Context.—Approximately 25% of patients with breast cancer develop cutaneous metastases. Sweat gland carcinomas (SGCs) account for about 0.05% of all cutaneous neoplasms. Cutaneous metastases of breast carcinoma (CMBCs) (especially the ductal type) can be difficult to distinguish from SGCs. Treatment and prognoses for these 2 types of tumors differ radically, making accurate histologic diagnosis crucial. Although a few studies attempt to differentiate these entities employing immunohistochemical (IHC) studies (some of which we review here), to date, no panel of IHC stains exists, to our knowledge, to distinguish these entities.
Objective.—To devise a panel of IHC stains to distinguish CMBC from SGC.
Design.—Twelve cases of ductal CMBCs (11 not otherwise specified type, and 1 basal phenotype), 11 cases of SGCs (5 eccrine carcinomas, 3 porocarcinomas, and 3 microcystic adnexal carcinomas), 2 benign sweat gland neoplasm cases, and 2 primary breast cancer cases were retrieved and analyzed with the following IHC panel: mammaglobin, gross cystic disease fluid protein (GCDFP) 15, p63, basal cytokeratins (CK5, CK14, and CK17), androgen receptor, and PAX5.
Results.—The p63 was only weakly expressed in 1 of 12 CMBC cases (8.3%), whereas it was strongly expressed in 10 of 11 SGC cases (90.9%) (P < .001). Basal cytokeratins demonstrated a similar immunoprofile in the SGC group, with 10 of 11 cases (90.9%) expressing all 3 markers, and a variable immunoprofile in the CMBC group with 0% (CK14) (P < .001) to 16.7% (2 of 12 cases; CK5 and CK17) (P < .001) expression. Mammaglobin was expressed in 8 of 12 cases (66.7%) of CMBC.
Conclusions.—Together, these 5 IHC stains were combined to make a panel that was 100% sensitive and 91% specific in distinguishing between CMBC and SGC.
Approximately 25% of patients with breast cancer may develop cutaneous metastases.1,2 Sweat gland carcinomas (SGCs) account for 0.05% of all cutaneous neoplasms.3 Clinically, the presentation of these 2 entities is often distinct. Ductal cutaneous metastases of breast carcinoma (CMBCs) present as multiple lesions in patients with a previous diagnosis of primary breast carcinoma (PBC), whereas SGC presents as a single cutaneous lesion in patients with no known history of breast cancer. However, CMBCs can be difficult to distinguish from SGCs histologically, and complex clinical circumstances can further complicate a diagnosis. Several recent case reports continue to demonstrate this potential diagnostic pitfall.4,5 We recently received 2 diagnostically challenging cases presenting similar challenges.
A 75-year-old woman with a history of bilateral mastectomies for infiltrating ductal carcinoma of both breasts, colectomy for carcinoma of the rectum, and hysterectomy with bilateral salpingo-oophorectomy for benign disease presented with skin lesions on her back 7 years after her bilateral mastectomies. Although the morphology of the PBC and the skin lesions was similar, immunohistochemical (IHC) studies were performed on all lesions. For the right PBC, the tumor cells demonstrated strong positivity for cytokeratin (CK) 7, estrogen receptor (ER), and ERBB2 (formerly HER2/neu) and focal positivity for progesterone receptor (PR). For the left PBC, the tumor cells were strongly positive for CK7, ER, and PR. Both PBCs were negative for CK903, S100, cross cystic fluid protein–15 (GCDFP-15), and CK20. On hematoxylin-eosin stain, the skin lesion, which demonstrated areas of tight intermingling between carcinoma and a spiradenocylindroma, was more compatible with a malignant neoplasm arising in the background of a benign mixed tumor rather than metastasis to the benign tumor (Figure 1). In the benign component, CK7 was positive only in the internal mature cells and negative in basal cells, ERBB2 demonstrated weak and less than 1+ positivity, and CK903 was strongly positive. CK20, ER, PR, and S100 were all negative in the tumor cells. In the malignant component, CK903 was strongly positive, whereas CK7 and S100 were focally positive. CK20, GCDFP-15, ER, PR, and ERBB2 were all negative. Although these findings supported the diagnosis of a primary eccrine carcinoma (EC) arising in a spiradenocylindroma, the possibility of a CMBC to the benign neoplasm could not be ruled out. Molecular studies performed on the paraffin-embedded tissues showed multiple allelic losses in the PBC and no genetic alterations in the EC. The patient subsequently developed additional metastases to lymph nodes and additional skin sites from the EC. Although a definitive diagnosis was rendered in this case, a simplified IHC panel would have benefited both the pathologist and patient.
An 84-year-old woman with a history of bilateral PBC 20 years earlier presented with a skin lesion of the right breast. The morphology of the previous PBC was not consistent with the new skin lesion, which showed an infiltrating carcinoma (Figure 2, A). Numerous IHC stains were performed with the following results: the tumor cells were positive for p63, AE1/AE3, CK5/6, and epithelial membrane antigen, were focally positive for CK7 (shown in Figures 2, B through F, respectively), and were negative for S100, ER, PR, CK20, GCDFP-15, and mammaglobin. Based on this information alone, it was unclear whether the lesion was a CMBC or an SGC, making the development of a treatment plan difficult.
As the treatment and prognosis for CMBC and SGC differ significantly, developing a panel of IHC markers to differentiate these 2 entities would be of high clinical value. Numerous studies have evaluated these entities either individually6–15 or comparatively using various (IHC) stains,16–23 but there has been no study, to our knowledge, to determine the most useful IHC panel for differentiating CMBC from SGC.
Several IHC stains have shown promise in their ability to differentiate CMBC and SGC. Mammaglobin, a protein frequently overexpressed in breast cancer,11 has not been thoroughly examined in either CMBC or SGC. GCDFP-15, a glycoprotein originally isolated in human breast gross cystic fluid, although present in most primary and metastatic breast cancers, has also been reported to be expressed in a few cases of SGC.16,24 Androgen receptor (AR), although often found in high-grade invasive breast cancer, has also been identified in some SGCs.7–9,13
The basal cytokeratin stains (CK5, CK14, and CK17), recently shown to be expressed in the basal phenotype of breast carcinoma, were not present in most other types of PBC.25,26 These stains have also been shown to be present in a few SGC cases.14,19,20 However, with the exception of CMBC with a basal phenotype, these markers are more likely to be seen in SGC; p63, a homologue of the p53 gene and expressed primarily by basal and myoepithelial cells of the skin, would be more likely to be seen in cases of SGC than in cases of CMBC.10,21,22
In addition, 1 of the authors (M.R.-R.) noticed a strong cytoplasmic and/or membranous staining with PAX5, a B-cell–specific activator protein of the basal layer of healthy epithelium, in an unrelated research study and decided to use the antibody in the current study, hypothesizing that the staining pattern might be similar in SGC. The PAX5 expression has been previously studied in B lymphocytes, in most small cell carcinomas, and in Merkel cell carcinomas.27,28
The goal of this study was to construct a panel of IHC stains that would be highly sensitive and specific in distinguishing these 2 morphologically similar entities.
MATERIALS AND METHODS
A MEDLINE (US National Library of Medicine, Bethesda, Maryland) search was performed from 1995 to 2009 to identify studies similar to our own, comparing primary SGC to CMBC (or in 1 study, to PBC) using IHC stains. We selected a panel of 8 IHC stains to investigate their staining patterns in CMBC and SGC. We did not repeat IHC studies that appeared in previous studies to be less useful.
A retrospective search for CMBC and SGC cases using a Natural Language Search was performed in the University of Pittsburgh Medical Center (Pittsburgh, Pennsylvania) CoPath Plus database (Cerner DHT, Inc, Waltham, Massachusetts) for the 9 years from 1998 to 2007. Based on the results of the database search, 27 cases were retrieved from the University of Pittsburgh Medical Center archives for the study: 12 cases of ductal CMBC (44%), which included 11 cases with morphologies not otherwise specified and 1 case of basal phenotype; 11 cases of SGC (41%), which included 5 cases of EC, 3 cases of porocarcinoma (PC), and 3 cases of microcystic adnexal carcinoma (MAC); and 4 additional, randomly selected cases (15%), which included 2 primary cutaneous adnexal benign neoplasms (a poroma and an apocrine adenoma) and 2 cases of PBC.
The patients in PBC and metastatic breast cancer groups were all women, whereas the SGC group consisted of both men and women. The mean ages for patients with ductal CMBC and SGC were 57 years and 73.6 years, respectively (Table 1). The difference in ages was significantly different. Although all the patients with CMBC (12 of 12; 100%) had documented PBCs, only 1 (case 1, described above) of the 11 patients with SGC (9%) had a history of PBC.
Hematoxylin-eosin slides were reviewed; a representative tumor block was selected; sections were obtained on formalin-fixed, paraffin-embedded tissues; and samples were analyzed with the following IHC stains: mammaglobin, GCDFP-15, p63, CK5, CK14, CK17, AR, and PAX5.
All IHC stain analysis was performed on the Ventana BenchMark, XT (Ventana Medical Systems, Inc, Tucson, Arizona) using the iView DAB Detection System (Ventana Medical) with commercially available antibodies according to standard protocols (Table 2). All negative and positive controls demonstrated appropriate immunolabeling.
The IHC stain results were semiquantitated as follows: AR, PAX5, and p63 were nuclear stains, and strong nuclear positivity was considered positive staining. The intensity of immunostaining was graded as negative (0), weak (1), moderate (2), and strong (3), and the proportion of positive staining cells was recorded as 0% through 5% (1), 6% through 20% (2), 21% through 80% (3), and greater than 80% (4).
Cytoplasmic staining was considered positive for GCDFP-15 and mammaglobin, and the intensity of immunostaining was graded as weak, moderate, or strong.
Basal cytokeratins—CK5, CK14, and CK17—were scored as positive if any cytoplasmic or membranous staining was observed in the tumor cells.
A strong cytoplasmic and/or membranous staining pattern was considered positive for PAX5.
A statistical analysis of the positive and negative results was performed with R statistical software 2.10.1 (R Project for Statistical Computing, Wien, Austria). The SGCs were all compiled into one group for the statistical analysis. Comparisons of the IHC staining profiles in the CMBC and SGC groups were performed using a z test to calculate the P values for each IHC stain. Subsequently, a simple computer algorithm was created to examine the effectiveness of a combination of the near statistically significant and statistically significant biomarkers.
Wallace et al16 investigated the IHC staining characteristics of 15 cases (from 12 patients) of CMBC and compared them to a series of primary eccrine tumors, including 8 MACs. The authors16 concluded that using standard IHC markers, such as ER, PR and GCDFP-15, would not reliably distinguish these 2 populations.
Wick et al17 examined 59 cases of ductal PBCs and compared them with 27 cases of ductal SGC, which were further described as demonstrating eccrine morphology (23 cases; 85%) and apocrine differentiation (4 cases; 15%). Using pancytokeratins, carcinoembryonic antigen, S100 protein, GCDFP-15, ER, PR, and c-erbB-2 protein (ERBB2) IHC stains, their findings17 concluded that the infrequency of GCDFP-15 in eccrine sweat gland tumors as well as the paucity of carcinoembryonic antigen in breast carcinomas could be useful in predefined differential diagnostic settings involving these 2 entities with the appropriate clinicopathologic information provided. We included this study in our review because of its similarity to the current study, even though it focused on distinguishing PBC from SGC, rather than distinguishing CMBC from SGC.
Busam et al18 studied 30 cases of CMBC compared with 42 primary SGC cases for their expression of ER, PR, and epidermal growth factor receptor (EGFR). Several additional histologic types of SGC were examined in this study,18 including apocrine, hidradenocarcinoma, mucinous, and basaloid carcinomas, which were not included in our study. Only the 3 types of tumors included in our study—MAC, PC, and EC—are presented in Table 3. Their results18 suggested that the expression of EGFR may be diagnostically helpful in differentiating these 2 groups of tumors, whereas ER and PR continued to show no significant difference between the 2 groups.
Plumb et al19 used CK5/6 to differentiate primary cutaneous adnexal neoplasms, including 3 MACs, from cutaneous metastatic lesions, including 17 CMBCs. Only 2 of 17 CMBC cases (12%) displayed strong positive staining with CK5/6, with 6 cases (35%) having weak positive staining, whereas all 3 MACs (100%) expressed strong positive staining. This difference, although not evaluated specifically in the study,19 revealed the diagnostic potential of CK5/6 in differentiating these lesions.
Qureshi et al20 also examined the diagnostic potential of several IHC stains, including p63, CK5/6, CK7, and CK20, to differentiate metastatic carcinomas from primary cutaneous adnexal neoplasms. Several benign adnexal neoplasms as well as malignant neoplasms, such as hidradenocarcinoma, adenoid cystic carcinoma, sebaceous carcinoma, digital papillary adenocarcinoma, syringomatous carcinoma, and mucinous carcinoma, were included in their study.20 Only their examinations of PCs and CMBCs were included in Table 3. Three cases of PC were all strongly positive for CK5/6 and p63, whereas CK7 demonstrated positivity in less that 10% of cells for 2 cases, and CK20 was completely negative in all cases. The 6 CMBC cases were diffusely positive for CK7 and were negative for CK5/6, p63, and CK20. This study20 demonstrated the diagnostic potential of CK5/6 and p63 in differentiating these 2 entities.
Ivan et al22 assessed the utility of p63 antibody for differentiating primary cutaneous adnexal neoplasms and adenocarcinoma metastatic to the skin. In addition to several benign adnexal tumors, the authors22 analyzed 4 MACs and 3 ECs (1 of the latter with mucinous differentiation) as well as a case of hidradenocarcinoma and 2 cases of trichilemmal carcinomas. Of the MACs and ECs, only the ECs with mucinous differentiation were negative for nuclear positivity of p63, and none of the CMBCs demonstrated nuclear positivity, confirming the usefulness of this marker in distinguishing SGC from CMBC.
Finally, Liang et al23 investigated the use of podoplanin to differentiate metastases to skin from various organ sites, including the breast, from primary skin adnexal carcinomas. In their study,23 the authors examined 11 cases of metastatic breast cancer to the skin, all of which (100%) were completely negative for podoplanin. They also examined a total of 40 primary skin adnexal carcinomas, only 2 of which (5%; a case of adenoid cystic and a case of poorly differentiated adnexal carcinoma) were negative for podoplanin. However, because the distinction between SGC and metastatic adenocarcinomas may be equivocal in many cases, the only SGCs included in their study were the 6 PCs, which were all (100%) positive for podoplanin. The authors23 suggest that additional studies may be necessary in the future to evaluate more SGCs for podoplanin. From that limited study,23 podoplanin appeared to show promise in distinguishing SGC from CMBC.
In our study, the only 2 SGCs that were positive for AR were 2 of the 3 PCs (67% of the PCs; 18% of all SGCs). Although most adnexal carcinomas were negative for GCDFP-15, only 1 of the 2 ECs (50%), ex spiradenocylindroma,29 was positive. That same carcinoma was positive for mammaglobin as well, while staining negative for all other stains. The second EC ex spiradenocylindroma was positive for mammaglobin, while also staining positive for p63, CK5, CK14, and CK17.
Although most of the CMBC cases in this study were negative for the basal CK markers, 3 of the 12 cases (25%) were positive for either or both CK5 and CK17. One case that was positive for both of these immunostains was morphologically of a basal phenotype.
Table 4 provides a summary of IHC profile staining pattern in CMBC and SGC. The p63 was only weakly expressed in 8.3% (1 of 12) of the CMBC cases, whereas it was strongly expressed in 90.9% (10 of 11) of the SGC cases (P < .001). All 3 basal CKs were expressed in 90.9% (10 of 11) of the SGC cases. In comparison, CMBC cases demonstrated a staining profile of 0% (0 of 12) for CK14 and 16.7% (2 of 12) for CK5 and CK17. One case of CMBC (8.3%) expressed both CK and CK17, whereas 2 more cases of CMBC (16.7%) were positive for either CK5 or CK17. The difference in basal CK staining among the SGC cases and the CMBC cases was statistically significant for all 3 stains (P < .001). Mammaglobin expression was seen in 66.7% (8 of 12) of the CMBC cases, compared with 18.2% (2 of 11) of the SGC group (P = .06), which was marginally statistically significant. The difference in expression for both AR or GCDFP-15 in the 2 groups was not statistically significant.
The appearance of a typical CMBC is demonstrated in Figure 3, A, whereas a typical SGC (a PC) is demonstrated in Figure 3, B. Using these 2 cases as examples, the most prominent staining pattern for each group is represented in Figure 4. The CMBCs were generally positive for mammaglobin and negative for p63, CK5, CK14, and CK17 (Figure 4, A through E, respectively), whereas SGCs (represented in the figure by a PC) were generally negative for mammaglobin and positive for p63, CK5, CK14, and CK17 (Figure 4, F though J, respectively).
None of the CMBC or SGC tumor cells (0%) demonstrated nuclear staining for PAX5 (as shown in the right inset of Figure 5, B). However, 54.5% (5 of 11) of the cases of SGC expressed a distinct cytoplasmic and/or membranous staining pattern (Figure 5, B). Figure 5, A, demonstrates the faint cytoplasmic blush seen in ductal CMBC (left inset), as well as the benign basal layer epithelium showing strong cytoplasmic/membranous staining (right inset).
Only 5 out of the 8 stains (63%) examined by accepted criteria demonstrate statistically significant, or near statistically significant, results. Incorporating those 5 IHC markers into a sum score system, we constructed a panel to predict the disease represented in each case. The conditions set were based on the assumption that breast cancer was usually expected to demonstrate the following IHC staining profile: mammaglobin+, p63−, CK5−, CK14−, CK17−.
For each condition that was not met, 1 point was added. If the score was less than 3 of 5 (0, 1, or 2; <60%), the case was defined as CMBC; if it was greater than or equal to 3 of 5 (3, 4, or 5; ≥60%), it was defined as SGC. Using this sum score system with these conditions, 12 of 12 patients with CMBC (100% sensitivity) were correctly identified as were 10 of 11 patients with SGC (91% specificity) (Table 5).
After reviewing several studies that also attempted to differentiate breast cancer and SGC, we attempted to identify the most specific antibodies to differentiate these 2 neoplasms. The ER, PR, CK7, and CK20 stains were not effective in differentiating these entities. The GCDFP-15, carcinoembryonic antigen, EGFR, CK5/6, podoplanin, and p63 stains all showed potential based on previous studies. We further investigated 3 of these 6 IHC stains (GCDFP-15, CK5/6, and p63). In addition, we incorporated 5 additional IHC stains (CK14, CK17, AR, mammaglobin, and PAX5), which had not been previously employed, to our knowledge, in differentiating these lesions. Our limited panel did not include carcinoembryonic antigen, EGFR, or podoplanin. Carcinoembryonic antigen and EGFR had not shown as much promise in the studies reviewed as other IHC stains we wished to include. The promising utility of podoplanin (published after the completion of our study) was unknown during our investigation.
Our study demonstrated a sustained potential of CK5/6 (or CK5 in our study) and p63 in distinguishing CMBC from SGC. The GCDFP-15 stain did not reveal a statistically significant difference between the staining patterns of CMBC and SGC. Ultimately, combining mammaglobin, p63, CK5, CK14, and CK17, we constructed the IHC panel described above that consistently differentiates CMBC from SGC in our cases.
Numerous difficulties have hindered researchers in the identification of a clinically useful IHC panel to distinguish these entities: the paucity of material, varying morphologic appearances of the entities, and differences in IHC staining techniques across laboratories are only a few. At the outset of this study, the number of cases identified was few and reflected the rarity of these neoplasms. In addition, there was considerable heterogeneity among the groups of tumors examined. Within the classifications of CMBC and SGC, rare subtypes existed that were challenging to evaluate. For example, only one case of basal-phenotype CMBC was included in our study. The IHC staining pattern of that CMBC subtype included known positivity for the basal cytokeratins (CK5, CK14, and CK17).26 Although that one case expressed positivity for both CK5 and CK17, it still fulfilled the criteria of the panel for classification of the neoplasm as a CMBC. Another rare entity with possible confounding IHC staining would be metaplastic breast carcinoma. Although its morphologic characteristics can be quite distinct, it has been reported to be positive for p63.10 The staining pattern of that entity for basal CK has not been extensively evaluated. Both of these subtypes of breast cancer warrant further investigation of their unique IHC staining patterns.
For SGC, 2 cases of EC ex spiradenocylindroma were included. These extremely rare neoplasms demonstrated areas of apocrine differentiation. Undoubtedly, further study of a larger cohort of SGC cases with apocrine differentiation would be desirable. However, of these 2 EC ex spiradenocylindroma, only 1 did not fulfill enough conditions to be defined as an SGC by our immunopanel (see Table 5, SGC-2). The other EC ex spiradenocylindroma (case 1, described above) fulfilled the criteria of the IHC panel and was correctly identified as an SGC (see Table 5, SGC-11). The 5-stain IHC panel would have significantly reduced the diagnostic difficulty initially encountered by the pathologists involved with that case. Although these rare subtypes of CMBC and SGC have slightly different IHC staining patterns compared with most cases within these categories, the IHC panel correctly classified the entities in all but one case.
Applying the conditions of the IHC panel to case 2 (described above) also pointed to a more definitive diagnosis. Mammaglobin was negative, whereas p63 and CK5/6 were positive, which fulfilled 3 of the conditions of the IHC panel and identified the lesion as an SGC. The case pathologist commented that the morphology of this new lesion was not completely consistent with primary breast lesion but agreed that because breast carcinomas may progress into poorly differentiated forms over time, a breast carcinoma could not be completely ruled out. With the addition of the IHC panel to the morphologic examination and clinicopathologic information, the case pathologist was more confident in favoring a diagnosis of SGC. The IHC panel was useful in distinguishing CMBC from SGC in cases with classic morphologies as well as those with unique characteristics that yielded broader differential diagnoses.
In addition to organizing a diagnostically useful IHC panel, we present other interesting findings. The basal-phenotype CMBC cases have the potential to metastasize to the skin, apart from other known metastatic sites, such as brain and bone. Furthermore, the percentage of mammaglobin expression in CMBC appeared similar to its previously reported expression in PBC.11 This finding may indicate preservation of this marker from PBC to the metastases.
Also, the novel interpretation of PAX5 in a cytoplasmic and/or membranous staining pattern may provide a specific marker for tumors of adnexal origin. The known role of PAX5 as a transcription factor was functionally consistent with its commonly described nuclear localization in B lymphocytes. However, the cytoplasmic and/or membranous staining pattern was striking and raises the possibility of alternative functions in cellular pathways. This staining pattern warrants further investigation.
We formed a highly sensitive and specific IHC panel, composed of mammaglobin, p63, and 3 basal cytokeratins, with sufficient power to aid in the differentiation between CMBC and SGC. We recommend the use of this panel to differentiate most cases of these 2 entities in routine clinical practice.
We thank Cary Sipos, HT (ASCP) and Kim McManus, HT (ASCP) for their technical assistance. We would also like to thank Jay S. Raval, MD (Department of Pathology, University of Pittsburgh Medical Center) and Darice Y. Wong, PhD (Department of Bioengineering, University of California, Los Angeles) for thoughtful discussions and critical reviews of earlier versions of this manuscript. Financial support for this article was provided through the University of Pittsburgh Medical Center's Department of Pathology.
From the Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania (Drs Rollins-Raval, Chivukula, Jukic, and Dabbs); and the Department of Biostatistics, University of Pittsburgh (Dr Tseng). Dr Jukic is now with the Division of Dermatopathology, Department of Pathology and Laboratory Medicine, James A. Haley Veterans' Hospital, Tampa, Florida.
The authors have no relevant financial interest in the products or companies described in this article.
Presented in part at the annual meeting of the United States and Canadian Academy of Pathology, Denver, Colorado, March 5, 2008.