Abstract
Context.—Immunohistochemistry (IHC) has become an important tool in the diagnosis of brain tumors.
Objective.—To review the latest advances in IHC in the diagnostic neuro-oncologic pathology.
Data Sources.—Original research and review articles and the authors' personal experiences.
Data Synthesis.—We review the features of new, useful or potentially applicable marker antibodies as well as the new uses of already established antibodies in the area of diagnostic neuro-oncologic pathology, focusing on the use of IHC for differential diagnosis and prognosis. We discuss (1) placental alkaline phosphatase, c-Kit, and OCT4 for germinoma, (2) α-inhibin and D2-40 for capillary hemangioblastoma, (3) phosphohistone-H3 (PHH3), MIB-1/Ki-67, and claudin-1 for meningioma, (4) PHH3, MIB-1/Ki-67, and p53 for astrocytoma, (5) synaptophysin, microtubule-associated protein 2, neurofilament protein, and neuronal nuclei for medulloblastoma, (6) INI1 for atypical teratoid/ rhabdoid tumor, and (7) epithelial membrane antigen for ependymoma. All the markers presented here are used mainly for supporting or confirming the diagnosis, with the exception of the proliferation markers (MIB-1/Ki-67 and PHH3), which are primarily used to support grading and are reportedly associated with prognosis in certain categories of brain tumors.
Conclusions.—Although conventional hematoxylin-eosin staining is the mainstay for pathologic diagnosis, IHC has played a major role in differential diagnosis and in improving diagnostic accuracy not only in general surgical pathology but also in neuro-oncologic pathology. The judicious use of a panel of selected immunostains is unquestionably helpful in diagnostically challenging cases. In addition, IHC is also of great help in predicting the prognosis for certain brain tumors.
Although conventional hematoxylin-eosin (H&E) staining is crucial for diagnosis, diagnostic neuropathology has benefited in the last 2 decades from the incorporation of, and recent advances in, immunohistochemistry (IHC). A number of markers for IHC have been developed in the area of diagnostic neuro-oncology, since glial fibrillary acidic protein (GFAP), the antibody against which is currently most commonly used in practice, was found by Eng et al1 in 1971 and was later reported as a useful marker antigen for astroglial cells by Kleihues et al2 in 1987.
In general, brain tumors are classified into 2 major groups, primary and metastatic, and the primary brain tumors can be classified further into 3 groups: neuroepithelial (eg, astrocytic, oligodendroglial, ependymal, choroid plexus, neuronal, and pineal parenchymal tumors), nonneuroepithelial (eg, meningioma, nerve sheath tumors, malignant lymphoma, pituitary adenoma, and germ cell tumors), and others (ie, tumors of unknown origin, eg, capillary hemangioblastoma). In neuropathology practice, we routinely use several useful IHC markers that are relatively sensitive and specific for some of these tumors (eg, GFAP for astrocytomas, synaptophysin for neuronal tumors); however, none of these are diagnostic (ie, no absolute sensitivity and specificity).
There have been many recent publications in the area of IHC in brain tumor pathology, with several articles relating to new specific antibodies.3–10 This review will discuss the features of new, reportedly sensitive and specific marker antibodies as well as new uses of already established antibodies in the areas of adult and pediatric, diagnostic neuro-oncology practice, based on recently published reports and our own experience.
GERMINOMA: PLACENTAL ALKALINE PHOSPHATASE, C-KIT (CD117), AND OCT4
Germinoma occurs predominantly in the pineal and suprasellar regions and is composed of lobules or sheets of uniform cells with large vesicular nuclei, prominent nucleoli, well-defined cell boundaries, and abundant clear cytoplasm, admixed with lymphoplasmacytic infiltrates. Given that, characteristically, intracranial germinomas are highly radiosensitive and chemosensitive, allowing for a high cure rate with radiation alone or cisplatin-based chemotherapy followed by low-dose radiotherapy, an accurate diagnosis is critical for patient management. The histologic features are virtually diagnostic when the specimens are sufficient for evaluation and are well preserved without artifact. Immunohistochemistry is of particular use in cases when either the specimen is very small or the lymphocytic infiltrate is predominant.11 As with ovarian dysgerminomas and testicular seminomas, intracranial germinomas are known to show immunoreactivity for placental alkaline phosphatase (PLAP) in a surface membrane or, somewhat less commonly, diffuse cytoplasmic distribution.12 This antigen is a cell surface glycoprotein and is normally expressed in syncytiotrophoblasts and primordial germ cells.13 Although this marker is the mainstay in current neuropathology practice, it has its shortcoming in that PLAP labeling is not a constant feature with variable sensitivity, intensity, and extent of reactivity,3,12,14 and it can sometimes be difficult to interpret, especially in the cases with heavy inflammatory cell infiltrates and in specimens that were previously frozen.12
The c-kit proto-oncogene encodes a receptor tyrosine kinase that is required in normal spermatogenesis.15 Expression of c-Kit (CD117) has been reported on the cell surface in almost all gonadal seminomas/dysgerminomas (Figure 1, A) but very rarely in nonseminomatous germ cell tumors.15,16 Takeshima et al17 and Sakuma et al18 reported that they studied 16 cases of intracranial germinomas, respectively, and c-Kit was diffusely expressed on the surface of germinoma cells in all cases examined. In addition, Takeshima et al17 reported that stem cell factor (SCF), a specific ligand of c-Kit, was also expressed on the cell surface, the staining pattern of which was identical to that of c-Kit. CD30 and c-Kit (CD117) used in combination are known to be useful to distinguish between embryonal carcinoma and seminoma in the gonads.15 However, to our knowledge, the expression of these markers has not been studied in combination in their intracranial counterparts.
Germinoma. A, c-Kit (CD117) immunostaining. Note membranous positivity. B, OCT4 immunostaining. Note nuclear positivity (original magnification ×400). Figure 2. Capillary hemangioblastoma. Inhibin A immunostaining. Stromal cells show cytoplasmic immunoreactivity (original magnification ×400). Figure 3. Meningioma. A, Phosphohistone-H3 immunostaining. Mitotic cells are immunoreactive with characteristic finger-like projections. (Courtesy of Gregory N. Fuller, MD, PhD, Houston, Tex.) B, Claudin-1 immunostaining. Note crisp, punctate/granular immunoreactivity (original magnification ×400). Figure 4. Atypical teratoid rhabdoid tumor. INI1 immunostaining. Note negative staining in tumor cells. Immunoreactivity is seen in the intratumor vascular endothelial cells, which serve as an internal control (original magnification ×200). (Reprinted with permission from Am J Surg Pathol10 [Figure 1A]. Copyright 2004, Lippincott Williams & Wilkins.)
Germinoma. A, c-Kit (CD117) immunostaining. Note membranous positivity. B, OCT4 immunostaining. Note nuclear positivity (original magnification ×400). Figure 2. Capillary hemangioblastoma. Inhibin A immunostaining. Stromal cells show cytoplasmic immunoreactivity (original magnification ×400). Figure 3. Meningioma. A, Phosphohistone-H3 immunostaining. Mitotic cells are immunoreactive with characteristic finger-like projections. (Courtesy of Gregory N. Fuller, MD, PhD, Houston, Tex.) B, Claudin-1 immunostaining. Note crisp, punctate/granular immunoreactivity (original magnification ×400). Figure 4. Atypical teratoid rhabdoid tumor. INI1 immunostaining. Note negative staining in tumor cells. Immunoreactivity is seen in the intratumor vascular endothelial cells, which serve as an internal control (original magnification ×200). (Reprinted with permission from Am J Surg Pathol10 [Figure 1A]. Copyright 2004, Lippincott Williams & Wilkins.)
OCT4, also known as POU5F1, OCT3, or OTF3, is a nuclear transcription factor expressed in early embryonic cells and germ cells.19,20 This factor is involved in the regulation and maintenance of pluripotency of these cells19–22 and has been shown to be essential for embryonic stem cell formation and self-renewal.23,24 Cheng et al25 reported that OCT4 was expressed in all 33 cases of ovarian dysgerminomas examined, including metastases, while no immunoreactivity was noted in all 111 cases of ovarian nondysgerminomatous tumors with the exception of 4 of 14 clear cell carcinomas of ovary that showed focal (<10%) positivity. On the other hand, Jones et al26 examined 91 cases of primary testicular neoplasms and reported that there was near 100% staining of the seminoma and embryonal carcinoma cells for OCT4 in all 64 cases of adult mixed germ cell tumors examined, while the other germ cell tumor components (yolk sac tumor, choriocarcinoma, and teratoma) showed no staining. In these 2 studies, the main finding is not only that OCT4 is a highly sensitive and specific marker of ovarian dysgerminoma, and testicular seminoma and embryonal carcinoma, respectively, but also that the OCT4 staining pattern was nuclear (in contrast to PLAP and c-Kit, which show characteristically cell membrane staining), with uniformly strong staining intensity and staining extent of greater than 90%. Hattab et al3 conducted a comparative immunohistochemical study of intracranial germinomas using OCT4 and PLAP with control cases, and concluded that OCT4 is a highly specific and sensitive marker for primary intracranial germinomas (100% sensitivity for OCT4 vs 92% for PLAP). As with the ovarian and testicular counterparts in the previous studies cited above, OCT4 demonstrated characteristically diffuse and strong nuclear staining in the germinoma cells (Figure 1, B),3 which is more easily interpreted than the membranous pattern seen with PLAP immunostaining, especially in very small specimens. Since no intracranial embryonal carcinomas were included in this study,3 OCT4 may not be specific for intracranial germinomas; in other words, intracranial embryonal carcinomas should be excluded with H&E-stained sections and probably with CD30 immunostaining if the tumor is OCT4 positive.
CAPILLARY HEMANGIOBLASTOMA: α-INHIBIN (INHIBIN A) AND D2-40
Capillary hemangioblastomas (CHBs) are considered by the World Health Organization (WHO) to be grade 1 tumors of uncertain histogenesis, composed of stromal cells and abundant capillaries and commonly involving the cerebellum and spinal cord.27 The histological differential diagnosis of CHB includes metastatic clear cell renal cell carcinoma (CRCC), paraganglioma, angiomatous meningioma,28 and capillary hemangioma. The histological distinction of CHB from metastatic CRCC has long been recognized as a particular difficulty because of striking morphologic similarities between them. This difficulty may be compounded in patients with von Hippel-Lindau (VHL) disease, an autosomal dominant disorder caused by germline mutations of the VHL tumor suppressor gene, in which both CHB and CRCC are among the most commonly encountered tumors. In view of the possibility of both CHB and metastatic CRCC to the central nervous system (CNS) occurring synchronously, metachronously, or both, and tumor-to-tumor metastasis (CRCC metastasizing to CHB)29–31 in patients with VHL disease, their distinction is of particular importance and cannot be overemphasized, since the prognostic and therapeutic significance is completely different. Capillary hemangioblastoma is a benign tumor and generally has a benign course following resection, whereas metastatic CRCC in the brain carries a dismal prognosis32 and may require more aggressive treatment after surgery. Given that medical history and conventional histological examination with H&E staining alone cannot reliably distinguish between these 2 entities, IHC is crucial for differential diagnosis, and there have been a number of IHC studies to address these concerns.
In general, renal cell carcinomas are immunoreactive for epithelial markers, such as epithelial membrane antigen (EMA) and low-molecular-weight cytokeratins (eg, CAM 5.2),33 whereas CHBs are negative. On the other hand, the stromal cells of CHB have been reported to show variable patterns of immunoreactivity for neuron-specific enolase (NSE), S100 protein, VHL protein, and several growth factors27,34,35; however, all of these immunohistochemical reactions are nonspecific and would not exclude other possibilities in the differential diagnosis of CHB.29 The combined use of the immunohistochemical markers mentioned above is of help for differential diagnosis and often allows for a definitive diagnosis. Since loss of immunophenotype is sometimes encountered during tumor progression, and a definite subset (10%–30%) of CRCC is negative for EMA and CAM 5.2,36,37 other more “specific” immunohistochemical markers for CHB are needed.
Inhibin, a dimeric 32-kd glycoprotein belonging to the transforming growth factor β family and composed of an α (inhibin A) and a β subunit, is produced mainly by ovarian granulosa cells and testicular Sertoli cells.38 Inhibin A is expressed in the sex cord–stromal tumors and adrenal cortical tumors.39,40 Hoang and Amirkhan4 reported in 2003 that immunoreactivity for inhibin A was demonstrated in all 25 cases of hemangioblastoma with cytoplasmic expression in the stromal cells (Figure 2), in contrast to all 19 cases of renal cell carcinoma, including both primary and metastatic, none of which were positive, and concluded that inhibin A was a helpful marker in distinguishing CHB from metastatic CRCC. In addition, this study included 11 cases of CHB from 8 patients with VHL disease, and there was no difference in the inhibin A staining pattern between the sporadic CHB and those associated with VHL disease. A recent study performed by Jung and Kuo5 demonstrated that CD10 membranous immunoreactivity was seen in all 21 cases of CRCC (5 metastatic, 16 primary) examined, whereas all 22 cases of CHB were negative. They also showed that 91% (20/22) of cases of CHB and 24% (5/21) of cases of CRCC expressed inhibin A, and concluded that, in addition to inhibin A, CD10 was a superior marker for the differential diagnosis of CHB (negative) and metastatic CRCC (positive).
D2-40, a novel monoclonal antibody that was initially raised against an oncofetal antigen M2A,41 was recently introduced to diagnostic pathology to help identify lymphatic endothelium.42,43 Apart from lymphatic endothelium, D2-40 has been reported to be immunoreactive in mesotheliomas44,45 and, in the CNS, in choroid plexus epithelium, ependymal cells, subependymal areas, and the leptomeninges.6 A recent study conducted by Roy et al6 revealed that all 23 cases of CHB examined expressed D2-40 membranous immunoreactivity in the stromal cells with strong intensity in 19 cases (83%), whereas all 28 cases of CRCC (8 metastatic, 20 primary) failed to show immunoreactivity. There were 3 cases of CHB with VHL disease included in this study, and no difference was seen in D2-40 staining of CHB in patients with or without VHL disease. Although D2-40 appears to be a very useful marker, we have not experienced constant positivity in several cases of hemangioblastoma examined in our laboratory.
In summary, based on the recent studies, inhibin A and D2-40 are sensitive and specific markers for CHB, while EMA, CAM 5.2, and CD10 mark CRCC. Combined use of at least one of these markers from each group helps to distinguish CHB from metastatic CRCC in patients with or without VHL disease.
We have recently experienced a case of angiomatous meningioma, histologically closely mimicking CHB, which is generally considered to be one of the histological differential diagnoses of CHB.34 Diffuse EMA immunoreactivity and negative inhibin A staining supported the diagnosis of a meningioma and excluded that of a CHB. Metastatic CRCC was included in the differential diagnosis based on the diffuse EMA staining; however, CD10 and CAM 5.2 were negative.
Paraganglioma may mimic CHB and is included in the differential diagnosis of CHB; however, neuroendocrine markers, such as chromogranin and synaptophysin, are usually positive in paragangliomas.
Capillary hemangiomas of the CNS are very rare and show histological features similar to those of lobular capillary hemangiomas of the skin as well as the capillary hemangiomas of infancy,46 with a fibrous pseudocapsule.47 Most examples have been documented recently and appear to arise more commonly in spinal cord.46–51 Central nervous system capillary hemangiomas may mimic CHB, especially a vascular dominant CHB, given that this subtype of CHB may feature a lobular architecture with feeding vessels and a compact growth pattern.52 In order to exclude this unusual subtype of CHB in which the stromal cell component is small to minimal, we should use IHC with markers for the stromal cells (eg, S100 protein, NSE, inhibin A, D2-40) whenever making a diagnosis of CNS capillary hemangiomas. There were several recently reported cases of CNS capillary hemangiomas with no immunohistochemical analysis with these markers.49–51
MENINGIOMA: PHOSPHOHISTONE-H3, MIB-1/KI-67, AND CLAUDIN-1
Meningiomas are among the most common primary neoplasms of the CNS, comprising between 13% and 26% of intracranial tumors.53 According to the WHO classification, most meningiomas are benign and can be graded into WHO grade 1, and certain histological subtypes are associated with greater likelihood of recurrence and/or aggressive behavior and correspond to WHO grades 2 and 3.53 Quantifying the proliferative potential is of help in predicting the biologic behavior. One of the WHO criteria other than particular histological subtypes in the assignment of grade in meningiomas is the number of mitotic figures (MFs) per 10 high-power fields (HPFs, defined as 0.16 mm2) in the areas of highest mitotic activity; that is, no less than 4 mitoses per 10 HPFs in atypical (WHO grade 2) meningiomas, and no less than 20 mitoses per 10 HPFs in anaplastic (WHO grade 3) meningiomas.53
Given that it is difficult to distinguish MFs on H&E-stained slides from similar chromatin changes occurring in apoptotic cells or secondary to crush, distortion, pyknosis, or necrosis,54 as the diagnostic criterion, MF is not as subjective as usually considered. Ribalta et al7 studied a mitosis-specific antibody against phosphorylated histone H3 (PHH3) (which is negligible during interphase but reaches a maximum during mitosis55,56) in 54 cases of meningiomas and showed a robust positive correlation between MF counts performed on traditional H&E-stained slides and those performed on anti-PHH3–immunostained slides. Labeled nuclei are characterized by multiple finger-like projections of immunoreactivity (Figure 3, A). This robust staining of mitoses with a negative background makes mitoses “stand out.” They concluded that anti-PHH3 immunostaining facilitated rapid reliable grading of meningiomas according to WHO 2000 criteria by permitting quick focus of attention on the most mitotically active tumor area(s) for quantitation and by allowing easy and objective differentiation of MFs from apoptotic and distorted/pyknotic nuclei.7 Although it is uncertain whether or not this seemingly more sensitive and specific method is more useful than the conventional method to predict the biological behavior of meningiomas and further studies are required for this issue, there is no doubt that this immunostaining allows for rapid and accurate identification of MF.
Apart from MFs, an anti–Ki-67 monoclonal antibody MIB-1, which is immunoreactive for the nuclei of cells in non-G0 phases (ie, G1, S, G2, and M phases) of the cell cycle, is commonly used as a useful ancillary study in routine surgical/neuropathological practice to assess the proliferative potential in a given neoplasm. The MIB-1 labeling index (MIB-1 LI), which is calculated as a percentage of the MIB-1–positive cells to the total number of tumor cells, in meningiomas has been reported to correlate well with histologic grade57,58 and clinical tumor recurrence.58,59 Although MIB-1 LI is not included in the WHO criteria to grade the meningiomas, it is of particular help in tumors that are histologically “on the fence” with regard to tumor grade. Amatya et al57 studied 146 cases of meningiomas immunohistochemically and reported that the mean MIB-1 LI of benign, atypical, and anaplastic meningiomas was 1.5%, 8.1%, and 19.5%, respectively. They also reported that p53 immunoreactivity (p53 labeling index) correlated with the histologic grade as well. Nakasu et al58 investigated the predictability of tumor growth, assessed radiologically by tumor doubling time, and recurrence in 139 cases of meningiomas by MIB-1 IHC, using 2 different counting methods; that is, counting in the area of the highest MIB-1 labeling (HL method) and counting in randomly selected fields (RS method). They reported that the MIB-1 LI measured by both methods showed a significant correlation with tumor grade, growth speed, and recurrence rate. Interestingly, they pointed out that focal accumulation of MIB-1–positive cells in meningiomas was not likely to correlate with their biologic aggressiveness and concluded that the RS method was a better predictor of recurrence and tumor growth in meningiomas than the HL method when counting manually. There have been several studies describing the cutoff point of MIB-1 LI for recurrence, although it varied from report to report.58–61 Nakasu et al58 suggested approximately 2% and 3% in the RS and HL methods, respectively. Matsuno et al60 studied the MIB-1 LI of recurrent (n = 54) and non-recurrent (n = 73) groups of meningiomas using the HL method, and mentioned that a MIB-1 LI of 3% was a cutoff point for recurrence, especially within the first 10-year follow-up periods, although there was a marked overlap of values between the groups. Ho et al61 studied 83 cases of meningiomas with at least a 10-year follow-up by IHC using the HL method and reported that the MIB-1 LI of 10% was a cutoff point for recurrence. Perry et al59 studied prognostic significance of variable parameters by IHC using image analysis in 425 cases of meningiomas, and reported that the MIB-1 LI of 4.2% or more was strongly associated with a decreased recurrence-free survival rate in gross, totally resected meningiomas.
Meningiomas, particularly the fibroblastic type, may be difficult to distinguish from schwannomas with routine H&E-stained sections, especially when located in the cerebellopontine angles and intradural, extramedullary regions of spinal canal. In these tumors, EMA immunoreactivity may be faint and/or focal. Winek et al62 mentioned that because of overlap in S100 protein and EMA reactivity, these markers were unreliable in differentiating meningioma from acoustic schwannoma.
Claudin-1, an integral structural protein of tight junctions, has recently been used as a marker of perineurial cells and has been reported to be often a more robust marker than EMA to distinguish soft tissue perineuriomas from its mimics.63 Bhattacharjee et al8 conducted a comparative IHC study using claudin-1 and EMA in 20 and 10 cases of meningioma and schwannoma, respectively. They reported that claudin-1 and EMA expression was observed in 85% and 100% of meningiomas, respectively, and of the 10 schwannomas, 2 cases showed focal, nonmembranous staining for EMA, while none were positive for claudin-1. They also stressed that the immunoreactive pattern of claudin-1 was unique in the crisp, punctate/ granular membranous reaction (Figure 3, B), which was visually more favorable in contrast to the faint or weak membranous pattern of expression seen in EMA. They concluded that claudin-1 was a very useful adjunct to EMA in meningiomas with equivocal morphologic features or with weak/focal EMA expression, and that the lack of claudin-1 expression by schwannomas was very useful in the context of differential diagnosis with fibroblastic meningiomas, particularly of cerebellopontine angle tumors. A recent report by Hahn et al9 showed similar results, although the sensitivity was lower, with 21 (53%) of 40 meningiomas being immunoreactive and all other tumors being negative.
ASTROCYTOMA: PHH3, MIB-1/KI-67, AND p53
Diffusely infiltrating astrocytomas include (low-grade) diffuse astrocytoma (WHO grade 2), anaplastic astrocytoma (WHO grade 3), and glioblastoma (WHO grade 4), according to the current WHO classification.64 Distinguishing between WHO grade 2 and 3 infiltrating astrocytomas is particularly important for patient management as well as for prognosis. By current WHO guidelines, this distinction is made primarily by assessment of the proliferation activity of neoplastic cells. There are a few studies demonstrating a significant positive correlation between MIB-1 LI and tumor grade in diffusely infiltrating astrocytomas, classified according to the WHO 2000 classification system.65,66 A recent large retrospective study of grade 2 and 3 astrocytomas by Colman et al, evaluating the utility of PHH3 staining for determining proliferative activity, demonstrated that the PHH3 mitotic index (per 1000 cells) was significantly associated with the standard mitotic count (mitoses per 10 HPFs) and with the MIB-1 LI and had specific technical advantages over the MIB-1 LI67 because the latter showed significant interlaboratory variability, depending on staining conditions.68 For a practical usage of this marker, they pointed out that the antigenicity seemed to have decreased after 3 to 5 years in their samples. With regard to prognosis, they reported that the PHH3 index was an independent predictor of survival after adjusting for relevant clinical variables in multivariate analysis. Interestingly, they listed specific cutoffs to separate the patients into 2 groups with survival times similar to those established in a previously reported series for grade 2 and grade 3 astrocytomas. The cutoffs (grade 2 vs 3, respectively) were as follows: PHH3 index (≤4 vs >4 per 1000 cells), MIB-1 LI (≤9% vs >9%), and mitoses per 10 HPFs (≤3 vs >3).67 Although the current WHO guidelines do not define such a cutoff to distinguish between grade 2 and 3 astrocytomas, this information on the proliferation markers, in addition to the variable morphologic parameters, can be very useful to grade the tumors.
One of the most common diagnostic dilemmas in neuro-oncologic pathology is a distinction between benign reactive astrocytic lesions (ie, gliosis) and low-grade astrocytomas, especially with small biopsies (eg, stereotactic biopsies). In general, there should be negligible or very low levels of Ki-67/MIB-1 immunoreactivity in the setting of gliosis.68 This pattern of immunoreactivity in gliosis and that seen in some low-grade astrocytomas may overlap. Wild-type p53 is involved in regulation of the cell cycle as well as apoptosis,69 and it has been demonstrated to suppress cell transformation.70 The wild-type p53 protein has a short half-life (5–30 minutes) because of its rapid turnover, and is not normally detectable by standard immunohistochemical methods.71 Mutation of the gene usually leads to the production of a functionally impaired or altered protein, which retards degradation and thus can be detected via immunohistochemical staining. On the other hand, p53 immunoreactivity is sometimes unaccompanied by gene mutations. This pattern can be seen in the settings, such as binding of wild-type p53 by various oncoproteins (eg, mdm-2)72 and the result of epigenetic changes.73,74 Yaziji et al75 reported that p53 (monoclonal antibody, DO-7; Dako, Carpinteria, Calif; dilution, 1:60) immunoreactivity was seen in 12 (54%) of 22 low-grade astrocytomas and 5 (9%) of 56 reactive astrocytic lesions, all 5 being cases of progressive multifocal leukoencephalopathy. Given unusual p53 immunoreactivity seen in astrogliosis in their study, they concluded that p53 immunostain can help to differentiate reactive from neoplastic astrocytic lesions. In contrast, Kurtkaya-Yapicier et al76 conducted a similar study of 60 nonneoplastic lesions, including gliosis, infarction, demyelination, progressive multifocal leukoencephalopathy, and herpes simplex virus encephalitis, and 50 astrocytomas of WHO grades 2 to 4, using p53 antibody (monoclonal, DO-7; Dako; dilution, 1:200). They showed p53 immunoreactivity in all lesions examined, although the reactivity was low-level in most instances, and concluded that it was not a reliable indicator of astrocytic neoplasia. We believe that this distinction is still best handled on histologic grounds with the clinical, radiologic, and operative findings, although immunohistochemical staining for Ki-67 and p53 may be of help if the expression is significantly high.
MEDULLOBLASTOMA: SYNAPTOPHYSIN, MICROTUBULE-ASSOCIATED PROTEIN 2, NEUROFILAMENT PROTEIN, AND NEURONAL NUCLEI
The medulloblastoma (MDB) is defined as a malignant, invasive embryonal tumor of the cerebellum, preferentially occurring in children and adolescents, with a propensity for leptomeningeal dissemination. Medulloblastomas are the second most frequent brain tumors in childhood after pilocytic astrocytomas, and account for approximately 15% of all pediatric brain tumors.77 The median age at diagnosis is 9 years.77 Medulloblastomas have been chiefly subtyped as classic, nodular (desmoplastic), and large cell/anaplastic based on histologic appearances.78 Of these, the large cell/ anaplastic variant is known to be associated with worse prognosis.78,79
Although MDBs are derived from embryonal precursor cells with a capacity for divergent differentiation, neuronal differentiation is most consistently seen.80,81 This is usually incipient in that it is restricted to the expression of neuronal markers, with rare cases showing overt ganglionic or mature neuronal cells. Synaptophysin has proven to be a reliable marker of neuronal differentiation and is detected in virtually all cases on frozen sections,80,81 with 70% to 80% of cases being positive in paraffin sections.82 Microtubule-associated protein 2 antibody mirrors synaptophysin but has a more intense granular or punctate pattern of reactivity. It is often helpful in those cases where the synaptophysin staining is weak or equivocal. Neurofilament proteins of low and intermediate molecular masses (68 kd and 160 kd, respectively), are expressed in proliferating medulloblastoma cells,80 but high-molecular-mass neurofilament protein (200 kd) is only expressed in the tumor cells with advanced neuronal differentiation and overt ganglionic or neuronal morphology.83 Neuronal nuclei immunoreactivity is seen focally in the nuclei of cells with advanced neuronal differentiation.84
Glial (astrocytic) cell differentiation in MDBs is restricted to small foci of tumor cells without evidence of progressive differentiation to mature astrocytes. When strict criteria are applied for true tumor cells with glial (astrocytic) differentiation, excluding entrapped reactive astrocytes, the incidence is seen in up to 13% of cases.83 True tumor cell glial differentiation is defined as a typical medulloblastoma cell with hyperchromatic nucleus and scant cytoplasm, showing GFAP immunoreactivity restricted to perinuclear cytoplasm, and is associated with a poor prognosis.85
Mesenchymal, epithelial, and melanotic markers are seen in rare variants of MDBs. It should be noted that focal expression of epithelial markers (keratins and EMA) is otherwise rare in MDBs, and in particular, in the case of large cell/anaplastic subtype, raises the possibility of atypical teratoid/rhabdoid tumor.86
ATYPICAL TERATOID/RHABDOID TUMOR: INI1
Atypical teratoid/rhabdoid tumors (ATRTs) may form a histological spectrum from pure rhabdoid to atypical teratoid/rhabdoid tumors,87,88 and occur most commonly in young children (Figure 4). More than 90% of cases are diagnosed before age 5 years.89 Since their initial description, they have been reported to occur throughout the neuraxis, but the posterior fossa remains a preferred site, in particular the cerebellopontine angle. Neither clinical presentation nor neuroimaging distinguishes the ATRT from medulloblastoma.90 Microscopically, ATRTs are very cellular tumors that show marked regional heterogeneity, with primitive neuroepithelial, rhabdoid, epithelial, and mesenchymal components. There is often a fibrovascular stroma separating lobules and sheets of tumor cells. The cellular morphology varies from smaller, primitive neuroepithelial type cells, with hyperchromatic nuclei resembling those of a medulloblastoma, to large cells with eosinophilic, pale, or clear cytoplasm and large round nuclei with more open chromatin and prominent nucleoli. Immunohistochemistry reflects the morphological heterogeneity of the tumor. There is immunoreactivity for a range of mesenchymal, epithelial, and neuroectodermal markers, but the tumors are consistently negative for the germ cell markers. Vimentin is consistently expressed. Expression of EMA, keratin, smooth muscle actin, and GFAP is also frequently observed. S100 protein, synaptophysin, chromogranin, neurofilament protein, desmin, and HMB-45 may be variably and focally expressed in ATRTs. Glucose transporter protein 1 (GLUT-1) is expressed by ATRTs, and suggests origin from a stem cell.91 There is a high frequency of monosomy 22 in CNS ATRTs. Molecular cytogenetic screening has shown deletions of chromosome 22q11.2; this region contains the hSNF5/INI1 gene. Most CNS, renal, and extrarenal rhabdoid tumors show homozygous inactivation of INI1 by deletion and/or mutation of the INI gene, with decreased or absent expression at the RNA or protein level. Immunohistochemistry with antibody to INI1 (with absent nuclear staining in tumor cells in ATRT) has been shown to correlate with molecular findings in ATRT.10 The INI antibody may be more useful in analysis of tumors with indeterminate histologic and immunophenotypic profiles, since negative staining (albeit with preserved nuclear expression in normal components such as endothelial cells) is intuitively not as desirable an end result as compared with positive staining. In the diagnosis of ATRTs, a panel of immunohistochemical markers which include vimentin, EMA, keratins, smooth muscle actin, GFAP, and synaptophysin is likely to help confirm the diagnosis in the context of the appropriate morphological appearance. INI1 immunostaining can be used in those cases having indeterminate histological features, or in small biopsies which may not be representative of the morphological heterogeneity typical of ATRTs.
EPENDYMOMA: EMA
Ependymomas account for 3% to 9% of all neuroepithelial tumors.92 They are most often seen in children, adolescents, and young adults, but can be seen in older age groups. In children, they are the third most common CNS tumors after astrocytomas and medulloblastomas.77 Posterior fossa tumors are more frequent than supratentorial, and spinal cord ependymomas occur in older age groups than pediatric tumors. Ependymomas of the fourth and lateral ventricles occur in a ratio of 6:4; their location and growth pattern in the fourth ventricle may influence prognosis.93 Microscopically, ependymomas are well-demarcated, moderately cellular tumors with a monomorphic nuclear morphology. Characteristic features include perivascular pseudorosettes, which are nearly always seen in these tumors, and “true” ependymal rosettes. The latter feature is much less frequently seen in ependymomas and is characterized by clusters of ependymal cells arranged around a lumen with some resemblance to the central canal of the spinal cord. Immunohistochemistry in ependymomas reveals their dual nature with glial fibrillogenesis and GFAP expression, and epithelial with EMA expression. GFAP is variably positive in ependymomas, with the tumor cells and processes forming the perivascular pseudorosettes being most consistently positive. Normal mature ependymal cells do not express GFAP, but reactive and neoplastic ependymal cells reacquire the developmentally repressed ability to express GFAP which occurs from the 15th week of gestation, but is lost in adulthood. Other intermediate filaments such as vimentin and desmin are also expressed in neoplastic ependymal cells. EMA expression is useful in that it is consistently and widely expressed in well-differentiated tumors; anaplastic examples were not immunoreactive in one study,94 but in our experience are at least focally expressed. The pattern of EMA immunoreactivity typically is seen as dotlike perinuclear cytoplasmic reaction. Histological features by themselves are not reliable predictors of biological behavior, likely due to tumor heterogeneity. However, ependymomas with 2 or more of the features of hypercellularity, mitotic figures, elevated MIB-1 LI, microvascular proliferation, and necrosis are likely to show aggressive behavior.95 According to the recent study of ependymomas from 103 consecutive patients, Wolfsberger et al96 demonstrated that extent of resection and MIB-1 LI were independent prognostic factors on multivariate analysis. They defined its median value of 20.5% as cutoff point, and showed low (<20.5%) or high (≥20.5) MIB-1 LI predicted favorable (≥5 years' survival) or unfavorable (<5 years) patient outcome at 79% and 70%, respectively.
CONCLUSIONS
We presented several new IHC markers for supporting and at times confirming the morphologic diagnosis of adult and pediatric brain tumors. These include OCT4 for germinoma, α-inhibin (inhibin A) and D2-40 for CHB, claudin-1 for meningioma, microtubule-associated protein 2 and neuronal nuclei for MDB, and INI1 for ATRT. Of particular importance is the differential diagnosis of CHB from metastatic CRCC because of completely different prognostic as well as therapeutic significance. We stressed the combined use of inhibin A (and D2-40) and epithelial markers (EMA, CAM 5.2, and CD10) for this distinction. INI1 is unique in its negative nuclear staining in ATRTs, and can be used in those cases having indeterminate histological features or in small biopsies that may not be representative of the morphological heterogeneity typical of ATRTs. Another new marker, PHH3, is a mitosis-specific marker, and enables us to provide a quick focus on the most mitotically active areas within the tumor and to facilitate rapid, reliable grading in meningiomas. In astrocytomas, this antibody can be of particular help to differentiate between grade 2 and 3 tumors, compared with MIB-1/Ki-67. With this particular marker, mitotic index will be further investigated in the neoplasms, for which proliferation potential is of relevance to tumor grading and prognosis. We also showed the current data on MIB-1/Ki-67 LI in prognosis in meningioma, astrocytoma, and ependymoma.
From a practical point of view, an accurate diagnosis of brain tumors is usually possible after careful assessment of routine microscopic features with sufficient clinical and radiological information. Although conventional H&E staining is a mainstay for pathologic diagnosis, IHC has played a major role in differential diagnosis and in improving the diagnostic accuracy in neuro-oncologic pathology. The judicious use of a panel of IHC, whose selection was based on the differential diagnosis rendered after the initial assessment, is unquestionably helpful in diagnostically challenging cases. In addition, IHC is also reportedly of great help to grade and to predict the prognosis in certain brain tumors.
References
The authors have no relevant financial interest in the products or companies described in this article.
Author notes
Reprints: Hidehiro Takei, MD, Department of Pathology, Baylor College of Medicine, One Baylor Plaza, Suite 286A, Houston, TX 77030-3498 ([email protected])