Carcinoma of unknown primary is a particularly difficult clinical problem with up to 10–15% of all solid cancers presenting in this manner (Nystrom et al., 1977;Greenberg and Lawrence, 1988). Many tumours are poorly differentiated, giving no hint of their origin, while the histological appearance of moderate to well-differentiated adenocarcinomas and squamous cell carcinomas are generally not specific for any organ. Determining the site of the primary carcinoma is of paramount importance because treatment of carcinomas arising at different sites varies widely. The widespread use of antibodies with broad specificity to the cytokeratins has allowed the identification of many poorly differentiated tumours as epithelial tumours. Additionally, there are antibodies that help to subclassify the origin of tumours, particularly prostate-specific antigen (PSA) and prostate alkaline phosphatase (PrAP) in the case of prostate cancer, α-fetoprotein (AEP) in the case of hepatocellular carcinoma, breast cystic disease fluid in the case of breast cancer, and thyroglobulin (TGB) and calcitonin in the case of carcinomas of the thyroid. However, these antibodies have limited utility because metastases from these tumours without knowledge of the primary are rare. For most carcinomas, including those originating in the lung, colon, and urothelium, specific antibodies have not been available to aid in their identification.
Cytokeratins comprise a family of at least 20 different proteins distributed in a tissue-specific manner (Franke et al., 1982;Moll et al., 1982). Tumours that divide from these different tissues tend to maintain their tissue-specific cytokeratin profile (Cooper et al., 1985;Ueda et al., 1993). As such, determination of the cytokeratin profile of a particular carcinoma can be a useful aid in determining the primary site of tumours presenting as metastatic disease from clinically unknown primary sites of origin.
Among the most useful cytokeratins are CK7 and CK20. CK7 is found in the glandular epithelium and epithelial tumours of lung, ovary, endometrium and breast, but is not found in gastrointestinal epithelium (Osborn et al., 1986Ramaekers et al., 1987, 1990Vojtesek et al., 1990;van Niekerk et al., 1991;van de Molengraft et al., 1993;Baars et al., 1994). Conversely, CK20 is expressed principally in the normal glands and epithelial tumours of the gastrointestinal tract, urothelium, and Merkel cells (Moll et al., 1990, 1992, 1993, 1995Miettinen, 1995;Chan et al., 1997). In the last several years, antibodies to both CK7 and CK20 have been developed which are applicable to paraffin-embedded tissue. Because of the differential expression patterns of CK7 and CK20 among the different types of human epithelium, these antibodies can be used to subclassify epithelial tumours from different organs in defined clinical settings. The most useful settings are those in which the differential diagnosis includes colon versus lung, colon versus ovary, and urothelial (transitional cell) carcinoma versus other tumours. Lung cancer is a prevalent diagnosis in a work-up of metastasis of unknown origin. CK7 and CK20 staining may help in finalizing the diagnosis of lung cancer.
Described here are clinical settings in which CK7 and CK20 were used to help clarify diagnostic problems. We distinguish tumours that are CK7+/CK20–, CK7–/CK20+, CK7+/CK20+ and CK7–/CK20–; CK7+/CK20– tumours are most consistent with lung cancer.
Materials and methods
The files of the Department of Pathology at the Brigham and Women's Hospital (Boston, Massachusetts, USA) were searched between 1 January 1997 and 1 July 1998. Diagnoses were confirmed by retrieval of pathology reports, review of haematoxylin and eosin (H and E) stain and immunohistochemical sections, and review of the clinical records.
Sections (4 μm) from formalin-fixed, paraffin-embedded tissues were mounted on glass slides. Those used for staining with CK7 and CK20 required a 10-minute protease digestion prior to immunohistochemical staining. Immunohistochemical staining was performed using the avidin–biotin–peroxidase complex technique (Hsu et al., 1981). The antibodies used were CK7 (Dako, Carpinteria, California, USA, mouse monoclonal, 1:100 dilution) and CK20 (Dako, mouse monoclonal, 1:100 dilution), prostate-specific antigen (PSA) (Dako, rabbit polyclonal, 1:300 dilution), prostate alkaline phosphatase (PrAP) (Gift of Edie Heyderman, Guys and St Thomas Medical and Dental School, London, UK, mouse monoclonal, 1:10 dilution), carcinomoembryonic antigen (CEA) (Dako, rabbit polyclonal, 1:200) and chromogranin (Boehringer Mannheim, Indianapolis, Indiana, USA, mouse monoclonal, 1:500 dilution). The sections were incubated with the primary antibodies at room temperature for 60 minutes, followed by biotinylated rabbit anti-mouse immunoglobulin for 30 minutes and then avidin–biotin–peroxidase complex for 30 minutes. The colour reaction was developed with 0.1% diaminobenzidine. Appropriate positive and negative controls were used.
One hundred and forty-four cases of metastatic carcinoma of unknown primary to bone, lung or liver were identified. There were 75 cases in which clinical evaluations did not help in the determination of the site of origin, and thus CK7 and CK20 were used. Forty-five of 75 cases were CK7+/CK20–; 15 cases were CK7–/CK20–; 9 cases were CK7–/CK20+; and 6 cases were CK7+/CK20+.
A representative case taken from the group of 45 cases that were CK7+/CK20– is shown in Fig. 1
. These cases represented lung cancer as the primary tumour. The patient in Fig. 1 was a 66-year-old male smoker with an elevated alpha-fetoprotein (AFP) level, a liver mass and a left scapular lesion. Chest CT scan showed no significant abnormalities. Other clinical work-up was completely negative. The patient underwent biopsy of the left scapular lesion to obtain tissue for diagnosis. Histologically, the tumour was composed of nests of large pleomorphic cells embedded in a desmoplastic stroma (Fig. 1). There was single-cell necrosis and the tumour cells had hyperchromatic nuclei with inconspicuous nucleoli and a high nuclear/cytoplasmic ratio. Immunohistochemistry revealed the tumour cells to be strongly positive for CK7 and negative for CK20 (Fig. 1), chromogranin, CEA, PSA, PrAP and AFP. Mucicarmine stain was negative for intracellular mucin as well. The final pathologic diagnosis was metastatic carcinoma, most likely primary of the lung.
This case illustrated several points. An elevated AFP level is not specific for hepatocellular carcinoma but may be expressed by carcinomas originating in different organs (Flint and Lloyd, 1992). While positive staining for AFP and polyclonal CEA in a so-called ‘canalicular’ pattern are helpful in establishing the diagnosis of hepatocellular carcinoma, a large number of hepatocellular carcinomas do not express either antigen (Flint and Lloyd, 1992;Askin, 1993). Additionally, hepatocellular carcinomas are negative in 77–93% of cases for both CK7 and CK20, while they are usually positive for CK8 and CK18 (Flint and Lloyd, 1992). In this case, positivity for CK7 and the negative result for AFP and CEA argued strongly against a hepatocellular carcinoma. In a man, it is important to rule out prostate carcinoma as a source of an occult primary cancer. Prostate carcinoma was excluded in this case by the negative prostate markers and the positivity for CK7 as the majority of adenocarcinomas of the prostate are CK7+/CK20+ (Erlandson, 1994). Virtually all colonic carcinomas express CK20 and are negative for CK7. The combination of CK7 positivity and CK20 negativity rules out a poorly differentiated colonic primary. Since most primary lung carcinomas are CK7 positive and CK20 negative, the tumour in this case was thought to be a metastasis from an occult lung primary.
Case 2 in Fig. 1 is a representative example of the nine cases that were CK7–/CK20+. This 71-year-old male non-smoker had been diagnosed with a Duke's B2 adenocarcinoma of the rectosigmoid. He underwent a surgical resection, and 4 years later had a routine X-ray and CT scan of the chest showing a solitary pulmonary nodule in the interior segment of the right upper lobe, measuring 1.1 × 1.2 cm. The patient underwent a thorascopic wedge resection of the pulmonary nodule. Gross examination of the specimen revealed a well-demarcated 0.5-cm nodule, while histological examination showed a moderately differentiated adenocarcinoma composed of glands with a garland and cribriform growth pattern and conspicuous necrosis (Fig. 1). The tumour cells were tall, with basically located, hyperchromatic nuclei and abundant eosinophilic cytoplasm. The surrounding lung parenchyma was unremarkable. Immunohistochemistry revealed the tumour cells to be diffusely and strongly positive for CK20 and negative for CK7 (Fig. 1). The surrounding normal lung parenchyma was strongly positive for CK7 and negative for CK20.
While the histological picture of adenocarcinoma of the colon is characterized by glands with a cribriform growth pattern lined by tall neoplastic cells with basally located nuclei and a frequent association with necrosis, this appearance is not specific and primary pulmonary adenocarcinoma may have the same appearance (Burt and Goudie, 1979;Shah et al., 1985). Due to prognostic and therapeutic implications, it was important in this case to determine whether the ‘coin lesion’ was primary to the lung or represented an isolated metastasis from the patient's known colonic cancer (Wilson et al., 1986;Sheahan et al., 1990;Pavelic et al., 1991). The combination of CK20 positivity and CK7 negativity is diagnostic in this setting of a metastasis from the patient's original colonic primary. It should also be noted that the uninvolved pulmonary parenchyma had the opposite cytokeratin profile (CK7 positive, CK20 negative).
Case 3 in Fig. 1 is one of the six cases that were CK7+/CK20+. This was a 65-year-old female smoker who had developed haematuria and dysuria, prompting a cytoscopic biopsy. Cytoscopy revealed a large bladder mass measuring 10 cm in diameter, as well as a mass along the anterior portion of the vagina consistent with a transitional cell carcinoma of the bladder. A chest X-ray and a CT scan revealed a pulmonary nodule 2–3 cm in diameter in the right lower lung. A final needle aspirate of one of the pulmonary lesions was performed for diagnosis.
Cytological preparations showed numerous neoplastic epithelioid cells while a cell block consisted of sheets of neoplastic epithelioid cells with eosinophilic cytoplasm, vesicular nuclei and single, prominent, centrally located nucleoli (Fig. 1). Immunohistochemistry showed the cells to be focally positive for CK7 and CK20 (Fig. 1). Primary pulmonary carcinomas generally have strong staining for CK7 and may have weak staining for CK20. While not all transitional cell carcinomas stain positively for CK20, positive focal staining for both CK7 and CK20 favours a transitional cell carcinoma (Erlandson, 1994). Given the histological appearance of the tumour and the clinical history, a final diagnosis of metastatic transitional cell carcinoma was made.
There is considerable morphologic overlap between carcinomas that arise in different organs. In many instances the clinical, radiographic and histological features are sufficient to determine the origin of tumours, but in other situations they are not. For instance, although colonic adenocarcinomas tend to have a characteristic histological appearance with the areas of ‘dirty’ necrosis (necrosis with characteristic collections of cellular debris), pulmonary adenocarcinomas with similar features have been described (Flint and Lloyd, 1992). Metastatic colonic adenocarcinomas in the lung can simulate bronchoalveolar carcinomas (Askin, 1993), making it challenging in some instances to differentiate these neoplasms by conventional histological methods. This problem is not only seen in the lung but can be a problem in other organs where primary neoplasm may be simulated by metastases. Determining the origin of metastatic carcinomas to bone can be difficult if not impossible for the same reasons.
Ultrastructural studies are only occasionally helpful in these situations (Ghadially, 1985;Erlandson, 1994). The presence of myelinosomes or surfactant-like bodies is good evidence of lung origin but these structures are not seen in most pulmonary adenocarcinomas. Microvilli with glycocalyceal bodies and filamentous core rootlets suggest colonic origin in the differential diagnosis of adenocarcinoma of unknown origin but these findings are not specific for colonic adenocarcinoma.
Many attempts have been made to separate carcinomas of unknown origin through the use of immunohistochemical techniques. While there have been some conspicuous successes such as with the use of PSA or PrAP, which are very specific and sensitive for prostatic adenocarcinomas (Shah et al., 1985), and TGB and calcitonin, which are very specific for the carcinomas of the thyroid (Burt and Goudie, 1979;Wilson et al., 1986), most antibodies have not proven to be specific. For example, CEA was originally thought to be specific for colonic adenocarcinomas; however, this antibody reacts with a wide variety of carcinomas from different sites (Sheahan et al., 1990;Pavelic et al., 1991;Daya et al., 1992). Likewise, positive HAM 56 immunoreactivity was thought to exclude the possibility of colonic primary, but it has been shown that up to 39% of these tumours show positive staining with this antibody (Sheahan et al., 1990;Pavelic et al., 1991;Daya et al., 1992;Loy and Abshier, 1993;Loy and Calaluce, 1994). CO19-9-9 is reactive with most gastrointestinal carcinomas; however, it is not specific and is seen in many endometrial carcinomas as well (Ellis and Hitchcock, 1988;Gatalica and Miettinen, 1994).
Broad-specificity antisera against cytokeratins may have been used to separate poorly differentiated epithelial tumours from non-epithelial tumours with similar histology, however, these antibodies are not useful in the subclassification of epithelial tumours according to the site of the origin. The development of antibodies specific to individual cytokeratin molecules has begun to help distinguish between epithelial tumours that originate in different organs. Low molecular weight cytokeratins are more characteristic of pulmonary tissue, whereas higher molecular weight cytokeratins are characteristics of colonic tissue. CK7 and CK20 have proven to be the most useful of these antibodies (Table 1
To date, CK7 immunostaining has been seen in 95–100% of pulmonary adenocarcinomas and 0–21% of colonic adenocarcinomas, but when found in the colon it is usually focal. CK20 reactivity is present in 8–15% (of the total) of pulmonary adenocarcinomas and 97% of colonic adenocarcinomas. While the usage of either CK7 or CK20 alone could lead to erroneous conclusions in the different diagnosis of carcinoma of lung versus colon, immunostaining with both markers allows the origin of the tumour to be pinpointed in 86% of primary pulmonary adenocarcinomas (CK7+/CK20–) and 77% of primary colonic carcinomas (CK7–/CK20+) (Loy and Calaluce, 1994). The CK7+/CK20+ and CK7–/CK20– phenotypes are non-specific in this differential diagnosis.
Recent work supports the use of CK7 and CK20 in other clinical scenarios as well. CK20 positively is strongly supportive of a diagnosis of Merkel cell carcinoma in the differential diagnosis of small carcinoma of uncertain origin (Miettinen, 1995;Chan et al., 1997). Additionally, the CK7+/CK20– phenotype is supportive of pancreaticobiliary carcinoma over ampulla of vater/duodenal carcinoma, which is usually CK7–/CK20+ (Alexander et al., 1997).
The use of monoclonal antibodies to CK7 and CK20 can be quite helpful in determining the primary site of carcinoma of unknown origin in defined clinical situations. The most useful settings are in the differential diagnosis of pulmonary versus colonic adenocarcinoma and ovarian (non-mucinous) versus colonic adenocarcinoma. Acceptance and usage of these antibodies will undoubtedly clarify the limitations of these techniques while researchers will surely discover new uses for these important reagents.
This work was in part supported by NIH grant CO753-948-04 and Lowe's Center for Thoracic Oncology (RS).
1. Alexander J , Krishnamurthy S , Kovacs D ( 1997 ). Cytokeratin profile of extrahepatic pancreaticobiliary epithelia and their carcinomas . Appl Immunohistochem 5 : 216 – 222 .
2. Askin F ( 1993 ). Something old? Something new? Second primary or pulmonary metastasis in the patient with known extrathoracic carcinoma
. Am J Clin Pathol 100 : 4 – 5 .
3. Baars J , De Ruijter J , Smedts F et al. ( 1994 ). The applicability of a keratin 7 monoclonal antibody in routinely Papanicolaou-stained cytologic specimens for the differential diagnosis of carcinomas . Am J Clin Pathol 101 : 257 – 261 .
4. Burt A , Goudie R ( 1979 ). Diagnosis of primary thyroid carcinoma
by immunohistological demonstration of thyroglobulin . Histopathology 3 : 279 – 286 .
5. Chan J , Suster S , Wenig B et al. ( 1997 ). Cytokeratin 20 immunoreactivity distinguishes Merkel cell (primary cutaneous neuroendocrine) carcinomas and salivary gland small cell carcinomas from small cell carcinomas of various sites . Am J Surg Pathol 21 : 226 – 234 .
6. Cooper D , Schermer A , Sun T ( 1985 ). Classification of human epithelia and their neoplasms using monoclonal antibodies to keratins: strategies, applications, and limitations . Lab Invest 52 : 243 – 256 .
7. Daya D , Nazerali L , Frank G ( 1992 ). Metastatic ovarian carcinoma
of large intestinal origin stimulating primary ovarian carcinoma
. A clinicopathologic study of 25 cases . Am J Clin Pathol 97 : 751 – 758 .
8. Ellis I , Hitchcock A ( 1988 ). Tumour marker immunoreactivity in the adenocarcinoma . J Clin Pathol 41 : 1064 – 1067 .
9. Erlandson R ( 1994 ). Diagnostic Transmission Electron Microscopy of Tumors Raven Press , New York .
10. Flint A , Lloyd R ( 1992 ). Pulmonary metastases of colonic carcinoma
. Distinction from pulmonary adenocarcinoma . Arch Pathol Lab Med 116 : 39 – 42 .
11. Franke W , Schmid E , Schiller D et al. ( 1982 ). Differentiation-related patterns of expression of proteins of intermediate-size filaments in tissues and cultured cells . Cold Spring Harbor Symp Quant Biol 46 ( Part 1 ): 431 – 453 .
12. Gatalica Z , Miettinen M ( 1994 ). Value of carcinoma
antigens CA 19–9 and CA 15–3 in the immunohistochemical evaluation of metastatic carcinoma
of unknown primary
. Appl Immunohistochem 2 : 205 – 208 .
13. Ghadially F ( 1985 ). Diagnostic Electron Microscopy of Tumours Butterworth , London .
14. Greenberg B , Lawrence H ( 1988 ). Metastatic cancer with unknown primary
. Med Clin North Am 72 : 1055 – 1065 .
15. Hsu S , Raine L , Fanger H ( 1981 ). The use of antiavidin antibody and avidin-biotin-peroxidase complex in immunoperoxidase techniques . Am J Clin Pathol 75 : 816 – 821 .
16. Loy TS , Abshier J ( 1993 ). Immunostaining with HAM56 in the diagnosis of adenocarcinomas . Mod Pathol 6 : 473 – 475 .
17. Loy T , Calaluce R ( 1994 ). Utility of cytokeratin immunostaining in separating pulmonary adenocarcinomas from colonic adenocarcinomas . Am J Clin Pathol 102 : 764 – 767 .
18. Miettinen M ( 1995 ). Immunohistochemical marker for gastrointestinal, urothelial, and Merkel cell carcinomas . Mod Pathol 8 : 384 – 388 .
19. Moll R , Franke W , Schiller D et al. ( 1982 ). The catalog of human cytokeratins
: patterns of expression in normal epithelia, tumors, and cultured cells . Cell 31 : 11 – 24 .
20. Moll R , Schiller D , Franke W ( 1990 ). Identification of protein IT of the intestinal cytoskeleton as a novel type I cytokeratin with unusual properties and expression patterns . J Cell Biol 111 : 567 – 580 .
21. Moll R , Lowe A , Laufer J et al. ( 1992 ). A new histodiagnostic marker detected by monoclonal antibodies . Am J Pathol 140 : 427 – 447 .
22. Moll R , Zimbelmann R , Goldschmidt M et al. ( 1993 ). The human gene encoding cytokeratin 20 and its expression during fetal development and in gastrointestinal carcinomas . Differentiation 53 : 75 – 93 .
23. Moll I , Kuhn C , Moll R ( 1995 ). Cytokeratin 20 is a general marker for cutaneous Merkel cells while certain neuronal proteins are absent . J Invest Dermatol 104 : 910 – 915 .
24. Nystrom J , Weiner J , Heffelfinger-Juttner J et al. ( 1977 ). Metastatic and histologic presentations in unknown primary
cancer . Semin Oncol 4 : 53 – 58 .
25. Osborn M , van Lessen G , Weber K et al. ( 1986 ). Differential diagnosis of gastrointestinal carcinomas by using monoclonal antibodies specific for individual keratin polypeptides . Lab Invest 55 : 497 – 504 .
26. Pavelic Z , Pavelic L , Pavelic K et al. ( 1991 ). Utility of anti-carcinoembryonic antigen monoclonal antibodies for differentiating ovarian adenocarcinomas from gastrointestinal metastasis to the ovary . Gynecol Oncol 40 : 112 – 117 .
27. Ramaekers F , Huysmans A , Schaart G et al. ( 1987 ). Tissue distribution of keratin 7 as monitored by a monoclonal antibody . Exp Cell Res 170 : 235 – 249 .
28. Ramaekers F , van Niekerk C , Poels L et al. ( 1990 ). Use of monoclonal antibodies to keratin 7 in the differential diagnosis of adenocarcinomas . Am J Pathol 136 : 641 – 655 .
29. Sheahan K , O'Brien M , Burke B et al. ( 1990 ). Differential reactivities of carcinomoembryonic antigen (CEA) and CEA-related monoclonal and polyclonal antibodies in common epithelial malignancies . Am J Clin Pathol 97 : 157 – 164 .
30. Shah N , Tuttle S , Strobel S et al. ( 1985 ). Prostatic carcinoma
metastatic to bone: sensitivity and specificity of prostate-specific antigen and prostatic acid phosphatase in decalcified material . J Surg Oncol 29 : 265 – 268 .
31. Ueda G , Sawada M , Ogawa H et al. ( 1993 ). Immunohistochemical study of cytokeratin 7 for the differential diagnosis of adenocarcinomas in the ovary . Gynecol Oncol 51 : 219 – 223 .
32. van de Molengraft F , van Niekerk C , Jap P et al. ( 1993 ). OV-TL 12/30 (keratin 7 antibody) is a marker of glandular differentiation in lung cancer
. Histopathology 22 : 35 – 38 .
33. van Niekerk C , Jap P , Ramaekers F et al. ( 1991 ). Immunohistochemical demonstration of keratin 7 in routinely fixed paraffin-embedded human tissues . J Pathol 165 : 145 – 152 .
34. Vojtesek B , Staskova Z , Nenutil R et al. ( 1990 ). A panel of monoclonal antibodies to keratin no. 7: characterization and value in tumor diagnosis . Neoplasm 37 : 333 – 342 .
35. Wang NP , Zee S , Zarbo RJ ( 1995 ). Coordinate expression of cytokeratins
7 and 20 defines unique subsets of carcinomas . Appl Immunohistochem 3 : 99 – 107 .
36. Wilson N , Pambakian H , Richardson T et al. ( 1986 ). Epithelial markers in thyroid carcinoma
: an immunoperoxidase study . Histopathology 10 : 815 – 829 .