classification, the term “multiforme” was removed and these tumors were classified simply as “glioblastoma.”⁴ Since 2016, ongoing discoveries in molecular pathology have led to better understanding and further refinement of the classification. Given the importance of the molecular markers, an international group of leading neuropathologists (all directly involved in establishing the WHO 2016 classification) formed the cIMPACT-NOW to provide updates before the release of the next WHO classification scheme. One of the major updates by the consortium (cIMPACT-NOW) was identification of molecular clues for recognizing histologically lower-grade diffuse IDH-wild-type astrocytic gliomas, behaving as GBM.

The histologic findings of necrosis and/or MVP (Fig 3C, -D) are the traditional pathognomic hallmarks for diagnosis of GBM. However, even if these pathologic changes are absent, most histologically lower-grade (WHO 2/3) diffuse IDH-wild-type gliomas in adult patients behave as WHO grade 4 lesions.^1,3 Despite the absence of necrosis and MVP, these tumors have an aggressive clinical course, with overall patient survival times not significantly different from the “histopathological” glioblastomas. This led to the proposal of new molecular criteria that could be used to upgrade these low-grade tumors. These were briefly labeled as “diffuse astrocytic glioma, IDH-wild-type, with molecular features of GBM, WHO grade IV,” by cIMPACT, however, are now simply called “glioblastoma.”^3,5 The new CNS5 WHO classification defines glioblastoma as “diffuse, astrocytic glioma that is IDH-wild-type and H3-wild-type and has one or more of the following histologic or genetic features: MVP, necrosis, TERT promoter mutation, EGFR gene amplification, +7/−10 chromosome copy-number changes (CNS WHO grade 4).”² By definition, IDH-wild-type GBM lacks immunostaining for IDH1 and is negative for H3.3 G34 mutations and H3 K27 alterations. Nuclear immunostaining for ATRX is retained in most tumors.^1,3 GBM has also traditionally been divided into primary and secondary; the former arising de novo (90%) in older patients and the latter developing from a pre-existing lower-grade tumor (10%) in a younger population.⁷ With the development of molecular criteria, these concepts have been outdated along with removal of the term “glioblastoma multiforme” in favor of glioblastoma.⁸ Diffuse astrocytoma, IDH-wild-type, CNS WHO grade2/3 without molecular or histopathological features of glioblastoma is rare and is no longer regarded as a tumor type in the new classification. If there are no histopathological or molecular features of GBM in an IDH-wild-type glioma, testing for other mutations should be done (Fig 4). This includes H3.3 G34 mutation seen in diffuse hemispheric glioma, H3.3 G34-mutant (WHO grade 4) and H3 K27 alterations seen in diffuse midline glioma (WHO grade 4). H3 G34–mutant diffuse hemispheric gliomas can be excluded by immunohistochemistry for H3.3 p.G35R (G34R) or H3.3 p.G35V (G34V) mutation or by H3-3A (H3F3A) sequencing.^1⇓-3,5

• Download figure
• Open in new tab
• Download powerpoint

FIG 4.

Chart illustration of categorization of WHO (5th) 2021 Classification of IDH-wild type diffuse astrocytic glioma. The chart highlights the role of molecular markers (TERT, EGFR, +7/−10) for diagnosis of GBM in the absence of classic histopathological findings (necrosis and MVP). The diagram also highlights the role of H3.3 G34 mutation and H3 K27 alteration in IDH-wild-type gliomas and segregation of these mutated entities from GBM.

Detailed knowledge about these mutations is beyond the scope of practice for the neuroradiologist. However, it is important to have a basic understanding of the major molecular markers, as they are frequently discussed in multidisciplinary tumor boards and correlated with radiographic patterns. MVP is one of the major histologic features of glioblastoma featuring rapid new blood vessel growth (Fig 3D) and forms the basis for the anti-vascular endothelial growth factor therapies such as bevacizumab (Avastin). The other major histologic feature of glioblastoma is necrosis. There are several postulated mechanisms by which necrosis develops in glioblastoma secondary to a microenvironment that is acidic, low in oxygen and glucose. As a result, the proliferating tumor cells migrate away from the hostile microenvironment resulting in the palisading appearance on histology (Fig 3C). IDH-wild-type gliomas have a higher incidence of intratumoral microthrombi compared to IDH-mutated gliomas due to the expression of tissue factor and podoplanin. This contributes to necrosis and an increased risk of venous thromboembolism.⁹ As per the WHO guidelines, absence of immunoreactivity for IDH1 p.R132H is sufficient to diagnose IDH-wild-type GBM in a patient aged ≥55 years, who has a histologically classic GBM not located in midline structures and no history of a pre-existing lower-grade glioma. This practical method is feasible because the probability of a noncanonical IDH mutation is <1% in GBM from patients aged ≥55 years. In patients aged <55 years, or in patients with a history of lower-grade glioma, negative IDH1 p.R132H immunostaining should be followed by DNA sequencing for less common IDH1 or IDH2 mutations. When no IDH mutations are detected by sequencing, tumors are classified as GBM, IDH-wild-type.^3,10,11 However, genetic analysis in the form of chromosomal microarray or NGS is routinely performed at major centers to better understand tumor biology and for targeted therapies. Molecular and cytogenetic testing in neuro-oncology includes NGS, chromosomal microarray, and MGMT promoter methylation status evaluation. Chromosomal microarray provides high-resolution assessment of copy number variations across the genome, is easier to use with less complicated and less labor-intensive sample preparation than NGS. NGS evaluates mutations and rearrangements in 160 genes, including most abnormalities described by the WHO with rapid drop in cost over the last few years. MGMT promoter methylation status may be evaluated by multiple methods, and the testing platform with most prospective clinical trial validation is methylation-specific polymerase chain reaction (evaluating downstream CpG dinucleotides sites).¹²

TERT promoter mutation is seen in approximately 50%–60% of adults with diffuse glioma, primarily identified in oligodendroglioma (>95%) and glioblastoma. TERT promoter mutation has been associated with more aggressive behavior in GBM, however shows better outcome in low-grade gliomas. TERT results in maintenance of telomere length and plays a major role in tumorigenesis by allowing cancer cells to repress cellular senescence. Overall, approximately 60% of patients with GBM have modification of EGFR, most frequently in the form of amplification. EGFR encodes a tyrosine kinase receptor with amplification resulting in overactivation of multiple pro-oncogenic signaling pathways.^13,14 Tumors with frequent EGFR alterations with co-occurrence of PTEN and TERT promoter mutations are more frequently multifocal.⁷ Whole chromosome 7 gain (trisomy 7) and whole chromosome 10 loss (monosomy 10) are the most frequent chromosomal copy-number aberrations in GBM, usually occur as a combination, and are commonly associated with EGFR amplification.¹⁵ Although not a part of the WHO essential diagnostic criteria for glioblastoma, MGMT promoter methylation status is listed as a desirable criterion. It is one of the leading determinants of prognosis and predictor of response to alkylating agents such as temozolomide (TMZ). MGMT promoter methylation is an independent prognostic marker for overall survival in GBM with more than 90% of longer-term surviving patients with GBM having MGMT promoter methylation/hypermethylation. Patients with MGMT promoter methylation/hypermethylation benefit more from adjuvant TMZ with more frequent pseudoprogression on imaging. Lack of MGMT promoter methylation is associated with resistance to TMZ.^16,17 Few human neoplasms are as morphologically heterogeneous as GBM, and this variability can be seen in both the molecular and histopathological categories of GBM. Histologic varieties of GBM such as giant cell glioblastoma, epithelioid glioblastoma, and gliosarcoma, were added as “variants/subtypes” in the 2016 WHO classification. Although these morphologic varieties are still acknowledged in the new CNS5 edition, these are no longer defined as “subtypes” of GBM.^1,4 For example, epithelioid GBM is a histologic subtype of glioblastoma defined by a mostly sharply demarcated aggregate of large epithelioid cells, which is prognostically a more favorable tumor of children and young adults and shows prominent genetic overlap (BRAF p.V600E mutation) with pleomorphic xanthoastrocytoma (Online Supplemental Data).

The most common imaging appearance of GBM is a heterogeneous mass centered in the supratentorial white matter with central necrosis, areas of hemorrhage with thick irregular enhancement, extensive peritumoral edema, and mass effect. Histologic features of MVP (neoangiogenesis) and necrosis are the major drivers for the these radiographics findings, and in the absence of these pathologic changes, molecular GBM may show only mild or patchy contrast enhancement or may lack central necrosis (Online Supplemental Data).^1⇓–3 In one of the recent studies by Guo et al, it has been shown that molecular GBM was diagnosed at a younger age, had higher incidence of epilepsy, and was less likely to have contrast enhancement and intratumoral necrosis on MR imaging. Importantly, despite the lower likelihood compared to their histopathological counterparts, enhancement and necrosis were observed in most molecular GBM patients (78.8% and 63.6%, respectively) (Online Supplemental Data). Also, a small proportion of patients with histologic GBM showed absence of contrast enhancement and necrosis (<5% and 15%, respectively) (Online Supplemental Data).¹ Early studies have shown that MR imaging features can predict molecular features of glioblastoma with absent histopathological changes by using a wide range of models by using distinct imaging lexicons and radiomic features. Park et al¹⁸ used the standardized feature set of Visually AcceSAble Rembrandt Images (VASARI) and radiomics extracted from the ROIs on T2WI and postcontrast T1 images to predict molecular subtype of GBMs. Studies have particularly focused on the imaging signature of EGFR mutations given the high potential for gene-based targeted therapies. Studies have shown that EGFR-mutated tumors have higher rCBV, lower ADC, higher fractional anisotropy, lower T2/FLAIR signal, and more variable spatial pattern; however, recognizing the limitation that individual assessment of these features is not sufficient to identify the mutation.¹⁹ Ivanidze et al²⁰ in a comparison of TERT promoter mutated and nonmuated GBM showed that significantly fewer TERT promoter mutated GBM had nonenhancing tumor that crossed midline. No other imaging features were significantly different between the TERT promoter mutated and nonmutated group. It is important to note that none of these studies have conclusive imaging findings to differentiate low-grade (WHO 2/3) glioma with or without molecular features of GBM. There is, however, robust ongoing research by using radiogenomics, primarily texture analysis, for prediction of these molecular changes, as targeted therapies take center stage in treatment of gliomas. Radiomics is a rapidly growing field that utilizes computational algorithms to extract macro- and micro-scale subtle subvisual cues embedded in routine MR imaging that enables distinction of molecularly distinct GBM phenotypes. Genomewide RNA expression analysis has become a routine tool in biomedical research and RNA sequencing data can be used to noninvasively predict immune subtypes of adult diffuse gliomas by using a biologically interpretable radiomic signature from MR imaging. Newer, more powerful analytical methods such as gene set enrichment analysis (GSEA) are now being used to interpret gene expression data.²¹ In recent years, several notable studies have contributed to this field utilizing advanced image analysis techniques, such as morphologic and texture analysis methods, for increasing diagnostic accuracy. Choi et al²² used RNA sequencing with GSEA to calculate the radiomics features from the contrast-enhancing tumor and peri-tumoral edema regions and classify GBM into 3 different subtypes, with distinctly different prognosis and survival outcomes. Deep convolutional neural networks are used to extract features of the MR images, which are then fed to latest state-of-the-art classifier models like the “support vector machine” with highly accurate results. Tumor microenvironment is a key target area for the next-generation treatment strategies for gliomas. Radiomics has the potential to be used as a noninvasive approach to predict tumoral immunologic microenvironment and predict response to immunotherapies.^22,23