Immunohistochemistry in the Diagnosis and Classification of Breast Tumors

Ductal and lobular carcinomas, both in situ and invasive, have different clinical behaviors and implications; hence, the distinction between ductal and lobular lesions is clinically important. Ductal carcinoma in situ (DCIS) is usually managed surgically with the goal of complete removal of the lesion. Lobular carcinoma in situ (LCIS), except for florid and pleomorphic variants, is managed conservatively if there are no concerning pathologic or radiologic features.¹  Compared with invasive ductal carcinoma (IDC), invasive lobular carcinoma (ILC) has distinct molecular aberrations and is less sensitive to chemotherapy. Clinically, patients with ILC have an increased propensity for multifocal and multicentric disease, and they are less likely to undergo breast-conserving surgery.^2,3 

Histologically, ductal carcinoma is characterized by cell-to-cell cohesion, a defined tumor cell border, and tumor cells arranged in tubules, trabeculae, and papillary architectures. Lobular carcinoma usually shows small monotonous, discohesive cells without the formation of tubules or papillae. In ILC, tumor cells loosely disperse through or are arranged in a single-file pattern in the fibrous connective stroma. Although these histologic criteria can distinguish most ductal versus lobular carcinomas, a challenge may be presented by poorly differentiated carcinomas, in situ carcinoma with a solid growth pattern, or the pleomorphic variant of lobular carcinoma, which is composed of large, pleomorphic tumor cells with or without necrosis, mimicking the ductal phenotype.⁴  Sometimes, DCIS and LCIS can coexist, which makes diagnosis problematic.

The hallmark of lobular carcinoma in the breast is loss of expression of the intercellular adhesion molecule E-cadherin, which is encoded by the CDH1 gene located on chromosome 16q22.1. E-cadherin is a calcium-dependent transmembrane protein composed of extracellular, transmembrane, and intracellular domains. It forms homodimers with E-cadherin molecules on adjacent cells through extracellular domains to mediate cell-cell adhesion. The intracellular domain is linked to the actin cytoskeleton via α-, β-, γ-, and p120 catenins.^5,6  This cadherin-catenin system is important in the organization and integrity of most epithelial tissues. Loss of E-cadherin function or a dysfunctional cadherin-catenin complex are common findings in lobular neoplasia. Genetic or epigenetic alterations, such as loss of heterozygosity at 16q, inactivating mutations or promoter hypermethylation, or homozygous deletion of CDH1, are common molecular findings for this abnormality.^7–11  IHC stains for E-cadherin and other members of the cadherin-catenin complex, particularly p120 catenin and β-catenin, have been widely used in breast pathology to help to distinguish ductal from lobular carcinomas in histologically problematic cases.^5,6,12–14  Both DCIS and IDC cells typically exhibit membrane expression of E-cadherin, p120 catenin, and β-catenin. In contrast, the cells of LCIS and ILC show loss of cell membrane expression of E-cadherin and β-catenin, in conjunction with cytoplasmic expression of p120 catenin.

There are pitfalls in using E-cadherin IHC staining alone to differentiate ductal versus lobular carcinomas. On one hand, aberrant E-cadherin expression in lobular carcinoma has been reported in the range of 2% to 16%.^{5,12,13,15,16}  The most common aberrant E-cadherin IHC stain pattern is focal, fragmented, or beaded membrane staining. Circumferential membrane staining of tumor cells with reduced or attenuated intensity can also be seen in lobular carcinoma. In some cases, cytoplasmic staining may exhibit a dotlike, perinuclear Golgi-type pattern. Rarely, a wild-type pattern with strong, diffuse, circumferential membrane staining, similar to that seen in ductal carcinomas, has been reported in lobular carcinomas.^5,12  Further studies showed that most lobular carcinomas with aberrant E-cadherin expression demonstrate evidence of impaired integrity of the cadherin-catenin complex.^12–16  These cases frequently show cytoplasmic staining of p120 catenin and loss of membrane staining of β-catenin. Thus, using p120 catenin or β-catenin in conjunction with E-cadherin and interpreting in accordance with histologic features can be helpful to categorize a carcinoma as a ductal or lobular phenotype. On the other hand, some ductal carcinomas may show reduced or even complete absence of membrane expression of E-cadherin, which could be misinterpreted as lobular phenotype.^17–19  Reduced or absent of E-cadherin expression in nonlobular carcinomas has been reported to be associated with aggressive features, such as larger tumor size and higher tumor grade or basal-like and triple-negative phenotypes. ^17–19  Taken together, the presence (or absence) of E-cadherin expression in an in situ or invasive carcinoma with classic lobular (or ductal) morphology cannot preclude its classification as lobular (or ductal) carcinoma. E-cadherin IHC stain should be avoided for a lesion that clearly exhibits lobular or ductal morphologic features on histologic examination to avoid misinterpretation.

Spindle cell lesions of the breast constitute a wide spectrum of entities, ranging from reactive tumorlike lesions to benign neoplasms and low- and high-grade malignant neoplasms.^20–24  These entities are rare and can have overlapping morphologic features, especially on small biopsies, which pose diagnostic challenges. Accurate classification and diagnosis are of crucial importance to ensure appropriate management. IHC studies play an important role in the workup of these lesions. The common entities from each category are discussed below.

This group comprises lesions with similar histomorphology but different clinical characteristics. The major entities include reactive spindle cell nodule or exuberant scar, nodular fasciitis (NF), inflammatory pseudotumor or inflammatory myofibroblastic tumor (IMT), and fascicular variant of pseudoangiomatous stromal hyperplasia (PASH). They can be generally viewed as benign fibroblast and myofibroblast proliferation. Thus, these lesions share a common IHC phenotype with positive staining for vimentin, CD34, Bcl-2, and CD99. Myofibroblasts also express myogenic markers such as α–smooth muscle actin (SMA), desmin, and hormone receptors. The spindle cells are negative for cytokeratin (CK), S-100, and nuclear β-catenin expression. Clinical history, histologic and IHC features, and molecular assays can be helpful to distinguish them.

A reactive spindle cell nodule or exuberant scar arises postoperatively or after biopsy or fine-needle aspiration. It is usually a small painless breast nodule (ranging from 1.5 to 9 mm in greatest dimension).²⁵  Histologically, this lesion is well circumscribed and composed of intersecting fascicles of plump bland-appearing spindle cells with a vague storiform architecture set in a myxoid edematous stroma containing numerous thin-walled vessels and acute and chronic inflammation. Depending on the duration of the lesion, either extravasated red blood cells or hemosiderin-laden macrophages and multinucleated giant cells can be seen in the stroma. Mitotic figures are absent or very rare (0–1/10 high-power fields).²⁵  The scar can express SMA but is negative for CKs, p63, CD34, and nuclear β-catenin. A rare variant of reactive spindle cell nodule called cellular spindled histiocytic pseudotumor was first described in 2012.²⁶  It represents an exaggerated histiocytic reaction to fat necrosis in the breast. Clinically, the majority of patients have a history of ipsilateral breast carcinoma and prior therapeutic irradiation. The radiographic features can be suspicious for malignancy. Histologically, it shows bland spindled cells with histiocyte-like features, admixed with chronic inflammatory cells and multinucleated giant cells. A more typical feature of fat necrosis is frequently present at the periphery. The lesional cells express the histiocyte-associated markers CD163, CD11c, and CD31. They are negative for CKs.²⁶  Awareness of this entity could help prevent its misdiagnosis as other spindle cell lesions. More recently, cases have been reported as mammary ALK-positive histiocytosis, either as localized disease to the breast or as systematic disease involving the breast.^27–29  Microscopically, it can appear to be similar to a cellular spindled histiocytic pseudotumor of the breast. However, the spindled and epithelioid histiocytes in this entity diffusely express ALK and harbor a highly recurrent KIF5B-ALK gene fusion, indicating a neoplastic nature. Mammary ALK-positive histiocytosis should be distinguished from other spindle cell lesions as patients could benefit from ALK inhibitor treatment.

NF of the breast has similar clinical and histologic features to NF at other locations.³⁰  It usually presents as a rapidly growing, reactive, self-limiting, mass-forming lesion in the breast. Histologically, early lesions are characterized by proliferation of mitotically active spindled myofibroblasts embedded in a myxoid to fibrous stroma that usually contains extravasated red blood cells and lymphocytic infiltrates (tissue culture–like appearance). Advanced lesions are less cellular and have more collagenized stroma. NF has a similar IHC phenotype to that of a reactive spindle cell nodule/exuberant scar. Molecular studies demonstrate USP6 gene rearrangement in NF, which can be used for definitive diagnosis.³¹ 

Inflammatory pseudotumor or IMT of the breast is considered a reactive lesion to either mechanical trauma or unknown stimuli. It usually presents as a painless, circumscribed, firm mass. Microscopically, IMT shows bland-appearing spindled myofibroblasts intermingled with a significant number of lymphocytes and plasma cells.³²  An IHC study may show expression of SMA and ALK-1 in 50% to 60% cases, with focal expression of desmin and CKs. The spindle cells are negative for p63, S-100, and CD34.

PASH sometimes presents as a palpable or radiographic well-circumscribed, nontender breast mass. Morphologically it is characterized by proliferation of myofibroblasts forming slitlike pseudovascular spaces.³³  However, in the fascicular variant of PASH, the spindle cells can arrange in fascicles with interspersed keloidlike collagen fibers, with no evidence of pseudovascular spaces, which makes it difficult to distinguish from other bland spindle cell lesions, especially myofibroblastoma of the breast.³⁴  The spindle cells in PASH express CD34, Bcl-2, and progesterone receptor (PR); they variably express SMA, desmin, and calponin. The spindle cells do not express CKs, S-100, estrogen receptor (ER), and the endothelial markers CD31 and ERG.

This group includes tumors that are relatively specific to the breast and those typically occurring in soft tissues of other locations that can be occasionally encountered in breast parenchyma.

Myofibroblastoma is a benign tumor of the mammary stroma composed of fibroblasts and myofibroblasts. It usually presents as a well-circumscribed, slow-growing, painless mass that is typically seen in men, as well as in postmenopausal women.³⁵  Histologically, the lesion has circumscribed margins and consists of bland-appearing spindle cells arranged haphazardly or in short intersecting fascicles with bands of intervening collagen fibers. Mitoses are absent or rare. Entrapment of breast ducts or lobules is usually not seen. Five morphologic patterns of myofibroblastoma have been described: lipomatous, myxoid, fibrous/collagenized, epithelioid/deciduoid, and palisading/Schwannian-like.^36,37  The wide variety of morphologies could lead to it being misdiagnosed as other soft tissue tumors or even as epithelial neoplasms. The characteristic IHC features of myofibroblastoma includes coexpression of desmin and CD34, along with ER, PR, and androgen receptor (AR); and loss of RB1 expression in tumor cells.³⁸  SMA, calponin, CD10, Bcl-2, and CD99 are variably expressed in tumor cells as well. The tumor cells do not express CKs, p63, S-100, STAT6, or nuclear β-catenin.

Some benign spindle cell soft tissue neoplasms can rarely occur in the breast, such as leiomyoma, spindle cell lipoma, solitary fibrous tumor, schwannoma/neurofibroma, and myxoma. These tumors are histologically and immunohistochemically identical to their counterparts in soft tissue. Awareness of these possibilities and keeping them in the differential diagnosis could help in choosing appropriate IHC panels for the following workup.

In leiomyoma, the spindle cells form interlacing fascicles and show eosinophilic cytoplasm and elongated nuclei with blunt ends. They diffusely coexpress SMA, desmin, and h-caldesmon. Spindle cell lipoma has a mixture of spindle cells, a mature fatty component, ropey collagen fibers, and floretlike multinucleated cells. The spindle cells diffusely express CD34 and show RB1 loss.³⁸  The adipocytes in spindle cell lipoma express S-100. A solitary fibrous tumoris composed of round to spindle-shaped cells that are haphazardly arranged (patternless), and traversed by bands of collagen and thin-walled branching blood vessels. Immunohistochemically, it is characterized by diffuse and strong expression of STAT6, in addition to CD34, CD99, and Bcl-2.³⁹  The spindle cells in schwannoma and neurofibroma show wavy nuclei. Schwannoma can have classic Antoni A and Antoni B areas, verocay bodies, and hyalinized vessels. S-100 and SRY-related HMG-box 10 (SOX10) are diffusely expressed in schwannoma, while they are focally expressed in neurofibroma. Myxoma shows stellate-shaped spindle cells in a myxoid stroma. The spindle cells can focally express SMA or calponin.

Primary desmoid fibromatosis of the breast parenchyma is a locally aggressive nonmetastatic lesion with a high recurrence rate.^40,41  It frequently originates from the fascia of pectoral muscles or Cooper ligaments. This lesion may be visualized on mammography or appear as an ill-defined mass on ultrasound examination, mimicking carcinoma. Microscopically, fibromatosis shows bland spindle cells aligned parallel to one another and arranged in long, sweeping fascicles in a fibrous stroma with variable collagen deposition. The lesion has an infiltrating border, with entrapped fat and lymphocyte cuff at the periphery. On IHC, fibromatosis cells show immunoreactivity for SMA, along with nuclear staining for β-catenin (in more than 80% of cases). They are negative for CKs, p63, CD34, demsin, S-100, and hormonal receptors (Figure 1, A and B). It is important to be aware that although β-catenin nuclear staining is a characteristic feature of fibromatosis, it can be seen in other lesions of the breast, including phyllodes tumor (PT) and fibromatosis-like metaplastic breast carcinoma (MBC).⁴² 

Fibromatosis-like MBC is another bland-looking spindle cell proliferation that is morphologically similar to desmoid fibromatosis. However, it is by far the most important differential diagnosis of bland-looking breast spindle cell lesions because patients may have local recurrence and occasional metastasis.⁴³  Focal cytologic atypia and areas of epithelioid differentiation (more rounded nuclei with eosinophilic cytoplasm and sharper cell borders, arranged in small cohesive clusters) is highly suggestive of fibromatosis-like MBC.^44,45  Although not specific, scattered inflammatory infiltrate composed of lymphocytes and plasma cells with occasional lymphoid follicle can be seen at the edges of the tumor. Sometimes, a diagnosis of malignancy solely relies on the demonstration of CK expression on IHC. The tumor cells are positive for CKs or p63 (Figure 1, C and D). They often express SMA but are typically negative for CD34, ER, PR, and human epidermal growth factor receptor 2 (HER2).

The most common high-grade spindle cell neoplasms are metaplastic spindle cell carcinoma (SpCC) and malignant phyllodes tumor (MPT). Other rare entities, such as primary breast spindle cell sarcoma and metastatic high-grade spindle cell neoplasm, are included in the differential diagnosis when the tumor has no morphologic and IHC evidence of epithelial differentiation and no “leaflike” structure, or the patient has a history of soft tissue sarcoma or metastatic sarcomatoid tumors.

Metaplastic SpCC is a type of high-grade MBC that usually presents as a large firm breast mass.⁴⁶  Histologically, the tumor can be pure high-grade malignant spindle cells or mixed spindle cells with other malignant epithelial cells.⁴⁷  The spindle cells can form long or short fascicles, or storiform or haphazard growth patterns. Moderate-to-severe cytologic atypia and mitotic figures are easily identified. Foci of necrosis may also be present. In mixed metaplastic SpCC, a squamous or conventional invasive breast carcinoma (IBC) component tends to be located at the tumor periphery. Metaplastic spindle cells can also be a component of metaplastic carcinoma with heterologous mesenchymal differentiation, in which the mesenchymal components include chondroid, osseous, rhabdomyoid, and even neuroglial differentiation. Although the presence of malignant epithelial components makes the diagnosis of metaplastic carcinoma straightforward, in pure metaplastic SpCC, an IHC analysis revealing the expression of epithelial and myoepithelial markers is crucial for making the correct diagnosis.^47,48  In this situation, a broad-spectrum of CKs should be applied, including pancytokeratin (AE1/AE3 and MNF116), high-molecular-weight CKs (CK903, CK5/6, CK14, and CK17), and low-molecular-weight CKs (CAM5.2, CK7, and CK19).⁴⁹  Myoepithelial marker p63 is often positive in metaplastic SpCC. SOX10 has been reportedly positive in 46% (6 of 13) of the spindle cell component of MBCs.⁵⁰  GATA binding protein 3 (GATA3) expression has been shown to be significantly higher in metaplastic SpCC than in other spindle cells lesions of the breast.⁵¹  Recently, trichorhinophalangeal syndrome 1 (TRPS1) was recognized as a highly sensitive and specific marker for breast cancer, and its expression is retained in MBCs, whereby GATA3 can be lost (Figure 2, A through C).^52,53  The tumor cells in SpCC are typically negative for CD34, ER, PR, and HER2.

Stromal-rich MPT is a major differential diagnosis for metaplastic SpCC.⁵⁴  Although the presence of an epithelial cleft is pathognomonic of MPT, it may present as a pure spindle cell tumor, particularly on core needle biopsy in which only the stromal component is represented. In challenging cases, an IHC panel including CKs, p63, CD34, CD117, and Bcl-2 can be used to distinguish these 2 entities. CD34 is a useful marker in the diagnosis of PT. Although CD34 expression is less strong in MPT than in benign and borderline PT, the majority of MPTs show some degree of positivity, a feature not seen in metaplastic SpCC.^55,56  In contrast, the expression of CD117 and Bcl-2 increases in MPT compared with that in benign and borderline PT.^56,57  It is also important to be aware that, although CKs and p63 are frequently expressed by MBCs, their expression level may be reduced in high-grade tumors and focal expression can be seen in the stroma of some MPTs (Figure 2, D through F).^58,59  In addition, TRPS1 is also highly expressed in MPT, which limits its utility in differentiating MPT with metaplastic SpCC of the breast.⁶⁰  For problematic cases, a thorough sampling of excision specimens is advised to identify relevant diagnostic “leaflike” architectural components (for MPT) or the carcinoma component (for metaplastic SpCC) for making the correct diagnosis.

Primary breast sarcoma is extremely rare.^61,62  Theoretically, most types of soft tissue sarcomas can arise in the breast. Their histologic and IHC features remain like those of tumors occurring elsewhere. Among the primary breast sarcomas, radiation-associated or secondary angiosarcoma is the most common.⁶²  It usually is a high-grade lesion composed of solid sheets of spindle and epithelioid cells with variable degrees of vasoformation. Mitotic activity, necrosis, and blood lakes are commonly seen. Secondary angiosarcoma is associated with a very poor prognosis.^63,64  For this reason, a high degree of suspicion for this lesion and use of IHC for vascular markers, including CD31, CD34, D2-40, and ERG, is recommended. The majority of radiation-associated angiosarcoma is associated with high-level amplification of MYC at 8q24, which can be used to distinguish other atypical vascular lesions and primary angiosarcoma of the breast.⁶⁵  Other than angiosarcoma, the most common histology of primary breast sarcoma is reported to be undifferentiated pleomorphic sarcoma, followed by leiomyosarcoma and giant cell sarcoma.⁶²  In our practice, before considering primary breast spindle cell sarcoma as the diagnosis, pathologists must work thoroughly to exclude the much more common metaplastic SpCC and MPT of the breast.

Metastatic malignancies to the breast from extramammary primary origins are reported to account for between 0.2% and 1.1% of malignant tumors of the breast. However, they represent the first sign of malignancy in up to one third of cases.³⁵  Metastatic high-grade spindle cell neoplasms are rare and can mimic those arising from the breast primarily. Patient clinical history, unusual tumor morphologic features, lack of in situ carcinoma, triple-negative status, and multiple or bilateral lesions can be clues to the correct diagnosis. By far the most reported noncarcinomatous metastasis to the breast is melanoma.⁶⁶  Melanoma may show spindle cell morphologic features as well. Carefully looking for intranuclear inclusions, frequent mitotic activity, and melanin pigmentation in tumor cells is important for raising suspicion. Other relatively common spindle cell tumors that metastasize to the breast include leiomyosarcoma, rhabdomyosarcoma, and sarcomatoid renal cell carcinoma.^67–70  IHC plays an essential role in classifying tumor origins. Table 1 lists some useful immunostains for differential diagnosis of common metastatic spindle cell neoplasms to breast with primary breast metaplastic SpCC and MPT. In our clinical practice, a targeted IHC panel should include markers that are predicted to be both positive and negative in each entity in the differential diagnosis. In addition, being aware of the IHC caveats is important to avoid misdiagnosis. For example, SOX10 and S-100 can label both triple-negative breast carcinoma (TNBC), including metaplastic SpCC, and metastatic melanoma, which presents a significant potential diagnostic pitfall.⁷¹ 

Invasion in breast cancer is defined as breaching the myoepithelial layer and basement membrane into the surrounding stroma.³⁵  The presence of myoepithelial cells around proliferating epithelial cells of breast lesions determines the difference between benign breast lesions with infiltrative growth patterns and invasive carcinoma, as well as the difference between in situ and invasive carcinoma. The 2 most common applications of IHC for this purpose are determining the differentiation of sclerosing benign lesions from invasive carcinoma and classifying carcinoma as invasive or in situ.^72–75  In both situations, the demonstration of intact myoepithelial cells is used as evidence of a noninvasive process in most cases. Myoepithelial cells can be visualized readily in normal breast ducts and acini on H&E-stained sections. When these structures are distorted by stromal reactive changes such as fibrosis or inflammation, it may be difficult to visualize myoepithelial cells and IHC can play an essential role.

A variety of benign breast lesions are categorized as benign sclerosing lesions, which include sclerosing adenosis, radial scars, and microglandular adenosis. They are characterized by the proliferation of glandular elements distorted by fibrous or fibroelastotic stroma. Imaging and histopathologic features of these lesions may mimic invasive carcinomas, especially low-grade cancers such as tubular carcinoma.

Sclerosing adenosis is the most common histologic variant of adenosis and has a broad spectrum of presentations mimicking invasive carcinoma on imaging and histopathologic evaluation, especially on small-core needle biopsies.^74–77  Sclerosing adenosis is composed of small glandular proliferation and spindled myoepithelial cells with stromal sclerosis distorting and compressing the glands. On mammography, sclerosing adenosis may present as microcalcifications, architectural distortion, or rarely, as a mass lesion. Sclerosing adenosis arises in terminal ductal lobular units and maintains lobular architecture at low power, with rounded and well-defined nodules. Sometimes, the lobulocentric growth pattern is not evident and proliferating glands show a dispersed growth pattern into adipose tissue, mimicking invasive carcinoma. Unlike invasive carcinoma, in sclerosing adenosis, no cytologic atypia, mitotic activity, or stromal desmoplasia is present. The epithelial cells may show metaplastic changes, which are most commonly the apocrine type. Apocrine metaplasia cells may show cytologic atypia that should not be confused for carcinoma. Myoepithelial cells are easily identified in some cases but can be inconspicuous in others. IHC staining for myoepithelial markers may be necessary to identify intact myoepithelial cells in these cases (Figure 3, A through D).⁷⁷  Ductal lumens may have secretions and microcalcifications, along with hyalinization or fibrosis of the surrounding stroma. Sclerosing adenosis may show an irregular pattern, with involvement of the adipose tissue and even nerves, which can be misinterpreted as perineural invasion of invasive carcinoma. Demonstration of myoepithelial cells in these glands will aid in the diagnosis.

These lesions are characterized by a fibroelastotic stromal core from which ducts and lobules radiate. The radiating ducts and lobules may display varying degrees of epithelial hyperplasia, metaplastic changes, or ectasia, and the ducts around the central scarred zone may have a distorted angular appearance. Typically, the proliferating ducts demonstrate a dual ductal epithelial and myoepithelial layer. It is accepted that the term radial scar refers to lesions less than 1 cm, while complex sclerosing lesion is used to describe lesions 1 cm or larger. These lesions are frequently associated with hyperplasias (either ductal or lobular) and in situ carcinomas. The most important lesion from which these sclerosing lesions should be distinguished is an invasive carcinoma, especially tubular carcinoma. Although the diagnosis is straightforward on H&E-stained sections in most cases, the differential diagnosis might be challenging if there is a superimposed in situ neoplasia. The demonstration of a myoepithelial component by IHC stains may assist in this differential diagnosis.^74–77 

Microglandular adenosis is a rare variant of adenosis characterized by the proliferation of small, round, and uniform ducts in a haphazard fashion. The glands are lined by a single layer of monotonous, cuboidal luminal cells that are surrounded by basement membrane but lack myoepithelial cells.^78,79  The haphazard growth pattern and absence of the myoepithelial layer mimic IBC. It may clinically present as a palpable mass or as an incidental finding on core needle biopsies performed for other imaging abnormalities. In contrast with other forms of adenosis, the proliferating glands are distributed in a nonlobulocentric pattern and lack a myoepithelial cell layer. Intraluminal periodic acid-fast and mucicarmine positive secretions as well as microcalcifications may be present. The microglandular adenosis cells display strong immunoreactivity for CKs, S-100, and cathepsin-D. In addition, they are negative for ER and PR, HER2 overexpression, or epithelial membrane antigen (EMA). In contrast, tubular carcinoma cells are characteristically positive for ER and PR, and EMA.^74–77 

DCIS of the breast is a nonobligatory precursor to IBC and is defined as malignant proliferation of epithelial cells that are confined to preexisting breast ducts and lobules.³⁵  The extension of carcinoma cells beyond the basement membrane defines invasion. In most cases, the distinction between in situ and invasive carcinoma is straightforward on H&E-stained sections. However, there are a variety of growth patterns of invasive carcinoma that may mimic in situ carcinoma. Invasive cribriform carcinoma, invasive papillary carcinoma, or adenoid cystic carcinoma may display growth patterns that are remarkably similar to those of in situ carcinoma. Conversely, lobular unit involvement in DCIS can cause distortion marked by stromal reactive changes, which can mimic invasive carcinoma. IHC stains for myoepithelial cells can be extremely helpful to establish the diagnosis (Figure 4, A through C).

Although there are many markers with varying sensitivity and specificity for identifying myoepithelial cells in clinical practice, there is no single marker that is 100% specific and sensitive. The most commonly used myoepithelial markers are p63, p40, SMA, smooth muscle myosin heavy chain (SMM-HC), and calponin (Table 2).^80–86 

Both p63 and p40 are members of the p53 gene family; p40 is a truncated isoform of p63. Both are expressed exclusively in the nuclei of myoepithelial cells in normal breast tissue. In ductal carcinoma, the expression of these markers can be discontinuous with “gap” areas that should not be overinterpreted as invasion. In addition, p63 and p40 can be expressed in invasive carcinoma cells showing squamous or basal differentiation, such as low-grade adenosquamous cell carcinoma, squamous cell carcinoma, MBC with squamous differentiation,⁸⁵  and basal-like TNBC.⁸³  SMA and SMM-HC both identify contractile elements and are sensitive markers of the myoepithelium.⁸⁰  In addition, both markers stain stromal myofibroblasts, vascular smooth muscle, and pericytes. In general, SMM-HC displays less reactivity in myofibroblasts than SMA. Calponin is also a sensitive myoepithelial marker that identifies contractile elements expressed in differentiated smooth muscle cells. Calponin is strongly expressed in normal myoepithelial and vascular smooth muscle cells. Overall, calponin has been reported to be more specific but less sensitive than SMA.

Other markers, such as S-100, specific CKs (CK5/6, 14), maspin, and CD10, also stain myoepithelial cells, but the staining is not specific and is not optimally sensitive for diagnostic use. In clinical practice, to demonstrate the presence or absence of myoepithelial cells, a panel-based approach of at least 2 markers is recommended. Markers that most effectively combine sensitivity, specificity, and ease of interpretation include p63 and either SMM-HC or SMA.

For newly identified ER/PR-positive and/or HER2-positive carcinoma in the breast, if there is no specific cancer history such as ovarian serous carcinoma, it will be taken for granted that it is a primary breast cancer. No diagnostic marker is needed in this situation. However, for a high-grade triple-negative carcinoma with no in situ component in the breast, a diagnostic marker may be needed to confirm breast origin and rule out possible metastasis from other organs. On the other hand, for metastatic carcinoma identified in the liver, lung, or other organs in women, breast origin is usually considered in the differential diagnosis, even for patients without history of breast cancer, because of the high prevalence of breast carcinoma and possibility of presence of occult primary tumors in breast. Therefore, sensitive and specific breast markers are required for establishing or ruling out a breast origin, which will directly impact patient treatment.

Over the past half century, multiple diagnostic markers have been identified and tested in breast carcinoma. Gross cystic disease fluid protein-15 (GCDFP-15), mammaglobin, GATA3, SOX10, and TRPS1 are the breast carcinoma markers that are currently used in clinical diagnosis. These markers can be used for different subtypes of IBC and generally can be classified into 3 groups.

Different phenotypic carcinomas usually express markers reflecting their cell origin, which is the theoretical foundation for developing an organ-specific tumor marker. IBC with luminal differentiation (IBC-LD) can express some markers that expressed in the luminal cells, but not in the basal/myoepithelial cells of the normal breast duct, such as the hormonal receptors ER, PR, and AR; tyrosine kinase receptor HER2; and low-molecular-weight cytokeratin 7. Therefore, IBC-LD generally includes ER-positive IBC, HER2-positive IBC, and some TNBCs, such as AR-positive apocrine carcinoma and triple-negative ILC.

GCDFP-15, also known as BRST-2, was first described by Mazoujian et al in 1983.⁸⁷  Subsequent extensive studies showed relatively high expression of GCDFP-15 in ER-positive and HER2-positive IBCs (around 40%–70%) but low expression in TNBC (10%–30%).^88–90  In addition, GCDFP-15 is an apocrine marker with a strong relationship to AR status, and highly expressed in carcinoma with apocrine features. Mammaglobin, a 10-kDa secretory protein, was initially identified by Watson et al in 1996.⁹¹  Its expression in breast carcinoma is markedly increased compared to that in normal breast luminal cells. Mammaglobin exhibits a similar expression pattern and shows comparable sensitivity to that of GCDFP-15: relatively high in ER-positive and HER2-positive IBCs (around 40%–70%) but low in TNBC (10%–30%).^92–94  In the past decade, with the identification and use of another breast marker, GATA3, GCDFP-15 and mammaglobin were gradually replaced due to the following reasons: (1) GATA3 exhibits higher sensitivity for IBC than GCDFP-15 and mammaglobin. GCDFP-15- and mammaglobin-positive IBC are generally positive for GATA3; it is rare to see GCDFP-15–positive and mammaglobin-positive but GATA3-negative IBC; and (2) GATA3 usually shows uniform nuclear staining on IHC, especially in ER-positive and HER2-positive IBC (Figure 5, A through H). On the contrary, GCDFP-15 and mammaglobin exhibit cytoplasmic staining, and most of time, the staining is focal or patchy in tumor cells. Therefore, in small tumor samples, such as core needle biopsy, it can be difficult to determine whether the focal staining is truly positive or nonspecific background staining.

GATA3 belongs to the GATA family members of zinc finger transcription factors. It was reported to be an essential regulator for mammary gland luminal cell differentiation.^95,96  Several early studies showed that GATA3 expression is closely related to ER expression, with relatively low sensitivity in all IBCs (<50%) using GATA3 antibody clone HG3-31.^97,98  However, using a different antibody clone, L50-823, later studies demonstrated high sensitivity (>90%) of GATA3 expression in all primary and metastatic IBCs.^98,99  In ER-positive IBC, the sensitivity of GATA3 expression is up to 100%; in TNBC, its sensitivity varies from 20% to 60%.^100–103  Due to its higher expression rate, GATA3 gradually becomes the first choice of breast markers in the differential diagnosis workup. However, GATA3 is not specific for the breast. Besides its expression in normal breast luminal cells, GATA3 is also expressed in normal urothelial cells, apocrine sweat glands, salivary glands, skin squamous cells, and T-lymphocytes. Therefore, GATA3 is also the most sensitive marker for urothelial carcinoma and is highly expressed in many other malignancies, such as salivary ductal carcinoma, skin adnexal tumor, squamous cell carcinoma, mesothelioma, and T-cell lymphoma.^104,105  Although these tumors are not commonly listed in the differential diagnosis with breast cancer, the expression of GATA3 in a subset of ovary and lung adenocarcinomas may make the differential diagnosis challenging.^106,107 

The major disadvantage of GATA3, GCDFP-15, and mammaglobin is their limited use in TNBCs. Because most TNBCs are basal-type IBCs, these luminal markers are negative in more than 80% of metaplastic TNBCs and more than 50% of nonmetaplastic TNBCs.^{102,103,108,109}  Therefore, a diagnosis of metastatic TNBC would be challenging in the lung, liver, or other organs, as negative staining of these luminal markers does not exclude breast origin.

IBC with basal differentiation (IBC-BD) can express some markers that are expressed in the basal/myoepithelial cells but not in the luminal cells of the normal breast duct, such as tyrosine kinase receptor EGFR and high-molecular-weight cytokeratin 5. IBC-BD will not express hormonal receptors (ER, PR, and AR) or tyrosine kinase receptor HER2. Therefore, IBC-BD is generally triple-negative IBC. Gene expression profiling demonstrates that more than 70% TNBCs are basal type (Figure 6, A through H).¹¹⁰ 

SOX10 is a transcription factor that is involved in the differentiation of neural crest cells into glial cells and melanocytes. It was originally used as a melanoma and pan-schwannian marker to diagnose melanoma and nerve sheath tumors. Subsequent studies showed that SOX10 is also an excellent basal/myoepithelial marker. It is expressed in normal myoepithelial cells of the breast duct, salivary duct, sweat gland, and other glands with myoepithelial cells, with corresponding expression in multiple carcinomas with a basal/myoepithelial component, such as epithelial-myoepithelial carcinoma, myoepithelial carcinoma, and adenoid cystic carcinoma.^111–113  In 2013 studies of the breast, Cimino-Mathews et al⁵⁰  and Ivanov et al¹¹⁴  found that SOX10 is highly expressed in triple-negative and basal-like IBCs. Multiple recent studies showed that SOX10 is rarely expressed in ER-positive and HER2-positive IBC (<5%), while its expression in TNBC varies from 40% to 70%, which is slightly higher than that of GATA3 (20%–60%). Interestingly, SOX10 and GATA3 are found to be expressed in different portions of TNBC cells, and SOX10 is inversely correlated with GATA3 expression in TNBCs.^115,116  Therefore, SOX10 has been used by some institutions for clinical diagnosis of TNBCs over the last several years, as most GATA3-negative TNBCs are positive for SOX10, which, to some extent, resolves the diagnostic problem of GATA3-negative TNBCs. The main diagnostic pitfall for using SOX10 in this setting is differentiating TNBC from melanoma, because it can be expressed in both malignancies. Therefore, a pan-cytokeratin stain should always be included in this diagnostic workup. Due to the nature of SOX10 as a basal/myoepithelial marker, but not a true breast marker, there is a concern for its specificity as breast diagnostic marker. It is well known that some other basal cell markers, such as CK5/6 and p63, can also be expressed in most TNBCs, but they are not used clinically to diagnose TNBC. For this reason, our institution prefers not to include SOX10 in the workup to differentiate IBCs from carcinomas of other origins.

In 2020, it was reported that TRPS1 is a highly sensitive and specific breast marker, especially for TNBC.⁵²  Since then, it has been rapidly implemented for clinical diagnosis in the United States and other countries. TRPS1 is one of the 7 GATA family members of zinc finger transcription factors and plays an essential role in the development of cartilage, bone, and hair follicles. Loss of the TRPS1 gene can cause a hereditary autosomal dominant disorder, trichorhinophalangeal syndrome, that is characterized by distinctive facial features and bone and joint malformations.^117–120 

TRPS1 was first found to be associated with breast cancer by Radvanyi et al in 2005.¹²⁰  By using Affymetrix GeneChip microarray analysis, TRPS1 was found to be one of the 15 genes that were highly expressed in RNA level in breast cancer. Later studies showed that TRPS1 was expressed in more than 90% of both ER-positive and ER-negative breast cancer samples. In addition, it was reported as an essential regulator for the growth and differentiation of normal mammary epithelial cells and a lineage-specific transcription factor that is required for breast cancer cell survival,^121–123  although it is a little surprising that TRPS1 was originally thought to be a suppressor gene for breast cancer and a regulator (upstream of GATA3) of luminal progenitor.^122–125 

In 2020, through mining the Cancer Genome Atlas database, Ai et al⁵²  found that TRPS1 is highly expressed in not only ER-positive luminal IBC but also in HER2-positive and basal-type IBC. Compared to GATA3, TRPS1 shows similar sensitivity in ER-positive (98% versus 95%) and HER2-positive (87% versus 88%) carcinomas but exhibits significantly higher expression rate in TNBCs: 86% versus 21% in metaplastic TNBCs and 86% versus 51% in nonmetaplastic TNBCs. TRPS1 is also a highly specific marker for IBC, with little to no expression in other tumor types, including GATA3-positive urothelial carcinoma, lung adenocarcinoma, colon and gastric adenocarcinomas, pancreatic adenocarcinoma, cholangiocarcinoma, hepatocellular carcinoma, renal cell carcinoma, endometrial carcinoma, and melanoma. These findings have been confirmed in a tissue microarray study including 479 breast cancers and 1234 cancers of other organs. In 2021, Yoon et al¹²⁶  reported our experience of comparing TRPS1 with GATA3 and SOX10 in the clinical diagnosis of primary and metastatic TNBC. TRPS1 is expressed in more than 99% of triple-negative IDCs and ILCs, which is significantly higher than that of GATA3 (52%) and SOX10 (64%). TRPS1 is also highly expressed in metaplastic carcinoma (>90%) and is very useful in diagnosing primary and metastatic IBC in cytologic specimens.^53,126–128  Although TRPS1 has only been used in clinical diagnosis for less than 2 years, it demonstrates superior performance to GATA3 and SOX10. TRPS1 is expressed in almost all GATA3-positive and SOX10-positive IBC cases, indicating that it is a marker for both IBC-LD and IBC-BD (Figures 5 and 6).

It should be noted that TRPS1 is not expressed in some special types of TNBCs. TRPS1 and GATA3 are both negative in acinic cell carcinoma, most cribriform adenoid cystic carcinomas, and neuroendocrine carcinoma of the breast.¹²⁶  Apocrine carcinoma is the only type of TNBC we found so far that is GATA3-positive but TRPS1-negative (Figure 7, A through H).¹²⁶  Therefore, we recommend using TRPS1 and GATA3 in complement of each other in diagnosing TNBC in clinical practice. Furthermore, TRPS1 has been reported to be expressed in a small portion of ovarian carcinomas. Because the commonly used gynecologic marker PAX8 could also be expressed in some breast carcinomas, a differential diagnosis of breast and gynecologic carcinoma may be challenging, even with a workup panel including PAX8, TRPS1, and GATA3. A newly identified marker, SOX17, can be helpful in this situation. SOX17 was found to be highly expressed in ovarian and endometrial carcinomas (>90%) but not in breast carcinoma (Figure 8, A through H).¹²⁹ 

It is well known that a poorly differentiated carcinoma may lose organ-specific markers, which will lead to a diagnosis of “carcinoma of unknown origin,” even after extensive workup. Although several sensitive markers have been identified and used, patient history is still the most important clue for diagnosis of metastatic breast carcinoma. If there was a previous tumor, comparing its morphologic features and immunoprofile will greatly facilitate the correct diagnosis in a metastatic setting. In another situation, some peculiar carcinomas may express tumor markers that do not belong to their origins. For example, some breast cancers may express the lung marker TTF-1 (<5%), gastrointestinal marker CDX2 (<5%), and gynecologic marker PAX8 (<10%). A thorough workup panel for diagnosing metastatic breast carcinoma should include the above-mentioned breast markers TRPS1 and GATA3 and other organ-specific markers such as PAX8, CDX2, and TTF-1, which, with the help of clinical history and presentation, will eventually exclude other possibilities and confirm the breast origin.

In summary, IHC plays an important role in the diagnosis and classification of a variety of breast lesions. However, H&E evaluation always remains the principal step in diagnostic breast pathology. The selection of appropriate IHC markers, knowing the sensitivity and specificity of each marker, and standardization of technical aspects are all necessary in our clinical practice.

We would like to thank Ann Sutton of the Research Medical Library and Kim-Anh T. Vu of Department of Pathology at The University of Texas MD Anderson Cancer Center for their assistance in editing this document.