Immunophenotypic Characterization of Germ Cell Tumor–Derived Primitive Neuroectodermal Tumors: Evidence for Frequent Neuronal and/or Glial Differentiation

This study was approved by the Institutional Review Board of Indiana University–Purdue University at Indianapolis. A computerized text search of surgical pathology reports at the Indiana University Pathology Laboratory was performed for the phrase “primitive neuroectodermal tumor.” The reports were reviewed to determine if the diagnosis represented a testis GCT-derived PNET. According to current recommendations, the PNET was required to occupy at least a 4× microscopic field to distinguish it from a teratoma.¹⁶  Hematoxylin-eosin slides from all available cases with a component of a testis GCT-derived PNET were reviewed by 2 study authors to determine if any neuronal or glial differentiation was present in association with the undifferentiated PNET. Glial differentiation was recognized by cells resembling glia of the CNS with abundant eosinophilic cytoplasm and eccentric, euchromatic nuclei. Neuronal differentiation, in contrast to the primitive neural cells of undifferentiated PNET, showed larger cells with a moderate amount of eosinophilic cytoplasm and round to ovoid nuclei with relatively smooth nuclear contours. If a tumor morphologically demonstrated neuronal or glial differentiation in association with a testis GCT-derived PNET, IHC stains for glial fibrillary acidic protein (GFAP), S100 protein, synaptophysin, chromogranin A (CGA), and SOX11 were performed on any available material (ie, unstained slides or formalin-fixed, paraffin-embedded tissue block), if not already performed for the original diagnosis.

The IHC staining for GFAP, S100, CGA, synaptophysin, and SOX11 was performed using a polymer-based method (Envision FLEX, Dako, Carpenteria, California) and diaminobenzidine as the chromogen on a Dako automated immunostaining instrument (Table 1). All the authors analyzed all slides. The IHC stains were scored based on the quantity and intensity of the positive reaction. For quantity, 0 = negative staining; 1 = ≤10%; 2 = 11% to 50%; and 3 = >50%. For intensity, 0 = negative staining; 1 = weak; 2 = moderate; and 3 = strong. Any amount or intensity of staining was considered positive.

A total of 54 testis GCT-derived PNETs with available hematoxylin-eosin slides were initially identified by the computerized archival search. Upon review of the hematoxylin-eosin slides, 13 of the 54 tumors (24%) had neuronal and/or glial differentiation (designated as a differentiated component) in association with the PNET. A total of 10 of these 13 cases had sufficient remaining available material to complete the entire panel of GFAP, S100, synaptophysin, CGA, and SOX11. The remaining 3 cases either had some of the IHC stains previously performed and available for review or sufficient available material to perform a partial IHC panel.

The clinical findings are summarized in Table 2. The 13 patients ranged from 16 to 55 years of age (mean, 35 years) at the time of surgical resection of the specimen containing the testis GCT-derived PNET. The PNET was located in the testis in 6 cases, the retroperitoneum or pelvis in 5, the mediastinum in 1, and both the testis and retroperitoneum in 1. Most of the men (n = 9; 69%) had a nonseminomatous mixed GCT of the testis, whereas 1 had a pure seminoma, and 1 had only embryonal rhabdomyosarcoma. The embryonal rhabdomyosarcoma was presumably another SM. In 2 men, the primary testis tumor pathology was unknown. The pathologic stage of the testis tumor was pT1 in 4 men, pT2 in 6, and unknown in 2.

All 13 cases demonstrated a component of glial or neuronal differentiation in association with the undifferentiated PNET. The undifferentiated PNET components were typically composed of relatively small, atypical cells with scant cytoplasm, pleomorphic, hyperchromatic nuclei, and showed frequent mitotic figures and apoptotic bodies (Figure 1, A). These cells were present in diffuse sheets and formed occasional rosettes, pseudorosettes, and tubules (Figure 1, B). The areas of glial differentiation showed cells with more abundant, eosinophilic, and often fibrillar cytoplasm and typically less nuclear hyperchromasia (Figure 1, C). Differentiated neuronal components were composed of large cells with a moderate amount of cytoplasm (ie, less cytoplasm than the differentiated glial cells but more than the undifferentiated PNET cells; Figure 1, D). The undifferentiated and differentiated components were often closely admixed, and some of the differentiated components clearly demonstrated an admixture of both differentiated neuronal and glial cells (Figure 1, E). Two cases (2 of 13; 15%) showed high-grade cytologic atypia in the differentiated glial component, mitotic figures, and microvascular proliferation, and these areas occupied greater than a 4× microscopic field. Based on these findings, these cases were also considered to harbor a SM (ie, glioblastomas; Figure 1, F).

The IHC results are summarized in Table 3. SOX11 was the most sensitive IHC marker, with positive staining, typically strong and diffuse, in the undifferentiated PNET component of all tumors (n = 10; 100%; Figure 2, A) and was only rarely positive in the differentiated components (1 of 10; 10%), with the single positive case showing only focal and weak expression in differentiated glia (Figure 2, B). Synaptophysin was slightly less sensitive in the undifferentiated PNET components (12 of 13; 92%; Figure 2, C) compared with SOX11 and was frequently positive in the differentiated components (5 of 13; 38%; Figure 2, D). The heterogeneous staining of synaptophysin in the differentiated components is attributable to the variation in neuronal versus glial differentiation. Chromogranin A was less sensitive for highlighting the undifferentiated PNET component than either SOX11 or synaptophysin (5 of 11; 45%; Figure 2, E). Chromogranin A was also usually negative in the differentiated components (2 of 11; 18%). Glial fibrillary acidic protein (3 of 12; 25%) and S100 (2 of 12; 17%) were typically negative in the undifferentiated PNET components and were usually positive in the differentiated PNET components (GFAP: 10 of 12, 83%; S100: 10 of 13; 77%; Figure 2, F), a finding attributable to glial differentiation in the latter.

Testis GCTs and their SMs present a challenging group of malignancies with a wide spectrum of morphologic findings. One well-known type of SM is PNET, which was relatively recently described in 1983, soon after it was reported in the ovary.^11,17,18  The tumor is usually derived from teratoma; however, occasionally a pure PNET without any other GCT component can occur.¹²  It has been speculated that these tumors represent overgrowth of PNET with complete replacement of other GCT components.

Michael et al¹⁰  reported the first large study and presented a comprehensive account of their morphologic and clinicopathologic findings.¹⁰  To date, only a few large series have been published in the English-language literature,^6,10,19  and the factors that contribute to their formation and knowledge of their spectrum of differentiation are not established. What is clear is that testis-derived PNETs virtually always resemble CNS-type PNETs rather than peripheral-type PNETs. As such, they lack chromosome 22 abnormalities and are morphologically characterized by scant cytoplasm, hyperchromasia, and frequent mitotic figures and apoptotic bodies while being arranged in primitive tubules, rosettes, and pseudorosettes.^4,8–12  Additionally, the 2 senior authors of this study have reviewed cases that appeared to show early neuronal and glial differentiation admixed with undifferentiated PNET, although there is scant discussion of this phenomenon in the literature. This anecdotal observation was the basis for the present study. Additionally, Matoso et al¹⁵  recently demonstrated the potential of testicular GCTs to rarely show neuroglial differentiation; 3 of their 13 patients had PNET arising in a teratoma in the orchiectomy specimen. These 3 patients had metastases with neuroglial differentiation in retroperitoneal lymph node dissection specimens that lacked any undifferentiated PNET.¹⁴  Subsequently, Murati Amador and Matoso¹⁴  documented a patient who had concurrent PNET and neuroglial differentiation in an orchiectomy specimen. To our knowledge, this was the first documented case of PNET with neuroglial differentiation in the English-language literature. In the present study, we found that 13 of 54 cases of PNET (24%) showed light microscopic evidence of neuronal and/or glial differentiation. Thus, approximately one-quarter of all cases of testis GCT-derived PNET show at least focal neuronal and/or glial features.

Regarding IHC, we found that SOX11 is a sensitive marker for testis GCT-derived PNET because all undifferentiated PNETs were positive. SOX11 has been shown to be expressed in CNS embryonal tumors (eg, medulloblastoma).^13,20  Because PNETs arising in association with GCTs are also CNS-type, it intrigued us to test this marker in testicular PNETs.^20,21  Synaptophysin was only slightly less sensitive, with 92% of cases being positive. Interestingly, SOX11 expression was typically lost as the neoplastic cells differentiated toward glial or more mature neuronal tissue, with only 1 case having focal, weak expression of SOX11 in the differentiating component. At the same time, as SOX11 expression was diminished in the differentiated tissue components, GFAP and S100 expression increased. In summary, SOX11 is a sensitive IHC marker for the undifferentiated components of PNET, but it is lost as differentiation occurs, whereas GFAP and S100 expression is gained.

Considering both the clinical setting and the histomorphologic features of testis GCT-derived PNET with neuronal or glial differentiation, the differential diagnosis is limited. Nevertheless, if scant tissue (eg, core needle biopsy of a retroperitoneal mass) or no clinical history is available, the diagnosis may not be obvious. The differential diagnosis may include other tumors with PNET-type (so-called small round blue cell) morphology and pure undifferentiated PNET if the differentiated components are inconspicuous or scant. Ewing sarcoma is typically composed of small cells with scant clear cytoplasm arranged in irregular lobules by areas of fibrosis, and one of the senior authors has reviewed a case that was determined to be a true Ewing sarcoma arising in the testis, molecularly confirmed by a positive EWSRI break-apart probe. Thus, molecular studies for EWSR1 gene and i(12)p abnormalities are helpful in separating Ewing sarcoma from testis GCT-derived PNET.¹²  Embryonal rhabdomyosarcoma may occur either as an SM or as a primary paratesticular malignancy, and it may resemble PNET.^4,8,9  Morphologic features, including tubular structures, rosettes, and absence of rhabdomyoblasts, in combination with IHC (eg, lack of expression of muscle markers), support PNET. Nephroblastoma may be the most problematic differential diagnostic consideration based purely on morphologic features, but an IHC panel of WT1, PAX-8, and SOX11 is helpful in separating these entities, because WT1 and PAX-8 are expressed in most nephroblastomas but not in PNET,^22,23  and although we anticipate that SOX11 would be negative, we are not aware it has been studied in nephroblastoma. An IHC panel of SOX11, GFAP, and S100 may be useful in highlighting the differentiated neuronal or glial components of tumors in which these are not clearly present.

It is unclear what, if any, clinical significance the presence of neuronal and glial differentiation within PNET confers. Metastatic PNET has a poor prognosis, and patients may relapse despite chemotherapy.²⁴  If total surgical resection is accomplished, or if the PNET is confined to the testis, the prognosis is more favorable. Clinical outcome data were not included in our study or in the case report by Murati Amador and Matoso¹⁴  of neuroglial differentiation in a testis PNET. Matoso et al¹⁵  did include clinical outcome data in their study of testicular GCTs with neuroglial differentiation, but that study did not include any cases of PNET with differentiated components, and clinical follow-up was only available in 6 patients. Nevertheless, they did document that 2 patients died of disease, 1 with a tumor consistent with a WHO grade IV gliosarcoma and the other with a neoplasm resembling developing CNS without significant cytologic atypia. Two of our cases had areas qualifying for glioblastomas, but follow-up was not available. Given the inherent poor prognosis of metastatic PNET, it is possible that neuronal or glial differentiation does not influence the prognosis, but further study is necessary.

In conclusion, we identified the presence of glial and neuronal differentiation in approximately one-quarter of testis GCT-derived PNET, and SOX11, S100, and GFAP may be useful IHC markers to help identify these different components. Further clinical studies are required to determine the possible prognostic significance.