Immunohistochemistry Update in Dermatopathology and Bone and Soft Tissue Pathology

BRCA1-associated protein 1 (BAP1) is a tumor suppressor protein whose activity is dependent on nuclear localization. BAP1-inactivated melanocytic tumors (BIMTs) are characterized by a unique morphology consisting of an intradermal nodular proliferation of large epithelioid melanocytes with abundant eosinophilic cytoplasm, scant melanin pigment, and distinct cell borders. An associated conventional nevus component is seen in most cases. Most BIMTs have an indolent clinical course; however, because they are often the presenting sign in BAP1 tumor predisposition syndrome, diagnosis of BIMT is important in order to identify patients who may require germline testing.¹ 

Multiple studies have found loss of nuclear expression of BAP1 by IHC to be a reliable marker for BIMTs, with most cases having nuclear loss in the epithelioid component and retention of nuclear staining in the conventional nevus component (if present; Figure 1, A). A subset of tumors, particularly sporadic cases, have clumped perinuclear BAP1 staining in addition to loss of nuclear BAP1 expression.¹  Coexisting BRAF V600E mutations are seen in most BIMTs and can be detected with VE1 IHC; cytoplasmic staining for VE1 is seen in both the epithelioid and conventional nevus components.^2,3  Rare cases of BIMTs with alternative molecular abnormalities, including NRAS mutations or RAF1 fusions, have been reported.^4–6 

A molecularly confirmed case of BIMT with retained nuclear BAP1 expression has been described. Further testing found that although there was heterozygous loss of the BAP1 gene locus, the remaining BAP1 allele possessed a small variant mutation; although this mutant BAP1 protein was biologically nonfunctional, it still accumulated in the nucleus, resulting in retained nuclear BAP1 staining by IHC.⁷  Therefore, molecular testing should be considered when there is retained nuclear BAP1 expression but a strong morphologic suspicion for BIMT. In addition, BIMTs without conventional epithelioid morphology have been reported, indicating that these tumors have a wider morphologic spectrum than initially described.^8,9  Therefore, use of BAP1 may be indicated in such cases if there is clinical suspicion for BAP1 tumor predisposition syndrome.

Insulinoma-associated protein 1 (INSM1) is a zinc-finger transcription factor involved in the development of neuroendocrine cells. Antibodies targeting INSM1 are useful for identifying neuroendocrine and neuroepithelial tumors, including Merkel cell carcinoma (MCC), a cutaneous neuroendocrine neoplasm most often arising in the sun-damaged skin of older, fair-skinned adults. MCC is more aggressive than its histologic mimics, but it is also highly responsive to treatment with checkpoint inhibitors; therefore, correct and prompt diagnosis is essential for appropriate treatment. INSM1 is highly sensitive for MCC (93%–100%); most cases demonstrate strong, diffuse nuclear positivity, staining more than 75% of tumor cells (Figure 1, B). Other neuroendocrine markers, such as synaptophysin and chromogranin, are less sensitive (88%–96% and 32%–86%, respectively) and have more variable staining.^10–12  Rarely, MCC is negative for INSM1 but positive for other neuroendocrine markers; therefore, a panel of neuroendocrine markers may be the best approach when evaluating for neuroendocrine differentiation.¹² 

Common histologic mimics, such as basal cell carcinoma, basaloid squamous cell carcinoma, sebaceous neoplasms, melanoma, and B-cell lymphoma, show at most weak, focal staining for INSM1.^11,12  However, INMS1 is not entirely specific; strong, often diffuse positivity in seen in extraskeletal myxoid chondrosarcoma, a subset of Ewing sarcoma and angiosarcoma, and a case of BCOR::CCNB3 sarcoma. Strong but focal staining has been reported in cases of ossifying fibromyxoid tumor, Ewing sarcoma, soft tissue myoepithelioma, synovial sarcoma, and chordoma.^13–15  Of note, INSM1 staining cannot distinguish MCC from metastatic neuroendocrine carcinoma from other sites.^10–12 

Preferentially expressed antigen in melanoma (PRAME) is a tumor-associated antigen found in normal sebaceous glands, testis, ovarian, placental, adrenal, and endometrial tissues, as well as a variety of malignant neoplasms, including melanoma.^16–18  Overall, diffuse nuclear PRAME expression may be helpful in distinguishing between benign and malignant melanocytic tumors (Figure 1, C); however, it has notable limitations. A small subset of melanomas is negative for PRAME, and a small subset of nevi is positive for PRAME. Thus, PRAME should not be used as a standalone binary “benign versus malignant” stain but rather as an additional ancillary tool for evaluating challenging melanocytic lesions. PRAME expression should be interpreted in conjunction with the microscopic morphology and the clinical context.¹⁷ 

Diffuse PRAME expression (defined in most studies as PRAME staining in >75% of lesional cells) occurs in most cutaneous melanomas and melanoma subtypes, including superficial spreading melanoma (92.5%), acral melanoma (89.3%–94.4%), nodular melanoma (90%), and lentigo maligna melanoma (88.6%).^16,17,19  However, only 56% of spindle cell melanomas and 35% of desmoplastic melanomas are diffusely positive.^16,20  PRAME is also useful in the assessment of nodal and distant metastasis and in distinguishing between melanoma and nodal nevi. PRAME expression in more than 50% of tumor cells has a 94% sensitivity and 100% specificity for lymph node metastatic melanoma (versus nodal nevus).^21,22 

In melanocytic lesions with Spitzoid or intermediate histology, PRAME has variable sensitivity (67%–91.7%) and specificity (93.5%–100%).^23–25  A threshold of staining in at least 60% of tumor cells has been proposed as a threshold for malignancy.²⁶  In addition, the use of “hot spot” staining for PRAME may be helpful in such cases.²⁷  However, because rare false-positive cases have been reported, PRAME should be interpreted with caution, and in the context of morphology and clinical presentation.¹⁷ 

Diffuse PRAME expression is not seen in most nonmelanoma skin cancers; however, rare MCCs and sarcomatoid squamous cell carcinomas have positivity in more than 50% of tumor cells.^20,28  It may prove to be a helpful marker in distinguishing melanoma from clear cell sarcoma and perivascular epithelioid cell tumor (PEComa); the latter 2 entities are negative for PRAME or show only weak, nondiffuse PRAME expression (with the exception of 1 malignant PEComa that showed strong diffuse PRAME expression).^8,14  However, relatively few cases of clear cell sarcoma and PEComa have been evaluated with PRAME IHC in the literature to date. PRAME appears to have more limited utility in relation to other soft tissue tumors because diffuse PRAME expression can be seen in a subset of poorly differentiated angiosarcoma, malignant peripheral nerve sheath tumor (MPNST), atypical fibroxanthoma, and pleomorphic dermal sarcoma.^20,24,29  Lastly, high-intensity PRAME expression may be seen in scattered isolated melanocytes in the dermal-epidermal junction of background nonlesional epidermis or within solar lentigines; caution must be used to not mistake these for melanoma cells.^16,28  Of note, strong cytoplasmic (not nuclear) PRAME staining is also seen in normal sebaceous glands and in a subset of sebaceous neoplasms; this may represent artifactual staining rather than true PRAME protein expression.^16,27 

Myxoid liposarcoma (MLS) is a translocation-associated sarcoma, characterized by DNA damage-inducible transcript 3 (DDIT3) gene rearrangements/fusions not found in other tumors. Initial studies of an antibody targeting the N-terminal region of DDIT3 demonstrate high sensitivity (96%–100%) and specificity (94%–96%) for MLS, including cases with high-grade morphology and therapy-related changes (Figure 2, A).^30–32  Nearly all cases of MLS have strong, diffuse nuclear staining (>50% of tumor cells) for DDIT3, with some false negatives attributed to the use of older or decalcified tissue.^31,32  Although DDIT3 staining has been reported in a variety of other sarcomas (including non-MLS liposarcomas, myxoid sarcomas, round cell sarcomas, and radiation-associated sarcomas), those non-MLS tumors showed only focal DDIT3 staining (usually 5%–10% of tumor cells or less, but rarely up to 25%).^30–32  Diffuse DDIT3 staining (in >50% of tumor cells) has so far proven to be 100% specific for MLS.^31,32  DDIT3 is a promising marker with the potential to confirm the diagnosis of MLS in difficult cases without the need for molecular testing.³² 

Giant cell tumor of bone (GCTB) is characterized by mutations of the histone H3.3 encoding gene, H3F3A (H3 histone family member 3A), at position G34. An antibody targeting the most common mutation, G34W (present in 95% of cases of GCTB) has a high sensitivity (91%–98%) and near-perfect specificity for the diagnosis of GCTB.^33–35  The vast majority of GCTBs demonstrate strong and diffuse nuclear positivity in neoplastic mononuclear cells; nonneoplastic cells, such as osteoclast-like giant cells, are negative (Figure 2, B).^33–39  However, weak and/or patchy staining can rarely be seen, particularly in cases treated with denosumab.³⁴  Decalcification, particularly by harsh methods, may lead to decreased or negative staining in GCTB.³³ 

H3.3 G34W mutations also occur in the malignant counterpart of GCTB, as well as a small subset of predominately giant cell–rich osteosarcomas.^33,36  In addition, 1 case of undifferentiated high-grade pleomorphic sarcoma showed positivity of G34W IHC; however, the authors noted that this may represent a malignant GCTB with sarcomatous overgrowth.³⁵  Alternate H3.3 mutations, such as G34R and G34V, which are relatively more common in GCTB arising in the small bones of the hands and feet, are not detected with G34W IHC.^33,35  Antibodies targeting these alternate mutations have been developed, but they are not widely available.³⁵ 

MPNST is a rare, aggressive sarcoma, which can arise sporadically or in the setting of neurofibromatosis type 1 (NF1) or prior radiation therapy. Conventional MPNST (spindle MPNST) can be difficult to diagnose, especially in sporadic cases, because it is usually composed of hypercellular “herringbone” fascicles of atypical spindle cells, morphologic features that may be seen in a variety of other soft tissue tumors. Additionally, unlike benign nerve sheath tumors, conventional (spindle cell) MPNST is often paradoxically negative for S100 and SOX-10 (or shows only weak/focal expression of these markers). A subset of spindle MPNST demonstrates loss-of-function mutations in polycomb repressive complex 2 (PRC2), leading to loss of trimethylation of histone 3 at lysine position 27, which can be reliably detected using a monoclonal antibody (H3K27me3; Figure 2, C). H3K27me3 loss has been identified in 48% to 76% of MPNSTs, and with a higher incidence in high-grade (80%–81%), sporadic (72%–95%), and radiation-associated (91%–100%) cases.^40–43  Of note, epithelioid MPNST does not show loss of H3K27me3 staining; instead, it usually has loss of INI1 (SMARCB1) staining. Because epithelioid MPNST is composed of sheets of markedly atypical epithelioid cells that strongly express S100 and SOX-10, it can usually be easily distinguished from conventional spindle cell MPNST.

H3K27me3 loss can occur in other tumors, such as neurofibroma, synovial sarcoma, spindle cell melanoma, and other types of radiation-associated sarcomas.^40–44  In addition, both spindle MPNST and histologic mimics can have weak and/or patchy staining, which is difficult to interpret. A lack of standardized criteria for interpretation also contributes to significant intraobserver variability.^41,42 

Because of its relatively low sensitivity and specificity, loss of H3K27me3 by IHC is supportive, but it is not sufficient by itself to definitively confirm the diagnosis of MPNST. Loss of dimethylation of H3K27 (H3K27me2), which is more specific to PRC2 loss and MPNST, is a promising antigenic target, and it requires further study.⁴⁵ 

Activating rearrangements involving neurotrophic tyrosine kinase receptor genes (NTRK1, NTRK2, NTRK3) occur in a variety of tumors, including infantile fibrosarcoma and the still-developing category of NTRK-rearranged spindle cell neoplasms. Identification of these alterations is of growing clinical importance because of the advent of targeted therapy available for TRK, such as larotrectinib and ertectinib.^46,47 

Pan-TRK monoclonal antibody targets the C-terminal region conserved in NTRKs 1, 2, and 3, and has an overall sensitivity of 75% to 100% and specificity of 81% to 98%. Sensitivity for NTRK3 rearrangement is lower (55%–94%) than for rearrangements of NTRK1 (86%–100%) and NTRK2 (89%–100%).^48–52  Distinct staining patterns are also observed; NTRK1- and NTRK2-rearranged tumors show cytoplasmic staining, and NTRK3-rearranged tumors have nuclear staining, with or without cytoplasmic staining.⁴⁸  Most NTRK-rearranged tumors have moderate-to-strong staining in more than 50% of cells; cases with less intense and/or less diffuse staining are more difficult to interpret. The threshold for pan-TRK positivity is not well defined.^48–52 

However, pan-TRK immunostaining may also be seen in a variety of soft tissue tumors that are negative for NTRK gene rearrangements. Diffuse positivity for pan-TRK can be seen in tumors with YWHAE or BCOR rearrangements, such as ossifying fibromyxoid tumor with ZC3H7B::BCOR fusion and clear cell sarcoma of the kidney.^53,54  This is because molecular alterations of YWHAE/BCOR result in upregulation of the NTRK3 gene product.⁵⁴  Solitary fibrous tumor was positive for pan-TRK in 100% of tested cases (n = 15) in 1 study, usually with a strong and diffuse cytoplasmic staining pattern. Of note, SFT often expresses BCOR by IHC.⁵⁴  Ewing sarcoma, synovial sarcoma, and leiomyosarcoma are pan-TRK positive in a small subset of cases.^54,55  Most other sarcomas are usually pan-TRK negative, although a variety of rare exceptions have been reported.^54,55  Because of these factors, pan-TRK immunohistochemical positivity is not sufficient for definitive identification of NTRK rearrangements; confirmatory molecular testing is recommended, especially if the patient will receive targeted therapy with a TRK inhibitor.⁵⁵ 

Dermatofibrosarcoma protuberans (DFSP) is a translocation-associated sarcoma, characterized by COL1A1::PDGFB gene fusion (in the vast majority of cases) or alternate fusions involving PDGFD.^56–58  Although molecular testing is not required for diagnosis in classic cases of DFSP, it may be helpful to confirm the diagnosis in cases with nonconventional morphology, including fibrosarcomatous DFSP, myxoid DFSP, or giant cell fibroblastoma (a unique morphologic variant of DFSP usually seen in children). Molecular testing may also be helpful to confirm or exclude a diagnosis of DFSP in bland CD34⁺ cutaneous spindle cell lesions when the sample is small or limited. RNA chromogenic in situ hybridization (CISH) targeting a portion of PDGFB (PDGFB CISH) is an emerging marker for DFSP, with the potential to decrease the need for molecular confirmation. A recent study found diffuse nuclear overexpression of PDGFB in 92% of conventional DFSPs (24 of 26) and 100% of fibrosarcomatous DFSPs (diffuse overexpression was defined as more than 5 puncta or a single aggregate of chromogen in greater than 50% of neoplastic cells; Figure 2, D).⁵⁹  Limited staining (less than 25% of tumor cells) was only seen in 2% (7 of 300) of other tumors evaluated, which included desmoplastic melanoma, MPNST, pleomorphic dermal sarcoma, and angiosarcoma. One limitation of PDGFB CISH is its inability to detect alternate fusions of PDGFD. Of the 2 cases of DFSP that were negative for PDGFB CISH, 1 had an alternate fusion involving PDGFD.⁵⁹  Additional studies are needed to further evaluate the utility of this marker.

Synovial sarcoma is a translocation-associated sarcoma characterized by a t(X;18)(p11;q11) translocation, resulting in a fusion of SS18, and SSX1, SSX2, or SSX4. Traditional immunohistochemical markers, such as keratins, EMA, CD99, BCL-2, and TLE-1, are not specific for synovial sarcoma, limiting their utility.^60,61  Molecular testing for SS18 gene rearrangements has been the gold standard for diagnosis; however, cases with atypical fluorescence in situ hybridization patterns or variant fusion transcripts may result in false negatives.^62,63  Recently, 2 promising antibodies for synovial sarcoma have been identified. SS18-SSX targets the gene fusion protein and has near-perfect specificity but is less sensitive (87%–95%; Figure 2, E). SSX-CT targets the C-terminal region, which is homologous in SSX1, SSX2, and SSX4; it has near-perfect sensitivity but is less specific (93%–96%).^60–64 

Most synovial sarcomas demonstrate strong, diffuse nuclear positivity (>75% of tumor cells) for both markers. However, a weaker and/or focal staining occurs in a minority of cases. Positive staining has not been observed in any other tumors, including a variety of benign and malignant soft tissue neoplasms, sarcomatoid carcinoma, and mesothelioma.^60–63 

Potential pitfalls include nonspecific staining in necrotic tissue and nonneoplastic cells, and false-negative staining in decalcified specimens.⁶²  In most cases, strong, diffuse positivity for both SS18-SSX and SSX-CT is sufficient to confirm the diagnosis of synovial sarcoma. Cases that are negative for either marker require molecular testing.

There are a few additional noteworthy markers for rare bone and soft tissue tumors that could not be comprehensively discussed in this paper because of length but that pathologists should still be aware of. These will be summarized concisely below.

Chondroblastoma is a rare benign cartilage tumor of the bone. The vast majority of chondroblastomas (95%) have a p.K36M mutation in the histone H3.3 encoding genes, either H3F3B (most common) or H3F3A.⁶⁵  In multiple studies, 100% of chondroblastomas with confirmed H3F3 p.K36M mutation showed positive nuclear expression of K36M by IHC (the small subset of chondroblastomas that lack the mutation show negative staining).^65–67  Rare cases of clear cell chondrosarcoma may have H3F3 p.K36M mutation, and these will show positive K36M immunostaining.⁶⁵  A wide range of other bone and soft tissue tumors and carcinomas have been tested for K36M immunostaining and were negative.⁶⁵  K36M is a highly sensitive and specific immunohistochemical marker for chondroblastoma that can be used on small-needle biopsies or in cases with unusual clinical or pathologic features.^65,66 

A subset of round cell sarcomas show fusions involving the CIC gene (most commonly CIC::DUX4).^68,69  These CIC-rearranged sarcomas may mimic Ewing sarcoma or other round blue cell tumors microscopically but have distinct clinical implications, including aggressive clinical behavior and the potential for developing resistance to chemotherapy.^68,69  One study found strong diffuse nuclear expression of ETV4 immunostain in 100% of CIC-rearranged sarcomas (n = 17); of the 110 morphologic mimics tested, 6 cases showed focal weak ETV4 staining and the remainder were completely negative for ETV4 staining.⁶⁹  Another study evaluated 40 cases of CIC-rearranged sarcoma and 200 morphologic mimics for immunohistochemical expression of ETV4 and WT1.⁶⁸  Diffuse moderate-to-strong ETV4 expression was seen in 90% of CIC-rearranged sarcomas (as well as in 5% of the morphologic mimics), and WT1 expression was seen in 95% of CIC-rearranged sarcomas. In summary, diffuse ETV4 expression was 90% sensitive and 95% specific for CIC-rearranged sarcoma, whereas WT1 expression was 95% sensitive and 81% specific. Coexpression of ETV4 and WT1 was seen in 85% of CIC-rearranged sarcomas, suggesting that the combination of these markers may be particularly helpful in distinguishing CIC-rearranged sarcomas from other round blue cell tumor mimics.⁶⁸ 

BCOR::CCNB3 fusion sarcoma is a recently described entity that may arise in bone or soft tissue and may display a combination of round cell and/or spindle cell morphology.^70–72  The tumor cells may be arranged in cords, small nests, or microcystic structures with myxoid background change.⁷⁰  Some display collagenized stroma.⁷⁰  Cases with round cell features can be confused with other round cell sarcomas, such as Ewing sarcoma, CIC-rearranged sarcomas, and small cell osteosarcoma.⁷⁰  Those with spindle cell features may mimic synovial sarcoma or MPNST.⁷⁰ BCOR::CCNB3 fusion sarcomas are consistently positive for BCOR IHC, usually with strong diffuse nuclear staining.⁷⁰  CCNB3 immunostain is also positive in most cases, although a small subset of molecularly confirmed cases are negative for CCNB3 by IHC.⁷¹  Other markers often expressed by BCOR::CCNB3 fusion sarcomas include SATB2, cyclin D1, TLE1, Bcl-2, CD99, and CD117.^70,72  PAX-8 may also be positive.⁷¹ 

Biphenotypic sinonasal sarcoma is a rare low-grade sarcoma of the sinonsal tract characterized by PAX3 rearrangements (usually partnered with MAML3).⁷³  These tumors are infiltrative and have a propensity for local recurrence but not metastasis. They are composed of uniform spindle to ovoid cells arranged in fascicles, often with an ectatic staghorn hemangiopericytoma-like vascular pattern. Potential morphologic mimics include monophasic synovial sarcoma, MPNST, solitary fibrous tumor, sinonasal hemangiopericytoma, cellular schwannoma, and spindle cell rhabdomyosarcoma.^73,74  Biphenotypic sinonasal sarcoma typically displays a “biphenotypic” combination of neural (S100) and myogenic (smooth muscle actin, desmin, and, rarely, myogenin) marker expression by IHC. Interestingly, they are negative for SOX-10. They are consistently positive for PAX3 by IHC, whereas the morphologic mimics listed above are almost always negative; Jo et al⁷³  found PAX3 IHC to be 100% sensitive and 98% specific for biphenotypic sinonasal sarcoma versus histologic mimics. PAX8 IHC is also consistently positive in biphenotypic sinonasal sarcoma (possibly due to cross-reactivity with PAX3), although it is not as specific because it is also variably expressed in many histologic mimics.⁷³ 

In conclusion, IHC is an important adjunctive tool for select entities in dermatopathology, bone pathology, and soft tissue pathology, giving pathologists evidence of their behavior, underlying molecular alterations, and line of differentiation. Many of these recently described stains demonstrate higher sensitivity and specificity than older immunohistochemical markers. In select scenarios, immunohistochemical stains may even be able to partially or completely replace the need for certain molecular tests. However, because limitations and pitfalls still exist in IHC, correlation with morphologic and clinical findings continues to be a crucial diagnostic skill for the modern anatomic pathologist.