An Update on the Application of Newly Described Immunohistochemical Markers in Soft Tissue Pathology

ERG is a member of the ETS family of transcription factors, which also include ETS-1; Friend leukemia integration site 1 (Fli-1); NERF-2; and TEL; and which are expressed by vascular endothelial cells and defined by a conserved DNA-binding ETS domain that forms a winged helix-turn-helix structural motif.¹  ERG regulates endothelial cell differentiation, angiogenesis, and expression of several endothelial-specific antigens, and it is also required for embryonic stem cells to differentiate into endothelial cells.^2–6  Detection of ERG expression by immunohistochemistry was first described in a subset of prostatic adenocarcinomas, where it was shown to correlate with ERG overexpression via chimeric fusion proteins generated from oncogenic translocations, with TMPRSS2-ERG being the most common variant, seen in approximately 50% of all prostate-specific antigen–screened prostate cancers detected in the United States.^7–12 

A large study evaluating ERG as an immunohistochemical marker for vascular tumors showed that nuclear positivity for ERG was present in the endothelia of all hemangiomas and lymphangiomas examined.¹³  In addition, virtually all subtypes of hemangioendotheliomas were positive for ERG, including kaposiform hemangioendotheliomas, retiform hemangioendotheliomas, and epithelioid hemangioendotheliomas. Expression of ERG was also seen in all Kaposi sarcomas evaluated, and in 96 of 100 angiosarcomas (Figure 1).¹³  A wide variety of other nonvascular and nonepithelial mesenchymal, neuroectodermal, and hematopoietic tumors were also evaluated, and the vast majority were negative for ERG, with the exception of 7 of 10 blastic extramedullary myeloid tumors (70%) and 2 of 29 Ewing sarcomas (7%).¹³  Another study evaluated ERG in the context of differentiating cutaneous angiosarcoma from other cutaneous neoplasms that arise in sun-damaged skin and may enter the histologic differential diagnosis with angiosarcoma, specifically squamous cell carcinoma, malignant melanoma, and atypical fibroxanthoma. In that study, nuclear ERG expression was 100% sensitive and specific for angiosarcoma, with all 23 cases of angiosarcoma showing distinct nuclear staining, whereas all other tumors evaluated (15 poorly differentiated squamous cell carcinomas, 17 melanomas, 12 atypical fibroxanthomas, and 5 leiomyosarcomas) were negative.¹⁴  From these studies,^13,14  ERG is positive in greater than 95% of angiosarcomas, with a greater sensitivity for angiosarcoma than CD31 and CD34, which are markers routinely used to evaluate for angiosarcoma, and ERG usually shows a diffuse pattern of nuclear staining, which facilitates its interpretation in this context.

Expression of ERG occurs in a small subset of Ewing sarcomas (see above), and strong nuclear ERG expression has been found to correlate with EWSR1-ERG rearrangement, which is present in a small percentage of Ewing sarcomas compared with the more characteristic EWSR1-FLI1 rearrangement.¹⁵  TLS/FUS-ERG fusion transcripts occur in acute myeloid leukemia and likely account for the detection of ERG positivity in acute myeloid leukemia tissue infiltrates.^13,16,17  Positivity for ERG has also been reported to occur in a significant percentage of epithelioid sarcomas, with 41 of 109 cases (38%) showing ERG positivity in one study.¹⁸  However, ERG reactivity in epithelioid sarcoma is likely dependent on antibody choice, as a subsequent study showed that an antibody to the N-terminus of ERG resulted in positivity in 19 of 28 cases (68%), whereas an antibody to the C-terminus showed focal positivity in only 1 of 29 cases (3%).¹⁹  Whereas CD34 expression is frequent in epithelioid sarcoma, other more specific endothelial markers, such as CD31, are negative, and the characteristic loss of INI1/SMARCB1 expression typically found in epithelioid sarcomas should allow for the distinction of epithelioid sarcoma from vascular tumors in the vast majority of cases.¹⁸ 

Other than prostatic adenocarcinoma, very few epithelial malignancies have been reported to show ERG immunoreactivity—1 (of 42) pulmonary large cell undifferentiated carcinoma and 1 (of 27) pleural epithelioid type mesothelioma.¹³  ERG gene fusions have also been found in myxoid liposarcoma, although ERG immunoreactivity in myxoid liposarcomas has not been detected thus far.^13,16 

ERG is therefore a useful marker for confirming endothelial differentiation in both benign and malignant neoplasms, but expression can also be seen in a subset of epithelioid sarcomas and a small percentage of Ewing sarcomas, as well as approximately 45% to 50% of prostatic carcinomas.

Cutaneous angiosarcoma arises in four typical clinical settings—chronically sun-damaged skin, particularly the scalp or face; sporadic visceral angiosarcoma; in the setting of chronic lymphedema (eg, after mastectomy in Stewart-Treves syndrome); and in areas of prior therapeutic radiation, such as for the management of breast carcinoma.^20–29  The last two groups are considered “secondary” angiosarcoma. Atypical vascular proliferations, which have been described under various nomenclature designations, occur in areas of prior radiation, are often seen in association with angiosarcoma, and in some cases may be difficult to distinguish from vasoformative angiosarcoma.^{22,23,30–40}  MYC proto-oncogene is a transcription factor located on the long arm of chromosome 8 (8q24.21) and is implicated in cellular proliferation, differentiation, and apoptosis.⁴¹  Nuclear expression of MYC occurs in the vast majority of secondary angiosarcomas (Figure 2), but it is only very rarely seen in primary angiosarcoma, and it is not detected in atypical or benign vascular lesions occurring in irradiated skin.^42–45 

An array-based comparative genomic hybridization study of 22 cases of angiosarcoma (8 primary; 14 secondary to irradiation or chronic lymphedema) found that 10 secondary angiosarcomas (9 associated with radiation therapy and 1 with lymphedema) showed 16 recurrent alterations, with high-level amplifications on 5q35.3 (2), 8q24.21 (8), and 10p12.33 (6), whereas no recurrent alterations were seen in primary angiosarcomas.⁴²  Given that MYC in the region of 8q24.21 was a likely candidate for amplification, in the same study, fluorescence in situ hybridization (FISH) analysis of 28 primary angiosarcomas and 33 secondary angiosarcomas showed high-level MYC gene amplification in 18 secondary angiosarcomas (55%; 16 irradiated, 2 lymphedema) but not in primary angiosarcomas (P < .001).⁴²  Another study has shown high-level MYC amplification in 100% (20 of 20) of secondary angiosarcomas, as well as coamplification of FLT4 in 25% (5 of 20) of cases.⁴³  Atypical postradiation vascular proliferations and other benign vascular lesions of the breast are negative for both MYC amplification and MYC overexpression by immunohistochemistry, whereas MYC positivity is consistently detected in postradiation angiosarcomas.⁴⁴  A very recent study found that a subset of non–radiation-associated cutaneous angiosarcomas (9 of 38 cases evaluated) showed detectable expression of MYC by immunohistochemistry and that a subset of these cases showed high-level MYC gene amplification by FISH, although a small number of MYC immunohistochemistry–negative cases also showed MYC gene amplification.⁴⁵ 

MYC immunohistochemistry is therefore useful in differentiating atypical or benign vascular lesions occurring in irradiated skin from secondary postradiation angiosarcomas.

Amplification and overexpression of MDM2 is characteristic of well-differentiated and dedifferentiated liposarcoma, and both FISH and immunohistochemistry have become extremely useful tests for confirming the diagnosis of these tumor types, in conjunction with evaluation of CDK4 expression.⁴⁶  MDM2 protein is encoded by a gene at chromosome 12q14.3-q15 and acts as an inhibitor of the tumor suppressor effects of p53. Amplification of MDM2 can be detected by FISH, and overexpression of MDM2 protein can be detected with mouse monoclonal antibodies (clones 2A10 and 1F2; Figure 3). CDK4 is encoded by a gene at chromosome 12q13 and functions in cell cycle progression—CDK4 inhibits the retinoblastoma-1 (RB1) gene, and similar to MDM2 is overexpressed in well-differentiated and dedifferentiated liposarcoma (Figure 3). This phenomenon reflects the common presence of ring or giant marker chromosomes that contain amplified material from the q13-15 region of chromosome 12, where both the CDK4 and MDM2 genes are located. For both MDM2 and CDK4, only nuclear staining is considered positive. Staining in well-differentiated liposarcoma is often limited in extent and may be present in scattered nuclei only, in contrast to dedifferentiated liposarcoma, where nuclear staining is usually more diffuse.^46,47  Benign lipomatous lesions are negative for both MDM2 and CDK4, as are pleomorphic liposarcoma and myxoid liposarcoma.

Perhaps the most useful role for these two markers is in confirming a diagnosis of dedifferentiated liposarcoma (Figure 3, D through F) when faced with an otherwise nondescript spindle cell or pleomorphic sarcoma in the retroperitoneum, and when a well-differentiated component is not seen. In addition, these markers are often helpful in the distinction between benign lipomas, particularly those with prominent fat necrosis, and atypical lipomatous tumor (well-differentiated liposarcoma), particularly when the latter shows very minimal atypical histologic features.^46–48  However, it must be noted that histiocytes, such as those present in areas of fat necrosis, are frequently MDM2 positive, emphasizing the need for concurrent CDK4 evaluation. A small percentage of well-differentiated/dedifferentiated liposarcomas are negative for MDM2 and CDK4 by immunohistochemistry, with one study showing approximately 3% (3 of 105) and 10% (10 of 105) of cases of well-differentiated/dedifferentiated liposarcomas negative for MDM2 and CDK4, respectively, and in such cases where the histologic and clinical/radiologic findings are suggestive of this diagnosis, FISH for MDM2 amplification should be considered.⁴⁶  It is also important to note that overexpression of MDM2 may also be seen in other spindle cell neoplasms—in up to 64% (21 of 33) of malignant peripheral nerve sheath tumors, 42% (10 of 24) of myxofibrosarcomas, and 29% (12 of 41) of embryonal rhabdomyosarcomas—but when combined with coexpression of CDK4, very few of these tumors will express both of these markers.⁴⁶  In addition, staining for MDM2 and CDK4 is usually less extensive in these tumor types than in dedifferentiated liposarcoma.

Several other tumor types show amplification of MDM2 and CDK4, with corresponding overexpression at the protein level. Intimal sarcoma, a malignant neoplasm arising from large central vessels, most commonly the pulmonary artery, shows MDM2 amplification and overexpression in 75% (6 of 8) of cases.^49,50  Similarly, cardiac sarcoma, previously considered as representing an otherwise unclassified spindle cell sarcoma, has also been found to show MDM2 and CDK4 amplification in most cases, suggesting that it is closely related to intimal sarcoma. This has allowed for improved classification of cardiac sarcomas, and it appears that intimal sarcoma represents the most common primary cardiac sarcoma.⁵¹ 

Recent studies have also shown that amplification of MDM2 and CDK4 occurs in 67% (10 of 15) of low-grade osteosarcomas (parosteal and central low-grade osteosarcomas) and only 12% (16 of 130) of high-grade osteosarcomas, with corresponding rates of immunohistochemical expression of these proteins.⁵²  This finding has been shown to be of diagnostic utility; in one study all low-grade osteosarcomas expressed MDM2 and/or CDK4, usually diffusely and with moderate or strong intensity, whereas expression of these markers in benign morphologic mimics of these tumors was extremely limited, with expression seen in only one case of bizarre parosteal osteochondromatous proliferation.⁵³  Because expression of MDM2 and CDK4 is generally limited to low-grade osteosarcomas and their dedifferentiated counterparts, expression of these two markers in an otherwise nondistinctive high-grade sarcoma of bone may therefore suggest evolution from (dedifferentiation of) a low-grade osteosarcoma.⁵⁴ 

A recurrent NAB2-STAT6 fusion gene has very recently been identified in the vast majority of solitary fibrous tumors (SFTs), both benign and malignant, using a variety of techniques, including whole-exome and transcriptome sequencing.^55–58  STAT6 is a member of the STAT family of cytoplasmic transcription factors, which regulate gene expression by transmitting signals to the nucleus and binding to specific DNA promoter sequences. STAT6 modulates signaling by interleukin 4 and interleukin 13 in the immune system.⁵⁹  STAT signaling is important for normal cellular processes, embryonic development, innate and adaptive immune function, and regulation of cell differentiation, growth, and apoptosis. NAB2 is a transcriptional corepressor, a regulator of the early growth response 1 (EGR1) transcription factor. The NAB2-STAT6 gene fusion results in variable truncation of the repressor domain of NAB2 with replacement by the transcriptional activation domain of STAT6, and it is thought that the resulting fusion protein translocates to the nucleus, where it acts as a transcriptional activator, inducing expression of EGR target genes and resulting in increased proliferation.^57,60  In the early discovery studies, STAT6 was shown to be overexpressed at the protein level in SFT, whereas NAB2 showed less specific staining compared with other tumor types.⁵⁸  A large study of STAT6 expression in 231 tumors showed that nuclear expression of STAT6 is highly sensitive for SFT, with expression seen in more than 95% (59 of 60) of cases examined (Figure 4). Expression of STAT6 was limited in other tumors and was most commonly seen in a subset of dedifferentiated liposarcoma, as well as in 1 (of 10) deep fibrous histiocytoma, in which staining was weak and focal.⁶⁰  Other studies have since confirmed the high sensitivity and specificity of STAT6 for the diagnosis of SFT, with staining seen in a subset of dedifferentiated liposarcomas, and rarely in other tumor types.^60–63,193  The antibodies used differed in some of these studies, accounting for the additional cytoplasmic staining described by some authors; however, nuclear expression is expected in SFT.^61–63  Interestingly, a very recent study has shown a correlation between specific fusion types and morphology, with tumors with NAB2ex4-STAT6ex2/3 fusion genes corresponding to classic SFTs that arise most often in the pleura of older patients, show diffuse fibrosis, and typically have a benign clinical course. Tumors with the less common NAB2ex6-STAT6ex16/17 fusion occur in younger patients, usually in deep soft tissue, are more cellular, and are associated with aggressive clinical behavior.⁶⁴ 

Perhaps the greatest potential pitfall in the use of STAT6 for evaluating spindle cell lesions, particularly those in an intra-abdominal/retroperitoneal location, is the presence of STAT6 expression in approximately 7% to 15% of dedifferentiated liposarcomas, which likely represents amplification of the 12q13 region in this tumor type, where STAT6 is located close to MDM2 and CDK4.^60,62,193  The staining pattern in dedifferentiated liposarcoma is variable and may be focal or diffuse, weak or medium to strong in intensity, unlike the generally strong diffuse pattern seen in solitary fibrous tumor. In addition, unlike the predominantly nuclear pattern of staining seen in solitary fibrous tumor, both cytoplasmic and nuclear expression is common in dedifferentiated liposarcoma.¹⁹³  Diffuse expression of MDM2 and CDK4 helps favor a diagnosis of dedifferentiated liposarcoma, and if doubt persists, FISH for MDM2 amplification may also be useful. Because STAT6 and NAB2 are located close together on the long arm of chromosome 12, FISH to demonstrate rearrangement of the genes is technically challenging and not diagnostically useful. The identification of the NAB2-STAT6 fusion gene in meningeal hemangiopericytoma is further evidence that these lesions are in fact morphologic variants of solitary fibrous tumor.⁵⁸  STAT6 nuclear positivity helps distinguish meningeal hemangiopericytomas/solitary fibrous tumor from histologic mimics, because expression was not detected in any (n = 87) meningiomas and the vast majority of potential mimics in a large series of meningeal tumors evaluated for STAT6 expression.⁵⁸ 

MUC4 is a useful marker for low-grade fibromyxoid sarcoma (LGFMS) and sclerosing epithelioid fibrosarcoma (SEF; Figure 5). The mucin 4 (MUC4) gene was found to be significantly upregulated in LGFMS compared with other tumor types through gene expression array analysis.⁶⁵  MUC4 is a transmembrane glycoprotein that is normally expressed on many epithelial surfaces, where it is thought to serve a protective role, and is also involved in cell proliferation and survival through interacting with the ErbB/HER2 family of growth factor receptors.^66–71  A large study evaluating MUC4 as an immunohistochemical marker for LGFMS evaluated whole-tissue sections of 309 soft tissue tumors, including 49 LGFMSs (all of which showed FUS gene rearrangement by FISH), and found expression of MUC4 in all cases of LGFMS, with a diffuse cytoplasmic pattern of staining in tumor cells (Figure 5).⁷²  However, it should be noted that very rare cases of LGFMS may be negative for MUC4,⁷³  and if strong suspicion for LGFMS persists, FISH for FUS or EWSR1 rearrangement should be considered.

Sclerosing epithelioid fibrosarcoma, another fibroblastic neoplasm, shows morphologic and molecular overlap with LGFMS, and hybrid tumors showing features of both exist. A subset of SEF has been shown to contain the FUS-CREB3L2 gene fusion or FUS gene rearrangements, characteristic of LGFMS.^74–81  Similar to LGFMS, strong diffuse cytoplasmic expression of MUC4 is seen in SEF, occurring in 69% (20 of 29) to up to 90% (9 of 10) of “pure” forms, and virtually all tumors (12 of 12) showing hybrid features of both LGFMS and SEF (Figure 5).^82,83  Recent studies have shown that EWSR1-CREB3L1 is the predominant gene fusion in SEF; this fusion gene occurs uncommonly in LGFMS.^83,84 

MUC4 positivity is also seen in synovial sarcoma, predominantly in the glandular component of the biphasic subtype, with more limited expression in the spindle cell component of either biphasic or monophasic synovial sarcoma. Focal positivity has also been identified in ossifying fibromyxoid tumors (5 of 17; 29%), epithelioid GISTs (2 of 10; 20%), and some myoepithelial carcinomas (1 of 10; 10%).⁸²  However, unlike the diffuse strong pattern seen in LGFMS or SEF, staining in these tumor types is generally limited in extent, with either just scattered positive cells or focal staining. Importantly, regarding SEF, in which a poorly differentiated carcinoma may fall into the differential diagnosis, it should be remembered that MUC4 expression is seen in a variety of different carcinomas, such as pancreaticobiliary carcinomas, breast carcinoma, and colonic adenocarcinoma. Helpful features to support a diagnosis of SEF include the presence of characteristic dense sclerotic stroma, within which the tumor cells are embedded and may have pale or clear cytoplasm; the tumor cells of SEF are generally negative for cytokeratins, and in some cases FISH for EWSR1 and FUS gene rearrangement may also help support a diagnosis of SEF.

Gastrointestinal stromal tumors typically harbor gain-of-function mutations in KIT or PDGFRA, and KIT immunoreactivity is important in the diagnosis of most GISTs; however, approximately 5% to 10% of GISTs are negative for KIT, usually those with PDGFRA mutations.^85–88  Gene expression profiling using cDNA microarrays identified overexpression of DOG1 in GIST relative to other tumor types, and immunoreactivity to DOG1 was shown in 98% (136 of 139) of GISTs, regardless of KIT or PDGFRA mutation status.⁸⁹  Expression of DOG1 is typically both cytoplasmic and membranous (Figure 6). Further studies have confirmed that DOG1 is a highly sensitive marker for GIST, being present in more than 87% of cases (986 of 1040 cases⁹⁰ ; 370 of 425 cases⁹¹).^89–91  Importantly, DOG1 expression is found in up to 96% of KIT-negative GISTS (24 of 25 cases⁹² ; 10 of 28 cases⁹³ ; 9 of 10 cases⁹⁴), the most common of which are gastric tumors with PDGFRA mutations and epithelioid cytomorphology.^92–94  Approximately 3% (27 of 1040) of all GISTs are negative for both KIT and DOG1, and in cases where the diagnostic suspicion remains high for GIST, mutational analysis for KIT and PDGFRA mutations is indicated to confirm the diagnosis: 46% (11 of 24) of DOG1-negative GISTs have KIT or PDGFRA mutations.⁹⁰  The specificity of DOG1 for GIST is also relatively high, particularly among other mesenchymal neoplasms that may mimic GIST. Other mesenchymal tumors that show rare and/or focal DOG1 positivity include synovial sarcoma (2.5% to 16%, in 1 of 39 cases⁹¹  and 6 of 37 cases,⁹⁰  respectively), uterine type retroperitoneal leiomyomas (5 of 42; 12%),⁹⁰  leiomyosarcoma (1 of 326; 0.3%),⁹¹  and some perivascular epithelioid cell tumors (PEComas).^90,91 

DOG1 is also known as Anoctamin-1 (Ano-1), transmembrane protein 16A (TMEM16A), overexpressed in oral (squamous cell) carcinoma 2 (ORAOV2), and tumor-amplified and overexpressed sequence 1 (TAOS1).^95,96  DOG1 positivity is also common in esophageal squamous cell carcinomas and gastric adenocarcinomas, particularly intestinal type, and is less commonly seen in colorectal adenocarcinomas.⁹⁰  Infrequent positivity for DOG1 has been reported in a variety of other tumors, including endometrioid adenocarcinoma, acinic cell carcinoma, desmoplastic melanoma, malignant peripheral nerve sheath tumor, Ewing sarcoma, and glomus tumor.^97,98  DOG1 has also recently been found to be positive in the cellular areas of 100% (9 of 9) of chondroblastomas and may have utility in diagnosing this lesion.⁹⁹  The combination of both KIT and DOG1 reactivity is most useful in clinical practice in the diagnosis of GIST.¹⁰⁰ 

Approximately 15% of GISTs in adults and more than 90% in children lack KIT and PDGFRA mutations, so-called wild-type GISTs.^101–103  Wild-type GISTs also include those tumors arising in the Carney triad, the Carney-Stratakis syndrome, and neurofibromatosis 1 (NF1).¹⁰⁴  Carney triad is a nonhereditary syndrome that typically occurs in young females and is characterized by gastric GIST, paraganglioma, and pulmonary chondroma; Carney-Stratakis syndrome is inherited in an autosomal dominant fashion and consists of the dyad of gastric GIST and paraganglioma.^105–108  Defects in the succinate dehydrogenase (SDH) metabolic pathway have been found to occur in most pediatric GISTs, a subset of adult “wild-type” GIST, and GISTs occurring in Carney-Stratakis syndrome and Carney triad.¹⁰⁹  These tumors form a clinicopathologically distinct subset of wild-type GISTs known as “SDH-deficient GIST.” These tumors arise exclusively in the stomach, particularly the antrum, and show a characteristic multifocal and plexiform growth pattern, and usually a purely or predominantly epithelioid cytomorphology (Figure 7, A and B). Succinate dehydrogenase–deficient GIST is typically positive for KIT and DOG1. Unlike KIT mutant GIST, lymph node metastases are common in SDH-deficient GIST, and these tumors tend to be imatinib resistant, yet often show an indolent clinical course.^{103,110–112}  Succinate dehydrogenase is an enzyme complex composed of four subunit proteins (A, B, C, and D) that is localized in the inner mitochondrial membrane and is involved in the oxidation of succinate to fumarate in the citric acid cycle and the electron transport chain.^113,114  SDHB and SDHA are normally ubiquitously expressed in a granular cytoplasmic pattern. Germline mutations in SDHA, SDHB, SDHC, and SDHD were first identified in pheochromocytoma-paraganglioma syndrome, and they result in the destabilization of the SDH complex and loss of SDHB expression by immunohistochemistry (Figure 7, C).^115–117  However, somatic mutations in any of these genes result in loss of SDHB expression, and loss of expression is also seen in tumors where no identifiable mutations are detected, suggesting that dysfunction of the pathway may have arisen by other mechanisms or by other mutations that are not detectable by current methods. Loss of SDHB by immunohistochemistry defines the group of SDH-deficient GIST^118–121  and has been shown to occur with an estimated frequency of 7.5% among all gastric GISTs.¹¹⁰  Although loss of SDHB expression is detected in 42% (22 of 53) of wild-type GISTs, GISTs with KIT or PDGFRA mutation consistently show intact SDHB expression.¹²¹ 

Germline mutations in SDHA, SDHB, SDHC, and SDHD occur in some, but not all, patients with SDH-deficient GIST.^{109,120,122,123}  Germline mutations in SDHA represent the most commonly mutated gene, and loss of expression of SDHA by immunohistochemistry has been shown to reliably predict the presence of SDHA mutations.¹²⁴  Interestingly, although germline mutations have been identified in association with Carney-Stratakis syndrome, mutations in SDH genes have not been found in patients with Carney triad, and the mechanism for deficient SDH function in Carney triad is uncertain at this time.^{119,125–127}  Wild-type GIST associated with NF1 do not show loss of SDHB expression.^119,121,128 

Given the significant prognostic and predictive implications with regards to clinical course, selection of appropriate therapy (SDH-deficient GIST responds better to second- or third-generation tyrosine kinase inhibitors), and implications for screening of other tumors in both the patient and family members, SDHB immunohistochemistry should be considered for gastric GISTs with an epithelioid cytomorphology and multinodular or plexiform growth pattern. If expression is lost in tumor cells, with normal endothelial cells and inflammatory cells acting as a positive internal control, the patient should be referred for genetic counseling. In addition, if SDHB expression is lost, additional loss of staining for SDHA suggests the presence of SDHA mutation.

INI1 is the protein product of the gene hSNF5/INI1/SMARCB1, located on the long arm of chromosome 22 (22q11.2). INI1 is a core subunit of the SWI/SNF ATP-dependent chromatin remodeling complex, and is ubiquitously expressed in the nuclei of normal cells. INI1 is thought to function as a tumor suppressor. Loss of INI1 function (with corresponding loss of protein expression) can result from mutations or deletions of the INI1 gene.¹²⁹  Loss of INI1 expression is observed in several tumor types. Abnormalities in INI1 were first described in malignant rhabdoid tumors (renal, extrarenal, and atypical teratoid/rhabdoid tumor of the central nervous system), and loss of INI1 expression is seen in nearly all of these tumors. Although most are sporadic mutations, a small group of infants with malignant rhabdoid tumor have germline mutations in INI1/SMARCB1.¹³⁰  Loss of INI1 expression is also seen in approximately 93% (127 of 136) of epithelioid sarcomas, both conventional and proximal-type (Figure 8, A and B).¹³¹  In contrast to malignant rhabdoid tumors, where mutations in INI1 are the most common cause of loss of function, chromosomal deletions are usually identified in epithelioid sarcomas.¹³²  In addition to malignant rhabdoid tumors and epithelioid sarcoma, loss of INI1 expression is also seen in 50% (12 of 24) of epithelioid malignant peripheral nerve sheath tumors and in a subset of myoepithelial carcinomas (Figure 8, C and D)—particularly those arising in children (9 of 22; 40%; versus 2 of 22; approximately 10% in adults).^131,133 

Other tumor types reported to show variable loss of INI1 expression include extraskeletal myxoid chondrosarcomas (4 of 24; 17%),¹³⁴  poorly differentiated chordoma, undifferentiated hepatoblastoma, and renal medullary carcinoma.^135–138  Approximately 60% of patients with the rare hereditary syndrome of familial schwannomatosis have germline mutations in INI1/SMARCB1, which is associated with a “mosaic” pattern of protein loss by immunohistochemistry.¹³⁹  Reduced, but not complete loss of, INI1 expression has been reported in synovial sarcoma; the significance of this finding is uncertain.^140,141 

TLE1 is a relatively recently described marker of synovial sarcoma. Gene expression profiling studies have shown a major association of the Wnt signaling pathway with synovial sarcoma and overexpression of members of the TLE (Transducin-Like Enhancer of split) family of genes, particularly TLE1, which is one of four TLE genes that encode human transcriptional repressors homologous to the Drosophila corepressor groucho, and is an important component of the Wnt signaling pathway.^142–150  TLE proteins are involved in multiple developmental processes, including lateral inhibition, segmentation, sex determination, eye development, neuronal differentiation, and hematopoiesis.^{149,151–154}  An early large study using tissue microarrays showed strong positive nuclear staining for TLE1 in 91 of 94 molecularly confirmed synovial sarcomas (97%), including both the epithelial and spindle cell components of biphasic synovial sarcoma as well as poorly differentiated variants (Figure 9).¹⁵⁰  Other studies have also shown high sensitivity and specificity of TLE1-positive immunohistochemical staining for molecularly confirmed synovial sarcoma.^155–157  However, one study evaluating 163 soft tissue and bone tumors using whole sections found TLE1 positivity in 18 of 20 synovial sarcomas (90%), but also in 53 of 143 nonsynovial sarcomas (37%).¹⁵⁸ 

Regarding other tumors, TLE1 expression is most commonly seen in peripheral nerve sheath tumors and solitary fibrous tumors. Terry et al¹⁵⁰  reported 16 of 88 malignant peripheral nerve sheath tumors (MPNSTs; 18%) with any staining for TLE1, but only 4 tumors (5%) showed more than weak staining. Another study has shown a slightly lower rate of expression of TLE1 in synovial sarcoma (82%; 60 of 73).¹⁵⁹  In that study nuclear reactivity for TLE1 was also observed in 15% (7 of 47) of MPNSTs and 8% (4 of 49) of solitary fibrous tumors, but it was usually only weak in these tumor types. Occasional staining for TLE1 has also been described in clear cell sarcoma, high-grade chondrosarcoma, Ewing sarcoma, rhabdomyosarcoma, GIST, myxofibrosarcoma, and leiomyosarcoma.^150,158  TLE1 expression has also been reported in malignant mesothelioma.¹⁶⁰  However, in most of these tumors, staining is focal and weak to moderate in intensity, unlike the strong diffuse nuclear pattern characteristic of synovial sarcoma. Within nonneoplastic tissue, TLE1 expression has been reported to occur in basal keratinocytes, adipocytes, perineurial cells, endothelial cells, and mesothelial cells.¹⁵⁸ 

TFE3 is a member of the microphthalmia (MiT) family of transcription factors, which includes MiTF, TFEB, TFEC, and TFE3, all of which share a common structure consisting of a helix-loop-helix leucine zipper dimerization motif, a transactivation domain, and a basic region involved in DNA binding.¹⁶¹  The most well-studied MiT transcription factor is MiTF, which is important in the development of melanocytes (survival, growth, and migration), melanogenesis, and osteoclast development; the function of the other MiT family members is less well defined at this time.¹⁶¹  The utility of MiTF in clinical practice is limited by its low specificity: nuclear expression of MiTF is seen in primary and metastatic melanomas, as well as in some PEComas, clear cell sarcomas, and histiocytic proliferations or neoplasms.^162,163 

Although TFE3 is ubiquitously expressed in humans, native TFE3 protein is usually not detected by routine immunohistochemical methods. Nuclear expression of TFE3 is seen in a variety of different tumors, most of which harbor TFE3 gene fusions, including alveolar soft part sarcoma, Xp11 translocation renal cell carcinoma, “melanotic” Xp11 translocation renal cell carcinoma, and a subset of PEComas and epithelioid hemangioendotheliomas.^164–171  Although virtually all alveolar soft part sarcomas and Xp11 translocation renal cell carcinomas will show diffuse nuclear TFE3 expression (Figure 10), only a subset of PEComas will show nuclear immunoreactivity for TFE3, with a subset of those cases also harboring a TFE3 gene rearrangement.^170,172  Moreover, PEComas with a TFE3 gene fusion have been found not to harbor the tuberous sclerosis complex (TSC2) alterations characteristic of conventional PEComas, and expression of TFE3 in PEComa is mutually exclusive to expression of MiTF.¹⁷³  Although limited by small numbers, cutaneous PEComas do not appear to show reactivity for TFE3 by immunohistochemistry or TFE3 rearrangement by FISH.¹⁷⁴ 

Very recently, a YAP1-TFE3 fusion gene has been detected in a subset of epithelioid hemangioendothelioma, with corresponding detectable TFE3 nuclear reactivity by immunohistochemistry.¹⁷¹  Interestingly, epithelioid hemangioendotheliomas with this fusion gene show distinct morphologic appearances, in that they typically have voluminous eosinophilic cytoplasm, show well-formed vascular channels and focally solid growth, and usually arise in young adults.¹⁷¹ 

TFE3 immunohistochemistry is therefore useful in confirming a diagnosis of alveolar soft part sarcoma and Xp11 translocation renal cell carcinoma, and is also expressed in epithelioid hemangioendotheliomas with a distinctive fusion gene and clinicopathologic features, as well as a small subset of PEComas.

SOX10 is a nuclear transcription factor normally expressed in neural crest cells that is crucial for differentiation of Schwann cells and melanocytes.^175–177  Nuclear staining is seen in normal melanocytes, Schwann cells, secretory cells of the eccrine coil, myoepithelial cells, and acinar cells of salivary gland tissue.^178–180  Expression of SOX10 is correspondingly seen in tumors showing neural crest differentiation (ie, melanocytic and nerve sheath tumors), as well as a subset of myoepithelial and salivary gland–type tumors (acinic cell carcinomas, adenoid cystic carcinomas, epithelial-myoepithelial carcinomas, myoepitheliomas/myoepithelial carcinomas, and mixed tumors/pleomorphic adenomas).^179–182  In addition to staining most benign nerve sheath tumors (ie, neurofibroma and schwannoma), recent studies have shown that SOX10 is expressed in 27% to 50% of MPNSTs (21 of 78 cases¹⁷⁹ ; 38 of 77 cases¹⁸²), a sensitivity similar to that of S100 protein for this tumor type.^179,182  SOX10 expression is also detected in a large majority of melanocytic neoplasms, including benign nevi (blue, neurotized, dysplastic, Spitz, and nodal capsular) and malignant melanoma (conventional, spindled, desmoplastic, and metastatic), with one study showing SOX10 positivity in 97% of melanomas (76 of 78)¹⁸²  and another study showing SOX10 positivity particularly in desmoplastic melanoma (7 of 9; 78%),¹⁷⁹  as well as clear cell sarcoma (4 of 7; 57%¹⁷⁹ ; malignant melanoma of soft parts).^179,182,183  Expression of this marker should therefore be interpreted in context, and in general it is not useful in distinguishing melanocytic from neural tumors. However, it can be useful to confirm a diagnosis of melanoma when other melanocytic markers are negative, in the right histologic and clinical context (Figure 11). Karamchandani et al¹⁷⁹  reported expression of SOX10 in 26 of 26 granular cell tumors (100%). The specificity of SOX10 for tumors showing neural crest differentiation (ie, melanocytic and nerve sheath tumors) is greater than that for S100 protein because expression of SOX10 in other mesenchymal and epithelial tumors is limited; expression is found in a subset of myoepithelial and salivary gland tumors, diffuse astrocytomas, and some ductal breast carcinomas.^181–183  Sustentacular cells of paraganglioma/pheochromocytoma and a subset of carcinoid tumors also express SOX10.¹⁸² 

Cancer-testis antigens are a family of antigens normally expressed in adult testicular germ cells, but which have been found to be aberrantly expressed in a variety of tumors. These antigens elicit both humoral and cell-mediated immune responses and are becoming attractive targets for immune-based cancer therapies.^184,185  NY-ESO-1 is encoded by the CTAG 1B gene and was initially discovered by serologic analysis of a cDNA expression library of a patient's esophageal carcinoma and autologous serum, and it has since been found to be expressed in melanoma, sarcomas, and various other carcinomas.^186,187 

NY-ESO-1 is expressed in up to 80% (20 of 25 cases¹⁸⁸ ; 38 of 50 cases¹⁸⁹) of synovial sarcomas (both monophasic and biphasic types) by immunohistochemistry, with strong reactivity seen in both the spindle cell and glandular components, and with a predominantly cytoplasmic pattern of staining.^188,189  Expression appears to be limited among other spindle cell tumors, with 10% (2 of 20) of dermatofibrosarcoma protuberans, 3% (1 of 34) of MPNSTs, and 1% (2 of 155) of GISTs showing diffuse positivity. NY-ESO-1 may therefore be useful in the distinction of synovial sarcoma from other spindle cell tumors.

NY-ESO-1 has also been found to be frequently expressed in myxoid liposarcoma,^190,191  with a lower frequency seen in other liposarcoma subtypes.¹⁹¹  In a study of 138 tumors, including myxoid liposarcoma and other myxoid neoplasms, some of which may fall into the differential diagnosis, such as extraskeletal myxoid chondrosarcoma, 36 of 38 cases of myxoid liposarcoma (95%) were positive for NY-ESO-1, whereas all other tumor types were negative.¹⁹²  NY-ESO-1 may therefore also be useful in the distinction of myxoid/round cell liposarcomas from other myxoid neoplasms.

In summary, many new diagnostic immunohistochemical markers for soft tissue tumors have been described in the last 10 years. This has resulted in improved classification of tumors, identification of clinically and pathologically distinct groups of tumors with prognostic and predictive implications, most notably SDH-deficient GIST, and has also likely reduced the need for additional ancillary molecular studies in routine surgical pathology practice. However, the use of all of these markers requires careful clinical correlation and knowledge of the spectrum of staining in other tumor types, as no one marker is 100% sensitive or specific for a given diagnosis. This article has reviewed the main clinical uses of several newly described immunohistochemical markers that have proved to be of significant clinical utility, along with potential pitfalls and indications where additional molecular studies may be warranted.