Objective.—Poorly differentiated metastatic carcinoma may be difficult to distinguish histologically from high-grade astrocytic malignant neoplasms, particularly on small open or stereotactic biopsy specimens. Previous authors have reported that a subset of glioblastoma multiforme (GBM) variably stains with cytokeratin immunomarkers. The authors examined a panel of epithelial and keratin antibodies by paraffin immunohistochemistry to evaluate the immunophenotype of GBM for these markers and to determine what combination of immunostains would be optimal in distinguishing GBM from metastatic carcinoma.
Methods.—Twenty-three patients with GBM (age range, 19–86 years; mean, 63.4 years; 14 men and 9 women) and 22 patients with metastatic carcinoma (age range, 26–77 years; mean, 58.1 years; 7 men and 15 women) to the brain were studied with a panel of immunostains, including glial fibrillary acid protein (GFAP), Ber-EP4, antikeratin monoclonal antibodies AE1/3, and antibodies to CAM 5.2 and cytokeratins 7 (CK7) and 20 (CK20). Sites of origin for the metastatic tumors included lung (n = 11), breast (n = 5), endometrium (n = 1), prostate (n = 1), colon (n = 1), presumed kidney (n = 1), and unknown (n = 2).
Results.—All GBMs stained positive for GFAP (100%), and all but 1 (95.7%) stained positive for cytokeratins AE1/3. Only rare focal immunoreactivity was observed in a single case of GBM with CAM 5.2 (4.3%), CK7 (4.3%), and CK20 (4.3%). Immunoreactivity with Ber-EP4 was not observed in any of the GBMs (0.0%). All cases of metastatic carcinoma stained positive with cytokeratins AE1/3 (100%) and CAM 5.2 (100%). Variable staining was observed in carcinomas with CK7 (17 of 22, 77.3%), Ber-EP4 (11 of 22, 50.0%), and CK20 (9 of 22, 40.9%). Three metastatic carcinomas showed rare GFAP-positive staining cells (13.6%).
Conclusions.—Based on the aforementioned results, a combination of immunostains, including GFAP and cytokeratin CAM5.2, may be the most useful in differentiating poorly differentiated metastatic carcinoma from GBM. A significant number of GBMs stain with some cytokeratin markers, in particular cytokeratins AE1/3. Because of the poor specificity of cytokeratins AE1/3 in distinguishing metastatic carcinoma from GBM, it should not be used to differentiate the 2 entities.
Brain biopsies performed specifically for diagnostic purposes rather than for tumor removal are often small. Metastatic carcinoma, the most common malignant tumor seen in the central nervous system, may occasionally be difficult to distinguish histologically from a high-grade astrocytic malignant neoplasm, the most common primary malignant tumor in the central nervous system. Glioblastoma multiforme (GBM) is characterized histologically by the presence of nuclear atypia, high mitotic rate, and the presence of necrosis, all characteristics often encountered in poorly differentiated metastatic carcinoma.1 The absence of well-differentiated regions of metastatic carcinoma may make diagnosis of metastatic carcinoma versus GBM difficult by light microscopic examination alone. Further complicating the issue are reports of so-called epithelioid GBM lesions that can resemble carcinoma or melanoma more than glioma.2,3 Immunohistochemical stains may be useful to the pathologist in these limited specimens. However, previous authors have reported that a subset of GBM stains with certain cytokeratin immunomarkers, especially with antikeratin cocktails.4–8 Using immunomarkers to differentiate metastatic carcinoma from GBM may lead to an erroneous diagnosis of metastatic carcinoma if the pathologist is unaware that GBMs can stain with certain cytokeratin immunomarkers.8
There is relatively little information in the current literature describing the pattern of cytokeratin subset staining or selected antiepithelial, noncytokeratin antibodies in GBM. The current study examines a panel of readily available antiepithelial and anticytokeratin immunostains, including glial fibrillary acid protein (GFAP), cytokeratins AE1/3, CAM 5.2, cytokeratin 7 (CK7), cytokeratin 20 (CK20), and Ber-EP4 to determine what combination of immunostains is optimal in the evaluation of this differential diagnosis.
MATERIALS AND METHODS
Twenty-three cases of GBM and 22 cases of metastatic carcinoma to the brain were evaluated. Diagnoses were determined on hematoxylin-eosin–stained slides only in each of the cases evaluated. The source of metastasis was identified on review of the medical record or previous tissue diagnoses. In 2 patients, the primary source of metastasis was not identified. Routine tissue sections were generated from formalin-fixed (minimum 12 hours), paraffin-embedded material.
Slides generated for immunohistochemical staining were sectioned at 6-μm intervals and stained using an automated method (Ventana ES, Tucson, Ariz) involving an avidin-biotinylated immunoperoxidase method. The following immunoperoxidase stains were performed on each case: GFAP (1:600 dilution, Dako Corporation, Carpinteria, Calif), which reacts with the 51-kd intermediate filament protein commonly seen in astrocytomas9,10; CK7 (1:40, Dako), which reacts with a 54-kd cytokeratin intermediate filament protein as indicated by immunoblotting of cytokeratin-enriched cytoskeletal preparations isolated from human OTN 11 ovarian carcinoma cells and other cell lines11; CK20 (1:20, Dako), which reacts with the 46-kd cytokeratin intermediate filament protein as indicated by immunoblotting of cytokeratin-enriched cytoskeletal preparations isolated from villi of duodenal mucosa12; Ber-EP4 (1:10, Dako), which is a monoclonal mouse antihuman epithelial antigen antibody that reacts with 2 glycoproteins of 34 and 49 kd present on the surface and in the cytoplasm of all epithelial cells except the superficial layers of squamous epithelia, hepatocytes, and parietal cells13; CAM 5.2 (1:10, Becton Dickinson, San Jose, Calif), which is a mouse monoclonal antibody to cytokerain proteins 8 and 18, which are found in most epithelial cells with the exception of stratified squamous epithelium14,15; and cytokeratins AE1/3 (1:200, Boehringer Mannheim, Indianapolis, Ind), which is a mixture of monoclonal antibodies to human epithelial keratins with specificity for the 56.5-, 50-, 50′-, 48-, and 40-kd keratins of the acidic subfamily and all members of the basic subfamily.16,17 Protease treatment for antigen retrieval was performed with stains for cytokeratins AE1/3, CAM 5.2, CK7, and CK20. Positive staining was defined as any cytoplasmic or membranous staining in histologically malignant cells. Appropriate positive and negative controls were performed with each immunostain.
Twenty-three cases of GBM were evaluated in this study. The subjects consisted of 14 men and 9 women, who ranged in age from 19 to 86 years (mean age, 63.4 years). All the astrocytic tumors histologically fulfilled criteria for the diagnosis of GBM and were marked by frequent mitoses, nuclear atypia, vascular endothelial proliferation, and necrosis.18 One tumor demonstrated a prominent epithelioid appearance (epithelioid GBM).
The immunohistochemical staining results are summarized in the Table. All cases of GBM expressed GFAP, and all but one case of GBM stained positive for cytokeratins AE1/3 (Figure 1). In 20 of 23 GBMs, staining was positive for GFAP in at least 50% of malignant cells. Thirteen of 23 GBMs stained positive for cytokeratins AE1/3 in at least 50% of cells. Only 1 GBM stained for CAM 5.2, CK7, and CK20 (Figure 2). This single case contained only a single small focus of positive staining cells for all 3 cytokeratins. None of the GBMs stained positive for Ber-EP4.
Twenty-two cases of metastatic carcinoma to the brain were evaluated in this study. The patients consisted of 7 men and 15 women, ranging in age from 26 to 77 years (mean age, 58.1 years). Sites of origin were known in 20 cases and included lung (n = 11), breast (n = 5), endometrium (n = 1), prostate gland (n = 1), colon (n = 1), and presumed kidney (n = 1).
Only 3 cases of metastatic carcinoma stained with GFAP. The staining in each of these cases was focal and limited to less than 10% of malignant cells. All cases of metastatic carcinoma stained with cytokeratins AE1/3 and CAM5.2 (Figure 3). The metastatic carcinomas were widely reactive for these 2 immunostains, with positive staining present in most malignant cells present. The only case of metastatic carcinoma in which less than 50% of malignant cells present stained for CAM5.2 occurred in the single case of metastatic breast carcinoma. Half of the metastatic carcinomas stained positively with Ber-EP4. Most carcinomas, including all but one lung tumor, stained with CK7; CK20 staining was observed in less than half the carcinomas (40.9%).
Given the rather high incidence of metastatic carcinoma and GBM (most common primary glial neoplasm of central nervous system), both of these lesions are fairly commonly encountered in the routine surgical neuropathology practice. Being able to distinguish one from the other is important from a clinical management standpoint. In most cases, the histologic features of each of these lesions are distinctive enough to allow diagnosis. However, in this age of stereotactic biopsies, occasionally a neoplasm that is difficult to definitively classify as either metastatic carcinoma or high-grade glioma is encountered. Particularly problematic are tumors in which most of the small sample is necrotic and only small groups of viable cells remain. In addition, recognition of the so-called epithelioid variant of GBM may further complicate this issue. Immunohistochemistry has, in recent years, been useful in allowing the pathologist to distinguish tumors from each other. Classically, GFAP has been associated with astrocytic neoplasms, including GBM. Likewise, a variety of cytokeratin markers have been devised for the recognition of carcinomatous lesions.
Intuitively, one would assume that use of GFAP with one of these keratin markers would easily allow one to distinguish these 2 lesions in the occasional case when routine light microscopic evaluation proves insufficient. Unfortunately, a number of studies have documented cytokeratin immunoreactivity in glial neoplasms. In an immunohistochemical analysis of 30 paraffin-embedded astrocytic neoplasms, Cosgrove et al5 noted immunoreactivity with antikeratin monoclonal antibodies AE1/3 in 24 tumors (80%). In a series of 25 astrocytic tumors including 12 GBMs, Ng and Lo noted a positive immunostaining with cytokeratins AE1/3 in 56% of astrocytomas.7 Other studies have similarly shown cross-immunoreactivity of astrocytic tumors with intermediate keratin filaments.4,6,19 Recently, Kriho and colleagues20 looked for the presence of keratin in astrocytomas by immunofluorescence on frozen section tissue and by 1-dimensional and 2-dimensional immunoblotting using a panel of 7 monoclonal antibodies. They concluded that positive immunofluorescence staining of astrocytomas with 3 of the 7 keratin antibodies was due to cross-reactivity with a nonkeratin protein such as GFAP. Therefore, it seems that most of the immunoreactivity frequently observed in gliomas using cytokeratins AE1/3 is due to a nonspecific cross-reactivity, perhaps with other intermediate-molecular-weight filaments within the cytoplasm of astrocytic cells as opposed to reactivity with true keratin filaments. In our experience, most GBMs stained positive with cytokeratins AE1/3, making their utility in distinguishing carcinoma from glioma suboptimal.
Cytokeratin CAM5.2, in contrast, seems to be a more useful stain in that all the carcinomas examined in this study have stained positive and only one glioma showed focal immunoreactivity. In the 25 astrocytomas stained by Ng and Lo, none of them were reported to have stained with CAM5.2.7 Cosgrove et al4 noted CAM5.2 immunoreactivity in 1 of 29 astrocytic tumors studied. All 4 epithelioid GBMs reported by Rosenblum et al3 failed to stain with CAM5.2. It is our feeling that a combination of GFAP with cytokeratin CAM5.2 may be most useful in sorting out this differential diagnosis, although cross-reactivity of GFAP with keratin filaments may result in rare staining of carcinomas.21
Use of a variety of other markers, including EMA, CK7, CK20, and Ber-EP4, would appear to be less useful in distinguishing carcinoma from glioma, primarily because of the fact that these only stain a subset of carcinomas. Immunoreactivity with these stains and GBM appears to be least as limited as CAM5.2 in this study. Although the CK7 and CK20 markers may not be particularly helpful with regard to this differential diagnosis, particularly with a negative staining result, there may be some role for these stains in pinpointing a site of origin in a known metastatic tumor. Perry et al,22 in a recent study of metastatic adenocarcinomas to the brain, demonstrated the utility of CK7 and CK20 markers in helping to determine sites of origin.22 The CK7 expression is well known to be associated with carcinomas of lung, breast, ovarian, pancreatic, and bladder (transitional cell) origin. In contrast, CK20 is useful in staining adenocarcinoma of gastrointestinal origin, Merkel cell tumors, transitional cell carcinomas, and mucinous tumors of the ovary. Although one study noted positive staining with Ber-EP4 in 19 of 20 metastatic carcinomas and in none of the astrocytic tumors evaluated,23 only 50% of the metastatic carcinomas in our experience stained positive with this antibody.
Although none of the antibodies examined in this study are perfect, the best combination of immunomarkers to use in most easily differentiating carcinoma from GBM includes GFAP and cytokeratin CAM5.2. In the rare instances in which a neoplasm fails to stain with either antibody, a number of possible explanations may be forthcoming. Immunogenicity of the paraffin block needs to be confirmed. The possibility of a malignant glioma of oligodendroglial lineage may also account for a failure to stain with GFAP. Also, GFAP immunoreactivity is frequently focal because of the highly undifferentiated nature of cells that can comprise this tumor. In rare cases that stain with neither antibody, use of a more extensive panel of keratin markers may also be helpful in ruling out carcinoma. Because of its high degree of cross-immunoreactivity, use of cytokeratins AE1/3, particularly with regard to this differential diagnosis, is not advocated.
Special thanks to Denise Egleton for her help in the preparation of the manuscript.