Neurooncology

Abstract

General session 4 of a 2010 scientific symposium sponsored jointly by the Society of Toxicologic Pathology and the International Federation of Societies of Toxicologic Pathologists considered the problem of primary neoplasms of the nervous system, especially the brain and spinal cord. Primary tumors of the nervous system account for a small but persistent fraction of cancers, occur disproportionately in juveniles, and are notoriously difficult to treat. This session explored new ideas and techniques with respect to the etiopathogenesis, diagnosis, and treatment of neural tumors with an emphasis on the translation of animal models of brain neoplasia and toxicology investigations for candidate therapeutic agents to the human situation. Additionally, this session covered current thinking on pharmaceutical discovery and safety assessment that is unique to this area of oncology.

General session 4 of the 2010 scientific symposium on “Toxicologic Neuropathology,” sponsored jointly by the Society of Toxicologic Pathology (STP) and the International Federation of Societies of Toxicologic Pathologists (IFSTP), discussed current concerns with respect to primary neoplasia of the nervous system, especially the brain and spinal cord. Primary tumors of this system are uncommon cancers, but their tendency to occur disproportionately in juveniles and their resistance to common antineoplastic strategies and treatments requires that they remain a persistent focus of modern biomedical research. This session explored a spectrum of topics that encompassed current concepts regarding the pathogenesis, diagnosis, and treatment of neural tumors, with an emphasis on translating animal models of brain neoplasia to the human situation.

The first speaker, Dr. Klaus Weber (Harlan Laboratories, Itingen, Switzerland), presented a comprehensive review of the current classification system for nervous system neoplasms in laboratory animals. Neoplasms of the nervous system, whether spontaneous or induced, are rare in laboratory animals (Harlan 2008). In the commonly used laboratory species, primary neural cancers are seen most frequently in rats, followed by mice and hamsters; they are very rare in monkeys and dogs at the ages usually employed in laboratory studies. The morphology of neural tumors depends on the intrinsic functions and/or properties of the neoplastic cell type rather than the size or location of the tumor. Furthermore, interactions between the neoplasm and surrounding normal tissue, as well as regressive changes (e.g., neuronophagia and Rosenthal fiber formation), may produce diagnostic features that are typical of a particular kind of brain neoplasm. Although histological appearances are important in the diagnosis, immunohistochemical analyses may be required to confirm the lineage of origin. The incidence of neural neoplasms can vary with sex, age of tumor onset, and the location of the mass. For example, central nervous system (CNS) neoplasms are recorded at slightly higher incidences in male than female rats (Brix et al 2005; Harlan 2008). It was particularly noteworthy that Dr. Weber’s data indicated important strain differences in the incidences of specific tumor types. For instance, F344 (National Toxicology Program 2004a) and Sprague-Dawley rats have astrocytomas and gliomas at higher incidences than granular cell tumors, whereas the reverse is true with CRL:Han(WI) and Rcc:Han:WIST rats (Harlan 2008). Although the precise onset of spontaneous tumor development is difficult to establish in routine carcinogenicity studies, calculations using the time of diagnosis (day of death) reveal significant differences between tumor types in rats. With increasing animal age, the following statistically significant trends have been observed: oligodendroglial tumors occur earliest, followed by astroglial tumors, granular cell tumors (a type of meningioma), and other meningiomas. Overall, granular cell tumors are the most common CNS neoplasm in rats followed by glial cell tumors, whereas in mice, glial cell tumors are the most frequently reported class of primary neural neoplasia (National Toxicology Program 2004b). Central nervous system tumors are generally less common in mice compared to rats and may also include meningioma, pinealoma, pituicytoma, craniopharyngioma, teratoma, primitive multipotent tumors, schwannoma/neurofibroma, primary CNS lymphomas, and, in certain mouse strains, lipoma. In mice, the hamartomatous proliferative lesion “meningioangiomatosis” can be mistaken for a malignant meningioma. In rodents, primary and secondary nervous system tumors are seen most frequently in the brain compared to the spinal cord. Neural neoplasms are infrequent in the peripheral nervous system (PNS), with the exception of pheochromocytoma (affecting the adrenal medulla). The second most common recorded neoplasm in the PNS is the schwannoma, which may affect any organ. Other neural crest tumors (e.g., paraganglioma, gangliocytoma, melanoma/amelanotic melanoma) have also been recorded in the abdominal cavity, adrenal medulla, eye, skin, and less commonly, in the ganglia associated with particular organs (e.g., the thyroid gland).

Dr. Weber’s session took many attendees to our core training at the level of the light microscope by methodically dissecting the individual morphological features of a tumor and then eventually constructing an appropriate diagnosis. The recommended approach to the morphological dissection of laboratory animal brain neoplasms is to logically step through the various architectural and cellular alterations that make up the tumor. Level 1 architecture consists of the pathognomonic and core nonpathognomonic features of a tumor cell population. Appropriate emphasis was placed on the pathognomonic class of features as an intrinsic property of the tumor, such as nuclear palisading in schwannomas and psammoma bodies in meningiomas. Numerous examples of nonpathognomonic but still helpful morphological features were also presented, such as the radiating crowns of ependymomas, the “honeycomb” appearance of oligodendrogliomas, and the clustering of neoplastic astrocytes in various gliomas. Of course, no review of CNS tumors would be complete without a discussion of the difference between rosettes (i.e., a rose-like accumulation of tumor cells in a circular pattern) and pseudorosettes (i.e., tumor cells attaching to perivascular tissue). Level 2 architecture was described as a continuation of the nonpathognomic morphological dissection through recognition of essential features, such as perineuronal satellitosis and pseudopalisading of neoplastic cells around necrotic foci in astrocytomas. Finally, Level 3 architecture is a spectrum of reactive changes produced by non-neoplastic elements in the brain, such as hemorrhage and neuronophagia; these are usually nonpathognomic. A noteworthy example was the vascular convolutions observed at the periphery of oligodendrogliomas in the rat.

Once this tri-level scheme had been developed for the morphological descriptive process, Dr. Weber’s presentation took the audience on a tour of CNS tumors, complete with key histological features supplemented with immunohistochemical (IHC) diagnostic techniques. Examples of tumors that were discussed included the enigmatic rodent granular cell tumor with periodic acid-Schiff (PAS)–positive granules, astrocytomas that express the immunohistochemical (IHC) markers ED1 and RCA1, and oligodendrogliomas that are positive by IHC for carbonic anhydrase C. Other colorful examples of CNS tumors included ependymomas with characteristic rosettes, choroid plexus carcinomas with distinctive papillary formation, and meningiomas with prominent desmosome staining. Numerous other CNS and PNS tumor types were also reviewed.

The second speaker, Dr. Jeffrey Raizer (Feinberg School of Medicine, Northwestern University, Chicago, IL), extended the theme by discussing the challenges faced by programs engaged in drug development for neuro-oncology indications. Approximately 22,000 new cases of brain cancer arise per year in the United States (Brem et al. 2008). Of these brain tumors, astocytomas and glioblastomas in combination represent the largest tumor population; meningiomas are a close second (Grimm and Raizer 2008), and gliosarcomas are a rare but noteworthy CNS tumor (Han et al. 2010). Prognosis for a given neoplasm type closely correlates with the WHO grading scheme, with a wide range of median survival times (Louis et al. 2007). A trifecta of barriers has contributed to the very slow progress in treating brain cancers: (1) limited animal models, (2) a complex and poorly understood target organ, and (3) tumor cells with an infiltrative pattern that can evade treatment behind multiple anatomical barriers. Most oncologists treat gliomas with a combination of radiation, surgery, and chemotherapy. However, the sensitivity of normal neurons to radiation and the difficulty in identifying surgical margins currently makes chemotherapy the only hope for successful long-term treatment (Duran and Raizer 2007; Raizer et al. 2010).

Most existing animal models used to translate preclinical discoveries into human clinical trials are based on direct intracranial introduction of cultured tumor cells (often as human tissue xenografts) or by genetic engineering to create an accelerated incidence of “natural” tumor induction and progression by indigenous mouse neural cells. These models, coupled with standard human tumor xenografts in the subcutis and in vitro studies using cultured human brain tumor cells, form the foundation for the current research paradigm for finding new drugs to fight brain cancers in humans. Drugs directed against primary CNS neoplasia can be administered peripherally (oral or intravenous) or locally (into the tumor or the parenchyma adjacent to the tumor), and each of these routes has advantages and disadvantages. The ideal therapeutic agent would be a tumoricidal agent that is amenable to peripheral administration but would nonetheless be able to readily cross the relevant blood–neural barriers to preferentially accumulate within or near the vicinity of the cancer cells. Like the treatments for non-CNS cancers, therapies for primary CNS tumors are associated with multiple peripheral side effects (e.g., rash, myelosuppression, and diarrhea). In addition, neurologic side effects specific for patients with these brain malignancies (e.g., seizures, somnolence, and hearing loss) may develop according to either an acute or a chronic time course. The CNS, and to a lesser extent the PNS, are privileged sites protected from exposure to systemic toxicants, including anticancer agents, by the blood–brain and blood–nerve barriers. Thus, many agents developed for peripheral administration will not penetrate into the CNS owing to issues of size, molecular charge, and/or fat solubility. In addition, the abnormal vasculature within brain tumors, increased interstitial pressure within tumors, and molecular mechanisms of tumor resistance (e.g., expression of O-6-methylguanine-DNA methyltransferase [MGMT], P-glycoprotein) may interfere with therapeutic agent efficacy. Each of these issues alone and all of them in aggregate are reasons for the limited arsenal of therapeutic agents that is available to treat primary brain tumors in humans despite the initial promise that is held out in cell line data and current animal models.

The third talk, entitled “Identification of Therapeutic Targets for Glioma,” was delivered by Dr. Heidi Phillips (Genentech, Inc., South San Francisco, CA). Disease heterogeneity, particularly the divergence in gene expression among tumors and among distant sites within the same neoplasia, poses a key challenge in the development of effective therapies for brain tumors in human patients (Parsons et al. 2008). In recent years, molecular profiling has revealed the existence of distinct subtypes of high-grade gliomas, and numerous studies demonstrate prognostic significance of particular markers or gene expression signatures (Brix et al. 2008; Phillips et al 2006; Yan et al. 2009). To identify drug targets, recent work has focused on the identification of key molecular events associated with specific subsets (as defined by molecular markers) of these high-grade gliomas. For example, emerging work reveals that mutations in isocitrate dehydrogenase 1(IDH1) or IDH2 are present at extremely high frequencies in secondary glioblastoma multiforme (GBM), which occur in young patients and often have p53 mutations, as well as in low-grade gliomas that can eventually progress to GBM. Glioblastoma multiforme with activating alterations in PI3K/AKT pathway are associated with a poor prognosis; however, the IDH1 mutation is a strong positive prognostic indicator in GBM. Both histology and magnetic resonance imaging studies between neoplasms with IDH1 mutations and wild-type tumors demonstrate marked differences in the degree of necrosis, edema, cyst formation, vascular abnormalities, and even lobar location within the brain. In particular, IDH1 mutant GBM exhibit features that are common to less aggressive, lower-grade lesions (Dang et al. 2009). These findings suggest that the more aggressive therapeutic strategies that are the standard of care for the treatment of primary GBM are less likely to be effective for the less-aggressive gliomas that harbor IDH1 mutations. It is now becoming apparent that the presence of the IDH1 mutation in GBM tumors represents an improved prognosis based on these phenotypic differences and clinical outcomes. The role of IDH1 mutation in gliomagenesis and its potential as a drug target are areas of ongoing investigation.

In contrast, the majority of de novo cases of primary GBM, which typically affect older patients, have been known for some time to show activating alterations of the PI3K/AKT signaling pathway and often also PTEN mutations. Gene mapping studies of primary GBM within this subcategory have also revealed three distinct gene profiles: proliferative, mesenchymal, and proneural. These subdivisions can be further defined by their biomarker profiles, such as increased proliferating cell nuclear antigen (PCNA) expression in the proliferative group, increased vascular endothelial growth factor (VEGF) expression in the mesenchymal group, and increased epithelial growth factor receptor (EGFR) expression in the proneural group. Such clear molecular subcategorization of GBM clearly suggests that different therapeutic targets might selectively improve the efficacy of treatment for different subsets of GBM.

Further work on correlating gene expression patterns with clinical outcomes to various treatments has demonstrated marked differences in responsiveness to current therapies. As mentioned above, individuals whose tumors have IDH1 mutations did not benefit from increased frequency of current standard-of-care therapies. This observation begs the question as to whether or not primary and secondary GBM tumors are derived from the same cell of origin. This tumor heterogeneity may be associated with the tumor stem cell hypothesis, which states that within a given tumor, a specified subset of tumor cell precursors possesses the capability to self-renew and reconstitute the heterogeneous lineages of cancer cells that make up the tumor. The current cell surface marker that is thought to best identify the neural stem cell is CD 133 (Chen et al 2010). Current investigations and animal model development are ongoing to identify and characterize the role of this elusive cell within the milieu of nascent and advanced brain tumors.

The final speaker, Dr. Eric Holland (Memorial Sloan Kettering Cancer Center, New York, NY), addressed the comparative pathology of brain neoplasia in humans versus animal models. Malignant glial cell tumors are the most aggressive and difficult to treat neoplasms that arise in the brain. Most people diagnosed with malignant glioma will die within two years following diagnosis. Traditionally, gliomas have been categorized based solely on tumor histological features. However, recent expression studies have found that molecular signatures can be used to categorize these tumors into subclasses that more effectively predict a patient’s clinical outcome. The heterogeneity among tumors as well as at different locations within individual tumors makes understanding the molecular aspects of tumorigenesis extremely important (Huse and Holland 2010). The importance of distinguishing primary GBM tumors with PI3K/AKT activation, EGFR amplification, and PTEN mutations versus secondary GBM tumors with p53 and IDH-1 mutations was again emphasized (Verhaak et al. 2010). In addition to alterations in the PI3K/AKT pathway, alterations of the SHH/Wnt and Notch pathways along with the platelet-derived growth factor (PDGF) receptor are also considered potential targets of pathogenic importance for gliomas.

The challenge for researchers today to advance our understanding of both glioma pathogenesis and its treatment hinges on the development of better animal models (Momota and Holland 2009). The complex function of the central nervous system and its unique anatomical barriers has thus far inhibited realistic chemically induced or implanted tumor models in mice. Current animal model development of gliomas is using a molecular approach by isolating and cloning previously described genes associated with this tumor class, inserting the gene (e.g., PDGF) into a murine retrovirus vector, and then injecting the vector into the frontal lobe of neonatal mice. Histopathology of these genetically engineered mouse models (GEMM) taken three to six months post-inoculation has demonstrated GBM tumors with a remarkable resemblance to the morphological characteristics of wild-type tumors from human patients. The chosen molecular mutation that is inserted means that GEMM of a glioma also recapitulate the molecular alterations observed in the human disease.

The final topic for discussion went back to the question of stem cells and their role in brain tumors. Data from flow cytometric examination of viral vector–induced gliomas demonstrated that nestin, a putative cancer stem cell marker, can be identified in a small subset of tumor cells known as side population cells. Furthermore, the expression of nestin can be altered when tumor-bearing mice are treated with routine cytotoxic agents indicated for GBM therapy. So where do these stem cells live? Confocal microscopy experiments addressing this question are starting to point to the increased presence of nestin-bearing and Notch-activated cells (i.e., likely cancer stem cells) in the perivascular space (Charles and Holland 2010). Why would a stem cell prefer to live at this precise site? Presumably they benefit from trophic factors that collect abundantly in this location as well as proximity to a steady supply of vascular-borne nutrients. One such nutrient or trophic factor might be nitric oxide, which is abundant in the perivascular space. In vitro data was shown in which cells exposed to nitric oxide donor molecules have activated Notch pathway signaling which, when subsequently transplanted to mice, results in significantly increased tumor incidence, which represents another putative small step in the complex and multifactor pathogenesis of these tumors.

Current therapeutic options for brain neoplasia often have limited efficacy or are associated with unacceptably severe adverse events. Additionally, despite distinct altered molecular pathways in primary versus secondary GBM, chemotherapeutic agents currently indicated for gliomas are not used in a pattern-specific manner. This is clearly recognized as a shortcoming in our current approach to brain cancer therapy. Targeted molecular therapies (e.g., EGFR) were described as ongoing and represent our best hope for improved efficacy in brain cancer therapy. Toxicities experienced by brain cancer patients are reminiscent of other neurotoxicities described in the scientific literature. These effects include leukoencephalopathy, cerebral infarction/hemorrhage, aseptic meningitis, cerebral atrophy, and calcification. Such morphological indices of toxicity can be clinically manifested as various encephalopathies including delirium, dementia, headaches, ataxia, ototoxicity, vision loss, peripheral neuropathies, and/or seizures. Various non–nervous system adverse events are also possible following administration of agents that attack brain tumors, including myelosuppresion with cytotoxic drugs, hypertension with VEGF inhibitors, and hypersensitivity with monoclonal antibodies. Importantly for the toxicology professional in the field of pharmaceutical drug safety, it was noted that combination therapies will continue to be a key strategy in cancer therapy. Thus, further refinement of predicting and managing drug-associated toxicities will be increasingly important in the future.

Like other areas of cancer, these problems highlight our lack of knowledge of the complete pathogenesis of this family of diseases. Categorization and subsequent treatment of brain cancer is still largely based on histological categorization by traditional morphological criteria. It is widely agreed that histopathology is an essential starting point in assessing and treating brain tumors. However, successful treatment of neural cancer in the future will demand that these morphological features become better integrated with the myriad of genetic and biochemical alterations that characterize the various neoplastic cells that make up a tumor, and that the entire data set be more closely linked to the prognostic outcome. This session reaffirms that brain cancer researchers are slowly but continually moving toward this elusive goal. Toxicologic pathologists continue to be integral members of the research teams who are pursuing this goal.