We present 8 examples of a neoplasm with features of both astrocytoma and ependymoma that may represent a distinct clinicopathologic entity. The cerebral hemispheric tumors occurred in patients that were 3, 4, 12, 14, 15, 26, 30, and 37 years of age. All presented with seizures that, with the exception of 2, began in childhood. Magnetic resonance imaging studies showed ill-defined, T2-hyperintense, generally noncontrast-enhancing lesions that, although centered on the cortex or amygdala, extended into the underlying white matter for a short distance. Histologically, the variably infiltrative tumors were distinctively angiocentric with well-developed perivascular pseudorosettes in some cases. Longitudinal and/or circumferential orientations of perivascular cells were common also. The cells were uniform in their cytologic features from case to case and were bipolar in all but one case. A glial nature was inferred from immunoreactivity for GFAP, and ependymal differentiation was suggested by positivity for EMA in three cases and ultrastructural features in one. Overall, the tumors were biologically indolent except for one that recurred and ultimately proved fatal.
In 2002, we presented a series of 3 cerebral hemispheric tumors, which we designated “angiocentric bipolar astrocytoma” (1). Since that time, further study of these and an additional 5 cases have disclosed histologic, immunohistochemical, and ultrastructural expressions of ependymal differentiation as well. The lesion has no counterpart in the World Health Organization (WHO) classification.
Materials and Methods
Clinical data are summarized in the Table.
Case 1 was a 15-year-old girl who originally presented with partial complex seizures and a somewhat ill-defined T2-hyperintense, nonenhancing lesion centered on the right amygdala and medial temporal cortex (Fig. 1A). Abnormal T2 signal was also present in the subjacent white matter. Begun on and then tapered from anticonvulsive drugs, she remained seizure-free until the onset of olfactory hallucinations 4 years later. The lesion, which had not changed radiologically in this interval (Fig. 1B), was then totally resected. The surgeon noted an enlarged hippocampus and a darkened and somewhat indurated amygdala. There has been no clinical or radiologic evidence of recurrence during a 12-month postoperative period. The pathologic diagnosis was “tumor with astrocytic, ependymal, and neuronal lineages-considering a hamartomatous process, pending outside consultation.”
Case 2 was a 4-year-old girl with a 2-month history of intractable seizures and a 1- to 2-cm tumor largely within the gray matter of the right parietal lobe. The mainly intracortical tumor also involved the underlying white matter, where hyperintense signal was present in T2-weighted images. The lesion did not enhance after intravenous injection of gadolinium. The totally resected lesion was originally considered a well-differentiated glioneuronal tumor, and, later, the deeper portion was interpreted as grade III diffuse astrocytoma because of mitoses and a MIB-1 labeling index of 5%. A still later review suggested that the ganglion cells were trapped normal cells, not neoplastic elements. Conformal radiotherapy to the tumor site and then chemotherapy were administered. There has been no clinical or radiologic evidence of recurrence during a 4-year posttreatment period. Pathologic diagnoses rendered by various pathologists were “low-grade glioneuronal tumor,” “grade III diffuse astrocytoma,” and “astrocytoma-difficult to grade but concerned that it could behave as a grade III lesion.”
Case 3 was a 3-year-old boy who presented with a 3-month history of both grand mal seizures and radiologic documentation of a 2.5-cm, noncontrast-enhancing mass in the left occipital lobe. The lesion was largely intracortical with involvement of subjacent of white matter. It had not changed radiologically in a 3-month preoperative interval, nor did residual tumor enlarge in the 3 years after subtotal (90%) resection. Pathologic diagnoses were “astrocytoma, unusual diffuse, not in WHO (World Health Organization)” and “gliomatosis cerebri.”
Case 4 was a 14-year-old boy with a history of seizures with recent exacerbation and a lesion in a single gyrus of the right posterior frontal lobe. The nonenhancing lesion, with only slight mass effect, extended as bright signal in T2-weighted images into the subjacent white matter for a short distance. The lesion was totally resected after a 10-month period of magnetic resonance surveillance during which there was no change in the size. There was also no enlargement during 2 years of postoperative follow up. Pathologic diagnoses were “benign or low grade glioma with features suggestive of hamartoma” and “astrocytoma, but does not conform to classical diffuse astrocytoma.”
Case 5 was a 12-year-old girl with a history of intractable epilepsy for 9 years before excision of a right orbital-frontal mass and superior temporal lobe. Magnetic resonance imaging revealed a nonenhancing, largely intracortical lesion in the inferior frontal and adjacent temporal lobes, which raised the issue of dysembryoplastic neuroepithelial tumor, although the lesion extended into the white matter and lacked the common “bubbly” appearance of DNT in precontrast T1-weighted images. There has been no enlargement during 2 years after partial excision. Pathologic diagnoses were “tumor with some features of ependymoma and some of pleomorphic xanthoastrocytoma, and subependymal giant cell astrocytoma” and “astrocytic neoplasm with unusual angiocentricity.”
Case 6 was a 30-year-old woman with a history of seizures since age 5 associated with a lesion in the left anterior-medial temporal lobe. The patient was free of seizures 4 years after a total resection. The radiologic status is unknown. The pathologic diagnosis was “low grade astrocytoma, predominantly perivascular.”
Case 7 was a 26-year-old man with a 2-year history of seizures before partial removal of a noncontrast-enhancing lesion in the left frontal lobe. No postoperative treatment was administered. Generalized seizures reappeared 21 months later, at which point abnormal T2 signal filled the lobe. The recurrent lesion was partially resected. Based largely on the presence of mitoses, the diagnosis of anaplastic astrocytoma was rendered and the patient received a 6-week course of radiotherapy to 60 Gy in 30 fractions. Chemotherapy (procarbazine, carmustine, and vincristine) was administered, but the tumor ultimately progressed, claiming the patient's life 62 months after the initial resection. Pathologic diagnoses on the initial specimen were “possible congenital/dysplastic lesion or low grade astrocytoma or pilocytic lesion or spongioblastoma” and “glioma.” The recurrent tumor was called “anaplastic astrocytoma (grade III)” and “anaplastic astrocytoma with astroblastoma features.”
Case 8 was a 37-year-old man with seizures since age 19 and a mass in the right frontal lobe. Pathologic diagnoses were “papillary glioneuronal tumor” and “low grade glioma with features of ependymoma and astroblastoma.”
Formalin-fixed, paraffin-embedded tissue was used for routine histologic and immunohistochemical analysis. Standard methodology was used for histochemical and immunohistochemical staining. Sources and dilutions of antibodies used were as follows: glial fibrillary acidic protein (GFAP, 1:6000; DAKO, Carpinteria, CA), synaptophysin (100; DAKO), neurofilament protein (SM 31, 1:10; Sternberger, Lutherville, MD); NeuN (1:1000; Chemicon, Temecula, CA); chromogranin (prediluted; Ventana, Tucson, AZ); MIB-1 (Ki-67) (1:1000; DAKO); epithelial membrane antigen (prediluted; Ventana); p53 (1:2000; Ventana); CD99 (1:1000; Signet, Dedham, MA); CD45-Common Leukocyte Antigen (1:200; DAKO); CD68 (prediluted; Ventana); S-100 (prediluted; Ventana); tau (prediluted; Ventana); and Vimentin (prediluted; Ventana). Antibodies were detected using the avidin-biotin complex method, with diaminobenzidine as the chromogen with either the Ventana XT or Ventana Techmate system.
The most cellular areas with prominent perivascular pseudorosettes were identified in sections of case 1 and the recurrent tumor of case 7. Corresponding areas were removed from the paraffin blocks and processed by standard techniques.
Six of the 8 lesions diffusely infiltrated the parenchyma, where there were often gradations among paucicellular, infiltrating, and compact tumor that partially displaced surrounding tissues. One exception, case 6, had only perivascular cells. In all but the most compact areas, trapped cortical ganglion cells were prominent. These and neurofilament protein-positive axons attested to the infiltrating nature of the process.
The lesions were principally intracortical with single-cell infiltration, perivascular orientation, and frequent sub pial aggregation (Fig. 2A-C). In the last, tumor cells were often aligned perpendicular to the pial surface (Fig. 2C), but could be aligned horizontally. Perpendicular orientation created formations analogous to perivascular pseudorosettes of ependymoma or astroblastoma when the cells had broad processes and distinct cell borders (Fig. 2C). Other than the chance encounter of a neoplastic cell and a neuron, perineuronal satellitosis was absent except in the recurrent tumor of case 7, described subsequently.
All tumors were composed of monomorphous cells that were strikingly angiocentric with circumferential, radial, or longitudinal alignments to both large and small vessels (Figs. 2D-F, 3A). Even capillaries were affected. Tumor cells in cases 6 and 7 were concentrated around blood vessels with little if any intraparenchymal component (Fig. 2D), but the recurrent lesion in case 7 was diffusely infiltrative in other areas.
Circumferential or longitudinal orientation of tumor cells to vessels predominated in paucicellular areas, whereas radial alignment with perivascular pseudorosettes prevailed in more cellular regions. These rosettes contained finely fibrillar perivascular zones that closely resembled those of ependymoma, but in some areas, perivascular cells with stout processes and distinct cell borders created perivascular formations of the astroblastoma type defined by Bonnin et al (2). No true lumen-containing rosettes were present.
Seen only in case 8 were round, paranuclear structures consistent with “microlumens” filled with granular material. These were particularly apparent in perivascular tumor cells (Fig. 3B), but were visible also in independent clusters or even individual cells.
Present in 3 cases were compact masses with fascicular architecture that resembled Antoni A tissue of schwannomas (Fig. 3C, D). Smaller masses of this type appeared to evolve from perivascular fibrillar areas.
Seven tumors had uniform, strikingly monomorphous, usually slender cells (Figs. 2E, F3, 3D) that were well demonstrated in smear preparations (Fig. 3E). The long nuclei in seven cases with their speckled, crisp chromatin were remarkably similar from case to case in amount and distribution. The one exception to cytologic bipolarity was case 6 in which the tumor cells, all perivascular, were round.
The tissue removed at the second operation in case 7 contained paucicellular, angiocentric tumor identical to that in the original specimen removed 21 months earlier (Fig. 2D), but also had densely cellular tissue composed of nonpolar cells that, in some areas, surrounded neurons (Fig. 3F). Even retrospectively, there was nothing to distinguish the initial lesion from the other tumors in the series in terms of mitotic activity (none identified) or MIB-1 index (low) or degree of cytologic atypia (identical to other cases) to predict this evolution. The recurrent tumor had more mitoses and a higher MIB-1 index than the initial lesion, as discussed subsequently. Neither necrosis nor vascular proliferation was present.
Absent from all tumors were calcifications, Rosenthal fibers, eosinophilic granular bodies, vascular proliferation, vascular hyalinization, necrosis, and components of classic dysembryoplastic neuroepithelial tumor (intracortical nodules, “floating neurons,” oligodendrocyte-like cells, and diffuse mucinous background). Although there were many trapped cortical neurons, we did not identify small or large neurons that we felt were part of the neoplasm. There was a region in the recurrent lesion of case 7 in which cortical ganglion cells may have been somewhat large, but because their apical dendrites were uniformly directed to the cortical surface, we did not feel that they were dysplastic or neoplastic. Microcysts were not present, but case 8 had scattered cysts lined by low columnar epithelium. Reticulin was confined to vessels.
Mitoses were absent (in 50 high-power fields) in all but 3 cases. One mitosis was found in 50 fields in cases 1 and 2. Eleven were counted in 50 high-power fields in the most cellular area of the recurrent tumor in case 7. None were identified in the initial specimen.
Tumor cells were variably reactive for GFAP. Infiltrating cells were negative or only marginally positive, with the gliotic background parenchyma more darkly stained than the tumor. In contrast, perivascular tumor cells, or those in areas of high cellularity, were usually immunoreactive, at least to some degree (Fig. 4A). EMA staining was present focally on the surface of individual (isolated) cells in two cases, but was more intense, albeit still focal, along the surface of perivascular tumor cells, either where they contacted each other or interfaced with brain parenchyma (Fig. 4B). Minute EMA immunoreactive intracytoplasmic “dots” were present in 4 cases (Fig. 4B, C) and were especially prominent in case 8 in which they were larger and sometimes elliptical (Fig. 4C). They corresponded to structures seen in the hematoxylin & eosin-stained sections illustrated in Figure 3B.
The lesions were negative for synaptophysin, NeuN, chromogranin, S100 protein, and p53. Included neurons did not exhibit abnormal surface or diffuse perikaryal chromogranin labeling that would suggest the presence of neoplastic elements. Neurofilament protein-positive axons were present between tumor cells in both “diffuse” and more cellular and compact regions.
Bipolar tumor cell processes were vimentin-immunoreactive. Scattered activated microglia were immunopositive for CD68 and CD45, but tumor cells were not. Although ependymomas are reported to be positive for CD99 (3), immunostaining was negative, as it has been in ependymomas in our experience using the antibody used here.
Sections stained for MIB-1 were available for 4 cases. The overall index was approximately 1% for cases 2, 4, and 7 (initial lesion) and case 8. It was 5% in case 1 and as high as 10% in the recurrent lesion in case 7. MIB-1-positive cells were present in obvious tumor areas, but were also scattered in brain parenchyma in which it was often difficult to identify tumor cells in companion hematoxylin & eosin-stained sections. These MIB-1-positive cells contributed to the overall MIB-1 indices and in some cases appeared to have a higher index than the bipolar cells themselves. The nature of these scattered cells was unclear. They were negative for CD45.
Tissue preservation in case 1 was adequate to identify cells that contained intermediate filaments and were covered with basement membrane material where they contacted vessels. There were a few intercellular microlumens filled with microvilli and surrounded by cells joined by short “zipper-like” repeats of “intermediate” junctions (Fig. 5). Cilia were not observed in these sparse luminal structures. The cells of the recurrent tumor in case 7 were filled with intermediate filaments. No cell junctions, microvilli, or cilia were noted.
These 8 seizure-producing, monomorphous tumors appear to represent a distinct clinicopathologic entity not found in the World Health Organization classification system (4). Defining features were superficial location, angiocentricity, bipolar cells in most cases, cytologic monomorphism, and ependymoma-like features in an infiltrating lesion. The infiltration, with subpial accumulation in some cases, suggests an infiltrating “diffuse” astrocytoma, the original diagnosis in several cases. Perivascular formations, immunoreactivity for EMA, and ultrastructural features in one case, on the other hand, are those of ependymal or astroblastic differentiation, broadly defined. These terms were incorporated also into the descriptive diagnoses by some of the original pathologists. The observed microlumens constitute especially compelling immunohistochemical and ultrastructural evidence of ependymal differentiation (5).
Unlike diffuse astrocytomas, the cells were, overall, extremely monomorphous and strikingly bipolar so that the lesions would be difficult to place into the grade II “diffuse” astrocytoma category, even if there were no associated ependymoma-like features. In contrast to those of infiltrating astrocytomas, nuclei were generally elongated and the cells were in striking perivascular formations foreign to infiltrating astrocytomas. Although the cells were generally “piloid,” the degree of infiltration, especially the formation of subpial aggregates, is alien to pilocytic astrocytoma. In further distinction, the tumors were radiologically neither well-defined nor contrast-enhancing. Pilocytic astrocytomas are typically both. The lesions' cytologic features and orientation to vessels somewhat resemble features of “pilomyxoid astrocytoma” (PMA) (6), but lack the mucinous background and generally compact noninfiltrating nature of the latter. PMAs are also distinguishable radiologically because of their contrast enhancement (generally absent in our cases), by location (usually hypothalamus rather than cerebral hemispheres), and age of onset (usually in the very young) and histologic profile (piloid and mucin-rich). We note that suprasellar tumors have been reported by one group to exhibit ultrastructural characteristic, including some ependymal qualities, of differentiation along tanycytic lines (7, 8). The relationship of the present lesions far removed from the ventricular system to tanycytes is unclear, and no intracellular lumens have been demonstrated in PMAs.
Although there are occasional ependymomas with a zone of infiltration, the present lesions were first and foremost “diffuse” neoplasms, not ependymomas with focal tissue infiltration. In addition, although there were radially structured perivascular pseudorosettes characteristic of ependymoma, there were also extensive regions where perivascular tumor cells were aligned longitudinally or circumferentially in formations not typical of ependymoma. Furthermore, even in compact areas with perivascular formations, the present lesions, as a group, did not bear a close resemblance to classic cellular ependymoma. The cells' bipolarity is also at odds with cellular ependymoma. These solid regions resembled tanycytic ependymoma to a degree but were more disorganized and somewhat schwannoma-like in appearance. Radiologically, the lesions deviated distinctly from ependymoma because they were more consistent with an infiltrating neoplasm, that is, nonenhancing and somewhat ill-defined.
Could the lesions be variants of astroblastoma? Since the original description in 1930 (9), this term has been applied variably to infiltrating astrocytomas with a tendency to perivascular orientation (“astroblast formation”) and to well-circumscribed, often superficial masses with ependymoma-like features. The present tumor could perhaps be considered astroblastoma in the former sense, when radial perivascular rosettes were present, but less clearly so in infiltrating areas in which perivascular cells were not radially oriented.
By a second definition, astroblastoma is an ependymoma cousin with ultrastructural features such as microvilli and “zipper-like” runs of intermediate junctions (2, 10). These features suggested to Bonnin et al an origin from the tanycyte, a bipolar relative of the ependymal cell that in one form has a long process extending into the subependymal zone. Astroblastomas of this ilk are compact, roughly spherical, largely noninfiltrative, contrast-enhancing masses (11), that is, not tumors with the radiologic constellation of features in our cases. Vascular hyalinization, a frequent histologic feature of this form of astroblastoma, was also absent.
Although the present tumors technically fulfill at least some of its criteria, the lesion should not be subsumed under the DNT rubric in its ill-defined “non-specific form” (12). One case ultimately proved lethal.
Whatever its cytogenesis, this distinctive neoplasm appears to behave overall as a low-grade lesion. The 9-, 18-, and 25-year periods of preoperative seizures in cases 5, 8, and 6, respectively, and the 4-year period of radiologic stability of case 6, are certainly those of a biologically indolent lesion. The tumor in case 7, however, recurred as a more mitotically active lesion that ultimately killed the patient. Although it could be argued that this tumor, in an adult, was “diffuse” anaplastic astrocytoma, the histopathology of the initial specimen and areas of the recurrent lesion were identical to other cases in all aspects. The monomorphous angiocentric lesion is therefore not totally benign.