Best Practices in Diagnostic Immunohistochemistry: Myoepithelial Markers in Breast Pathology

At the light microscopic level, 2 distinct cell populations can be recognized lining the human mammary ductal and terminal ductolobular units: a luminally located layer of polarized epithelial cells and a basally located layer of myoepithelial cells (MECs).1 These cell populations display significant differences in normal function and protein expression profiles.2,3 Because the presence and distribution of MECs can significantly differ between the various breast proliferations, the use of immunohistochemical stains with relative specificity for MECs (myoepithelial markers) has become a staple of routine diagnostic breast pathology. The most common use of myoepithelial markers is to establish the presence or absence of an invasive carcinoma (Figures 1 and 2), a practice that is based on the fundamental principle that in situ carcinomas and nonneoplastic epithelial proliferations (Figure 3), in contrast to invasive carcinomas, retain a peripheral layer of MECs. Myoepithelial differentiation, however, may occur in neoplasms that have traditionally been considered as purely epithelial, and the antigens being recognized by some myoepithelial markers may also be present in stromal myofibroblasts, vascular smooth muscle cells, and even luminal/epithelial cells. These caveats (Figures 4 and 5) and exceptions, as well as related issues of marker specificity and sensitivity, underlie the diagnostic pitfalls that may be encountered with the use of myoepithelial markers in the pathologic evaluation of breast proliferations, especially in evaluating core biopsies. We review herein commonly used myoepithelial markers in breast pathology, as well as a selection of diagnostic settings where they may be useful.

Numerous antigens have been reported to display exclusive or preferential patterns of immunolocalization in breast MECs relative to their luminal/epithelial counterparts. These include CD109,4,5 caveolin 1 and 2,6,7 podoplanin,8 P-cadherin,9,10 maspin,11–13 nestin,14 p75,15–17 14-3-3 sigma (stratifin),18 smooth muscle actin (SMA),19 p63,20 CD10,21 smooth muscle myosin heavy chains (SMMHCs),22,23 h-caldesmon,23 calponin,23 S100,24–26 basal-type and high-molecular-weight cytokeratins,27 glial fibrillary acid protein,28,29 metallothionein,12 Wilms tumor 1 protein,30 CD44s,31 and vimentin,32 among others.3 A subset among these markers that have been most extensively studied (or are most widely available) are evaluated here.

In breast tissues, the contractile protein SMA has been demonstrated in the normal MEC in 88% to 100% of cases,19,33,34 in normal luminal/epithelial cells in 37% of cases,34 and in the MEC associated with benign lesions (such as sclerosing adenosis and radial scars) in 95.6% of cases.35 Although most invasive breast carcinomas are SMA negative, the diagnostic utility of this marker is limited by the frequent positivity of stromal myofibroblasts and vascular smooth muscle and pericytes, as well as subsets of the tumor cells in a number of histotypes (notably 10%–16.1% of invasive ductal carcinomas34,36).

Muscle-specific actin (MSA; also known as HHF-35) is comparable to SMA in its sensitivity for MECs,13 but shows a significantly lower specificity. In addition to the SMA-like patterns of cross-reactivity with stromal myofibroblasts, luminal/epithelial cells, and vessels, one study demonstrated at least weak positivity for MSA in 71% of invasive ductal carcinomas.34 

The contractile protein SMMHC is a structural component of myosin that is reportedly specific for smooth muscle cells, and its presence is thought to denote terminal smooth muscle differentiation.22,23,37 This marker stains nearly 100% of the MECs associated with normal breast ductules and benign breast proliferations, as well as vascular smooth muscle.38,39 As compared with SMA and MSA, SMMHC is significantly easier to interpret (Figure 2, B); only 8% of cases displayed significant cross-reactivity with myofibroblasts in one analysis.38 A subset of high-grade carcinomas was reported to be SMMHC positive in one study,22 whereas no cases of invasive carcinoma were found to be positive in another.39 The expression of SMMHC in MECs tends to be significantly reduced (relative to normal ductules) around ducts with ductal carcinoma in situ (DCIS), especially those of high histologic grade,17 which somewhat reduces its overall utility in distinguishing invasive from in situ lesions. In one study, the MECs associated with DCIS in 16% of cases were SMMHC negative.39 

The nuclear protein p63 is a homologue of p53 that is expressed in the basal epithelia of multiple organs. In the breast, p63 is positive in nearly 100% of normal MECs and those associated with benign proliferations (Figure 3, B).20,38 The advantages of this marker include (1) its nuclear staining pattern, which removes the interpretation difficulties that may be associated with the cross-reactivity for myofibroblasts seen with many of the other markers; (2) its lack of reactivity with myofibroblasts and vessels20,38; and (3) high sensitivity. Disadvantages include (1) the corollary to the stated advantages of nuclear staining: that is, the potential for the appearance of a discontinuous staining pattern, which may, in certain circumstances, result in a morphologic impression that MECs are absent; these apparent gaps in the periductal MEC layer (Figure 5) were seen in 10% of DCIS cases in one study38; (2) positivity in tumor cells, including in 15.7% to 23% of invasive ductal carcinomas,4,40 85% to 100% of adenoid cystic carcinomas, and the majority of metaplastic carcinomas20,43; and (3) the potential for a temporal reduction in its expression within archival material.44 

The common acute lymphoblastic leukemia associated antigen CD10 is strongly expressed in normal breast MECs and in MECs associated with noninvasive proliferations.39 This marker also stains stromal myofibroblasts, albeit with a lesser intensity than is seen with MSA or SMA, and does not stain vessels. Overall, CD10 is less sensitive than MSA, SMMHC, and MSA in identifying MECs in the neoplastic setting. The expression of CD10 in the MECs associated with DCIS is reduced (relative to normal MECs) in 34% of cases,17 and CD10 may be undetectable in up to 8% of DCIS cases.39 CD10 may be rarely expressed in the tumor cells of invasive ductal carcinomas45,46 and in a subset of poorly differentiated mammary sarcomas.47 

Calponin is a contractile element that is expressed in differentiated smooth muscle cells. Similar to most of the aforementioned markers, calponin is highly sensitive for MECs.23 Calponin is strongly expressed in normal breast MECs, vascular smooth muscle, and the MECs associated with noninvasive proliferations.38 Calponin-positive myofibroblasts can be detected in up to 74% of breast proliferations,38 but the staining pattern is generally light, patchy, and nonobscuring, in contrast to MSA and SMA. Overall, calponin is more specific but is less sensitive than SMA.48 It displays less cross-reactivity for myofibroblasts than SMA and MSA and similar levels as compared with CD10. Calponin is expressed, at least focally, in 33% of invasive ductal carcinomas,4 most cases of collagenous spherulosis,41 a subset of poorly defined mammary sarcomas,47 and poorly differentiated carcinomas.

The S100 protein is variably expressed in normal MECs and the MECs associated with noninvasive proliferations (Figure 3, C).12,13,25,35 However, this marker is also expressed in the lesional cells of such a substantial proportion of invasive and in situ carcinomas, and is so unreliably demonstrable in in situ carcinomas that its diagnostic utility as a myoepithelial marker is limited.24,26,34 S100 may also be expressed in normal luminal/epithelial cells and those associated with benign proliferations.49 Nonetheless, in addition to its traditional uses (to establish nerve sheath differentiation, for example), S100 expression may be supportive of a diagnosis such as microglandular adenosis (MA) and its associated proliferations.35,50 

Antibodies to a number of basal-type and high-molecular-weight cytokeratins, including cytokeratin (CK) 1, CK5, CK5/6, CK10, CK14, and CK17, react variably with breast MECs. The sensitivities of CK5/6, CK14, and CK17 for MECs are, in our experience, comparable to those of SMA, MSA, p63, and SMMHC, whereas 34βE12 (which recognizes CK1, CK5, CK10, and CK14) exhibits considerably lower sensitivity. These markers have the added advantage of showing no positivity in stromal myofibroblasts and vessels. Their principal drawback is a lack of specificity. Up to 32% of high-grade DCIS51 and 38% of invasive carcinomas52 are positive for at least one basal-type and high-molecular-weight cytokeratin. Subpopulations of the luminal/epithelial cells in the normal breast as well as in nonneoplastic lesions such as sclerosing adenosis and usual ductal hyperplasia (Figure 1, A) are also typically positive for basal-type and high-molecular-weight cytokeratins, especially CK5/6 (Figure 1, B) and 34βE12. Additionally, their diagnostic sensitivity and specificity tends to be somewhat more variable between different laboratories. Maximizing their diagnostic utility therefore requires a detailed knowledge of their patterns of reactivity in normal luminal epithelial cells and in neoplasms.

The calcium-dependent intercellular adhesion molecule P-cadherin is reliably expressed in normal breast MECs and the MECs associated with noninvasive breast proliferations.9,13 This marker shows no significant cross-reactivity with stromal myofibroblasts, luminal/epithelial cells, and vessels.9 P-cadherin, however, is also expressed in 20% to 40% of invasive carcinomas10,53 and in 25% of DCIS.54 

The p75 neurotrophin receptor (nerve growth factor receptor, p75^NTR or p75) consistently stains, in a membranous and cytoplasmic manner, normal breast MECs and the MECs associated with noninvasive breast proliferations.15–17 It is also expressed in vessels, nerves, and scattered epithelial/luminal cells in usual ductal hyperplasia.15,16 This marker shows minimal staining for the stromal cells associated with some benign proliferations; the rare periductal p75-positive stromal cells seen in some in situ carcinomas are generally not difficult to interpret.16 p75 is positive in 4.5% of pan-grade invasive breast carcinomas, but is expressed in substantially higher proportions of cancers with a basal-like phenotype,52 with which the expression of this marker bears a strong correlation.16 

Maspin consistently stains, in a nuclear and cytoplasmic manner, normal breast MECs and the MECs associated with noninvasive breast proliferations.11–13 It also stains subsets of epithelial cells, typically with significantly lesser intensity.12 There is no cross-reactivity with stromal myofibroblasts or vessels.12 Maspin expression is seen in 27.4% of invasive breast carcinomas,55 including most myoepithelial neoplasms,56 and 9.6% of DCIS cases.57 In one study, maspin expression was seen in more than 90% of all invasive lesions.58 Practically, it is most useful in studying MECs in normal tissue than in premalignant or malignant disease.

The vast majority of diagnostic scenarios in which there are multiple differential considerations can be successfully resolved by paying detailed attention to morphologic features. Nevertheless, myoepithelial markers are valuable diagnostic adjuncts, especially in core biopsies, and can facilitate the accurate diagnostic categorization of a given proliferation. Because of the aforementioned cross-reactivity patterns and the fact that lesional foci are typically minute, none of the myoepithelial markers enjoy a 100% sensitivity and specificity for MECs (Table). As such, at least 2 markers should be used to evaluate any given focus. Markers SMA (Figure 4), MSA, calponin, p63 (Figures 3, C and 5), maspin, p75, and SMMHC (Figure 2, B) are probably largely comparable in their sensitivities for MECs for scenarios in which the question is invasive carcinoma versus an entirely benign process, with the important caveat that sclerosing lesions of the breast may show patchy or no staining in a significant subset.59 For diagnostic questions of invasive versus in situ carcinoma (Figure 5), however, it should be noted that many of the markers show reduced staining in the MECs bordering the in situ lesions, and that this reduction differs between the markers.17 In one recent study, the proportion of DCIS cases showing reduced expression of the markers (relative to normal MECs) were as follows: 76.5% (SMMHC), 34.0% (CD10), 30.2% (CK5/6), 17.4% (calponin), 12.6% (p63), 4.2% (p75), and 1% (SMA).17 As previously noted, cross-reactivity patterns with stromal myofibroblasts, vascular cells, and tumor cells significantly differ between the markers. Regarding cross-reactivity for stromal myofibroblasts, the markers can be divided into 3 groups: (1) markers that show no significant cross-reactivity (p63, basal-type and high-molecular-weight cytokeratins, P-cadherin, and maspin); (2) markers that may show cross-reactivity but that generally do not result in interpretation problems (SMMHC, p75, calponin, and CD10); and (3) markers whose high frequency and intensity of cross-reactivity frequently cause interpretation problems (MSA and SMA). MSA, SMA, SMMHC, calponin, and p75 are expressed in vascular smooth muscle and/or pericytes, whereas basal-type and high-molecular-weight cytokeratins, maspin, and P-cadherin are generally negative in these cells. Therefore, whether the background stroma appears “exuberant” (vessel- and myofibroblast-rich) can be a consideration in the choice of myoepithelial marker. Finally, all the myoepithelial markers may be expressed in a subset of DCIS, invasive ductal carcinomas (the so-called basal-like carcinomas), as well as other carcinomas displaying myoepithelial or basal-like differentiation.52 Therefore, the diagnostic significance that is assigned to the presence of cells positive for a myoepithelial marker in a given proliferation should only be within the morphologic and immunophenotypic context of the background in which it is arising. Because myoepithelial markers are commonly used to establish the absence of MECs in a putative invasive carcinoma, it is important to ensure that the foci of interest, which are often small, are still present on the section that was stained. The most common diagnostic scenarios in which myoepithelial markers are used are outlined below.

This is the most common scenario in which myoepithelial markers are used. First, small epithelial tributaries adjacent to DCIS (typically high grade) may display a haphazard architectural pattern (Figure 1, A), surrounding stromal sclerosis (Figure 2, A) and/or an inflammatory infiltrate, raising the possibility of microinvasion. Furthermore, in situ carcinomas may involve architecturally infiltrative processes such as sclerosing adenosis and radial scars, resulting in an invasive appearance. Even when uninvolved by in situ carcinoma, these 2 and other benign processes may superficially mimic an invasive carcinoma. Myoepithelial markers (basal-type and high-molecular-weight cytokeratin CK5/6 is shown in Figure 1, B) are useful in all of these scenarios to establish the presence or absence of MECs around the epithelial nests of interest. The lack of myoepithelial layer will help in identifying morphologically subtle or inapparent invasive foci. As illustrated in Figure 2, B, the lack of myoepithelial layer (SMMHC) highlights invasive foci, which is otherwise not readily apparent. Myoepithelial markers highlighting nuclei (p63) have a distinctive advantage in certain settings (Figure 5).

The distribution patterns of MECs in breast papillary lesions and the diagnostic utility of myoepithelial markers in their classification have recently been reviewed.60,61 In the lesional papillae of conventional papillomas, MECs are present within the papillae and at the peripheries of the spaces in which they reside. In encapsulated papillary DCIS, MECs are absent in these 2 areas. In papillomas with atypia or DCIS, as well as papillary DCIS, MECs are absent in the atypical areas but are present at the peripheries of the involved spaces. In solid papillary carcinoma, MECs are absent within lesional papillae and are variably present in the peripheries of the involved spaces.60 Because the stroma associated with papillary lesions are often exuberant and/or sclerotic, nuclear stains such as p63 and maspin should be included in a myoepithelial marker panel for their evaluation. Figure 4 illustrates a sclerosing intraductal papilloma, where SMA stain highlights not only the MECs but also the background myofibroblasts within the sclerotic stroma.

A number of breast tumors manifest true myoepithelial differentiation, including adenomyoepithelioma and myoepithelioma, as well as the salivary glandlike tumors such as adenoid cystic carcinoma. These should not be confused with the so-called basal-like variant of breast cancers,62 a subtype of breast cancer brought to attention through gene expression array studies. These latter invasive ductal carcinomas may be prognostically and predictively significant and have significant epidemiologic associations such as the BRCA-1 mutation carrier status.52,63 There is, however, no consensus on an immunohistochemical definition of the basal-like phenotype.52,63 Whether or not the expression of luminal keratins is allowable, whether “triple negativity” is a requirement, and the precise complement of positive and negative immunoreactions that most closely correlates with the basal-like gene expression profile remain areas of investigation. Most authors include the expression of basal-type CKs, notably CK5 or CK5/6, in their definitions (as seen in Figure 1, B), but many of the other myoepithelial markers outlined above have also been used in various combinations.52,63 In one study, the expression of CK5 and/or CK14 defined the basal-like phenotype in invasive ductal carcinomas irrespective of the expression (or lack thereof) of the other markers.64 Depending on the myoepithelial marker that was used, between 15% and 20% of invasive mammary carcinomas appear to fall into the triple-negative immunophenotypic group, with the latter including most typical and atypical medullary carcinomas, metaplastic carcinomas, conventional myoepithelial-type carcinomas (such as low-grade adenosquamous carcinomas, malignant adenomyoepithelioma, and myoepithelial carcinomas), and others.52,63 

Microglandular adenosis (Figure 3) is a potentially mass-forming lesion whose constituent glands display open lumina with eosinophilic secretions and an infiltrative growth pattern (Figure 3, A).33,65–67 Although Diaz et al65 reported a layer of actin-positive cells around the ducts of MGA, other authors have consistently found that MGA ducts lack a myoepithelial layer.35,66 ,Figure 3 illustrates this, where the MGA is positive for S100 (Figure 3, B) but lacks a myoepithelial layer, as shown by a lack of p63 (Figure 3, C). The absence of a myoepithelial layer in MGA and its infiltrative growth pattern may result in an appearance that simulates well-differentiated invasive ductal carcinomas, including the tubular variant, and conventional myoepithelial markers are expectedly of limited value in this diagnostic scenario. Features that are significantly more likely to be seen in MGA than in well-differentiated invasive ductal carcinomas include (1) S100 immunopositivity (Figure 3, C), (2) estrogen receptor immunonegativity (Figure 3, D), (3) lack of background desmoplasia, (4) glands of uniform size and shape with eosinophilic secretions and that are lined by cells with vacuolated cytoplasm, and (5) presence of a basement membrane (laminin- or type IV collagen–positive). Microglandular adenosis with atypical areas or those that show morphologic transitions with invasive carcinomas are generally clearly identifiable as such.50,67 

In core biopsies, adenoid cystic carcinoma can occasionally be difficult to distinguish from collagenous spherulosis, invasive cribriform carcinoma, cribriform DCIS, or invasive tubular carcinoma, and myoepithelial markers are useful for these diagnostic scenarios. Adenoid cystic carcinoma and collagenous spherulosis may both be positive for SMA, S100, and p63.41 Calponin and SMMHC are expressed in collagenous spherulosis but not in adenoid cystic carcinoma.41 Furthermore, c-kit expression is present in >95% of adenoid cystic carcinoma but not in collagenous spherulosis, invasive cribriform carcinoma, or invasive tubular carcinoma.41,42 Neither invasive tubular carcinoma nor invasive cribriform carcinoma expresses conventional myoepithelial markers,12,35,59 and nests of cribriform DCIS display an outer rim of MECs.

Contemporary analysis of MECs and their differential protein expression has provided interesting, albeit preliminary, insights into the mechanisms of development and progression of breast cancers.2,68,69 In routine diagnostic surgical pathology, however, myoepithelial markers are most commonly used to supplement morphologic impressions in the classification of breast proliferations, especially in core biopsies. Although numerous myoepithelial markers are available, they differ in their sensitivity, specificity, and ease of interpretation, which may be attributed, to a large extent, to the variable immunoreactivity of these markers in stromal myofibroblasts, vessels, luminal/epithelial cells, and tumor cells. To avoid the pitfalls that are potentially associated with the use of myoepithelial markers, a panel-based approach of multiple markers is recommended. Markers that most effectively combine sensitivity, specificity, and ease of interpretation include SMMHCs, calponin, p75, p63, P-cadherin, basal CKs, maspin, and CD10. These markers, however, display varying cross-reactivity patterns and variably reduced expression in the MECs bordering in situ carcinomas.

Several specific caveats in the use of myoepithelial markers in breast pathology are worth emphasizing. Because these markers are negative markers (looking for absence), the laboratory should use adequate quality controls and titer to optimal dilutions (know the analytic sensitivity) for a given platform. In addition, pathologists should be aware of myoepithelial mimics (marker dependent: eg, SMA in sclerosing intraductal papilloma; Figure 4), and incomplete staining: for example, immunostaining with p63 results in widely spaced nuclear staining (Figure 5), sometimes leading to the impression of negative staining because of sampling.

In summary, several factors should go into the choice of a myoepithelial marker, including published evidence on its diagnostic utility, its availability, optimal reactivity that has been achieved in a given laboratory, and the specific diagnostic scenario. Furthermore, because a demonstration of the absence of MECs is often the diagnostic goal with use of these markers, particular attention must be paid to the time-honored principles of diagnostic immunohistochemistry, including ensuring that lesional foci remain on the sections that were stained, that positive and negative controls show expected patterns of reactivity, and that the final diagnosis is compatible with the hematoxylin-eosin–derived impression. When its use is deemed necessary, immunohistochemistry for MECs in breast pathology is most effective when conceptualized as supplemental, rather than central to routine morphologic interpretation.