Atypical teratoid/rhabdoid tumors are malignant embryonal tumors characterized by the presence of rhabdoid cells, genetic alterations affecting the SMARCB1 gene (hSNF5/INI1), and a poor prognosis. Whether INI1 plays a role in the pathogenesis of other central nervous system tumors is uncertain. We report on cases of 2 young children with unusual intracranial nonrhabdoid neuroectodermal tumors within and around the third or fourth ventricle that are characterized by cribriform strands and trabeculae and well-defined epithelial membrane antigen-immunopositive surfaces and show INI1 protein loss. Histological and immunohistochemical features did not correspond to established tumor types, including atypical teratoid/rhabdoid tumors, medulloepithelioma, choroid plexus carcinoma, and ependymoma. Fluorescence in situ hybridization analyses failed to identify chromosomal alterations affecting the SMARCB1 locus, but sequencing revealed a homozygous 4-bp duplication in exon 4 (492duplCCTT) in one of the tumors. Both children responded well to conventional adjuvant therapy protocols and are alive and in complete remission longer than 5 years postoperatively. We suggest that cribriform neuroepithelial tumor (CRINET) is a nonrhabdoid ventricular tumor that shows loss of tumoral INI1 protein and has a relatively favorable prognosis.
Genetic alterations play an increasing role in the diagnosis of malignant central nervous system (CNS) tumors and may define prognostic relevant subsets of patients (1). Atypical teratoid/rhabdoid tumors (AT/RTs) are characterized by the presence of rhabdoid tumor cells and loss of INI1 protein expression caused by deletions of chromosome 22q and/or inactivating mutations of the SMARCB1 gene (hSNF5/INI1) located on chromosome 22q11.2 (2-4). Atypical teratoid/rhabdoid tumors respond poorly to chemotherapy and carry a dismal prognosis; more aggressive multimodality treatment schedules may be beneficial (5).
The immunohistochemical demonstration of INI1 protein loss is generally thought to be characteristic of AT/RT (6). The demonstration of this feature has resulted in the reclassification of many pediatric brain tumors that initially had been categorized as supratentorial primitive neuroectodermal tumor, medulloblastoma, or choroid plexus carcinoma as AT/RT (7, 8). Lack of INI1 protein expression has also been reported in a fraction of primitive neuroectodermal tumors without rhabdoid phenotype, but it has been suggested that these biologically aggressive tumors most likely represented AT/RT in which rhabdoid features had been missed because of sampling bias (7). It has become evident that some nonrhabdoid extracranial tumors such as epithelioid sarcoma (9) and schwannoma (10) may carry genetic alterations of SMARCB1, resulting in INI1 protein loss. Such alterations have not previously been described in nonrhabdoid CNS tumors, however. Here, we report on cases of 2 children with unusual neuroepithelial tumors that show distinct and nonrhabdoid histological features, loss of INI1 protein expression, and have had a relatively favorable prognosis.
A 26-month-old girl presented with a large contrast-enhancing tumor in the vicinity of the third ventricle (maximal diameter, 25 mm). Upon partial resection, the tumor was initially diagnosed as a choroid plexus carcinoma and treated according to CPT-SIOP-2000 (etoposide, cyclophosphamide, and vincristine). Starting at 40 months of age, radiotherapy was administered (total dose, 54 Gy). On repeated lumbar puncture and magnetic resonance imaging, there has been no evidence of spinal dissemination of the tumor. At the time of writing (72 months after the surgery), the patient is alive and in complete remission.
A 10-month-old boy presented with a large cerebellar tumor in the vicinity of the fourth ventricle with irregular contrast-enhancement and a maximal diameter of 56 mm that completely obstructed the fourth ventricle (Fig. 1). After gross total resection, the patient received chemotherapy according to HIT-SKK-2000 (including methotrexate, cyclophosphamide, and vincristine). He is alive and in complete remission 85 months postoperation.
Formalin-fixed paraffin-embedded specimens from both tumors that had been sent for consultation to the Institute of Neuropathology, Münster, were available for examination. In addition to hematoxylin and eosin and periodic acid-Schiff staining, immunohistochemistry for epithelial membrane antigen (EMA), vimentin, synaptophysin, microtubule-associated protein 2 (MAP2C), cytokeratins (KL1 and MNF116), glial fibrillary acidic protein (GFAP), choroid plexus marker Kir7.1 (11) (kindly provided by Dr. Hirose, Department of Biological Sciences, Tokyo Institute of Technology, Tokyo, Japan), and other antibodies was performed as previously described (11). Expression status of INI1 protein was determined as described (6, 11). Briefly, after deparaffinization and boiling for antigen retrieval (citrate buffer, pH 9), 2-µm sections were stained with a monoclonal antibody against INI1 (BAF47; 1:200, BD Biosciences, San Jose, CA) and the avidin-biotin complex method on an automated staining system (TechMate, Dako, Glostrup, Denmark). Fluorescence in situ hybridization was performed on interphase cells from tissue sections using 4 probes for the SMARCB1 locus in 22q11.2 (RP11-1112A23, RP11-71G19, RP11-911F12, and RP11-76E8) and a chromosome 11 centromere probe as a control (Abbott Laboratories, Abbott Park, IL). Probe preparation, sample pretreatment, hybridization, and evaluation were performed according to previously described protocols (12). Mutational analysis of SMARCB1 was done on genomic DNA-derived from paraffin-embedded tissues. All 9 coding exons and flanking intronic sequences of SMARCB1 were amplified by polymerase chain reaction using primers chosen from published sequences (13). All polymerase chain reaction products were directly sequenced using an ABI 3100 automatic sequencer (Applied Biosystems, Foster City, CA).
The histopathologic features of both tumors were remarkably similar (Fig. 2). The highly cellular tumors were composed of relatively small undifferentiated cells arranged in cribriform strands and trabeculae of varying thickness forming well-defined surfaces (Figs. 2C-F). The tumor cells nuclei had dense chromatin and lacked prominent nucleoli; the cytoplasm was slightly eosinophilic and ill defined. Cells gathered in strands often showed elongated nuclei, but there was no stratification. There were up to 2 mitoses per high-power field, and tumor necrosis was seen. Areas with more compact growth in which single-layered tumor cells were arranged around lumina formed true rosettes were also present (Figs. 2G, H). Importantly, rhabdoid tumor cells with prominent nucleoli, larger eosinophilic cytoplasm, and/or inclusions were absent. No external limiting membrane could be demonstrated on periodic acid-Schiff staining.
Immunohistochemistry and Molecular Genetics
Both tumors showed distinct expression of EMA highlighting surfaces (Fig. 3A); they also expressed vimentin, MAP2C, and synaptophysin (Figs. 3B-D); there was only focal cytokeratin staining (Fig. 3E). Focal expression of S100 was also seen in Case 1. Staining for GFAP and potassium channel Kir7.1 (a highly and specifically expressed marker of choroid plexus tumors 11) was absent. No tumoral expression of neurofilament, NeuN, neuron-specific enolase, chromogranin A, actin, desmin, p53, placental alkaline phosphatase, or β-human chorionic gonadotropin was observed (data not shown). The Ki67/MIB1 proliferation index accounted for 30% and 28% in Cases 1 and 2, respectively. Loss of nuclear INI1 protein expression of tumor cells with retained immunoreactivity in vascular cells and lymphocytic infiltrates was observed in both cases (Fig. 3F).
Interphase fluorescence in situ hybridization analyses using probes spanning and flanking the SMARCB1 locus in 22q11.2 failed to identify imbalances or translocation breakpoints in either case. Sequencing of the SMARCB1 gene revealed a homozygous duplication within exon 4 (492duplCCTT), resulting in a premature stop codon in Case 2; no mutation of SMARCB1 was detected in the other. These data are summarized in the Table.
These 2 patients had unusual neuroectodermal tumors with distinct cribriform nonrhabdoid histological features and loss of INI1 protein expression, and they have had a relatively favorable prognosis. Although rhabdoid differentiation of tumor cells was not encountered, lack of tumoral INI1 protein expression in both tumors and associated with a homozygous mutation within exon 4 of SMARCB1 (492duplCCTT) in 1 tumor initially raised the suspicion that these tumors might represent AT/RT. Rhabdoid cells are classically described in AT/RT, but these tumors can show a variety of histological features and not all cases contain identifiable rhabdoid cells. Cases of AT/RT with unusual histological features such as ependymoblastic rosettes and focal yolk sac tumor formation are on record (14), but the predominance of surface-forming tumor strands as encountered in the present cases has heretofore not been described in AT/RT.
Embryonal tumors without rhabdoid features lacking INI1 protein expression have been reported, and it has been suggested that they most likely represented AT/RT lacking rhabdoid features because of sampling bias (7). In contrast to our cases, follow-up of those cases confirmed an aggressive biologic behavior typically associated with AT/RT. Unlike most patients with AT/RT (15), our patients have responded remarkably well to conventional therapy regimens; this further argues against the likelihood that these are cases of AT/RT.
Routine use of INI1 staining in embryonal brain tumors has been advocated (6, 7), but only few centers have implemented such a policy during a longer period. It is therefore reasonable to suggest that INI1 loss may also be identified in other nonrhabdoid tumor entities.
The presence of surface-forming tumor strands raises the possibility that these cases might represent medulloepithelioma. Since the initial description by Bailey and Cushing in 1926 (16), the definition of CNS medulloepithelioma has been a matter of debate, and this rare pediatric brain tumor remains an ill-defined entity (17). Even if one argues that the surfaces and strands encountered in our cases resemble “neural tube-like” strands, both tumors neither show the elongated surfaces and pseudostratification classically seen in medulloepithelioma nor tumor cells resting on a periodic acid-Schiff-positive external limiting membrane (17). Furthermore, we did not observe INI1 protein loss in any of 12 medulloepitheliomas of whom 9 had been previously published (18-20) (data not shown).
One of the cases had been sent for consultation because a diagnosis of “choroid plexus carcinoma” had initially been considered. The tumor lacking well-defined papillary structures and cytological features of relatively small tumor cells, in our opinion, makes this possibility unlikely. Furthermore, in both cases, the expression of cytokeratins was only observed focally, and membranous expression of potassium channel Kir7.1 was lacking. Distinct EMA staining of surfaces is characteristic of ependymomas (21) but is usually not encountered in choroid plexus tumors. Although loss of INI1 protein expression has been reported in some cases diagnosed as choroid plexus carcinoma (6, 22), on re-evaluation, most of these historical cases are considered to be AT/RTs (8). We also have found that INI1 protein expression is generally preserved in choroid plexus tumors (11).
In one of our cases, the diagnosis of ependymoma had been suggested by the referring pathologist. Indeed, the diagnosis of anaplastic ependymoma could be considered, especially because the surfaces and rosettes of both tumors showed distinct EMA immunoreactivity as encountered in a fraction of ependymomas (21). The complete lack of GFAP immunoreactivity and the presence of distinct synaptophysin and MAP2C staining, which are usually absent in ependymal tumors (23), argue against this possibility, however. Moreover, genetic alterations affecting INI1 protein expression have not been observed in large series of ependymomas (24-26).
Because of their distinct histological, immunohistochemical, and clinical features, the 2 tumors do not easily fit into any category of the current World Health Organization classification of CNS tumors (27). We propose the designation cribriform neuroepithelial tumor (CRINET) for this type of tumor. Because they represent only a small fraction of pediatric referral cases at our institution, CRINET is certainly very rare. In view of the similarity with more frequent pediatric brain tumor entities, such as ependymomas and choroid plexus tumors, however, it seems likely that other cases may have been overlooked. Furthermore, recognition and better characterization of CRINET could be of importance because our limited experience suggests that despite INI1 protein loss, these tumors might respond well to conventional chemotherapy regimens. If our patients' tumors had been classified as AT/RT solely based on INI1 protein expression status, more aggressive multimodality treatment schedules would have been considered, potentially resulting in overtreatment. Interestingly, cases of intraventricular tumors with formation of ribbons and tubular structures have been previously reported. These tumors had been initially diagnosed as ependymoma (28) but were later reclassified as AT/RT mainly based on the absence of INI1 protein expression (29). They also were associated with long survival. Another striking similarity is that (as in our Case 2) genetic alterations affecting exon 4 of the SMARCB1 gene were identified (29).
To conclude, CRINET is a nonrhabdoid intraventricular and periventricular CNS tumor showing loss of tumoral INI1 protein and relatively favorable prognosis.
The authors thank Drs Anne Jouvet (INSERM 842, Université de Lyon), Maria-Beatriz S. Lopes (Department of Pathology, University of Virginia Health System), Lucy B. Rorke-Adams (Department of Pathology, Children's Hospital of Philadelphia, University of Pennsylvania School of Medicine), and Dirk Troost (Department of Neuropathology, Academic Medical Centre, University of Amsterdam) for generously sharing archival medulloepithelioma tissues for examination of INI1 protein expression status. The MAP2C immunohistochemistry was kindly performed at the Institute of Neuropathology Bonn (Drs Pietsch and Becker). Claudia Becher, Barbara Riesmeier, and Dorit Schuster provided expert technical assistance.
- fluorescent in situ hybridization
- central nervous system tumor
- choroid plexus
- fourth ventricle
- heart ventricle
- tissue membrane
- neoplasms, neuroepithelial
- neuroectodermal tumors
- rhabdoid tumor
- neoplasms, embryonal
- complete remission