General Histopathological Features

Light microscopic examination of hematoxylin and eosin-stained sections is the basis of most classification and grading systems for nervous system tumors. The approach to nervous system neoplasia follows the standard approach to classifying human tumors from all parts of the body; thus, many neuropathological terms come from the discipline of general pathology. This lexicon includes the terms most frequently used in clinical practice.

Palisading

Palisading, or lining up of tumor cells, is characteristic of schwannomas, in which collections of palisaded nuclei are termed Verocay bodies. The term describes a collection of neoplastic cells arranged in parallel arrays, or palisades. In glioblastomas, palisading (also known as pseudopalisades) occurs around necrosis, with serpiginous zones of necrosis lined by hypercellular viable tumor nuclei (Fig. 52B.1).

Fig. 52B.1 Glioblastoma. Nuclear pseudopalisading surrounding foci of central necrosis (hematoxylin-eosin stain, ×100).

Rosettes

The term rosette is confusing because of its application to several histological structures. The two most commonly encountered rosettes are Homer Wright rosettes and perivascular rosettes. A Homer Wright, or neuroblastic, rosette is a ring of cells surrounding neuropil (i.e., delicate fibrillary processes) and represents axon formation by primitive neuronal elements. Because these Homer Wright rosettes are identical to those encountered in the neuroblastomas of the peripheral nervous system, their presence is often taken as evidence of neuroblastic differentiation. They are associated with primitive neuroectodermal tumors (PNETs) such as medulloblastoma and pineoblastoma, where they are typically a focal finding. Sometimes, the terms central or cerebellar neuroblastomas are used in reference to CNS tumors with extensive Homer Wright rosette and neuropil formation.

Characteristic of perivascular rosettes is a peripheral ring of tumor nuclei surrounding a central blood vessel with a nuclear-free eosinophilic zone in between (Fig. 52B.2). The nuclear-free zone derives from tapering cellular processes that radiate from the tumor cells to the vessel. Perivascular rosettes are most typical of ependymomas, and glial fibrillary acidic protein (GFAP) immunostains highlight the perivascular processes. Occasionally, similar axon-bearing structures appear in neuronal tumors as well (e.g., neurocytomas, pineocytomas, various forms of PNET).

Fig. 52B.2 Ependymoma. Perivascular pseudorosette characterized by a fibrillary-appearing perivascular nuclear-free zone (hematoxylin-eosin stain, ×200).

Microvascular Proliferation

Microvascular (endothelial and pericytic/smooth muscle cell) proliferation represents an important and interesting response to tumor-associated and hypoxia-associated angiogenesis factors. It occurs in either benign or malignant CNS neoplasms, although it is more common in the malignant. In pilocytic astrocytoma, the new vessels may take on a delicate glomeruloid or “wicker-works” configuration, usually with a single-layered flat endothelial lining. In malignant tumors such as glioblastoma, these vessels are also glomeruloid but additionally display multilayering, with enlarged or “hyperplastic” endothelial cells (Fig. 52B.3). In either case, these neovascular structures lack the tight junctions normally contributing to the blood-brain barrier. Therefore, they typically leak intravascular contrast material, thus accounting for the tumor-associated enhancement detected radiologically.

Fig. 52B.3 Glioblastoma. Focus of endothelial hyperplasia with glomeruloid multilayered vessel (hematoxylin-eosin stain, ×400).

Frozen Sections and Touch Imprints/Smears

Intraoperative frozen sections are more difficult to interpret than fixed tissue sections, but their use has the advantage of speed, requiring only a few minutes to prepare. This technique is requested for several reasons, including simple curiosity and a desire to provide rapid feedback to patients and their families. However, the most important reasons for performing this technique are to ensure that the pathologist obtains a representative sample for permanent sections and to provide information that may alter the surgical procedure. For example, a neurosurgeon may opt to stop at a limited biopsy for a suspected lymphoma, aggressively resect for an ependymoma, or send additional material to the microbiology laboratory for an abscess. Because the process of tissue freezing produces significant cytological artifacts, another popular technique has been the preparation of touch imprints, or smears, from fresh tissue. Although the underlying architecture is lost, the tissue requirements are minimal, and the cytological preservation is excellent (Fig. 52B.4). This may be critical for diagnoses based primarily on nuclear cytology, such as reactive gliosis versus low-grade diffuse glioma. It is also important to realize that the frozen section technique results in substantial tissue loss and morphological artifacts in the thawed residual specimen. Therefore, it is best to omit its use in limited biopsies for which the surgical procedure will not be altered. In other words, it is imperative to save at least some optimally fixed nonfrozen tissue for final diagnosis on sections from the fixed embedded material.

Fig. 52B.4 Diffuse astrocytoma on an intraoperative smear. Cytological preservation is superior to that of a typical frozen section, and cytoplasmic processes are easier to discern. The oval- to spindle-shaped hyperchromatic nuclei are typical of diffuse astrocytoma (hematoxylin-eosin stain, ×400).

Electron Microscopy (Ultrastructural Pathology)

Electron microscopy has been virtually supplanted by immunohistochemistry in most pathology laboratories. Nevertheless, it can be extremely valuable in the diagnosis of certain tumors and remains the gold standard for a few. For example, it may be the only method capable of proving ependymal differentiation. Ultrastructural pathology is used to gain insight into forms of differentiation primarily by visualizing specific organelles and other components of the cytoplasm and cell surface (e.g., intermediate filaments, neurosecretory granules, synapses, pinocytotic vesicles, intercellular junctions, cilia, microvilli, basement membrane) and is therefore of particular utility in the diagnosis of poorly differentiated neoplasms.

Immunohistochemistry

Immunohistochemistry is an ancillary diagnostic technique for detecting protein expression within tumor nuclei, cytoplasm, and cell membranes. It is used most commonly to determine lines of cellular differentiation, proliferation indices, oncogene overexpression, or losses of tumor suppressor expression. Monoclonal antibody technology and recently improved antigen-retrieval methods have greatly expanded the versatility of this technique in routine formalin-fixed paraffin-embedded tissue, although a number of pitfalls remain, and considerable experience is necessary to make accurate determinations. Commercial antibodies vary greatly in terms of sensitivities, specificities, and clinical utilities. Most laboratories use the immunoperoxidase staining technique with horseradish peroxidase and a brown diaminobenzidine dye because staining is permanent and the reaction is visible by conventional light microscopy. The application of several antibodies is usual to study the more common CNS tumors.

Methods of Assessing Cell Proliferation

The simplest and least expensive method for estimating cellular proliferation is the mitotic index, usually expressed as either the average or the maximal attainable count per 10 high-power fields (HPF). Difficulties stem from pyknotic nuclei mimicking mitoses in poorly preserved specimens, variable field sizes among microscopes, and lack of uniform methods for counting. Underestimates of the proliferation index are common as well, given that the cells undergoing mitosis represent only a small fraction of cycling cells within a tumor.

Most proliferation antigens are nuclear constituents that are manifested during one or more proliferative phases in the cell cycle, allowing the recognition of cycling cells by routine immunohistochemistry. These markers are already valued in the diagnosis and management of brain tumors. The monoclonal antibody, Ki-67 (also known as MIB-1), binds to a human nuclear protein in the growth fraction (G1, S, G2, and M phases of the cell cycle). In most tumor types, the Ki-67 or MIB-1 labeling index increases with increasing degree of malignancy.

Molecular Diagnostics

As stated, tumors arise when alterations occur in key growth regulatory molecules. The expectation is that measurements of such molecular alterations will change the way tumors are classified. Molecular changes are detectable at the genomic DNA, messenger ribonucleic acid (mRNA), and protein levels. Some tumors are associated with “signature” cytogenetic abnormalities such as chromosomal translocations and deletions. However, abnormalities at the submicroscopic or molecular levels (e.g., gene rearrangements, single-gene defects) that may affect oncogenes and tumor suppressor genes are undetectable by routine karyotyping (conventional cytogenetics). The most common and practical approaches to detect deletions of chromosomal regions or specific tumor suppressor genes include loss of heterozygosity (LOH), fluorescence in situ hybridization (FISH), array comparative genomic hybridization (aCGH), and quantitative polymerase chain reaction (PCR) techniques. For detection of oncogene amplification, FISH and quantitative PCR are the most practical methods. Thus far, no specific translocations of fusion transcripts associated with primary CNS neoplasms exists, as is the case in hematopoietic and soft-tissue tumors.

Techniques such as aCGH and oligonucleotide or cDNA expression profiling are useful for surveying entire genomes for abnormalities. However, the primary use for these techniques is research rather than diagnostic pathology.

Few molecular diagnostic tests have become routinely used or standard of care in neuro-oncology. The most notable and most widely used is chromosome 1p and 19q testing as a prognostic/management tool for patients with oligodendroglial tumors. FISH has the advantage of simplicity, morphological preservation, minimal tissue and purity requirements, and a lack of necessity for microdissection or matching blood/nonneoplastic tissue. However, it requires some experience for accurate interpretation, especially in cases with aneuploid populations of tumor cells. It also utilizes large probes (100–300 kb) and is insensitive to very small deletions. In selected cases, FISH has been used to evaluate epidermal growth factor receptor (EGFR) amplification to distinguish small cell glioblastoma from anaplastic oligodendroglioma. It is likely that many more applications of molecular diagnostics will become incorporated into diagnostic neuropathology laboratories in the near future, but it is also important to note that specific immunohistochemical assays may “substitute” for molecular diagnostic approaches in certain situations. For example, immunohistochemical absence of the product of the INI1 gene on chromosome 22q is now widely accepted as a sensitive and specific marker for atypical teratoid/rhabdoid tumors (AT/RTs) (Judkins et al., 2004). Similarly, because IDH1 is a frequently mutated gene in lower-grade gliomas and secondary glioblastoma (Yan et al., 2009), antibodies to the most common mutant form of the IDH1 protein have demonstrated clinical utility in the pathological assessment of infiltrating glioma (Camelo-Piragua et al., 2009).

Diffuse Astrocytoma (WHO Grade II)

Referred to loosely as well-differentiated, low-grade, diffuse, grade II, or simply astrocytoma, this slow-growing tumor regularly undergoes malignant progression to anaplastic astrocytoma and glioblastoma. Despite often-quoted survivals of 5 to 10 years, a wide range exists among individual patients. The median age at the time of diagnosis is approximately 35 years, and young patients have a considerably better prognosis than older patients. Radiographically, these tumors are ill-defined nonenhancing cerebral masses. A much smaller proportion of diffuse astrocytomas occur in the spinal cord, cerebellum, and brainstem. Brainstem gliomas deserve special mention because these tumors typically present in childhood, expand the pons, and have a uniformly poor prognosis regardless of the histological grade.

Grossly, diffuse astrocytomas are poorly circumscribed tumors in which the CNS parenchyma variably expands, but the overall disruption of anatomy is minimal. Given its diffusely infiltrative nature and general lack of solid mass formation, large portions of the tumor may appear invisible to the naked (or radiographic) eye (i.e., microscopic disease). Also, this infiltrative growth pattern makes therapy a challenge because it has not been possible to treat microscopic foci of disease adequately with focal therapies such as surgery and radiation. Particularly widespread infiltration involving multiple lobes and even the brainstem (gliomatosis cerebri) is rare and tends to have a poor prognosis regardless of histological grade.

Microscopically, a slight to moderate increase in cellularity occurs, and the primary distinction from reactive astrocytosis rests on the finding of cytological atypia. Diffuse astrocytoma cells can take on various morphological appearances, including fibrillary, gemistocytic, and protoplasmic types. Fibrillary morphology is most common and consists of irregular elongated hyperchromatic nuclei, either appearing “naked” in an otherwise fibrillary background or displaying discernible cytoplasmic processes (see Fig. 52B.4). GFAP immunostains highlight the latter, although this is neither absolutely sensitive nor specific, because (1) some astrocytomas harbor minimal quantities of GFAP-positive cytoplasm, (2) interpretation may be difficult owing to staining of nonneoplastic astrocytic elements or high background in general, and (3) other gliomas display GFAP immunoreactivity. By definition, grade II astrocytomas lack mitotic activity, microvascular proliferation, and necrosis. Ancillary staining for MIB-1 (Ki-67) generally reveals a low proliferative index as well.

Recently, a large proportion of low-grade gliomas (including 90% of grade II astrocytomas) were found to have mutations in IDH1 and IDH2, even though the mutation was initially described in 12% of (largely secondary) glioblastomas (Parsons et al., 2008). IDH1 encodes isocitrate dehydrogenase, a component of the citric acid cycle. The role of IDH1 in glioma genesis remains uncertain, although recent work suggests that this loss of function mutation may cause accumulation of substrate, triggering HIF signaling pathways (Zhao et al., 2009). Others have proposed that tumorigenesis may be a consequence of 2-hydroxyglutarate formation (Dang et al., 2009). IDH1 mutations in low-grade glioma are tightly correlated with earlier age at diagnosis and tumor grade, appear to confer prognostic benefit beyond tumor grade, and are frequently associated with O⁶-methylguanine-DNA methyltransferase (MGMT) methylation (Sanson et al., 2009).

Chromosome 17p losses and mutations of TP53 are common molecular genetic events in diffuse astrocytomas, seen in approximately half of cases (Louis, 2006). Those with gene mutation usually display strong nuclear immunostaining for p53 protein, although the association is imperfect, since the protein is stabilizable by other mechanisms, and many immunopositive cases lack TP53 mutations. Those cases that evolve to glioblastoma (secondary glioblastoma) typically retain evidence of TP53 mutation and may even show additional TP53 gene mutations or clonal expansion of the tumor cells with mutation. On the other hand, de novo, or primary, glioblastomas rarely display such mutations. Other common alterations involve the overexpression of platelet-derived growth factor receptor alpha (PDGFα) in astrocytomas of all grades and losses of chromosome 22.

Anaplastic Astrocytoma (WHO Grade III)

The clinical presentation and radiographic features of anaplastic astrocytoma are similar to those of grade II astrocytoma, except that the mean age at presentation is approximately a decade later, and some cases show contrast enhancement. Average survival is reduced to 3 years, although as with grade II astrocytoma, great individual variability exists; patient age is one of the most powerful prognostic variables.

Histologically, anaplastic astrocytomas are more cellular than grade II astrocytomas and display a greater degree of proliferation. The presence of mitotic figures primarily defines these tumors. One astrocytoma variant that commonly presents at the anaplastic or grade III level is the gemistocytic (“stuffed cell”) astrocytoma. Characterized by strongly GFAP-positive cells with eccentric bellies of eosinophilic cytoplasm, this tumor type has a high incidence of progression to glioblastoma. Interestingly, the small-cell astrocytoma elements in the background primarily constitute the proliferating element. Another feature that may help distinguish anaplastic astrocytomas from grade II astrocytomas is the generally higher Ki-67 or MIB-1 proliferative index on immunostains (Giannini et al., 1999a).

Anaplastic astrocytomas share the high frequency of TP53, IDH1, and IDH2 mutations seen in grade II astrocytomas. Inactivation of the retinoblastoma protein (Rb) cell-cycle regulatory pathway is also common and may be primarily responsible for the increased proliferation observed histologically. Homozygous deletion of the CDKN2A/p16 gene on 9p is the most common mechanism for disabling this pathway, although deletions of the RB gene and CDK4 amplifications are alternative aberrations (Louis, 2006). Amplifications of the EGFR gene and losses of PTEN on 10q occur less frequently in anaplastic astrocytomas than in glioblastomas and are virtually absent in tumors containing an IDH1 mutation (Sanson et al., 2009). When present, EGFR amplification may suggest either a worse prognosis or the possibility of undersampling/undergrading in a glioblastoma (Smith et al., 2001).

Glioblastoma (WHO Grade IV)

Glioblastoma (previously known as glioblastoma multiforme) is the most common malignant primary brain tumor in adults, accounting for approximately 50% of all gliomas. The peak age at onset is 50 to 60 years, roughly a decade later than anaplastic astrocytoma. Unfortunately, several decades of basic and clinical research have had little impact on the clinical outcomes associated with glioblastoma, and the average survival remains at approximately 1 to 1.5 years following radiation therapy and chemotherapy.

Glioblastoma most commonly occurs in the deep white matter, basal ganglia, or thalamus and is rarely found in the cerebellum or spinal cord. The gross and microscopic appearance is heterogeneous. Replacing the affected portion of the brain is a single mass that grossly may appear deceptively well circumscribed but microscopically infiltrates widely, often spreading to the opposite hemisphere via the corpus callosum (butterfly lesion). Multifocal tumors may occur and in most cases likely represent separate regions of malignant transformation within a widely disseminated lower-grade astrocytoma such as gliomatosis cerebri. In advanced stages, the tumor may extend into the meninges or the ventricle. Seeding of the neuraxis as multiple implants on the brain or ventricular surfaces is an atypical growth pattern, and extracranial metastases are extremely rare. The cut surface has a variegated appearance (Fig. 52B.5) characterized by central yellow or white zones of necrosis and hemorrhage surrounded by a hyperemic ring (endothelial hyperplasia) and “edematous” brain with variable mixtures of vasogenic edema, gliosis, and tumor infiltrates. Microscopically characterizing glioblastomas are all the features of anaplastic astrocytoma plus endothelial hyperplasia or necrosis. The endothelial hyperplasia is thickened or glomeruloid vessels with multilayering (see Fig. 52B.3), representing a form of tumor-induced angiogenesis, a potential target for novel therapies. The necrosis is often associated with a characteristic serpiginous distribution and associated nuclear pseudopalisading (see Fig. 52B.1). Several histological variants exist, including giant cell glioblastoma, small cell glioblastoma, and gliosarcoma. Characteristic of the latter is a sarcomatous element, currently felt to arise most likely from mesenchymal metaplasia within a glioblastoma. No significant clinical differences are identifiable in these variants when compared with conventional glioblastoma, although gliosarcomas more frequently occur in the temporal lobes.

Fig. 52B.5 Glioblastoma (cut in axial imaging plane). This patient’s right hemispheric glioblastoma demonstrates central necrosis, a deceptively demarcated hyperemic rim (endothelial hyperplasia), and an infiltrative component that extends well beyond the grossly discernible margins (note expansion of right thalamus, with blurring of gray/white junctions).

Glioblastomas are among the best genetically characterized CNS tumors and often harbor numerous cytogenetic and molecular genetic aberrations (Louis, 2006). Mutations of TP53, IDH1, and IDH2 are less common than in diffuse astrocytomas or anaplastic astrocytomas and occur more often in secondary glioblastomas. Arising in younger patients after malignant transformation of a lower-grade precursor, TP53 mutations occur at greater frequency in giant cell glioblastomas (Louis et al., 2001). In contrast, 30% to 40% of glioblastomas harbor EGFR amplifications, often with an associated constitutively activating mutation, and these occur mainly in the primary, or de novo, glioblastoma (Louis et al., 2001). Defining the latter is the lack of a precursor lesion, typically in an older patient with rapid clinical onset. EGFR amplifications are also particularly common (≈70%) in the small-cell variant. These tumors are composed of small round cells with minimal cytoplasm and a remarkably brisk mitotic index. Owing to the rounded, uniform nuclei, a misdiagnosis of anaplastic oligodendrogliomas may occur, although small cell glioblastoma share the demographic features and poor prognosis of glioblastomas in general and do not harbor the 1p and 19q deletions seen in oligodendrogliomas. Alterations of the p16/CDK4/RB pathway are nearly universal in glioblastoma, and many cases harbor chromosome 10 losses (mostly primary glioblastomas). The PTEN gene located on chromosome 10q23 mutates in a subset of these and may have prognostic significance in certain patient populations. Additional tumor suppressor genes on both the long and short arms of chromosome 10 are suspected in the remaining group of PTEN wild-type tumors with demonstrable LOH. The list of additional glioblastoma-associated alterations is growing rapidly. Recent large-scale sequencing studies have shown that glioblastomas have somatic mutations of multiple genes (Parsons et al., 2008; TCGA Network, 2008). The studies have confirmed our knowledge of the incidence of mutated and/or amplified genes such as EGFR, PTEN, RB1, and PIK3CA and showed higher prevalence for mutations in TP53, NF1, ERBB2, and PIK3R1 (TCGA Network, 2008) as well as new candidate genes involved in tumorigenesis, such as IDH1 (Parsons et al., 2008). Nevertheless, the focus of intensive research is on mechanisms of invasion, cellular migration, proliferation, angiogenesis, apoptosis deregulation, and therapeutic resistance (Louis, 2006).

Early evidence suggests that molecular analysis of glioblastomas may have prognostic and predictive relevance. For example, methylation of the MGMT gene promoter, leading to down-regulation of MGMT expression, appears prognostic of improved survival in patients with glioblastoma treated with the alkylating agent, temozolomide (Hegi et al., 2005). However, emerging data suggest that prognostic benefit associated with MGMT methylation in glioblastoma may be unrelated to the mode of therapy administered (Rivera et al., 2010). Furthermore, resistance to alkylating chemotherapy in glioblastoma is associated with therapy-related inactivation of MSH6, a DNA mismatch repair enzyme, allowing increased mutagenesis and the emergence of a hypermutated phenotype (Cahill et al., 2007; Yip et al., 2009).

Molecular approaches to global-expression profiling potentially provide powerful means to distinguish glioblastomas from other malignant gliomas that follow distinct clinical courses (Nutt et al., 2003). Moreover, gene-expression profiling of glioblastoma has progressed from augmenting information provided by histological examination to characterization of molecular subtypes based on different gene expression patterns in each group (Sulman et al., 2009). For example, The Cancer Genome Atlas (TCGA) Research Network published a molecular classification of glioblastoma based on expression of signature genes (Verhaak et al., 2010). Of the four subgroups identified (named based on prior work and genes expressed as proneural, neural, mesenchymal, and classic subtypes), the proneural group showed distinctive clinical characteristics including overrepresentation among younger patients, mutations in IDH1, and prior or concomitant areas with low-grade histology—indicating that these are secondary glioblastomas (Verhaak et al., 2010). There also appears to be a CpG island methylator phenotype (CIMP) of glioma belonging to the same proneural molecular subgroup that is common among low-grade gliomas, with a high proportion showing mutations in IDH1 (Noushmehr et al., 2010). While perhaps allowing more coherent selection of potential targets for molecular therapy and validating the role of surrogate markers such as IDH1 for identifying prognostically significant subgroups of glioma, molecular profiling remains largely investigative to date.