Glioblastoma

Morphologic and Molecular Genetic Diversity

C. Ryan Miller, MD, PhD; Arie Perry, MD

Context.Glioblastoma (GBM), the most common primary intracranial malignancy, is a morphologically diverse neoplasm with dismal prognosis despite multimodality therapy. Only 3 distinct morphologic variants of GBM are currently recognized by the current World Health Organization classication scheme, including GBM, giant cell GBM, and gliosarcoma. Additional variants, some of which have signicant morphologic overlap with tumors that have more favorable prognosis and treatment response rates, particularly anaplastic oligodendroglioma, have been described since its publication in 2000 and may be included in the next classication. iffuse, inltrative gliomas are the most common primary intracranial neoplasms, accounting for 40% of all primary and 78% of all malignant central nervous system tumors.1 More than 80% of these tumors are considered high-grade (grades III and IV) when diagnosed according to the current World Health Organization (WHO) classication, a system based on morphologic evidence of differentiation along astrocytic, oligodendroglial, or mixed lineages. Classication and grading criteria for astrocytic neoplasms are well established, having been rened periodically during the last 80 years from the classication schemes of Bailey and Cushing,2 Kernohan,3 Ringertz,4 Nelson,5 St. Anne-Mayo,6,7 and WHO.8,9 The WHO 2000 criteria are currently the most widely accepted scheme and are based largely on the St. AnneMayo classication. For example, it uses mitotic activity to distinguish anaplastic astrocytoma (AA), WHO grade III, from its low-grade counterpart, diffuse astrocytoma, WHO grade II. Median overall survival (OS) for patients with AA is typically 3 to 5 years.10 Microvascular proliferation (MVP), loosely dened to include endothelial hypertrophy, endothelial hyperplasia, and glomeruloid vessels, and/or necrosis are distinguishing features of WHO grade IV astrocytoma, the latter being synonymous with the commonly used term glioblastoma multiforme or simply glioblastoma (GBM) in the current WHO scheme.10 Despite
Accepted for publication October 25, 2006. From the Division of Neuropathology, Washington University School of Medicine, St Louis, Mo. The authors have no relevant nancial interest in the products or companies described in this article. Reprints: Arie Perry, MD, Division of Neuropathology, Department of Pathology and Immunology, Washington University School of Medicine, 660 S Euclid, Campus Box 8118, St Louis, MO 63110 (e-mail: aperry@wustl.edu). Arch Pathol Lab MedVol 131, March 2007

Objective.To summarize the morphologic and molecular genetic diversity of both well-established and novel GBM variants and outline our approach to these heterogeneous neoplasms and their distinction from other diffuse, high-grade gliomas. Data Sources.Published literature and our own experience in an active academic diagnostic surgical neuropathology practice were reviewed. Conclusions.Precise subclassication of GBM is required for accurate prognostication and appropriate treatment planning. (Arch Pathol Lab Med. 2007;131:397406) advances in microsurgical techniques, radiation, and chemotherapy, GBM is still associated with a median OS of approximately 1 year.6 In its most mature and prognostically signicant manifestation, MVP is recognized as multilayered tufts of hyperplastic and mitotically active endothelial cells, accompanied by smooth muscle cells and pericytes. The presence of MVP in GBM has been correlated with hypoxia-induced neoangiogenic and inltrative phenotypes, the former of which is characterized by expression of cytokines such as vascular endothelial growth factor.11,12 Microvascular proliferation is both temporally and spatially linked to the presence of necrosis, particularly pseudopalisading necrosis.11,12 These morphologic criteria are highly reproducible and serve as useful prognostic factors for OS.1315 Glioblastoma may arise through 2 distinct pathways of neoplastic progression. Tumors that progress from lowergrade (II or III) astrocytic tumors, termed secondary or type 1 GBMs, typically display both well-differentiated and poorly differentiated foci. Secondary GBMs develop in younger patients (fth to sixth decade), with time to progression from lower-grade lesions ranging from months to decades. In contrast, primary, type 2 GBMs develop in older individuals (sixth to seventh decade), have short clinical histories (less than 3 months), and arise de novo without any evidence of a lower-grade precursor. Primary and secondary GBMs also harbor distinct molecular genetic abnormalities: Primary GBMs are characterized by relatively high frequencies of EGFR amplication, PTEN deletion, and CDKN2A (p16) loss, whereas secondary GBMs often contain TP53 mutations, especially those involving codons 248 and 273 or G:CA:T mutations at CpG sites.16 The clinicopathologic and molecular genetic features of primary and secondary GBM have been recently reviewed in detail by Ohgaki and colleagues.16
GBM VariantsMiller & Perry 397

Table 1.
Feature

Clinicopathologic and Molecular Genetic Features of Glioblastoma Multiforme (GBM) Variants*

Tumor Primary16 Secondary16 Fib Gem GCA GC32 GS43 SCA57 GBM-O57

Peak age (decade) 67 56 67 67 67 56 67 67 56 Circumscription ... ... / Nuclear irregularity ... ... / / Perinuclear haloes ... ... Cytoplasmic granules ... ... Multinucleated cells ... ... / / / / / / Reticulin meshwork ... ... / / PVI ... ... / / / / / Mucin-lled microcysts ... ... / 1p del, % ... ... ... ... ... ... ... 3 24 19q del, % ... ... ... ... ... ... 6 43 1p/19q codel, % ... ... ... ... ... ... ... 0 22 EGFR amp, % 36 8 ... ... ... 5 4 63 7 9p (CDKN2A) del, % 31 19 ... ... ... 0 37 ... ... 10q (PTEN ) del, % 70 4 ... ... ... 5 ... 89 27 EGFR amp/10q del, % ... ... ... ... ... ... ... 56 4 TP53 mut, % 28 65 ... ... ... 89 23 ... ... PTEN mut, %43 32 4 ... ... ... 27 38 ... ... * Fib indicates brillary; Gem, gemistocytic; GCA, granular cell astrocytoma; GC, giant cell; GS, gliosarcoma; SCA, small cell astrocytoma; GBM-O, GBM with oligodendroglial features; . . ., no data available; , absent; / , infrequent; to , increasing frequency; PVI, perivascular inammation; del, deletion; codel, codeletion; amp, amplication; and mut, mutation. Sources of molecular genetic frequencies are referenced by column except PTEN mutation (row).

VARIANTS OF GBM The term GBM was coined by Mallory in 191417 and solidied in the lexicon of surgical neuropathology by Bailey and Cushing in 1926.2 As the original designation multiforme implies, GBMs comprise a morphologically highly heterogeneous neoplasm. In other words, the cellular composition, even within a single tumor, can vary widely and mixed histologic features are typical.18 Three GBM variants are recognized as distinct clinicopathologic entities in the current WHO classication: conventional GBM, giant cell GBM (GC-GBM), and gliosarcoma (GS). In addition, other variants of GBM with distinct clinicopathologic and genetic features have been described (Table 1). In the present review, we discuss the cytologic heterogeneity that may be encountered in the variants of GBM and outline our approach to their subclassication. Conventional GBM Even within the conventional GBM category, the cellular composition is heterogeneous and may include brillary, gemistocytic, and/or occasional giant cells (GCs). Neoplastic brillary astrocytes contain enlarged, elongated to irregularly shaped, hyperchromatic nuclei, scant cytoplasm, and variable glial brillary acidic protein (GFAP) immunoreactive processes that form a loose, brillary matrix (Figure 1). Such tumors may also harbor mucin-rich microcystic spaces. Because of cross sectioning of spindled or cylindrical nuclei, neoplastic brillary astrocytes may also appear to have occasional rounded nuclear proles, raising the differential diagnosis of high-grade oligodendroglial (HGO) neoplasms (WHO grade III anaplastic oligodendroglioma [AO, O3] and WHO grade III anaplastic mixed oligoastrocytoma [AOA, MOA3]). Nevertheless, a signicant fraction of cells should have classic cytology (eg, round uniform cells with sharp nuclear membranes and bland chromatin) before one invokes an oligodendroglial component in the diagnosis. Gemistocytes were rst described by Franz Nissl as glia (gemaestete glia) with voluminous cytoplasm.19 Gemisto398 Arch Pathol Lab MedVol 131, March 2007

cytic astrocytes contain plump, glassy, eosinophilic GFAPimmunoreactive cytoplasm; eccentric, irregularly shaped, hyperchromatic nuclei; and short cytoplasmic processes (Figure 2, A and B). Once thought to represent a reactive component, these cells have been found to harbor TP53 mutations20 and cytogenetic abnormalities (chromosome 7p gains and 10q losses)21; therefore, they are now thought to represent a true neoplastic component. Perivascular lymphocytic inltrates are also a common but nonspecic nding in these tumors (Figure 2, A). Although gemistocytes may be found in all grades of astrocytoma (II, III, and IV), gemistocytic astrocytoma is currently recognized as a distinct variant only of diffuse astrocytoma (WHO grade II). In this context, the WHO denes a gemistocytic astrocytoma as an astrocytoma composed of more than 20% gemistocytes.10,22 The prognostic impact of gemistocytic cytology in astrocytomas has been an active area of investigation during the last 40 years.2226 Various reports have quantied the percentage of gemistocytes present in a tumor, a feature termed the gemistocytic index, and sought to correlate its signicance with patient outcomes. Gemistocytic astrocytoma (WHO grade II) with greater than 5% gemistocytes have been reported to progress more rapidly to GBM.27 Some groups have suggested that these tumors behave more aggressively than their non-gemistocytic grade II counterparts and thus deserve a grade III designation. However, the prognostic signicance of the gemistocytic index in grade II and III astrocytomas still remains unclear and no studies have examined the independent signicance of this factor in tumors with the grade IV features of MVP and/or necrosis. This work and its historical underpinnings has been summarized in a recent review by Tihan and colleagues.26 Granular cell astrocytomas are rare astrocytic neoplasms composed of sheets of large, deceptively bland cells with abundant granular, periodic acid-Schiffpositive cytoplasm (Figure 3).10,2831 Most granular cell astrocytomas are GFAP positive and may nonspecically display
GBM VariantsMiller & Perry

Figure 1. Fibrillary glioblastoma multiforme. Tumor cells contain enlarged, cigar-shaped or irregularly shaped, hyperchromatic nuclei and scant cytoplasm. This tumor also harbors pseudopalisading necrosis with peripheral palisading of tumor cells around a central core of necrosis (hematoxylin-eosin, original magnication 200). Figure 2. Gemistocytic glioblastoma multiforme. A, Neoplastic gemistocytes contain plump, glassy, eosinophilic cytoplasm; eccentric, irregularly shaped, hyperchromatic nuclei with occasional small nucleoli; and indistinct processes that form a dense, brillary background. B, Perivascular inammatory inltrates are particularly common in these tumors (hematoxylin-eosin, original magnications 400 [A] and 200 [B]). Figure 3. Granular cell glioblastoma multiforme. These tumors are composed of sheets of large cells with bland oval nuclei and abundant granular cytoplasm. These tumor cells are occasionally mistaken for macrophages (hematoxylin-eosin, original magnication 400). Figure 4. Giant cell glioblastoma multiforme. A, This tumor is composed of markedly enlarged neoplastic astrocytes with multiple, often bizarre, hyperchromatic nuclei (hematoxylin-eosin, original magnication 200). B, Fluorescence in situ hybridization analysis shows marked gains of chromosome 19 (19q13, green, and 19p13, red) in 2 apparent giant cells within this eld; other chromosomes were similarly gained, consistent with polyploidy (original magnication 1000).

cytoplasmic epithelial membrane antigen but no cytokeratin immunoreactivity.28,29 Similar to astrocytomas with nongranular cytology, these tumors may also harbor TP53 mutations, high-frequency loss of heterozygosity at 9p, 10q, and 17p, and less frequent loss of heterozygosity at 1p and 19q.30 Ultrastructural studies have shown that the cytoplasmic granularity correlates with increased lysosomes, evident as either dense or multivesicular bodies.28,29 Brat and colleagues28 reported the largest series of such tumors to date (22 cases, including 4 grade II, 7 grade III, and 11 grade IV tumors) and found that these tumors were more aggressive than non-granular cell astrocytomas of the same grade. Given their average survival of 7.6 months, it may indeed be appropriate to consider them variants of GBM. Giant Cell GBM Giant cell GBM constitutes approximately 5% of GBMs and is recognized as a distinct clinicopathologic entity in the WHO 2000 classication.10 Although occasional neoplastic GCs may be found in conventional GBM, these cells are a dominant cytologic component in GC-GBM. As the name implies, the tumor cells are markedly enlarged and bizarre, often appearing multinucleated (Figure 4, A). Giant cell GBMs are typically well-circumscribed masses that occur in younger patients (fth decade). Their molecular genetic features include relatively high frequencies of
Arch Pathol Lab MedVol 131, March 2007

TP53 mutations (75%90%) and PTEN deletion (5%30%), whereas EGFR amplication and CDKN2A deletion are rare in comparison to conventional GBMs; therefore, GCGBMs contain clinical and molecular genetic features of both primary and secondary GBMs and as such occupy an intermediate position between these two.10,32,33 Several reports have demonstrated a slightly better prognosis for GC-GBM than for conventional GBM.10,34 Nevertheless, the prognosis remains poor overall. As shown in Table 2, this has similarly been our experience at Washington University School of Medicine. Among 453 supratentorial GBMs newly diagnosed between 1990 and 2005, only 7 (1.5%) were GC-GBMs. These patients were approximately 10 years younger than those with nonGCGBMs. Median OS was 11.4 months for GC-GBM, compared with 9.3 months for nonGC-GBM. However, this difference was not statistically signicant (log-rank P .48) after adjusting for patient age and surgery type (resection vs stereotactic needle biopsy). A potential molecular mechanism for GC formation has recently been described.35,36 The process of endoreduplication results in multiple cycles of DNA replication associated with a failure of cell division, thus causing markedly enlarged polyploid nuclei with extensive chromosomal gains (Figure 4, B). Using antibodies to specic phosphorylation sites on vimentin and GFAP catalyzed by various kinases, including cdc2, CF, Rho, protein kinase C,
GBM VariantsMiller & Perry 399

Table 2. Prognostic Impact of Morphologic Variants in 453 Newly Diagnosed, Adult ( 20 y) Supratentorial Glioblastoma Multiformes (GBMs) at Washington University School of Medicine (19902005)*
Tumor Characteristic GBM GC-GBM GS SC-GBM GBM-O

n 388 7 % 85.7 1.5 Mean age, y 61.6 52.1 Resection, % 64 57 Median OS, mo 9.3 11.4 Log-rank P (vs GBM) ... .48 * P adjusted for patient age and surgery type. GC-GBM indicates giant cell GBM; with oligodendroglial features; and . . ., no data available.

10 21 27 2.2 4.6 6.0 59.1 57.6 47.9 90 91 93 7.6 14.3 27.4 .33 .95 .01 GS, gliosarcoma; SC-GBM, small cell GBM; GBM-O, GBM

and aurora-B, investigators have shown that the molecular alteration that causes failure of cytoplasmic cleavage in cultured multinucleated GCs is loss of aurora-B kinase function.35,36 Importantly, loss of aurora-B kinasemediated phosphorylation was conrmed in human GC-GBM samples.35 However, aurora-B kinase has been shown to be overexpressed in astrocytomas in general, with both mRNA and protein levels correlating with WHO grade.37 Therefore, the molecular basis for aurora-B kinase dysfunction (mutation, gene deletion, methylation) in GCs and its overexpression in non-GC-GBM remains to be established. Gliosarcoma and Adenoid GBM Gliosarcoma constitutes approximately 2% of GBMs and is likewise recognized as a distinct clinicopathologic entity in the WHO 2000 classication.10 These tumors are characterized by their well-circumscribed, biphasic growth pattern with clearly identiable glial and meta-

plastic mesenchymal components (Figure 5, A and B). The glial component of GS may show any of the aforementioned cytologic features and is typically immunoreactive for GFAP (Figure 5, D). The mesenchymal component may also show a wide variety of morphologic appearances, with evidence of differentiation along broblastic, cartilaginous (Figure 5, E and F), osseous, smooth and striated muscle, and adipose lines.10 These areas are generally characterized by dense extracellular matrix deposition, the latter evident on reticulin-stained sections (Figure 5, C, upper) and contrasting with the typically reticulin-poor glial component (Figure 5, C, lower). In addition to sarcomatous differentiation, epithelial metaplasia may also occur in either GS or conventional GBM, including areas of keratinizing squamous or glandular differentiation. Such cases often invoke a differential diagnosis of metastatic carcinoma. The latter is illustrated in Figure 6, A, an adenoid glioma that was immunoreactive for both GFAP (Figure 6, B) and low-molecular-weight cytokeratins (CK7;

Figure 5. Gliosarcoma. A, These tumors show a well-circumscribed, biphasic growth pattern with clearly identiable glial (right) and metaplastic, nonglial components (left) (hematoxylin-eosin, original magnication 200). B, These components are often intermingled (hematoxylin-eosin, original magnication 100). C, A reticulin stain is helpful in identifying the 2 distinct components: The glial component is reticulin poor (center), whereas the sarcomatous component is reticulin rich (top and bottom) (original magnication 100). D, The glial element expresses glial brillary acidic protein (immunoperoxidase, original magnication 100). E and F, The sarcomatous component may also show evidence of other mesenchymal elements, such as cartilaginous differentiation (hematoxylin-eosin, original magnications 400 [E] and 200 [F]). 400 Arch Pathol Lab MedVol 131, March 2007 GBM VariantsMiller & Perry

the differences in OS have been statistically insignicant in most series. This too has been our experience at Washington University School of Medicine (Table 2). Among the 453 newly diagnosed, adult supratentorial GBMs, only 10 (2.2%) were GSs. These patients were of similar age to those with no sarcomatous component. Median OS was 7.6 months for GS, compared with 9.3 months for conventional GBM. This difference was not statistically signicant (log-rank P .33) after adjusting for patient age and surgery type. Metaplasia in noncentral nervous system epithelial tumors is not infrequently encountered in general surgical pathology. Extensive work in developmental and tumor molecular cell biology has elucidated potential molecular mechanisms responsible for epithelial-mesenchymal transition, a phenomenon postulated to be essential for systemic metastasis in noncentral nervous system epithelial tumors. The existence of epithelial-mesenchymal transition is an area of intense debate within the cancer research community and has been extensively reviewed elsewhere.3942 The possibility of an analogous role of epithelial-mesenchymal transition in GS is currently unexplored. Regardless, it is clear that the metaplastic component in GS is neoplastic and frequently harbors cytogenetic and molecular abnormalities similar to those found in the glial component. In this regard, GSs are genetically similar to primary GBMs: 20% to 40% of tumors harbor TP53 mutations, PTEN deletions, and CDKN2A deletions. The only exception to this generalization is the relative infrequency of EGFR amplication in these tumors.10,43 The notion that the sarcomatous element represents mesenchymal metaplasia analogous to that seen in carcinosarcomas elsewhere in the body comes from data demonstrating the presence of 1 or more identical genetic aberrations in both elements from individual tumors; the latter supports a monoclonal process rather than a collision tumor.4345 Similarly, identical genetic alterations have been described in microdissected portions from the glial and epithelial elements of adenoid GBMs and GSs, suggesting that this too represents metaplasia.46,47 OTHER VARIANTS OF GBM Additional variants of GBM have been described beyond those published in the 2000 WHO classication scheme. These include small cell astrocytomas (SCAs), glioblastoma with oligodendroglial features (GBM-O), and GBM with primitive neuronal (primitive neuroectodermal tumor [PNET]like) features. In contrast to granular cell astrocytoma, GC-GBM, and GS, all of which pose little diagnostic difculty for the experienced neuropathologist on routine histopathologic examination, these variants have signicant morphologic overlap with other recognized clinicopathologic entities, particularly HGO neoplasms. Identication of signature molecular genetic alterations, specically codeletion of material on chromosomes 1p and 19q, and its association with responsiveness to adjuvant therapy and improved prognosis, has made accurate histopathologic and molecular genetic characterization of these neoplasms of paramount importance for proper treatment planning and precise prognostication.4851 Codeletion of 1p/19q has been shown to occur most frequently (50%80%) in morphologically classical AO.52,53 A subset ( 20%) of AOA also harbor 1p/19q codeletion,50 although its prognostic utility remains less well dened for these neoplasms. Even when histopathologic
GBM VariantsMiller & Perry 401

Figure 6. Adenoid glioblastoma multiforme. This tumor is an extremely rare variant with epithelial (glandular) metaplasia (A). The glial component (B) is intermingled with the epithelial component, which was immunoreactive for a variety of cytokeratins, including CK7 (C) (hematoxylin-eosin, original magnication 200 [A]; immunoperoxidase, original magnications 200 [B] and 400 [C]).

Figure 6, C). Overall, these tumors are exceedingly rare and are thus far not recognized as a distinct clinicopathologic entity by the WHO. Several reports have demonstrated a slightly better prognosis for GS than for conventional GBM.10,38 However,
Arch Pathol Lab MedVol 131, March 2007

Figure 7. Morphologic features of small cell astrocytoma. These tumors are composed of a monotonous population of cytologically bland cells with uniform oval nuclei with minimal hyperchromasia and minimal discernible cytoplasm (A and B) and may show secondary structuring such as perineuronal satellitosis (C). Delicate chicken-wire capillary vasculature is common (D) (hematoxylin-eosin, original magnications 400 [A] and 200 [B, C, and D]).

classication is rened by ancillary molecular genetic analyses, diagnostic discrepancies remain frequent for the HGO neoplasms, even among experienced neuropathologists.11,1315,54 Small Cell Astrocytoma Small cell astrocytoma is a variant of GBM with considerable morphologic overlap with AO on routine hematoxylin-eosinstained sections.55,56 The histopathologic hallmarks of this tumor are its bland nuclear cytology and striking mitotic activity (Figure 7, A and B). It is composed of a large ( 80%) monotonous collection of astrocytes with small, uniformly oval nuclei with mild hyperchromasia and minimal discernible cytoplasm. Like other diffuse gliomas, SCA can show cortical inltration and secondary structuring, including perineuronal satellitosis (Figure 7, C). The tumor vasculature is sometimes composed of a delicate, chicken-wire capillary network (Figure 7, D). Immunohistochemical studies are a useful adjunct in diagnosing SCA: Tumor cells typically contain long, thin, GFAP-positive cytoplasmic processes (Figure 8, A) and a markedly elevated MIB-1 (Ki-67) labeling index (Figure 8, B).
402 Arch Pathol Lab MedVol 131, March 2007

In contrast to AO, in which median OS ranges from 5 to 10 years, the clinical behavior of SCA is much more aggressive, with median OS virtually indistinguishable from conventional GBM.55,56 It is for this reason that accurate diagnosis of SCA is imperative for accurate prognostication and proper treatment planning. Our experience with 21 SCAs (4.6% of 453 newly diagnosed, adult supratentorial GBMs) at Washington University School of Medicine is listed in Table 2. These patients were of similar age to those with conventional GBMs. Their median OS was 14.3 months and was not statistically signicantly different (log-rank P .95) from that of conventional GBM after adjusting for patient age and surgery type. Identication of characteristic molecular genetic alterations in SCA has provided an additional diagnostic aid in challenging cases. In contrast to AO, SCAs rarely harbor deletion of either 1p or 19q and we have never encountered codeletion of these chromosomes (Table 1). Rather, SCAs show chromosome EGFR amplication and chromosome 10q (PTEN) deletion, with frequencies even higher than conventional GBMs (Table 1). Combined EGFR amplication and 10q deletion is likewise more prevalent in SCA compared with conventional GBM. We have recently
GBM VariantsMiller & Perry

Figure 8. Immunohistochemical and genetic features of small cell astrocytoma. A, Tumor cells typically contain long, thin glial brillary acidic proteinpositive cytoplasmic processes (immunoperoxidase, original magnication 400). B, Mitotic activity and MIB-1 (Ki-67) labeling index is markedly elevated (immunoperoxidase, original magnication 200). C and D, Fluorescence in situ hybridization analyses for chromosomes 7 (EGFR region in red, centromere for chromosome 7 in green) and 10q (PTEN in green and DMBT1 in red) typically reveal evidence of EGFR gene amplication (C) and 10q deletion (D, single red and green signals), respectively (original magnications 400 [C] and 1000 [D]).

analyzed the diagnostic utility of molecular genetic analysis in the diagnosis of SCA.57 Among 239 diffuse, highgrade gliomas (HGGs, including AAs, GBMs, AOs, and AOAs) analyzed by uorescence in situ hybridization (FISH) for EGFR amplication, including 108 SCAs, the sensitivity and specicity for the diagnosis of GBM were 54% and 94%, respectively. EGFR amplication was equally sensitive (92%) and slightly more specic (64%) for SCA (Figure 8, C). Among 222 HGGs with 106 SCAs that were analyzed for 10q abnormalities by FISH, the sensitivity and specicity of 10q deletion for the diagnosis of GBM were 81% and 71%, respectively. 10q deletion was slightly more sensitive (90%) but slightly less specic (64%) for the diagnosis of SCA (Figure 8, D). Because these molecular genetic abnormalities may occasionally be found in other HGGs, FISH analysis does not serve as a replacement for routine histopathologic examination but only as an objective adjunct to conrm the initial morphologic impression of SCA or GBM in general.
Arch Pathol Lab MedVol 131, March 2007

GBM With Oligodendroglial Features The current WHO classication of HGO neoplasms contains signicant ambiguities. The 2 forms of HGO neoplasms, AO and AOA, are distinguished from their lowgrade counterparts, oligodendroglioma and oligoastrocytoma, by the presence of any one or more of the following histologic features: brisk mitotic activity, microvascular proliferation, and necrosis. No signicant weight is given to any of these features, which has given rise to the following debate within the neuro-oncology community: Do grade IV variants of HGO neoplasms exist and, if so, what diagnostic features constitute such entities? We recently performed a retrospective analysis of prognostic factors in 1093 newly diagnosed, adult supratentorial HGGs specically to address this issue.57 This study included 81 AAs, 581 GBMs (including 120 SCAs, 13 GSs, and 2 GC-GBMs), 215 AOAs, and 216 AOs diagnosed according to current WHO 2000 criteria. We considered the diagnosis of an AOA only if distinct astrocytic (Figure 9,
GBM VariantsMiller & Perry 403

We therefore consider WHO grade III AOA with necrosis to constitute a distinct grade IV variant of AOA and routinely designate such neoplasms as GBM-O. An alternative, equally valid designation would be highly anaplastic oligoastrocytoma, grade IV (MOA4). These tumors portend a signicantly better prognosis than other variants of GBM, in line with the relatively better prognosis for oligodendroglial neoplasms in general. In contrast to the nding of codeletions in roughly 85% of our AOs, deletion of either 1p (24%), 19q (43%), or combined 1p/19q (22%) is relatively infrequent in GBM-O (Table 1). Nonetheless, these frequencies are nearly identical to those we have encountered in mixed oligoastrocytomas overall. Interestingly, these tumors more frequently show solitary 19q deletion (21% of 58 analyzed tumors). Among 586 diffuse HGGs analyzed by FISH for 1p/19q abnormalities, the sensitivity and specicity of solitary 19q deletion for the diagnosis of AOA and GBM-O were 25% and 93%, respectively. Thus, the presence of solitary 19q deletion may be a useful molecular genetic feature to conrm the impression of a GBM-O after routine histopathologic examination, but its infrequent occurrence in these tumors precludes reliance on this feature in establishing the diagnosis. GBM With Primitive Neuronal (PNET-like) Features Rare examples of high-grade glioneuronal tumors or GBMs combined with cerebral neuroblastoma/supratentorial PNET have been previously published but remain relatively poorly characterized.58 Similarly, we have encountered a number of such cases, often as consults with the differential diagnosis including an unusual GBM versus a supratentorial PNET with extensive glial differentiation (Figure 10, A). In support of the former interpretation, the majority of our cases have occurred in adults and had foci resembling conventional GBM, sometimes with foci of low-grade glioma in either the same or a prior biopsy. The primitive neuronal (PNET-like) element often looks like a separate clone with nodules of markedly increased cellularity, high nuclear-cytoplasmic ratios, markedly hyperchromatic cells, remarkably high mitotic and Ki-67 indices (close to 100% in some, Figure 10, D), and convincing evidence of neuronal or neuroblastic differentiation, such as Homer-Wright rosettes (Figure 10, B) and expression of neuronal antigens (eg, synaptophysin [Figure 10, C], Neu-N, neurolament protein, neuron-specic enolase). To date, too few cases have been reported to determine the behavior or optimal therapy, although most of the patients we have encountered have died in less than a year, analogous to GBM in general. CONCLUSION Glioblastoma multiformes are morphologically diverse neoplasms with poor prognosis. Precise subclassication is becoming increasingly important, as several of the newly dened variants described since the publication of the WHO classication in 2000 have signicant morphologic overlap with tumors that have more favorable prognosis and treatment response rates. Some of the variants described in this article have recently been added to the upcoming WHO 2007 scheme as distinctive patterns of GBM. More widespread application of the principles described in this review may permit more accurate prognostication and appropriate treatment planning in the future.
GBM VariantsMiller & Perry

Figure 9. Glioblastoma multiforme with oligodendroglial features (GBM-O or mixed oligoastrocytoma, grade IV). GBM-O show evidence of both astrocytic (A) and oligodendroglial (B) differentiation, with either geographically distinct (A) or intermingled (not shown) components. Necrosis, particularly pseudopalisading necrosis (A), distinguishes these tumors from their grade III counterparts (AOA) (hematoxylineosin, original magnications 100 [A] and 200 [B]).

A) and oligodendroglial (Figure 9, B) components were evident, either as geographically separated (Figure 9) or intermingled (not shown) areas. We found that the presence of any type of necrosis (Figure 9, A) in an otherwise grade III AOA was a strong predictor of poor survival independent of patient age (hazard ratio 2.4, 95% condence interval 1.53.7, Cox P .001): median OS for AOA patients whose tumors featured necrosis was 22.8 months, compared with 86.9 months for those without necrosis (log-rank P .001). No signicant effect of necrosis on median OS (89.8 months) was evident in patients with pure AO (log-rank P .32). Moreover, the median OS of AOA with necrosis was closer to GBM (9.8 months) than to AOA without necrosis. Survival data for a subset of these GBM-O cases seen at Washington University School of Medicine is presented in Table 2. In this cohort of 27 patients (6% of 453 GBMs seen between 1990 and 2005), patient age was similar to that seen in GC-GBM, approximately 10 to 15 years younger than conventional GBM. Their median OS was 27.4 months, signicantly longer than that for conventional GBM after correction for patient age and surgery type (P .01).
404 Arch Pathol Lab MedVol 131, March 2007

Figure 10. Glioblastoma multiforme with neuronal peripheral neuroectodermal tumorlike features. These tumors contain both astrocytic (A) and primitive neuronal (B) components. The latter element is characterized by Homer-Wright rosettes (B) and synaptophysin immunoreactivity (C). Marked proliferative activity is evident (D) (hematoxylin-eosin, original magnications 200 [A] and 400 [B]; immunoperoxidase, original magnications 100 [C] and 200 [D]).

Supported in part by a cancer biology training grant from the National Cancer Institute, Bethesda, Md (grant T32CA009547 to C.R.M.).
References
1. Central Brain Tumor Registry of the United States: Statistical Report: Primary Brain Tumors in the United States, 1998-2002. Hinsdale, Ill: Central Brain Tumor Registry of the United States; 2006. 2. Bailey P, Cushing H. A Classication of the Tumors of the Glioma Group on a Histogenetic Basis With a Correlated Study of Prognosis. Philadelphia: JB Lippincott Co; 1926:175. 3. Kernohan JW, Mabon RF, Svien HJ, Adson AW. A simplied classication of gliomas. Proc Staff Meet Mayo Clin. 1949;24:7175. 4. Ringertz J. Grading of gliomas. Acta Pathol Microbiol Scand. 1950;27:51 64. 5. Nelson DF, Nelson JS, Davis DR, Chang CH, Grifn TW, Pajak TF. Survival and prognosis of patients with astrocytoma with atypical or anaplastic features. J Neurooncol. 1985;3:99103. 6. Daumas-Duport C, Scheithauer B, OFallon J, Kelly P. Grading of astrocytomas: a simple and reproducible method. Cancer. 1988;62:21522165. 7. Kim TS, Halliday AL, Hedley-Whyte ET, Convery K. Correlates of survival and the Daumas-Duport grading system for astrocytomas. J Neurosurg. 1991;74: 2737. 8. Kleihues P, Burger PC, Scheithauer BW, Zulch KJ. Histological Typing of Tumours of the Central Nervous System. 2nd ed. Berlin, Germany: Springer-Verlag; 1993:xii, 112. 9. Zulch KJ. Histological Typing of Tumors of the Central Nervous System. Geneva, Switzerland: World Health Organization; 1979. 10. Kleihues P, Cavenee WK. Pathology and Genetics of Tumours of the Ner-

vous System. Lyon, France: IARC Press; 2000:314. World Health Organization Classication of Tumours. 11. Brat DJ, Castellano-Sanchez A, Kaur B, Van Meir EG. Genetic and biologic progression in astrocytomas and their relation to angiogenic dysregulation. Adv Anat Pathol. 2002;9:2436. 12. Brat DJ, Van Meir EG. Glomeruloid microvascular proliferation orchestrated by VPF/VEGF: a new world of angiogenesis research. Am J Pathol. 2001;158: 789796. 13. Aldape K, Simmons ML, Davis RL, et al. Discrepancies in diagnoses of neuroepithelial neoplasms: the San Francisco Bay Area Adult Glioma Study. Cancer. 2000;88:23422349. 14. Coons SW, Johnson PC, Scheithauer BW, Yates AJ, Pearl DK. Improving diagnostic accuracy and interobserver concordance in the classication and grading of primary gliomas. Cancer. 1997;79:13811393. 15. Prayson RA, Agamanolis DP, Cohen ML, et al. Interobserver reproducibility among neuropathologists and surgical pathologists in brillary astrocytoma grading. J Neurol Sci. 2000;175:3339. 16. Ohgaki H, Dessen P, Jourde B, et al. Genetic pathways to glioblastoma: a population-based study. Cancer Res. 2004;64:68926899. 17. Mallory FB. The Principles of Pathologic Histology. Philadelphia, Pa: WB Saunders Co; 1914:1, 11677. 18. Burger PC, Kleihues P. Cytologic composition of the untreated glioblastoma with implications for evaluation of needle biopsies. Cancer. 1989;63:2014 2023. 19. Nissl F. Zur histopathologie der paralytischen Rindenerkrankung. Histol Histopathol Arb Grosshirn. 1904;1:315. 20. Reis RM, Hara A, Kleihues P, Ohgaki H. Genetic evidence of the neoplastic nature of gemistocytes in astrocytomas. Acta Neuropathol (Berl). 2001;102:422 425.

Arch Pathol Lab MedVol 131, March 2007

GBM VariantsMiller & Perry

405

21. Kros JM, Waarsenburg N, Hayes DP, Hop WC, van Dekken H. Cytogenetic analysis of gemistocytic cells in gliomas. J Neuropathol Exp Neurol. 2000;59: 679686. 22. Krouwer HG, Davis RL, Silver P, Prados M. Gemistocytic astrocytomas: a reappraisal. J Neurosurg. 1991;74:399406. 23. Watanabe K, Peraud A, Gratas C, Wakai S, Kleihues P, Ohgaki H. p53 and PTEN gene mutations in gemistocytic astrocytomas. Acta Neuropathol (Berl). 1998;95:559564. 24. Kosel S, Scheithauer BW, Graeber MB. Genotype-phenotype correlation in gemistocytic astrocytomas. Neurosurgery. 2001;48:187193; discussion 193 184. 25. Avninder S, Sharma MC, Deb P, et al. Gemistocytic astrocytomas: histomorphology, proliferative potential and genetic alterationsa study of 32 cases. J Neurooncol. 2006;78:123127. 26. Tihan T, Vohra P, Berger MS, Keles GE. Denition and diagnostic implications of gemistocytic astrocytomas: a pathological perspective. J Neurooncol. 2006;76:175183. 27. Watanabe K, Tachibana O, Yonekawa Y, Kleihues P, Ohgaki H. Role of gemistocytes in astrocytoma progression. Lab Invest. 1997;76:277284. 28. Brat DJ, Scheithauer BW, Medina-Flores R, Rosenblum MK, Burger PC. Inltrative astrocytomas with granular cell features (granular cell astrocytomas): a study of histopathologic features, grading, and outcome. Am J Surg Pathol. 2002;26:750757. 29. Geddes JF, Thom M, Robinson SF, Revesz T. Granular cell change in astrocytic tumors. Am J Surg Pathol. 1996;20:5563. 30. Castellano-Sanchez AA, Ohgaki H, Yokoo H, et al. Granular cell astrocytomas show a high frequency of allelic loss but are not a genetically dened subset. Brain Pathol. 2003;13:185194. 31. Chorny JA, Evans LC, Kleinschmidt-DeMasters BK. Cerebral granular cell astrocytomas: a Mib-1, bcl-2, and telomerase study. Clin Neuropathol. 2000;19: 170179. 32. Meyer-Puttlitz B, Hayashi Y, Waha A, et al. Molecular genetic analysis of giant cell glioblastomas. Am J Pathol. 1997;151:853857. 33. Peraud A, Watanabe K, Schwechheimer K, Yonekawa Y, Kleihues P, Ohgaki H. Genetic prole of the giant cell glioblastoma. Lab Invest. 1999;79:123129. 34. Shinojima N, Kochi M, Hamada J, et al. The inuence of sex and the presence of giant cells on postoperative long-term survival in adult patients with supratentorial glioblastoma multiforme. J Neurosurg. 2004;101:219226. 35. Fujita M, Mizuno M, Nagasaka T, et al. Aurora-B dysfunction of multinucleated giant cells in glioma detected by site-specic phosphorylated antibodies. J Neurosurg. 2004;101:10121017. 36. Maeda K, Mizuno M, Wakabayashi T, et al. Morphological assessment of the development of multinucleated giant cells in glioma by using mitosis-specic phosphorylated antibodies. J Neurosurg. 2003;98:854859. 37. Araki K, Nozaki K, Ueba T, Tatsuka M, Hashimoto N. High expression of Aurora-B/Aurora and Ipll-like midbody-associated protein (AIM-1) in astrocytomas. J Neurooncol. 2004;67:5364. 38. Galanis E, Buckner JC, Dinapoli RP, et al. Clinical outcome of gliosarcoma compared with glioblastoma multiforme: North Central Cancer Treatment Group results. J Neurosurg. 1998;89:425430. 39. Thiery JP. Epithelial-mesenchymal transitions in tumour progression. Nat Rev Cancer. 2002;2:442454.

40. Lee JM, Dedhar S, Kalluri R, Thompson EW. The epithelial-mesenchymal transition: new insights in signaling, development, and disease. J Cell Biol. 2006; 172:973981. 41. Larue L, Bellacosa A. Epithelial-mesenchymal transition in development and cancer: role of phosphatidylinositol 3 kinase/AKT pathways. Oncogene. 2005;24:74437454. 42. Tarin D, Thompson EW, Newgreen DF. The fallacy of epithelial mesenchymal transition in neoplasia. Cancer Res. 2005;65:59966000; discussion 6000 6001. 43. Reis RM, Konu-Lebleblicioglu D, Lopes JM, Kleihues P, Ohgaki H. Genetic prole of gliosarcomas. Am J Pathol. 2000;156:425432. 44. Actor B, Cobbers JM, Buschges R, et al. Comprehensive analysis of genomic alterations in gliosarcoma and its two tissue components. Genes Chromosomes Cancer. 2002;34:416427. 45. Paulus W, Bayas A, Ott G, Roggendorf W. Interphase cytogenetics of glioblastoma and gliosarcoma. Acta Neuropathol (Berl). 1994;88:420425. 46. Ozolek JA, Finkelstein SD, Couce ME. Gliosarcoma with epithelial differentiation: immunohistochemical and molecular characterization: a case report and review of the literature. Mod Pathol. 2004;17:739745. 47. du Plessis DG, Rutherfoord GS, Joyce KA, Walker C. Phenotypic and genotypic characterization of glioblastoma multiforme with epithelial differentiation and adenoid formations. Clin Neuropathol. 2004;23:141148. 48. Bauman GS, Ino Y, Ueki K, et al. Allelic loss of chromosome 1p and radiotherapy plus chemotherapy in patients with oligodendrogliomas. Int J Radiat Oncol Biol Phys. 2000;48:825830. 49. Cairncross JG, Ueki K, Zlatescu MC, et al. Specic genetic predictors of chemotherapeutic response and survival in patients with anaplastic oligodendrogliomas. J Natl Cancer Inst. 1998;90:14731479. 50. Perry A, Fuller CE, Banerjee R, Brat DJ, Scheithauer BW. Ancillary FISH analysis for 1p and 19q status: preliminary observations in 287 gliomas and oligodendroglioma mimics. Front Biosci. 2003;8:a19. 51. Smith JS, Perry A, Borell TJ, et al. Alterations of chromosome arms 1p and 19q as predictors of survival in oligodendrogliomas, astrocytomas, and mixed oligoastrocytomas. J Clin Oncol. 2000;18:636645. 52. Burger PC. What is an oligodendroglioma? Brain Pathol. 2002;12:257 259. 53. Burger PC, Minn AY, Smith JS, et al. Losses of chromosomal arms 1p and 19q in the diagnosis of oligodendroglioma: a study of parafn-embedded sections. Mod Pathol. 2001;14:842853. 54. Perry A. Oligodendroglial neoplasms: current concepts, misconceptions, and folklore. Adv Anat Pathol. 2001;8:183199. 55. Perry A, Aldape KD, George DH, Burger PC. Small cell astrocytoma: an aggressive variant that is clinicopathologically and genetically distinct from anaplastic oligodendroglioma. Cancer. 2004;101:23182326. 56. Burger PC, Pearl DK, Aldape K, et al. Small cell architecturea histological equivalent of EGFR amplication in glioblastoma multiforme? J Neuropathol Exp Neurol. 2001;60:10991104. 57. Miller CR, Duhnam CP, Scheithauer BW, Perry A. Signicance of necrosis in grading of oligodendroglial neoplasms: a clinicopathologic and genetic study of newly diagnosed high-grade gliomas. J Clin Oncol. 2006;24:54195426. 58. McLendon RE, Provenzale J. Glioneuronal tumors of the central nervous system. Brain Tumor Pathol. 2002;19:5158.

406 Arch Pathol Lab MedVol 131, March 2007

GBM VariantsMiller & Perry