Laminin and type VII collagen distribution in different types of human lung carcinoma: correlation with expression of keratins 14, 16, 17 and 18

The expression patterns of basement membrane components and keratin intermediate filament proteins were studied in normal human bronchial epithelium and 56 lung carcinomas using monoclonal antibodies to laminin, type VII collagen and the individual keratins 14, 16, 17 and 18. In normal lung, laminin and type VII collagen were present between the epithelium and the lamina propria of bronchi and bronchioles. Keratin 14 was expressed in the basal cells, keratin 17 in the basal and some suprabasal cells and keratin 18 in the columnar cells of the bronchi and bronchioles. Keratin 16 was not present in normal bronchial epithelium. Laminin was found in all subtypes of lung carcinoma, but type VII collagen was present only in squamous cell carcinomas, where it showed a reduction in expression with decreasing differentiation. Type VII collagen was not identified in adenocarcinomas, small cell carcinomas or carcinoids. Antibodies to basal cell keratins 14 and 17 also displayed positivity only in squamous cell carcinomas, although no correlation with the degree of differentiation could be observed. Keratin 16 appeared to be a marker of the squamous phenotype, rather than of hyperproliferation. The keratin 18 marker for columnar epithelial cells showed a reaction pattern opposite to that of the basal cell keratins, being extensively present in adenocarcinomas, small cell carcinomas and carcinoids, with less expression in squamous cell carcinomas. This study shows a correlation between the presence of type VII collagen and the basal cell keratins 14 and 17, and a negative correlation between these components and keratin 18. These findings are likely to be useful in identifying lung cancer subtypes.     

Intracellular keratins in normal and pathological bronchial mucosa

Summary The distribution of intracellular keratins was investigated in normal bronchial epithelium and in several morphologically distinct forms of respiratory tract carcinomas. This study was performed with two different experimentally produced antisera against normal human stratum corneum keratin and against keratin protein of MW 67000 dalton, using indirect immunofluorescence and immunoperoxidase methods on tissue sections and cell suspensions.In normal bronchial epithelium, the basal cells were strongly labelled by both antisera. The ciliated columnar cells appeared devoid of cytokeratins in tissue sections but were strongly labelled with both antisera in cell suspensions. The goblet cells remained negative in every case. In squamous metaplasia of the bronchus, all epithelial cells were unevenly stained with both antisera.Among tumours, only the squamous cell carcinomas were strongly labelled by both antisera. Primary lung adenocarcinoma appeared weakly positive, whereas metastatic lung carcinomas, undifferentiated lung carcinomas, oat cell tumours, carcinoid tumours were negative.The immunocytochemical determination of keratins appears to be of value in the study of normal and abnormal epithelial differentiation, in the diagnosis of poorly differentiated carcinomas and in their distinction from metastatic tumours of the lung.     

Immunocytochemical localization of neuron specific enolase in small cell carcinomas and carcinoid tumours of the lung

Carcinoid tumours and small cell carcinomas of the lung share many characteristics with normal neuroendocrine cells. While carcinoid tumours contain many dense-cored neurosecretory granules and are frequently argyrophil, small cell carcinomas are poorly granulated and rarely argyrophil, which casts doubt on their neuroendocrine nature. Immunostaining of the enzyme neuron specific enolase (NSE) was recently used to demonstrate the neuroendocrine components of the lung including nerves and neuroendocrine cells. We therefore used NSE immunostaining to investigate neuroendocrine differentiation in 79 lung tumours, including 18 bronchial carcinoids and 31 small cell carcinomas, and compared these results with those obtained with silver stains. Thirteen of the 18 carcinoids were reactive to silver, all other types being negative. NSE-immunoreactivity occurred in 16 carcinoids and 18 small cell carcinomas. None of the squamous cell carcinomas, large cell anaplastic carcinomas and adenocarcinomas examined showed NSE-immunoreactivity. Radioimmunoassay of extractable NSE from 10 fresh lung tumours correlated well with the immunostaining results, demonstrating large amounts in two small cell carcinomas (334 and 517 ng/mg protein) and three carcinoids (152, 908, and 1143 ng/mg protein). Values were much lower for four squamous cell carcinomas (31-44 ng/mg protein) and one large cell anaplastic carcinoma (30 ng/mg protein) and were accounted for by the presence of NSE-positive nerves and neuroendocrine cells in the surrounding lung. NSE immunostaining is a useful marker of neuroendocrine differentiation in lung tumours and should prove particularly valuable in the diagnosis of small cell anaplastic tumours and their metastases.     

P63 in pulmonary epithelium,pulmonary squamous neoplasms,and other pulmonary tumors

p63 is a p53-homologous nuclear protein that appears to play a crucial role in regulation of stem cell commitment in squamous and other epithelia. In this study, p63 expression was examined in benign lung and in neoplasms of pulmonary origin. Eighty sections from routinely fixed and processed archival bronchoscopic biopsy or lobectomy specimens were pretreated with citric acid (pH 6.0) for antigen retrieval, then incubated overnight with anti-p63 monoclonal antibody 4A4. Slides were stained using a streptavidin-biotin kit and diaminobenzidine as chromagen, and were counterstained with hematoxylin. In normal lung, p63 intensely stained nuclei of bronchial reserve cells but did not stain ciliated cells, alveolar epithelial cells, or nonepithelial cells. The lower strata of squamous metaplastic bronchial epithelium stained positively. All squamous-cell carcinomas stained positively (n = 30). In some well-differentiated carcinomas, staining was found at the periphery of tumor nests but was negative in central zones showing squamous maturation. Poorly differentiated carcinomas showed very high proportions (80% to 100%) of p63-positive nuclei. All small-cell carcinomas were p63 negative (n = 9). Staining of bronchioloalveolar carcinomas (n = 7) and adenocarcinomas (n = 23) was variable: some tumors showed no detectable staining, others showed heterogeneously positive staining. Adenosquamous carcinomas (n = 5) displayed a unique basalar staining pattern. Carcinoid tumors were almost entirely negative (n = 5). We conclude that p63 is expressed in benign bronchial stem cells, in neoplastic cells with either squamous differentiation or squamous differentiating potential, and in a subpopulation of adenocarcinomas. p63 immunostaining may also aid in some histopathologic distinctions, such as in small biopsies where the differential diagnosis is poorly differentiated squamous carcinoma versus small-cell carcinoma. A stem cell biology-based classification system for squamous carcinomas is proposed.     

Cytokeratin and neurofilament in lung carcinomas.

Three monoclonal antibodies, one directed against cytokeratin (clone 80) and two directed against neurofilament (clones 2F11 and 3G6), were used in the study of a series of 77 lung carcinomas by immunohistochemical staining. The anti-cytokeratin antibody, a very broadly reacting antibody directed against an antigenic determinant common to a great number of cytokeratins, was applicable on frozen sections. The two anti-neurofilament antibodies, directed against the 70 kD protein (clone 2F11) and the 160 kD and 200 kD proteins (clone 3G6) of neurofilament, were applicable on both frozen sections and paraffin sections. The staining results on the lung carcinomas indicate that all types of tumors studied, including small-cell anaplastic carcinoma, are markedly positive for cytokeratin. Frozen sections of five and formalin-fixed and paraffin-embedded sections of six other small-cell anaplastic carcinomas were negative with both anti-neurofilament monoclonal antibodies. One poorly differentiated squamous cell carcinoma positive with anti-neurofilament clone 2F11 but negative with clone 3G6. This distribution of cytoskeletal proteins demonstrates the epithelial differentiation of all types of lung carcinomas. Neuroendocrine differentiation of lung carcinomas as found in the small-cell anaplastic types does not result in expression of neurofilament proteins.     

Use of monoclonal antibodies to keratin 7 in the differential diagnosis of adenocarcinomas.

Monoclonal antibodies (MAbs) to specific keratin subtypes were prepared and characterized by immunoblotting and immunohistochemical assays on human cell cultures and normal and malignant human tissues. Chain-specific MAbs to keratin 7 (RCK 105, OV-TL 12/30) and keratin 18 (RGE 53, RCK 106, CK18-2), as well as broadly cross-reacting keratin MAbs (RCK 102, OV-TL 12/5) could be shown to react with different types of human epithelial tissues and were therefore tested for their usefulness in the differential diagnosis of carcinomas. The two broad-spectrum antibodies stained virtually all of the more than 350 carcinomas tested, especially when combined, and distinguished them from most nonepithelial tumors. The keratin 18 MAbs distinguished adenocarcinomas (which are keratin 18 positive) from most squamous cell carcinomas (which are generally keratin 18 negative). The MAbs to keratin 7 could be shown to recognize specific subtypes of adenocarcinoma and could, for example, distinguish between ovarian carcinomas (keratin 7 positive) and carcinomas of the gastrointestinal tract (keratin 7 negative), or between transitional cell carcinomas (keratin 7 positive) and prostate cancer (keratin 7 negative). In general, malignancies showed the expected keratin reactivity pattern as concluded from the keratin pattern of its cell of origin or its type of differentiation. The use of an extended series of malignancies did, however, also illustrate that exceptions to this rule exist. For example, certain antibodies to keratin 18 stained tumor areas in squamous cell carcinomas of the lung. Also a certain percentage of tumors, which generally showed no keratin 7 expression, were positive with RCK 105 or OV-TL 12/30. On the other hand, a certain percentage of tumors, which were generally positive for keratin 7, did not show a staining reaction with these MAbs. Furthermore subtle differences between reactivity patterns of different MAbs recognizing the same keratin protein were observed, both in the normal and malignant human tissues, indicating that specific keratin epitopes may be masked in certain tissues and that unmasking of such epitopes can occur with malignant progression. This phenomenon may be of some use in a further subtyping of carcinomas, especially those of the gastrointestinal tract. Despite these exceptional staining patterns, the keratin MAbs described above have proved to be useful tools in the characterization of epithelial tumors in routine histopathology and cytopathology, in which they add to a more refined diagnosis of (adeno)carcinomas.     

Expression and localization of thrombomodulin in preneoplastic bronchial lesions and in lung cancer

Thrombomodulin (TM) is an endothelial surface glycoprotein that acts as a natural anticoagulant. It inhibits thrombin and accelerates the activation of the anticoagulant protein C. TM has been detected in dermal keratinocytes, where it is associated with terminal differentiation. It can also be detected in various types of squamous malignant neoplasms and in malignancies of endothelial and mesothelial origin, such as Kaposi's sarcoma or malignant mesothelioma, but is absent in pulmonary adenocarcinomas (AC). Seventy-two lung tumour specimens [33 squamous cell carcinomas (SQCC), 23 AC, 1 large cell carcinoma, 8 small cell lung cancers (SCLC) and 7 multidifferentiated tumours (MT)] were analysed immunohistochemically by staining with an anti-TM antibody in order to assess TM expression. All of the SQCC stained positively for TM. In contrast, only 9 AC and 4 MT and none of the SCLC showed positive anti-TM staining. Seven hyperplastic bronchial epithelial specimens and eight preneoplastic bronchial lesions (five cases of moderate dysplasia, two cases of severe dysplasia and one case of carcinoma in situ) were used as controls.Normal or hyperplastic areas of bronchial epithelium revealed no positive reaction. However, a distinct positive anti-TM staining pattern related to the degree of keratiniziation of dysplastic lesions was seen. The present results suggest that anti-TM immunostaining is a useful marker for squamous cell carcinoma in the differential diagnosis of pulmonary carcinoma, also indicating keratinocyte differentiation in dysplastic bronchial epithelium.     

Carcinomas of the lung: an ultrastructural and immunocytochemical study

Fifty-two primary carcinomas of the lung were studied by electron microscopy and by an immunoperoxidase method, using an anti-human keratin antiserum. The results were compared with light microscopic observations. One-third of the carcinomas of the lung showed ultrastructural evidence of both glandular and squamous differentiation. The group of small cell carcinomas was found to be particularly heterogenous ultrastructurally with only three out of eight tumors showing neurosecretory-type granules. Indirect immunoperoxidase staining revealed presence of a keratin-type protein in the vast majority of carcinomas, including foci of small cell carcinomas. Our studies emphasize the heterogeneity and frequent intermixing of the four major categories of lung carcinomas: squamous cell carcinomas, adenocarcinomas, small cell carcinomas, and large cell carcinomas. It is suggested that all these tumors might be derived from pluripotential "reserve" bronchial or bronchioalveolar cells. The segregation of small cell carcinomas from other groups continues to be justified on pragmatic grounds because these carcinomas constitute a group of predominantly small fast-replicating cells amenable to chemotherapy.     

Are keratin proteins a better tumor marker than epithelial membrane antigen? A comparative immunohistochemical study of various paraffin-embedded neoplasms using monoclonal and polyclonal antibodies

Epithelial membrane antigen and keratin proteins represent markers of epithelial differentiation that may be detected in routine formalin-fixed, paraffin-embedded tissues. Eighty-seven neoplasms, including 48 adenocarcinomas of various types, squamous and transitional cell carcinomas, small-cell anaplastic carcinomas, carcinoid tumors, mesotheliomas, hepatomas, melanomas (metastatic), adrenal cortical carcinomas, germ cell tumors, and extramammary Paget's disease, were assessed to determine the relative effectiveness of these antigens as tumor markers. Immunoperoxidase studies were performed using monoclonal antibodies to epithelial membrane antigen and monoclonal (combined AE1 and AE3) and polyclonal (bovine muzzle keratins) antibodies to keratin proteins. In more than half the cases (50/87%), both markers yielded comparable results. However, in 29 cases (33%), keratin proteins were clearly superior to epithelial membrane antigen as a tumor cell marker. Particular discrepancies were apparent for some gastrointestinal adenocarcinomas, squamous cell carcinomas, hepatomas (hepatocellular type), spindle cell components of mesotheliomas, and carcinoid tumors. Epithelial membrane antigen represented a better marker in eight cases (9%), mainly for small-cell anaplastic carcinomas and some renal cell and pulmonary adenocarcinomas. Adrenal cortical carcinomas, melanomas, and seminomas were nonimmunoreactive for both antigens. Epithelial membrane antigen and keratin proteins represent useful complementary markers in diagnostic surgical pathology.     

The significance of alpha-fetoprotein and other tumour markers in differential immunocytochemistry of primary liver tumours

One hundred benign and malignant primary liver tumours were screened immunocytochemically for alpha-fetoprotein (AFP), alpha 1-antitrypsin, alpha-human chorionic gonadotropin, carcinoembryonic antigen (CEA), keratin and vimentin. Alpha-fetoprotein was found in 16/63 (24%) hepatocellular carcinomas and in two hepatoblastomas. When comparing tissue positivity for AFP with tumour differentiation, grade 1 hepatocellular carcinomas were found to be negative, while 21% of grade 2, 36% of grade 3 and 16% of grade 4, respectively, stained positively. Alpha-fetoprotein positive cells were present in 9/10 hepatocellular carcinomas with serum levels exceeding 5000 ng/ml, but were absent in 17 tumours with serum AFP levels below 5000 ng/ml. All tumours other than hepatocellular carcinomas and hepatoblastomas were AFP negative. Carcinoembryonic antigen was present in 72% of cholangiocarcinomas, but was demonstrated in only one hepatocellular carcinoma. This exception was a combined hepatocellular-cholangiocarcinoma in which CEA expression was restricted to the cholangiocellular part. Alpha 1-antitrypsin was found in 4/63 hepatocellular carcinomas, in 2/2 fibrolamellar carcinomas and in 2/18 cholangiocarcinomas. Alpha-human chorionic gonadotropin was detected in one hepatocellular carcinoma and was strongly expressed in both fibrolamellar carcinomas. Weak staining for keratin was seen in most tumours with hepatocellular differentiation. All cholangiocarcinomas, in contrast, were strongly labelled with the keratin antibody. Co-expression of keratin and vimentin was observed in seven poorly differentiated hepatocellular carcinomas and three cholangiocarcinomas as well as in the two hepatoblastomas. The findings suggest that AFP is a diagnostic but rather insensitive immunocytochemical marker for hepatocellular differentiation in malignant liver tumours; CEA and keratin may help in discriminating cholangiocarcinomas from hepatocellular carcinomas.     

Immunoreactivity for epithelial and neuroendocrine antibodies are useful in the differential diagnosis of lung carcinomas

The histologic classification of pulmonary neoplasms can have important implications regarding appropriate management of patients. Although the histologic classification of lung tumors is predominantly based on morphology, ancillary studies such as immunohistochemistry can be used in difficult cases, and the diagnosis of large cell neuroendocrine carcinoma requires confirmation of neuroendocrine differentiation by immunohistochemistry or electron microscopy. We immunostained 142 lung tumors for B72.3, keratin 34betaE12, keratin 7, keratin 14, keratin 17, synaptophysin, and chromogranin to determine the utility of neuroendocrine markers and epithelial markers in the differential diagnosis. Among neuroendocrine carcinomas (small cell carcinoma and large cell neuroendocrine carcinoma), 84% (37 of 44) were chromogranin positive, 64% (21 of 36 small cell, 6 of 6 large cell neuroendocrine) were synaptophysin positive, 5% (2 of 43) were keratin 34betaE12 positive, 9% (4 of 44) were keratin 7 positive, and 5% (2 of 37) of small cell carcinomas and 50% (3 of 6) of large cell neuroendocrine carcinomas were B72.3 positive. Among non-neuroendocrine carcinomas, 5% (5 of 98) were chromogranin positive, 3% (3 of 96) were synaptophysin positive, and 97% (95 of 98) were positive for either keratin 34betaE12 or keratin 7 and 99% (97 of 98) were positive for either keratin 34betaE12, keratin 7 or B72.3. An antibody panel consisting of keratin 7, keratin 34betaE12, chromogranin, and synaptophysin separated 132 of 141 tumors (94%) into distinct groups. We conclude that immunostaining with both neuroendocrine markers and epithelial markers can be useful in the differential diagnosis of lung neoplasms.     

The limited difference between keratin patterns of squamous cell carcinomas and adenocarcinomas is explicable by both cell lineage and state of differentiation of tumour cells.

AIM: To study the differentiation of epithelial tissues within their histological context, and to identify hypothetically, on the basis of keratin pattern, the putative tissue origin of a (metastatic) carcinoma. METHODS: Using well characterised monoclonal antibodies against individual keratins 7, 8, 18, and 19, which are predominantly found in columnar epithelia, and keratins 4, 10, 13, and 14, predominantly expressed in (non)-keratinising squamous epithelia, the keratin patterns for a series of 45 squamous cell carcinomas and 44 adenocarcinomas originating from various epithelial tissues were characterised. RESULTS: The predominant keratins in all adenocarcinomas proved to be 8, 18, and 19. In addition, these keratins were also abundantly present in squamous cell carcinomas of the lung, cervix, and rectum and, to a lesser extent, of the larynx, oesophagus, and tongue, but not in those of the vulva and skin. Keratins 4, 10, 13, and 14 were present in almost all squamous cell carcinomas, but also focally in some of the adenocarcinomas studied. CONCLUSIONS: There is a limited differential expression of distinctive keratin filaments between squamous cell carcinomas and adenocarcinomas. Apparently, squamous cell carcinomas that originate from columnar epithelium by squamous metaplasia gain the keratins of squamous cells but retain the keratins of columnar epithelial cells. However, the simultaneous expression of two of three squamous keratins (4, 10, and 13) identifies a squamous cell carcinoma, and thus might be useful in solving differential diagnostic problems.     

Monoclonal anti-keratin (AE1) reactivity in routinely processed tissue from 166 human neoplasms

A large number of human neoplasms were tested for their keratin expression in routinely processed tissues by a simple, three-stage immunoperoxidase method using a broadly reactive monoclonal anti-keratin antibody AE1, which recognizes a number of keratin polypeptides distributed in a wide variety of epithelia. All carcinomas, with the exception of hepatocellular, adrenocortical, and basal cell carcinomas and occasional renal cell, pulmonary small-cell, and pulmonary large-cell anaplastic carcinomas, reacted with this antibody irrespective of differentiation, in most instances displaying staining of strong or moderate intensity in the majority of tumor cells. Equivocal results were obtained in some seminomas and dysgerminomas. Malignant melanoma, large-cell lymphoma, Hodgkin's disease, malignant histiocytosis, and stromal mesenchymal elements in all tumors did not show any reactivity with AE1. Even after routine processing, the determinant detected by AE1 is conserved and restricted to epithelial neoplasms. This suggests that AE1 would be valuable in the diagnostic distinction of anaplastic carcinoma from lymphoma and melanoma in routinely processed tissues.     

IMMUNOREACTIVITY FOR HEPATOCYTE GROWTH FACTOR/SCATTER FACTOR AND ITS RECEPTOR,met, IN HUMAN LUNG CARCINOMAS AND MALIGNANT MESOTHELIOMAS

Paraffin sections from 29 lung carcinomas (28 primary and 1 metastatic) and 9 pleural malignant mesotheliomas were immunostained with antisera to human hepatocyte growth factor/scatter factor (HGF/SF) and its receptor, met. For HGF/SF, immunoreactivity was demonstrated in all 9 mesotheliomas, 9 of 12 adenocarcinomas, and 7 of 10 squamous cell carcinomas. None of seven cases of small cell anaplastic carcinoma was positive. The adenocarcinomas frequently showed enhanced luminal staining, suggesting possible secretion of HGF/SF, and this pattern of staining was also seen occasionally in bronchial epithelium adjacent to the tumour. Stromal fibroblasts also showed immunoreactivity for HGF/SF in 6/8 cases of mesothelioma but in only 3/12 adenocarcinomas, 1/10 squamous cell carcinomas, and 1/4 small cell anaplastic carcinomas. All tumours stained for met, usually strongly. The staining was mainly cytoplasmic in nature, but some plasma membrane staining was usually evident. Adenocarcinomas showed strong luminal membrane staining, as did adjacent, histologically normal bronchial epithelium. This study demonstrates the presence of HGF/SF and met in most of the tumour types described, particularly mesotheliomas, and suggests that the HGF/SF/met signalling system may play a role in the development of these tumours, either by autocrine or by paracrine mechanisms.     

Expression of occludin,a tight-junction-associated protein,in human lung carcinomas

Occludin is a tight-junction-associated transmembrane protein, and previous observations suggested that occludin might play a crucial role in the formation and maintenance of organized tubular structures. Based on these observations, we explored the possible role of occludin immunostaining in the diagnosis of lung carcinomas. A total of 68 lung carcinomas and surrounding normal lung tissues were studied. A formalin-fixed, paraffin-embedded section from each tumor was stained with a new anti-occludin monoclonal antibody raised in our laboratory. In normal lung tissues, the anti-occludin antibody strongly stained the apicoluminal borders of the bronchial/bronchiolar epithelia and bronchial glands as a dot or short line. The antibody also stained the intercellular borders of alveolar epithelia. In cancer cells that faced lumina of all adenocarcinomas, regardless of grade, including bronchioloalveolar carcinomas, occludin showed an expression pattern identical to that of the normal bronchial and alveolar epithelia. Occludin reactivity was not noted in any cases of squamous cell carcinoma, large cell carcinoma, small cell carcinoma, or large cell neuroendocrine carcinoma. The results of the present study suggest that occludin can serve as an immunohistochemical indicator of the true glandular differentiation that forms tubulo-papillary structures in human lung carcinoma tissues.     

Involucrin in lung tumors. A specific marker for squamous differentiation

Involucrin is a precursor of the cross-linked envelope protein or marginal band present in human stratum corneum. This study uses immunohistochemical techniques for localization of involucrin in histologic sections from 91 lung tumors in order to evaluate the usefulness of involucrin as a tumor marker in lung neoplasms. Although involucrin is absent from bronchial epithelium, it is expressed in cultured tracheal epithelial cell colonies and in bronchial mucosa with squamous metaplasia. Involucrin was present in all 25 cases of squamous and adenosquamous carcinoma. Staining was focal in 12 cases of squamous cell carcinoma and was most marked in the larger neoplastic cells in the center of squamous cell nests. Only two of 20 cases of adenocarcinoma revealed focal staining for involucrin, and these cases may represent adenosquamous variants. Six of 12 cases of large cell undifferentiated carcinoma stained for involucrin, indicating squamous differentiation, and seven cases of malignant mesothelioma were negative. Isolated involucrin-positive cells were present in two of 16 cases of small cell anaplastic carcinoma and one of 11 carcinoid tumors, identifying variants of neuroendocrine tumors with dual differentiation. Patterns of localization of involucrin in paraffin and frozen sections were compared with staining for cytokeratins in parallel sections. Immunohistochemical localization of involucrin comprises a specific marker for squamous differentiation in lung tumors.     

Immunohistology of anaplastic thyroid carcinoma. A study of 43 cases

Forty-three cases of large cell anaplastic thyroid carcinoma were examined with various antisera. Four histological patterns were identified: spindle cell, giant cell, trabecular and squamous. In 38 cases the epithelial origin was demonstrated with various epithelial markers: 11 cases stained positively for thyroglobulin, 19 for T3 and/or T4, 35 for fat globule membrane antigens, 28 for keratin, 29 for lactoferrin, and one for calcitonin. Five cases were negative for all epithelial markers but could not be characterized further since, except for vimentin, they did not have mesenchymal markers. The immunohistochemical proof of the epithelial nature of anaplastic thyroid tumours is given by staining with anti-keratin and anti-fat globule membrane antigen. In addition the detection of lactoferrin seems useful, but not that of thyroglobulin which was present in only 4% of the spindle cell tumours and in 32-56% of the other histological types. Thirty carcinomas were positive for vimentin; the co-expression of vimentin and epithelial markers seems frequent in thyroid anaplastic carcinomas.