Perhaps the most frequent use of diagnostic immunohistochemistry (IHC) in the surgical pathology laboratory pertains to breast biopsies by virtue of the volume and sheer difficulty of cases. In addition to using IHC for diagnostic problems with breast biopsies, breast biopsies lend themselves to the frequent use of IHC for prognostic and predictive tests. In addition, the diagnosis of breast carcinoma in the metastatic setting requires thorough knowledge of IHC, along with pitfalls for each stain.
This chapter addresses diagnostic issues involving stromal invasion, papillary lesions, atypical proliferative lesions, discrimination of ductal and lobular neoplasia, and identification of breast tumor types, Paget disease of the breast, fibroepithelial lesions, and metastatic breast carcinoma. The diagnostic section is followed by theranostic applications in breast cancer and a discussion regarding immunogenomics.
Myoepithelial Cells and Assessment of Stromal Invasion
Epithelial lesions of the breast are not only the most frequent lesions encountered by the surgical pathologist but also are the greatest source of concern in the differential diagnosis of benign versus malignant lesions. The lesion categories that typically need to be differentiated include nonneoplastic proliferative lesions versus malignant lesions, in situ carcinoma versus invasive malignancy, and pseudoinvasive lesions (adenosis, radial scar, sclerosing lesions, etc.) versus invasive malignancies. In addition, atypical ductal epithelial hyperplasia (ADH), papillary lesions, and microinvasive carcinoma (invasive focus less than or equal to 1 mm) lend themselves to IHC clarification in many instances.
In all of these diagnostic situations, it is the presence of the myoepithelial cell (MEC) in intimate relationship with the epithelial cells of the lesion that determines the difference between in situ and invasive disease and between benign pseudoinvasive lesions and invasive carcinoma. Microglandular adenosis (MGA), a distinct nonorganoid benign form of adenosis, is the only known exception to this statement (discussed later). The presence of MECs that envelop ductal-lobular epithelium, situated on the epithelial basal lamina, has always been considered to be the important criterion that separates invasive from noninvasive neoplasms. MECs can be visualized rather easily in normal breast ductules and acini, but when these structures dilate and fill with proliferating cells or are compressed, it is virtually impossible to visualize them on hematoxylin and eosin (H&E) stain. Replacing the yesteryear antibodies to S100 protein, high-molecular-weight keratin (HMWK), smooth muscle actin (SMA), calponin, and smooth muscle myosin heavy chain (SMMHC) are the more sensitive and specific antibodies to cytoplasmic components of MECs, along with the nuclear marker p63 ( Table 19.1 ).
Antibodies to S100 protein are not sensitive or specific for MEC and stain MEC in an erratic manner. The use of antibodies to maspin and CD10 has been tempered by the fact that they stain a variety of cell types, including luminal cells of the terminal duct lobular unit and tumor cells.
Cytokeratin cocktail antibodies, in addition to CK14 and CK17, identify MEC, but they also immunostain acinar cells, which makes it difficult to differentiate MECs because of their proximity to the acinar cells. Anti-SMAs react with stromal myofibroblasts in addition to MECs and thus are not specific for MECs. The cross-reaction with myofibroblasts makes it difficult to identify MECs specifically, especially in ductal carcinoma in situ (DCIS), in which there may be periductal stromal desmoplasia.
Although anti-SMA and muscle-specific actin HHF-35 stain MECs in the majority of benign breast lesions, there is substantial cross-reaction with stromal myofibroblasts, especially with SMA. Calponin and SMMHC are two antibodies that are more specific for MECs. SMMHC is a structural component (200 kD) unique to smooth muscle cells, which functions within the hexagonal array of the thick-thin filament contractile apparatus. Calponin, a 34-kD polypeptide, modulates actomyosin adenosine triphosphatase (ATPase) activity in the smooth muscle contractile apparatus and is unique to smooth muscle. In their analysis of 85 breast lesions, Werling and colleagues found that calponin and SMMHC always detected MEC in benign lesions, and SMMHC stained myofibroblasts in 8% of cases compared with calponin, which stained 76% of cases. It is also our experience that SMMHC and calponin are excellent antibodies, but calponin does stain stromal myofibroblasts to a greater extent than SMMHC.
p63, a homolog of the tumor suppressor protein p53, is used as a multitasker in multiple organs for the detection of MEC, basal cells (prostate), myoepithelial differentiation (breast metaplastic carcinoma and salivary gland tumors), and as a marker for squamous differentiation. The advantage of p63 in the diagnosis of stromal invasion is that it is present only in the nucleus, which renders it most specific for MEC in the breast, and it does not stain myofibroblasts. Some have used a cocktail of dual staining for SMMHC and p63 together. In our experience, using SMMHC and p63 is optimal for discerning MECs on difficult breast biopsies, especially diagnostic core biopsies ( Figs. 19.1 to 19.5 ). Distinguishing DCIS from invasive carcinoma on core biopsy can be crucial, as almost all patients with invasive carcinoma will have a sentinel lymph node (SLN) biopsy. An isoform of p63 (ΔNp63) is recognized by antibody p40. Studies comparing p63 and p40 reactivity in breast tissue and lesions have shown similar staining patterns for both antibodies.
An important pitfall to note is that around 5% of DCIS cases (especially the DCIS in the background of papillary lesion) completely lack MECs using any antibody ( Fig. 19.6 ). In these situations, critical appraisal of the histologic section is crucial to arrive at the correct diagnosis. For a lesion to be diagnosed as invasive carcinoma, the tumor cells should show “frank” infiltration in addition to lack of MECs in the periphery. It is also important to remember that p63 nuclear immunostaining results in apparent “gaps” of immunostaining because staining of cytoplasm of the MEC does not occur ( Fig. 19.7 ). Any nuclear staining around nests of tumor cells can be construed as evidence of the presence of MECs. Special care must be taken to exclude the nuclear staining of tumor cells around the periphery of neoplastic ducts, as p63 stains tumor cells in approximately 10% of cases (see Fig. 19.7 ).
Lesions that are especially difficult on core biopsies include the distinction of carcinoma in situ from invasive carcinoma in the presence of prominent periductal stromal desmoplasia (“regressive changes”) or heavy lymphoid infiltrates, lobular growth of rounded sheets of tumor cells (to the pathologist’s eye, invasion is “all or none”), infiltrating cribriform carcinoma, sclerosing adenosis (with or without DCIS involvement), cancerization of lobules, radial scars with stromal elastosis-desmoplasia, tubular carcinoma, and sclerosing papillary lesion. The optimal MEC antibodies needed to attack these difficult cases include both SMMHC and p63 ( Figs. 19.8 to 19.12 ).
A significant pitfall for misinterpretation of MEC antibodies such as calponin and even SMMHC is that these antibodies may immunostain the microvasculature around tumor nests. Initial examination of such a case with SMMHC will reveal immunostaining hugging the tumor nests, suggesting the presence of MECs ( Fig. 19.13 ). Examination at higher magnification will reveal the microvasculature around the tumor nests. When one then examines the p63, it is negative (see Fig. 19.13 ).
IHC for MEC is useful to help discriminate the three dominant benign lesions of the breast—sclerosing adenosis, MGA, and tubular carcinoma ( Table 19.2 )—but a detailed morphologic study of the lesion is essential. The MECs are seen by IHC in all forms of adenosis except the microglandular form, the only benign lesion that is known not to contain MECs. In addition to the distinct nonorganoid morphology of MGA, tubular adenosis, described by Lee and colleagues, may mimic both MGA and carcinoma but differs from MGA by containing MECs. MGA is positive with S100 protein, whereas sclerosing adenosis and tubular carcinomas are S100 negative.
The presence of MEC enveloping proliferating and sclerosing breast lesions is indicative of a benign or noninvasive process.
A combination of cytoplasmic SMMHC and nuclear p63 antibodies are the best discriminators for the presence of MECs, especially in desmoplastic-sclerotic proliferations.
MEC antibodies may be confirmatory for diagnosing microinvasive carcinoma (i.e., invasive carcinomas measuring no greater than 1 mm in largest dimension).
MGA: SMMHC−, p63−, S100+, ER−.
Pitfall: Immunostaining of myofibroblasts and vascular walls with SMMHC; and p63 occasionally stains neoplastic cells.
MEC , Myoepithelial cell; MGA , microglandular adenosis; SMMHC , smooth muscle myosin heavy chain.
|Diagnosis||Histology||Myoepithelial Cells||Collagen IV||Other Immunohistochemistry|
|Tubular carcinoma||Invasive tear-drop shape tubules, apical snouts, desmoplasia||Absent||Absent||EMA+, ER/PR+|
|Microglandular adenosis||Round glands in fat lined by flat to cuboidal epithelium. Inspissated secretions within glands||Absent||Present||S100+, EMA−, ER/PR−, GCDFP15−|
|Tubular adenosis||Tubules sectioned longitudinally and lacks lobulocentric distribution||Present||Present||S100−|
|Sclerosing adenosis||Lobular growth pattern, epithelial cell atrophy and lobular fibrosis||Present (relative abundance)||Present||S100−|
Immunohistochemistry of Papillary Lesions
Papillary lesions range from benign papilloma to atypical papilloma to papillary carcinoma (in situ and invasive). There are several reports on the use of MEC markers to distinguish between different categories. A papillary lesion is classified as a papilloma if there is a uniform layer of MECs in the proliferating intraluminal component of the lesion, whereas the absence of MECs would suggest an atypical papillary lesion. Atypical papilloma is a term often used when ADH overgrows the papilloma. These atypical areas generally lack MECs by immunoperoxidase examination. However, the absence of staining for HMWKs such as 34βE12, CK5, and CK5/6 within the proliferative component is more helpful in classifying a lesion as atypical ductal hyperplasia or in situ carcinoma. The distinction between a papilloma, atypical papilloma, and papillary DCIS (either de novo or DCIS involving papilloma) is not that problematic in the majority of cases, and a diagnosis can be made using morphology and IHC staining ( Fig. 19.14 ). The more difficult and confusing area is the distinction between a large cystic in situ lesion and an invasive carcinoma. In situ papillary carcinoma has been referred to by different names in the literature. The term intracystic papillary carcinoma has been used for a single mass-forming cystic lesion with malignant papillary proliferation. Papillary DCIS is a term that has been used for more diffuse lesions. The use of MEC markers to assess invasion in these lesions has yielded variable results. In an immunohistochemical study of papillary breast lesions, Hill and Yeh found consistent staining patterns in cases originally diagnosed as papilloma or invasive papillary carcinoma, but found variable staining in cases diagnosed as intraductal papillary carcinomas. Of the nine intraductal papillary carcinomas in their series, four cases showed unequivocal basal MECs by IHC, one case showed partial discontinuous staining, and four cases were predominantly negative for basal MECs. The authors found that lesions originally classified as intraductal papillary carcinoma but lacking basal MECs by IHC were uniformly large, expansile, papillary lesions with pushing borders and a fibrotic rim. The authors hypothesized that such lesions form a part of the spectrum of progression intermediate between in situ and invasive disease and suggested that these lesions should be termed encapsulated carcinoma . Collins and colleagues have also favored such designation. In some recent reviews of papillary lesions, an attempt has been made to classify the lesions in a uniform fashion using morphology and IHC. The papillary lesions are now classified as papilloma, papilloma with ADH (atypical papilloma), papilloma with DCIS, papillary DCIS, intracystic papillary carcinoma (encysted or encapsulated papillary carcinoma), solid-papillary carcinoma, and invasive papillary carcinoma.
An intraductal papilloma with usual ductal hyperplasia shows the presence of MECs within and around the lesion, and the proliferative component is positive for HMWK CK5. Atypical papilloma/papilloma with DCIS/papillary DCIS show a peripheral layer of MECs, but the proliferative component is negative for myoepithelial markers and CK5.
The problem in diagnosis arises from the fact that intracystic and solid-papillary carcinomas have the morphology of an in situ lesion, but they lack the presence of MECs around the periphery ( Fig. 19.15 ). Intracystic papillary carcinomas generally retain strong expression for collagen IV completely around the lesion ( Fig. 19.16 ), suggesting that these are noninvasive lesions. However, it is to be noted that collagen can also be laid down around the invasive front of carcinoma, but true invasive cancers generally show a weak and discontinuous type of staining. We do not recommend using collagen IV in routine diagnostic testing, due to inconsistent staining and interpretation issues. Conceptually, encapsulated and solid papillary carcinomas are somewhere in the spectrum between in situ and invasive carcinomas, but the clinical behavior of these lesions is more akin to in situ disease. Therefore these lesions are best managed as a variant of in situ carcinoma for practical purposes. However, it is extremely important to analyze the resection specimen on these lesions in entirety by histologic evaluation due to the not-so-infrequent presence of frank invasion (invasion within or beyond the fibrotic rim) in the periphery of these lesions. We believe it is the presence of these minute foci of invasive carcinoma that are responsible for occasional metastatic disease reported with intracystic papillary carcinomas. The MEC staining pattern for each papillary lesion is summarized in Table 19.3 .
|Papillary Lesions||Myoepithelial Markers (p63 and Smooth Muscle Myosin Heavy Chain)||CK5 or CK5/6||Clinical Behavior|
|Papilloma||Positive within and around ducts||Positive in proliferative epithelium||Benign|
|Papilloma with ADH/DCIS or papillary DCIS||Reduced/absent within, positive around ducts||Negative within proliferative epithelium||Risk for invasive malignancy|
|Encapsulated papillary carcinoma||Reduced/absent within, mostly negative around||Negative within proliferative epithelium||Similar to DCIS, unless frankly invasive|
|Solid papillary carcinoma||Reduced/absent within, mostly negative around||Negative within proliferative epithelium||Similar to DCIS, unless frankly invasive|
|Solid papillary neoplasm with reverse polarization||Reduced/absent within, mostly negative around||Positive in lesional epithelium||Generally indolent clinical course|
There are two other lesions that have papillary architecture, but are usually not considered in the differential diagnosis of papillary lesions due to their uncommon occurrence. The first is an adenomyoepithelioma, which, as the name suggests, shows a mixture of glandular and MECs. Adenomyoepitheliomas often arise in a background of adenosis or a papillary lesion. A pancytokeratin stain highlights glandular elements, and the MECs (which are often in abundance) are highlighted by myoepithelial markers. However, the reactivity for myoepithelial markers varies from case to case, and it is not uncommon to see patchy loss of some myoepithelial markers in the periphery of the lesion. This latter finding should not be misconstrued as an evidence for invasion. Most adenomyoepitheliomas are clinically benign, with only occasional cases reported to undergo malignant transformation. Malignancy in adenomyoepithelioma is diagnosed based on morphologic features and infiltration into the surrounding tissue rather than based on IHC stains. The second uncommon lesion is the so-called breast tumor resembling tall cell variant of papillary thyroid carcinoma (BTRPTC). It has characteristic cytomorphologic features, but often comes across as a solid-papillary lesion. Clinically it is identified as an incidental mass lesion, 1 to 3 cm in largest dimension, which shows solid-papillary nests of bland appearing proliferating epithelium. The epithelium can show grooves and occasional nuclear inclusions. The lesional nuclei are arranged away from the stromal aspect (known as reverse polarization), and hence it is also named solid papillary neoplasm with reverse nuclear polarization (SPNRNP). The proliferating epithelium stains strongly for CK5 (staining similar to usual ductal hyperplasia), but is only weak and patchy positive for ER. There is no staining for myoepithelial cells around the lesion (staining similar to EPC and SPC). The exact clinical course of this lesion is uncertain, but limited clinical experience suggests an indolent clinical course. The lesion may locally recur if incompletely excised. The potential for distant recurrence is questionable. BTRPTC is a distinct lesion with a characteristic staining pattern that needs to be distinguished from other established papillary lesions for proper patient management.
The MECs and CK5 are present in the proliferative cellular component of a papilloma, but are absent in the areas of atypical ductal hyperplasia or DCIS.
MECs are uniformly present around the periphery of the lesion in a papilloma, atypical papilloma, papilloma with DCIS, papillary DCIS, but absent at the periphery of intracystic and solid-papillary carcinomas.
Caution is advised in diagnosing invasion based on MEC antibodies in a papillary lesion on a core biopsy-recommend complete excision for assessing invasion.
DCIS , Ductal carcinoma in situ; MEC , myoepithelial cell.
Proliferative Ductal Epithelial Lesions and In Situ Carcinoma
Differences in cytokeratin expression have been described between hyperplasia and DCIS. The antibody 34βE12 recognizes CK1, CK5, CK10, and CK14, and these keratins are typically found in myoepithelium and squamous epithelium. Normal breast MECs and proliferative duct epithelium express 34βE12 ( Fig. 19.17 ). The expression is lost in ADH. Low to intermediate grade DCIS is also largely negative for 34βE12 (see Fig. 19.17 ). Most low/intermediate grade DCIS are uniformly positive for CAM5.2, reflecting a shift away from HMWKs to the more simple keratins 8 and 18. The 34βE12 immunostaining profile for DCIS and ADH is very similar and cannot be used to help distinguish DCIS from ADH, but can be an aid to histomorphology in separating DCIS from florid ductal epithelial hyperplasia (DEH) in difficult cases. CK5/6 (clone D5/16B4) antibody has also been shown to be largely negative in DCIS. Similar results have been obtained with CK5 antibody. This expression of HMWKs (or basal keratins) in usual hyperplasia with loss in ADH and DCIS suggests that atypical lesions try to acquire a more “luminal” phenotype. Adding to the same theme, usual hyperplasia is generally negative or patchy positive for estrogen receptor (ER), but atypical hyperplasia and low/intermediate grade DCIS are often strongly and diffusely ER positive. So in a lesion with ambiguous morphology for ADH, a combination of CK5 and ER may be helpful in rendering a more definitive diagnosis. A CK5+ and ER low/negative immunophenotype of the proliferative component would favor usual hyperplasia, whereas the opposite (CK5−, ER strongly positive) profile would favor ADH/DCIS.
There are a few pitfalls for using these IHC stains for making a diagnosis of ADH/DCIS. First of all, this panel is not valid for columnar cell lesions, as even benign columnar cell changes strongly express ER. Second, apocrine lesions (benign/atypical/or malignant) are generally negative for ER, and are either negative for CK5 or show variable reactivity for CK5. Finally, the basal-like DCIS are almost always positive for CK5 and negative for ER. Another aspect to remember is that native luminal epithelium—that is, the normal epithelium that rests upon the myoepithelium—is almost always negative for CK5 and should not be interpreted as “atypia.” The results for CK5 should be interpreted in the intraductal proliferative component for atypia assessment. Therefore CK5 and ER should be used in conjunction with defined morphologic criteria for diagnosing ADH/DCIS (see Fig. 19.17 ).
The diagnosis of atypia in papillary lesions is also very challenging. Fortunately, the same cytokeratin patterns of immunostaining hold up for the differential of ADH/DCIS in a papilloma versus florid hyperplasia in a papilloma.
Cytokeratin antibody 34βE12, CK5/6, and CK5 intensely stain florid ductal hyperplasia of the breast, which may be useful in separating florid hyperplasia in ducts or papillomas from ADH/DCIS.
Both ADH and DCIS lack 34βE12, CK5/6, and CK5 antibody staining and cannot be distinguished by IHC.
Estrogen receptor expression is diffuse and strong in ADH and low-grade DCIS, but is either negative or patchy in ductal hyperplasia.
ADH , Atypical ductal epithelial hyperplasia; DCIS , ductal carcinoma in situ; IHC , immunohistochemistry.
Tumor Type Identification by Immunohistochemistry
Cell Adhesion: Ductal Versus Lobular Carcinoma
Based on cell cohesiveness, the two broad categories of breast carcinoma (invasive or in situ) are ductal and lobular types. Ductal carcinoma in situ increases the risk of invasive malignancy at the local site, whereas lobular carcinoma in situ (LCIS) is considered a marker of generalized increased risk of invasive malignancy, although some recent data suggest precursor properties for lobular carcinoma in situ. Invasive ductal carcinomas (IDCs) are often unifocal lesions compared with invasive lobular carcinoma, which are often multifocal and/or more extensive than what is estimated on clinical and mammographic examination. Distant metastases from ductal carcinoma preferentially involves the lung and brain, whereas metastases from lobular carcinoma more often involve the peritoneum, bone, bone marrow, and visceral organs of gastrointestinal and gynecologic tracts. In spite of the listed differences, at present, with the combined multimodality therapy, there appears to be no difference in disease-free and overall survival between ductal and lobular carcinomas. However, there are enough significant differences in patient preoperative evaluation and subsequent treatment that an accurate diagnosis is warranted at the time of core biopsy. At some breast cancer centers, a preoperative (before lumpectomy or mastectomy) magnetic resonance imaging (MRI) of the breast is performed to evaluate the extent of disease with a core biopsy diagnosis of invasive lobular carcinoma. A core biopsy diagnosis of ductal versus lobular carcinoma is also important if the patient will be treated by neoadjuvant chemotherapy. Although therapy response is better predicted with tumor receptor status rather than morphologic tumor type, studies do show response to neoadjuvant chemotherapy (NACT) only in a subset of ductal cancers with no or minimal effect on lobular cancers. Therefore pathologists have to strive hard to give the best diagnosis possible for current management and also for the future as specific therapies become available. One study has shown the relative effectiveness of aromatase inhibitor letrozole over tamoxifen for patients with lobular carcinoma.
Strong E-cadherin (ECAD) membranous staining has been long used to define ductal carcinomas. Ductal carcinomas (in situ or invasive) retain membranous ECAD because they do not show homozygous mutation/silencing of the ECAD gene. Mutation of the ECAD gene either leads to a mutant protein that loses its adhesive properties or there is not enough protein to function as an adhesive molecule.
The ECAD gene, CDH1 , is a large gene located on 16q22.1. The ECAD protein has an intracytoplasmic portion, an intramembranous portion, and an extracellular domain. Cell-to-cell adhesion through ECAD is also critically dependent on the subplasmalemmal cytoplasmic catenin complexes (α, β, γ, and p120 isoforms) that link ECAD to the actin cytoskeleton of the cell. Abnormalities of the catenins or ECAD gene expression can result in a variety of immunohistochemical ECAD pathologies. Lobular carcinomas studied at the genetic level have often shown ECAD mutation that accounts for the loss of cohesiveness of the tumor cells. The majority of these mutations have been found in combination with loss of heterozygosity (LOH) of the wild type ECAD locus (16q22.1), a hallmark of classical tumor suppressor genes. Immunohistochemically, this correlates with either a complete absence of the ECAD protein or abnormal localization (apical or perinuclear). This abnormal localization may be dependent on the type of mutation. Truncation mutations produce an ECAD product that is inept at binding to neighboring cells, resulting in a histologic pattern of widely dyshesive cells that are completely negative for ECAD by IHC (e.g., classic infiltrating lobular carcinoma [ILC]) ( Fig. 19.18 ). Loss of membrane staining may be associated with a granular cytoplasmic immunostaining ( Fig. 19.19A and B ) that represents cytoplasmic solubilization of a portion of the truncated protein. Proximal truncation mutations may result in the inability of ECAD to bind to the catenin complex, resulting in a short ECAD represented by focal or dot-like membrane immunostaining (see Fig. 19.19C ). Patients with focal staining of LCIS cells with ECAD may have an ipsilateral risk of carcinoma akin to DCIS. Mutations in the catenin complex can also lead to dysfunctional ECAD and loss of membrane staining. Although deletions of the CDH1 gene as a result of LOH are seen in ductal carcinomas, they are not early events and are not usually associated with the point mutations seen in lobular neoplasia.
In the majority of cases, ECAD staining is unequivocal (positive or negative), and can be solely used in distinguishing ductal from lobular carcinomas. In a minority (~15% cases) of cases, the stain may be difficult to interpret. Another stain that could be used in such situations is p120. This stain represents p120 catenin, which binds with ECAD on the internal aspect of the cell membrane to form the cadherin-catenin complex ( Fig. 19.20A ). The complex is essential for the formation of intercellular tight junctions and is composed of an external domain of calcium-dependent ECAD and an internal domain of ECAD to which is bound the α, β, and p120 catenins. The α and β catenins are complexed with the carboxy-terminal cytoplasmic tail of ECAD, whereas the p120 catenin is anchored to ECAD in a juxta-membranous site. p120 is actively involved in the status of cell motility, ECAD trafficking, ECAD turnover, promotion of cell junction formation, and regulation of the actin cytoskeleton. The binding of p120 to ECAD stabilizes the complex and increases the half-life of membrane ECAD by slowing the normal turnover of ECAD that normally occurs by cellular endocytosis. p120 that is bound to ECAD exists in equilibrium, with a small cytoplasmic pool of p120. When ECAD is absent, the cytoplasmic pool of p120 increases. Therefore, in normal ducts and ductal carcinomas, p120 shows a membranous pattern of staining (see Fig. 19.20B and C ). In contrast, lobular carcinomas (with absent or nonfunctional ECAD) show strong cytoplasmic p120 immunoreactivity (see Fig. 19.20D and E ). This positive cytoplasmic staining for lobular carcinoma is much easier to interpret than ECAD negative staining. A combination of ECAD and p120 drastically reduces the number of ambiguous diagnoses and better delineates (or diagnosed with increased confidence) the category of mixed ductal and lobular carcinoma (see Fig. 19.20F ). These mixed carcinomas comprise no more than 10% of all breast carcinomas and probably arise due to “late” ECAD inactivation within a ductal carcinoma. In contrast, loss of ECAD protein occurs very early in lobular carcinogenesis. Lack of ECAD staining and strong p120 cytoplasmic staining is observed in all morphologically characterized lobular carcinoma in situ and atypical lobular hyperplasias ( Fig. 19.21 ). In addition, lack of ECAD within minimal epithelial proliferation in the breast terminal duct lobular unit defines atypical lobular hyperplasia and distinguishes it from mild ductal hyperplasia. This distinction is important because patients with ALH are typically referred to a risk clinic. Although some studies have advocated surgical excision with core needle biopsy diagnosis of atypical lobular hyperplasia, most breast care providers recommend referral to high-risk clinic rather than surgical excision. The IHC stains support the notion that the term lobular hyperplasia has no significance in breast pathology.
Immunohistochemical and molecular methods not only aid with the diagnostic issues, but to some extent they also demand that one looks at the morphology that has been learned in a new light. All invasive breast carcinomas that infiltrate in a “single-file” pattern with a low nuclear grade are not lobular carcinomas, as IDCs also have this pattern.
The morphologic assessment of a ductal or lobular phenotype is not without controversy, and has limitations. The classic ILC is composed of small cells with bland cytology and some “plasmacytoid” features. The growth pattern is completely dyshesive. Breast tissue can grossly appear normal (as well as the mammogram) yet show widespread dyshesive carcinoma of the classic type. These tumors are uniformly ECAD negative and associated with specific patterns of systemic metastases. IDC can also show patterns seen in ILC—for example, “single-filing” of tumor cells, targetoid patterns, and regional dyscohesiveness. Such patterns may be confusing but are readily resolved with ECAD immunostaining. There are subgroups of morphologically “indeterminate” lobular/ductal phenotypes. ECAD separates these groups distinctly in most cases and demonstrates the existence of mixed lobular-ductal phenotypes in a minority of cases. ECAD stains MECs, a pitfall for misinterpretation of LCIS as DCIS (see Fig. 19.18A ).
The morphologic reproducibility of distinguishing IDC from ILC and LCIS from DCIS is less than optimal. There can be substantial variation in the interpretation of ILC versus IDC and LCIS versus DCIS. For this reason alone, ECAD and similarly p120 IHC could be justified to aid in correctly classifying these lesions.
ECAD stain is a useful diagnostic adjunct in cases with indeterminate morphology.
p120 further enhances diagnostic accuracy by being a “positive” stain for lobular carcinoma.
Lobular lesions are characteristically negative for ECAD and demonstrate intense cytoplasmic immunoreactivity for p120.
Aberrant ECAD immunostaining may be cytoplasmic or punctate/membranous, depending on the type of EACD mutation. p120 catenin in these instances proves the lobular phenotype.
Normal ducts and ductal lesions demonstrate membranous staining for ECAD and p120.
ECAD , E-cadherin.
Lobular Carcinoma Variants and Former Lobular Variants
Pleomorphic Lobular Carcinoma
Described by Bässler in 1980 and further detailed by Weidner, Eusebi, and Reis-Filho, the genetic, immunohistologic, and clinical features have been sufficiently detailed to recognize invasive pleomorphic lobular carcinoma (PLC) and pleomorphic lobular carcinoma in situ (PLCIS) as a distinct clinicopathologic entity. Based on cell cohesiveness, PLC and PLCIS are basically a subtype of lobular carcinoma. The histologically recognizable PLC and PLCIS are almost always ECAD-negative (or show aberrant staining) and demonstrate strong cytoplasmic immunoreactivity for p120. Histologically, these show grade 3 nuclei with a dyshesive pattern of growth in both in situ and infiltrating varieties ( Fig. 19.22 ). The in situ component may be discovered on mammograms as calcifications. The core biopsies demonstrate in situ dyshesive grade 3 nuclei, with some cases showing comedonecrosis and calcification. A comprehensive analysis of 26 PLCs revealed a closer association between PLC and classic lobular carcinoma than between PLC and ductal carcinoma. The authors analyzed 26 cases of PLC, 16 cases of classic lobular carcinoma, and 34 cases of IDC by IHC, array comparative genomic hybridization (aCGH), fluorescence in situ hybridization (FISH), and chromogenic in situ hybridization (CISH). Comparative analysis of aCGH data suggested the molecular features of PLC (ER/PR+, ECAD-negative, 1q(+), 11q(−), 16p(+) and 16q(−)) were more closely related to those of classic ILC than IDC. However, PLCs also showed some molecular alterations that are more typical of high-grade IDC than ILC (p53 and HER2 positivity in some cases, 8q(+), 17q24–q25(+), 13q(−), and amplification of 8q24, 12q14, 17q12, and 20q13). Some of these IDC-like alterations may be responsible for the aggressive biology of PLC.
Sneige and colleagues studied 24 cases of PLCIS by IHC and found them to be universally positive for ER (100%). They also showed frequent p53 reactivity (25%) and moderate to high proliferative activity; in addition, HER2 positivity was seen in 1/23 cases (4%). Of these 24 cases, 14 were associated with PLC, which showed similar IHC profile. Our experience with PLC and PLCIS is also very similar. Although HER2 overexpression/amplification may be seen in PLCs, the HER2+ rate is not very high, as previously reported, and PLCs are ER+ in the majority of cases, although expression levels may vary from case to case. Because there is high likelihood for developing invasive carcinoma in the vicinity of PLCIS, these lesions should be managed similar to DCIS.
Described originally by Fisher in 1977 as a lobular growth pattern with tiny tubules and single-filing characteristic of lobular carcinoma, the prognosis was described as intermediate between that of pure tubular carcinoma and ILC. This lesion had been categorized as a variant of ILC because of the small cells and characteristic ILC pattern of single-filing, and targetoid infiltration.
Wheeler as well as Esposito documented uniform membranous ECAD immunostaining in the tubules and lobular-appearing components ( Fig. 19.23 ), and discovered that pure LCIS and mixed LCIS/DCIS predominate in these lesions. The combination of small, rounded tubule profiles with infiltrating lobular-like patterns that are ECAD+ is a ductal immunoprofile.
The term histiocytoid breast carcinoma (HBC) was coined by Hood and colleagues due to tumor cells resemblance to histiocytes. In 1983, Filotico described a case of lobular-appearing carcinoma with histiocytic features. Subsequent reports assumed that this variant was of lobular type by virtue of the characteristic infiltrating pattern. Immunohistologic studies have provided further clarification. Gupta and colleagues studied 11 cases of histiocytoid carcinoma and found that 8 of 11 cases lacked ECAD and 8 of 11 had LCIS. The authors concluded that histiocytoid carcinoma lacked distinct morphologic and clinical features and has an immunohistochemical profile that is consistent with both ductal and lobular differentiation. The current evidence suggests that HBC is a morphologic pattern that can be observed in ductal, lobular, and apocrine tumors, and is not a distinct entity by itself.
Immunohistochemistry for Identifying Special Types of Breast Carcinomas
Invasive Micropapillary Carcinoma—Use of Epithelial Membrane Antigen
Tight clusters of neoplastic cells surrounded by clear spaces characterize the invasive micropapillary carcinoma. The cell clusters are devoid of fibrovascular cores (unlike papillary carcinoma) and often display tubular structures in the center. The stroma is typically described as “spongy,” with little or no desmoplasia of the surrounding tissue. Some ductal carcinomas of “no special type” (NST) also show clear spaces around neoplastic cells, which are likely due to retraction of the intervening fibrotic stroma and should not be confused with micropapillary morphology. Fortunately, the distinction between true micropapillary carcinoma and NST carcinoma with retraction artifact can be easily made by epithelial membrane antigen (EMA) (or MUC1) stain. Ductal carcinoma of NST shows an apical or cytoplasmic staining with EMA. In contrast, invasive micropapillary carcinomas show accentuation of the basal surface (stroma facing) of the neoplastic cells ( Fig. 19.24 ). This “reverse polarity” of the neoplastic cells is a characteristic feature of invasive micropapillary carcinoma. The distinction between a ductal NST and micropapillary carcinoma may not be clinically very significant, because stage for stage, there is no significant difference between the two entities. However, micropapillary morphology (even when small) is highly predictive of lymph node metastases, and these tumors also more commonly tend to involve the skin and chest wall. The incidence of axillary lymph node metastasis has been reported to be as high as 95%. However, we believe the actual incidence of axillary lymph node involvement in routine clinical testing is close to 60%. Acs and colleagues showed that even partial reverse cell polarity, defined as prominent linear EMA reactivity on at least part of the periphery of tumor cell clusters, has the same implication as micropapillary differentiation, and these tumors may represent part of a spectrum of invasive micropapillary carcinoma. A confident diagnosis of a true micropapillary carcinoma on a core biopsy helps the surgeon to plan appropriate management, namely special attention and clinical exam of the axilla, to perform fine needle aspirate if axillary nodes are slight enlarged (but not definitely suspicious), and to request intraoperative frozen section, even if the lymph node is grossly negative.
“Basal-Like” Carcinoma—Use of Basal Cytokeratins
Routine hormone receptor (HR) protein and HER2 oncoprotein analysis on invasive breast carcinomas in the last decade has delineated clinically significant subgroups. One such group is that of “triple negative” tumors—that is, tumors that are negative for all three biomarkers, ER, PR, and HER2. These tumors have been known to be clinically aggressive, and therapeutic options are limited because these are not amenable to HR-based therapy or HER2-targeted therapy. The so-called basal-like breast carcinomas constitute at least 80% of the “triple negative” tumors. The “basal-like” subtype was initially recognized by gene expression profiling studies. Basal-like carcinomas are histologically characterized by high Nottingham grade, geographic necrosis, good circumscription, and mild to moderate host lymphocytic response. These tumors generally show reduced (not absent) ECAD membranous expression, but show strong membranous p120 immunoreactivity (unpublished data), and therefore we consider them as subtypes of ductal carcinomas. Basal-like carcinomas are characteristically “triple negative” and show expression of basal-type cytokeratin (CK5/6, CK14, CK17), epidermal growth factor receptor (EGFR), vimentin, and p53. Often, a panel of basal-type cytokeratins and EGFR in triple negative tumors is used to identify basal-like carcinomas ( Fig. 19.25 ). The antibodies used to identify the basal-like variant are an example of “genomic application” of IHC. This details fundamental immunohistologic profiles, used as surrogate markers, that reflect a genomic profile. Antibody to CK5 (clone XM26) is more sensitive (but equally specific) than CK5/6 (clone D5/16B4) for identifying basal-like breast carcinomas. The immunohistochemical studies have also confirmed the existence of in situ carcinoma of basal phenotype. Gene expression studies have consistently identified basal-like carcinomas to have poor prognosis. These tumors occur in both premenopausal and postmenopausal patients; however, identifying basal-like carcinoma in the young premenopausal patient may suggest the presence of hereditary breast and ovarian carcinoma syndrome. Although there are no specific chemotherapeutic drugs currently available to treat these patients, data (especially on PARP inhibitors) are emerging, and it is important to recognize these tumors as therapies become more refined.
Metaplastic Carcinoma—Use of Keratins, Melanoma, and Vascular Markers
Metaplastic carcinoma comprises a group of heterogeneous neoplasms that exhibit pure epithelial or a mixed epithelial and mesenchymal phenotypes. Diagnosis is not problematic when there is a recognizable component of metaplastic carcinoma—that is, an obvious adenocarcinoma, adenosquamous or squamous cell carcinoma, or osseous or chondroid differentiation. The most problematic cases are the ones that predominantly show spindle cell morphology without an obvious epithelial or DCIS component ( Fig. 19.26 ). This is usually the issue on a core biopsy rather than on an excision specimen. IHC stains can be helpful in this situation. A panel composed of multiple keratin stains (CAM5.2, AE1/3, 34βE12, CK5, and CK7) and EMA is more useful than a single keratin. Another sensitive and specific marker for spindle cell metaplastic carcinoma is p63, and should always be included in the panel. Vimentin expression in the tumor does not exclude a spindle cell carcinoma. Vimentin expression has been found in 50% of hormone-independent cell lines, and because metaplastic carcinomas are usually negative for receptors, vimentin expression is actually expected. If all the keratins, EMA, and p63 fail to show any immunoreactivity on a core biopsy, complete excision of the lesion should be recommended. In many cases, an epithelial component is present only focally. Although every effort should be made to prove an atypical/malignant-appearing spindle cell lesion to be a metaplastic carcinoma, the differential diagnosis also includes melanoma, angiosarcoma, and a stroma of phyllodes tumor. At least two melanoma markers should be performed. S100 is a very sensitive melanoma marker, but has been reported to stain between 20% and 50% of metaplastic breast carcinomas and therefore is not the best stain for this differential diagnosis. Strong keratin reactivity or multiple keratin positivity would also exclude a melanoma. However, CAM5.2 positivity alone is not enough to exclude a melanoma unless it is strong and diffuse. Another significant malignant lesion with which metaplastic carcinoma can be confused is an angiosarcoma. These tumors may occur after radiation treatment or de novo. It is obvious to think about angiosarcoma in a malignant spindle or epithelioid lesion of the breast if there has been a prior history of radiation treatment. However, in absence of such a clinical history, the lesion should be extensively examined by available IHC stains. More than one vascular marker should be used due to the heterogeneous expression of vascular markers. Of the three commonly used vascular markers (CD31, CD34, and Factor VIII), CD31 is generally considered as the most specific vascular endothelial marker, but occasional weak staining of carcinomas has been described. We have also seen staining of carcinoma cells and histiocytes with CD31. It is a diagnostic pitfall, especially in small samples. A diagnosis of a de novo primary angiosarcoma of the breast should be made only if there is unequivocal IHC staining for vascular markers, negative staining for p63 and HMWKs, and appropriate histology of the lesion. Other vascular markers that can be positive in angiosarcoma include D2-40 and FLI-1. In absence of staining for epithelial markers, the possibility of phyllodes stroma should also be considered. The stroma of phyllodes tumor is often (not always) positive for CD34, but is negative for other vascular markers. p63 is often negative in the phyllodes tumor. However, Cimino-Mathews reported p63 and p40 reactivity in 57% and 29% of malignant phyllodes tumor ( n = 14), respectively. In the same publication, focal cytokeratin reactivity was also reported in 21% of malignant phyllodes tumor. In our experience this is uncommon, but we suggest taking overall morphology and results of all IHC markers into account for making a definitive diagnosis. In summary, a malignant spindle cell lesion is a metaplastic carcinoma, unless proven otherwise. A panel comprising multiple keratins, EMA, p63, melanoma, and vascular markers is required in the workup of a malignant spindle cell lesion.
Other Spindle Cell Neoplasms (Myoepithelial and Mesenchymal Tumors)
Tumors of the breast in which there is MEC differentiation predominantly include adenomyoepithelioma, myoepithelioma, and myoepithelial cell carcinoma (MECC). Although the majority of myoepithelial tumors are benign, occasional tumors may exhibit aggressive behavior in the form of carcinoma or myoepithelial carcinoma. The typical immunostaining pattern of the myoepithelial components of these tumors is strong cytoplasmic staining for 34βE12, CK5, and nuclear p63. Tumor cells are typically positive, with S100 protein (90%), and may be positive with muscle markers such as calponin (86%), muscle-specific actin, desmin (14%), and α-SMA (36%). Occasional cells exhibit immunostaining with glial fibrillary acidic protein (GFAP). The presence of smooth muscle markers and immunostaining for GFAP is more in keeping with pure myoepithelial differentiation, as opposed to metaplastic carcinomas (discussed previously), which are largely negative for these markers. Expression of SMA is very nonspecific and is not a definitive marker for muscle differentiation. Metaplastic carcinomas of the breast (carcinosarcoma, spindle cell carcinoma, sarcomatoid carcinoma) have an immunoprofile very similar to myoepithelial differentiation, as they regularly coexpress weak cytoplasmic CAM5.2 for low molecular weight keratins, strong cytoplasmic immunostaining for HMWK 34βE12, CK5/6 or CK5, vimentin, and nuclear immunostaining for p63 (90%). However, GFAP and SMMHC are largely negative. Immunostaining with the muscle markers is most indicative of a pure myoepithelial neoplasm as opposed to a metaplastic carcinoma. The immunoprofile of metaplastic carcinoma is shared to a great degree with myoepithelial neoplasms, with some investigators suggesting that the MEC is the progenitor cell for metaplastic carcinomas. Leibl also demonstrated that the experimental myoepithelial markers CD29 and 14-3-3 sigma stain metaplastic carcinomas, supplying further evidence of the myoepithelial nature to these tumors. The literature suggests that using the term myoepithelial carcinoma versus metaplastic carcinoma is a matter of semantics and may not have any clinical significance.
Myoepithelial tumors need to be separated from the rare primary spindle cell sarcoma of the breast, which may include fibrosarcoma (vimentin positive), leiomyosarcoma and rhabdomyosarcoma (positive with muscle markers), synovial sarcoma (positive with CK7 and CK19), malignant nerve sheath tumors (patchy S100+ and vimentin positive), and the so-called malignant fibrous histiocytomas (vimentin positive). Although each of these tumors may have characteristic light microscopic features, immunostaining patterns may be useful in the diagnostic distinction ( Table 19.4 ). Primary liposarcomas (S100+) of the breast are rare tumors that may arise in a preexisting phyllodes tumor (CD34+ stroma).
When a spindle cell lesion in the breast demonstrates bland cytomorphologic features, then the differential diagnosis includes myofibroblastoma, fibromatosis, fibromatosis-like metaplastic carcinoma, inflammatory myofibroblastic tumor, nodular fasciitis, scar tissue, leiomyoma, and cellular pseudoangiomatoses stromal hyperplasia (PASH). Most of these lesions have characteristic morphology, but IHC stains are frequently used to confirm cellular differentiation. The myofibroblastoma of the breast ( Fig. 19.27 ) is distinguished immunohistochemically from myoepithelial tumors by lack of immunostaining for keratins, S100 protein, and other myoepithelial markers. Myofibroblastomas demonstrate CD34+ cells in at least 50% of the cases. Myofibroblastomas are often positive for HRs and show variable reactivity for actin and desmin. Myofibroblastomas have been previously referred to as solitary fibrous tumor of the breast, but immunohistochemical and genetic studies fail to find the link. Solitary fibrous tumors are often STAT6 positive, and mammary myofibroblastomas have been found to be negative. The genetic studies on myofibroblastomas have shown partial monosomy of 13q and 16q with deletion of the region 13q14 in greater than 50% cases. These findings are similar to 13q and 16q rearrangements in spindle cell lipoma. Moreover, it is not uncommon to find adipose tissue within myofibroblastomas supporting the concept that myofibroblastomas are likely related to spindle cell lipomas. PASH, which could be part of fibroadenoma or phyllodes tumor, is also positive for CD34. In contrast, fibromatosis involving the breast is negative for CD34, but demonstrates abnormal (nuclear) localization of β-catenin. It is to be noted that nuclear β-catenin expression has been documented in phyllodes tumor and metaplastic carcinomas, but expression is diffuse and strong in fibromatosis compared with other diagnoses. Scar tissue can be extremely difficult to distinguish from fibromatosis, but is thought to be negative for nuclear β-catenin expression. Fibromatosis-like metaplastic carcinoma is distinguished from fibromatosis by expression of keratin stains and p63 reactivity in the former. In addition to the characteristic morphology, inflammatory myofibroblastic tumor shows reactivity for actin and ALK protein. Leiomyomas are extremely rare in the breast, but are expected to be positive for smooth muscle markers, actin, desmin, and caldesmon. Patchy actin and desmin reactivity can be seen in many other myofibroblastic lesions, and therefore diagnosis of leiomyoma requires appropriate morphology and uniform strong expression for muscle markers.
Micropapillary carcinomas can be confidently identified by using EMA (or MUC1) that demonstrates “reverse polarity” of the tumor cells.
Basal-like carcinomas are “triple negative” and show immunoreactivity for basal cytokeratins and EGFR.
Basal-cytokeratins (especially CK5), along with p63, are very sensitive markers for identifying spindle cell metaplastic carcinomas.
Although muscle differentiation in myoepithelial carcinomas separates them from spindle cell metaplastic carcinomas, they likely represent two different spectra of one entity.
Adenomyoepitheliomas are biphasic tumors that resemble (and are actually a variant of) intraductal papilloma at one extreme, and show pure spindle cell myoepithelioma at the other extreme.
Noncarcinomatous spindle cell proliferation in a breast core biopsy sample includes both benign (fibromatosis, myofibroblastoma) and malignant lesions (stromal component of malignant phyllodes tumor, melanoma, or primary sarcomas).
EGFR , Epidermal growth factor receptor.
Paget Disease of the Breast
The presence of malignant epithelium (with breast cancer immunophenotype) within the nipple epidermis is termed Paget disease of the breast . It is often the result of intraepidermal spread of malignant cells from underlying ductal carcinoma in situ. The presence of an associated mass lesion indicates the presence of invasive carcinoma in addition to DCIS.
Paget disease of the breast is manifested as CK7+ malignant cells infiltrating the epidermis of the nipple. Tumor cells are conspicuous by their infiltrative “shotgun” pattern, large size, abundant cytoplasm, signet-ring forms, and sometimes mucin positivity ( Fig. 19.28 ).
The majority (>90%) of the underlying breast carcinomas are ductal in nature. In a nipple/areola biopsy, or in cases where underlying carcinoma could not be documented, the differential for Paget disease includes a melanoma and squamous cell carcinoma in situ (Bowen disease). The single best stain for this differential diagnosis is CK7, which is positive in almost all cases of Paget disease (see Fig. 19.28 ). Cells of Paget disease are also positive for HER2 (in ~80%–90% of cases), and this correlates to the IHC expression of underlying breast carcinoma, which is often a HER2-positive high-grade DCIS with apocrine differentiation and comedonecrosis. Additional stains that can be positive in Paget disease are gross cystic disease fluid protein-15 (GCDFP-15+), mammaglobin (MGB), carcinoembryonic antigen (CEA+), and HRs. But it is important to remember that estrogen and progesterone receptors (PR) are not good markers of Paget disease. Although Paget disease is a manifestation of underlying breast carcinoma, most often the carcinoma is a DCIS with comedonecrosis, and these tumors are frequently negative for HRs. GATA3 is also positive in Paget disease, but it also stains background keratinocytes. This limits the usefulness of GATA3 in diagnosing Paget disease. If the possibility of a melanoma is entertained, then at least two melanoma markers should be used because S100 can be positive in about 18% of Paget disease. However, it should be noted that malignant melanoma on the nipple is extraordinarily rare. Pagetoid squamous carcinoma (Bowen disease) of the breast is rare and can be distinguished from Paget disease. Cells of Bowen disease are negative for CK7, and squamous nature of the cells can be confirmed by CK5 or CK5/6 and p63 stains, whereas Paget shows reverse result for these antibodies.
Toker cells are CK7+ and may be present in the skin of the normal nipple, but generally they are inconspicuous compared with Paget cells and are cytologically bland and do not cause diagnostic problems. It has been suggested that Toker cells may be the origin of intraepithelial Paget cells, based on similarity of immunophenotypes. In cases of florid papillomatosis of the nipple, some CK7+ cells may be found in the epidermis, a pitfall to be aware of in diagnosing Paget disease of the nipple. In addition, the intraepidermal portion of nipple ducts can be a pitfall for intraepidermal CK7+ cells.
Pseudo-Paget disease may on occasion be seen in the major ducts. Large histiocytes infiltrate the epithelium and impart a picture simulating Paget disease. These large cells are CK7 negative and strongly positive for CD68 ( Fig. 19.29 ).
Most often positive in Paget disease: CK7, HER2
Other positive but less helpful stains: GCDFP-15, MGB, CEA, ER, PR, GATA3
Pitfall: CK7+ cells in the epidermis in cases of florid nipple duct papillomatosis, Toker cells, or intraepithelial extension of lactiferous duct cells
Pagetoid Bowen disease: CK7 negative, CK5 and p63+
Melanoma: Keratin negative, melanoma markers positive
Detection of Lymphatic Space Invasion
Lymphovascular space invasion in breast carcinoma is an independent predictor of axillary lymph node metastases, which in turn is one of the most important prognostic factors in breast carcinoma. One recent study has shown that peritumoral lymphatic space invasion (and not blood vessel invasion) was determinant of lymph node metastasis. In addition, identification of tumor emboli within dermal lymphatics is also important for correlation purposes in cases of inflammatory carcinomas. However, the pitfalls of interpretation of lymphatic channels in paraffin-embedded breast tissue are well known. Retraction artifacts, ducts with misplaced epithelium, and artifactual displacement of cells commonly complicate the interpretation of biopsy samples. D2-40 shows high sensitivity and specificity for normal lymphatic channels in a variety of tissues. D2-40 stains the lymphatic endothelium crisply and intensely, but does not stain the normal vascular endothelium ( Fig. 19.30 ). It is highly sensitive and specific in identifying lymphatic space invasion. In the breast, D2-40 stains lymphatic channels with a crisp, intense membrane staining of lymphatic endothelium. The D2-40 shows a smudgy immunostaining pattern with MECs and reactive stromal myofibroblasts. It is a pitfall, and this faint to occasionally moderate staining around the periphery of a small duct may be mistaken for lymphatic space invasion; however, it is important to remember that lymphatic vessels are stained very intensely with D2-40 (see Fig. 19.30C ).
Sentinel Lymph Node Examination
Historically, complete axillary lymph node dissection (CALND) had been performed with lumpectomy or mastectomy specimens primarily for staging purposes, providing information that was used to determine adjuvant chemotherapy. The CALND may not change the course of the disease, although with removal of involved axillary nodes, the control of local recurrence in the axilla is easier. The morbidity associated with this procedure is substantial in terms of limitation of arm motion, arm pain, and chronic lymphedema.
The concept of an SLN was spawned by Cabanas in his study of penile carcinoma. The pioneering studies of sentinel lymph node metastasis (SLNM) originated with the study of melanoma patients; the goal was to spare these patients the morbidity of large regional lymph node dissections. Patients with melanomas who had SLN surgery were found to have a relatively orderly progression of lymph node metastases, with the SLN receiving the initial deposits of metastatic cells, followed by metastases in more distal lymph node groups. The same rationale was subsequently used for breast cancer patients. The SLN is identified by injecting a radioisotope and blue dye before planned surgical excision. The SLN, identified by a combination of visual inspection for blue dye and intraoperative scanning for radioactivity, is harvested and submitted for pathologic study. The rationale is that for patients who are SLN-negative, a further morbid procedure of axillary cleanout is unnecessary, but for SLN-positive patients, an axillary dissection is indicated for proper staging and possibly to provide better control for local recurrence. The controversy in this approach arises from several valid questions:
What is the natural history of micrometastatic (MM) disease in the axilla?
Is MM SLN disease an obligate pathway to clinically manifested local recurrence in the axilla?
Is MM SLN disease an indication for adjuvant chemotherapy?
How should the excised SLN be examined pathologically?
Does MM SLN disease affect overall survival?
What are the biologic parameters of MM disease that can predict the behavior of the disease in an individual patient?
Is it possible to recognize “benign transport” of epithelial elements in an SLN?
These are interesting and provocative questions for the care of the breast cancer patient. The American Joint Commission on Cancer defines micrometastasis as a cluster of cells that are no larger than 2 mm. One study with more than 10 years of follow-up concluded that micrometastases are associated with a small but statistically significant decrease in tumor-free survival and overall survival when compared with truly node-negative cases, but they are not an independent prognostic factor. The size of the metastatic deposit, taken together with tumor size and other factors, may additionally stratify patients at risk for further disease.
In most institutions, SLN biopsy with lumpectomy or mastectomy as indicated has become the standard of care. The vast majority of SLN metastases are found in the first three SLNs that are submitted. In addition, with completion and reporting of American College of Surgeons Oncology Group (ACOSOG) Z0011 study data, axillary lymph node staging has further changed. This study randomized patients undergoing lumpectomy for invasive cancer to have either completion axillary dissection or no axillary dissection for up to two positive sentinel nodes. Follow-up on these two randomized groups showed no statistical difference in disease-free and overall survival. The lack of worse prognosis in patients who did not receive complete axillary dissection is likely related to the beneficial effect of radiation therapy after lumpectomy and also due to the fact that in approximately 60% cases of positive sentinel node(s), it is only the sentinel node(s) that is/are positive. In view of the Z0011 data, extensive sampling (i.e., multiple levels of the lymph node) and use of highly sensitive methods for identifying metastatic tumor seem unnecessary.
Sentinel Lymph Node Immunohistochemistry
For the surgical pathologist, the appropriate triage and examination of the SLN is of utmost importance, but even here some controversy exists. When SLN mapping procedure began to be the standard of care a few years ago, the SLNs were histologically examined on multiple levels and cytokeratin stains on at least two levels. Since then, more experience has been gained with the procedure and the reporting of SLNs. It was soon realized that a majority of micrometastases (metastases between 0.2 and 2 mm) can be identified by H&E alone, and IHC for cytokeratin stains generally highlight isolated tumor cells (tumor cell aggregates ≤0.2 mm). Although the exact clinical significance of isolated tumor cells, and even micrometastases, remains uncertain, studies have shown that they both are associated with non-SLN positivity in approximately 10% of cases, especially when the tumor size is larger than 1 cm (pT stage 1C or more). However, completion axillary dissection with 1 to 2 positive sentinel nodes in lumpectomy patients does not translate into improved disease-free or overall survival, as shown in Z0011 trial. If the primary breast carcinoma is of ductal type, it would be difficult (not impossible) to identify isolated tumor cells by H&E stain, and most pathologists would agree that they would be able to identify micrometastases ( Fig. 19.31A ). Therefore cytokeratin stains on SLN do not add any significant information beyond H&E stain in a primary ductal cancer. However, there are significant differences, when the primary breast tumor shows a lobular morphology. Due to single cell infiltration, small (micro) metastases of lobular carcinoma (of the classical type) in a lymph node are extremely difficult to identify (see Fig. 19.31B and C ). Occasionally, cytokeratin stains would identify macrometastases not readily apparent on H&E stain. Cserni and colleagues have reported that sentinel node positivity detected by IHC in lobular carcinomas was associated with further nodal metastases in 12 of 50 (24%) cases. Therefore it is not unreasonable (although not required) to perform cytokeratin stains on SLNs in cases of lobular carcinoma. As far as lymph node examination after neoadjuvant therapy is concerned, routine cytokeratin staining of the lymph node is not required. However, IHC staining can be performed if findings are equivocal or suspicious on H&E staining.
When performing cytokeratin immunostaining of SLNs, one should use a cocktail such as AE1/AE3 cocktail ; CAM5.2 is less desirable because of the manner in which it stains dendritic cells in the lymph node. MM cells occur in small clusters less than 2 mm in diameter within the lymph node or subcapsular sinus, and they need to be distinguished from the dendritic appearance of the interstitial reticulum cells of the lymph node, which are also keratin positive. Studies with larger numbers of patients are needed to discern if the site of lymph node micrometastasis (peripheral sinus vs. parenchyma of lymph node) is clinically significant.
Aggregates of breast epithelial cells in the subcapsular sinus of axillary lymph nodes have been described by Carter and associates as occurring as a result of “mechanical transport” after a breast biopsy. Some impugn the core biopsy itself or the breast massage that follows isotope/dye injection as sources of mechanical displacement of cells into the SLN. Solitary keratin-positive cells may be transported to the SLN, and the histologic feature often associated with true benign transport is the association of CK-positive cells with altered red blood cells and hemosiderin and macrophages ( Fig. 19.32 ). Diaz and colleagues described benign epithelial tissue in skin dermal lymphatics and SLN(s) from patients with pure DCIS. This lends morphologic documentation to the concept of “benign mechanical transport.” The distinction between benign transport and “true” metastasis is easy if the cells in lymph node appear “benign,” but there is no objective way to distinguish benign transport from true metastasis when the cells appear cytologically malignant.
Intraoperative Molecular Testing of Sentinel Lymph Node
In the past few years, a few studies have shown the usefulness of intraoperative molecular tests in determining metastatic disease. These are reverse transcriptase polymerase chain reaction assays (RT-PCR), which use a completely closed system and are fully automated from RNA extraction to final interpretation. One such assay was called the GeneSearch Breast Lymph Node (BLN) Assay (Veridex LLC, Warren, New Jersey), which was approved by the US Food and Drug Administration (FDA) for axillary lymph node testing. The GeneSearch BLN assay was composed of sample preparation kit, all reagents required for performing RT-PCR, and protocol software to be used with the Cepheid SmartCycler System (Sunnyvale, California). According to the company, the test was optimized for detecting metastatic disease larger than 0.2 mm. The test analyzed the expression of CK19 and MGB genes. The studies showed high sensitivity, specificity, positive, and negative predictive values for the test. Overall this molecular assay was very much comparable to the frozen section examination, permanent sections, and even IHC. However, just like any other test, one should be aware of the false-positive and false-negative test results, as well as the usefulness and pitfalls of a particular test. Because molecular tests are not morphologic assays, one has to be extremely careful with any sources of contamination. A cutting bench metastasis (“floater”) can be easily recognized on an H&E-stained slide as such, but will give a false-positive result by RT-PCR, and there will be no definite way to identify this as an error. The SLNs identified in the axillary tail may contain a minute amount of breast tissue in the surrounding adipose tissue, which may also give a false-positive result. Therefore lymph nodes should be completely trimmed of the adipose tissue before sectioned for the molecular analysis. Moreover, fat interferes with the assay itself, and this could be an issue when the SLN is diffusely replaced by adipose tissue. Occasionally, a benign epithelial inclusion (>0.2 mm) within the lymph node could also be a source of a false-positive result ( Fig. 19.33 ). Given the significance of the treatment decision based on a positive SLN result (CALND, which cannot be undone) and several sources of false-positive result with molecular tests, we believe that currently there are insufficient data to replace the morphologic methods with molecular assay. At present, we suggest that a positive molecular result should be confirmed by morphology either by frozen or permanent sections before a final decision is made. In contrast, a negative result is highly valuable given the very high negative predictive value of the molecular tests . Notably, the commercial assay mentioned earlier (GeneSearch BLN assay) was taken off the market due to a variety of reasons, including workflow problems and lack of interest in identifying micrometastasis and isolated tumor cells not seen with routine testing.
Section the lymph node perpendicular to the long axis or longitudinally at 2-mm intervals; examine with H&E (no levels required).
For primary ductal carcinomas, AE1/AE3 keratin stain should be avoided.
AE1/AE3 can be performed for lobular carcinomas, as even large tumor aggregates may be missed on H&E examination alone.
Ninety-seven percent of all SLN metastases will be found in the first three SLNs when multiple SLNs are submitted.
Intraoperative molecular tests are comparable to morphologic examination, but there are potential sources of false-positive results.
Molecular tests to identify micrometastatic tumor in the sentinel lymph node seems unnecessary for individual patient management.
IHC is not routinely required for postneoadjuvant therapy lymph nodes but can be used if suspicious cells are identified.
H&E , Hematoxylin and eosin; IHC , immunohistochemistry; SLN , sentinel lymph node.
Systemic Metastasis of Breast Carcinoma
The diagnosis of breast carcinoma at a metastatic site requires a careful histologic examination, review of all prior case material, and immunohistologic evaluation of tumor cells. If patient had a prior history of breast cancer, it is valuable to know if it showed ductal or lobular morphology. Comparison to prior tumor is helpful in making the correct diagnosis in majority of the cases. Immunohistologic evaluation is mainly required in cases of carcinoma of unknown origin. CK7 and CK20 have been generally used in this evaluation to narrow the differential diagnosis. Breast carcinomas are generally CK7+ and CK20−; however, similar cytokeratin profile is seen in a number of other carcinomas including from the lung and gynecologic tract. However, it is to be noted that not all breast carcinoma will show CK7+/CK20− profile, as approximately 8% of breast cancers are negative for CK7.
Gross cystic disease fluid protein-15 (GCDFP-15) has been used for several years as the most specific marker of breast carcinoma ; however, its sensitivity in formalin-fixed paraffin-embedded (FFPE) tissue is less than optimal. Originally described by Pearlman and colleagues and Haagensen and associates, the prolactin-inducing protein identified by Murphy and coworkers has the same amino acid sequence as GCDFP-15 and is found in abundance in breast cystic fluid and any cell type that has apocrine features. The latter, in addition to breast, includes acinar structures in salivary glands, apocrine glands, and sweat glands, and in Paget disease of skin, vulva, and prostate. Homologous-appearing carcinomas of the breast, skin adnexa, and salivary glands demonstrate a great deal of overlap immunostaining with GCDFP-15. Aside from these immunoreactivities, most other carcinomas show no appreciable immunostaining. Breast carcinoma metastatic to the skin (or locally recurrent) may be difficult to distinguish from skin adnexal tumors. Wick and associates, in a study of the overlapping morphologic features of breast, salivary gland, and skin adnexal tumors, found that GCDFP-15 was infrequently found in eccrine sweat gland carcinomas, a paucity of CEA was found in breast carcinomas, and ERs were largely absent in salivary duct carcinomas. The positive predictive value and specificity for detection of breast carcinoma with GCDFP-15 have been reported up to 99%. The sensitivity for the GCDFP-15 antibodies has been reported to be as high as 75% for tumors with apocrine differentiation, but the overall sensitivity is 55%, and only 23% for tumors without apocrine differentiation. The sensitivity is even worse when it comes to core biopsy, because the pattern of staining for GCDFP-15 is often patchy.
Because the specificity of GCDFP-15 antibodies for breast carcinoma is so high, this antibody is often used in a screening panel in the appropriate clinical situation, which often turns out to be the presentation of a woman with metastasis of unknown primary or a new lung mass in a patient with a history of breast cancer. Others have demonstrated the utility and specificity of GCDFP-15 antibodies in the distinction of breast carcinoma metastatic in the lung. However, Striebel and colleagues demonstrated GCDFP-15 immunoreactivity in 11 of 211 (5.2%) lung adenocarcinomas. This study again stresses the importance of a panel, rather than an individual stain in determining site of origin of a metastatic tumor. On a similar note, WT1 (a specific marker of ovarian serous carcinoma) nuclear expression is seen in a subset of breast carcinoma that demonstrates mucinous differentiation. However, the expression is generally weak to moderate in contrast to ovarian serous carcinoma, where the expression is generally strong and diffuse. ERs and PRs may be helpful in cases with history of receptor positive breast cancer; however, a large proportion of gynecologic tumors are positive for HRs. HRs have also been reported to be positive in nonbreast and nongynecologic sites. MGB has been described to be a more sensitive marker than GCDFP-15 for diagnosis of breast carcinoma. The MGB gene is a member of the uteroglobin family that encodes a glycoprotein that is associated with breast epithelial cells. The immunostaining pattern is cytoplasmic, analogous to GCDFP-15. If the weak equivocal staining is disregarded (as it is not helpful in determining site of origin in “real life”), the sensitivity of MGB is between 50% and 60% compared with less than 30% for GCDFP-15. We have seen that even in cases positive for both GCDFP-15 and MGB, the percentage of cells and intensity of staining is much higher with MGB than with GCDFP-15 ( Fig. 19.34 ). Studies have suggested association of MGB staining with HR positivity, but HR-negative tumors can also be MGB positive. Therefore MGB may be useful in identifying breast tumors negative or low (patchy) positive for receptors. The drawback of using MGB is its lack of specificity. It is noteworthy that MGB stains a substantial number (~40%) of endometrioid carcinomas, and occasional melanomas. With respect to distinction of breast carcinoma from skin adnexal tumor or ductal salivary gland tumor, both MGB and GCDFP-15 are unreliable because these tumors have similar IHC profile. In spite of some nonspecificity, it appears that a combination of GCDFP-15 and MGB is better than GCDFP-15 alone in diagnosis of metastatic breast cancer.
Another positive breast marker that can be used is NYBR1, the reactivity for which correlates with HRs. NYBR1 is generally negative in müllerian tumors and in other CK7+/CK20− tumors that can mimic breast carcinomas.
A useful negative marker in the workup of tumors suspected to be of breast origin is PAX8. This antibody is almost always negative in breast carcinoma, but is positive in a substantial proportion of tumors of müllerian origin. In contrast to WT1, the PAX8 reactivity in müllerian tumors is not type-specific. PAX8 is also positive in thyroid tumors, renal tumors, GI carcinoid tumors, and thymic carcinomas.
Another sensitive marker of breast carcinoma is GATA-binding protein 3 or GATA3. A survey of epithelial tumors suggests diffuse and strong reactivity, mainly in breast and urothelial carcinomas. Weak, patchy reactivity (or less than diffuse-strong) is seen in most squamous cell carcinomas and also in müllerian adenocarcinomas. Within breast cancer, almost all ER+ tumors are also GATA3+ and most show diffuse strong reactivity for GATA3. Approximately 70% of ER negative breast cancers are positive for GATA3; however, reactivity is more variable, with most tumors showing weak to moderate reactivity in varying the number of tumor cells. Although not entirely specific, GATA3 is a sensitive and a very useful marker of breast carcinoma.
In determining or confirming the breast as the site of origin, sometimes a broad range of stains are performed, which results in identification of reactivity with unusual stains. One such example is breast carcinoma reactivity for SOX10. SOX10 is a neural crest transcription factor crucial for specification, maturation, and maintenance of Schwann cells and melanocytes. In the normal breast, SOX10 marks MECs. In breast carcinomas, SOX10 reactivity is mainly seen in triple negative and metaplastic carcinomas (66% positive), suggesting myoepithelial differentiation in these tumors. To avoid problematic and confusing results during evaluation of an undifferentiated tumor, we suggest stepwise use of specific markers based on lesion morphology and clinical context.
Diagnostic confirmation requires use of a panel.
Usual breast carcinoma immunoprofile is CK7+, GATA3+, GCDFP-15+, mammaglobin+, ER+, NYBR1+, CK20 negative, TTF-1 negative, WT1 negative, and PAX8 negative.
GCDFP-15 is the most specific marker of breast carcinoma.
GATA3 is the most sensitive marker of breast carcinoma.
Mammaglobin also stains endometrioid adenocarcinomas (up to 40% cases) and rare melanomas.
Up to 30% of breast carcinomas may be negative for both GCDFP-15 and mammaglobin.
Salivary gland carcinomas and skin adnexal carcinomas have immunoprofile similar to breast carcinomas.
Fibroepithelial tumor is a term used for biphasic tumors that contain both an epithelial and stromal component. Fibroadenomas and phyllodes tumors comprise the majority of fibroepithelial tumors. No immunohistochemical stains are required for the diagnosis of fibroadenoma. However, the intracanalicular variant and cellular subtypes need to be distinguished from benign phyllodes tumor.
Fibroadenoma and Phyllodes Tumor
Phyllodes tumors are biphasic neoplasms that are distinguished from fibroadenomas mainly on morphologic grounds. Unlike fibroadenoma, phyllodes tumors are relatively large, heterogeneous neoplasms, and histologically show a “leaf-like” architecture and periglandular stromal condensation. The variably cellular spindled-cell stroma has been proven by clonal analysis to be the neoplastic compartment. Immunohistochemical reports have shown the majority, regardless of grade, to express CD34 in the stromal compartment, consistent with fibroblastic and/or myofibroblastic differentiation, which is supported by ultrastructural studies. The World Health Organization (WHO) has adopted a three-tiered classification of benign, borderline, and malignant phyllodes tumors. Benign and borderline tumors can recur (especially with positive or close margins); metastases are generally seen only with malignant phyllodes tumor. Benign phyllodes tumors are histologically akin to cellular intracanalicular fibroadenomas but with stromal heterogeneity, leaf-like architecture, periglandular stromal condensation, and/or 1 to 3 mitoses per 10 high-power field (HPF). Due to low proliferation activity (1 to 3 mitotic figures per 10 HPF), the benign phyllodes tumors often need to be distinguished from fibroadenomas. There is a great deal of morphologic overlap between cellular fibroadenomas and benign phyllodes tumor, but unfortunately, immunohistochemical stains are also of limited value in this differential diagnosis. Nevertheless, investigators have tried proliferation markers for this distinction. Jacobs and colleagues found significantly higher stromal proliferation indices, such as Ki-67 (marker of all phases of cell cycle) and topoisomerase II-α (marker of G2-M phase), in phyllodes tumors compared with fibroadenomas on core needle biopsy. In this report, however, Ki-67 index ranged from 0.4% to 4.4% (average 1.6%) in fibroadenomas and from 0% to 18% (average 6%) in benign phyllodes tumors. Thus the margin of error in determination of the proliferation index is relatively small, and given the subjectivity involved in its estimation, may not be entirely reliable to distinguish between the two lesions ( Fig. 19.35 ).
Malignant phyllodes tumors are akin to sarcomas with stromal expansion, pleomorphism, and mitotic rates often greater than 10 per 10 HPF, and borderline tumors lie somewhere in between benign and malignant phyllodes tumor. Numerous immunohistochemical studies have shown correlation of biomarker expression with tumor grade. Studies targeting Ki-67 with the monoclonal antibody MIB-1 have shown progressively increased expression with tumor grade. Ki-67 labeling indices range from 1% to 5% in benign tumors, 6% to 16% in borderline tumors, and 12% to 50% in malignant tumors in published reports ( Fig. 19.36 ). Similarly, tumor suppressor gene p53 is increasingly expressed with tumor grade, though less consistently than Ki-67. More recently, the expression of proteins with targeted therapy implications in phyllodes tumors have been explored. Chen and colleagues first reported c-kit expression in the stroma of phyllodes tumors in 2000, and found c-kit expression to be preferentially expressed in histologically malignant phyllodes tumors. Since then, several additional studies have reported increased c-kit expression in malignant phyllodes tumors compared with benign and/or borderline tumors. However, whether c-kit expression in these tumors infers susceptibility to the KIT-receptor tyrosine kinase inhibitor imatinib mesylate is doubtful, as activating c-kit mutations have yet to be reported. Of interest is the study by Djordjevic and Hanna that suggests c-kit expression in fibroepithelial tumors to be related to the presence of mast cells. The authors have argued against any appreciable true stromal cell c-kit staining in fibroepithelial tumors. EGFR has also been studied in phyllodes tumors, with most reports correlating increased stromal expression with tumor grade as well as chromosome 7 polysomy ( Fig. 19.37 ). Prognostication solely based on histologic categories, however, has proved problematic in some cases, as histologically benign phyllodes tumors may recur locally as higher-grade tumors with subsequent metastases, and many histologically malignant tumors neither recur nor metastasize. The primary aim of most immunohistochemical studies has thus been to correlate biomarker expression with recurrence and/or patient outcome rather than diagnostic category. Unfortunately, most reports have failed to do so and are conflicting. For example, one study reported an inverse relationship between Ki-67 expression and overall survival in multivariate analysis of 117 phyllodes tumors, whereas others have not corroborated this finding. Similarly, although p53 expression appears to correlate with tumor grade as noted previously, it has failed to consistently predict tumor recurrence. The potential prognostic value of EGFR is yet to be determined, given the modest body of literature that exists on the subject to date.
Overall, studies aimed at developing better prognostic markers in phyllodes tumors have suffered from low sample size, a lack of patient follow-up data, and lack of reproducibility. The most significant consistently reported variables in prediction of phyllodes tumor behavior remains histologic characteristics, in particularly the presence of stromal overgrowth, and adequate surgical resection margins. The Singapore General Hospital nomogram for phyllodes tumor is a useful tool for estimating recurrence based on four key morphologic features (AMOS criteria)—namely, stromal atypia (A), stromal mitotic activity (M), stromal overgrowth (O), and surgical margin status (S).
Periductal Stromal Tumor
Periductal stromal tumor was initially described by Burga and Tavassoli as a distinct entity from phyllodes tumors, though histologically identical, except lacking the intracanalicular or “leaf-like” pattern. Like phyllodes tumors, however, the stromal cells express CD34, and thus some have proposed they are best regarded as a phyllodes tumor variant that lacks the classic leaf-like architecture rather than a distinct entity.
Other Fibroepithelial Lesions
The so-called myoid hamartoma can also be considered a fibroepithelial lesion, as it has both stromal and epithelial components. Some authors regard it as a variant of fibroadenoma, or as adenosis tumor with myoid metaplasia, which may be regarded as a variant of mammary hamartoma. These tumors show positive stromal immunoreactivity for SMA, desmin, myosin heavy chain, and vimentin ( Fig. 19.38 ), and are negative for S100.
Phyllodes tumor stroma is CD34+, a finding that is useful in the workup of spindle cell lesion in a core biopsy.
Ki-67 may supplement grading of phyllodes tumor in addition to morphology and counting of mitotic figures.
Ki-67 proliferation index does not reliably distinguish between fibroadenoma and benign phyllodes tumor.
Molecular analyses so far have also been inconclusive in distinguishing fibroadenomas from benign phyllodes tumors.
Periductal stromal tumor is likely a variant of phyllodes tumor and also has CD34+ stroma.
Paraffin-embedded tissue, used for the primary morphologic diagnosis of breast carcinoma, also lends itself to a variety of antibody tests that not only shed a great deal of light on the biology of the disease but also serves as a cutting-edge medium for the development of tests that may have an impact on how the disease is treated.
Pathologic features of breast carcinoma that have prognostic value include:
Lymph node status
Histologic type of tumor
These parameters, which should be included in each pathologist’s surgical pathology report, have been thoroughly studied and found to have significant prognostic value for the patient’s clinical course and response to therapy. Immunohistochemical stains for ERs, PRs, and HER2 are the most common stains performed for prognostic and predictive information. We have come a long way from ligand binding assays (LBAs) to IHC for analyzing ER and PR. ER and PR LBAs have been validated by long-term follow-up for clinical use, with established cutoffs for positive results. Although there are several studies with excellent correlation between the two tests, over the years there have been concerns about the quality, reproducibility, and accuracy of IHC to study markers that predict response to targeted therapy. The preanalytical and analytical factors that affect IHC results should be carefully taken into account while interpreting these results. In this section, we discuss the IHC tests that are performed routinely, in that they may have direct, immediate therapeutic implications.
Needle biopsy of the breast and fine needle aspiration (FNA) cytologic techniques are the most common methods of making the diagnosis of breast carcinoma. All of the diagnostic and prognostic tests (SMMHC, muscle-specific actin, ER, PR, Ki-67, p53, HER2, and so on) can be applied to these small core biopsies and yield reliable results. Of all these tests on core biopsy specimens, only progesterone gives a substantial number of false-negative results because of the greater heterogeneity of immunoperoxidase staining in tissue. However, caution is advised on performing prognostic assays (ER, PR, HER2) on FNA material. Apart from preanalytical factors related to fixation, one cannot be absolutely sure about the presence of invasive carcinoma in the cytology material and may erroneously report ER, PR, and HER2 on in situ carcinoma.
Recognition that estrogen ablation had an impact on groups of patients with breast cancer and that clinical responsiveness correlated with the expression of the estrogen receptor were seminal events in the treatment of patients with breast cancer.
ER and PR bind hormones that exert their effects in the nucleus. Nuclear immunostaining for both receptor proteins can be demonstrated in normal breast acini, which serve as internal controls for the testing procedure. In general, approximately 15% to 20% of the luminal epithelial cells in a duct or lobule stain with ER and PR. However, nuclear staining in normal breast tissue is heterogeneous and may vary with the menstrual cycle. One of the effects of estrogen is to induce the PR, and thus the coordinate expression of both hormone receptors in the same cell reflects the fidelity of the ER/PR axis in the cell. In carcinomas of the breast, most PR+ tumors are also ER+, and ER−, PR+ tumors account for less than 1% of all breast cancers. In general, patients with positive PRs have a significantly longer disease-free survival than patients who are PR−.
Since the early 1990s, the immunohistochemical assay determination of ER/PR levels has replaced the dextran-coated charcoal (DCC) method, also called the LBA. The DCC method, the gold standard for many years, suffered significant drawbacks—namely, (1) tumor sampling error; (2) heavy reliance on obtaining tissue immediately on termination of the blood supply to the tumor, usually in the operating room; (3) normal tissue contamination; and (4) analytic error. Some of the advantages of the IHC method include (1) histologic documentation of the tumor tissue to be assayed; (2) appreciation of the heterogeneity of ERs/PRs in tumor cell nuclei; (3) rapid assessment of the tissue for ERs/PRs by semiquantitative method; (4) rapid turnaround time; (5) lower cost; and (6) the ability to use minute quantities of tissue.
Despite the widespread use of IHC for HR determination, the lack of standardization of preanalytic and analytic variables, scoring schemes, and threshold for determining HR positivity remain a concern.
This mainly relates to tissue fixation—that is, cold ischemic time, time in fixative, and type of fixative used. Cold ischemic time is defined as the time taken to place the specimen in fixative after it has been removed from the patient’s body. The College of American Pathologist (CAP) and American Society of Clinical Oncology (ASCO) recommend that the specimen have cold ischemic time of 1 hour or less if used for receptor testing in breast cancer. However, there are only a few studies that exist on this subject. Neumeister and colleagues found no evidence for loss of antigenicity with time-to-fixation for ER, PR, HER2, or Ki67 in a 4-hour time window. However, with a bootstrapping analysis, the authors did observe a trend toward loss for ER and PR. This particular study was mainly done on tissue specimens represented on tissue microarray using AQUA technology and assessed the effect of ischemia for a relatively short time beyond the 1 hour limit recommended by ASCO. In a study of 97 patients with paired core biopsy and resection specimens, Li and colleagues studied the impact of cold ischemic time (ranging from 64 minutes to 357 minutes) on ER immunohistochemical staining. Although the difference in the percentage of ER staining between core biopsy and resection was not statistically significant, they did identify a trend of decreased ER staining with cold ischemic time of greater than 2 hours. In our own study of cold ischemic time, nonrefrigerated samples were affected more by prolonged cold ischemic time than refrigerated samples. Significant reduction in IHC staining for HRs did not result until 4 hours for refrigerated samples and 2 hours for nonrefrigerated samples. Because most studies with clinical outcome have been performed using FFPE, the current CAP/ASCO recommendation is to use 10% neutral phosphate-buffered formalin (NBF), and the tissue should be fixed for 6 to 72 hours. If an alternative fixative or fixation method is used, it has to be validated with standard fixation before it is implemented in clinical testing. Yildiz-Aktas showed that formalin fixation for up to 96 hours does not affect semiquantitative receptor results. Most laboratories are currently performing HR analysis on core biopsy samples, and there has been some misconception regarding formalin exposure time on smaller samples. Tissue permeation and fixation are not synonymous. Fixation is a chemical reaction that takes time, and therefore both smaller and larger tissues have to be fixed for similar amounts of time. However, larger resection specimens need to be sectioned for efficient permeation of formalin. Impact of formalin fixation time on ER staining has been elegantly shown by Goldstein and colleagues. The authors studied tissue sections from 24 known strongly ER+ tumors that were fixed for 3, 6, 8, and 12 hours, and 1, 2, and 7 days. They used a semiquantitative score (Q score with a range of 0 to 7) to determine ER expression. With 40 minutes of antigen retrieval, the mean Q score at 3 hours was 2.46 and reached a plateau of 6.70 at 8 hours. The mean Q score at 7 days was 6.60. However, with 25 minutes of antigen retrieval, the mean Q score at 3 hours was 1.75, progressively increased with time to 6.62 at 10 hours and then declined with time to 3.79 at 7 days. This classic example shows that optimum formalin exposure time for ER determination is 8 hours and that antigen can be retrieved with increasing retrieval times for over exposed tissue, but an underfixed tissue is completely useless for biomarker study.
Another issue of concern for HRs and other biomarker testing is the use of rapid or alternate processors. It is strongly recommended that breast cancer specimens should be processed in conventional processors. If alternate processors or alternate solutions are being used, the procedures need to be validated for ER results against similar samples that are processed in conventional processors. There also appears to be a need for documenting the type of fixative and the formalin exposure time on pathology reports. Having the fixation time for each breast tissue sample might prove valuable for interpreting and troubleshooting aberrant and/or unexpected ER results. Another important issue is the choice of antibody. Currently the literature is mostly available for three different antibody clones 1D5, 6F11, and SP1 for ER. Most studies suggest a high degree of concordance among these antibodies; however, subtle differences are present depending on the cutoff value for a positive result. All the commercially available antibodies for ER assessment in breast carcinoma target only ER-α isoform. Although another isoform ER-β exists, its role in breast cancer is not well defined. There are currently conflicting data concerning its potential role as a prognostic or predictive factor. Unlike ER-α, ER-β expression is seen in a variety of tissues other than breast and female pelvic organs. In addition, similar to HER2 or any other biomarker assay, controls should ideally be placed on each test slide.
An equally important component for HR assays is the postanalytical (interpretation) portion of the tests; however, scoring schemes and threshold for determining HR positivity remain a concern. A critically important statement from the NIH Consensus Conference of December 2000 suggested that “any nuclear expression of HRs should be regarded as a positive result and render a patient eligible for hormonal therapy.” This statement was supported by the studies performed by Pertschuk and colleagues and Cheang and colleagues, where both documented a greater than 1% of cells cut off for the 1D5 and SP1 ER antibody clones, respectively. The current CAP/ASCO guidelines on HR testing also recommend ≥1% of positive cells as a positive result for ER and PR. ASCO/CAP also recommends that the pathologist quantify the HR test results using one of several methods.
Quantitative results of the IHC method correlate closely with biochemical results and are predictive of prognosis. Veronese and colleagues, in a study using ER1D5, found that immunostaining was predictive of response to tamoxifen in 65 homogeneously treated patients, and was a discriminator for relapse-free and overall survival. Barnes and coworkers and Goulding and colleagues confirmed that the ER by IHC correlated better than the DCC method, and the results were strongly related to patient outcome, regardless of the method of immunoscoring.
The effect of quantitation and establishing that ER is a continuous variable (and not bimodal) was clearly shown by Harvey and colleagues in 1999 using the Allred score ( Fig. 19.39 ) in their series of 1982 primary breast cancer cases. An Allred score of 0, 2, 3, 4, 5, 6, 7, 8 was seen in 26%, 3%, 6%, 10%, 16%, 19%, 16%, and 4% cases, respectively. The authors demonstrated that there was a linear correlation between the Allred score and ER content, as measured by LBA. This study also showed differences in disease-free survival based on the Allred score. Even the current RT-PCR assay for HR has shown a broad dynamic range of HRs that are present in tumor cells. These tests do not demonstrate a bimodal distribution of HR results, as suggested by a few studies. Although the data from the NSABP-B14 clinical trial showed that the level of ER expression (as measured by RT-PCR) has little prognostic significance in the absence of endocrine therapy, the expression level has clinical significance when patients are treated with tamoxifen. When the tamoxifen-treated cohort was examined, patients in the top two tertiles showed benefit, whereas the lower tertile showed the same outcome as the placebo control group. The results suggest that the benefit from tamoxifen was limited to the patients with the higher level of expression. The degree of ER and PR staining has consistently been shown to identify groups of patients with significantly different risks of overall survival, time to recurrence, and treatment response to hormonal therapy. Moreover, a combination of standard histologic and semiquantitative immunohistochemical results for prognostic/predictive markers in breast cancer can help prognosticate individual patient risk and also has the potential to predict for standard chemotherapy benefit. Therefore ER and PR results should be reported using a semiquantitative method. At our institution, we prefer the histochemical score (H-score) method of reporting, as it has a broad dynamic range compared with the Allred method of reporting. H-score is given as the sum of the percent staining multiplied by an ordinal value corresponding to the intensity level (0 = none, 1 = weak, 2 = moderate, 3 = strong). With four intensity levels, the resulting score ranges from 0 (no staining in the tumor) to 300 (diffuse intense staining of the tumor). We also recently showed excellent interobserver concordance for H-score among pathologists. The reason for good concordance is likely the way the H-score is calculated and the slightly increased time spent by the pathologist evaluating the slide to calculate the score ( Fig. 19.40 ). Because the likelihood of benefit correlates with the amount of HR protein in tumor cells, we recommend using a semiquantitative criterion in addition to a positive or negative result that clearly states the percentage and intensity of ER and PR staining within tumor cells.
Routine H-scoring for HRs definitely has advantages but also poses some challenges in terms of pathologist training and performance review. Automated image analysis systems have been suggested as an alternative to human scoring. However, more performance studies are still needed. The image analysis systems that are currently used for analyzing immunohistochemical ER/PR slides are incapable of reporting H-score results on the entire tumor section. Most systems provide quite accurate assessment of percentage positive cells in the field of view, which is generally a very small area of the tumor unless the whole slide is scanned and interpreted. Whole slide scanning takes time, and even if it is performed, the final onus is on the pathologist to make sure the invasive tumor cells are counted by the computer for reporting. Improvement in image analysis technology may overcome some of these aspects. Image analysis systems using fluorescent probes have also been described as a method of identifying immunoreactive tumor cells with an increased interobserver reproducibility. However, the challenge of correctly identifying invasive carcinoma for reporting under dark field microscopy and increased time in doing such analysis can further impede incorporation into diagnostic pathology. Furthermore, there has been an effort to use mRNA expression levels of ER/PR through quantitative reverse transcription polymerase chain reaction as an alternative to IHC. The increased use of onco type DX and MammaPrint (TargetPrint) diagnostic testing, as ordered by clinicians, may increase the popularity of these methods. However, as noted by the ASCO/CAP committee, there is a paucity of data on the concordance between mRNA-based assays and IHC-based clinical validation studies. In regard to this issue, we have reported good agreement between IHC semiquantitative H-score results and Genomic Health onco type DX quantitative ER/PR quantitative reverse transcription polymerase chain reaction results, with IHC being slightly more sensitive than polymerase chain reaction for both ER and PR. There is no reason to replace IHC with polymerase chain reaction, and moreover, due to the speed, higher sensitivity, preservation of morphology, and widespread use in every pathology laboratory, IHC is more desirable.
As far as PR IHC is concerned, Press and colleagues compared 12 different PR antibodies and found that PgR636 and PgR1294 stained the highest percentage of breast carcinomas known to be positive by the biochemical assay (95%–98%), and they exhibited the highest concordance with the biochemical assay (88%–90%). Later, Mohsin and colleagues performed clinical validation of the progesterone receptor by IHC. Using a positive result of 1% of cells with PR clone 1294 (Dakocytomation, Carpinteria, California) and the Allred scoring system, it was demonstrated that PR was better than the LBA to predict disease-free survival and overall survival.
Progesterone nuclear staining by the IHC method is more heterogeneous than ER and may be a cause of false-negative results, especially in core biopsies or needle aspirates. We have seen similar results on some core biopsies; as a result, if a negative ER and PR result on a core biopsy specimen is obtained, the test is repeated on the excisional breast specimen, particularly when the HRs are negative in tumors that are expected to be positive, such as grade 1 and 2 ductal cancers or lobular, tubular, and mucinous carcinomas.
The sensitivity and specificity of other PR antibodies, namely 1A6 and the more recently available rabbit monoclonal antibodies (1E2), are similar to the previously used antibodies. There are two isoforms for PR (PRA and PRB), and both isoforms are recognized by the commercially available antibodies. Our in-house data indicate that there are far fewer ER-negative/PR-positive cases with the 1E2 clone (1% of cases) compared with clone 1A6 (5%–6%).
Although less importance is given to PR compared with ER in the clinical literature, it is increasingly recognized that the amount of PR expression in ER+ tumors provides significant prognostic and predictive information. In a comparative/correlative study by Clark and colleagues, semiquantitative PR expression had a huge impact on the onco type DX recurrence score that was independent of grade. Cheang and colleagues also showed the significance of PR and even proposed to modify the definition of luminal A tumors based on the amount of PR expression. Therefore it is critical that laboratories perform and interpret PR testing in a manner similar to ER testing.
Preanalytic, analytic, and postanalytic factors can significantly alter hormone receptor results.
There are clinically validated studies for ER clones SP1, 1D5, and 6F11, all of which demonstrate a positive cutoff value of greater than 1% of cells staining, linked to response to tamoxifen.
Some form of semiquantitation (H-score, Q score, Allred score) should be performed on all positive cases, as semiquantitation is therapeutically important.
Hormone receptors should be repeated on resection specimen if tumor was double negative on core biopsy sample and the tumor histology suggests a positive result.
The ERBB2 ( HER2 ) gene was originally called NEU , as it was first derived from rat neuro/glioblastoma cell lines. Coussens and colleagues named it HER2 because its primary sequence was very similar to human epidermal growth factor receptor ( EGFR or ERBB or ERBB1 ). Semba and colleagues independently identified an ERBB-related but distinct gene, which they named as ERBB2 . DiFiore and colleagues indicated that both NEU and HER2 were the same as ERBB2 . Akiyama and colleagues precipitated the ERBB2 gene product from adenocarcinoma cells and demonstrated it to be a 185-kD glycoprotein with tyrosine kinase activity. In 1987, 2 years after its discovery, clinical significance of HER2 gene amplification was shown in breast cancer. We now know that approximately 15% to 20% of breast cancers demonstrate HER2 gene amplification and/or protein overexpression. In absence of adjuvant systemic therapy, HER2-positive breast cancer patients have a worse prognosis (i.e., a higher rate of recurrence and mortality), clearly demonstrating its prognostic significance. An even more important aspect of determining HER2 status is its role as a predictive factor. HER2 positivity is predictive of response to anthracycline and taxane-based therapy, whereas the benefits derived from nonanthracyclines and nontaxane therapy may be inferior. It is also important to note that HER2 positive tumors generally show relative resistance to all endocrine therapies; however, this effect may be more toward selective endocrine receptor modulators like tamoxifen and less likely toward estrogen depletion therapies like aromatase inhibitors. Most important, the availability of HER2 targeted therapy brought this biomarker at the forefront of theranostic testing for breast cancer. Trastuzumab is a humanized monoclonal antibody to HER2 that was approved by the US FDA in 1998 for use in metastatic breast cancer. Trastuzumab improves response rates, time to progression, and survival when used alone or in combination with chemotherapy in the treatment of metastatic breast cancer. Although approved for use in metastatic cancer, several prospective randomized clinical trials have shown large therapeutic benefits from trastuzumab in early stage breast cancers. The same paradigm has also shifted to neoadjuvant chemotherapy using trastuzumab in HER2 positive tumors.
Given the enormous therapeutic benefit derived from trastuzumab and other anti-HER2 therapies in HER2 positive tumors, it is absolutely critical that an accurate determination of HER2 status be made on each case. Due to its prognostic and predictive value, HER2 status should be determined on all newly diagnosed invasive breast cancers, which is also recommended in the ASCO/CAP guidelines. These guidelines provide a detailed review of literature and recommendations for optimal HER2 testing. The issues ensuring reliable HER2 testing can be divided in three categories: preanalytical, analytical, and postanalytical. All three issues are equally important, and just like HR testing require a commitment to continuous quality improvement.
The current CAP/ASCO recommendation is to use 10% NBF, and the tissue should be fixed for 6 to 72 hours. If an alternative fixative or fixation method is used, it has to be validated with standard fixation before it is implemented in clinical testing. Although the guidelines stress more regarding overfixation, we believe it is underfixation that seems to be the real problem with HER2 testing. The antigen can be retrieved by various methodologies, and the enzymatic digestion times for in situ hybridization can be altered if the tissue is overfixed, but nothing can be done if the tissue is underfixed. Overfixation may become an issue with alcohol fixation, which can lead to antigen diffusion, but is generally not an issue with formalin fixation. We have validated tissue fixation times up to 96 hours for performing HRs and HER2 testing on breast carcinoma at our institution. The effect of underfixation on biomarker testing has been nicely shown by Goldstein and colleagues, as discussed previously, using ER as an example. In our own study of impact of cold ischemic time on breast cancer prognostic/predictive factors, we reported that significant reduction in immunohistochemical staining generally does not result for up to 4 hours for refrigerated samples and for up to 2 hours for nonrefrigerated samples. In addition, HER2 staining is less commonly impacted compared with HR staining. Recently, Portier and colleagues studied the impact of cold ischemic time on tumors subjected to varying time periods of cold ischemia (45 cases with <1 hour of cold ischemia, 27 cases with 1 to 2 hours of cold ischemia, 6 cases with 2 to 3 hours of cold ischemia, and 6 cases with >3 hours of cold ischemia). The tumors were assessed for HER2 gene and protein via in situ hybridization assay and IHC. They concluded that cold ischemic time of up to 3 hours has no deleterious effect on the detection of HER2 via in situ hybridization or immunohistochemistry. It should also be noted that CAP/ASCO guidelines for fixation times were addressed to resection specimens, but there is no reason to believe that these cannot be applied to needle core biopsies. As a matter of fact, the guidelines should remain the same, irrespective of the size of the specimen. This is because tissue permeation (which is roughly 1 mm/h) is not equal to fixation. It is true that formalin will permeate core biopsy samples faster and make it harder for sectioning, but actual fixation or chemical reaction of aldehyde cross-linking takes time and is independent of specimen size.
This refers to the actual testing protocol, including IHC equipment, reagents, competency of the staff performing IHC, use of appropriate controls, and finally the type of antibody used. The last issue of the type of antibody used deserves special mention. The very first clinical trial assay for assessing the effect of trastuzumab on metastatic breast cancer used CB11 and 4D5 antibodies for determining HER2 status. In these studies, only patients with 2+ or 3+ scores were eligible to receive trastuzumab. Retrospective analyses have revealed therapeutic benefit in cases with either 3+ score or HER2 amplification by FISH. Only 24% of 2+ cases showed amplification by FISH. At the time of FDA approval of trastuzumab, a polyclonal antibody (HercepTest, Dako Corporation) was compared with the clinical trial assay antibody CB11 using the same scoring criteria. HercepTest received FDA approval based on its 79% concordance with CB11. It was soon realized and shown by studies that HercepTest had a slightly higher false-positive rate than other monoclonal antibodies (CB11, TAB250) when compared with FISH. Even to this day, several different antibodies are being used, but all IHC 2+ cases are sent for reflex FISH testing, which in a majority of cases resolves the clinical dilemma about HER2 status. A more reliable rabbit monoclonal antibody 4B5 is now more widely used. In a study by Powell and colleagues, rabbit monoclonal 4B5 demonstrates sharper membrane staining with less cytoplasmic and stromal background staining than CB11. The major advantage of 4B5 was its excellent interlaboratory reproducibility (kappa of 1.0).
This involves interpretation criteria, reporting methods, and quality assurance measures, including competency of the interpreting pathologist. The literature regarding HER2 IHC testing would suggest that 2+ score is the most problematic ; however, we believe that it is the 1+ and 3+ scores which are often misinterpreted that has grave clinical consequences. Almost all laboratories would do FISH for HER2 gene copy number assessment when the IHC score is 2, but would skip HER2 FISH testing for 0, 1+, or 3+ scores. There are ample data that HER2 FISH has great correlation with response to trastuzumab treatment. Therefore, a 2+ HER2 IHC score (seen in approximately 25% cases) coupled with HER2 FISH has no adverse clinical consequences. In contrast, a false-positive HER2 IHC 3+ score would result in inappropriate (ineffective, expensive, and potentially harmful) therapy. Similarly, a false-negative HER2 IHC 1+ score would not be tested by FISH. We urge that utmost care should be exercised in interpreting HER2 IHC results. The best strategy is to keep a low threshold for scoring a case as 2+ and a high threshold for scoring a case as 3+. Some laboratories perform only FISH testing, but there are drawbacks of performing FISH as the only test, because assessment of heterogeneity is better appreciated at low power on bright field microscopy rather than at high power examination that is required for FISH.
Last but not the least, a quality assurance program should be in place for the laboratories that perform HER2 testing. Quality control procedures for HER2 IHC should include the laboratory statistics of percentage of positive cases and the percentage of IHC cases that are amplified by FISH. Periodic laboratory assessment of these correlations is essential for quality reporting. Rigorous adherence to quality, tissue fixation times, control tissues/cell lines, and improved interobserver interpretation agreement or image assisted analysis is preferable. The CAP/ASCO guidelines recommend participation in proficiency testing program specific to each method used.