Significance of immunohistochemistry for neuro-oncology. VI. Occurrence, localization and distribution of glial fibrillary acid protein (GFAP) in 820 tumors

In a retrospective study 820 tumors were immunohistochemically examined with anti-GFAP. All 224 astrocytomas and 105 of 112 glioblastomas were, at least focally, positive. 72% of ependymomas and 64% of oligodendrogliomas contained tumor cells which expressed GFAP. In such entities the reaction is dependent on the histologic subtype. Only 26 of 114 medulloblastomas (22.8%) demonstrated scattered GFAP positive cells. GFAP was also demonstrated in the CNS in gangliogliomas, monstrocellular sarcomas, 3 of 6 PNET, one non-classifiable tumor in a child, 1 plexus papilloma, in scattered stromal cells in 15 of 26 hemangioblastomas as well as in the mature glial component of intracranial germ cell tumors. Outside of the CNS there was evidence of GFAP in 3 cases with nasal glial heterotopy and in the myxoidal part of a pleomorphic salivary gland adenoma. Neoplasms which proved negative to GFAP in our series included purely neural differentiated tumors meningioma, neurolemmomas, chordomas, paragangliomas, sarcomas, lymphomas, melanomas and carcinoma metastases. Separating GFAP-positive reactive astrocytes from the actual tumor cells has proved to be a problem in the routine use of GFAP in differential diagnosis. Absence of an immunohistochemical response does not exclude a tumor of glial origin. Tissue samples which are too small, particularly in the case of anaplastic astrocytomas and glioblastomas can give false negative results.     

A simple method for the isolation of GFAP and its use for the study of brain tumors

A simple method is described in this paper for the production of a polyclonal antiserum against GFAP. The antiserum was tested on 212 primary brain tumours which had been selected from biopsy and autopsy material of the Institute of Pathological Anatomy at the Medical Academy of Erfurt, GDR. 52 of 81 astrocytomas (64%) and 26 of 47 glioblastomas (55%) gave GFAP-positive results. GFAP-negative responses were primarily recorded from tumours with severe anaplasia. GFAP was found in all 22 ependymomas tested. Epithelioid ependymomas, however, exhibited lower immunological reactions than tanycytic variants. Isomorphic oligodendrogliomas, meningiomas, medulloblastomas, and brain metastases of carcinomas were GFAP-negative. The possibility is discussed in some detail of falsely negative results on account of too little biopsy material or insufficient fixation of tumour tissue.     

100例中枢神经系统肿瘤的胶质纤维酸性蛋白免疫组化观察

对100例中枢神经系统肿瘤进行GFAP免疫组化观察。结果表明,GFAP主要分布于各种类型的星形细胞瘤,多形性胶质母细胞瘤,混合性胶质瘤及部分室管膜瘤。GFAP染色强度与瘤细胞分化程度有关。本文还探讨了室管膜瘤与少突胶质瘤存在GFAP问题。研究表明GFAP可作为诊断神经胶质瘤的一种有用标记物。     

Cell proliferation and glial fibrillary acidic protein in brain tumors

The results of histoautoradiographic and immunohistochemical studies of biopsy specimens of 15 brain tumours are reported. The specimens were labeled with 3H-thymidine using an in vitro technique. Meningiomas, oligodendrogliomas and well differentiated astrocytomas showed a median S-phase fraction of about 1%. In contrast, the labeling indices of 4 from 7 anaplastic astrocytomas were higher (2.1, 3.0, 3.5, 11.4). With increasing degree of malignancy the proliferative heterogeneity of the tumours increases. In every glioma varying amounts of glial fibrillary acidic protein (GFAP) were detected immunohistochemically (PAP technique). In 3 high-grade gliomas (2 glioblastomas, 1 anaplastic astrocytoma) an inverse relation of the investigated parameters (high S-phase fraction, low GFAP expression) was found. An exact prediction on biological behaviour of an individual tumour by GFAP detection immunohistochemically is not possible, because a high GFAP content can be detected also in some malignant tumours. However, the 3H-thymidine labeling indices of viable parts of the tumours, probably reflecting the growth fraction seem to be clinically important parameters, especially in respect to the prognosis.     

Importance of immunohistochemistry for neuro-oncology. III. Demonstration of glial fibrillary acidic proteins (GFAP) in anaplastic astrocytomas and glioblastomas

Expression of gliofibrillary acidic protein in 23 anaplastic astrocytomas and 33 glioblastomas has been investigated and correlated with tumor behavior as reflected in both the length of the preoperative history and in the post-operative survival time. Three degrees of positive immunoreactivity to anti-GFAP can be distinguished: positive GFAP reaction in more than 2/3 of cells; in 1/3 to 2/3 of all cells; in less than 1/3 of all cells; negative reaction. All anaplastic astrocytomas and 27 of 33 glioblastomas showed GFAP positive reactions. The proportion of highly reactive tumors is higher by anaplastic astrocytomas than by glioblastomas (7 of 33). For both astrocytomas and glioblastomas there is a tendency for a decrease in the expression of GFAP to be associated with a shorter preoperative history and with a shorter survival time. This is more prominent for astrocytomas than for glioblastomas. This finding supports the opinion expressed in previous publications that the GFAP expression is reversely related to the level of tumor anaplasticity.     

Neurospecific proteins: potentials and prospects for their use in morphological research on tumors

Data on the application of neurospecific proteins S-100, GFAP, D2 glycoprotein and neuron-specific enolase (NSE) in the differential tumor diagnosis are reviewed. S-100 protein and GFAP are found in well differentiated astroglial tumors. S-100 protein can be used as melanoma and Schwannoma specific marker. In malignant CNS tumors there is a decrease of S-100 protein content up to its complete disappearance, while the content of GFAP is variable. D2 glycoprotein is detected in gliomas and medulloblastomas, being absent in other brain tumors. NSE is invariably present in apudomas and was also found in the majority of investigated astrocytomas, ependymomas, glioblastomas and in some medulloblastomas.     

Keratin expression in astrocytomas: an immunofluorescent and biochemical reassessment

 Several studies have shown that immunoenzymatic staining of formalin-fixed, paraffin-embedded astrocytomas with keratin antibodies frequently yields positive labelling, but no biochemical evidence of keratin expression in astrocytomas has been reported. We have investigated the presence of keratin in astrocytoma and normal brain tissues both by immunofluorescence on frozen sections and by 1D and 2D immunoblotting using seven monoclonal antibodies that, collectively, recognize most keratin polypeptides. Four of these antibodies did not stain neural tissues by immunofluorescence and were also negative by immunoblotting. The remaining three keratin antibodies stained normal brain and/or a high proportion of astrocytomas. Two of these three antibodies only stained glial fibrillary acidic protein (GFAP)-positive cells, while the third only stained GFAP-negative cells. 1D and 2D immunoblotting analysis showed that positive immunofluorescence staining of normal brain and/or astrocytomas seen with these three keratin antibodies was due to cross-reactivity with non-keratin proteins, such as GFAP. These results demonstrate that, contrary to earlier suggestions, keratin polypeptides are not frequently expressed in astrocytomas. Our studies also emphasize that keratin antibodies should be used cautiously for the differential diagnosis of undifferentiated gliomas from tumours of non-glial origin. Received: 28 October 1996 / Accepted: 19 February 1997     

An immunohistochemical study on the distribution of glial fibrillary acidic protein, S-100 protein, neuron-specific enolase, and neurofilament in medulloblastomas

In order to clarify the differentiation of medulloblastomas, the authors studied on the morphological features and immunohistochemical expression of glial fibrillary acidic protein (GFAP), S-100 protein, neuron-specific enolase (NSE), and neurofilament (NF) in 31 medulloblastomas. GFAP was detected only in a small number of tumor cells of 5 medulloblastomas; S-100 protein in both small tumor cells and some so-called spongioblastic cells in 16 medulloblastomas; NSE in the more abundant tumor cells and the matrix in 28 medulloblastomas; NF in a few tumor cells of 12 medulloblastomas; GFAP and NF in 2 medulloblastomas, but each of them in different tumor cells. These results suggest that medulloblastomas have a capacity of differentiation along neuronal and/or glial lines. The conventional morphological markers of differentiation in medulloblastomas such as spongioblastic cells and Homer Wright rosettes were not necessarily compatible with expression of immunohistochemical markers such as GFAP or NF. NSE and S-100 protein seem less valuable markers of differentiation because they were detected in both neuronal and glial elements. But NSE, which was observed in most medulloblastomas, might have a value as a marker for medulloblastomas.     

Neurospecific Proteins in the Serum of Patients with Brain Tumors

Neurospecific proteins S-100 and GFAP were measured in the serum of 145 patients with neural tumors and 69 healthy individuals. In patients with glyoblastomas, the concentrations of S-100 and GFAP were significantly higher than in patients with anaplastic astrocytomas, benign meningiomas, and brain metastases and in healthy individuals. Serum S-100 concentrations in patients with anaplastic astrocytomas, benign meningiomas, and brain metastases were similar; significant difference from the control was found only for patients with cerebral metastases. A specific feature of GFAP was high incidence of its detection in patients with glioblastomas (83%) compared to other groups of patients with neural tumors and healthy volunteers who demonstrated practically zero level of this protein. These findings attest to the possibility of using S-100 as an additional biochemical criterion of brain involvement in tumor patients and GFAP as a glioblastoma marker.     

AN IMMUNOHISTOCHEMICAL STUDY ON THE DISTRIBUTION OF GLIAL FIBRILLARY ACIDIC PROTEIN, S-100 PROTEIN, NEURON-SPECIFIC ENOLASE, AND NEUROFILAMENT IN MEDULLOBLASTOMAS

In order to clarify the differentiation of medulloblastomas, the authors studied on the morphological features and immunohistochemical expression of glial flbrillary acidic protein (GFAP), S-100 protein, neuron-specific enolase (NSE), and neuroftlament (NF) in 31 medulloblastomas. GFAP was detected only in a small number of tumor cells of 5 medulloblastomas; S-100 protein in both small tumor cells and some so-called spongloblastic cells in 16 medulloblastomas; NSE in the more abundant tumor cells and the matrix in 28 medulloblastomas; NF in a few tumor cells of 12 medulloblastomas; GFAP and NF in 2 medulloblastomas, but each of them in different tumor cells. These results suggest that medulloblastomas have a capacity of differentiation along neuronal and/or glial lines. The conventional morphological markers of differentiation in medulloblastomas such as spongioblastic cells and Homer Wright rosettes were not necessarily compatible with expression of immunohistochemical markers such as GFAP or NF. NSE and S-100 protein seem less valuable markers of differentiation because they were detected in both neuronal and glial elements. But NSE, which was observed in most medulloblastomas, might have a value as a marker for medulloblastomas.     

Expression of Matrix Metalloproteinases and Their Tissue Inhibitors in Human Brain Tumors

In this study, we investigated the expression patterns of 15 matrix metalloproteinases (MMPs) and three tissue inhibitors of metalloproteinase in gliomas, medulloblastomas, and normal brain tissue. By Northern blot analysis we found increased levels of mRNAs encoding for gelatinase A, gelatinase B, two membrane-type MMPs (mt1- and mt2-MMP), and tissue inhibitors of metalloproteinase-1 in glioblastomas and medulloblastomas. We observed a significant increase of mt1-MMP, gelatinase A, gelatinase B, and tissue inhibitors of metalloproteinase-1 in glioblastomas as compared with low-grade astrocytomas, anaplastic astrocytomas, and normal brain. In medulloblastomas, the expression of mt1-MMP, mt2-MMP, and gelatinase A were also increased, but to a lesser extent than that observed in glioblastomas. These data were confirmed at the protein level by immunostaining analysis. Moreover, substrate gel electrophoresis showed that the activated forms of gelatinases A and B were present in glioblastomas and medulloblastomas. These results suggest that increased expression of mt1-MMP/gelatinase A is closely related to the malignant progression observed in gliomas. Furthermore, the present study demonstrates, to our knowledge for the first time, that medulloblastomas express high levels of MMP.     

Molecular markers of primitive neuroectodermal tumors and other pediatric central nervous system tumors. Monoclonal antibodies to neuronal and glial antigens distinguish subsets of primitive neuroectodermal tumors

Seventy-one tumors of the central nervous system in children were studied immunohistologically. Thirty-seven were classified histologically as PNETs, of which 35 were located in the cerebellum (medulloblastomas), one in the cerebrum, and one in the spinal cord. The 34 non-PNETs included five ependymomas, seven gangliogliomas, 15 astrocytomas, and seven tumors of other histology. We used monoclonal antibodies specific for neurofilament (NF) triplet proteins, for microtubule associated protein 2 and tau protein and for glial fibrillary acidic protein (GFAP) and myelin basic protein. In addition, a monoclonal antibody to epithelial membrane antigen was applied. The presence or absence of these antigens defined four major groups of PNETs: 1) PNETs not otherwise specified (10 cases), 2) PNETs with neuronal differentiation (eight cases), 3) PNETs with astrocytic differentiation (six cases), and 4) PNETs with both neuronal and astrocytic differentiation (12 cases). One case showed ependymal differentiation. The pattern of expression of NF isoforms in PNETs was reminiscent of that seen during normal mammalian development, such that phosphorylated NF-H was only present in combination with NF-M and NF-L. Among the other central nervous system tumors, all astrocytomas and gangliogliomas were positive for GFAP, and the gangliogliomas also expressed all NF isoforms. Three atypical teratoid tumors and two rhabdoid tumors showed strong positivity for epithelial membrane antigen and also for GFAP. We conclude that the differentiation antigens described here serve to distinguish PNETs from other pediatric central nervous system tumors and to identify subsets of PNETs. Accordingly, PNETs represent a heterogeneous group of pediatric brain tumors capable of neuronal and glial differentiation.     

Aberrant localization of the neuronal class III beta-tubulin in astrocytomas

BACKGROUND: The class III beta-tubulin isotype (betaIII) is widely regarded as a neuronal marker in development and neoplasia. In previous work, we have shown that the expression of betaIII in neuronal/neuroblastic tumors is differentiation dependent. In contrast, the aberrant localization of this isotype in certain nonneuronal neoplasms, such as epithelial neuroendocrine lung tumors, is associated with anaplastic potential. OBJECTIVE: To test the generality of this observation, we investigated the immunoreactivity profile of betaIII in astrocytomas. DESIGN: Sixty archival, surgically excised astrocytomas (8 pilocytic astrocytomas, WHO grade 1; 18 diffuse fibrillary astrocytomas, WHO grade 2; 4 anaplastic astrocytomas, WHO grade 3; and 30 glioblastomas, WHO grade 4), were studied by immunohistochemistry using anti-betaIII monoclonal (TuJ1) and polyclonal antibodies. A monoclonal antibody to Ki-67 nuclear antigen (NC-MM1) was used as a marker for cell proliferation. Antibodies to glial fibrillary acidic protein (GFAP) and BM89 synaptic vesicle antigen/synaptophysin were used as glial and neuronal markers, respectively. RESULTS: The betaIII immunoreactivity was significantly greater in high-grade astrocytomas (anaplastic astrocytomas and glioblastomas; median labeling index [MLI], 35%; interquartile range [IQR], 20%-47%) as compared with diffuse fibrillary astrocytomas (MLI, 4%; IQR, 0.2%-21%) (P <.0001) and was rarely detectable in pilocytic astrocytomas (MLI, 0%; IQR, 0%-0.5%) (P <.0001 vs high-grade astrocytomas; P <.01 vs diffuse fibrillary astrocytomas). A highly significant, grade-dependent relationship was observed between betaIII and Ki-67 labeling and malignancy, but this association was stronger for Ki-67 than for betaIII (betaIII, P <.006; Ki-67, P <.0001). There was co-localization of betaIII and GFAP in neoplastic astrocytes, but no BM89 synaptic vesicle antigen/synaptophysin staining was detected. CONCLUSIONS: In the context of astrocytic gliomas, betaIII immunoreactivity is associated with an ascending gradient of malignancy and thus may be a useful ancillary diagnostic marker. However, the significance of betaIII-positive phenotypes in diffuse fibrillary astrocytomas with respect to prognostic and predictive value requires further evaluation. Under certain neoplastic conditions, betaIII expression is not neuron specific, calling for a cautious interpretation of betaIII-positive phenotypes in brain tumors.     

Antigenic staining patterns of human glioma cultures: primary cultures, long-term cultures and cell lines

Summary The immunocytochemical staining patterns of cultured glioma cells were investigated. Fifty nine individual cases were stained at differentin vitro ages for glial fibrillary acidic protein, fibronectin, galactocerebroside, HNK-1/Leu 7, A2B5, vimentin, factor VIII and A4. Histologically, the cases were composed of eight low-grade astrocytomas, 11 high-grade astrocytomas, four low-grade oligodendrogliomas, seven high-grade oligodendrogliomas and 29 glioblastomas. The 45 cases were analysed within the first 3 weeks of culture, many of them as primary cultures. In 11 cases stainings were performed repeatedly at intervals of up to 6 months.Glial fibrillary acidic protein staining was positive in most of the early cultures of astrocytomas (low and high grade) and glioblastomas; expression in more than 50% of the cells was found in 1 of 5 low-grade astrocytomas, 5 of 11 high-grade astrocytomas and 14 of 29 glioblastomas. Two of the high-grade astrocytomas were stained once more after 6 weeks in culture and were found to be only 1% positive for glial fibrillary acidic protein but strongly positive for fibronectin. The same was true for five of the glioblastoma cases. Two of these cases remained glial fibrillary acid protein positive and developed into stable permanent cell lines. Only one case started with 1% of glial fibrillary acidic protein positive cells and later developed into a 99% glial fibrillary acidic protein positive cell line. Neither HNK-1/Leu 7 expression nor A2B5 staining appeared to have a relationship to the glial fibrillary acidic protein staining. It was observed that glial fibrillary acidic protein and HNK-1/Leu 7 were both 100% in some cases but that later one of the two antigens disappeared but not the other. The amount of glial fibrillary acidic protein staining does not allow the prediction of A2B5 staining.The study shows that initiation of primary cultures on an extracellular matrix yields more glial fibrillary acidic protein positive cells in primary cultures than have been found in other studies. It is concluded that only a rigid standardization of culture conditions will ensure the validity of comparisons ofin vitro data obtained in primary cultures.     

Nestin expression in neuroepithelial tumors

Nestin is a marker of early stages of neurocytogenesis. It has been studied in 50 neuroepithelial tumors, mostly gliomas of different malignancy grades, by immunohistochemistry, immunofluorescence, immunoblotting, and confocal microscopy and compared with GFAP and Vimentin. As an early marker of differentiation, Nestin is almost not expressed in diffuse astrocytomas, variably expressed in anaplastic astrocytomas and strongly and irregularly expressed in glioblastomas. Negative in oligodendrogliomas, it stains ependymomas and shows a gradient of expression in pilocytic astrocytomas. In glioblastomas, Nestin distribution does not completely correspond to that of GFAP and Vimentin with which its expression varies in tumor cells in a complementary way, as confirmed by confocal microscopy. Tumor cells can thus either derive from or differentiate toward the neurocytogenetic stages. Hypothetically, they could be put in relation with radial glia where during embriogenesis the three antigens are successively expressed. Completely negative cells of invasive or recurrent glioblastomas may represent malignant selected clones after accumulation of mutations or early stem cells not expressing antigens.     

Expression of urokinase-type plasminogen activator receptor (uPAR) in primary central nervous system neoplasms.

The cellular receptor for urokinase-type plasminogen activator receptor (uPAR) is a member of the glycosylphosphatidylinositol (GPI) anchored protein family. It is a specific cell surface receptor for its ligand, urokinase-type plasminogen activator, which catalyzes the formation of plasmin from plasminogen to generate the proteolytic cascade and leads to the breakdown of the extracellular matrix. uPAR has been shown to correlate with a propensity to tumor invasion and metastasis in several types of non-central nervous system tumors. In this study, the authors examined the immunohistochemical expression of uPAR in 65 primary brain tumors (5 pilocytic astrocytomas, 5 diffuse astrocytomas, 6 anaplastic astrocytomas, 8 glioblastomas, 5 oligodendrogliomas, 4 oligoastrocytomas, 6 anaplastic oligoastrocytomas, 4 gangliogliomas, 4 ependymomas, 5 medulloblastomas, 6 schwannomas, 5 meningiomas, 2 atypical meningiomas). The specimens were evaluated for intensity of immunostaining (0-3 scale), cellular localization of staining, and specific or unique patterns of staining. Some degree of uPAR expression was observed in all tumors. A significant positive correlation (P = 0.0006) between tumor grade and staining intensity was identified within the astrocytoma/glioblastoma subgroup, suggesting a possible correlation with anaplastic change and propensity to tumor invasion. Expression of uPAR in nonmalignant, noninvasive tumors such as schwannoma and meningioma suggests that uPAR may have other biologic functions in addition to promotion of tumor invasion.     

Products of cells cultured from gliomas. VII. Extracellular matrix proteins of gliomas which contain glial fibrillary acidic protein

Extracellular matrix (ECM) components of two glial fibrillary acidic protein positive (GFAP+) glioma lines U251 and UM6 were studied by silver stain, morphometry, immunofluorescence, enzyme-linked immunosorbent assay, and biosynthetic labeling. Both GFAP+ lines expressed the following qualitative features in common with previously studied GFAP-negative gliomas: (a) laminin, (b) type IV collagen, (c) extracellular fibrils of silver-reducing collagen (d) pattern of reactivity with lectins. Quantitative differences in GFAP+ glioma proteins included less collagen and more laminin than GFAP-negative gliomas. Sparse collagen of GFAP+ gliomas aggregated as extracellular masses. Individual cells of UM6 simultaneously expressed GFAP and mesenchymal ECM components. Results show qualitative similarities of ECM expression among GFAP+ and negative gliomas suggesting a common lineage of these two glioma cell types and universal expression of two epithelial components of ECM, laminin and type IV collagen, among cultured gliomas. Moreover, there is a diversity of quantity and type of ECM proteins of GFAP+ gliomas with the U251 line most restricted in its expression of ECM components and with UM6 manifesting markers of epithelial and mesenchymal lineage. This diversity suggests a capacity for regulation of phenotypic expression of ECM beyond that explained simply by the presence of two cell types of different lineage.     

Glial fibrillary acidic protein in medulloblastomas and other embryonic CNS tumours of children

Summary Investigation of GFAP in 50 medulloblastomas showed a few GFAP-positive tumour cells in 5 cases only; 17 tumours were negative, and 28 showed a pseudopositivity, i.e. GFAP-bearing cells were identified as reactive or degenerating astrocytes, intermingled with tumour elements. A high GFAP content was seen in 2 small-cell gliomas of the cerebellum, whereas 3 pineoblastomas, 2 neuroblastomas of CNS, and one medulloepithelioma were negative.GFAP is a very good method for identificating astrocytes, but does not seem to be reliable for identifying the origin of undifferentiated tumours such as medulloblastomas. In these neoplasms glial differentiation is lacking or extremely rare, GFAP-positivity being mostly an artifact. The investigation of small tumour samples or the positivity of a single cell are inadequate data for a correct evaluation of the findings, especially taking in mind that GFAP of degenerated astrocytes can be phagocytised by cells other than glial (e.g., macrophages, epithelial and meningioma cells). The importance of carefully checking the whole structure of the tumour is stressed, GFAP positivity or negativity being not a sufficient criterion for its nosological classification.     

Morphological maturation of tumor cells induced by ethylnitrosourea (ENU) in rat brains. I. On the tumors by administration of ENU in the late gestational stage

Sequential changes in the development of ethylnitrosourea (ENU)-induced rat brain tumors were examined histologically, immunohistochemically, electron microscopically and autoradiographically. In 47 Sprague-Dawley rats transplacentally administered ENU, 95 brain tumors developed, including 76 microtumors less than 1mm in diameter. Microtumors were found mainly in the paraventricular area, but some were found in the peripheral brain tissue. They were composed of small tumor cells which had round dark nuclei and scanty cytoplasm immunohistochemically negative for Leu 7 and glial fibrillary acidic protein (GFAP). The 19 macrotumors were mature gliomas, 3 of which histologically corresponded to oligodendrogliomas and 16 to mixed gliomas. The tumor cells of the former had small round nuclei with distinct perinuclear halos and a small amount of cytoplasm positive for Leu 7. The latter were chiefly composed of polygonal cells having large round nuclei and rich cytoplasm positive for GFAP. An autoradiographic study using 3H-thymidine revealed that the labeling index of the tumor cells was high in mixed gliomas and microtumors, but low in oligodendrogliomas. It may be concluded that the constituent cells of microtumors correspond to glioblasts or migrating neuroglias, which gradually mature to form oligodendrogliomas or astrocytomas.     

Neuroblastic Differentiation of Metastases of Medulloblastoma to Extracranial Lymph Node: An Ultrastructural Study

Extracranial metastases of a poorly differentiated medulloblastoma in a 12-year-old girl were studied by thin section transmission electron microscopy and immunohistochemistry. The primary tumor did not show any differentiation as revealed by immunohistochemistry. On the contrary, the metastatic tumor cells and their processes disclosed features of neuroblastic differentiation when examined ultrastructurally: microtubules, dense core vesicles, and abortive synaptic ribbons. Several dystrophic neurites containing altered subcellular organelles were also found. Furthermore, few processes contained concentric arrays of paired membranes. This report is the first to clearly show the neuronal differentiation of extracranial metastases of poorly differentiated medulloblastoma. We speculate that metastases of medulloblastomas outside the neuraxis behave analogously to medulloblastoma explants cultured in vivo.