Immunohistochemistry (IHC) is the use of antibody-based reagents for localization of specific epitopes in tissue sections. Over the past several decades, IHC has become a powerful tool to assist the surgical pathologist in many clinically critical settings. It is important to recognize that IHC has two components, each with its own strengths and weaknesses. These components may be thought of as the “hardware” (i.e., antibodies, detection systems) and the “software” (i.e., analytic processes). No matter how selective the antibodies or how powerful the detection system, the method fails if the analytic tools are inadequate. This chapter focuses on the antibody component of the hardware and some of the analytic processes in which IHC assists in the diagnosis of soft tissue neoplasms.
It cannot be overemphasized that IHC is an adjunctive diagnostic technique to traditional morphologic methods in soft tissue pathology, as in any other area of surgical pathology. It is critical to recognize that the diagnosis of many soft tissue tumors does not require IHC (e.g., osteocartilaginous tumors), and that there are no markers or combinations of markers that will reliably distinguish benign from malignant tumors (e.g., nodular fasciitis from leiomyosarcoma). Furthermore, reliable specific markers do not yet exist for certain mesenchymal cell types and their tumors, and a subset of soft tissue tumors is better defined by the tumor’s molecular rather than immunophenotypic profile, and techniques other than IHC, such as cytogenetic or molecular genetic studies, may prove more valuable in this setting. Lastly, it is important to acknowledge that a subset of soft tissue tumors defy classification, even with exhaustive IHC or genetic study.
The expression of certain antigens, or clusters of antigens, is characteristic of some tumors. Whereas thousands of monoclonal and polyclonal antibodies are available to assist in tumor diagnosis, only a small subset has proved to be of practical value in the diagnosis of soft tissue neoplasms. Tables 6.1 and 6.2 present an overview of the markers discussed in the following sections; the question marks highlight the gaps in our understanding of the cellular biology of many soft tissue tumors.
|Antibodies to:||Expressed by:|
|Keratins||Carcinomas, epithelioid sarcoma, synovial sarcoma, some angiosarcomas and leiomyosarcomas, mesothelioma, rhabdoid tumor|
|Vimentin||Sarcomas, melanoma, lymphoma, some carcinomas|
|Desmin||Benign and malignant smooth and skeletal muscle tumors|
|Glial fibrillary acidic protein||Gliomas, some schwannomas|
|Neurofilaments||Neural and neuroblastic tumors|
|Smooth and skeletal muscle actins (HHF35)||Benign and malignant smooth and skeletal muscle tumors; myofibroblastic tumors and pseudotumors|
|Smooth muscle actin (1A4)||Benign and malignant smooth muscle tumors; myofibroblastic tumors and pseudotumors|
|Caldesmon||Smooth muscle tumors, glomus tumors|
|Myogenic nuclear regulatory proteins (myogenin, MyoD1)||Rhabdomyosarcoma|
|S-100 protein||Melanoma, benign and malignant peripheral nerve sheath tumors, cartilaginous tumors, normal adipose tissue, Langerhans cells, many others|
|SOX10||Melanoma, benign and malignant peripheral nerve sheath tumors, myoepithelial tumors|
|Epithelial membrane antigen||Carcinomas, epithelioid sarcoma, synovial sarcoma, perineurioma, meningioma, anaplastic large cell lymphoma|
|CD31||Benign and malignant vascular tumors|
|Von Willebrand factor (factor VIII–related protein)||Benign and malignant vascular tumors|
|CD34||Benign and malignant vascular tumors, solitary fibrous tumor, epithelioid sarcoma, dermatofibrosarcoma protuberans, GIST|
|CD99||Ewing sarcoma/primitive neuroectodermal tumor, some rhabdomyosarcomas, some synovial sarcomas, lymphoblastic lymphoma, mesenchymal chondrosarcoma, small cell osteosarcoma, many others|
|MUC4||Low-grade fibromyxoid sarcoma|
|CD30||Anaplastic large cell lymphoma, embryonal carcinoma|
|CD68 and CD163||Macrophages, fibrohistiocytic tumors, granular cell tumors, various sarcomas, melanomas, carcinomas|
|Melanosome-specific antigens (HMB-45, melan-A, tyrosinase, microphthalmia transcription factor)||Melanoma, PEComa, clear cell sarcoma, melanotic schwannoma|
|MDM2/CDK4||Atypical lipomatous tumor and dedifferentiated liposarcoma|
|Claudin-1||Perineurioma, synovial sarcoma, epithelioid sarcoma|
|GLUT-1||Perineurioma, infantile hemangioma|
|SMARCB1 (INI1)||Expression lost in extrarenal rhabdoid tumor, epithelioid sarcoma, some epithelioid MPNSTs, and extraskeletal myxoid chondrosarcomas|
|CD117 and DOG1||GIST|
|Tumor Type||Normal Cell Counterpart||Useful Markers|
|Angiosarcoma||Endothelium||CD31, CD34, FLI-1, ERG, von Willebrand factor|
|Kaposi sarcoma||Endothelium||CD31, CD34, VEGFR3, podoplanin, HHV8 LANA|
|Epithelioid sarcoma–like (pseudomyogenic) hemangioendothelioma||Endothelium||FOSB|
|Leiomyosarcoma||Smooth muscle||Muscle (smooth) actins, desmin, caldesmon, myosin heavy chain|
|Rhabdomyosarcoma||Skeletal muscle||MyoD1, myogenin; muscle (sarcomeric) actins; desmin|
|Ewing sarcoma||?||CD99, FLI-1, ERG, NKX2.2|
|Ewing-like CIC -rearranged sarcoma||?||ETV4, WT-1|
|Ewing-like BCOR -rearranged sarcoma||?||BCOR, CCNB3, SATB2|
|Synovial sarcoma||?||Keratin, EMA, TLE1|
|Epithelioid sarcoma||?||Keratin, CD34, SMARCB1 (INI1) loss|
|Malignant peripheral nerve sheath tumor||Nerve sheath (e.g., Schwann cell, perineurial cell)||S-100, EMA, claudin-1, GLUT-1, SOX10, H3K27me3 loss|
|Malignant melanotic schwannian tumor||?Schwann cell||S100 protein, SOX10, melanoma markers, loss of PRKAR1A|
|Liposarcoma||Adipocyte||S-100 protein, MDM2, CDK4|
|Osteogenic sarcoma||Osteocyte||SATB2, osteocalcin|
|Myofibroblastic lesions (e.g., nodular fasciitis)||Myofibroblast||Smooth muscle actins|
|Gastrointestinal stromal tumor||Interstitial cells of Cajal||CD117a (c-kit), CD34, DOG1|
|Solitary fibrous tumor||?||CD34, STAT6|
|Glomus tumors||Glomus cell||Smooth muscle actins, caldesmon, type IV collagen|
|Angiomatoid (malignant) fibrous histiocytoma||?||Desmin, EMA, CD68|
|Alveolar soft part sarcoma||?||TFE3|
|Perivascular epithelioid cell neoplasms (PEComas)||?||Smooth muscle actins, melanocytic markers|
The intermediate filaments comprise the major component of the cytoskeleton and consist of five major subgroups—vimentin, keratins, desmin, neurofilaments, and glial fibrillary acidic protein (GFAP)—and a small number of minor subgroups (e.g., nestin, peripherin). Ultrastructurally, the intermediate filaments appear as wavy unbranched filaments that often occupy a perinuclear location in the cell. The original thinking that intermediate filament expression was restricted to specific cell types (e.g., keratins in carcinomas, vimentin in sarcomas) is now well recognized as an oversimplification. The following sections on intermediate filaments concentrate not only on the normal pattern of expression of these proteins, but also on the situations where intermediate filaments show “anomalous expression.”
Vimentin, a 57-kilodalton (kDa) intermediate filament protein, is expressed in all mesenchymal cells. Vimentin is ubiquitously expressed in all cells during early embryogenesis and is gradually replaced in many cells by type-specific intermediate filaments. In some mesenchymal tissues, vimentin is typically coexpressed along with the cell type–specific intermediate filaments (e.g., desmin and vimentin coexpression in muscle cells, vimentin and GFAP in some Schwann cells). Vimentin is expressed in virtually all mesenchymal tumors and is thus of minimal value in identifying particular tumors. Given the frequent coexpression of vimentin along with keratin in carcinomas, vimentin expression is also of little value in the immunohistochemical distinction of carcinomas from sarcomas. Vimentin immunoreactivity has been touted as a good marker of tissue preservation. However, vimentin expression, similar to that of all the intermediate filaments, is rather hardy and may remain present in tissues in which all other immunoreactivity has been lost. The absence of vimentin expression may occasionally be a clue to the diagnosis of rare vimentin-negative mesenchymal tumors, such as alveolar soft part sarcoma and perivascular epithelioid cell neoplasms. In general, there is no value in performing vimentin immunostains on any spindle cell neoplasm.
Keratins, also known as cytokeratins , the most complex members of the intermediate filament protein family, are a collection of more than 20 proteins. The keratins may be grouped by their molecular weights (40–67 kDa) into acidic and basic subfamilies or by their usual pattern of expression in simple or complex epithelium ( Fig. 6.1 ). In practice the keratins are most often seen as low-molecular-weight (LMW) keratins (generally keratins 8, 18, and 19) and high-molecular-weight (HMW) keratins (generally keratins 1, 5, 10, and 14). Keratins are highly sensitive markers for identifying carcinomas and are generally employed as markers distinguishing epithelial/mesothelial from nonepithelial tumors (i.e., lymphomas, sarcomas, melanomas) ( Fig. 6.2 ). Over the past decade it has become clear that keratin expression is not restricted to carcinomas.
Sarcomas with “True” Epithelial Differentiation: Epithelioid Sarcoma and Synovial Sarcoma
Among the sarcomas are two patterns of keratin expression. A small subset of sarcomas display true epithelial differentiation, as defined by usual expression of keratin and other epithelial proteins, such as the desmoplakins and occludin (e.g., synovial sarcomas, epithelioid sarcomas). A larger group of tumors occasionally display “anomalous” keratin expression (i.e., keratin expression by cells and tumors without true epithelial differentiation). Synovial sarcomas and epithelioid sarcomas are the best, if not the only, examples of sarcomas manifesting true epithelial differentiation ( Fig. 6.3 ). Expression of both LMW and HMW keratin isoforms is seen in both synovial sarcoma and epithelioid sarcoma, confirming the presence of true epithelial differentiation. Antibodies to specific keratins, such as keratins 7 and 19 for synovial sarcoma and keratins 5/6 in epithelioid sarcoma, may also be diagnostically useful in select cases.
Anomalous Keratin Expression
Anomalous keratin expression is typically characterized by immunostaining (even under optimal technical conditions) in only a subset of the target cell population. In these cells, keratin is present in only a portion of the cytoplasm, often yielding a “perinuclear” or “dotlike” pattern of immunostaining. This dotlike pattern is not always an indication of anomalous keratin, however, because it is typically seen in some neuroendocrine carcinomas (e.g., small cell carcinomas, Merkel cell tumors) and extrarenal rhabdoid tumors ( Fig. 6.4 ). In addition, it is rare to find keratins other than those corresponding to the Moll catalog 8 and 18 (corresponding to positivity with antibodies CAM5.2 or 35βH11) in tumors manifesting anomalous keratin expression.
Contrary to earlier suggestions, anomalous keratin expression is not a universal feature of sarcomas. It is instead a feature of a limited subset of nonepithelial tumors, particularly smooth muscle tumors, melanomas, and endothelial cell tumors; as such, it may serve as a clue to the diagnosis of these tumors. Interestingly, the normal cell counterparts of some of these tumors (i.e., smooth muscle cells and endothelial cells) have been found to express keratins in nonmammalian species, and in our experience these are frequently keratin positive in routinely processed sections, particularly with newer, more sensitive antibodies, such as the OSCAR wide-spectrum keratin antibody.
Smooth Muscle Cells and Smooth Muscle Tumors
Frozen sections of the smooth muscle cell–rich myometrium of the uterus (along with myocardial cells) were first reported to “react with” various antikeratin antibodies. Brown et al. and Norton et al. verified these findings using slightly different techniques, although Norton failed to find corroborative biochemical evidence of keratin expression by these smooth muscle cells. Gown et al. first presented biochemical documentation of true “anomalous” keratin expression of keratins 8 and 18, in which immunostaining was corroborated by Western blots; it was further documented by two-color immunofluorescence studies of myometrial smooth muscle cells grown in vitro. Subsequent studies have shown that at least 30% of leiomyosarcomas manifest keratin.
Despite that many studies completed during the mid-1980s concluded that melanomas were vimentin-positive, keratin-negative tumors, Zarbo et al. first confirmed the keratin positivity of many melanomas and demonstrated the positive immunostaining as a function of tissue preparation and fixation, with 21% of cases positive in frozen sections but far fewer in formalin-fixed, paraffin-embedded (FFPE) sections. They also performed one- and two-dimensional gel electrophoresis with immunoblotting, confirming that keratin 8 was expressed by the tumor cell population. More recently, Romano et al. documented keratin expression, using the OSCAR and AE1/AE3 antibodies, in up to 40% of epithelioid melanomas and in a smaller percentage of spindle cell melanomas. Although anomalous keratin expression was previously thought to be more common in metastatic than in primary melanomas, this does not seem to be the case.
Early reports suggested that vascular tumors manifesting epithelioid histologic features (e.g., epithelioid hemangioendothelioma, epithelioid angiosarcoma) express keratin in most cases ( Fig. 6.5 ). The largest published series of angiosarcomas of deep soft tissue documented keratin expression in about one-third of cases. Aberrant expression of keratins is essentially always present in epithelioid sarcoma–like (pseudomyogenic) hemangioendotheliomas.
Small Blue Round Cell Tumors
A surprising number of tumors in the category of “small, blue round cell tumors” of childhood typically coexpress keratin in a pattern similar to that of anomalous keratin expression. These tumors include Ewing sarcoma, rhabdomyosarcoma, Wilms tumor, and desmoplastic small round cell tumor of childhood. Expression of LMW keratin isoforms may be seen in almost 25% of Ewing sarcomas, usually confined to less than 20% of the neoplastic cells. Expression of HMW keratins is much less common and is restricted to the rare “adamantinoma-like” variant.
Keratin Expression in Other Sarcomas
The literature is replete with reports of keratin expression in other sarcomas, including undifferentiated pleomorphic sarcoma, chondrosarcoma, osteosarcoma, and malignant peripheral nerve sheath tumors. Nonetheless, in our experience, keratin expression in these tumors is exceedingly rare. Aberrant expression of keratins is typically seen in the recently described superficial CD34-positive fibroblastic tumor, where it is an important clue to the correct diagnosis.
It is important to remember that immunoreactivity and true antigen expression are not necessarily synonymous. Several factors can theoretically account for positive keratin immunostaining in tumors without true keratin expression. This includes the use of antibodies at inappropriately high concentrations, potentially altered specificities following the use of heat-induced epitope retrieval techniques (“antigen retrieval”), and cross-reactivity of antikeratin antibodies with other proteins, such as GFAP in gliomas and some schwannomas. Punctate keratin staining from exogenous keratin (dandruff) is an additional pitfall ( Fig. 6.6 ), as is the presence of reactive submesothelial fibroblasts in or around many tumor types ( Fig. 6.7 ). By employing an approach to IHC that includes a panel of antibodies, the pathologist can generally avoid misinterpretation that might result from “misbehavior” of one antibody.
Epithelial Membrane Antigen
Epithelial membrane antigen (EMA) is an incompletely characterized antigen that is present in a group of carbohydrate-rich, protein-poor, HMW molecules present on the surface of many normal types of epithelium, including those in the pancreas, stomach, intestine, salivary gland, respiratory tract, urinary tract, and breast. Among normal mesenchymal cells, EMA expression is limited to perineurial cells and meningeal cells. Uses for EMA are limited in sarcoma diagnosis. EMA expression is a more sensitive, but less specific, marker of poorly differentiated synovial sarcomas; it may be helpful in cases with only focal (or absent) keratin expression. Perineuriomas and malignant peripheral nerve sheath tumors with perineurial differentiation are characterized by a sometimes subtle expression of EMA along cell processes, as well as claudin-1, GLUT-1, and type IV collagen expression ( Fig. 6.8 ). Ectopic meningiomas, as with their meningeal counterparts, are characterized by EMA and vimentin expression in the absence of keratin expression. Patchy expression of EMA (along with desmin and CD68) is seen in approximately 50% of angiomatoid fibrous histiocytomas. EMA expression has been documented in a significant subset of genetically confirmed low-grade fibromyxoid sarcomas, a potentially serious pitfall, given the morphologic similarities between these tumors and perineuriomas. In our experience, patchy EMA immunoreactivity may be seen in a fairly broad range of mesenchymal tumors, emphasizing the need to employ antibodies to EMA as part of a panel of immunostains.
Markers of Muscle Differentiation
There are three types of muscle differentiation. The first is skeletal muscle differentiation, as recapitulated in rhabdomyoma and rhabdomyosarcoma. The second is “true” smooth muscle differentiation, reflected in leiomyoma and leiomyosarcoma. The third is “partial” smooth muscle differentiation, as seen in the myofibroblasts that constitute a significant population of cells in healing wounds and the stromal reaction to tumors. There is also a subset of soft tissue tumors (e.g., nodular fasciitis, myofibroblastoma), the phenotype of which greatly resembles myofibroblasts rather than true smooth muscle cells. The principal markers of muscle differentiation are the intermediate filament desmin, the various actin isoforms, caldesmon, and the myogenic regulatory proteins.
Desmin is the intermediate filament protein associated with both smooth and skeletal muscle differentiation; it is rarely expressed by myofibroblasts and their corresponding tumors. In skeletal muscle, desmin is localized to the Z zone between the myofibrils, where it presumably serves as binding material for the contractile apparatus. In smooth muscle, desmin is associated with cytoplasmic dense bodies and subplasmalemmal dense plaques. Desmin may also be expressed by nonmuscle cells, including the fibroblastic reticulum cell of the lymph node, the submesothelial fibroblast, and endometrial stromal cells. Desmin is among the earliest muscle structural genes expressed in the myotome of embryos, and some regard it as the best single marker for the diagnosis of poorly differentiated rhabdomyosarcoma. Although the early literature on desmin questioned its sensitivity in formalin-fixed, deparaffinized tissue sections, more recent studies have confirmed its excellent sensitivity. In our experience, with the use of heat-induced epitope retrieval techniques and modern antibodies such as D33, desmin is the most sensitive marker of skeletal and smooth muscle differentiation in terms of both the fraction of tumors so identified and the fraction of tumor cells in given tumors that are positive. Desmin expression is present in almost 100% of rhabdomyosarcomas of all subtypes, including very poorly differentiated ones ( Fig. 6.9 ).
Desmin expression is apparently not as specific for muscle tumors as originally thought, because it has also been described in Ewing sarcoma, desmoplastic small round cell tumors, neuroblastoma, mesothelial cells and tumors, the blastemal component of Wilms tumor, giant cell tumors of the tendon sheath, and ossifying fibromyxoid tumors of soft parts; in none of these contexts is there thought to be true muscle differentiation ( Fig. 6.10 ). Desmin expression, along with more limited expression of MyoD1 and myogenin, is a surprisingly common finding in mesenchymal chondrosarcoma. Expression of desmin along with EMA and CD68, in the absence of other muscle markers, is highly characteristic of angiomatoid fibrous histiocytomas ( Fig. 6.11 ). Desmin expression may be seen in up to 24% of melanomas and may rarely be seen in schwannomas. Definitive identification of tumors manifesting skeletal muscle differentiation therefore may require the additional use of antibodies to myogenic transcription factors such as myogenin and MyoD1 (see later).
Actin, a ubiquitous protein, is expressed by all cell types; high concentrations of actins and unique isoforms, however, help make actin a marker of muscle differentiation. In general, actins can be grouped into muscle and nonmuscle isoforms, which differ by only a few amino acids in a protein with a molecular weight of 43,000 kDa. It has nevertheless been possible to generate antibodies specific to muscle actins versus nonmuscle actins and to specific actin isotypes with respect to the various muscle types (e.g., smooth muscle vs. skeletal muscle). Although early literature cites the “specificity” of antiactin polyclonal antibodies for muscle cells, in most of these studies it is a quantitative rather than a qualitative phenomenon; that is, muscle cells have much more actin than many other cells, and demonstration of positivity is determined on this basis alone. Whereas there are monoclonal antibodies that can identify all actin isoforms (i.e., the C4 clone ), given the sensitive immunohistochemical techniques available, this antibody cannot be used to distinguish muscle from nonmuscle actins. The antibody HHF35, which has been widely used to identify muscle cells and tumors, displays specificity for all muscle (vs. nonmuscle) actins. Antibody 1A4 is a monoclonal antibody that specifically identifies smooth muscle actin isoforms and thus can distinguish smooth from skeletal muscle cells and tumors.
Smooth muscle actin isoforms are also expressed by myofibroblasts, and the characteristic pattern of actin expression in these cells may help distinguish them from true smooth muscle cells. In general, myofibroblasts show expression of smooth muscle actin only at the periphery of their cytoplasm (“tram-track” pattern), in contrast to the uniform cytoplasmic expression in smooth muscle ( Fig. 6.12A ). On occasion, this tram-track pattern is a clue that a myofibroblastic process is present (e.g., fasciitis) rather than a leiomyosarcoma. For unknown reasons, the 1A4 smooth muscle actin antibody occasionally shows aberrant nuclear immunoreactivity, which should not be interpreted as evidence of smooth muscle actin expression in a tumor ( Fig. 6.12B ). Antibody asr-1 is monoclonal and specifically identifies sarcomeric actins (skeletal, cardiac); it identifies rhabdomyosarcoma but not leiomyosarcoma. One should be aware that some rhabdomyosarcomas, particularly the paratesticular spindle cell type, can express low levels of smooth muscle actins.
Heavy caldesmon (h-caldesmon) is a calcium-binding protein involved in the regulation of smooth muscle contractility. The sensitivity of h-caldesmon for smooth muscle tumors is lower than that of antibodies to smooth muscle actin, but h-caldesmon is not expressed by myofibroblastic tumors. H-caldesmon expression is frequently present in glomus tumors and in myopericytic neoplasms of various types ( Fig. 6.13 ) . h-Caldesmon is generally best used for the distinction of poorly differentiated smooth muscle tumors from myofibroblastic tumors, rather than as a stand-alone marker of smooth muscle cells.
Myogenic Transcription Factors
Myogenic regulatory proteins (i.e., transcription factors of the MyoD [myogenic determination] family) play a critical role in the commitment and differentiation of mesenchymal progenitor cells to the myogenic lineage and subsequent maintenance of the skeletal muscle phenotype. MyoD1 and myogenin are members of the basic helix-loop-helix family of DNA-binding myogenic nuclear regulatory proteins; the other members include Myf5 and MRF4. These genes encode transcription factors, whose introduction into nonmuscle cells in culture can initiate muscle-specific gene expression and muscle differentiation. In addition, such regulatory factors are expressed much earlier in the normal skeletal muscle differentiation program than structural proteins such as desmin, actin, and myosin; indeed, expression of these myogenic regulatory proteins leads to activation of the latter. Antibodies to both MyoD1 and myogenin, but not the other myogenic nuclear regulatory proteins, have been studied in terms of diagnosing rhabdomyosarcoma. Both MyoD1 and myogenin are expressed in more than 90% of rhabdomyosarcomas of all subtypes. Antibodies to both MyoD1 and myogenin show excellent specificity (see Fig. 6.9C ). There is only a single report of nuclear immunoreactivity for MyoD1 in FFPE sections in a pleomorphic liposarcoma. Four alveolar soft part sarcomas have been demonstrated to express MyoD1 by IHC on frozen sections and by Western blot. Cytoplasmic immunoreactivity for MyoD1 has been reported in a small number of nonrhabdomyosarcomas, including primitive neuroectodermal tumor, Wilms tumor, and undifferentiated sarcoma. Only nuclear immunoreactivity for MyoD1 should be taken as evidence of skeletal muscle differentiation, because the epitope recognized by the most commonly used antibody to MyoD1, 5.8A, includes amino acid sequences with close homology to the class 1 major histocompatibility antigen and transcription factors E2A and ITF-1, 123 suggesting that cytoplasmic immunoreactivity may represent a cross-reaction rather than true MyoD1 expression.
As previously noted, both MyoD1 and myogenin are expressed by more than 95% of embryonal rhabdomyosarcomas (ERMS) and alveolar rhabdomyosarcomas (ARMS), including the well-differentiated spindle cell variant of ERMS and the solid variant of ARMS. In general, ARMS express very high levels of myogenin and comparatively less MyoD1, whereas ERMS show the opposite pattern, or equal levels of expression. The recently described spindle cell/sclerosing variant of RMS typically shows very strong expression of MyoD1, but much more limited myogenin expression. Pleomorphic rhabdomyosarcoma are less frequently MyoD1 or myogenin positive, with a recent large series documenting expression in only 53% and 56% of cases, respectively. In addition, pleomorphic rhabdomyosarcoma may show only a small percentage of tumor cells that are positive.
Although the available evidence appears to strongly support the view that MyoD1 and myogenin expression are highly specific for rhabdomyoblastic differentiation, it is important to realize that their expression does not obligate a diagnosis of rhabdomyosarcoma. In our experience and that of others, expression of MyoD1 and/or myogenin may be seen in a variety of rare tumors with rhabdomyoblastic differentiation, including Wilms tumors with myogenous differentiation, neuroendocrine carcinoma (including Merkel cell carcinoma) with rhabdomyoblastic differentiation, malignant glial tumors with myoblastic differentiation, malignant peripheral nerve sheath tumors with rhabdomyoblastic differentiation (malignant Triton tumor), and teratomas with rhabdomyoblastic differentiation. Expression of both MyoD1 and myogenin has recently been demonstrated in mesenchymal chondrosarcomas, likely indicative of true skeletal muscle differentiation in this tumor secondary to the HEY1-NCOA2 fusion ( Fig. 6.14 ). It is also important not to mistake myogenin and MyoD1 expression in degenerating and regenerating skeletal muscle for tumor cell expression, particularly in the setting of diffuse skeletal muscle infiltration by a nonmyogenous “small blue round cell tumor,” such as lymphoma.
PAX7 is a paired-box transcription factor essential for the developmental specification of skeletal muscle satellite cells ( Fig. 6.15A and B ). Recently, antibodies to PAX7 have been shown to be sensitive markers of embryonal, spindle cell/sclerosing, and pleomorphic rhabdomyosarcomas, present in more than 70% of cases ( Fig. 6.15C and D ). Interestingly, PAX7 expression is much less common in alveolar rhabdomyosarcoma (19%) and is confined to tumors harboring the PAX3-FOXO1A fusion. PAX7 expression has also been demonstrated to be present in all Ewing sarcoma, but not in other nonrhabdomyosarcoma small round cell tumors.
Myoglobin and Other Less Commonly Used Markers
Antibodies to myoglobin, an oxygen-binding heme protein found in skeletal and cardiac muscle but not smooth muscle, were the first markers used in the immunohistochemical diagnosis of rhabdomyosarcoma. Unfortunately, myoglobin is present in demonstrable amounts in fewer than 50% of rhabdomyosarcomas; it may be identified in nonmyogenous tumor cells that are infiltrating skeletal muscle and phagocytosing myoglobin. Commercially available myoglobin antibodies have a high level of nonspecific, “background” staining, which may be difficult to distinguish from true myoglobin expression. This is in distinct contrast to desmin and the myogenic regulatory protein, which do not diffuse. We do not use antibodies to myoglobin in our routine practice. Other muscle markers that have been used for diagnosing rhabdomyosarcoma include antibodies to myosin, creatine kinase subunit M, and titin. In general, these alternative markers suffer from a lack of sensitivity and/or specificity, and their use cannot be recommended.
Recommendations for Use of Muscle Markers
In summary, for identifying skeletal muscle differentiation, the myogenic regulatory proteins myogenin and MyoD1 are the most specific; antibodies to desmin and muscle actins (i.e., HHF35) are of high sensitivity but are not skeletal muscle specific. For identification of smooth muscle differentiation (e.g., in leiomyosarcomas), antibodies to desmin and muscle actins (i.e., antibody HHF35) or smooth muscle α-actin (e.g., antibody 1A4) are the best markers of smooth muscle differentiation. For identifying myofibroblasts (e.g., the type of differentiation present in lesions such as nodular fasciitis), antibodies to desmin are useful only for distinguishing myofibroblasts from true smooth muscle cells, because the former (in contrast to the latter) generally do not express desmin. Both cell types express smooth muscle actins, however, although myofibroblasts generally express the latter in a characteristic wispy or tram-track pattern of immunostaining that, on higher resolution, can be demonstrated to correspond to the peripheral bundles of actin filaments, which are the hallmark of this cell type. Myofibroblasts also manifest little or no expression of the smooth muscle and myoepithelial cell-associated proteins caldesmon and smooth muscle myosin heavy chain, and these markers may help to distinguish myofibroblastic and true smooth muscle proliferations.
Markers of Nerve Sheath Differentiation
The S-100 protein is a 20-kDa acidic calcium-binding protein, so named for its solubility in 100% ammonium sulfate. The protein is composed of two subunits, α and β, which combine to form three isotypes. The α-α isotype is normally found in myocardium, skeletal muscle, and neurons; the α-β isotype is present in melanocytes, glia, chondrocytes, and skin adnexa; and the β-β isotype is seen in Langerhans and Schwann cells.
Immunohistochemically, S-100 protein can be demonstrated in a large number of normal tissues, including some neurons and glia; Schwann cells; melanocytes; Langerhans cells; interdigitating reticulum cells of lymph nodes; chondrocytes; myoepithelial cells and ducts of sweat glands, salivary glands, and the breast; serous glands of the lung; fetal neuroblasts; and sustentacular cells of the adrenal medulla and paraganglia ( Figs. 6.16 and 6.17 ). In the immunohistochemical diagnosis of soft tissue neoplasms, S-100 protein is of most value as a marker of benign and malignant nerve sheath tumors and melanoma. S-100 protein is strongly and uniformly expressed in essentially all schwannomas. The finding of uniform S-100 immunoreactivity may be a valuable clue to the diagnosis of cellular schwannoma, because malignant peripheral nerve sheath tumors usually show only patchy, weak expression of S-100, 60,153,156,157 and fibrosarcomas would not be expected to be S-100 positive ( Fig. 6.18 ). S-100 protein expression is much more variable in neurofibromas than in schwannomas. S-100 protein expression is seen in 40% to 80% of malignant peripheral nerve sheath tumors.
However, as may be inferred from the long list of normal tissues that express this protein, significant S-100 protein expression may be seen in a subset of nonneural tumors included in the differential diagnosis of malignant peripheral nerve sheath tumors: synovial sarcoma, rhabdomyosarcoma, leiomyosarcoma, and myoepithelioma. Other tumors that may express S-100 protein include adipocytic tumors, chondrocytic tumors, ossifying fibromyxoid tumors, and chordoma. Malignant melanomas of all types, including the desmoplastic and sarcomatoid variants, are almost always strongly positive for S-100 protein. Uniform, strong S-100 protein expression may be a valuable clue that a melanoma is present rather than a malignant peripheral nerve sheath tumor of skin or soft tissue because, as noted, S-100 protein expression in malignant peripheral nerve sheath tumors tends to be weaker and more patchy. Our experience is that approximately 2% to 3% of melanomas (more often in the metastatic setting) are negative for S-100 protein; additional immunostaining for a melanosome-specific marker such as gp100 protein (identified by antibody HMB-45 166 or melan-A ) is essential for arriving at the correct diagnoses in these patients.
SOX10 is a transcription factor involved in neural crest development and differentiation of neural crest cells into melanocytic and schwannian lineages. Nonaka et al. first demonstrated SOX10 expression by IHC in melanocytic and schwannian neoplasms, noting expression in greater than 85% of melanomas of all subtypes, 60% of clear cell sarcomas, more than 93% of neurofibromas of all subtypes, 100% of schwannomas, and 30% of malignant peripheral nerve sheath tumors ( Fig. 6.19 ). They also noted SOX10 expression in sustentacular cells of various neuroendocrine tumors, but in no other epithelial or mesenchymal tumor studied. Karamchandani et al. reported similar findings. In general, expression of SOX10 parallels that of S100 protein in melanomas and malignant peripheral nerve sheath tumors. SOX10 expression is also common in myoepithelial tumors. In general, other S100 protein-positive tumors, such as ossifying fibromyxoid tumor and chordoma, are SOX10 negative.
The claudins are a family of approximately 20 homologous proteins that help determine tight junction structure and permeability. Claudins appear to be differentially expressed in tissues, with claudin-1 expression relatively widespread among epithelia and claudin-3 expression confined to lung and liver epithelia. Claudins are integral transmembrane proteins that complex with other transmembrane proteins such as junctional adhesion molecule (JAM) and occludin and interact with scaffolding proteins such as ZO-1, ZO-2, and ZO-3. Among normal mesenchymal tissues, claudin-1 expression appears to be limited to perineurial cells. In the appropriate histologic context, claudin-1 is a useful marker of perineuriomas, present in 20% to 90% of perineuriomas, but not in other tumors in this differential diagnosis, such as neurofibromas, schwannomas, low-grade fibromyxoid sarcoma, desmoplastic fibroblastoma, dermatofibrosarcoma protuberans, or fibromatosis (see Fig. 6.8 ). Aberrant, nonpolarized expression of claudin-1 and other tight junction–related proteins is seen in a significant number of synovial sarcomas and Ewing sarcoma/primitive neuroectodermal tumor (ES/PNET).
GLUT-1 is the erythrocyte-type glucose transporter protein, which plays a particular role in transporting glucose across epithelial and endothelial barrier tissues. Expression of GLUT-1 protein has been shown to be a consistent feature of normal perineurial cells and benign and malignant perineurial tumors. However, in soft tissue and bone tumors, GLUT-1 expression is frequently identified adjacent to foci of necrosis, presumably representing upregulation of this protein within hypoxic zones, secondary to upstream activation of proteins such as hypoxia-inducible factor 1α, and thus GLUT-1 is a highly nonspecific marker of perineurial differentiation in malignant-appearing lesions. Among vascular tumors, expression of GLUT-1 protein is seen in essentially all juvenile capillary hemangiomas, but not in other pediatric vascular tumors, such as vascular malformations and kaposiform hemangioendothelioma ( Fig. 6.20 ). Its expression in vascular tumors appears unrelated to the proliferative activity of the lesions.
The product of the pseudoautosomal MIC2 gene, CD99 is a transmembrane glycoprotein of 30 to 32 kDa (p30/32). Its exact function is unknown, although it appears to play a role in cellular adhesion and regulation of cellular proliferation. The MIC2 gene is expressed and the CD99 antigen produced in almost all human tissues, although the level of expression varies significantly. Normal tissues that usually display strong CD99 expression include cortical thymocytes and Hassall corpuscles, granulosa and Sertoli cells, endothelium, pancreatic islets, adenohypophysis, ependyma, and some epithelium, including urothelium, squamous epithelium, and columnar epithelium.
The most important use of antibodies to CD99 is for immunohistochemical diagnosis of Ewing sarcoma (ES). Many studies show that more than 90% of ES express CD99, with a characteristic membranous pattern ( Fig. 6.21 ). Despite early claims that CD99 expression was also specific for ES, this clearly is not true. It is particularly important to recognize that a significant subset of other small blue round cell tumors considered in the differential diagnosis of ES may express this antigen. CD99 expression is seen in more than 90% of lymphoblastic lymphomas, 20% to 25% of primitive rhabdomyosarcomas, more than 75% of poorly differentiated synovial sarcomas, approximately 50% of mesenchymal chondrosarcomas, and in rare cases of small cell osteosarcomas and intraabdominal desmoplastic round cell tumor. CD99 expression is not a feature of neuroblastomas or esthesioneuroblastomas. In general, “Ewing-like” primitive sarcomas of the type showing CIC or BCOR rearrangements typically show only patchy and weak expression of CD99, a finding that may be an important clue to the diagnosis of these rare sarcomas.
IHC analysis of CD99 expression plays a limited role in the diagnosis of pleomorphic or spindle cell soft tissue neoplasms. As noted, many synovial sarcomas express CD99, which may be helpful in discriminating them from malignant peripheral nerve sheath tumors and fibrosarcomas. Expression of CD99 may also be seen in solitary fibrous tumors, mesotheliomas, leiomyosarcomas, and undifferentiated pleomorphic sarcomas. In general, use of CD99 in the diagnosis of spindle cell neoplasms has been superseded by the development of much more specific immunohistochemical (e.g., STAT6 in solitary fibrous tumor) or molecular (e.g., SS18 FISH in synovial sarcoma) markers.
CD56 (Neural Cell Adhesion Molecule)
The 140-kDa isoform of the neural cell adhesion molecule, CD56, is an integral membrane glycoprotein that mediates calcium-independent homophilic cell-cell binding. CD56 is expressed by many normal cells and tissues, including neurons, astrocytes, and glia of the cerebral cortex and cerebellum, adrenal cortex and medulla, renal proximal tubules, and follicular epithelium of thyroid; gastric parietal cells; cardiac muscle; regenerating and fetal skeletal muscle; pancreatic islet cells; and peripheral nerve. CD56 is also ubiquitously expressed on human natural killer (NK) cells and on a subset of T lymphocytes.
As might be expected from this long list of CD56-positive normal tissues, CD56 expression is widespread among sarcomas. Soft tissue tumors that often express CD56 include synovial sarcoma, malignant peripheral nerve sheath tumor, schwannoma, rhabdomyosarcoma, leiomyosarcoma, leiomyoma, chondrosarcoma, and osteosarcoma. For this reason, examination of CD56 expression is not helpful when evaluating spindle cell soft tissue tumors. However, CD56 expression may be useful for evaluating primitive small blue round cell tumors, particularly in combination with CD99. CD56 expression is seen in only 10% to 25% of ES and in rare lymphoblastic lymphomas, compared with almost 100% of neuroblastomas, poorly differentiated synovial sarcomas, alveolar and primitive embryonal rhabdomyosarcomas, small cell carcinomas, Wilms tumors, and mesenchymal chondrosarcomas. The absence of CD56 expression may be a clue to the diagnosis of ES in cases where results with more specific positive markers such as CD99, CD45, keratin, and desmin are equivocal. Demonstration of CD56 expression may also be of some value in the diagnosis of ossifying fibromyxoid tumor, particularly when S-100 protein expression is weak or absent.
Synaptophysin and Chromogranin A
A transmembrane glycoprotein found in presynaptic vesicles, synaptophysin plays an important role in the packaging, storage, and release of neurotransmitters and functions as a membrane channel protein. It is routinely expressed by neural, endocrine, and neuroendocrine cells, as well as adrenal cortical cells and tumors. Synaptophysin expression is typically confined to neural and neuroendocrine tumors, including neuroendocrine carcinomas of various grades, Merkel cell carcinomas, paragangliomas, neuroblastomas, and esthesioneuroblastomas. Synaptophysin expression is relatively uncommon in “neuroectodermal” tumors, such as ES. Aberrant synaptophysin expression may also be seen in alveolar rhabdomyosarcomas, a minority of extraskeletal myxoid chondrosarcomas, and rare melanomas. Aberrant expression of synaptophysin and CD56 is one of the defining features of the recently described composite hemangioendothelioma with neuroendocrine marker expression ( Fig. 6.22 ).
A calcium-binding protein member of the granin family, chromogranin A is stored in the dense core granules of neural and neuroendocrine cells. Chromogranin A expression in normal tissues is generally similar to that of synaptophysin, although it is not expressed by adrenocortical cells. Aberrant expression of chromogranin A is much less common than that of synaptophysin and when present tends to be confined to very few cells.
NKX2.2 is a homeodomain-containing transcription factor that plays an important role in both neuroendocrine and glial differentiation; it has also been identified as a gene necessary for the oncogenic transformation of Ewing sarcoma, and one regulated by the fusion gene, EWSR1-FLI1 , that results from the t(11;22) characteristic of this tumor. NKX2.2 appears to represent a highly sensitive marker of ES in the context of small blue round cell tumors, with sensitivity of about 90%. However, NKX2.2 is fairly nonspecific and may also be expressed by olfactory neuroblastomas, mesenchymal chondrosarcomas, small cell carcinomas, CIC-DUX4 sarcomas, and melanomas. With the increasingly widespread availability of molecular genetic testing for ES-associated fusion events, it is difficult to see a role for NKX2.2 immunohistochemistry, except in rare instances.
Markers of Melanocytic Differentiation
Monoclonal antibody HMB-45 identifies the Pmel 17 gene product, gp100. This gene product is a component of the premelanosomal/melanosomal melanogenic oxidoreductive enzymes and as such is melanosome specific but not melanoma specific. HMB-45 is positive in the unusual myomelanocytic tumors that consist of the perivascular epithelioid cell family of tumors (angiomyolipoma, clear cell tumor of lung, lymphangioleiomyomatosis, soft tissue, and bone PEComas), but it has not convincingly been shown to react with any tumor that does not contain melanosomes. HMB-45 is generally negative in nevi and resting melanocytes but is expressed in approximately 85% of melanomas ( Fig. 6.23 ). Fewer than 10% of desmoplastic melanomas are HMB-45 positive.
Melan-A, the product of the MART1 gene (melanoma antigen recognized by T cells), is a 20- to 22-kDa component of the premelanosomal membrane. Its function is unknown. Like HMB-45, melan-A is a marker of melanosomes, not melanomas; it is also present in perivascular epithelioid cell tumors (PEComas) ( Fig. 6.24 ). Unlike HMB-45, melan-A is positive in resting melanocytes and nevi. Melan-A is expressed by approximately 85% of epithelioid melanomas and has been reported in up to 50% of desmoplastic melanomas, although the true rate is almost certainly much lower and is probably the same as for HMB-45. Melan-A is present in some HMB-45–negative melanomas, and vice versa. Interestingly, the most widely used antibody to melan-A, A103, is also useful for the diagnosis of adrenocortical and other steroid-producing tumors. A103 has reproducible cross-reactivity with an unknown epitope present in these tumors, and it may be helpful for distinguishing adrenocortical carcinomas from renal cell carcinoma.
Microphthalmia Transcription Factor
Microphthalmia transcription factor (MiTF), the product of the microphthalmia ( mi ) gene located on chromosome 3p14.1, is a transcription factor critical for melanocyte development. Mutations in mi were first described in mice in the 1940s, and it was subsequently shown in humans that heterozygous mi deficiency results in Waardenburg syndrome IIa, clinically defined by skin pigmentation abnormalities, bilateral hearing loss, and a white forelock. Biochemical studies have shown that MiTF, the protein encoded by mi , can transactivate several downstream gene promoters, including genes ultimately responsible for melanin biosynthesis: tyrosinase and related pigmentation enzymes TRP-1 and TRP-2. MiTF is expressed in essentially all resting melanocytes and nevi. Both melanocyte-specific and nonspecific isoforms of this protein exist, although the commercially available antibodies to MiTF are not specific for the melanocytic isoforms, in our experience, despite claims to the contrary.
MiTF was initially described as highly sensitive and specific marker of melanoma and is expressed in more than 90% of epithelioid melanomas. Sarcomatoid melanomas are less often positive (40%), and true desmoplastic melanomas are infrequently positive (<5%). MiTF is also expressed in almost all clear cell sarcomas ( Fig. 6.25 ). MiTF expression may also be seen in PEComas of all types. Because the commercially available antibodies are not specific for the melanocytic isoforms, MiTF expression is not limited to melanomas and may be seen in leiomyosarcomas, atypical fibroxanthomas, atypical lipomatous neoplasms, and very rare carcinomas. Thus, MiTF is best used for the confirmation of S-100 protein–positive, HMB-45/melan-A/tyrosinase–negative tumors suspected of being melanomas. MiTF expression in the absence of S-100 protein expression is not diagnostic of melanoma. Demonstration of MiTF expression may be of some value in the diagnosis of cellular neurothekeoma, particularly in its distinction from plexiform fibrohistiocytic tumor.
PNL2 is a monoclonal antibody that was generated to a fixative-resistant melanocytic antigen. Initial studies employing PNL2 demonstrated it to have sensitivity and specificity for melanoma comparable to that of HMB-45 and antibodies to melan-A, with the sole exception of neutrophils, which were also identified with antibody PNL2. Furthermore, like the latter two antibodies, PNL2 did not label desmoplastic melanomas. A larger more recent series of more than 1000 tumors showed that PNL2 was positive in greater than 85% of epithelioid melanomas and positive in PEComas; all other nonmelanocytic tumors were negative. PNL2 seems to represent a melanoma marker of comparable utility to HMB-45 and antibodies to melan-A.
Tyrosinase is an enzyme involved in the synthesis of melanin. Antibodies to tyrosinase have sensitivity and specificity comparable to that of HMB-45 and melan-A. In general, we reserve the use of tyrosinase for cases strongly suspected of representing melanoma, which are negative with HMB-45 or melan-A.
Markers of Endothelial Differentiation
A number of markers have been used to demonstrate endothelial differentiation, paralleling the progression in our concept of soft tissue tumors manifesting endothelial differentiation. Endothelial markers include commonly used markers such von Willebrand factor (vWF, factor VIII–associated protein, often erroneously referred to as factor VIII), CD34, CD31, as well as more recently described markers such as FLI-1/ERG, HHV8 LANA protein, and podoplanin. The pattern of expression of these markers in endothelial tumors is much like an overlapping Venn diagram, in which most tumors express all these markers, but some express only a subset. Whereas early studies suggested that markers such as vWF could differentially identify vascular versus lymphatic endothelium, more recent studies have demonstrated that both vascular and lymphatic endothelium express all three markers, and novel markers such as podoplanin, Prox1, and vascular endothelial growth factor receptor-3 (positive on lymphatic endothelium and negative on vascular endothelium) are required to make this distinction.
Von Willebrand Factor (Factor VIII-Related Antigen)
The von Willebrand factor was the first endothelium-specific marker employed in diagnostic immunohistochemical studies. vWF was first used to identify the nature of the vinyl chloride–induced sarcomas of liver and subsequently was shown to be a marker of vascular tumors of multiple sites, including but not limited to those of the central nervous system, gastrointestinal tract, and breast. vWF is the least sensitive of the vascular markers and is positive in 50% to 75% of vascular tumors. Although vWF expression in theory is absolutely specific for vascular tumors, technical problems limit its usefulness. vWF is not only produced by endothelial cells but also circulates in the serum; therefore it can be found often in zones of tumor necrosis and hemorrhage ( Fig. 6.26 ). Given the availability of much better endothelial markers, we see little role for vWF immunohistochemistry in current practice.
CD34 (Human Hematopoietic Progenitor Cell Antigen)
The function of CD34, a 110-kDa transmembrane glycoprotein, is thought to be related to cell-cell adhesion. It is expressed on hematopoietic stem cells, endothelium, the interstitial cells of Cajal, and a group of interesting dendritic cells present in the dermis, around blood vessels, and in the nerve sheath. CD34 is expressed in more than 90% of vascular tumors and is a highly sensitive marker of Kaposi sarcoma (KS). As shown by the previous list of normal tissues, CD34 expression is not limited to vascular tumors. CD34 expression is well documented in dermatofibrosarcoma protuberans, solitary fibrous tumors ( Fig. 6.27 ), malignant peripheral nerve sheath tumors, gastrointestinal stromal tumors, and epithelioid sarcomas. CD34 is expressed by approximately 50% to 60% of epithelioid sarcomas, but less than 2% of carcinomas. CD34 expression may be valuable for distinguishing keratin-positive epithelioid angiosarcomas and epithelioid sarcomas from carcinoma. Diffuse CD34 immunoreactivity, in combination with much more limited aberrant keratin expression, is characteristic of superficial CD34-positive fibroblastic tumor.
CD31 (Platelet Endothelial Cell Adhesion Molecule-1)
CD31 is generally considered to be the most sensitive and specific endothelial marker. It is expressed in more than 90% of angiosarcomas, hemangioendotheliomas, hemangiomas, and KS and probably much less than 1% of carcinomas ( Fig. 6.28 ; see also Fig. 6.5 ). CD31 expression is not seen in any nonendothelial tissue or tumor, with the notable exception of macrophages and platelets ( Fig. 6.29 ). CD31 expression in intratumoral macrophages may result in the misdiagnosis of a nonvascular tumor as a vascular tumor, if one is not aware of this potential pitfall. The CD31 expression seen in macrophages is distinctly granular, compared with the intense cytoplasmic and linear membranous staining of endothelium. CD31 expression has recently been reported in a small number of extremely rare primitive neoplasms showing CIC-DUX4 gene fusions; whether these tumors represent CIC -rearranged angiosarcomas or CIC-DUX4 sarcomas showing aberrant CD31 expression is not clear at this time ( Fig. 6.30 ) .
FLI-1 and ERG proteins
Members of the ETS family of transcription factors, FLI-1 and ERG are the best currently available nuclear markers of endothelial differentiation. Both FLI-1 and ERG are positive in more than 95% of endothelial neoplasms of all types and degrees of malignancy, including hemangiomas, hemangioendotheliomas, angiosarcomas, and KS (see Fig. 6.28B ). Data are conflicting on FLI-1 and ERG expressed in epithelioid sarcomas. Initial studies showed both markers to be absent in epithelioid sarcoma, assisting in their distinction from epithelioid forms of angiosarcoma. Stockman et al., however, found FLI-1 expression in essentially all epithelioid sarcomas; expression of ERG protein was also common using N-terminus antibodies, but was absent with C-terminus antibodies. Technical differences may account for these discrepant findings, with only rare examples of FLI-1–positive or ERG-positive epithelioid sarcomas in our practice. Rare melanomas, adenocarcinomas, and Merkel cell carcinomas have been reported to show focal FLI-1 positivity. ERG expression is present in almost 50% of prostatic adenocarcinoma, reflecting the presence of the prostate cancer–associated TMPRSS2-ERG fusion. ERG expression may also be seen in a small minority of ES/PNET, 70% of blastic extramedullary myeloid tumors, and rare large cell pulmonary carcinomas and mesotheliomas. FLI-1 is frequently present in subsets of mature lymphocytes, whereas ERG is not.
Markers of Lymphatic Endothelial Differentiation
Vascular Endothelial Cell Growth Factor Receptor 3
The platelet-derived growth factor (PDGF) family, including vascular endothelial growth factor (VEGF) and the closely related molecules VEGF-B, VEGF-C, and VEGF-D, play a significant role in angiogenesis and vascular permeability. VEGF-C plays a critical role in lymphangiogenesis; in transgenic mice, VEGF-C has the ability to induce both lymphatic endothelial proliferation and lymphatic vessel formation. Early studies suggested that VEGFR-3 was a highly sensitive and specific marker of KS, normal lymphatics, and lymphangiomas. Two subsequent large studies confirmed its superb (>95%) sensitivity for KS, but also noted expression in almost 50% of angiosarcomas, some with features suggestive of lymphatic differentiation, occasional hemangiomas, and almost all cases of kaposiform hemangioendothelioma, Dabska tumor, and retiform hemangioendothelioma. It is now generally conceded that VEGFR-3 expression is not specific for lymphatic differentiation in endothelial neoplasms. VEGFR-3 expression has not been extensively studied in nonendothelial neoplasms.
Podoplanin is a transmembrane glycoprotein normally expressed by lymphatic endothelial cells, glomerular podocytes, choroid plexus epithelium, type 1 alveolar cells, osteoblasts, and mesothelial cells. Expression of podoplanin is regulated by the homeobox gene PROX1 . Although podoplanin was once considered a relatively restricted marker of lymphatic endothelial-derived tumors, subsequent investigations have shown podoplanin expression in a wide variety of endothelial tumors and in many different types of mesenchymal, germ cell, and glial neoplasms. Antibodies to podoplanin are also of value in the differential diagnosis of mesothelioma (often positive) and adenocarcinoma (usually negative).
Expression of Prospero homeobox 1 protein (Prox1) is critical for normal lymphatic differentiation and is seen in the nuclei of developing and mature lymphatic endothelial cells. A few older studies evaluated Prox1 as a marker of neoplastic endothelial cells, with expression noted in KS, kaposiform hemangioendothelioma, and lymphangiomas. More recently, Miettinen and Wang evaluated Prox1 in a large number of endothelial and nonendothelial tumors, noting expression in almost 100% of KS, 48% of angiosarcomas, 56% of hemangioendotheliomas (including epithelioid, kaposiform, and retiform subtypes), and 42% of hemangiomas, most notably spindle cell hemangiomas (94%). They also showed Prox1 expression in smaller percentages of ES, paragangliomas, synovial sarcomas, non–small cell carcinomas of various primary sites, and pancreatic islet cell tumors. Although neither podoplanin nor Prox1 is specific for lymphatic endothelium, diffuse expression of both markers in the appropriate histologic context is generally considered indicative of lymphatic endothelial differentiation.
Human Herpesvirus 8 (HHV8) Latency-Associated Nuclear Antigen (LANA)
An infectious etiology for Kaposi sarcoma has long been suspected, and epidemiologic, serologic, and molecular genetic studies have identified a novel herpesvirus, HHV8, as the presumed causative agent of KS. HHV8 is known to latently infect endothelial cells, as well as peripheral blood monocytes and B lymphocytes in patients with KS. LANA is one of the most highly expressed proteins during latent HHV8 infection. The LANA protein is encoded by the open reading frame 73 (ORF73) of the HHV8 genome, where it tethers viral DNA to host heterochromatin and is thereby required for persistence of viral DNA in dividing cells. LANA also has an essential role in maintenance of the episomal DNA during latent infection and cell division and regulates gene expression in infected cells. Using different antibodies, multiple studies have shown that expression of LANA protein is highly sensitive and specific for KS ( Fig. 6.31 ). LANA expression has not been reported in any non-KS tumors, with the notable exceptions of primary effusion lymphoma and Castleman disease.
Recommendations for Use of Vascular Markers
In general, CD31 is the single best screening marker to assess for endothelial differentiation in tumors. FLI-1 and ERG are best used for the confirmation of endothelial lineage in tumors showing limited expression of CD31 or in cases containing large numbers of CD31-positive macrophages. CD34 should not be used as stand-alone endothelial marker. Table 6.3 summarizes the endothelial markers.
|CD34||Moderate||High||Epithelioid sarcoma, solitary fibrous tumor, DFSP, GIST|
|FLI-1||Moderate||High||Ewing sarcoma, small lymphocytes, lymphoblastic lymphoma|
|ERG||Moderate||High||Extramedullary myeloid tumor, subset of prostatic adenocarcinoma|
|Type IV collagen||Moderate||Moderate||Glomus tumors, nerve sheath tumors, smooth muscle tumors|
Markers of Gastrointestinal Stromal Tumors
The c-kit proto-oncogene product (CD117), a transmembrane receptor for stem cell factor, is normally expressed by mast cells, melanocytes, germ cells, various subsets of hematopoietic cells, and the interstitial cells of Cajal of the GI tract. CD117 is expressed by 85% to 95% of gastrointestinal stromal tumors (GISTs). In the GI tract, CD117 is a highly specific marker of GIST; it is not usually expressed by tumors in the differential diagnosis of a GI mesenchymal tumor, such as leiomyomas, leiomyosarcomas, and nerve sheath tumors ( Fig. 6.32 ). CD117 is expressed by melanocytic tumors, however, such as melanoma and clear cell sarcoma. CD117 expression may also be seen in a minority of Ewing sarcoma and PEComas. In any type of tumor, care should be taken not to mistake mast cells in a tumor for scattered CD117-positive tumor cells ( Fig. 6.33 ).
Anoctamin-1 (ANO1, DOG1, TMEM16A)
Anoctamin-1, more widely known as DOG1 (discovered on GIST-1), is a calcium-activated chloride channel normally expressed in a wide variety of tissues, including the interstitial cells of Cajal. Using expression profiling, West et al. found that DOG1 is selectively expressed by GIST among studied tumors. Subsequent studies of large numbers of GIST confirmed the high sensitivity (>94%) and relatively high specificity of DOG1 for GIST among mesenchymal tumors. Near-constant DOG1 expression has been shown in GIST from all anatomic locations, irrespective of CD117 expression or KIT mutational status ( Fig. 6.34 ), and in GIST arising in unusual clinical settings (e.g., pediatric, NF1-associated). Expression of DOG1, generally weaker in intensity or in a small number of cells than is typically seen in GIST, has also been reported in a small number of leiomyomas, synovial sarcomas, and adenocarcinomas, including acinar cell carcinomas of salivary gland origin. We have seen patchy and weak DOG1 expression in leiomyosarcomas arising in GI locations, emphasizing the need to use this antibody as part of a panel of immunohistochemical markers. Also, remember that not all GISTs are DOG1 positive. DOG1 expression may also be seen in low-grade fibromyxoid sarcoma and sclerosing epithelioid fibrosarcoma.
Immunohistochemistry as Surrogate for Tumor-Specific Molecular Alterations
The IHC studies applied to sarcomas described thus far are used to identify “cell type”–specific markers, that is, to identify the “normal cell counterpart” to the mesenchymal tumor in question. An emerging class of soft tissue tumors, however, does not appear to have a normal cell counterpart and represents a set of tumors instead characterized by specific genetic alterations, usually chromosomal translocations. The latter can result in the abnormal juxtaposition of two genes, resulting in the neoexpression of one of (or a portion of) the gene products. The alterations can sometimes be identified by IHC studies, which thus serve as a surrogate for the presence of the chromosomal translocation.
FLI-1 and ERG as Markers of Ewing Sarcoma
Ewing sarcoma is characterized by recurrent translocations involving the EWSR1 gene on 2q12, most frequently involving the FLI1 gene located on 11q22 (approximately 85% of cases) or the ERG gene located on 21q12 (5% of cases). Polyclonal and monoclonal antibodies to FLI-1 protein are positive in 70% to 90% of genetically confirmed ES, most often reflecting the presence of the EWSR1-FLI1 fusion protein. Interestingly, antibodies to FLI-1 are positive in a significant percentage of ES with known EWSR1-ERG fusions, reflecting protein homology between FLI-1 and ERG. In contrast, antibodies to ERG seem to be much more specific for ES/PNET containing ERG rearrangements, showing only weak positivity in rare cases known to have FLI1 rearrangements. FLI-1 expression is not generally seen in other tumors that enter this differential diagnosis, including rhabdomyosarcoma, mesenchymal chondrosarcoma, neuroblastoma, and Wilms tumor. However, lymphoblastic lymphomas are routinely FLI-1 positive, as are occasional cases of Merkel cell carcinoma, melanoma, and desmoplastic small round cell tumor. As noted earlier, both FLI-1 and ERG are routinely positive in endothelial tumors. ERG expression seems to be somewhat more restricted but may be seen in blastic extramedullary myeloid tumors and in a subset of prostatic adenocarcinomas.
WT-1 as Marker of t(11;22)(13;q24) Translocation of Desmoplastic Small Round Cell Tumor and CIC -Rearranged Sarcomas
Desmoplastic small round cell tumors (DSRCTs) are characterized in most cases by a specific translocation, t(11;22)(p13;q24), which fuses the EWSR1 and WT1 genes and produces a fusion protein containing the carboxy-terminus of WT-1. Antibodies directed against the carboxy-terminus of WT-1 are highly sensitive (>90%) and relatively specific markers of DSRCTs, among small blue round cell tumors. Although generally thought that the more common amino-terminus antibodies to WT-1 are negative in DSRCT, this is not always the case. Murphy et al. demonstrated immunoreactivity with the N-terminus WT-1 antibody in a genetically confirmed case of DSRCT carrying two novel fusion transcripts, both lacking WT1 exons 9 and 10 and one containing additional WT1 exons 3 to 7; this tumor also expressed full-length WT1 . Hung et al. also recently demonstrated N-terminus WT1 immunoreactivity in two DSRCTs.
It is important to realize that many rhabdomyosarcomas express cytoplasmic wild-type WT-1, which will be identified by both amino- and carboxy-terminus antibodies, and which should be rigorously distinguished from the nuclear positivity seen in DSRCTs. Wild-type WT-1 expression is also seen in Wilms tumor, although this is seldom in the differential diagnosis of DSRCT and is usually easily identified by routine microscopy.
Ewing-like sarcomas showing CIC rearrangements are also positive in more than 95% of cases for WT-1, using the amino-terminus antibody, typically in a strong nuclear and cytoplasmic pattern. This is likely caused by transcriptional upregulation of WT-1 in CIC -rearranged sarcomas. In contrast, WT-1 expression is not a feature of ES or BCOR -rearranged primitive sarcomas, and thus WT-1 may be very helpful in this differential diagnosis ( Fig. 6.35 ).
TFE3 as Marker of Alveolar Soft Part Sarcoma and TFE3-Rearranged Perivascular Epithelioid Cell Tumors
Alveolar soft part sarcoma (ASPS) is characterized in almost all cases by a tumor-specific der(17)t(X;17)(p11;q25) that fuses the TFE3 gene at Xp11 to the ASPSCR1 gene at 17q25, creating an ASPSCR1-TFE3 fusion protein. Antibodies directed against the carboxy-terminus of the TFE3 transcription factor have been reported to be highly sensitive and specific markers of ASPS ( Fig. 6.36 ). Although low levels of TFE3 expression are present in almost all normal tissues, strong nuclear expression of TFE3 has been suggested to be confined to tumors known to harbor TFE3 gene fusions, such as ASPS and rare pediatric renal carcinomas. However, we recently found that modern, highly sensitive epitope retrieval methods and detection systems seem to greatly reduce the utility of TFE3 immunohistochemistry, with nuclear TFE3 immunoreactivity demonstrating only a very poor correlation with TFE3 rearrangements detected by fluorescence in situ hybridization (FISH). TFE3 expression is also common in granular cell tumors and may be seen in some TFE3 -rearranged PEComas. In general, we have found FISH to be a much more specific marker for TFE3 rearrangements than IHC.
SMARCB1 (INI1) and SMARCA4 (BRG1) Expression Loss as Markers of Aberrations in SWI/SNF Chromatin-Remodeling Complex
The SWI/SNF chromatin-remodeling complex, ubiquitously expressed in all normal cells, is critical for nucleosome remodeling and transcriptional regulation. This complex is composed of several subunits, including SMARCB1 (INI-1), SMARCA4 (BRG1), SMARCC1 (BAF155), SMARCC2 (BAF170), and SMARCF1(ARID1a).
The SMARCB1 protein is the product of the hSNF5/INI1/SMARCB1/BAF47 gene, located on chromosome 22q11.2. Loss of SMARCB1 , either in the form of monosomy 22 or as homozygous deletions in the gene itself, has been strongly implicated in the pathogenesis of renal and extrarenal rhabdoid tumors as well as atypical teratoid/rhabdoid tumors of the central nervous system. On IHC, loss of SMARCB1 protein is seen in essentially all renal/extrarenal rhabdoid tumors ( Fig. 6.37 ) and more than 90% of epithelioid sarcomas of both conventional and proximal type. Loss of SMARCB1 expression is also seen in approximately 50% of epithelioid malignant peripheral nerve sheath tumors, where it may be a useful marker to distinguish this tumor from melanoma, in approximately 17% of extraskeletal myxoid chondrosarcoma, and in medullary carcinoma of the kidney, a tumor often associated with sickle cell trait. Loss of SMARCB1 expression is also seen in subsets of schwannomas, in particular those associated with schwannomatosis and epithelioid schwannomas.
A small subset of rhabdoid tumors and epithelioid sarcomas showing retained expression of SMARCB1 have also shown loss of SMARCA4 (BRG1) expression. SMARCA4 loss also characterizes an unusual primitive sarcoma of the thoracic cavity, so-called “ovarian small cell carcinoma of hypercalcemic type,” and a variety of undifferentiated carcinomas in several other locations. Loss of expression of SMARCF1 (ARID1) is most often associated with clear cell carcinoma of the ovary.
In practical terms, we report SMARCB1 and SMARCA4 results as “retained” (normal) and “absent” (abnormal) rather than “positive” and “negative.” Suspected cases of epithelioid sarcoma and rhabdoid tumor showing retained expression of SMARCB1 should be evaluated for SMARCA4 loss.
Anaplastic Lymphoma Kinase in ALK -Rearranged Tumors
Anaplastic lymphoma kinase (ALK) is a transmembrane tyrosine kinase first identified as part of the characteristic t(2;5) (NPM-ALK) translocation seen in anaplastic large cell lymphomas. In normal tissues, expression of ALK protein is restricted to the central nervous system. Inflammatory myofibroblastic tumors frequently contain chromosomal rearrangements that result in activation of the ALK gene, with subsequent overexpression of ALK protein in about 40% of cases ( Fig. 6.38 ). However, not all molecular alterations in the ALK locus may be reflected in IHC-detectable ALK protein, and FISH studies may prove more sensitive in this regard. However, overexpression of ALK may also be seen in a variety of other soft tissue tumors, including rhabdomyosarcoma, lipogenic tumors, ES, undifferentiated pleomorphic sarcoma, and leiomyosarcoma. In rhabdomyosarcoma, the presence of cytoplasmic ALK expression (more frequently seen in ARMS vs. ERMS) is associated with the presence of increased gene copy number and, less frequently, other gene alterations. ALK expression has been demonstrated to help distinguish inflammatory myofibroblastic tumors from low-grade myofibroblastic sarcoma, which does not express ALK.
ETV4 in CIC -Rearranged Primitive Sarcomas
A distinct subset of Ewing-like primitive round cell sarcomas shows rearrangements of the CIC (capicua transcriptional repressor) gene, typically as a result of t(4;19)(q35;q13.1) or t(10;19)(q26;q13.1) fusions. Gene expression profiling studies have shown overexpression of ETV4 , among other genes, as a characteristic feature of CIC -rearranged sarcomas. IHC for ETV4 has been shown to be positive in more than 90% of CIC -rearranged sarcomas and in only a very small minority of other primitive sarcomas. In particular, BCOR -rearranged primitive sarcomas have been consistently ETV4 negative ( Fig. 6.39 ).
BCOR and CCNB3 as Markers of Primitive Sarcomas Showing BCOR Gene Rearrangements and Internal Tandem Duplications
Another distinctive subset of Ewing-like primitive sarcomas is characterized by abnormalities involving the BCOR gene, which encodes the BCL6 transcriptional repressor. Although a variety of different genetic events have been identified in these tumors, including BCOR-CCNB3 , BCOR-MAML3 , BCOR internal tandem duplications, YWHAE-NUTM2B, and ZC3H7B-BCOR , these tumors seem to share similar transcriptional signatures, including high BCOR mRNA expression.
Pierron et al. first tested IHC for CCNB3 (a gene that encodes the testis-specific cyclin B3) as a marker of this group of sarcomas and found it was positive in 18 tested BCOR-CCNB3 sarcomas, but not in any of 25 BCOR-CCNB3 –negative tumors. Yamada, Ludwig, and Matsuyama and colleagues subsequently reported similar findings.
However, because the CCNB3 gene is not involved in subsets of BCOR -rearranged sarcomas or those showing instead internal tandem duplications of BCOR, more recent work has focused on BCOR IHC as a marker of these tumors. Kao et al. showed strong nuclear expression of BCOR protein in more than 95% of tumors showing BCOR-CCNB3 , BCOR-MAML3 , BCOR internal tandem duplications, and related YWHAE-NUTM2B tumors. They also showed BCOR immunoreactivity in 49% of synovial sarcomas, including all poorly differentiated types. With the exception of synovial sarcomas, however, BCOR IHC was found to be a specific marker of tumors with BCOR derangements. In a much larger series, Matsuyama et al. found BCOR expression in 100% of tested BCOR sarcomas and in only 4% of 412 other tumors, including some solitary fibrous tumors, ES, synovial sarcomas, small cell osteosarcomas, lymphomas, and small cell carcinomas. Other studies have also shown the high sensitivity of BCOR antibodies for BCOR sarcomas ( Fig. 6.40 ).
CAMTA1 as Marker of Epithelioid Hemangioendothelioma–Defining WWTR1-CAMTA1 Fusion
Epithelioid hemangioendothelioma is characterized at the genetic level by the reciprocal translocation t(1;3)(p36;q23-25), resulting in fusion of the WWTR1 and CAMTA1 genes. CAMTA1 (calmodulin-binding transcription activator 1) encodes a transcription factor normally expressed only in the brain. Expression of CAMTA1 by IHC, using a rabbit polyclonal antibody (Novus Biologicals, Littleton, Colorado) has been shown to be a highly sensitive and specific marker of epithelioid hemangioendothelioma, present in more than 85% of cases and not in morphologic mimics. Use of the correct CAMTA1 antibody is critical, since an earlier study by Yusifli and Kosemehmetoglu, using a different antibody, had shown very poor specificity for CAMTA1 IHC in the differential diagnosis of epithelioid hemangioendothelioma. CAMTA1 expression is seen in both conventional epithelioid hemangioendotheliomas and in cases showing high-grade cytology, greatly assisting in the distinction of the latter tumors from conventional epithelioid angiosarcoma ( Fig. 6.41 ).
FOSB as Marker of Epithelioid Sarcoma–Like (Pseudomyogenic) Hemangioendothelioma and Epithelioid Hemangioma
FOSB, a member of the Fos transcription factor family, is involved in a variety of biologic processes, including oncogenesis and adaption to stress. Alterations in FOSB were first implicated in the pathogenesis of certain types of endothelial neoplasia with the identification of SERPINE1-FOSB fusions in epithelioid sarcoma–like (pseudomyogenic) hemangioendotheliomas. Subsequently, FOSB rearrangements were identified in 9 of 46 studied epithelioid hemangiomas, in particular those with atypical morphologic features, including seven cases showing ZFP36-FOSB , one with WWTR1-FOSB , and one harboring FOSB rearrangement with an unknown partner. A subsequent study identified FOS rearrangements in 17 of 57 (29%) epithelioid hemangiomas, most often occurring in bone or showing cellular/solid histology; all tested cases of cutaneous “angiolymphoid hyperplasia with eosinophilia” were negative.
More recently, IHC for FOSB has been shown to be a useful surrogate for the identification of FOSB rearrangements in epithelioid sarcoma–like (pseudomyogenic) hemangioendothelioma and epithelioid hemangiomas ( Fig. 6.42 ). Expression of FOSB in epithelioid sarcoma–like hemangioendothelioma was first documented by Ide et al. and Sugita et al. and later confirmed in a larger series of 50 cases by Hung et al. Hung noted FOSB expression in more than 95% of epithelioid sarcoma–like (pseudomyogenic) hemangioendotheliomas, with weaker expression noted in occasional cases of angiosarcoma, epithelioid hemangioendothelioma, and nodular/proliferative fasciitis. This same study also showed FOSB expression in 54% of epithelioid hemangiomas, including all cutaneous cases. More recently, Ortins-Pena and Llamas-Velasco demonstrated FOSB expression in all studied epithelioid hemangiomas, including cutaneous cases, with absent expression in a variety of benign and malignant potential mimics.
STAT6 as Marker of NAB2-STAT6 Fusion in Solitary Fibrous Tumors
Solitary fibrous tumors (formerly known as “hemangiopericytomas”) of all anatomic locations, including the meninges, are now known to be caused by an inversion at the 12q13 locus, resulting in fusion of the NAB2 and STAT6 genes. IHC for STAT6 has been subsequently shown to be a robust marker for the diagnosis of solitary fibrous tumors, present in more than 95% of cases in several large series (see Fig. 6.27C ). We have found similar results in our clinical practice, although expression of STAT6 at times can be relatively weak and focal, necessitating careful examination of the stained slide and occasionally evaluation of more than one section. Importantly, only nuclear immunoreactivity should be taken as evidence in support of the diagnosis of solitary fibrous tumor, because cytoplasmic immunoreactivity is much less specific. In general, other CD34-positive tumors that may be confused with solitary fibrous tumor, such as genital stromal tumors, spindle cell lipomas, and dermatofibrosarcomas, are STAT6 negative. One notable exception is dedifferentiated liposarcomas, which may show nuclear immunoreactivity for STAT6, likely reflecting amplification of this locus as part of the 12q13 amplifications that typify dedifferentiated and well-differentiated liposarcomas.
H3K27me3 Loss as Marker of Malignant Peripheral Nerve Sheath Tumors
A significant subset of malignant peripheral nerve sheath tumors (MPNSTs) show loss-of-function mutations in the EED or SUZ12 genes, which encode the core subunit of the polycomb repressive complex 2 (PRC2). Loss of PRC2 activity leads to loss of trimethylation of histone H3 at lysine 27 (H3K27me3). Several studies have investigated loss of expression of H3K27me3 as a marker of malignant peripheral nerve sheath tumors. 443-449 Lee et al. first showed that complete loss of H3K27me3 expression is a universal feature of MPNST known to have PRC2 loss ( Fig. 6.43A and B ). A subsequent study by this same group showed H3K27me3 loss in 61% of MPNSTs, including 60% of neurofibromatosis 1 (NF1)–associated tumors and more than 90% of sporadic or postirradiation tumors. Retained expression was seen in neurofibromas and epithelioid MPNSTs. Cleven et al. found complete H3k27me3 loss in 34% of MPNSTs, including 41% of NF1-associated tumors, with retained expression in all tested neurofibromas and perineuriomas and all but one of tested schwannomas. Interestingly, this study also found H3K27me3 loss in 60% of synovial sarcomas and 38% of fibrosarcomatous dermatofibrosarcomas. Pekmezci et al. reported similar findings in MPNSTs, neurofibromas, synovial sarcomas, and fibrosarcomatous dermatofibrosarcomas. Importantly, however, complete loss of H3k27me3 expression has also been reported to be a feature of greater than 35% of melanomas, including the desmoplastic subtype, strongly suggesting that H3K27me3 IHC is not of value in this often-difficult distinction. Loss of H3K27me3 expression is not seen in epithelioid MPNST and thus is not of value in the distinction of this tumor from melanoma. We have observed H3K27me3 expression in the angiosarcomatous component of MPNST showing heterologous angiosarcomatous differentiation, supporting the clonal nature of these unusual lesions ( Fig. 6.43C and D )