The impact of diagnostic immunohistochemistry (IHC) for the surgical pathologist is legendary, and it is best appreciated when studying malignancies of unknown primary site. A cost-effective tool, IHC is performed in most hospital laboratories, is often automated, and provides for a rapid turnaround time, all desirable qualities for the pathologist. The number of antibodies that are available for diagnostic use rises exponentially each year, an attestation to the importance of ongoing research in this field. Since the first edition of this book, there has been a substantial addition of important antibodies that are especially useful in the work-up for patients with metastatic malignancy of unknown primary. Even with the larger armamentarium of antibodies, there remains a paucity of specific antibodies that allow for “100% unequivocal, definitive diagnosis” in every case. Indeed, it has been said that “it may be dangerous to base any distinction in tumor pathology primarily on the basis of the pattern of immunoreactivity of a given marker, no matter how specific it is purported to be.” This statement echoes the importance of histopathologic morphology that is the basis of diagnosis in surgical pathology. The standard tissue section is the starting point for raising questions that need to be answered for the patient when morphology alone is not enough, and IHC is perhaps the best method to obtain more information from the paraffin section.
Even the most specific antibodies (e.g., thyroid transcription factor-1 [TTF-1], PAX8, SOX10, WT1, or prostate-specific antigen [PSA]) are not entirely site specific, and we therefore resort to panels of antibodies that give statistical power to our morphologic diagnoses. Relevant diagnostic panels of antibodies change rapidly based on information from immunohistochemical studies, and we can expect this constant infusion of new data on antibody sensitivity and specificity to impart an uncomfortable state of chronic flux on the discipline of IHC. Nevertheless, change is often incremental in IHC, and the basics of separating the category of metastatic malignancy of unknown primary into the categories of carcinoma, melanoma, lymphoma, germ cell neoplasia, and sarcoma have stood the test of time.
Although the term cancer of unknown primary site (CUPS) is sometimes used interchangeably with carcinoma (cancers of epithelial differentiation) of unknown primary site, not all cancers of unknown primary site are epithelial in origin. This chapter reviews the triage and evaluation of all types of cancers, but the focus remains on the carcinomas, which form the predominant category (~90% to 95% cases) of CUPS.
Cancer of Unknown Primary Site: Clinical Aspects and Economic Considerations
Patients who present with CUPS by definition have no obvious identifiable primary site despite a careful clinical history, physical examination, radiologic imaging, and biochemical or histologic investigations. Studies of patients with malignancies have shown that CUPS accounts for 5% to 15% of all patients who present with a malignancy. The impact of recent improvements in radiologic imaging has reduced this percentage to 3% to 7% of patients who present with a CUPS diagnosis.
The clinical presentation in a CUPS case depends on a number of factors including age, gender, sites of involvement, and line of differentiation (viz., epithelial, mesenchymal, lymphoid, germ cell, melanocytic). It appears that the tumors that present as CUPS are biologically and clinically different from the known primary tumors that metastasize several years after diagnosis. CUPS patients fail to show any symptoms related to the primary tumor and demonstrate an unpredictable pattern of spread (i.e., a difference in frequency of involvement of a particular site than would be expected of a known primary tumor). The majority of CUPS cases present with multiple sites of involvement with a few presenting with only one or two sites of involvement. Based on the sites of involvement, several clinicopathologic entities have been characterized that are helpful in identifying the primary site.
The liver is one of the single largest repositories for metastatic malignancies of all types, especially for carcinomas. The most common malignancies metastatic to the liver are from the gastrointestinal (GI) tract, with colorectal carcinomas leading this group. Lung and breast carcinomas also commonly metastasize to the liver, as do pancreaticobiliary carcinomas. This entire group of adenocarcinomas may appear similar to primary cholangiocarcinoma of the liver and may simulate some hepatocellular carcinomas, particularly the less-differentiated hepatocellular carcinomas. Prostate carcinoma, although unusual, does metastasize to the liver and can be confused with cholangiocarcinoma. Thus, for hepatic metastases of unknown primary in women, colorectal, breast, and lung carcinomas are of primary consideration, whereas in men, colorectal, lung, and prostate carcinomas top the list. Malignant melanoma metastatic to the liver is not uncommon, with the highest frequency of liver metastases seen with primary eye melanomas. Pisharodi and associates, in a fine needle biopsy study of 200 malignant aspirates of the liver, found that 32% were hepatocellular carcinomas, 49.5% were readily diagnosed as metastatic carcinomas, and 18.5% were problematic. Of this latter group, IHC contributed to definitive diagnosis in half of the cases.
Along with the liver, the lung is a major repository for metastatic carcinomas, especially adenocarcinomas. Identification of the origin of an adenocarcinoma in the lung is a frequent, difficult, and challenging process for the surgical pathologist because adenocarcinomas not only are the most frequent primary lung tumor, but they are also the most common metastatic tumor found in the lung. Distinction among these tumor types can be especially challenging on scant biopsy materials, such as transbronchial biopsy or fine needle aspiration biopsy (FNAB). Clinical information regarding circumscription and the number of lesions is quite useful. Metastatic tumors often present as multiple circumscribed nodules, whereas primary lung tumors often show a dominant infiltrative lesion. It is important to identify those carcinomas that can be treated by chemotherapy or hormonal manipulation, or both, especially metastatic breast or prostate carcinomas.
In the brain, distinguishing metastatic adenocarcinoma or poorly differentiated carcinoma from a glial tumor is straightforward, although determining the source of the metastasis may be problematic, especially when the occult primary is unknown. Lung carcinomas are the most likely primary to be discovered subsequent to central nervous system (CNS) presentation, and other common primaries include breast, kidney, thyroid, and GI tract. Patients with other adenocarcinomas, such as ovarian, prostatic, and pancreaticobiliary, rarely present with brain metastases, because there is almost always evidence of widespread dissemination of these tumors before the occurrence of cerebral metastases. The site of origin of carcinoma remains unknown in up to 5% of patients. The survival of most patients with carcinomatous brain metastases is in the range of 3 to 11 months.
Patients presenting with skeletal metastases often have primary carcinomas in the lung, breast, kidney, or urogenital region, and imaging studies have been particularly useful in elucidating the primary tumor.
For patients who present with pleural effusions, the breast is the most common primary site for women, with the lung being the most common site for a primary tumor in men. Malignant lymphomas are seen in both sexes.
In women who present with malignant abdominal effusions (malignant ascites), common abdominal sites include müllerian primary (often tube and ovaries), whereas men with malignant ascites typically have primary tumor sites in the GI tract, predominantly in the colon, rectum, pancreas, or stomach. Patients with peritoneal carcinomatosis of nongynecologic origin most often have origins in the stomach, colon, or pancreas and have a median survival of 3 months.
For patients who present with primary lymph node metastases, there may be clues to the primary site of the tumor based on tumor morphology (adenocarcinoma, squamous cell carcinoma, or undifferentiated carcinoma) and the anatomic site of lymph node involvement.
In women presenting with adenocarcinoma in the axilla, the primary tumor is most often found in the ipsilateral breast. For the patient who presents with metastatic adenocarcinoma in the neck, the metastatic work-up will begin in the lung (males) or breast (females), although GI and prostate adenocarcinomas both show a predilection for the left side of the neck.
Undifferentiated carcinomas of the head and neck are the most common primary source for metastatic tumors in head and neck lymph nodes, and the majority of these are of squamous mucosal derivation. The prognosis for this group of patients rests largely on the nodal status, with patients having stage N3 lesions carrying a poor prognosis. For a squamous cell carcinoma involving the upper and mid-cervical lymph nodes, a thorough examination of the oropharynx, hypopharynx, nasopharynx, larynx, and upper esophagus by direct vision and fiberoptic nasopharyngolaryngoscopy with biopsy of any suspicious areas is more valuable than further pathologic examination. Advanced diagnostic techniques such as computerized tomography (CT) and positron emission tomography (PET) scans are helpful in determining the primary site. Occasionally, systematic random biopsies of mucosal sites such as nasopharynx, base of tongue, pyriform sinus, and tonsils may reveal an occult tumor. Metastatic squamous cell carcinoma involving lower cervical lymph nodes or at any other site except inguinal nodes is highly suspicious for a lung primary. In addition, occasionally, esophageal squamous cell carcinomas may preferentially involve the lower cervical lymph nodes. The vast majority of patients with squamous cell carcinoma involving the inguinal nodes generally have detectable primary tumor in the anogenital area. Therefore all women must undergo a thorough evaluation of the vulva, vagina, and the cervix, and all men should undergo careful inspection of the penis. Both sexes must also have examination of the anorectal region.
The current clinical approach is an attempt to identify favorable prognostic groups in patients with unknown primary tumors so that they can be managed appropriately. This group of tumors includes leukemia/lymphoma, germ cell tumors, small cell carcinoma of the lung, and carcinomas of the breast, ovary, endometrium, adrenal gland, thyroid, and prostate. When possible, it is useful to separate regional from distant metastases, because localized disease is more amenable to treatment. Other favorable clinical features that have been described include location of tumor in the retroperitoneum or peripheral lymph nodes, tumor limited to one or two metastatic sites, a negative smoking history, and young age. Kirsten and colleagues studied 286 patients with CUPS and concluded that the factors that predicted survival were lymph node presentation, good performance status, and body weight loss of less than 10%. Using a panel of antibodies to determine differentiation of tumors in 41 patients with CUPS, Van der Gaast and associates concluded that the immunohistochemical panel approach to uncover tumor origin is useful for selecting appropriate treatment of patients, especially those who may benefit from combination chemotherapy. Other immunohistochemical studies of CUPS have elucidated the origin of tumors in as few as 5% to as many as 70% of patients. However, most investigators have arrived at the same conclusion: for individual patient therapy, knowledge of site of origin improves patient survival. Furthermore, with the advent of targeted therapy (although, it is still in its infancy), the benefit of identifying the site of origin may be manifold. With the incorporation of next-generation sequencing in diagnostic testing, it may become less important to identify the site of origin and more useful to find actionable targets for an individual patient. However, finding the site of origin may still have genetic implications and patient and family need to know the tumor source.
The appropriate work-up for identifying a primary tumor depends on the patient’s clinical symptoms, age, history, gender, and the likelihood of finding the primary tumor. Patients with CUPS do poorly as a group, with a median survival of 6 to 11 months, and the importance of establishing the origin of the primary site guides therapeutic interventions of hormonal manipulation, chemotherapy, and radiation. The clinician must also take into account the economics of an extensive clinical work-up, as well as the inconvenience and discomfort to which the patient is subjected.
The economic considerations of clinical work-up in these patients have not been extensively studied. There are few data available on the cost-effectiveness of IHC in surgical pathology. Schapira and Jerrett analyzed the clinical work-ups in a group of 199 patients and concluded that the search for a primary neoplasm incurred an average cost of US$17,973, with only 19.6% of patients surviving for more than 1 year. As a matter of fact, IHC is probably undervalued and is likely a cost-effective maneuver in the study of CUPS. Radiologic studies themselves have limited value in the management of these patients, and prognosis is not affected. Even autopsies on some of these patients may not detect the primary tumor site because of small size, extensive dissemination, or regression due to therapy. In 1988, Le-Chevalier and coworkers studied 302 autopsy specimens from patients who presented with CUPS. The primary tumor site was located premortem in 27% of patients, at autopsy in 57% of patients, and remained unidentified in 16% of patients. The most common primary tumor sites in this study included pancreas, lung, kidney, and colon/rectum, a list that includes the two malignancies with the highest incidence in both men and women.
Diagnostic Approach to the Study of Cancer of Unknown Primary Site: Specimen Preparation
The goal of the surgical pathologist is to identify the line of differentiation of the tumor and identify those tumors that are within the “treatable” group of tumors, namely, carcinomas of breast, prostate, ovary, endometrium, and thyroid, as well as germ cell tumors and neuroendocrine carcinomas. Hormonal and antihormonal therapies are useful for patients with breast and prostate carcinomas. Neuroendocrine, thyroid, and germ cell tumors may be responsive to suppression by chemical agents. The therapeutic response of other carcinomas is less certain, but the identity of the carcinoma, if available, is useful to determine more useful therapeutic regimens for these patients prospectively. Some studies on patients with CUPS demonstrate that up to one-third respond to taxane-based therapies.
Tissue procurement is the first step in the work-up for tumors of unknown primary origin, and it is a common practice to obtain tissue by FNAB or core tissue biopsy. The sensitivity of FNAB for metastatic carcinoma in a series of 266 superficial lymph nodes was 96.5%, with no false-positive results and nine false-negative results. Tissue from both FNAB and core biopsies can be triaged for ancillary studies in the same manner. There is also great value in the immunocytochemical study of malignant effusions, and often these are the first samples available by virtue of therapeutic evacuation. Whatever the method of obtaining tissue, it is ideal to be able to monitor the process so that adequate tissue may be obtained to triage the patient’s problem appropriately, namely, triage of the specimen for immunohistology, electron microscopy, flow cytometry, and molecular-cytogenetic studies. If there is not enough tissue available for multiple studies, our recommendation would be to freeze some tumor sample after tissue has been collected for morphologic analysis, because fresh frozen tissue is the best sample for molecular analysis. If there is more than one core available for histology, it is best to place each core in a separate tissue block to maximize the tissue available for IHC and/or other ancillary studies.
Monitoring the tissue procurement process can be performed with frozen sections, immediate interpretation of FNAB, or tissue imprints. In addition to tissue procurement, the pathologist must define the problem by taking the patient’s age, gender, known risk factors, duration of symptoms, and clinical and radiologic findings. Based on this information and the morphologic appearance of the tumor, the quest for the study of tumor origin begins.
In surgical pathology and cytopathology, poorly differentiated carcinomas can be broadly classified as large cell undifferentiated, small cell undifferentiated, and spindle cell. The starting point for diagnostic interpretation is the standard hematoxylin and eosin (H&E) or Papanicolaou-stained slide. The importance of the histologic morphology should not be underestimated in arriving at a definitive diagnosis. Morphology is the foundation upon which the interpretation of all immunohistochemical studies rests.
In this chapter, the role of diagnostic IHC in diagnosing CUPS is emphasized, especially as it relates to adenocarcinoma/poorly differentiated carcinoma of unknown primary site and germ cell tumors, as they account for over 90% cases of CUPS. Specific tables are presented that aid in the differential diagnosis of tumors in specific anatomic sites. The role of molecular studies in combination with IHC for patients with CUPS is also discussed.
Determining Site of Origin: Stepwise Approach
Carcinomas form the predominant category (~90% to 95% cases) of CUPS and will therefore remain the main focus of this chapter. Because virtually all carcinomas show significant positivity for cytokeratins (CKs), carcinomatous differentiation becomes readily apparent when the tumor is diffusely CK positive. The simple and broad-spectrum CKs are the initial antibodies of choice for detecting carcinomatous differentiation. More specific subcategorization of the tumor origin is then possible using a variety of site-specific CKs as well as antibodies directed against various cellular products. It is a combination of these cellular antigens that may yield a cost-effective approach to tumor categorization.
The approach to definitive diagnosis of the patient with CUPS effectively follows four to five sequential steps:
Determine the cell line of differentiation using major lineage markers, including keratins, lymphoid, melanoma, germ cell, and sarcoma markers.
Determine the CK type or types of distribution in the tumor cells because some subsets of CKs are unique to certain tumor types.
Determine whether there is coexpression of vimentin.
Determine whether there is expression of supplemental antigens of epithelial or germ cell derivation, that is, carcinoembryonic antigen (CEA), epithelial membrane antigen (EMA), or placental alkaline phosphatase (PLAP). This step can be combined either with step 3 or step 5.
Determine whether there is expression of cell-specific products, cell-specific structures, transcription factors or receptors that are unique identifiers of cell types—for example, neuroendocrine granules, peptide hormones, thyroglobulin, PSA, prostate-specific membrane antigen (PSMA), NKX3.1, inhibin, gross cystic disease fluid protein-15 (GCDFP-15), GATA-binding protein 3 (GATA3), villin, uroplakin, TTF-1, transcription factor CDX2 or PAX8.
Step One: Screening Immunohistochemistry
An abbreviated first line panel to determine the line of differentiation should be composed of epithelial markers (pankeratin AE1/3 and CAM5.2, both used together), mesenchymal marker (vimentin), lymphoid marker (leukocyte common antigen or LCA), and melanocytic marker (S100). Although vimentin is included in the above panel, it is generally the least helpful; it should be interpreted with caution. Vimentin is considered a mesenchymal marker, but it may be expressed quite diffusely in many poorly differentiated carcinomas. As a matter of fact, vimentin coexpression in a carcinoma may point toward a specific primary site. Except for vimentin, a diffuse strong expression of any of the above markers is generally suggestive of a particular line of differentiation.
Line of Differentiation: Lymphoid
The above first line panel generally leads to a more extensive work-up. At this point, if the tumor is strongly positive for LCA and negative for keratins, further work-up is directed toward classifying the lymphoma using pan-B, pan T-cell (CD20, CD79a, and CD3), and other markers. If LCA is +/−, and pan-B, pan-T cell markers are negative, and the morphology is still suggestive of lymphoid neoplasm, it is not unreasonable to think about a myeloid neoplasm and perform myeloid markers for granulocytic sarcoma. This is one diagnosis that may be missed even after an expensive and extended immunohistologic evaluation. The stains that are helpful in demonstrating myeloid lineage are myeloperoxidase, chloroacetate esterase, lysozyme stains, and CD117 (also known as c-Kit or stem cell marker). CD43 and CD68 stains are also positive in granulocytic sarcomas.
Line of Differentiation-Melanocytic
A diffuse strong staining with S100, negative for CK in a tumor of unknown origin is good evidence that it may be a melanoma. However, this still needs to be confirmed by additional melanoma markers such as human melanoma black 45 (HMB-45), melan-A, tyrosinase, SOX10, or NKI/C3. This is because S100 is not a specific (although very sensitive) marker for melanoma. S100 is also expressed by some carcinomas and sarcomas (liposarcoma, chondrosarcoma, and neural tumors). Although the typical variants of these sarcomas are easy to diagnose on H&E alone, the unusual variants such as de-differentiated liposarcoma, mesenchymal chondrosarcoma, or a malignant peripheral nerve sheath tumor (MPNST) may pose a challenge to distinguish from melanomas. Therefore additional melanoma markers should be performed for definitive diagnosis. Similarly, a pathologist should also be careful in distinguishing melanoma from carcinoma based on only a limited number of immunostains. As mentioned earlier that some carcinomas may show strong S100 expression or expression for SOX10, an additional pitfall is that some melanomas may show polyclonal CEA and/or focal CAM5.2 keratin immunoreactivity. Therefore, caution is advised when the diagnosis is heavily based on immunohistochemical expression of markers.
Line of Differentiation: Mesenchymal
Strong vimentin expression in a nonmelanocytic, nonlymphoid neoplasm is generally an indication of it being a sarcoma. Most, but not all, sarcomas are negative for epithelial markers. Epithelioid sarcomas, as the name suggests, show some epithelial differentiation, and synovial sarcoma may have a well-defined biphasic pattern that shows strong staining for epithelial markers in the glandlike component. The sarcomas that need to be considered in a CUPS case are the ones that do not demonstrate a particular line of differentiation on morphology alone. These sarcomas may have small round blue cell tumor morphology (Ewing sarcoma, desmoplastic small round cell tumor and rhabdomyosarcoma), spindle and epithelioid cells (synovial sarcoma, clear cell sarcoma, angiosarcoma), and pure epithelioid cells (epithelioid sarcoma). Although a number of immunohistochemical stains are available to further classify a sarcoma into a defined category, many stains are not specific enough to provide a definitive diagnosis. Occasionally a high index of suspicion is required to make the correct diagnosis ( Fig. 8.1 ). However, IHC may be performed to narrow down the differential diagnosis and streamline the molecular tests that need to be ordered. For example, a small round blue cell tumor positive for CD99 and periodic acid–Schiff (PAS) in a child or young adult should be evaluated for Ewing sarcoma (EWS)-FLI1 fusion transcript ( Fig. 8.2 ). As mentioned earlier, fresh frozen tissue is the best sample for molecular testing; reverse transcription-polymerase chain reaction assays can also be performed on formalin-fixed, paraffin-embedded (FFPE) tissue. Triage for suspected sarcoma cases can be performed as shown in Table 8.1 .
|Sarcoma Type||Age/Site||Morphology||Special Stains/IHC||Ancillary Techniques for Confirmation of Diagnosis|
|Ewing sarcoma/PNET||Usually <30 years. Chest wall, extremities, retroperitoneum, pelvis. Metastases to lungs and bone.||Small round blue cell tumor.||PAS+, CD99+, FLI1+.||RT-PCR for EWS-FLI1, EWS-ERG, EWS-ETV1, EWS-E1AF, EWS-FEV, FUS-ERG. EWS translocation can also be shown by FISH with EWS break-apart probe.|
|Ewing-like BCOR-CCNB3 fusion positive sarcoma||Usually <20, male preponderance, bone and deep soft tissue||Ewing-like with angulated nuclei and spindle cell morphology.||CCNB3+, BCL2+, CD99+, CD117+||RT-PCR for BCOR-CCNB3|
|Ewing-like CIC-DUX4 fusion positive sarcoma||Median age ~33 years, soft tissue and bone||Atypia and proliferation greater than Ewing sarcoma.||CD99+ (focal, weak), WT1+||RT-PCR for CIC-DUX4|
|RMS-alveolar (A), embryonal (E), and pleomorphic (P)||A-RMS: 10–20 years. Extremities and perineum. |
E-RMS: 3–10 years. Prostate, paratesticular, orbit, nasal cavity.
P-RMS: 50+ years. Abdomen, retroperitoneum, chest wall, testes, extremities.
|Small round blue cells with alveolar growth pattern in alveolar RMS; round and spindle cells in embryonal; round, spindle, and pleomorphic cells in pleomorphic RMS.||Muscle specific actin (MSA)+, desmin+, myoglobin+, myogenin (most specific)+, myoD1+.||RT-PCR for PAX3-FKHR and PAX7-FKHR in alveolar RMS only.|
|Desmoplastic small round cell tumor||Young adults, often adolescent boys. Abdomen and pelvis, peritoneal implants.||Round/oval cells in desmoplastic stroma in classic cases, other cases with variable morphology.||Vimentin+, cytokeratin+, EMA+, desmin+, WT1+ (with C-terminal antibody).||RT-PCR for EWS-WT1.|
|Synovial sarcoma||Young adults. Extremities around large joints. Also other locations including kidney, lung and pleura.||Spindle cell or biphasic glandular and spindle cell pattern. Small round cells in poorly differentiated tumor.||EMA+, keratin+ (biphasic tumors), CD99+, BCL2+. Strong nuclear TLE1+.||RT-PCR for SYT-SSX1 and SYT-SSX2.|
|Clear cell sarcoma (melanoma of soft parts)||Young adults. Deep soft tissue with nodal and lung metastases. Occurs in proximity to tendon, fascia, aponeuroses.||Mixed epithelioid and spindle cells in nested growth pattern.||S100+, HMB-45+, melan A+, SOX10+.||RT-PCR for EWS-ATF1 or EWS-CREB1 in GI tract tumors (neither one seen in cutaneous melanoma).|
|Alveolar soft part sarcoma||Young adults—often females. Deep soft tissue. Lung metastases common.||Large polygonal cells, granular cytoplasm, prominent nucleoli, rare mitoses.||PASD+, TFE3+, desmin+ (50%).||Membrane-bound rhomboidal crystals by electron microscopy (EM). RT-PCR for ASPL-TFE3.|
|PEComas||40–50 years—usually females. Various visceral organs and soft tissue.||Epithelioid and spindle cells with perivascular arrangement, clear to granular cytoplasm.||HMB-45+, melan A+, but S100−. Muscle markers positive.||EM: glycogen, premelanosomes, occasional dense bodies.|
|Epithelioid sarcoma||Young adults. Deep soft tissue of extremities. Metastases to lung, lymph node and skin.||Epithelioid tumor cells, granuloma-like growth pattern.||Keratin+, EMA+, vimentin+, CD34+, CK5/6−, p63−, loss of SMARCB1 (INI1).||Nothing specific. EM may be helpful.|
|Vascular tumors||Adults. Soft tissue and various visceral organs.||Angiosarcoma: Epithelioid and spindle cell tumor, vasoformative areas. Epithelioid in hemangioendothelioma.||FVIII+, CD31+, CD34+, FLI1+, thrombomodulin+, patchy keratin+.||EM to identify endothelial cells is rarely required.|
|Leiomyosarcoma||Adults. Abdomen, pelvis and various other locations.||Spindle or epithelioid cells with areas of smooth muscle differentiation.||SMA+, HHF35+, desmin+, caldesmon+, patchy keratin+.||EM: smooth muscle differentiation.|
|Malignant peripheral nerve sheath tumor||Adults. NF1 patients (50%). Deep soft tissue in association with major nerve.||Spindle cells with neural differentiation, abundant mitosis, necrosis+/−. Rarely epithelioid.||S100+ (weak, patchy), CD56+, CD57+, SOX10+ (~30%–50%), PGP9.5+, CD99+. Negative for HMB-45, melan A, and vascular markers.||Negative for SYT-SSX1 and SYT-SSX2. |
EM: neural differentiation.
|Chordoma||Adults, usually males. Sacrococcygeal, thoracolumbar spine.||Physaliferous cells, vacuolated cytoplasm, mucoid stroma.||Brachury+ (nuclear), S100+, keratin+, CK7−/CK20−, EMA+.||EM rarely required.|
|Extraskeletal myxoid chondrosarcoma||Adults. Deep soft tissues of extremities. Metastases may be confused with myoepithelial type carcinomas.||Cords of spindle and epithelioid cells in myxoid stroma.||S100+, NSE+, synaptophysin+/−, keratin-, chromogranin−.||RT-PCR for EWS-CHN and TAF2N-CHN.|
|Angiomatoid fibrous histiocytoma||Extremities of children and young adults.||Nodular distribution of ovoid and spindle cells with blood filled cystic cavities, and surrounding dense lympho-plasmacytic infiltrate.||CD68+, calponin+, actin+ (<50%), desmin+ (~50%), EMA+ (<50%), keratin−, S100−.||RT-PCR for EWS-ATF1, EWS-CREB1, or FUS-ATF1.|
|Endometrial stromal sarcoma||Adult females. Abdominopelvic region. Distant metastases to lungs.||Oval/round to spindle cells. Vague resemblance to proliferative pattern endometrial stroma.||CD10+, ER+, BCL2−, CD34−, SMA and desmin positivity with smooth muscle differentiation.||FISH for 7p15 translocation better than RT-PCR for JAZF1-JJAZ (SUZ12) fusion.|
|Gastro-intestinal stromal tumor||Adults. GI tract. Abdominopelvic region. Metastases often to the liver.||Spindle or epithelioid cells.||DOG1+, CD117+, CD34+, often negative for S100, actin and desmin.||KIT activating mutations.|
Line of Differentiation: Epithelial
Carcinoma comprises approximately 90% cases of CUPS. Within the carcinoma category, the overwhelming majority of tumors are adenocarcinomas (~70%). The poorly differentiated carcinoma group comprises approximately 15% to 20%, and the remaining tumors represent either squamous cell carcinoma (5%) or neuroendocrine carcinomas (5%). CK stains are an excellent marker of epithelial differentiation and are strongly and diffusely expressed in carcinomas. However, examples of keratin positivity have been described in almost all tumor types including sarcomas, melanomas, and even lymphomas. Despite these disturbing reports, when an epithelioid tumor is overwhelmingly positive for pankeratin stains, a diagnosis of carcinoma must be seriously evaluated. The CKs are further discussed in step two.
Step Two: The Cytokeratins—An Overview
The soft epithelial keratin intermediate filaments comprise approximately 20 different keratin polypeptides. The polypeptides, numbered 1 through 20, comprise the type II (basic) keratins and the type I (acidic) keratins ( Table 8.2 ). This family of intermediate filaments is crucial in diagnostic IHC for the identification of carcinomatous differentiation and for identification of specific carcinoma subtypes.
|Type II (Basic) Keratin||Molecular Weight (kDa)||Typical Distribution in Normal Tissue||Type I (Acidic) Keratin||Molecular Weight (kDa)|
|CK1||67||Epidermis of palms and soles||CK9 |
|CK2||65||Epithelia, all locations||CK11||56|
|CK4||59||Nonkeratinizing squamous epithelia||CK13||51|
|CK5||58||Basal cells of squamous and glandular epithelia, myoepithelial, mesothelium||CK14 |
|CK6||56||Squamous epithelia, especially hyperproliferative||CK16||48|
|CK8||52||Basal cells of glandular epithelia, myoepithelial||CK18||45|
|Simple epithelia, most glandular and squamous epithelia (basal)||CK19||40|
|Simple epithelia of intestines and stomach, Merkel cells||CK20||46|
Keratin filaments are formed by tetrameric heteropolymers of two different keratins, two from type I and two from type II, to maintain cellular electrical neutrality. The vast majority of keratins are paired together as acidic and basic types, with rare exception. The classification and numbering system of the keratins is based on the catalog of Moll and associates.
There are 12 keratins with more acidic isoelectric points that form type I (acidic) keratins and 8 keratins with more basic isoelectric points, the type II (basic-neutral) keratins. The keratins are products of two gene families: most genes for type II keratins are localized on chromosome 12 and the genes for type I keratins are localized on chromosome 17. Within each group, the CKs are numbered consecutively from highest to lowest molecular weight in each group. In addition, most low-molecular-weight (LMW) keratins are typically found in all epithelia except squamous epithelium, whereas high-molecular-weight (HMW) keratins are typical of squamous epithelium.
The original methods for identification of the different keratin types in tissues relied on tedious biochemical methods, chiefly performed by Franke and Moll and their associates. More recently, the problem of keratin subtyping has been expedited by the development of numerous monoclonal keratin-specific antibodies. This development was crucial for the ease of keratin subtyping that is now indispensable to the surgical pathologist.
In the last few years, our knowledge on keratins has tremendously increased, with discovery of many new keratin genes (now numbered at 54 keratin genes); however, most newly discovered are expressed in hair follicles. There is even a new consensus nomenclature for mammalian keratins. There are now 28 type I keratin genes (17 epithelial and 11 hair keratins) and 26 type II keratin genes (20 epithelial and 6 hair keratins). The keratins that are functionally useful in determining site of origin have however remained limited (namely keratins 5, 7, 8, 14, 17, 18, 19, and 20).
The detection of keratin, and therefore carcinomatous differentiation, is possible in tumors with extensive necrosis. Judkins and colleagues studied a small number of tumors with necrotic areas, including carcinomas, melanomas and sarcomas, with a panel of antibodies and found that 78% of carcinomas stained with at least one antikeratin antibody in necrotic areas with 100% specificity.
Distribution of Keratin Antigens in Tissues
Simple Epithelial Keratins
Simple epithelial keratins are the first keratins to appear in embryonic development, as they are expressed in virtually all simple (nonstratified), ductal, and pseudostratified epithelial tissues. Because these keratins are widespread, they may be useful for the identification of epithelial differentiation. Almost all mesotheliomas and carcinomas, except squamous cell carcinomas, contain the simple keratins 8 and 18, and a few visceral organs such as liver contain only keratins 8 and 18.
Although identified by many keratin antibodies that recognize a cocktail of keratin peptides (e.g., pankeratin antibodies AE1 and AE3), CAM5.2 and 35BH11 recognize keratins 8 and 18 almost exclusively ( Fig. 8.3 ). This group of antibodies is perhaps the most commonly used to demonstrate the simple keratins in surgical pathology. Because simple keratins are widely distributed in most carcinomas, these antibodies are particularly useful in the initial approach to investigation for carcinomatous differentiation ( Table 8.3 ; see also Table 8.2 ).
|CK8||35BH11||Carcinomas of simple epithelium|
|CK8||CAM5.2||Carcinoma of simple epithelium|
|Pankeratin||AE1/AE3||Carcinomas of simple and complex epithelium|
|CK1/10||34B4||Squamous cell carcinoma|
|CK7||OV-TL 12/30||Nongastrointestinally derived carcinomas|
|CK20||K20||Most gastrointestinal carcinomas; mucinous ovarian, biliary, transitional, and Merkel cell carcinoma|
|CK19||RCK 108||Most carcinomas; many carcinomas with squamous component; myoepithelial cells|
|CK1/5/10/14||34betaE12||Basal cells of prostate; most duct-derived carcinomas|
|CK10/11/13/14/15/16/19||AE1||Most squamous lesions and many carcinomas|
The lowest molecular weight of the keratin group, CK19 is a simple keratin that has a distribution similar to keratins 8 and 18 and is also present in the basal layer of the squamous epithelium of mucosal surfaces and may be seen in epidermal basal cells. CK19 is a good screening marker for epithelial neoplasms because of its wide distribution in simple epithelia and in many squamous tissues. The monoclonal antibody AE1 (Boehringer-Mannheim, Indianapolis, Indiana) reacts with CK19 as does the AE1/AE3 cocktail (Boehringer-Mannheim). Also reacting in formalin-fixed tissues is a monoclonal antibody to CK19-RCK108 (DAKO, Carpinteria, California). In contrast, CK19 is mostly negative or rarely is seen focally in hepatocellular carcinoma.
CK7 is a 54-kDa type II simple keratin that has a restricted distribution compared with keratins 8 and 18. Its presence in many simple, pseudostratified, and ductal epithelia and mesothelia is similar in distribution to that of keratins 8 and 18. Much of the data in the literature on CK7 are based on the reactivity patterns of antibody OV-TL 12/30 (DAKO) in FFPE tissues. The OV-TL 12/30 antibody parallels the CK7 immunoreactivity with RCK 105, an antibody for use on frozen sections. In addition, predigestion with protease or heat-induced epitope retrieval (HIER) is required for OV-TL 12/30. The lack of, or extreme paucity of, CK7 distribution in tissues such as colonic epithelium, hepatocytes, and prostatic acinar tissue is used to diagnostic advantage. This antibody identifies transitional cell epithelium ( Fig. 8.4A ) but is predominantly negative in most squamous epithelia. The restricted topography of CK7 makes it especially useful in evaluating the origin of adenocarcinomas, as this keratin is present in most breast, lung, ovarian, pancreaticobiliary, and transitional cell carcinomas, but it is either absent or present in only rare cells in colorectal, renal, and prostatic carcinomas ( Box 8.1 , Table 8.4 ). CK7 stains squamous cell carcinoma and squamous dysplasias of the cervix. Although hepatocellular carcinomas are negative for CK7, it is expressed in the fibrolamellar variant. CK7 is also typically expressed in mammary and extramammary Paget disease. A diagnostic pitfall in the interpretation of CK7 is that CK7 stains subsets of endothelial cells of normal soft tissues, as well as endothelial cells in venules and lymphatics in intestinal mucosa, uterine exocervix, and lymphoid tissue.
Transitional cell carcinoma
Ovarian mucinous carcinoma
Non-small cell carcinoma of lung
Small cell carcinoma of lung
Breast carcinoma, ductal and lobular
Nonmucinous ovarian carcinoma
Pancreatic ductal adenocarcinoma
Squamous cell carcinoma of cervix
Merkel cell carcinoma
Squamous cell carcinoma, lung
Renal cell carcinoma
Some thymic carcinoma
|Lung, small cell carcinoma||43|
|Salivary gland, all tumors||100|
|Thyroid, all tumors||98|
|Bladder, transitional cell||88|
|Cervix, squamous cell||87|
|Head and neck, squamous cell||27|
|Esophagus, squamous cell||21|
|Germ cell, carcinoma||7|
Diagnostic Utility of Cytokeratin 7.
The specific diagnostic utility of CK7 lies in the fact that there are three dominant patterns of immunostaining:
Tumors that are characteristically strongly and diffusely positive include those of the salivary glands, lung, breast, ovary, endometrium, and bladder, as well as mesotheliomas, neuroendocrine tumors, pancreaticobiliary adenocarcinomas, and the fibrolamellar variant of hepatocellular carcinoma. CK7 is also typically expressed in tumor cells of mammary and extramammary Paget disease.
CK7 variably stains the tumor cells in biliary and gastric tumors.
Carcinomas that are almost invariably negative but may occasionally show rare CK7+ cells include hepatocellular carcinomas, duodenal ampullary carcinomas, colon carcinomas, renal, prostate, and adrenal cortical tumors.
Strong diffuse CK7 immunostaining is a valuable marker in the diagnostic work-up of a carcinoma and may be used as a starting point for further immunohistochemical study. Metastatic carcinomas in the lung that are CK7+ must be differentiated from a primary lung carcinoma with a panel of antibodies, and the IHC work-up will be dependent on the patient’s age, gender, and presenting findings. It is important to remember that CK7 may be expressed infrequently in certain tumors (see Table 8.4 ). In general, there is high fidelity of CK7 expression between primary and metastatic carcinomas.
CK20 is a 46-kDa LMW keratin that was discovered by Moll and associates. The tissue distribution of CK20 is limited predominantly to GI epithelium and its tumors, mucinous tumors of the ovary, and Merkel cell neoplasms. The limited distribution of CK20 in colorectal, pancreatic, and gallbladder carcinomas, Merkel cell carcinomas, and transitional cell carcinomas (see Fig. 8.4B ) is useful in the identification of this group of tumors in primary or even metastatic sites. When combined with the specific tissue distribution of other keratins such as CK7, it is possible to identify colon cancer metastases in the lung, distinguish pulmonary small cell carcinoma from Merkel cell carcinoma, and distinguish transitional cell carcinoma from other squamous cell carcinomas and poorly differentiated carcinomas. It is of importance to recognize that CK20 in this subgroup of tumors is most often distributed strongly and diffusely. Rare CK20+ cells may be seen in some other neoplasms. Up to 10% of primary pulmonary adenocarcinomas are not otherwise specified (NOS) and up to 25% of mucinous bronchioloalveolar types may show CK20+ cells. In addition, the controversial primary mucinous carcinoma of the lung (colloid carcinoma, goblet cell variant) shows CK20 immunostaining in about 50% of cases along with nuclear positivity for CDX2, a gut-specific marker. A very small percentage of müllerian and breast carcinomas may also show CK20 positivity.
Cholangiocarcinomas of liver are also positive; the central (large duct) carcinomas are more likely to have a high labeling index for CK20 in addition to CK7. The positive predictive value using the combination of CK7 and CK20 to predict the presence of metastatic carcinomas of colorectal or pancreaticobiliary origins in the liver, based on clinical outcomes, is close to 0.9. It is important to remember that CK20 may be expressed infrequently in certain tumors ( Table 8.5 ). In general, there is high fidelity of CK20 expression between primary and metastatic carcinomas (see Fig. 8.4C ).
|Skin, Merkel cell||78|
|Bladder, transitional cell||29|
|Head and neck, squamous cell||6|
|Cervix, squamous cell||0|
|Esophagus, squamous cell||0|
|Germ cell, carcinoma||0|
|Salivary gland, all tumors||0|
|Thyroid, all tumors||0|
|Lung, small cell||0|
|Lung, squamous cell||0|
Although the prominent expression of CKs is the essential element of epithelial differentiation, on occasion expression of other lineage-specific markers may cloud the issue. Such is the case of finding keratins in nonepithelial tissues (see later) and the rare observation of LCA (CD45) in some undifferentiated or neuroendocrine carcinomas and CD30 in embryonal carcinomas. The use of a panel of antibodies and the pattern and intensity of immunostaining is critically important in these confounding situations.
CAM5.2 and AE1/AE3: Broad coverage for detection of carcinomatous differentiation. Both should be used together for screening.
CK7 (+): Adenocarcinomas of breast, lung, ovary, endometrium, and pancreas; mesothelioma, urothelial carcinomas, thymic carcinomas; cervical squamous cell carcinoma; and fibrolamellar variant of hepatocellular carcinomas.
CK7 (negative/rare positive): Renal, prostate, adrenocortical, squamous (except uterine cervix), small cell carcinomas, and hepatocellular carcinomas.
CK20 (+): Colorectal, pancreas (60%), gastric (50%), cholangiocarcinoma (40%), mucinous ovarian, Merkel cell, and urothelial carcinomas (30%).
CK20 (negative/rare positive): Most breast, lung, and salivary gland carcinomas, hepatocellular, renal, prostate, adrenocortical, squamous, and small cell carcinomas.
See Box 8.1 for CK7/CK20 immunoprofile of various carcinomas.
Keratins of Stratified Epithelia: Complex Keratins
Keratins of HMW are observed in stratified epithelia and generally are not present in the simple visceral-type epithelia. Basal cells of prostate and myoepithelial cell populations of ducts and glandular tissue also contain an abundance of HMW type II keratins and LMW type I keratins. The antibody 34bE12 or (catalog number) keratin 903 (K903) identifies a cocktail of keratins including Moll types I, II, V, X, XI, and XIV/XV. The practical diagnostic use of this pattern of expression is to identify basal and myoepithelial cells in their respective organs. For example, the staining of myoepithelial cells around ductal carcinoma in situ or sclerosing adenosis can confirm a noninvasive lesion. This keratin of stratified type is also typically present in squamous epithelium and, using antibody K903, is a good antibody for detecting squamous differentiation in an otherwise poorly differentiated carcinoma.
These HMW structural keratins are also commonly seen in duct-derived epithelium (breast, pancreas, biliary tract, lung) and in transitional, ovarian, and mesothelial tissues. The degree of immunostaining of these tissues with HMW keratin antibodies is typically strong and diffuse, a feature that is helpful diagnostically, because HMW keratin immunostaining is seen only focally in visceral epithelial tissues such as colon, stomach, kidney, and liver.
Confirms the presence of basal cells of prostate ( Fig. 8.5 ).
Confirms the presence of myoepithelial cells in breast.
Present in basal cell layer of stratified and squamous epithelium.
Strong and diffuse in tumors of squamous epithelial differentiation.
Present in a wide variety of duct-derived carcinomas and mesotheliomas, and most neoplasms that demonstrate tonofilaments ultrastructurally.
Cytokeratin 5 and Cytokeratin 5/6.
CK5 and CK6 are basic (type II) polypeptides with molecular weight of 58 and 56 kDa, respectively. Most studies have been performed using antibodies to CK5/6 and have been found useful in the differential diagnosis of metastatic carcinoma in the pleura versus epithelial mesothelioma. Epithelial mesotheliomas are strongly positive in all cases ( Fig. 8.6 ), but up to 30% of pulmonary adenocarcinomas will show focal variable immunostaining.
Almost all squamous cell carcinomas, half of transitional cell carcinomas, and many undifferentiated large cell carcinomas immunostain with CK5/6 ( Table 8.6 ). CK5/6 has excellent sensitivity and specificity for the detection of squamous differentiation in poorly differentiated carcinomas. p63 is also seen with high frequency in squamous and transitional carcinomas, and when p63 is used with the CK5/6 antibody, it affords high sensitivity and specificity for squamous differentiation.
|Skin, squamous cell||100|
|Skin, basal cell||100|
|Salivary gland, all tumors||93|
|Bladder, transitional cell||62|
|Germ cell, carcinoma||0|
|Thyroid, all tumors||0|
Myoepithelial cells of the breast, glandular epithelium, and basal cells of the prostate express CK5/6, and some carcinomas of ovarian origin may display CK5/6.
Hyperplastic mesothelial cells can be seen on occasion in the sinuses of lymph nodes from the chest or cervical chain, and the differential diagnosis in this instance is metastatic carcinoma. The presence of strong, diffuse CK5/6 in the cells of these nests should aid in identifying them as mesothelial in origin, but this should be confirmed by using more specific markers for mesothelium such as calretinin and WT1.
There has been some renewed interest in these antibodies as both CK5 and CK5/6 are also used to identify the basal-like molecular class of breast cancer. We have previously shown that CK5 (clone XM26) is superior to CK5/6 (clone D5/16B4) antibody in identifying the basal-like phenotype of breast carcinoma with high sensitivity and specificity. Whether CK5 is superior to CK5/6 in other distinctions (mesothelioma vs. carcinoma; or in identifying squamous differentiation) needs to be investigated.
Good indicator of squamous and transitional cell differentiation.
Good discriminator of mesothelial differentiation versus adenocarcinoma in lung.
Positive in myoepithelial cells of breast and basal cells of prostate.
Sensitive and specific markers of basal-like phenotype of breast carcinoma.
Keratins in Nonepithelial Cells
Keratins have been documented by IHC, dot immunoblot, and polymerase chain reaction in several types of tumors in which there is no morphologic evidence of epithelial differentiation. This type of keratin immunostaining has been referred to as anomalous, aberrant, spurious, and unexpected.
The keratins most often found in these nonepithelial mesenchymal tissues or melanocytic lesions are keratins 8 and 18 and, less commonly, keratin 19. Antibodies that detect these LMW keratins have demonstrated positive immunostaining in a variety of FFPE mesenchymal tumors, including leiomyosarcomas (21% to 25%), fibrosarcoma (4%), liposarcoma (21%), rhabdomyosarcoma, MPNSTs (5%), some malignant fibrous histiocytomas (5%), GI stromal tumors, rare solitary fibrous tumors of pleura, angiosarcoma (33%), endometrial stromal sarcoma, and primitive neuroectodermal tumors. Keratin usually stains scattered cells in this group of tumors in traditional FFPE tissue, whereas carcinomas and sarcomatoid carcinomas are heavily and diffusely stained ( Fig. 8.7 ). In addition, keratin-positive soft tissue and bone tumors with partial epithelial differentiation are variably stained with keratin in the epithelial areas as expected. This group includes synovial sarcomas, epithelioid sarcoma, chordoma, MPNST, and adamantinoma of long bones. Although some of the soft tissue tumors may mimic metastatic carcinoma morphologically, the finding of sporadic cell immunostaining is unlike the strong, diffuse immunostaining seen in carcinomas, especially when using the broad-coverage antibodies. Frozen tissues fixed in acetone or alcohol, including alcohol-fixed cytologic specimens yields far more keratin-positive cells, and this can be confusing diagnostically, especially with cytologic specimens for which alcohol is a standard fixative for needle aspiration specimens.
Malignant melanoma also demonstrates immunostaining for keratins 8 and 18, but in FFPE tissues, the prevalence is around 1% of cases, with focal tumor cell staining. Frozen sections and alcohol-fixed melanomas show substantially more positive tumor cells than do formalin-fixed specimens, and it is important to recognize this to avoid misdiagnosing melanoma as a carcinoma, especially in alcohol-fixed cytologic preparations. The consensus regarding keratin immunostaining of nonepithelioid sarcomas and melanomas is that although the presence of keratin is real, as measured by molecular techniques and more sensitive immunohistologic methods (frozen sections, alcohol fixation), the observed nonexpression of keratin staining in these tumors in formalin-fixed tissue is desirable because of its diagnostic usefulness.
Truly “spurious” keratin immunoreactivity has been described in human glial tissue and in some human astrocytomas, especially with antibodies AE1 and 34BE12. In addition, the cocktail AE1/AE3 may cross-react with both normal and neoplastic astrocytes. The spurious keratin immunoreactivity is probably due to cross-reaction with glial cells containing glial fibrillary acidic proteins. This is an obvious pitfall for the misdiagnosis of metastatic carcinoma in the brain. The antibody CAM5.2 does not react with astroglial cells; thus it is best used to detect carcinomatous differentiation in the CNS.
Meningiomas, especially the “secretory variant,” may express keratin in up to one-third of cases.
Epithelial differentiation is simulated in lymph nodes with the LMW keratin-positive fibroblastic reticulum cells of the paracortex ( Fig. 8.8 ). These dendritic cells immunostain with CAM5.2, rarely with AE1/AE3, revealing an extensive network of extrafollicular dendritic processes in lymph nodes, tonsils, and spleen. These keratin-positive cells are a pitfall for the diagnosis of metastatic carcinoma, because the conventional wisdom had been that keratin-positive cells in a lymph node equated with metastatic carcinoma. The pitfall is twofold. When searching for keratin-positive micrometastases in patients with breast carcinoma, one must distinguish the dendritic processes from carcinoma cells that cluster in the subcapsular sinus. Also, needle aspirates and touch imprints of lymph nodes may contain keratin-positive cells without containing metastatic carcinoma; one must be aware of the morphologic features of the keratin-positive cells.
Keratin positivity has been described in plasma cells, plasmacytoma, and in anaplastic large cell lymphoma. For anaplastic large cell lymphoma, keratins may be detected in as many as 30% of cases and, along with some EMA-positive anaplastic lymphoma cells, the definitive diagnosis can be confusing. However, adherence to a broad-spectrum antibody for keratin immunoreactivity will show only focal rare staining at most in these lymphomas. Plasmacytomas likewise should be studied with broad-coverage antibodies in a panel that includes antibodies to CD138 and kappa/lambda light chains.
The majority of keratin immunostaining is performed on FFPE tissues. The duration of formalin fixation is a key factor when trying to optimize the technical performance of keratin immunoperoxidase stains. The fixation time is closely related to the time required for enzymatic predigestion. In general, tissue fixed in 10% formalin for more than 2 days requires greater antigen retrieval, with less time required for tissues fixed briefly (hours) in 10% formalin. Most, if not all, keratin antibodies require epitope retrieval (depending on antibody and fixation duration) for optimal keratin antibody performance.
Focal presence in many sarcomas (see text).
Focal rare presence in melanoma mainly with CAM5.2.
Plasma cells common; other lymphoid neoplasms rare.
Common in dendritic cells of lymph nodes mainly with CAM5.2.
Antibody AE1/AE3 may give spurious positive keratin result in astrocytic neoplasms.
Step Three: Carcinoma Subsets With Frequent Vimentin Coexpression
Mesenchymal and endothelial cells regularly immunostain with vimentin, and this immunostaining generally provides a measure of internal quality control for the quality of immunoreactivity. If there is no immunostaining of blood vessels or stromal cells by vimentin, it denotes significant damage to tissue antigens or other failure of the staining procedure.
Carcinomas in effusion specimens are universally positive for vimentin (presumably an in vivo fluid effect) and thus have no diagnostic utility.
Initially thought to be an intermediate filament restricted to mesenchymal cells, vimentin has been found in a diverse number of neoplasms, including a variety of carcinomas ( Box 8.2 ). Vimentin stains virtually all spindle cell neoplasms—mesenchymal spindle cell neoplasms and sarcomatoid carcinomas included. However, vimentin stains a subset of carcinomas regularly and to a significant degree, and this may be useful in the context of a panel of antibodies to narrow a differential diagnosis. The cellular vimentin immunostaining pattern is often a perinuclear band of reactivity, particularly for endometrioid adenocarcinomas. Carcinomas with frequent (more than 50% to 60%) and strong (more than 25% of cells) vimentin coexpression include spindle cell carcinomas, renal cell carcinomas (RCCs) (except the chromophobe variant), müllerian endometrioid adenocarcinomas and malignant mixed müllerian tumors, serous ovarian carcinomas, pleomorphic salivary gland tumors, “basal-like” breast carcinomas, and follicular thyroid carcinomas. Epithelial and sarcomatoid mesotheliomas also regularly demonstrate vimentin. Certain carcinomas may immunostain with vimentin but with lesser frequency (10% to 20%) and with far less intensity (<10% of cells). This group includes adenocarcinomas of colorectum, lung, breast, and prostate, and nonserous ovarian carcinomas.
Coexpression Common (>50%)
Renal cell carcinoma
Salivary gland carcinoma
Spindle cell carcinoma
Thyroid follicular carcinoma
Coexpression Uncommon (<10%)
Breast ductal-lobular carcinoma
Lung non-small cell carcinoma
Therefore the finding of substantial coexpression of vimentin in a metastatic carcinoma may aid in narrowing the differential diagnosis and adds value to the rest of the antibody panel ( Fig. 8.9 ).
Vimentin coexpression is especially useful in differentiating endometrial endometrioid carcinomas in uterine curettage specimens from endocervical adenocarcinomas, including the endometrioid variant of endocervical adenocarcinoma. Endometrial endometrioid carcinomas immunostain strongly for vimentin but endocervical carcinomas rarely stain (weak focal staining in up to 13% of endocervical carcinomas). However, with the current antigen retrieval techniques, moderate to occasionally strong vimentin expression may be seen in endocervical carcinomas, and a panel approach is more useful in this distinction.
Common in renal, endometrioid endometrial, salivary gland, follicular thyroid, and sarcomatoid (spindle cell) carcinomas, “basal-like” breast carcinomas, as well as in the stromal components of malignant mixed müllerian tumors.
May be seen in few cells in 10% to 20% of colorectal, lung, breast, prostate, and ovarian adenocarcinomas.
Not diagnostically useful in body cavity effusion specimens.
Epithelial and sarcomatoid mesotheliomas are usually vimentin positive.
Important internal quality measure for antigen assessment in any tissue.
Step Four: Supplemental Epithelial Markers
Although not specific for tissue lineage, these epithelial markers demonstrate characteristic immunostaining patterns for certain tissue types and therefore are useful to corroborate a diagnosis when used as part of a panel of antibodies. With the availability of more tissue specific markers and transcription factors, the usage of these supplemental markers has decreased overtime. Only the more commonly used markers are discussed.
CEA is a 180-kDa glycoprotein that is 50% carbohydrate, and there are many CEA antibodies available to a variety of CEA epitopes. The polyclonal antibodies commonly cross-react with tissue nonspecific cross-reacting antigens and biliary glycoprotein-1. The polyclonal (p) antibody is therefore used for demonstrating canalicular pattern in hepatocellular carcinoma, and monoclonal (m) antibodies are used for everything else. Although CEA is a sensitive marker, adenocarcinomas of colorectal origin cannot be distinguished from lung adenocarcinomas or ductal carcinomas of the breast because of the low specificity of CEA.
Primary adenocarcinomas of the lung are typically CK7+, CK20−, and CEA+, whereas colorectal carcinomas are CK7−, CK20+, and CEA+; ductal and lobular breast carcinomas are CK7+, CK20−, and often CEA+; and ovarian carcinomas are CK7+, CK20+/−, and CEA−. Neoplasms that typically are strongly positive for most CEA antibodies include adenocarcinomas of the lung, colon, stomach, biliary tree, pancreas, urinary bladder, endocervix, paranasal sinuses, sweat glands, and breast ( Box 8.3 ). The usefulness of CEA when used with keratins is to corroborate expected staining for CEA, whether positive or negative.
Carcinoembryonic Antigen (+)
a Pericanalicular pattern with polyclonal antibody.
Carcinoembryonic Antigen (−)
Neoplasms that are essentially negative with most CEA antibodies include adenocarcinomas of prostate, kidney, adrenal gland, and endometrium along with serous ovarian tumors, thyroid tumors (except medullary type) and mesotheliomas. However, squamous differentiation in endometrioid endometrial tumors often shows patchy reactivity for CEA. Liver cell-derived tumors are nonreactive with the monoclonal CEA antibodies but do react with the polyclonal antibodies in a distinct pattern of pericanalicular staining ( Fig. 8.10 ), as the polyclonal antibodies cross-react with the hepatic bile canalicular biliary glycoprotein-1. Adenocarcinomas of pulmonary, GI, thymic, endocervical, and pancreaticobiliary origin typically show strong, although variable, cytoplasmic immunostaining for CEA antibodies.
Immunostaining patterns of CEA in the liver are particularly useful. The epithelioid hemangioendothelioma (EH) of liver can mimic carcinoma to perfection. Demonstration of positive CD31/CD34 and factor VIII with variable (usually focal, sometimes diffuse) keratin and lack of CEA will separate this entity from hepatocellular carcinoma. However, it is important to remember that unlike normal liver, neoplastic liver sinusoids demonstrate the presence of immunoreactive CD34. This may be confused with EH, but it is useful in the differential diagnosis of primary liver neoplasm versus metastatic carcinoma and nonneoplastic liver, especially on small biopsy samples ( Fig. 8.11 ).
CEA(p)+ tumor (peri-canalicular pattern): Hepatocellular carcinoma.
Other CEA(m)+ tumors: Gastrointestinal, lung, breast, thymus, endocervical, primary cholangiocarcinoma.
CEA negative tumors: Prostate, renal, endometrial, adrenal, and serous ovarian tumors and mesothelioma.
Epithelioid hemangioendothelioma of liver (mimicker of hepatocellular carcinoma): CEA negative, but positive for vascular markers.
CEA , Carcinoembryonic antigen.
Epithelial Membrane Antigen
Encoded by the MUC1 gene on chromosome 1 and a derivative human antigen, EMA is a transmembrane glycoprotein of the breast mucin complex, and its expression is increased in carcinomas. Unlike normal breast in which EMA is present on the apical cell membrane, neoplasms demonstrate EMA on the entire circumference of the cell membrane. However, increased amounts of the large glycoprotein interfere with cell-to-cell and cell-to-matrix adhesion in neoplastic cells.
The utility of EMA antibody is in the detection of epithelial differentiation, as a supplement to the CK. Spindle cell, small cell, and large cell neoplasms may, on rare occasion, be stained with EMA but be only focally positive for CK.
There are several EMA antibodies available, each of which reacts to different epitopes of the large glycoprotein antigen; they include MAM-6, episialin, polymorphic epithelial mucin, CA15-3, DF3 antigen, and breast epithelial mucin. The EMA antibodies stain skin and adnexa, breast, lung, bile ducts, pancreas, salivary gland, urothelium, endometrium and endocervix, prostate ducts, thyroid, mesothelium, and neoplasms of these tissues ( Table 8.7 ). Many sarcomatoid carcinomas and epithelial and sarcomatoid mesotheliomas are positive. Reactive mesothelium may stain weakly compared with thick membranous staining of mesothelioma. Many types of adenocarcinomas immunostain with EMA and must be distinguished from mesothelial cells in effusions by using a panel of immunostains that includes CK5/CK6, CEA, LeuM1, and BER-EP4.
|Typically positive||Carcinomas: Skin and adnexa, breast, bile ducts, lung, pancreas, salivary gland, urothelium, endometrium, endocervix prostate, thyroid |
Noncarcinomatous lesions: Meningioma, mesotheliomas, solitary fibrous tumor
|Focal/patchy positive||Carcinomas: Sarcomatoid carcinomas |
Noncarcinomatous lesions: Plasma cell tumors, L&H cells of Hodgkin lymphoma, few cells of non-Hodgkin lymphoma, anaplastic large cell lymphoma, malignant peripheral nerve sheath tumors, synovial sarcoma, leiomyosarcoma
|Mostly negative||Germ cell tumors except choriocarcinoma, ovarian sex cord stromal tumors|
Subsets of normal and neoplastic hematopoietic cells express EMA, including plasma cells and erythroblasts, and neoplastic cells including the lymphocytic and histiocytic (L&H) cells (60% of cases) of lymphocyte-predominant Hodgkin lymphoma, 5% of B-cell lymphomas, 18% of T-cell lymphomas, and about 60% of anaplastic large cell lymphomas. The EMA antibodies do not have absolute sensitivity and specificity for carcinomas and therefore should always be used with a panel of CK and other corroborating antibodies such as LCA. However, in distinguishing ovarian stromal neoplasm from carcinoma, EMA is better than keratin. Most ovarian stromal neoplasms show focal to patchy keratin reactivity, but they are almost always negative for EMA.
In addition to epithelial neoplasms, a number of spindle cell tumors, sarcomas, CNS tumors, small round cell tumors, and a few germ cell tumors may be positive with EMA. These tumors include solitary fibrous tumors, meningiomas, ependymomas, malignant nerve sheath tumors, synovial sarcoma, leiomyosarcoma, malignant fibrous histiocytoma, epithelioid sarcoma, and chordoma. With the exception of the last two tumors mentioned, EMA immunostaining is focal.
Choroid plexus neoplasms and meningiomas show strong membranous EMA immunostaining. Germ cell tumors are largely negative except for variable EMA immunostaining in choriocarcinoma and teratoma, whereas the epithelial small round cell tumors of nephroblastoma and hepatoblastoma immunostain with EMA in the majority of cases.
Used as a supplement to detect epithelial neoplasms because cytokeratins may rarely be focally positive or negative in undifferentiated carcinomas.
EMA immunostaining may be membranous or cytoplasmic, or both.
Positive in meningioma and ependymoma.
Positive in renal cell carcinoma, negative in adrenocortical tumor.
Negative in malignant melanoma.
Negative in germ cell neoplasms except choriocarcinoma and teratoma.
EMA stains plasma cells, 60% of anaplastic large cell lymphomas, L&H cells of Hodgkin lymphoma, and a few T- and B-cell lymphomas.
Strongly positive in malignant mesothelioma, and some reactive mesothelial proliferations, and weak or negative immunostaining for normal mesothelial cells.
Almost always negative in ovarian stromal neoplasms.
EMA , Epithelial membrane antigen.
BER-EP4, Bg8 and MOC-31
These are markers of epithelial differentiation. These stains are most often used to distinguish adenocarcinomas from mesothelial proliferations ( Fig. 8.12 ). BER-EP4 and MOC-31 antibodies are directed against the epitope on glycoproteins present on the surface of glandular epithelial cells of endodermal derivation. Squamous epithelium of ectodermal derivation virtually never expresses BER-EP4. The characteristic staining with BER-EP4 and MOC-31 is membranous. Bg8 antibody is directed against the Lewis Y antigen and shows cytoplasmic staining in carcinomas. When BER-EP4, Bg8, and MOC-31 are combined with calretinin (nuclear and cytoplasmic staining in mesotheliomas) and WT1 (nuclear staining in mesotheliomas), the combination provides the best sensitivity and specificity for distinguishing adenocarcinoma from mesothelioma. MOC-31 is also very useful in distinguishing metastatic tumors from primary hepatocellular carcinoma when used in a panel format. Along with Hepatocyte Paraffin-1 (HepPar-1) and pCEA, MOC-31 permits distinction of metastatic carcinomas in liver from hepatocellular carcinoma 99% of the time.
Best epithelial markers to distinguish adenocarcinomas from mesothelioma.
Membranous reactivity for BER-EP4 and MOC-31 and cytoplasmic reactivity for Bg8.
Squamous cell carcinomas generally negative for BER-EP4.
MOC-31 is especially useful in distinguishing between primary liver carcinoma versus metastatic tumor.
Step Five: Focusing on Tumor Differentiation-Cell Specific Products
The tissue of origin of metastases can be narrowed to few sites with the panel approach of CKs, CEA, EMA, and vimentin. The use of additional antibodies to cell-specific products in most instances has a very high specificity for certain tissues, enabling the pathologist to “fine focus” the search for the origin of a metastasis.
The antibodies discussed here include neuroendocrine markers, thyroglobulin, TTF-1, calretinin, WT1, GCDFP-15, mammaglobin, hormone receptors, villin, CDX2, HepPar-1, glypican-3 (GPC3), Arginase-1, DPC4, prostate carcinoma antigens, uroplakin III (UPIII) and II, thrombomodulin (TM), RCC, CD10, PAX8 and PAX2, GATA3, melan-A, SOX10, inhibin, adrenal binding protein, germ cell tumor markers, and CD5.
Antibodies to neuroendocrine cell components are usually used in the context of trying to distinguish tumor cell types in specific organs such as lung, thyroid, colon, and adrenal gland. The antibodies are not typically used in the initial screening panel of the work-up of an undifferentiated tumor, and there is little literature that deals with this topic. It is critically important to use more than one antibody, because no single antibody has perfect specificity and sensitivity.
It is well known that a few “neuroendocrine cells” can be seen with IHC in a wide variety of carcinomas. This is not to be equated with a diagnosis of neuroendocrine carcinoma. Only after a complete account of the clinical findings, imaging studies, histologic studies, and immunohistochemical findings should a diagnosis be rendered.
The chromogranins (types A, B, and C) are a group of monomeric proteins that compose the major portion of the soluble protein extract of the neurosecretory granules of neuroendocrine cells; chromogranin A, with a molecular weight of 75 kDa, is the most abundantly distributed. There is a strong correlation between the chromogranin cellular immunostaining quantity and the number of neuroendocrine-type secretory granules seen at the level of electron microscopy.
The LK2H10 clone is a monoclonal antibody with abundant representation in the literature. Immunostaining intensity decreases with poor differentiation. The specificity of LK2H10 is close to 100%, but sensitivity is closer to 75%. Chromogranin reactivity is generally patchy compared with synaptophysin in neuroendocrine tumors.
Synaptophysin is a glycoprotein that is an integral part of the neuroendocrine secretory granule membrane and is recognized by monoclonal antibody (SY38) in a variety of neuroendocrine tumors. Synaptophysin is a broad-spectrum neuroendocrine marker, with higher sensitivity but lower specificity than antibody to chromogranin. Immunostaining for SY38 has also been documented to be most effective in identifying metastases of neuroendocrine type. Synaptophysin immunostaining alone is insufficient grounds for labeling a neoplasm neuroendocrine. When used in the context of appropriate morphology, synaptophysin is useful to identify neuroendocrine features.
Large cell undifferentiated neuroendocrine carcinoma (LCNEC) can present as CUPS, and it is easy to miss the diagnosis without applying the appropriate neuroendocrine markers. The correct diagnosis of LCNEC is an important distinction because it carries the same dismal prognosis as small cell carcinoma, whether in the lung or GI tract. Synaptophysin may be the most frequent positive marker in LCNEC.
In one study, small cell lung carcinomas were stained by synaptophysin in up to 79% of cases, whereas chromogranin was positive in 47% to 60% of cases, bombesin was positive in 45% of cases, and neuron-specific enolase (NSE) was seen in 33% to 60% of cases. In addition, synaptophysin may be seen in 8% of non-small cell carcinomas.
Neural Cell Adhesion Molecule (CD56) and Leu 7 (CD57)
These two antigens are very similar. CD56 is a glycoprotein expressed on neurons, glial tissue, skeletal muscle, and natural killer cells. The CD57 antigen of a human T-cell line generated a monoclonal antibody (HNK-1). The differentiation antigen of the T-cell line is indicative of a natural killer cell activity. The CD57 antibody also recognizes antigen of myelin-associated glycoprotein in the myelin of the central and peripheral nervous systems; CD57 reactivity has also been found in enterochromaffin cells, pancreatic islet cells, islet cell tumors, carcinoid tumors, pheochromocytomas, and small cell carcinoma of the lung. CD56 and CD57 lack the specificity of chromogranin and synaptophysin for detecting neuroendocrine neoplasms and therefore should be used as part of a panel that includes these antibodies.
The enolase enzymes comprise five different forms, each of which is composed of three homodimers and two hybrids. NSE is found in a variety of normal and neoplastic neuroendocrine cells and predominates in the brain. Originally believed to be a specific marker for neuroendocrine differentiation, it has subsequently been observed that NSE can be found in virtually any type of neoplasm, and because of this, it is a poor antibody to use to screen for neuroendocrine differentiation. Overall a poor marker for detection of neuroendocrine differentiation because of its lack of specificity, NSE may be useful in combination with other more specific antibodies, such as chromogranin and synaptophysin, for the appropriate neuroendocrine morphologic identification and documentation of immunostaining.
Peptide hormones are present in unique, sequestered tissues in the normal state and generally recapitulate the same hormone production in neoplasms. Endocrine neoplasms, with few exceptions, show a characteristic histologic pattern; therefore the study of hormone production is often of academic interest only.
Poorly differentiated endocrine neoplasms, depending on the site of origin, may produce characteristic peptide hormones. The group of poorly differentiated neuroendocrine tumors and their hormone production include islet cell tumors (insulin, glucagon, somatostatin, gastrin), pulmonary small cell carcinoma (bombesin in 45% of cases), and medullary thyroid carcinoma (calcitonin).
Cytokeratin Profile of Neuroendocrine Carcinoma
The CK profile of neuroendocrine carcinomas is somewhat distinctive in that virtually all are positive to some degree for CKs 8 and 18 (e.g., CAM5.2), sometimes positive for CK7 and negative with CK20 and HMW keratin (e.g., K903). Merkel cell carcinomas are characteristically positive for CK20 (67% of cases) and negative for CK7, which is the reverse for immunostaining of small cell carcinomas of lung (CK7+, CK20−).
Typical neuroendocrine keratin profile: CAM5.2+, and CK7+.
Chromogranin and synaptophysin complement each other as part of a diagnostic panel.
CD56, CD57 and other specific peptide (e.g., bombesin, glucagon) may supplement the preceding studies.
Subsets of peripheral neuroendocrine carcinomas may be TTF-1+ (see below).
Subsets of neuroendocrine carcinomas of gastrointestinal tract may be CDX2 (see later text).
Thyroglobulin, a 670-kDa heavily glycosylated protein, provides iodination sites for the production of thyroid hormones and is unique to the thyroid follicular epithelium. The great majority of thyroid carcinomas show immunostaining with thyroglobulin, although most of the positive cases are readily interpreted as follicular or papillary carcinomas. The undifferentiated anaplastic carcinomas are generally negative for thyroglobulin. Thyroid carcinomas (except medullary type) are almost always negative with monoclonal CEA antibody, which is a helpful feature in differential diagnosis. Thyroglobulin is often negative or may be seen as scattered positive cells in medullary carcinoma and, conversely, calcitonin-positive cells may be seen in poorly differentiated follicular carcinomas. Thyroglobulin may be seen in 10% to 25% of cases of leukemic blast cells in bone marrow.
Positive in almost all thyroid carcinomas (except medullary), with reduced immunostaining of poorly differentiated carcinoma to complete absence in anaplastic types.
Highly specific for thyroid carcinomas, with rare positivity in some leukemic blast cells.
Thyroid Transcription Factor-1 and Other Lung Markers
TTF-1, a nuclear tissue-specific protein transcription factor, is found in thyroid and thyroid tumors regardless of histologic type (except anaplastic type), as well as in lung carcinomas, including adenocarcinomas (75%), non-small cell carcinomas (63%), neuroendocrine and small cell carcinomas (>90%), and rare squamous cell carcinomas (<10%). Selectively expressed during embryogenesis in the thyroid, the diencephalon of the brain, and in respiratory epithelium, TTF-1 binds to and activates factors for surfactant protein derived from Clara cells. TTF-1 is rarely seen in carcinomas outside of the lung or thyroid ( Fig. 8.13 ). Neuroendocrine tumors of the lung, including typical and atypical carcinoids and large cell neuroendocrine carcinomas, are almost always positive with TTF-1, demonstrating a kinship with small cell carcinomas. Small cell and large cell neuroendocrine carcinomas from origins other than the lung are also frequently TTF-1 positive. These sites include prostate, bladder, cervix, GI tract, thyroid, and breast. However, Merkel cell carcinomas are TTF-1 negative.
The utility of TTF-1 becomes readily apparent in the differential diagnosis of primary versus metastatic carcinomas, especially in the lung or in effusions. CK7 and CK20, along with TTF-1 and CEA, are the antibodies that best discriminate primary lung carcinoma from carcinomas metastatic to the lung. In a study by Roh and Hong, the sensitivity of TTF-1 for metastatic lung carcinoma in lymph nodes was 69%.
The specificity of TTF-1 for pulmonary lesions was confirmed by Chang and colleagues. TTF-1 demonstrates cytoplasmic immunostaining of hepatocellular carcinomas in 71% of cases, but no nuclear immunostaining. In our opinion, nonnuclear staining is not diagnostically useful in work up of carcinoma of unknown origin.
In the last few years, the specificity of TTF-1 has been challenged. Siami and colleagues and Kubba and colleagues (both from M.D. Anderson Cancer Center) have shown TTF-1 (clone 8G7G3/1) reactivity in tumors of the endocervix, endometrium and ovary; however, the majority of the cases in their study showed only rare and focal staining. Robens and colleagues also showed TTF-1 reactivity in 2.4% of breast carcinomas.
Markers other than TTF-1 that are used for identifying lung carcinoma include surfactant apoprotein (PE10) and napsin A. Surfactant apoprotein is now less commonly used due to its low sensitivity, as it is positive in not more than 50% of lung carcinomas. Napsin’s sensitivity and specificity is similar to TTF-1 but show cytoplasmic reactivity in lung carcinomas. Just like TTF-1, Napsin A has been reported to stain some nonpulmonary carcinomas (renal clear cell, renal papillary, endometrial, hepatocellular carcinomas). In fact, Napsin A is diagnostically used to distinguish müllerian clear cell carcinoma from müllerian serous carcinoma. Müllerian clear cell carcinomas demonstrate cytoplasmic Napsin A in majority of cases. As far as renal carcinomas are concerned, Napsin A reactivity has been reported in up to one-third of renal clear cell carcinoma and two-thirds of renal papillary carcinomas.
Nuclear immunostaining with TTF-1 in all carcinomas of thyroid origin (except anaplastic type).
Nuclear immunostaining with TTF-1 in vast majority of carcinomas of the lung—adenocarcinomas (~70% to 80%), large cell neuroendocrine, and small cell (95%).
Small cell carcinomas of other sites (gastrointestinal tract, bladder, cervix, prostate) are frequently TTF-1+, although they are rare.
Merkel cell carcinomas are negative for TTF-1.
Focal to patchy TTF-1 expression in gynecologic tract tumors.
PE10 lacks sensitivity for identifying lung carcinomas.
Napsin A shows cytoplasmic reactivity in lung carcinomas and has comparable sensitivity and specificity to TTF-1.
Napsin A also stains müllerian clear cell carcinomas and renal carcinomas.
TTF-1 , Thyroid transcription factor-1.
Calretinin and Wilms Tumor Protein-1
Calretinin and Wilms tumor (WT1) protein are two positive mesothelial markers. Calretinin is a 29-kDa intracellular calcium-binding protein that has been described in a variety of cells, including neurons, steroid-producing cells, renal convoluted tubules, eccrine glands, thymic keratinized cells, and mesothelial cells. Ordonez showed the usefulness of calretinin in distinguishing mesotheliomas from pulmonary and nonpulmonary adenocarcinomas. Calretinin may be expressed in 8% of lung adenocarcinomas, but the expression is generally focal and weak. Focal weak expression may also be seen in carcinoma from other sites.
WT1 protein is expressed at high levels in kidney glomeruli, gonadal ridge of developing gonads, Sertoli cells of the testis, and both epithelial and granulosa cells of the ovary, suggesting a developmental role in both the genital system and kidney. WT1 nuclear expression is seen in normal mesothelium and mesothelioma, Wilms tumor (hence its name), desmoplastic small round cell tumor (with antibody to the carboxy-terminal end), and most notably in müllerian epithelial neoplasms (especially tubal/ovarian serous carcinoma— Fig. 8.14 ). The literature regarding endometrial serous carcinoma is contradictory, but it appears that nuclear expression may be seen in approximately 20% of endometrial serous carcinomas. However, the reactivity in endometrial serous carcinoma is often patchy moderate compared with diffuse strong in tubal/ovarian serous carcinomas. WT1 expression is not seen in endometrioid type tumors. Domfeh and colleagues showed weak to moderate WT1 expression in 64% of pure mucinous carcinomas of the breast and in 29% of breast mucinous carcinomas mixed with other subtypes.
Calretinin and WT1 are the two most sensitive mesothelial markers with high specificity.
Calretinin may be expressed in up to 8% of adenocarcinoma from various sites.
Ovarian/tubal serous carcinomas demonstrate diffuse strong WT1 nuclear expression.
The majority of endometrial serous carcinomas are negative for WT1; positive cases generally show patchy staining.
Pure and mixed mucinous breast carcinomas may be WT1+; however, staining is rarely as intense as in serous ovarian/tubal carcinomas.
Gross Cystic Disease Fluid Protein and Mammaglobin
Originally described by Pearlman and colleagues and Haagensen and associates, the prolactin-inducing protein identified by Murphy and coworkers has the same amino acid sequence as GCDFP-15 and is found in abundance in breast cystic fluid and any cell type that has apocrine features. The latter, in addition to breast, includes acinar structures in salivary glands, apocrine glands, sweat glands, Paget disease of skin, vulva, and prostate. Aside from these immunoreactivities, most other carcinomas show no appreciable immunostaining.
The positive predictive value and specificity of GCDFP-15 are both reported to be 99%. The sensitivity for the monoclonal antibody clone D6 (Cambridge Research Laboratories, Cambridge, MA) has been reported to be as high as 74%, but the experience of others has been closer to 40% to 50% and even lower. Similar results are obtained with the use of antibody BRST-2 (Signet Laboratories, Inc., Dedham, Massachusetts).
Because the specificity of GCDFP antibodies for breast carcinoma is so high, it is often used in a screening panel in the appropriate clinical situation, which often turns out to be the presentation of a woman with CUPS or a new lung mass in a patient with a history of breast cancer. Others have demonstrated the utility and specificity of GCDFP-15 antibodies in the distinction of breast carcinoma metastatic in the lung.
The absolute specificity of GCDFP-15 has also been challenged. Striebel and colleagues reported GCDFP-15 immunoreactivity in 5.2% (11/211) of pulmonary adenocarcinomas. These tumors were characteristically of mixed acinar and papillary types with abundant extracellular mucin production. However, 81% of these tumors coexpressed TTF-1, which would be helpful in their distinction from breast carcinomas.
The mammaglobin gene encodes a 93-amino acid protein that is largely confined to breast tissue. Han and colleagues developed antibodies to mammaglobin and found high sensitivity (84.3%) and specificity (85%) for the discrimination of breast carcinoma in lymph nodes. In contrast, the sensitivity and specificity for GCDFP-15 (BRST-2) expression in their study was 44.3% and 97.9%, respectively. Among nonbreast carcinomas, convincing mammaglobin expression is seen in endometrioid carcinomas (~40% cases) and sweat and salivary gland tumors. This nonspecificity of mammaglobin expression in endometrioid adenocarcinomas could be used diagnostically. Caution is advised in interpreting weak or equivocal immunoreactivity with mammaglobin because this pattern of staining can be seen in several nonbreast, nonendometrial carcinomas. With respect to breast carcinoma, mammaglobin is a more sensitive marker than GCDFP-15 ( Fig. 8.15 ). In an immunohistochemical study of nearly 200 consecutive breast carcinomas represented on a tissue microarray (that mimics small biopsy tissue with limited material), Gloyeske and colleagues reported GCDFP-15 reactivity in 26% of cases and mammaglobin reactivity in 52% of cases; 16% of tumors were positive for both GCDFP-15 and mammaglobin, 38% were negative for both GCDFP-15 and mammaglobin, 10% for GCDFP-15 only, and 38% for mammaglobin only. The results support using these stains in a panel, especially when only minimal diagnostic tissue is available for evaluation.
GCDFP-15 has >90% specificity but much lower sensitivity for breast carcinoma.
Mammaglobin has high sensitivity (~60%) for breast carcinoma, but the specificity is lower.
Both GCDFP-15 and mammaglobin should be included in the panel for diagnosing breast carcinoma (GCDFP-15+, mammaglobin+, GATA3+, CK7+/CK20−, ER/PR+, CEA+).
Mammaglobin expression in endometrioid carcinomas could be useful diagnostically.
Hormone Receptors (Estrogen and Progesterone)
Intuitively it would seem as though the estrogen receptor/progesterone receptor (ER/PR) would be confined to hormone-responsive tissues such as breast, but even the recent literature on this topic is controversial. Although some authors conclude that ER/PR is found only in subsets of breast carcinomas and carcinomas of the ovary and endometrium, others have observed mostly ER, and rarely PR, in carcinomas of the lung, stomach, and thyroid.
Vargas and colleagues demonstrated the estrogen-related protein p29 in 98% of non-small cell lung cancers by IHC, suggesting that the estrogen axis may be important in this group of malignancies. In the study by Vargas and associates, these same tumors were all negative with the commercially available antibody ER1D5 (DAKO). Survival of this group of patients differed for men versus women, suggesting some gender-specific p29-associated factor influence.
Dabbs and colleagues observed ER in pulmonary adenocarcinomas using antibody clone 6F11 (Ventana, Tucson, Arizona) with HIER. Nuclear ER was observed in 67% of lung adenocarcinomas, including the bronchioloalveolar variants, but this was not seen with antibody clone ER1D5 (DAKO). Therefore caution is advised in using ER clone 6F11 in isolation, in distinguishing between a primary lung and breast carcinoma. Gomez-Fernandez and colleagues reported focal ER reactivity in pulmonary adenocarcinomas in 7.6% cases with antibody 1D5, 14.1% with clone 6F11, and in 27.2% cases with SP1. In a systematic review of a number of studies evaluating ER/PR expression in tumors, Wei and colleagues concluded that hormone receptor expression can occasionally be seen in nonmammary, nongynecologic tract tumors. Nevertheless, diffuse strong staining for ER (clone 1D5 or rabbit monoclonal SP1) in the right clinical context and appropriate CK expression profile (CK7+/CK20−) is highly indicative of a breast or gynecologic primary tumor.
GATA-Binding Protein 3
GATA3 is a member of the GATA family of zinc finger binding transcription factors. GATA3 function is variable. Pandolfi and colleagues used a mouse model to elucidate GATA3 function. Mice with heterozygous GATA3 mutation were normal, but the mice with homozygous mutations died within 12 days post coitum and showed massive internal bleeding, growth retardation, brain and spinal cord abnormalities, and gross aberrations of fetal liver hematopoiesis. Other studies have shown that GATA3 expression is important in reproductive development and function, including the mammary gland. With respect to tissue expression, numerous immunohistochemical studies have been performed and the results show diffuse and strong nuclear expression predominantly in breast and urothelial bladder carcinomas. In a large systematic analysis of 2500 epithelial and non-epithelial neoplasms, Miettinen and colleagues identified high GATA3 expression not only in breast and urothelial cancers, but also in 98% of basal cell carcinoma (61 of 62 positive), 100% of choriocarcinomas (11 of 11 positive), 100% of yolk sac tumors (6 of 6 positive), 82% of paragangliomas (18 of 22 positive), and 92% of pheochromocytomas (22 of 24 positive). Squamous cell carcinomas from different sites show variable expression with highest seen in cutaneous (81%) and lowest in pulmonary (12%). GATA3 expression is also frequently seen in cutaneous adnexal tumors (100%), mesotheliomas (58%), chromophobe RCC (51%), and salivary gland tumors (43%). Miettinen and colleagues reported GATA3 expression in 37% cases of pancreatic carcinoma but others have found lower expression. In a more focused study, Clark and colleagues confirmed high expression in breast and urothelial carcinomas. Within breast cancers, ER+ tumors show strong and diffuse GATA3 expression in majority of the cases. Although 70% of ER− tumors are also positive for GATA3, the expression is commonly patchy weak to moderate. Gynecologic tumors are generally negative for GATA3 or show focal weak reactivity in up to 20% cases. Upper GI tract and pancreatic-biliary tract carcinomas are also mostly negative with weak expression in up to 10% of cases. Ellis and colleagues evaluated GATA3 expression in conventional ( n = 27) and signet ring cell ( n = 19) bladder adenocarcinomas and compared them with 32 gastric signet ring cell carcinomas. They found diffuse GATA3 reactivity in 41% of bladder signet ring cell carcinomas (SRCC), 7% of conventional bladder adenocarcinomas, and 0% reactivity in gastric signet ring cell adenocarcinomas.
Diffuse strong hormone receptor expression is suggestive of breast or gynecologic primary tumor.
Weak/moderate hormone receptor expression should be judged more carefully, taking into account the clinical presentation and results of other IHC stains.
Use ER and PR stains along with other specific stains and transcription factors for definitive diagnosis.
GATA3 is the most sensitive marker of breast carcinoma.
GATA3 also diffusely expressed in most urothelial bladder carcinomas.
Among nonepithelial tumors, GATA3 is frequently expressed in germ cell tumors and paragangliomas.
Villin is a calcium-dependent actin-binding cytoskeletal protein that is found in the brush border of the intestine and in the proximal renal tubular epithelium. A brush border is characteristic of colorectal carcinomas and is recognized at the ultrastructural level by the presence of microvilli with a dense core of microfilaments, core rootlets, and surface glycocalyx. Up to 33% of pulmonary adenocarcinomas may demonstrate microvillus rootlets by ultrastructure, and their presence correlates closely with villin immunostaining. Antibodies to villin are useful for identifying its molecular presence, which is in almost all colorectal carcinomas and in more than 90% of lung carcinomas that have microvillus rootlets. CKs are a necessary part of a diagnostic panel to distinguish lung and colon carcinomas, and 90% of lung adenocarcinomas are CK20−, whereas colorectal carcinomas are CK20+. Villin may stain hepatocellular neoplasms in a canalicular pattern similar to polyclonal CEA. Villin has also been reported to be expressed in carcinomas of the stomach, pancreas, and gallbladder, as well as in renal clear cell carcinoma and endometrial carcinomas.
CDX2 is a homeobox gene that encodes a transcription protein factor that guides development of intestinal epithelial cells from the region of the duodenum to the rectum. Discovered in 1983, the homeobox gene encodes proteins called homeodomains, which are very important in developmental processes of many multicellular organisms. The homeobox is a conserved DNA motif that encodes proteins that act as transcription factors, controlling the actions of other genes by binding to segments of DNA. The absence of CDX2 is a lethal event in utero, and heterozygotes have GI developmental abnormalities.
Barbareschi and colleagues, using clone CDX2-88, found a very high sensitivity and specificity for detection of colorectal carcinomas, with some CDX2 expression in other adenocarcinomas of the GI tract and in ovarian mucinous tumors. Useful in both paraffin tissue and cytology specimens, they concluded that CDX2 was highly sensitive and specific for intestinal differentiation ( Table 8.8 ). Other studies have confirmed the high specificity of CDX2 for intestinally derived adenocarcinomas, including the stomach, duodenum, gastroesophageal, pancreatic, and biliary tree. Colorectal and duodenal adenocarcinomas also tend to have a diffuse distribution of nuclear staining in a majority of cells, whereas adenocarcinomas from other intestinal sites tend to have staining in a minority of cells. CDX2 expression decreases dramatically in the subset of colon carcinomas that are “minimally differentiated” and are usually associated with mutations of the DNA mismatch repair genes.
Neuroendocrine carcinomas of intestinal derivation showed a focal pattern of nuclear immunostaining in only 42% of cases in one study. In the study by Babareschi and colleagues, well-differentiated neuroendocrine carcinomas of the ileum/appendix showed greatest expression, whereas rectal and upper GI tube tumors showed lower expression. In addition, 39% of neuroendocrine tumors from sites outside the GI tract, including the bladder, breast, uterus, salivary gland, prostate, and lung, showed low expression.
Not surprisingly, urinary bladder adenocarcinomas, derived from the intestinal urachus, are often CDX2+, as are urachal cysts in the bladder ( Fig. 8.16 ). Wang and colleagues studied the immunohistochemical distinction between primary adenocarcinomas of the bladder and secondary involvement of the bladder by colorectal adenocarcinoma. The key antibodies that permitted discrimination of these tumors were beta-catenin (clone 14, Transduction Labs, Lexington, KY), CK7, and TM. All colorectal tumors showed nuclear beta-catenin (bladder negative), were CK7−, and were negative for TM. In contrast, bladder adenocarcinomas were all TM+ and variably positive for CK7.
Importantly, other mucinous neoplasms with morphologic intestinal features, the “colloid” carcinoma of the lung with goblet cells (100%) and a subset of ovarian mucinous carcinomas (64%) are CDX2+. The majority of the colloid lung tumors are TTF-1+, which is a feature that allows distinction from metastatic colorectal mucinous carcinoma. Ovarian mucinous carcinomas may be separated from GI mucinous carcinomas by virtue of typical immunostaining for CK7 in the ovarian tumors.
A prior study has reported uterine cervical adenocarcinomas with intestinal features to be negative for both CDX2 and CK20 ; however, other studies have shown nuclear CDX2 immunoreactivity in up to 30% of cervical adenocarcinomas. This immunoreactivity is seen not only with müllerian mucinous or intestinal mucinous differentiation, but also in endometrioid tumors of the uterine cervix. However, dominant CK7 reactivity is useful in determining gynecologic origin.
CDX2 immunostaining may rarely be seen focally in prostate or thyroid carcinomas. Concomitant use of villin antibody adds specificity for intestinal differentiation. Although some CDX2 may be seen in nonintestinal carcinomas, villin is negative in these tumors. CDX2 does not immunostain liver, hepatocellular carcinoma, or carcinomas of kidney, breast, lung, or salivary gland. The specificity of CDX2 for metastatic colorectal carcinoma in the liver is enhanced by the concomitant use of a CK20+/CK7− profile, because CDX2 may be positive in upper GI carcinomas. Endometrioid carcinomas of the uterus or ovary may mimic colorectal carcinomas, and they may demonstrate nuclear CDX2, in which case a panel of antibodies that includes CK7, CK20, PAX8, villin, vimentin, and ER would be needed to discriminate from colorectal carcinomas ( Fig. 8.17 ).
The special AT-rich sequence binding protein 2 (SATB2) binds to specific nuclear matrix attachment regions and is involved in transcription regulation and chromatin remodeling. Immunohistochemical studies have shown SATB2 expression in wide variety of tissues and tumors, but the diagnostic application appears to be limited to GI lesions and bone tumors. In a carcinoma of unknown origin, SATB2 nuclear expression is supportive of a lower GI tract primary. In a study of 32 primary appendiceal mucinous tumors and 40 ovarian mucinous neoplasms, Strickland and colleagues identified SATB2 expression in 93.8% of appendiceal tumors and 2% of ovarian tumors. In a study of primary and metastatic tumors in the ovary, Moh and colleagues identified SATB2 staining in 0 of 22 mucinous cystadenomas, 4 of 12 mucinous cystadenomas associated with mature teratoma, 1 of 60 mucinous borderline tumors, 0 of 17 mucinous adenocarcinomas, 0 of 3 endometrioid borderline tumors, 0 of 72 endometrioid adenocarcinomas, 24 of 32 (75%) colorectal adenocarcinomas, 8 of 10 (80%) low-grade appendiceal mucinous neoplasms, and 4 of 4 (100%) high-grade appendiceal adenocarcinomas. Other metastatic tumors were negative for SATB2 (pancreatic, gastric, gallbladder, or endocervical origin). The authors concluded that that ovarian tumors with mucinous or endometrioid features that express SATB2 are unlikely to be of primary ovarian origin. Similar findings were reported by Perez-Montiel in a study of 106 mucinous tumors in the ovary. Our unpublished data also show SATB2 reactivity predominantly in colorectal and appendiceal tumors but only weak reactivity in a small number of müllerian, upper GI and pancreatic-biliary tract tumors. Because most mucinous tumors of gynecologic origin are negative for SATB2, the antibody can be helpful in the differential diagnosis of mucinous tumor in the ovary.
Apart from mucinous tumors of lower GI tract, SATB2 expression has also been identified in Merkel cell carcinoma. Fukuhara and colleagues reported SATB2 expression in 75% cases of Merkel cell carcinoma. In bone and soft tissue lesions, SATB2 nuclear expression is indicative of osteoblastic differentiation. It can be used to distinguish between hyalinized collagen and osteoid. However, SATB2 is unable to distinguish between chondrosarcoma and chondroblastic osteosarcoma. In a study of 42 osteosarcomas, 31 chondrosarcoma, and 371 genetically confirmed Ewing sarcoma, SATB2 was positive in 90.4% of osteosarcomas, 46.6% of chondrosarcomas, and in only 1.3% of Ewing sarcomas.