Tumors of soft tissue and bone are a diverse family, and categorization continues to evolve as more insight is gained into their patterns of differentiation and the underlying molecular pathogenesis of many distinct tumor types. As such, the classification of these neoplasms is an increasingly complex subject that requires at least a basic understanding of the biochemical attributes of the lesions in question. This chapter presents a practical summary of this topic; however, it does not purport to be an encyclopedic or all-inclusive treatise. In particular, Chapter 13 discusses tumors that are most common in the skin and subcutis, and we have excluded select lesions of soft tissues and bone that are morphologically distinct (i.e., they do not require immunohistochemistry for diagnosis). In the same vein, we emphasize that immunohistochemical (IHC) evaluation of mesenchymal neoplasms is merely an adjunct to thorough histologic evaluation, not a substitute for it. Furthermore, clinical and radiologic correlation and molecular studies all play significant roles in the evaluation of many soft tissue and bone lesions.
Biology of Antigens and Antibodies
Intermediate Filament Proteins
Intermediate filament proteins (IFPs) are structural components of all human cells, together with microfilaments and microtubules. They are 7 to 10 nm in diameter and are often arranged in skeins or bundles in the cytoplasm. Parallel aggregation of IFPs often is observed in epithelial cells that are rich in high-molecular-weight keratins, yielding the structures known to electron microscopists as tonofilaments or tonofibrils . Otherwise, the IFPs as a family are not morphologically distinguishable from one another at the ultrastructural level. Based on biochemical and functional grounds, they are composed of at least six distinct moieties: keratins, vimentin, desmin, neurofilament proteins (NFPs), glial fibrillary acidic protein (GFAP), and the lamins (nuclear envelope proteins). In this chapter, we will discuss the first five of these entities because they have been well characterized in diagnostic pathology.
All IFPs share structural homologies, but their precise natures vary considerably. Their molecular weights vary between 40 and 200 kD. IFPs also have dissimilar isoelectric pH values, and, more important, characteristic distribution patterns in nonneoplastic cells and human tumors. Two members of the IFP family, the keratins and NFPs, appear to be composed of heteropolymeric aggregates of two or more proteins, whereas the other members are homopolymers that contain only one protein isoform. The IFPs are encoded by multiple genes on various chromosomes (e.g., chromosomes 12q and 17q for the keratins, chromosome 2q for desmin, chromosome 10p for vimentin, and chromosome 17q for GFAP).
In keeping with their proposed cytoskeletal nature, IFPs initially were thought to serve a purely structural role in muscle cells. It was hypothesized that the function of these proteins was to keep other cytoplasmic proteins in proper relationship to one another and to anchor the cytoplasmic contractile apparatus to the cell membrane. However, subsequent developments in cell biology cast considerable doubt on this premise. The intermediate filaments are now known to serve a nucleic acid-binding function; moreover, they are susceptible to processing by calcium-activated proteases and are substrates for cyclic adenosine monophosphate-dependent protein kinases. Thus it has been proposed that all IFPs serve as modulators between extracellular influences that govern calcium flux into the cell, subsequent protease activation, and nuclear function at a transcriptional or translational level. Their morphologic associations with cell membranes and the perinuclear cytoplasm are consistent with this theory and relegate a cytoskeletal role to a secondary level. Current opinion favors the view that fibrils of the IFPs are formed to restrict the availability of their nucleic acid-binding domains in accord with cell-cycle activity, not as cellular “buttresses.” Although some IFPs insert into intercellular desmoplakin-containing desmosomes, which are responsible for maintaining tissue integrity, this fact does not necessarily assign a structural biologic role to these filaments. Intercellular junctions may be points of biochemical communication within the extracellular milieu. The distribution of each of the IFPs will be described in some detail; desmin will be discussed later in the section on markers of muscle differentiation.
As the essential IFPs of epithelial cells and epithelial neoplasms, keratins have a high degree of specificity and sensitivity for the diagnosis of carcinoma among malignant tumors. Cytokeratins typically are expressed by any given cell type in pairs, representing an acidic (type I) and a basic (type II) keratin. These vary in molecular weight from 40 to 67 kDa and have been given catalog designations by Moll and colleagues such that they are numbered within each respective type grouping from lowest to highest molecular weight. There are 12 type I keratins with acidic isoelectric points and 8 type II proteins with basic biochemical attributes. As described by Miettinen, cytokeratins tend to pair themselves during cell development so that a type I keratin is associated with a type II keratin that is 7 to 9 kD larger.
The particular keratin types that can be detected in given tissues or neoplasms follow predictable, known patterns of gene expression that serve, in part, to identify the cells containing them. With particular reference to nonepithelial cells, selected cytokeratins, CK8, CK18, and occasionally CK19, are demonstrable in the physiologic state, but special techniques such as acetone fixation, frozen section immunohistology, or amplified immunodetection methods are usually necessary to preserve or detect extremely low densities of these IFPs. Selected mesenchymal neoplasms may likewise exhibit keratin reactivity, which can be broader in scope. For example, CK1, CK7, CK8, CK13, CK14, CK18, and CK19 have all been observed in synovial sarcomas. Other soft tissue or bone tumors that are regularly cytokeratin-reactive include epithelioid sarcoma (EPS) (CK8, CK18, and CK19); chordoma (CK8, CK18, and CK19, with or without CK4 and CK5); myoepithelioma/myoepithelial carcinoma, including tumors previously known as parachordoma (CK8, among others); and adamantinoma (CK5 and CK19).
In reference to still other mesenchymal neoplasms, experience since the 1990s has shown that under some circumstances, selected tumors that are typically devoid of keratins may synthesize those IFPs in an aberrant fashion. Indeed, at this point, virtually all sarcomas have been reported to show this potential. Nevertheless, we would like to emphasize that aberrant keratin reactivity is most common in a narrow spectrum of malignant mesenchymal tumors, principally leiomyosarcoma (LMS); angiosarcoma, particularly epithelioid examples; and, to a lesser degree, Ewing sarcoma (ES) and alveolar rhabdomyosarcoma (A-RMS). In keeping with the comments mentioned earlier in this chapter, CK8, CK18, and CK19 are most often expected in this setting. It should be noted that aberrant IFP expression may also represent idiosyncrasies or flaws in IHC technique, wherein antikeratin antibodies are used at inappropriately high concentration or with especially sensitive detection procedures. Inasmuch as IFPs are structurally interrelated, cross-labeling of vimentin, desmin, NFPs, or GFAP can be obtained spuriously with many antikeratin reagents. A large body of clinical literature exists on the IFP profiles of human neoplasms. To benefit from the value of its contents in a diagnostic setting, the techniques used in obtaining these data must be retained. Thus in a practical sense, it is not advisable to substitute a new and improved immunodetection protocol for an old one without considering the effect the step has on the final interpretation. The diagnostic pathologist’s goal should not be to identify every molecule of a particular IFP, no matter how sparse, in any given tumor but rather to establish and maintain the windows of immunodetection to minimize the overlap between related moieties and maximize diagnostic utility. Attention to these maxims will lower the incidence of aberrant keratin expression in mesenchymal neoplasia when using routinely processed (formalin-fixed paraffin-embedded [FFPE]) clinical material.
NFPs are composed of three basic subunits with molecular weights of 68, 150, and 200 kD ; hence, they are larger than all other IFPs. Each NFP appears to be a separate gene product, rather than a derivative of the other two. Expression of this family of IFPs is correlated with differentiation of neurogenic blast cells into committed neurons in the developing embryo or in neoplasia, and each isoform is differentially expressed in different types of neurons. Another peculiarity of NFPs that is shared only by GFAP, another intermediate filament, is that each of the three neurofilament isoforms may be either phosphorylated or nonphosphorylated in vivo. Therefore antibodies to the NFPs may be specific for only one of those two configurations.
NFPs form a major component of the cytoskeleton of neurons and their axons. Expression is not seen in normal epithelium. Practically speaking, NFPs are generally not well detected in FFPE tissue, even with modern IHC methods and commercial antibodies. Among these, the Sternberger Monoclonal, Inc. (SMI) series of monoclonal antibodies and clone 2F11 are probably the most consistently active against routinely processed surgical pathology specimens. Among soft tissue neoplasms, NFP staining demonstrates axons in neurofibroma and ganglioneuroma; in the latter tumor, ganglion cells will also be positive for NFP. Neuroblastoma and its variants also express NFP, as do paragangliomas/pheochromocytomas and a subset of neuroendocrine tumors; the presence of a perinuclear dot-like pattern of staining is characteristically found in Merkel cell carcinoma, in which it may be a helpful diagnostic feature. NFP expression in other sarcomas is very limited.
Glial Fibrillary Acidic Protein
GFAP is another IFP that plays a somewhat limited role in the evaluation of soft tissue tumors. This 51-kD protein is the major component of astrocytes, ependymal cells, and retinal Müller cells and typically is not expressed by mature oligodendroglia. Nonglial tissues with putative GFAP reactivity include Schwann cells, myoepithelial cells, Kupffer cells, and some chondrocytes. It is therefore expected that selected neoplasms that include such elements, peripheral nerve sheath, myoepithelial, and cartilaginous tumors, may occasionally demonstrate immunolabeling for GFAP. As such, GFAP may be used as a second-line marker for malignant peripheral nerve sheath tumor (MPNST), although it shows only focal labeling in at most 30% of cases. Less frequently, GFAP is used to identify a component of Schwann cells in benign nerve sheath tumors. In addition, GFAP expression may help support the diagnosis of soft tissue myoepithelioma or myoepithelial carcinoma, being present in as many as 50% of myoepithelial tumors.
Vimentin is a 57-kD protein that was initially isolated from a mouse fibroblast culture. Its name derives from the Latin vimentum, describing an array of flexible rods. This IFP is the primordial member of the intermediate filament family because it is present in most, if not all, fetal cells early in development. Moreover, when two or more IFPs are coexpressed by a cell line or neoplasm, vimentin is virtually always one of them. Accordingly, vimentin is not considered to be cell-type specific. From the perspective of mesenchymal tumor pathology, it is of interest that vimentin shows a greater amino acid homology to desmin, NFPs, and GFAP than it does to the keratins.
The ubiquity of vimentin in soft tissues limits its diagnostic use in the setting of tumor pathology. However, it serves as a useful control marker function to ensure that the tissue has been properly preserved and processed, although other markers can also be used for this purpose. If vimentin cannot be easily detected in nonneoplastic endothelial cells, fibroblasts, and other mesenchymal elements routinely present in any tissue section, the reactivity or nonreactivity of accompanying neoplastic cells cannot be properly determined. Occasionally, the pattern of vimentin expression is also distinctive. For example, in malignant rhabdoid tumors (MRTs), this IFP usually assumes a densely globular cytoplasmic configuration, indenting the nuclei of the neoplastic cells.
Epithelial Membrane Antigen
Epithelial membrane antigen (EMA; MUC1) is a transmembrane glycoprotein encoded for by the MUC1 gene on chromosome 1. EMA is one of several human milk fat globule proteins (HMFGPs) derived from mammary epithelium. The HMFGPs vary greatly in molecular weight (51 to >1000 kD). They are glycoproteins that compose part of the plasmalemma of epithelial cells in areas of the cell membrane overlying tight junctions. In addition, because HMFGPs are packaged in the Golgi apparatus, globular labeling of this structure may be seen immunohistologically. The function of HMFGPs, including EMA, is still not absolutely certain. It is thought that they serve a role in secretion or, alternatively, that they provide a protective function for the cell.
The distribution of HMFGPs is such that many, but not all, nonneoplastic human epithelial cells express at least one member of this protein family. EMA expression is found in a variety of normal epithelial cells (breast, eccrine and apocrine glands, pancreas, urothelium, and renal collecting tubules) and nonepithelial cells (perineurial cells, plasma cells, arachnoid cells, ependyma, and choroid plexus). Exceptions include the gastrointestinal (GI) surface epithelium, endocervical epithelium, prostatic acinar epithelium, epididymis, germ cells, hepatocytes, adrenal cortical cells, rete testis, squamous cells of the epidermis, and thyroid follicular epithelium.
The most widely used monoclonal antibody to EMA is clone E29, to HMFPG. It labels a glycoprotein of approximately 450 kD. Neoplastic processes that may be EMA-immunoreactive are numerous. Synovial sarcoma, EPS, low-grade fibromyxoid sarcoma (LGFMS), perineuriomas, angiomatoid fibrous histiocytoma (AFH), epithelioid benign fibrous histiocytoma, chordoma, myoepithelioma/myoepithelial carcinoma, and some plasmacytomas are commonly labeled. Note that true EMA reactivity (i.e., that which generally equates with epithelial differentiation) must be localized to the cell membrane. Purely cytoplasmic labeling without a membrane component is a spurious pattern that should generally be ignored for diagnostic purposes. Expression of EMA is also seen in various carcinomas, mesothelioma, meningioma, and a subset of B-cell and T-cell lymphomas.
Other Epithelial Markers
When stains for standard epithelial determinants (e.g., keratin and EMA) yield equivocal results, it may be desirable to assess additional potential indicators of epithelial differentiation, especially if morphologic features suggest aberrant reactivity. The best known among the adjunctive epithelial markers are desmoplakin, desmoglein, and E-cadherin. The first two are desmocollins or elements of desmosomal complexes, which represent specialized intercellular anchoring structures. On the other hand, cadherins are calcium-dependent transmembranous intercellular adhesion molecules that are divided into three subclasses, E-, P-, and N-cadherin, and have distinctive immunologic specificities and tissue distributions. These molecules have subclass specificities in cell-cell binding and are involved in selective cellular adhesion. E-cadherin is typically associated with epithelial differentiation. Analysis of amino acid sequences, as deduced from the nucleotide sequences of complementary DNAs (cDNAs) that encode cadherins, has demonstrated that they share common sequences and are therefore regarded as a family of adhesion molecules with differential specificities. Although there are exceptions, concurrent immunoreactivity for desmoplakin or desmoglein and E-cadherin is generally restricted to epithelial neoplasms and mesenchymal tumors with epithelial characteristics, such as synovial sarcoma. One large study of epithelial-type and neural-type cadherin expression in soft tissue tumors showed expression of E-cadherin in 100% of biphasic synovial sarcomas (BSSs), 50% of monophasic synovial sarcomas (MSSs), and 8% of EPSs. N-cadherin expression was present in 100% of chordomas, 86% of BSSs, and 38% of EPSs. However, these markers are not widely used in the diagnosis of soft tissue tumors.
Markers of Muscle Differentiation
Desmin is a cytoplasmic IFP that is characteristically found in muscle cells and in neoplasms with myogenic differentiation. In smooth muscle cells, desmin is seen with cytoplasmic dense bodies and subplasmalemmal dense plaques; in striated muscle, desmin filaments are linked to sarcomeric Z disks. In both muscle types, desmin helps bind myofilaments into bundles.
In 1977, Small and Sobieszek were the first to recognize desmin as a distinct biochemical moiety. They found that it represented a residual filamentous protein in muscle cells that had been depleted of actin and myosin in vitro, and they assigned the provisional designation “skeletin” to it. It was observed to have an isoelectric point of approximately 4 and to be heat stable and insoluble in salt-rich solutions. Amino acid analysis revealed a high concentration of glutamate and aspartate and a significant chemical homology with glial filaments and NFPs. A notable finding in this study was that muscle cells depleted of skeletin (desmin) were still able to contract in response to adenosine triphosphate (ATP) and calcium. This point led the authors to conclude that the protein in question played no role in contractility but rather served to keep actin and myosin filaments associated and to anchor them to the plasmalemma.
These observations have been confirmed and expanded by others. Desmin is now known to have a molecular weight of 53 kD and a mass per unit of 36 to 37 kD/nm. It is composed of an N-terminal headpiece and a C-terminal tailpiece, both of which are nonhelical in conformation. These bracket an α-helical middle domain of approximately 300 amino acid residues. The former are greatly variable in biochemical constitution from species to species, but the helical segment is highly conserved, meaning that interspecies homology in this domain is striking. Similar to other intermediate filaments, desmin displays a 20- to 22-nm axial periodicity. Ip and Heuser showed that it forms heteropolymers that aggregate in a cross-linked, fibrillar, tetrameric fashion. The helical segment composition of desmin allows it to form coils with respect to the tertiary structure of the molecule. Hydrophobic amino acid residues are thereby exposed, which explains the ability of desmin to associate with nonhydrophilic nuclear and plasmalemmal membranes.
Desmin appears in developing striated muscle cells at the myotube-forming stage, in which myoblasts fuse with one another. It replaces vimentin, at least in large measure, because the latter is the intermediate filament first expressed by virtually all embryonic mesenchymal cells. Desmin filaments are oriented in a longitudinal fashion initially, but as the muscle cell matures, they become concentrated around Z disks. Not all muscle cells contain desmin, however; for example, mammalian aortic smooth muscle is typically negative for desmin, unlike the smooth muscle wall of most other blood vessels. Immunoelectron microscopic analyses have documented the binding of suitably specific antidesmins to the intermediate filaments of muscle cells and their neoplasms. There should be no cross-reactivity of such reagents with associated contractile proteins, such as actin and myosin; this is particularly important, because these three proteins appear to share some epitopes. The three best-characterized monoclonal antibodies to desmin are D33, DER-11, and DEB-5 . By the Western blot technique, they have been shown to recognize desmin epitopes between residues 324 and 415 and to have no cross-reactivity with other IFPs. These reagents show tissue specificity but species nonspecificity.
In general, desmin is, as expected, a specific marker for myogenic differentiation among soft tissue tumors. As such, it is seen in the majority of rhabdomyoma, leiomyoma, rhabdomyosarcoma (RMS), and LMS. Heterologous rhabdomyoblastic differentiation in other tumor types, such as MPNST (malignant Triton tumor), dedifferentiated liposarcoma (DDLPS), or carcinoma, will also show desmin positivity. Although the vast majority of leiomyomas express desmin, expression is seen in a smaller percentage of LMS (~70%), and expression in LMS can be limited in extent. Approximately 30% of cellular benign fibrous histiocytomas express desmin, presumably reflecting the myofibroblastic nature of fibrous histiocytoma. Other tumors that may have myofibroblastic features, and therefore may contain a subset of tumor cells that express desmin, include inflammatory myofibroblastic tumor (IMT), desmoid fibromatosis, angiomyofibroblastoma, and deep (“aggressive”) angiomyxoma. Reactivity for desmin may also be observed in neoplasms with divergent or uncertain phenotypes, including variable expression in perivascular epithelioid cell (PEC) tumors, so-called PEComas ; as much as 40% of ossifying fibromyxoid tumors (OFMTs) ; 50% of AFHs ; and approximately 80% of desmoplastic small round cell tumors (DSRCTs), often with a dot-like pattern.
Interdigitating reticulum cells of lymph nodes and a subset of reactive mesothelial cells are also positive for desmin. Other tumor types that occasionally express desmin include diffuse malignant mesothelioma and Wilms tumor.
Aside from desmin, the next most useful group of cytoplasmic determinants for defining myogenic differentiation is the protein family of the actins. The six major isoforms of these microfilamentous contractile polypeptides are designated skeletal muscle-α, smooth muscle-α, smooth muscle-γ, cardiac muscle-α, nonmyogenous-β, and nonmyogenous-γ. Alpha and gamma muscle isoforms are seen in tissues with pure myogenic differentiation, but they are also demonstrable in cells with myofibroblastic or myoepithelial features. The molecular weights of all these biochemical moieties cluster around 45 kD, and they may be labeled with antibodies that recognize conserved amino acid sequences or, alternatively, with isoform-selective reagents. From a diagnostic perspective, only the latter method is desirable. However, because of inevitable problems that arise in the immunohistologic detection of heteropolymeric proteins, even some of these antiactins are not truly specific for pure myogenic differentiation. This is particularly true of one commonly used commercial reagent, clone 1A4, which is widely known as anti-α-smooth muscle actin (SMA) . In practice, it is expressed in cell types other than smooth muscle, including myofibroblasts, myoepithelia, and others. In fact, nearly any neoplasm that shows spindle cell morphology may express SMA, including spindle cell (sarcomatoid) carcinomas. However, SMA is not detected in normal skeletal muscle; rarely, limited reactivity for SMA is seen in RMS. Myoepithelial tumors frequently express SMA, as do tumors of the PEComa family, which are characterized by dual myoid and melanocytic differentiation.
Another antibody, designated HHF-35, or anti-muscle-specific actin, shows more muscle-restricted immunoreactivity in routinely processed specimens. HHF-35 is expressed in both smooth muscle and skeletal muscle because it is targeted against the isoforms skeletal muscle-α, smooth muscle-α, and smooth muscle-γ. RMS, LMS, and benign smooth muscle and skeletal muscle tumors all express HHF-35. Myofibroblastic cells and tumors with myofibroblastic features, such as nodular fasciitis, show variable reactivity for HHF-35.
Smooth muscle myosin shows a similar staining profile to that of SMA, but the sensitivity appears to be lower, thus limiting its diagnostic utility. Sarcomeric actin is expressed in skeletal muscle tumors and to a lesser extent in smooth muscle tumors, but its use in clinical practice is limited because of low sensitivity and specificity.
Other Sarcomeric Contractile Proteins
The contractile mechanism in skeletal muscle is transduced by a complex of proteins that includes myosin II (molecular weight 460 kD), actin, tropomyosin (molecular weight 70 kD), and troponin. The troponin molecule has three subunits, troponin I, troponin T, and troponin C, with molecular weights between 18 and 35 kD. Myosin is an actin-binding protein that has two globular heads and an elongated tail. In particular, myosin II is composed of two heavy chains and four light chains (two phosphorylatable and two basic). These combine with N-terminal portions of the myosin heavy chains to form globular heads, each of which has an actin-binding site and an enzymatic locus that hydrolyzes ATP. The heads of the myosin molecules form cross-bridges to actin. Myosin molecules are configured in a symmetric fashion on either side of the center of the sarcomere. Sarcomeric thin filaments are polymers composed of two actin chains arranged in a double helix. Tropomyosin molecules, in turn, are situated in the groove between the two chains of actin, and troponins are interspersed along the tropomyosin. Troponin T melds other troponin components with tropomyosin; troponin I inhibits the interaction of myosin and actin, and troponin C contains binding sites for calcium in the initiation of muscle contraction. Actinin, a 190-kD moiety, binds actin to the Z lines of the sarcomere. Another protein, titin, connects Z lines to M lines and provides the base on which thick filaments may form. Because of their relatively poor sensitivity, the markers discussed in this section are used only in rare circumstances for diagnostic purposes.
Caldesmon is a cytoplasmic protein with two isoform classes, one of which is found predominantly in smooth muscle cells and other cell types with partial myogenic differentiation. High-molecular-weight isoforms, those with molecular weights between 89 and 93 kD, are capable of binding to actin, tropomyosin, calmodulin, myosin, and phospholipids, and they function to counteract actin-tropomyosin-activated myosin adenosine triphosphatase (ATPase). As such, they are mediators for the inhibition of calcium-dependent smooth muscle contraction.
Commercial antibodies to (“heavy”) h-caldesmon are of clinical utility in diagnostic surgical pathology. They appear to be relatively specific for smooth muscle differentiation and, as such, are useful adjuncts to desmin and actin immunostains in the evaluation of leiomyoma and LMS. The vast majority of leiomyomas express h-caldesmon, but expression in LMS is more variable, with one study showing reactivity for h-caldesmon in only 36% of LMS. Caldesmon expression is not entirely specific to smooth muscle tumors, however, because reactivity is also observed in the majority of gastrointestinal stromal tumors (GISTs), glomus tumors, and myopericytomas. In contrast, myofibroblastic cells do not express caldesmon.
Calponin is an actin-binding and tropomyosin-binding cytoskeleton-associated protein involved in the regulation of smooth muscle contraction. It is expressed in smooth muscle and myoepithelial cells and myofibroblasts. Reactivity for calponin is observed in leiomyoma, LMS, nodular fasciitis and other myofibroblastic tumors, and myoepithelial neoplasms, but GISTs tend to be negative for calponin.
Markers of Skeletal Muscle Differentiation
Myoglobin is a 17.8-kD protein found exclusively in skeletal muscle, where it forms complexes with iron molecules to transport oxygen. The concentration of this molecule is highest in muscles that undergo sustained contraction. Because myoglobin appears relatively late in the maturational sequence of striated muscle, it is typically undetectable immunohistologically in embryonic neoplasms that show differentiation toward that tissue. Accordingly, pleomorphic (adult-type) RMS and rhabdomyoma are the soft tissue tumors in which myoglobin is identified most often. Staining for myoglobin is cytoplasmic. Most available antibodies are polyclonal, and background staining often limits the clinical utility of this marker, as does its limited expression in RMS subtypes.
MYOD1 and Myogenin
A superfamily of transcription factors that regulates cell lineage-specific proliferation is represented in striated muscle by several moieties collectively known as the MyoD family . They are encoded by genes that reside on chromosomes 1, 11, and 12 and are part of a polypeptide complex called the basic helix-loop-helix (BHLH) motif, small proteins composed of 220 to 320 amino acids. Two members of this intranuclear protein group, MYOD1 (46 kD; MYF3) and myogenin (32 kD; MYF4), have been used since the 1990s as specific markers of striated muscle differentiation in human neoplasms. They activate their own transcription and that of other BHLH proteins and, in concert with the retinoblastoma (Rb) gene, govern exodus from the cell cycle and initiation of striated muscle differentiation. MYOD1 and myogenin are expressed by fetal skeletal muscle cells and regenerating skeletal muscle cells, but normal adult skeletal muscle is negative for both markers.
Because antibodies to MYOD1 and myogenin must gain access to the nucleus, they have been difficult to use in routine surgical specimens. However, modification of antigen retrieval solutions and utilization of heat-mediated epitope enhancement have allowed these reagents to enter diagnostic use. It is important to note that MYOD1 and myogenin are strictly localized to cellular nuclei; therefore background cytoplasmic labeling, a problem with some antibodies to MYOD1, must be ignored as a spurious pattern of staining. In contrast, stains for myogenin do not generally show this problem. In contrast to myoglobin, expression of MYOD1 and myogenin is greatest in less differentiated RMS, such as alveolar (greatest amount) and embryonal subtypes, and typically less in the pleomorphic subtype or in tumors that show cytodifferentiation after chemotherapy. Spindle cell RMS typically shows more extensive expression of MYOD1 than myogenin. The specificity of these markers for skeletal muscle differentiation is high. In other tumor types, expression is limited to heterologous rhabdomyoblastic differentiation in other tumors (e.g., MPNST and DDLPS). Care should be taken when interpreting MYOD1 and myogenin immunostains, because positivity limited to reactive skeletal muscle cells may lead to an erroneous diagnosis of RMS.
S100 protein derives its name from the fact that it is soluble in saturated (100%) ammonium sulfate solution. It was first isolated from the central nervous system (CNS) but is now known to have a wide distribution in human tissues, including glia, neurons, chondrocytes, Schwann cells, melanocytes, fixed phagocytic or antigen-presenting cells, Langerhans cells, myoepithelial cells, notochord, and various epithelia but especially in the breast, salivary glands, sweat glands, and the female genital system. S100 protein is dimeric in nature, with α and β subunits. Hence, it has three isoforms: S100ao (α-dimer), S100a (α-β isoform), and S100b (β-dimer). Each of the two subunits of this protein has a molecular weight approximating 10.5 kD, and the function of S100 protein is essentially that of a calcium flux regulator.
Both monoclonal antibodies and heteroantisera to S100 protein are available for diagnostic use. Some of the former reagents are monospecific for the α or β subunits; therefore they exhibit relatively narrower spectra of reactivity that than seen with polyclonal antisera. For example, β-subunit-specific antibodies preferentially label glial cells and Schwann cells. However, β-subunit-specific antibodies have not enjoyed widespread use among diagnostic pathologists, and hetero-antisera to S100 protein are most commonly used in clinical practice. In the proper context, as part of panels of antibodies designed to evaluate several possible lineages in a morphologically indeterminate neoplasm, reagents against S100 protein are still valuable indicators of schwannian or melanocytic differentiation in tumors of the soft tissues and bone. Expression of S100 protein is usually both nuclear and cytoplasmic. S100 protein is detected in more than 90% of clear cell sarcomas (CCSs) of tendons and aponeuroses, although expression may sometimes be limited in extent. Schwannomas are diffusely and strongly positive for S100 protein, whereas neurofibromas show more variable reactivity (i.e., usually in 50% to 80% of constituent cells). S100 protein expression in MPNST is usually limited, and only 30% to 50% of cases show demonstrable staining. S100 protein expression in normal adipocytes is variable; although S100 protein may sometimes highlight lipoblasts in liposarcomas, its utility in this regard is limited. Other soft tissue tumors that express S100 protein include OFMT (73% to 94%), extraskeletal myxoid chondrosarcomas (EMCSs) (at most 20%), synovial sarcoma (30%), and 5% to 10% of GISTs (most often observed in duodenal tumors). S100 protein expression is found in 90% of soft tissue myoepitheliomas and in around 75% of myoepithelial carcinomas. Among bone tumors, S100 protein is consistently detected in well-differentiated cartilaginous neoplasms and in as much as 80% of chordomas. Reactivity for S100 protein is also seen in as much as 40% of breast carcinomas and less frequently in renal cell carcinomas (RCCs) and carcinomas of müllerian origin.
S100 protein is also useful in supporting the diagnosis of some histiocytic and dendritic cell lesions. In Rosai-Dorfman disease, S100 protein highlights the lesional cells, with particularly strong staining in the cytoplasm, and it can make emperipolesis more easily appreciable. Langerhans cell histiocytosis is also consistently positive for S100 protein. Similarly, the tumor cells of histiocytic sarcoma often express S100 protein (in ~50% of cases), as do the extremely rare interdigitating dendritic cell sarcomas.
SOX10 (sex-determining region Y-related HMG-box) is a neural crest transcription factor that plays a critical role in the differentiation, maturation, and maintenance of Schwann cells and melanocytes. Expression of SOX10 is seen in neural crest derived cells, and nuclear staining is therefore seen in normal melanocytes, Schwann cells, myoepithelial cells, and acinar cells of salivary gland tissue. Tumors exhibiting neural crest differentiation, that is, melanocytic and nerve sheath tumors, granular cell tumors, as well as a subset of myoepithelial and salivary gland type tumors, show expression of SOX10.
SOX10 shows similar sensitivity for MPNST as S100 protein, with nuclear expression observed in 30% to 50% of MPNSTs. In contrast, the majority of benign nerve sheath tumors and more than 90% of melanocytic neoplasms (including CCS) show expression of SOX10. Expression of this marker should therefore be interpreted in context.
However, expression of SOX10 in other mesenchymal and epithelial tumors is limited, making it in many ways more helpful than S100 protein; expression has been reported to occur in diffuse astrocytomas, carcinoid tumors, and some ductal breast carcinomas. Sustentacular cells of paraganglioma/pheochromocytoma and a subset of carcinoid tumors also express SOX10. In practice, the greatest utility of SOX10 is its utility in confirming melanocytic differentiation, particularly in cases of metastatic melanoma with limited or no expression of S100 protein.
Potential Markers of Nerve Sheath Differentiation
CD56 (Neural Cell Adhesion Molecule) and CD57 (NKH1/LEU19 and LEU7/HNK-1)
CD56 and CD57 are membrane antigens expressed in peripheral blood mononuclear leukocytes, a proportion of which have natural killer (NK) activity. Their respective molecular weights are 140 and 95 kD. Antibodies in these cluster designations also react with several neural molecules that have a variety of molecular weights. Some of these moieties are associated with 5′-nucleotidase activity, whereas others are the neural cell adhesion molecule and myelin-associated glycoproteins (MAGs). The largest of the MAGs, MAG-72, is related structurally to the immunoglobulin superfamily gene products and to neural adhesion molecules and the autophosphorylation site of epidermal growth factor receptor (EGFR). MAGs are integral cell membrane proteins normally found in oligodendroglia. It is believed that their function involves the mediation of interaxonal or axonal-glial interaction during myelination. As such, their additional association with Schwann cells and neural neoplasms should not be surprising. Nevertheless, CD57 reactivity has also been documented in perineurial (non-schwannian) peripheral nerve sheath lesions.
In general, CD56 and CD57 have been used as potential markers not only of peripheral nerve sheath tumors (PNSTs) but also of ES, in which matrical proteins of neurosecretory granules and synaptic vesicles are thought to contain a target protein for the antibodies in question.
However, CD56 and CD57 are not restricted to nerve sheath cells or neuroectodermal elements among soft tissue tumors but rather are most often observed in those cell types. Synovial sarcoma, LMS, and some metastatic carcinomas may also express both of these markers. Because of their lack of specificity, the authors of this chapter do not use these markers in clinical practice for diagnosing soft tissue tumors.
Collagen Type IV and Laminin
Volumetrically, collagen type IV is the predominant component of basement membranes, regardless of which cell types they invest. This triple-helical molecule weighs 550 kD and has globular end regions and two noncollagenous domains. One of the latter is located 330 nm from the carboxy end of the molecule, where a bending point gives the moiety a hockey stick configuration. Type IV collagen differs from other collagen types in that it does not form fibrils, it shows interruptions of its helical structure, and it has a different amino acid constituency. Genes coding for the helical chains of this molecule are located on chromosome 13q.
Laminin is another important component of basement membranes. It is a 1000-kD molecule that binds to glycosaminoglycans, acting as a bridge for attachment of collagen type IV in basement membranes to the surrounding matrix. The exact location of laminin in basement membranes has been contentious. Some investigators claim that it is part of the lamina densa, others suggest that it resides in the lamina lucida, and yet others believe that it is codistributed between these two compartments. Beyond simple boundary and anchoring functions, laminin probably also influences intercellular interactions and contributes to alterations in cellular morphology.
In soft tissues, complete basement membranes are formed around endothelial, smooth muscle, and Schwann cells. Thus reagents directed against collagen type IV and laminin were traditionally included in antibody panels aimed at detecting those lineages, in particular, to help distinguish MPNSTs from some other spindle cell sarcomas. The use of both of these markers in current practice is diminishing as a result of the availability of more sensitive and specific markers.
The claudins are a family of approximately 18 proteins that play important structural and functional roles in tight junctions. They are transmembrane proteins that interact with other transmembrane proteins, such as junctional adhesion molecule (JAM) and occludin, as well as the scaffolding proteins ZO-1, ZO-2, and ZO-3. Members of the claudin family are differentially expressed in various cell types: claudin-3 is expressed primarily in lung and liver epithelia, whereas claudin-5 is expressed principally in endothelial cells; claudin-1 expression is widespread among epithelial cells, but in mesenchymal tissues, expression appears limited to perineurial cells. Claudin-1 expression has been reported in perineuriomas and occurs in 29% to 92% of tumors. Claudin-1-positive (perineurial) cells may also be found in neurofibromas, often at the periphery of the tumor, and in the capsule of schwannomas. Endothelial cells variably expressed claudin-1 without a consistent staining pattern; expression is not seen in normal Schwann cells, fat, smooth muscle, skeletal muscle, or fibroblasts. Claudin-1 is consistently negative in desmoplastic fibroblastoma, dermatofibrosarcoma protuberans (DFSP), and fibromatosis. Although claudin-1 expression has been reported in cases of LGFMS with perineurioma-like features, in the authors’ experience, LGFMS is consistently negative for claudin-1, irrespective of the histologic appearances (unpublished data); lack of demonstrable claudin-1 expression in LGFMS was also reported in the initial paper describing claudin-1 expression in perineurioma.
Markers of Endothelial Differentiation
Several markers associated with endothelial cells have been applied to the recognition of vascular neoplasms of soft tissue. The varying degrees of sensitivity and specificity of these markers are discussed below.
Factor VIII-Related Antigen (Von Willebrand Factor)
Factor VIII-related antigen, or von Willebrand factor (vWF), is a very large polymeric protein synthesized exclusively by endothelial cells and megakaryocytes. It consists of three multimeric subunits more than 10,000 kD in molecular weight; physiologically, they undergo proteolysis to yield substantially smaller fragments that can be found in plasma. The function of vWF is twofold. First, it forms circulating complexes with factor VIII coagulant protein, a 265-kD protein that effects the activation of factor X in the intrinsic coagulation pathway. Second, vWF plays a crucial role in platelet aggregation such that patients with low levels or dysfunctional variants of this protein have the clinical bleeding diathesis known as von Willebrand syndrome .
In the context of soft tissue pathology, vWF has been used principally to distinguish vascular neoplasms from their morphologic simulants. Because vWF is packaged within Weibel-Palade bodies (WPBs) in endothelial cells, it is logical to expect that immunoreactivity for that analyte would parallel the ultrastructural presence of such organelles. This is indeed the case, and because WPBs are rare in poorly differentiated endothelial neoplasms, this explains why the sensitivity of vWF is low (~10% to 15%) for the recognition of morphologically high-grade angiosarcoma. Accordingly, this marker is more consistently expressed in the spectrum of benign and locally aggressive endothelial tumors, such as hemangioma variants and hemangioendotheliomas. As would be expected given the cellular location of WPBs, expression of vWF is typically granular and cytoplasmic. Although expression of vWF is considered to be 100% specific for endothelial cells, because vWF is secreted into serum and can be found in fibrin thrombi and areas of hemorrhage, interpretation of this stain can be difficult; in fact, vWF has largely been abandoned in clinical practice because of the availability of more sensitive and more easily interpretable markers of endothelial differentiation, such as CD31 and V-ETS avian erythroblastosis virus E26 oncogene homolog (ERG).
CD34, the hematopoietic progenitor cell antigen, is recognized by several monoclonal antibodies that include My10/HPCA1, QBEND-10, and BI-3C5. CD34 is a 110-kD transmembrane glycoprotein that is expressed by embryonic cells of the hematopoietic system, including both lymphoid and myeloid cells and endothelial cells. Correspondingly, again in the setting of soft tissue tumors, CD34 is a potential indicator of vascular differentiation. It is a relatively sensitive marker for the endothelial lineage and recognizes more than 85% of angiosarcomas and Kaposi sarcomas (KSs). However, in the authors’ experience, reactivity for CD34 is observed in only approximately 60% of poorly differentiated angiosarcomas, which limits its diagnostic utility in this setting. Moreover, the specificity of CD34 is low, inasmuch as CD34 expression is seen in some LMS, PNSTs, and EPSs, which could potentially simulate variants of angiosarcoma or hemangioendothelioma. In addition, CD34 is so commonly detected in DFSP and its variants; in spindle cell lipoma, GIST, and in solitary fibrous tumor (SFT) that it is regularly used as an adjunct to the diagnosis of these tumor types. Although 70% to 90% of spindle cell GISTs express CD34, most frequently in rectal and esophageal tumors, at most only 50% of epithelioid GISTs and GISTs of the small intestine are CD34 positive.
Approximately 20% to 30% of retroperitoneal LMS expresses CD34. Diffuse CD34 expression is characteristic of DFSP and can help distinguish this tumor type from benign fibrous histiocytoma/dermatofibroma, which is typically negative for CD34 (although around 5% of benign fibrous histiocytomas may show focal staining). However, when fibrosarcomatous transformation of DFSP occurs, CD34 expression is often lost in the higher-grade component. CD34 expression is observed in 65% of soft tissue perineuriomas and highlights the delicate cytoplasmic processes of the lesional cells. Thus as endothelial markers, antibodies to CD34 are best used in a panel of reagents designed to account for these other diagnostic possibilities when appropriate.
CD31, also known as platelet-endothelial cell adhesion molecule 1 (PECAM-1), is a 130-kD transmembrane glycoprotein expressed by endothelial cells, megakaryocytes, platelets, and histiocytes. CD31 expression can also be found on some myeloblasts, plasma cells, and lymphocytes, and is recognized by the monoclonal antibody JC/70A. This marker is highly restricted to endothelial neoplasms among soft tissue tumors, where it shows a membranous pattern of staining, and its sensitivity is generally excellent. In our hands, more than 90% of angiosarcomas are CD31 positive, regardless of morphologic grade or subtype, and CD31 expression is consistently detected in hemangioma and hemangioendothelioma variants. It must be acknowledged, however, that KS is labeled more consistently for CD34 than for CD31. The greatest pitfall in the interpretation of CD31 expression in tumors is because CD31 also commonly labels macrophages, which may be present in large numbers, dispersed among tumor cells. Not surprisingly, CD31 expression is also seen in histiocytic sarcomas. Less specifically, limited expression of CD31 can occasionally be found in otherwise undifferentiated pleomorphic sarcomas and very rarely in carcinomas and mesotheliomas.
Friend Leukemia Virus Integration 1
Human friend leukemia virus integration 1 (FLI1) is a member of a family of transcription factor proteins that share a conserved DNA-binding region, the E26 transformation-specific (ETS) domain. The peptide sequence is 98 amino acid residues long, and it bears a molecular resemblance to the helix-turn-helix motif of DNA-binding proteins. FLI1 is a sequence-specific transcriptional activator involved in cell proliferation. It recognizes the DNA sequence 5′-C(CA)GGAAGT-3′ and is encoded by a gene on the long arm of chromosome 11 (11q24). In rodents, Fli1 is a common region for viral nucleic acid integration in virus-driven leukemogenesis. In all vertebrates, this gene also appears to act at the top of the transcriptional network that governs the development of hematopoietic precursors and endothelial cells. Landry and colleagues have shown that this effect occurs by regulation of the proximal promoter of the LMO2 gene in endothelia. FLI1 is usually expressed in the nuclei of lymphocytes and endothelial cells; thus care must be taken in the interpretation of this stain to not overinterpret the presence of FLI1-positive cells.
The best-known association between FLI1 and human tumors relates to its fusion with the EWSR1 gene on chromosome 22 (22q12) in most cases of ES. The resultant fusion protein may play a role in the evasion of cellular senescence. FLI1 positivity has also been found in melanoma, synovial sarcoma, Merkel cell carcinoma, and lung adenocarcinoma.
Folpe and colleagues were the first to study FLI1 immunoreactivity in human endothelial neoplasms, including hemangiomas, hemangioendotheliomas, angiosarcoma, and KS. They observed a sensitivity of 94% in that specific context and found no labeling of nonvascular tumors. However, more recently, Mhawech-Fauceglia and colleagues have documented FLI1 positivity in some carcinomas, lymphomas, and RMS as well. In a study of cutaneous tumors of the head and neck, all angiosarcomas examined showed strong diffuse nuclear expression of FLI1, but FLI1 reactivity of variable intensity was observed in 87% of squamous cell carcinomas, 92% of atypical fibroxanthomas, 59% of melanomas, and 20% of atypical intradermal smooth muscle neoplasms. In that study, the sensitivity of FLI1 for angiosarcoma was 100% with a specificity of only 29% if any staining in other tumors was considered positive and 76% if only moderate or strong staining was used. In aggregate, these data do not support the use of FLI1 as an adjuvant to the diagnoses of either ES or vascular neoplasms.
V-E26 Transformation-Specific Avian Erythroblastosis Virus E26 Oncogene Homolog
Similar to FLI1, ERG is a member of the ETS family of transcription factors. In human fetal tissues, ERG is also expressed in a nuclear pattern in a subset of primitive mesenchymal cells but subsequently becomes limited to vascular endothelium and a subset of normal hematopoietic stem cells/immature myeloid cells. ERG expression is an extremely sensitive and specific marker of endothelial differentiation, and a rabbit anti-ERG monoclonal antibody is commercially available. ERG is ubiquitously expressed in the nuclei of normal endothelium, which serves as an internal control. Nuclear expression of ERG has recently been reported in 100% of hemangiomas, 98% of epithelioid hemangioendotheliomas (EHEs), and 96% of angiosarcomas.
All cutaneous angiosarcomas expressed ERG in two large studies, regardless of the degree of morphologic differentiation, usually in a strong and diffuse pattern. The majority (90%) of deeply situated angiosarcomas were also reactive for ERG. In a study of a variety of different types of primary cutaneous tumors, some of which may mimic angiosarcoma, all atypical fibroxanthoma, squamous cell carcinoma, melanoma, and atypical intradermal smooth muscle neoplasms examined were negative for ERG.
Among other mesenchymal tumors, approximately 5% to 10% of ES show expression of ERG, specifically those with the EWSR1-ERG fusion gene. Variable staining for ERG occurs in EPS, depending on the antibody used. Antibodies against the N-terminus show expression in up to 68% of cases, whereas antibodies against the C-terminus result in expression in less than 5% of cases. Expression of ERG in EPS is therefore a potential pitfall in the differential diagnosis with EHE, but loss of INI1 expression in EPS can usually help resolve this differential diagnosis. In general, the monoclonal antibody for ERG should be used, as some antibodies react with other ETS family members (e.g., FLI1).
Using an antibody against the N-terminus, expression of ERG has also been described in a variety of different cartilaginous tumors, including soft tissue chondroma, chondromyxoid fibroma, chondroblastic osteosarcoma, clear cell chondrosarcoma (CCC), conventional chondrosarcoma, and a subset of chondroblastoma.
Regarding nonmesenchymal tumors, nuclear ERG expression is observed in approximately 50% of prostatic adenocarcinomas as a result of the presence of the TMPRSS2-ERG fusion gene in these carcinomas. Of note, ERG is not expressed by normal prostatic epithelium. Nuclear expression of ERG in other epithelial neoplasms is very rare; in the largest study to date of ERG expression, one large cell carcinoma of the lung and one pleural mesothelioma showed focal nuclear staining for ERG, out of 643 (549 nonprostatic) epithelial tumors examined. Expression of ERG also occurs in blastic extramedullary myeloid tumors. Although cytoplasmic ERG staining has been described in some nonendothelial tumors, we have not observed this pattern in our practice.
ERG is more sensitive and specific for vascular tumors than CD31, CD34, and FLI1.
More than 90% of all angiosarcomas are positive.
ERG is positive in approximately 50% of prostatic adenocarcinomas and, rarely, in other tumors that have an ERG -gene fusion (e.g., Ewing sarcoma).
Depending on the antibody used, expression may also be seen in epithelioid sarcoma and a variety of chondrogenic tumors.
ERG , E26 oncogene homolog, FLI1 , friend leukemia virus integration 1.
Glucose Transporter Type 1
Glucose transporter type 1 (GLUT-1) is an erythrocyte-type glucose transporter protein and a member of the facilitative cell-surface glucose transporter family. GLUT-1 is present on brain capillary endothelium where it functions in the transport of glucose across the blood-brain barrier. GLUT-1 also plays an important role in the cellular response to hypoxia as a downstream target of hypoxia-inducible factor 1-α (HIF1-α). GLUT-1 upregulation and subsequent overexpression of GLUT-1 receptors on the plasma membrane of various tumor cells are thought to allow escape from the apoptosis-inducing effects of a hypoxic environment. A rabbit polyclonal antihuman GLUT-1 is commercially available.
Constitutive expression of GLUT-1 has been documented in a variety of normal cell types, including placental trophoblast and perineurial cells, and upregulation of GLUT-1 expression has been reported in various carcinomas, including those of urothelial, breast, colonic, pancreaticobiliary, esophageal, renal, pulmonary, and ovarian surface epithelial origin, among others. In mesenchymal neoplasms, expression of GLUT-1 has been shown to be a constant feature of juvenile capillary hemangiomas, and its expression may be useful in distinguishing such tumors from various mimics, such as vascular malformations and kaposiform hemangioendothelioma (KHE). GLUT-1 expression has also been noted in a subset of soft tissue perineuriomas. However, expression appears to be widespread in mesenchymal neoplasms. In a large study of various mesenchymal tumors, Ahrens and colleagues found GLUT-1 expression in chordoma (100%), EPS (63%), GIST (14%), LMS (40%), ES (27%), synovial sarcoma (30%), and undifferentiated pleomorphic sarcoma (60%). Of note, staining in tumor cells adjacent to areas of necrosis was a common finding; therefore use of GLUT-1 as a discriminatory biomarker appears to be limited, but it may be helpful in selected settings.
Prospero-Related Homeobox 1
Prospero-related homeobox 1 (PROX1) is a nuclear transcription factor encoded for by the homeobox gene PROX1 , which functions in lymphatic development during embryogenesis. A monoclonal antihuman antibody against PROX1 has been reported in two studies of endothelial neoplasms. PROX1 was expressed in 58% of all vascular tumors examined, including 42% of hemangiomas, 100% of lymphangiomas, 47% of EHEs, 92% of retiform hemangioendotheliomas, 86% of cutaneous angiosarcomas of the head and neck, and 20% to 36% of deep soft tissue or visceral angiosarcomas. Reactivity for PROX1 was also found in 25% of ESs, 33% of paragangliomas, and 19% of synovial sarcomas, as well as in various carcinomas. The utility of PROX1 as a discriminatory marker for endothelial differentiation therefore appears to be limited by a lack of specificity and sensitivity; however, it may be of use in suggesting lymphatic differentiation in endothelial lesions.
Claudin-5 is a transmembrane tight-junction protein that functions as a barrier at epithelial and endothelial cell junctions. Expression of claudin-5 is seen in endothelium and in some epithelial cell types, such as glomerular podocytes. Although claudin-5 is a sensitive marker for detecting tumors with endothelial differentiation, it is nonspecific, because expression is also seen in many different epithelial tumors. Available antihuman antibodies for claudin-5 include a rabbit polyclonal antibody and a mouse monoclonal antibody (clone 4C3C2).
In normal tissues, claudin-5 is expressed by endothelial cells of vessels of varying caliber, but it tends to be strongest in capillaries, lymph node sinusoidal endothelium, and high endothelial venules. Lymphatics of the intestinal mucosa may stain weakly for claudin-5. Miettenin and colleagues evaluated expression patterns of claudin-5 in a large study of normal tissues, vascular tumors, mesenchymal neoplasms, and epithelial tumors. Nonendothelial tissues that express claudin-5 include hair follicle epithelium, sweat glands, surface epithelium of the luminal GI tract and pancreatic ducts, bile ducts, prostatic glandular epithelium, thyroid follicular epithelium, glomerular capsular and tubular epithelium, tonsillar crypt epithelium, and ductal epithelium of the breast. Histiocytic and lymphoid cells are negative for claudin-5.
The pattern of staining for claudin-5 is cytoplasmic and/or membranous, although epithelioid vascular tumors produce a predominantly membranous pattern of staining. Expression of claudin-5 is found in 96% of angiosarcomas, 97% of KSs, 88% of EHEs, 100% of retiform hemangioendotheliomas, and 100% of hemangiomas (capillary, cavernous, spindle cell, verrucous, intramuscular, capillary, juvenile, and venous) and lymphangiomas. In spindle cell hemangioma, the spindle cells are negative for claudin-5. In juvenile capillary hemangioma, staining is usually limited in extent; in EHE, reactivity may be variable within an individual tumor. Angiosarcoma typically shows strong diffuse expression in tumor cells regardless of the degree of differentiation, but in occasional cases, staining may be weak or focal.
Claudin-5 expression is also found to varying degrees in lung, gastric, colonic, pancreatic, prostatic, ovarian, and endometrial adenocarcinomas; low-grade ductal breast carcinomas; well-differentiated cutaneous squamous cell carcinomas; esophageal squamous cell carcinomas; large cell undifferentiated lung carcinoma; and in the epithelial component of BSS.
Thrombomodulin (CD141) is a 75-kD cytoplasmic glycoprotein that is distributed among endothelial cells, mesothelial cells, osteoblasts, mononuclear phagocytic cells, and selected epithelia. Its physiologic role is to convert thrombin from a coagulant protein to an anticoagulant. It has been shown that thrombomodulin is a sensitive indicator of endothelial differentiation, particularly in poorly differentiated malignant vascular neoplasms. KS is likewise immunoreactive for this marker. However, because of the potential presence of thrombomodulin in some metastatic carcinomas and most mesotheliomas, both of which may be confused with epithelioid angiosarcomas, it cannot be used as a single marker for vascular neoplasms.
Ulex europaeus I Agglutinin
Ulex europaeus I (UEAI) agglutinin is not an antibody reagent; instead it represents a lectin produced by the gorse plant. It recognizes the Fuc-α-1-2-Gal linkage in fucosylated oligosaccharides, which compose portions of various glycoproteins. In particular, the H blood group antigen and carcinoembryonic antigen (CEA) regularly bind to UEAI, as does another fucosylated protein expressed by endothelial cells. Biotinylated Ulex may be used as a histochemical reagent in surgical pathology or, alternatively, unlabeled lectin can be used, with its binding to tissue subsequently detected by application of biotinylated anti- Ulex and avidin-biotin-peroxidase complex. Because of the low specificity of UEAI for endothelial differentiation, as mentioned earlier in this discussion, and because in addition to vascular neoplasms, EPS and various metastatic carcinomas may also bind Ulex , UEAI is now rarely used in clinical practice.
The antibody D2-40 recognizes podoplanin, also known as AGGRUS, gp36, M2A, and T1A-2. It is a transmembrane glycoprotein encoded by a gene on the short arm of chromosome 1 (1p36.21) and is expressed in various tissues. Breiteneder-Geleff and colleagues originally described this moiety in glomerular podocytes of the rat, but it has since been observed in lymphatic endothelium, mesothelium, various epithelia, follicular dendritic cells, and germ cells in several species, including humans. Tumors that show differentiation toward these lineages may therefore also be podoplanin-positive, such as mesothelioma, seminoma, follicular dendritic cell sarcoma, and tumors of skin adnexa. Podoplanin is principally expressed during vertebrate development in lymphatic endothelial cells and is therefore thought to be a selective marker for lymphatic channels. Overexpression of podoplanin significantly increases endothelial cell adhesion, migration, and vascular lumen formation.
In the context of soft tissue neoplasia, podoplanin may be used as a determinant of lymphatic differentiation in the evaluation of vascular neoplasms, although expression is also found in a subset of angiosarcomas, KHE, and KS. This finding may reflect lymphatic differentiation within these tumors. Given the relative lack of specificity of D2-40 for vascular tumors, the utility of this marker is somewhat limited. However, immunolabeling with D2-40 may be helpful in delineating angiolymphatic invasion by carcinomas and melanomas.
Wilms Tumor 1
The Wilms tumor 1 gene, WT1, is located on the short arm of chromosome 11 (11p13). It encodes a protein that is a critical determinant of urogenital development and is expressed in more than 80% of nephroblastomas. With regard to osseous and soft tissue tumors, WT1 protein is not a specific marker and is potentially present in angiosarcoma, MPNST, synovial sarcoma, osteosarcoma, myxoid liposarcoma (MLPS), and CCS. Nevertheless, it is occasionally used in diagnostic practice, in a structured panel with other immunostains, as an adjunctive indicator of endothelial differentiation.
Vascular Endothelial Growth Factor Receptor 3
Vascular endothelial growth factor receptor 3 (VEGFR3) is a transmembrane protein also known as FMS-like tyrosine kinase 4 (FLT4). VEGFR3 is encoded by a gene at chromosomal locus 5q33. Initially thought to be expressed predominantly in lymphatic endothelia, VEGFR3 positivity is found in more than 90% of KS, papillary intralymphatic angioendothelioma (Dabska tumor), and retiform and KHE; 50% of angiosarcomas; and 15% of hemangiomas. VEGFR3 is also expressed by nonneoplastic capillaries within glomeruli and endocrine organs and in nonneoplastic capillaries in some carcinomas and sarcomas.
A variety of monoclonal antibodies and heteroantisera have been identified as histiocytic or fibrohistiocytic markers in paraffin sections. They may be useful in the evaluation of potential histiocytic proliferations or neoplasms of skin and soft tissue, such as juvenile xanthogranuloma, reticulohistiocytoma, histiocytic sarcoma, and in tumors with a histiocytic component. The targets of these reagents include moieties such as α 1 -antitrypsin, α 1 -antichymotrypsin, muramidase/lysozyme, cathepsin B, CD68, CD163, factor XIIIa (FXIIIa), and the HAM 56 antigen.
Although it is true that a majority of putative fibrohistiocytic neoplasms do label for some of the described markers, the specificity of many of these markers is poor. With the exception of CD163, which is highly specific for histiocytic differentiation, carcinomas, melanomas, and other sarcoma types also potentially express these markers with relatively high frequency. The current approach to the diagnosis of so-called fibrohistiocytic tumors is predominantly based on morphologic features, with the inclusion of an IHC panel as appropriate to exclude other diagnostic possibilities.
CD68 is a 110-kD lysosomal protein. Antibodies against CD68 include KP1 and PG-M1, and the staining pattern is cytoplasmic. Expression is seen in any lysosomal-rich cell and is therefore not limited to histiocytes. Normal cells that express CD68 therefore include histiocytes, osteoclasts, and other cells of monocytic lineage. CD68-positive cells are therefore frequently found within tumors, and care must be taken not to overinterpret positive staining in background tumor-infiltrating histiocytes. In addition to pure histiocytic tumors, it has been found that granular cell tumor, schwannoma, RCC, and melanoma may all express CD68.
CD163 is the most specific histiocytic (monocyte and macrophage) marker currently available for diagnostic use. A glycoprotein belonging to a scavenger receptor superfamily, CD163 may play a role in immune response, and it also functions as a hemoglobin scavenger. The commercially available monoclonal antibody 10D6 performs well in paraffin-embedded tissue, showing a predominantly membranous pattern of staining. CD163 expression is found in the majority of cases of sinus histiocytosis with massive lymphadenopathy (Rosai-Dorfman disease); histiocytic sarcoma; hemophagocytic lymphohistiocytosis; tenosynovial giant cell tumor; a subset of cases of Langerhans cell histiocytosis; and some acute myeloid leukemias with monocytic differentiation. In the same study, all cases of littoral cell angioma showed staining for CD163, whereas normal splenic littoral cells lacked staining. Expression in other soft tissue tumors is extremely limited, and epithelial tissues and tumors are typically negative for CD163.
Currently, clone 10D6 is the most specific histiocytic marker for paraffin tissue.
Lysozyme is a bacteriolytic enzyme present in granulocytes and histiocytes (monocytes/macrophages). Pure histiocytic lesions, such as juvenile xanthogranuloma and histiocytic sarcoma, show cytoplasmic expression of lysozyme in most cases, whereas the lesional cells of so-called fibrohistiocytic tumors, such as benign fibrous histiocytoma, are typically negative for lysozyme.
FXIIIa is a proenzyme involved in fibrin polymerization that is reportedly found on dermal dendritic phagocytic and antigen-presenting cells. The diagnostic utility of FXIIIa in surgical pathology is limited. Although once thought to be a useful marker in distinguishing benign fibrous histiocytoma from DFSP, it is now recognized that much of the staining seen in benign fibrous histiocytoma is due to expression in background nonneoplastic dermal cells. FXIIIa-positive stromal cells may be seen in DFSP but tend to be less prominent because of the greater cellularity of tumor cells in this lesion. Furthermore, whereas focal or multifocal reactivity for FXIIIa may be found in many cases of benign fibrous histiocytoma, the intensity of staining is usually greatest in background stromal cells, making interpretation difficult.
NKI/C3 is a monoclonal antibody that recognizes a melanoma-associated antigen in FFPE tissue samples. This melanoma-associated antigen is a glycoprotein located in cytoplasmic vacuoles, and it is strongly expressed in cells with a large population of melanosomes. Although NKI/C3 is positive in only a few normal tissues, it stains a wide spectrum of histiocytoid soft tissue neoplasms. Its greatest utility lies in its expression in cellular neurothekeoma, but NKI/C3 is also positive in many other cutaneous tumors, some of which may mimic cellular neurothekeoma, including juvenile xanthogranuloma, atypical fibroxanthoma, cellular benign fibrous histiocytoma, reticulohistiocytoma, and xanthoma. NKI/C3 is negative in epithelioid benign fibrous histiocytoma and Langerhans cell histiocytosis.
Markers of Melanocytic Differentiation
Human Melanoma Black 45
Human melanoma black 45 (HMB-45) is a 100-kD glycoprotein localized to immature melanosomes. Fetal melanocytes and activated adult melanocytes will therefore express HMB-45, whereas normal resting adult melanocytes and intradermal nevi are negative. The staining pattern is cytoplasmic and may be granular. Approximately 90% of primary melanomas are positive for HMB-45, but when spindled in morphology, the frequency of HMB-45 in melanoma is much lower; virtually all desmoplastic melanomas are negative for this marker. The lesional cells of blue nevi are also typically positive for HMB-45. Expression of HMB-45 in metastatic melanoma is variable, with reported rates of positivity between 70% and 90%.
Among soft tissue tumors, HMB-45 is a useful marker for identifying neoplasms of the PEComa family. Tumors in the PEComa family, including angiomyolipoma of kidney, are characterized by dual myoid and melanocytic differentiation; the latter can be identified by staining for melanocytic markers such as HMB-45, melan A (see below), and MITF. The frequency and percentage of HMB-45–positive cells in these tumors is quite variable, and some cases show only scattered positive cells.
HMB-45 is also expressed in CCS of tendons and aponeuroses, often with a stronger intensity of staining than that for S100 protein. However, confirmatory molecular studies may be needed to confirm the diagnosis of CCS over either primary or metastatic melanoma. HMB-45 positivity is also seen in melanotic schwannoma but not in MPNST. Expression of HMB-45 appears to be limited to the tumors listed above, making it a highly specific marker for melanocytic differentiation.
Melan A (also known as MART1, melanoma antigen recognized by T cells 1) is a protein coded for by the gene MLANA . MART 1 (clone M2-7C10) and melan A (clone A103) are two different antibody clones that recognize the same antigen. Melan A is expressed in the cytoplasm of normal melanocytes and adrenal cortical cells; the latter is only detectable with clone A103. Similar to HMB-45, expression of melan A is frequently seen in PEComas, which accounts for much of its use in soft tissue pathology. The other main use of melan A in soft tissue pathology is in the evaluation of possible melanocytic neoplasms. Melan A expression is found in approximately 80% of metastatic melanomas but is negative in desmoplastic melanomas. Expression of this marker is less specific for melanocytic differentiation than HMB-45; tumors of the adrenal cortex and ovarian and testicular Leydig cell tumors also express melan A (clone A103).
Microphthalmia-Associated Transcription Factor
Microphthalmia transcription factor (MITF) is a nuclear protein that functions as a transcriptional regulator of melanocytic genes, and it has a functional role in osteoclasts. The clinical utility of MITF in surgical pathology is limited by its low specificity. Nuclear expression of MITF is seen in melanocytes, primary and metastatic melanomas, and in some PEComas, CCSs, and histiocytic proliferations or neoplasms.
In soft tissue pathology, use of tyrosinase, a 75-kD glycoprotein that catalyzes the incorporation of tyrosine into melanin pigment, is for the most part limited to identification of melanoma. As much as 90% of metastatic melanomas are reported to show diffuse cytoplasmic staining for this marker. CCS expresses tyrosinase in 40% of cases. Unlike melan A and HMB-45, expression of tyrosinase is not frequently seen in PEComas.