Breast Cancer Research and Treatment 21: 201-210, 1992. © I992 Kluwer Academic Publishers. Printed in the Netherlands.
Report
I m m u n o h i s t o c h e m i c a l m a r k e r s in the identification of metastatic breast cancer
Paulo C. Cardoso de Almeida and Cynthia B. Pestana Centro de ReferOncia da Sauda da Mulher, Nutricdd, Alimentafdd e Desenvolvimento Infantil, Division of Pathology, Inst. Adolfo Lutz, Sgo Paulo, Brazil
Key words: breast neoplasms, immunohistochemistry, metastases, monoclonal antibodies, tumor antigens Abstract
A panel of nine monoclonal and polyclonal antibodies were tested regarding specificity for metastatic breast cancer. A hundred metastatic tumors were stained, 50 of breast origin and 50 of other origins. Antibodies used were anti-alpha-lactalbumin, anti-lactoferrin, anti-casein, E29 (Dako-EMA), anti-secretory component, anti-gross cystic disease fluid protein (GCDFP15), BRST1, BRST2, and MC5. Analyses of the results were performed using chi-square and logistic regression. Positivity for MC5, BRST1, BRST2, lactoferrin, EMA, and GCDFP15 was significantly higher in tumors of breast origin than in others (p < 0.05). Analyses of the whole panel indicated that GCDFP15 and MC5 were the best markers for identification of breast cancer metastases. When both were positive (58% of breast origin cases), the predicted probability of breast origin was 98%, compared to only 5% when both were negative. Comparison of anti-GCDFP15 with BRST2, a monoclonal antibody against the same protein, showed a slightly better sensitivity of the former, and a similar degree of specificity for breast tissue. In conclusion, a panel of antibodies can be used to securely differentiate metastatic breast cancer from other cancers in a large number of metastatic tumors of unknown origin.
Introduction
The introduction of immunohistochemistry and, more recently, of monoclonal antibodies, has brought a new tool in determining the origin of metastatic neoplasias, since several tissues seem to have specific antigens that allow their identification even after neoplastic transformation and spreading [1, 2]. Well known examples of tissuespecific antigens are thyroglobulin, prostate specific antigen, and prostatic acid phosphatase, which have been demonstrated to be reliable markers of thyroid and prostate origin, respectively, in metastatic carcinomas [3-5]. Although mammary
tissues are known to produce a great number of different proteins, few of them seem to be specific for breast tissue or neoplasia. Among the antigens used as markers of breast origin are alpha-lactalburain (ALA) [6-11], lactoferrin (LCF) [12-14], casein (CAS) [15-17], human milk fat globule membrane proteins (HMFGMP) [18-23], secretory component (SC) [24-26], and gross cystic disease fluid protein-15 (GCDFP15) [11, 27-29]. However, results using these markers are extremely variable due to different methods and antibodies employed. At least some of these discrepancies may also be related to tissue fixation and processing [5]. ALA, a milk protein reportedly synthesized in breast epi-
Address for offprints: RC. Cardoso de Almeida, R. Conselheiro Brotero, 1505, 01232 Sao Paulo, SP Brazil
202
PC Cardoso de Almeida and CB Pestana
thelial cells, is the most studied mammary antigen, but its value as a specific breast marker is presently controversial. Other monoclonal antibodies, such as MC5, produced against milk fat globule membrane, and BRST1, produced against a specific breast tumor cell line protein, have also been employed, but have seldom been tested in nonmammary tumors [20, 22, 30]. Most of the studies using breast markers were performed in normal or primary neoplastic tissue, and we believe that these results should be critically evaluated when metastatic tumor is concerned. Although there seems to be little variation in the immunophenotype of most of the proteins [31], tumor heterogeneity regarding a number of aspects such as morphology, DNA content, cytogenetic markers, growth rate, drug sensitivity, and also antigenic properties have been reported in neoplastic cellular populations [32, 33], including breast tumors [34, 35], and have been associated with cellular metastatic potential by several authors [36]. These findings suggest that at least some proteins may show phenotypic diversity in the metastatic lesions as compared to the primary tumors [37]. In the present study we have compared the immunohistochemical behavior of breast carcinoma metastases with non-mammary metastases, randomly selected, using a panel of nine antibodies: anti-ALA, anti-CAS, anti-LCF, anti-SC, antiGCDFP15, MC5, BRST1, BRST2, and E29. Statistical analyses were performed in order to evaluate their usefulness in the identification of breast origin in metastatic neoplasms.
Material and methods
Fifty metastatic ductal breast carcinomas were selected from the files of the Division of Pathology of the Silo Paulo University Medical School Hospital. As a control group, the last 50 metastatic tumors from other defined origins registered in these files were also analyzed. All breast tumors were lymph node metastases, except for one case which was an ovarian metastasis. Origin of the non-mammary tumors are presented in Table 1. All the speci-
mens had been fixed in buffered formalin and routinely processed. Immunohistochemical study was performed on 5 g m sections of the formalin-fixed and paraffinembedded tissue using the avidin-biotin-peroxidase method [38]. Polyclonal antisera and dilutions employed were: anti-ALA (1:100), anti-LCF (1:100), anti-SC (1:200), EMA absorbed anti CAS (1:300) [39], and anti-GCDFP15 (1:5000) [27]. All polyclonal antibodies were produced by Dako Corporation (Santa Barbara, CA) except anti-CAS and anti-GCDFP15 (kindly donated by Dr. G. Bussolatti, the latter antibody having been originally produced by Dr. D.E. Haagensen Jr and co-workers, Boston). Monoclonal antibodies employed were: E29 (Dako EMA, 1:50) [21, 40], BRST1 [41] (Cambridge Research Lab., Cambridge, MA; now commercialized by Signet Lab., clone CU-18), BRST2 [42], a monoclonal antibody against GCDFP-15 (Cambridge/Signet clone D6), and MC5 [34], a monoclonal antibody produced against HMFGMP (1:200, kindly donated by Dr. R. Ceriani). Briefly, after deparaffinization of histological sections, endogenous peroxidase was blocked with 0.3% H202 and non-specific binding with 10% nor-
Table 1. Origin and histologic type of the 50 non-mammary metastases
Neoplasia
Number of cases
Gastric adenocarcinoma Colon adenocarcinoma Pancreatic adenocarcinoma Thyroid papillary carcinoma Thyroid follicular carcinoma Lung adenocarcinoma Lung squamous cell carcinoma Larynx squamous cell carcinoma Hepatocarcinoma Bladder transitional cell carcinoma Kidney renal cell carcinoma Tongue squamous cell carcinoma Vulvar squamous cell carcinoma Skin Merckel cell carcinoma Ovarian serous papillary carcinoma Ovarian mucinous cystadenocarcinoma Adrenal cortical carcinoma Esophageal adenocarcinoma
12 11 5 2 2 2 2 2 2 2 1 1 1 l 1 1 1 1
ImmunohistochemicaI markers in breast cancer real goat serum or normal horse serum. Trypsinization improved positivity of control sections only for anti-E29, so that sections tested for this antibody were previously incubated with trypsin 0.25% (type II, Sigma Chemical Company, St. Louis, MO). The slides were incubated overnight with the primary antibody at 4 ° C and then consecutively incubated with biotinylated goat anti-rabbit IgG or horse anti-mouse IgG and the avidin-biotinperoxidase complex (Vector Laboratories, Burlingame, CA) for 30 min each at 37°C. Thorough washing in PBS was carried out between the various stages. The substrate used was 0.05% 3,3"-diaminobenzidine tetrahydrochloride (Sigma). Slides were counterstained with hematoxylin, dehydrated, and mounted in balsam. Positive controls consisted of known positive normal or neoplastic tissue. Negative controls were performed replacing the primary antibody with non-immunized sera. One to five slides were observed for each tumor, depending upon its size and availability of paraffin blocks. Observation of the slides was performed through an optical microscope without previous knowledge of the primary site of the neoplasia. Statistical analyses of the results were performed through a mainframe computer and the software packgage for biology BMDR using chi-square for simple comparison and stepwise linear logistic regression for interaction of all the various immunohistochemical markers.
Table 2. Tumor metastases positive for the nine markers employed
Marker
Mammary origin
Non-mammary origin
ALA LCF CAS E29 MC5 SC GCDFPI5 BRST2 BRST!
14 (28%) 19 (38%) 22 (44%) 49 (98%) 47 (94%) 28 (56%) 31 (62%) 17 (34%) 36 (72%)
15 (30%) 02 (04%) 15 (30%) 40 (80%) 27 (54%) 20 (40%) 01 (02%) 02 (04%) 12 (24%)
203
Results
All the cases that had at least some strongly immunostained neoplastic cells were considered positive. Faint and uniform staining in most of the cells was considered negative. Comparison of the different areas of the tumors was helpful in the interpretation of some otherwise questionable cases. The results are summarized in Table 2.
Alpha-lactalbumin (ALA ) Diffuse cytoplasmic staining in most of the cells in addition to background staining in the surrounding connective tissue, sometimes hampering interpretation, was observed in most of the positive cases. Variations in the methods (longer incubation with normal serum and use of bovine serum albumin) and in the antiserum had little effect in reducing background. Fourteen (28%) cases from the 50 breast metastases and 15 (30%) of the other tumors were positive. There was no statistical difference between the groups (22= 0.194, p> 0.25). Lactoferrin (LCF) Strong cytoplasmic staining was seen in the LCF positive tumors. While 19 (38%) of the 50 breast tumors showed immunoreactive cells, only 2 (4%) of the 50 other metastases were positive. This difference reaches statistical significance (Z2= 15.43, p< 0.0001). Casein (CAS) The positive tumors presented homogeneous cytoplasmic staining. Twenty-two (44%) of the mammary origin specimens were positive, while 15 (30%) from the control group showed immunoreactivity. This difference was not significant 0 ( = 1.54, p< 0.1). HMFGMP (E29 EMA) Cytoplasmic and membrane staining, especially around luminar spaces, was seen in most of the positive specimens. Forty nine cases (98%) from the breast origin group and 40 (80%) of the control group were positive for E29. Although close to the limit, this difference was statistically significant 0~2= 6.54; p< 0.025).
204
PC Cardoso de Almeida and CB Pestana
I m m u n o c y t o c h e m i c a l Markers in Metastatic C a r c i n o m a s percentage of positive cells
100%
l
80% 60%
40% 20%
0% ALA
LCF
[~
CAS
EMA
I
I
gO5
SC
Non breast origin
I
GCDFPI5 BRSTI
~
1
BRST2
Breast origin
Fig. 1. Immunocytochemicalmarkers in metastatic carcinomas.
HMFGMP (MC5) Neoplastic cells presented a membrane staining pattern in the positive cases, especially on the lining of intra or extra cellular lumens. Like E29, MC5 was positive in the majority of the metastases from the breast (94%). However, positivity in the other tumors was much lower, showing immunosmined ceils in only 27 (54%) of the 50 cases. This difference was highly significant 0 ( = 18.76, p< 0.0001).
the two groups was also extremely significant 0~2= 38.65, p< 0.0001). While reactivity for the mammary origin cases was 62% (31 cases), just a few stained cells could be observed in the only positive case (2%) of the control group, which was an hepatocarcinoma.
BRST2
0.1).
Positive neoplasms presented a staining pattern similar to the one observed with GCDFP15. Breast metastases were positive in 17 (34%) of the 50 cases, while only 2 (4%) showed immunoreactivity among the other tumors. The difference was significant (Z2= 12.73, p< 0.0005). The two positive metastases from other sites were the hepatocarcinoma, also positive for GCDFP15, and a papillary thyroid carcinoma, both with rare stained cells.
GCDFP15
BRST1
Great variability in the number of stained cells as well as in the intensity of the staining was seen with this antibody. Difference in the positivity between
Great variability in the cytoplasmic staining intensity of different cells in the same tumor was observed in the BRST1 positive cases. Thirty-six
Secretory component (SC) Positive neoplastic elements presented homogeneous cytoplasmic staining for SC. Twenty-eight (56%) cases from breast metastases and 20 (40%) cases from the other metastases were positive. This difference showed no significance 0~2= 1.76, p
0.2) LCF
MC5
GCDFP15
BRST1
Breast
Other
Observed
Predicted
Standard deviation
origin
origins
p r o p o r t i o n (%)*
probability (%)*
o f pred. prob.
-
-
-
+
0
2
0.00
9.80
8.69
+
-
-
-
0
1
0.00
17.83
18.46 3.58
.
1
20
4.76
4.17
-
. +
.
. -
-
6
16
27.27
27.48
8.75
-
+
-
+
9
9
50.00
48.60
11.19 23.65
+
+
-
-
1
1
50.00
65.38
-
+
+
+
8
1
88.89
97.17
3.00
-
-
+
+
1
0
!00.00
79.77
21.30
-
+
+
-
6
0
100.00
93.22
7.19
+
-
+
+
1
0
100.00
95.16
7.20
+
+
-
+
2
0
100.00
82.49
15.40
+
+
+
+
15
0
100.00
99.42
0.82
* O b s e r v e d p r o p o r t i o n a n d p r e d i c t e d probability o f breast origin
of breast origin in each of the combinations are shown in Tables 3 and 4.
(72%) of the mammary metastases, but only 12 (24%) of the other metastases were positive using this antibody. This difference is highly significant (;~2= 21.19, p< 0.0001).
Discussion
Assembly of results The importance of each antigen in differentiating mammary origin from non-mammary origin can be visualized in Fig. 1. Stepwise logistic regression analysis was performed using two degrees of restriction to select the group of markers which together could more efficiently predict breast origin of the neoplasia. The less restrictive analysis (removal limit p > 0.2) selected the antibodies anti-LCF, anti-GCDFP15, MC5, and BRST1, while a more restrictive one (removal limit p> 0.15) selected only MC5 and antiGCDFP15. The combined results obtained with these markers as well as the estimated probability
Alpha-lactalbumin (ALA ) Our results for ALA are within the broad range of positivity observed by different authors in primary and metastatic breast carcinoma, as well as in nonmammary tumors [6-11, 43-47]. Table 5 shows the variation of ALA positivity from 0 to 81% and illustrates the extremely discrepant conclusions found in the several studies. The most important cause of such a discrepancy seems to be antisera specificity. Lee et al. [10], testing primary and metastatic mammary tumors, could not verify absence or decrease of the staining when anti-ALA antisera were absorbed with purified ALA, sug-
Table 4. S u m m a r y o f statistical results o b t a i n e d t h r o u g h m o r e restrictive ' s t e p w i s e ' logistic r e g r e s s i o n ( r e m o v a l limit p > 0.15) MC5
G C D F P 15
B r e a s t origin
O t h e r origins
Observed proportion
P r e d i c t e d probability
Standard deviation of
(%)*
(%)*
pred. prob.
-
-
1
23
4.17
5.74
+
-
18
26
40.91
40.05
7.33
+
+
29
1
96.67
97.92
2.20
-
+
2
0
100.00
81.13
18.23
* O b s e r v e d p r o p o r t i o n and p r e d i c t e d probability o f breast origin
4.42
206
PC Cardoso de Almeida and CB Pestana
gesting that crossreaction may be occurring. This finding is also supported by Hall et al. [48], who, studying lactating and neoplastic mammary tissue, could detect mRNA for ALA in the former and not in the latter group, but were able to identify the protein in both using anti-ALA antibodies. The wide range of results observed in the literature as well as the difficult interpretation of our results due to strong background staining, indicate that ALA antisera commercially available have a great variability of specificity and sensitivity and should not be used as breast tissue marker. Production of monoclonal antibodies for ALA might clarify these conclusions and bring new results. Lactoferrin (LCF)
Few studies have dealt with the presence of LCF in human neoplasms. It was first described in breast carcinomas by Hurlimann etal. [12], who observed 29.5% positivity using immunofluorescence in paraffin sections. Employing the same methodology, Mason etal. [13] found 8.3% positivity. Similar results (7.1%) were reported by Charpin et al. [14] using immunoperoxidase. There are no studies demonstrating LCF in extramammary or metastatic breast tumors, except for the report of one positive gastric carcinoma. In our study, positivity in the breast metastases group was close to the value found by Hurlimann et al. [12]. Surprisingly, other metastases presented very low positivity,
suggesting that LCF could be an important tool in differentiating the two groups. These conclusions, however, should be confirmed for a greater number and variety of metastatic neoplasms. Casein (CAS)
Data concerning casein synthesis by human breast carcinomas is controversial in the literature. Several studies demonstrated presence of this protein in tissues and breast carcinomas using polyclonal antisera with variable positivity [15-17]. The specificity of the reagent has been questioned since contamination with other proteins such as EMA, ALA, and LCF has been described, probably due to the low immunogenicity of casein [49-50]. One study using two monoclonal antibodies against casein could detect this product in lactating but not in normal or neoplastic breast tissues [51]. These conclusions, however, do not exclude the possibility that mammary neoplasms synthesize modified casein which is not detectable by monoclonal antibodies but is reactive with polyclonal antisera. Other evidence for absence of casein in breast carcinomas is the study of Hall et al. [52], who was unable to demonstrate mRNA for casein in human breast neoplasms. We employed a polyclonal antiserum absorbed with EMA, but not with LCE The strong association between CAS and LCF (p< 0.0001) in our study suggests the possibility of cross-reaction. In any case, the difference in CAS positivity
Table 5. ALA positivity in human neoplasias. Summary of the studies using immunohistochemistry
Author
Method
Results
Bussolati et al. Walker
IF IP
71% breast carcinoma 51% ductal breast carcinoma
Bahu et al. Bailey et al. Clayton et al. Lloyd et al.
IF IP IP IP
Le Doussal et al. Lee et al. Doria et al. De las Mulas Wick et al. Our study
IP IP IP IP IP IP
53% 0% 77% 81% 76% 69% 19%
IF - Immunofluorescence. IP - Immunoperoxidase.
ductal breast carcinoma/63% breast metastases breast carcinoma ductal breast carcinoma/55% breast metastases breast carcinoma/0% other tumors breast carcinoma ductal breast carcinoma/62% breast metastases ovarian tumors
0% breast carcinoma and metastases 50% breast carcinoma/49% other carcinomas 28% breast metastases/30% other metastases
Immunohistochemical markers in breast cancer
between the mammary and non-mammary metastases groups was not significant, indicating that this antibody is not useful to identify metastatic breast cancer. HMFGMP
Studies with the monoclonal antibody E29 (Dako EMA) show a high frequency of positivity in all epithelial tissues, including neoplasms of various origins and different histological types [18, 19]. In the most extensive study using this antibody, Pinkus et al. [21] have reported 100% positivity for breast carcinomas, but 88% for adenocarcinomas from other sites. There are no reports of E29 behavior in metastatic tumors. In our study, positivity for breast origin neoplasms was 98%. High positivity (80%) was also found in the control group, but the difference between the two groups reaches significance, although marginally (p< 0.025). Therefore, E29 has limited usefulness in differentiating mammary from nonmammary metastases. If negative, it practically excludes the possibility of breast origin, but if positive, it only demonstrates the epithelial nature of the lesion. MC5, another monoclonal antibody prepared against HMGFMP, is not commercially available and is less extensively studied. Although Ceriani et al. [20] reported 98% and Russo [22] 100% positivity in breast carcinomas, no systematic analysis has been performed in other tumors or metastases. In our cases, MC5 positivity for breast metastases (94%) was similar to the results cited above. Other metastases showed lower immunoreactivity (54%) compared to the first group. This difference is statistically significant (p< 0.0001). Positivity in the control group, nevertheless, is too high to use MC5 alone as a definitive breast tissue marker. Secretory component (SC)
The first important study demonstrating SC in human breast neoplasms was performed by Harris et al. [24] using immunofluorescence. They found positivity for all breast carcinomas and metastatic adenocarcinomas, as well as for colonic, ovarian, and renal neoplasms. On the other hand, Walker [25] observed only 30% of positivity for breast carcinomas. Her explanation for the discrepancy be-
207
tween the results was the method of antibody preparation used by Harris et al. [24], which allowed contamination. Brooks et al. [26], using the immunoperoxidase technique, found 50% positivity for breast carcinomas and 75% for other epithelial tumors. Our results for breast metastatic tumors (56%) are in agreement with those of Brooks et al. [26], although immunoreactivity for other metastases was lower in our analysis. This difference was not significant, suggesting that SC is not useful in the identification of breast origin due to its low specificity and sensitivity. GCDFP15
Since its identification in 1976, several studies reporting GCDFP15 in human mammary carcinomas have been published. Mazoujian etal. [27] using anti-GCDFPI5 found 90% positivity for breast carcinomas and 75% for non-mammary tumors with apocrine features. Other tissues and neoplasms were negative. Their results were further confirmed by Eusebi et al. [28, 29]. Employing the same monoclonal antibody against GCDFP15 that we used (BRST2), Wick et al. [11] observed 72% positivity for breast carcinomas, including primary and metastatic tumors, and only 4% for other tumors. The studies above suggest high specificity of GCDFP15 for mammary origin neoplasia, in accordance with the present study where very few non-mammary metastases (2-4%) showed immunoreactivity. Positivity for breast tumors varied according to the antibody employed, with higher reactivity using the polyclonal (62%) than the monoclonal antibody BRST2 (34%). The difference in sensitivity may be due to the greater number of antigenic determinants detected by the polyclonal antiserum, allowing detection of modified proteins possibly synthesized by the neoplasms. GCDFP15 is the most reliable single marker of neoplasia of breast origin in this study. BRST1
The only immunohistochemical study using this antibody was performed by Mesa-Tejada et al. [30], who found 93% positivity for human breast carcinomas against 35% positivity for other neoplasms. In the present study, positivity for breast
208
PC Cardoso de Almeida and CB Pestana
origin tumors was lower (72%), but there was still a significant difference between the immunoreactivity of the mammary and non-mammary group (p< 0.0001). A possible explanation for this difference is that Mesa-Tejada et al. [30] tested primary neoplasms, while we analyzed metastatic tumors. Tissue fixation and processing may have also contributed to this difference. Summary
The linear logistic regression analyses using two degrees of restriction indicates that a panel of two or more antibodies increases the accuracy of the immunohistochemical method in the identification of breast origin carcinomas. Table 3 indicates that positivity for LCF, MC5, GCDFP15, and BRST1 corresponds to a 99.42% predicted probability of breast origin, compared to only 4.17% when all markers were negative. The more restrictive analysis (Table 4) (removal limit p> 0.15) showed that MC5 and GCDFP15 were the best markers for identification of breast metastases. When both were positive (58% of breast origin cases), the predicted probability of breast origin was 98%, compared to only 5% when both were negative. In this study, 54% of the specimens are part of one of these two groups. Therefore, in more than half of the cases, it is possible to predict with accuracy whether the metastatic neoplasia has originated from the breast. In conclusion, we suggest that a panel of antibodies can be used to securely differentiate metastatic breast cancer from other cancers in a large number of metastatic tumors of unknown origin. Anti-ALA, often employed as a breast specific marker, is unsuitable for this purpose. Anti-LCF, MC5, anti-GCDFP15, BRST1, and BRST2 are more often positive in mammary origin than in non-mammary origin carcinomas. MC5 and antiGCDFP15 used together are the most efficient for this identification. The most specific breast origin marker is anti-GCDFP15. Polyclonal antibody against this protein showed a slightly better sensitivity than the monoclonal BRST2 and a similar degree of specificity for breast origin tissue.
Acknowledgements We thank Prof. Dr. Thales de Brito for his suggestions and comments;Dr. G. Bussolati for his generous gift of antibodies anti-casein and antiGCDFP15, and for valuable advice during this study; and to Dr. R. Ceriani for kindly providing the MC5 monoclonal antibody. This study was supported by Fundag~o de Amparo a Pesquisa do Estado de Sad Paulo (FAPESP #87/0369-8).
References 1. DeLellis RA, Sternberger LA, Mann RB, Banks PM, Nakane PK: Immunoperoxidase techniques in diagnostic pathology. Am J Clin Pathol 71: 483-488, 1979 2. Krhler G, Milstein C: Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256: 495-496, 1975 3. Nadji M, Tabei SZ, Castro A, Ming Chu T, Murphy GR Wang MC, Morales AR: Prostatic specific antigen. An immunohistologic marker for prostatic neoplasms. Cancer 48: 122%1232, 1981 4. Albores-Saavedra J, Nadji M, Civantos E Morales AR: Thyroglobulin in carcinoma of the thyroid: an immunohistochemical study. Human Pathol 14: 62-66, 1983 5. Nadji M, Morales ZR: Immunoperoxidase Techniques: A Practical Approach to Tumor Diagnosis. American Society of Clinical Pathology Press, Chicago, pp. 50-51 and 64-65, I986 6. Bussotati G, Ghiringello B, DiCarlo F, Aimone V, Voglino GF: Lactalbumin synthesis in breast cancer tissue. Lancet ii: 352, 1977 7. Walker RA: The demonstration of alpha lactalbumin in human breast carcinomas. J Pathol 129: 37-42, 1979 8. Bahu 1LM,Mangkomkanok-Mark M, Albertson D, Fors E, Molteni A, Battifora H: Detection of alpha-lactalbumin in breast lesions and relationship to estrogen receptors and serum prolactin. Cancer 46: 1775-1780, 1980 9. Clayton F, Ord6fiez NG, Hanssen GN, Hanssen H: Immunoperoxidase localization of lactalbumin in malignant breast neoplasms. Arch Pathol Lab Med 106: 268-270, 1982 10. Lee AK, DeLeUis RA, Rose PP, Herbert-Stanton T, Tallberg K, Garcia C, Wolfe HJ: Alpha-lactalbumin as an immunohistochemical marker for metastatic breast carcinomas. Am J Surg Pathol 8: 93-100, 1984 11. Wick MR, Lillemoe TJ, Copland GT, Swanson PE, Manivel JC, Kiang DT: Gross cystic disease fluid protein-15 as a marker for breast cancer: immunohistochemical analysis of 690 human neoplasms and comparison with alpha-lactalbumin. Hum Pathot 20: 281-287, 1989 12. Hurlimann J, Lichaa M, Ozzello L: In vitro synthesis of immu-
Immunohistochemical markers in breast cancer
13. 14.
15.
16.
17.
18.
19.
20.
21.
22. 23.
24. 25.
26.
27.
28.
29.
noglobulins and other proteins by dysplastic and neoplastic human mammary tissues. Cancer Res 36: 1284-1292, 1976 Mason DY, Taylor CR: Distribution of tranferrin, ferritin, and lactoferrin in human tissues. J Ctin Pathol 31: 316-327, 1978 Charpin C, Lachard A, Pourrean-Schneider N, Jacquemier J, Lavaut MN, Andonian C, Martin PM, Toga M: Localization of lactoferrin and nonspecific cross-reacting antigen in human breast carcinomas. An immunohistochemical study using the avidinbiotin-peroxidase complex method. Cancer 55: 2612-2617, 1985 Bussolati G, Pich A, Alfani V: Immunofluorescence detection of casein in human mammary dysplastic and neoplastic tissues. Virchows Arch A (Pathol Anat Histol) 365: 15-21, 1975 Pich A, Bussolati G, Carbonara A: Immunocytochemical detection of casein and casein-like proteins in human tissues. J Histochem Cytochem 24: 940-947, 1976 Fortt RW, Gibbs AR, Williams D, Hansen J, Williams I: The identification of 'casein' in human breast cancer. Histopathology 3: 395-406, 1979 Heyderman E, Steele K, Ormerod MG: A new antigen on the epithelial membrane: Its immunoperoxidase location in normal and neoplastic tissue. J Clin Pathol 32: 35-39, 1989 Gatter KC, Alcock C, Heryet A, Pulford KA, Heyderman E, Taylor-Papadimitriou J, Stein H, Mason D: The differential diagnosis of routinely processed anaptastic tumors using monoclonal antibodies. Am J Clin Pathol 82: 33-43, 1984 Ceriani RL, Hill DL, Osvaldo L, Kandell C, Blank EW: Immunohistochemical studies in breast cancer using monoclonal antibodies against breast epithelial cell components and with lectins. In: Russo J (ed) Immunocytochemistry in Tumor Diagnosis. Martinus Nijhoff, Boston, 1985, pp. 233-263 Pinkus GS, Kurtin PJ: Epithelial membrane antigen - A diagnostic discriminant in surgical pathology. Hum Pathol 16: 929940, 1985 Russo J: Immunocytochemistry in Tumor Diagnosis. Martinus Nijhoff, Boston, 1985, pp. 207-232 Russo J, Russo IH: Immunocytochemical markers in breast cancer. In: DeLellis R (ed) Advances in Immunohistochemistry. Raven Press, New York, 1988, pp. 431-475 Harris JP, Caleb MH, South MA: Secretory component in human mammary cancer. Cancer Res 35: 1861-1864, 1975 Walker RA: Differentiation of human breast carcinomas: An immunohistological study of appropriate and inappropriate protein production. J Patho1135: 87-95, 1981 Brooks JJ, Ernst CS: Immunoreactive secretory component of IgA in human tissues and tumors. Am J Clin Patho182: 660-665, 1984 Mazoujian G, Pinkus GS, Davis S, Haagensen DE Jr: Immunohistochemistry of a gross cystic disease fluid protein (GCDFP-15) of the breast. A marker of apocrine epithelium and breast carcinomas with apocrine features. Am J Pathol I10: 105112, 1983 Eusebi V, Betts CM, Haagensen DE, Gugliotta P, Bussolati G, Azzopardi JG: Apocrine differentiation in lobular carcinoma of the breast: a morphologic, immunologic and ultrastructural study. Hum Pathot 15: 134-140, 1984 Eusebi V, Millis RR, Cattani MG, Bussolati G, Azzopardi JG:
30.
31.
32. 33.
34.
35.
36. 37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
209
Apocrine carcinoma of the breast. A morphologic and immunocytocbemical study. Am J Patho1123: 532-541, 1986 Mesa-Tejada R, Keydar I, Ramanarayan M, Ohno T, Fenoglio C, Spiegelman S: Detection in human breast carcinomas of an antigen immunologically related to group-specific antigen of mouse mammary tumor virus. Proc Natl Acad Sci USA 75: 1529-1538, 1978 Esteban JM, Battifora H: Tumor immunophenotype: comparison between primary neoplasm and its metastases. Mod Pathol 3: 192-197, 1990 Kerbel RS: Implications of immunological heterogeneity of tumors. Nature 280: 358-360, 1979 Hart IR, Fidler IJ: The implications of tumor heterogeneity for studies on the biology and therapy of cancer metastasis. Biochem Biophys Acta 651: 37-50, 1981 Ceriani RL, Peterson JA, Blank E: Variability in surface antigen expression of human breast epithelial cells cultured from normal breast, normal tissues peripheral to breast carcinomas, and breast carcinomas. Cancer Res 44: 3033-3039, 1984 Kim U: Pathogenesis and characteristics of spontaneously metastasizing mammary carcinomas and the general principle of metastasis. J Surg Oncol 33: 151-165, 1986 Fidler IJ: Review: Biologic heterogeneity of cancer metastases. Breast Cancer Res Treat 9: 17-35, 1987 Arklie J, Taylor-Papadimitriou J, Bodmer W, Egan M, Mitlis R: Differentiation antigens expressed in the lactating breast are also detected in breast cancers. Int J Cancer 28: 23-29, 1981 Hsu SM, Raine L, Fanger H: Use of avidin biotin peroxidase complex (ABC) in immunoperoxidase techniques. A comparison between ABC and unlabeled antibody (PAP) procedures. J Histochem Cytochem 29: 577-580, 1981 Ormerod MG, Bussolati G, Sloane JP, Steele K, Gugliotta P: Similarities of antisera to casein and epithelial membrane antigen. Virchows Arch (Pathol Anat) 397: 327-333, 1982 Cordell J, Richardson TC, Pulford KAF, Ghosh AK, Gatter KC, Heyderman E, Mason DY: Production of monoclonaI antibodies against human epithelial membrane antigen for use in diagnostic immunocytochemistry. Br J Cancer 52: 347-353, 1985 Mesa-Tejada R, Palakodety RB, Leon JA, Khatcherian AS, Greaton C J: Immunocytochemical distribution of a breast carcinoma associated glycoprotein identified by monoclonal antibodies. Am J Patho1130: 305-314, 1988 Mazoujian G, Parish T, Haagensen D: Immunoperoxidase localization of GCDFP-15 with mouse monoclonal antibodies versus rabbit antiserum. J Histochem Cytochem 36: 377-382, 1988 Bailey AJ, Sloane JP, Trickey BS, Ormerod MG: An immunocytochemical study of alpha lactalbumin in human breast tissue. J Patho1137: 13-23, 1982 Le Doussal V, Zangerle PF, Collette J, Spyratos F, Hacene K, Briere M, Gest J, Franchimont P: Immunohistochemical detection of alpha-Iactalbumin in breast lesions. Eur J Cancer Clin Oncol 20: 1069-1078, 1984 Lloyd RV, Foley J, Judd WJ: Peanut lectin agglutinin and alphalactalbumin. Binding and immunohistochemical localization in breast tissues. Arch Pathol Lab Med 108: 392-395, 1984 De las Mulas J, Morales C, Lopez A, Toro M: Value of alpha-
210
47.
48.
49.
50.
PC Cardoso de Almeida and CB Pestana lactalbumin as tissue marker for breast carcinomas. Breast Diseases-Senologia 2: 53-58, 1987 Doria MI Jr, Adamec T, Talerman A: Alpha-lactalbumin in 'common' epithelial tumors of the ovary. An immunohistochemical study. Am J Clin Pathol 87: 752-756, 1987 Hall L, Craig RK, Davies MS, Ralphs DNL, Campbell PN: AIpha-lactalbumin is not a marker of human hormone-dependent breast cancer. Nature 290: 602-604, 1981 Weir HM, MacKinlay AG, Nash AR, Sarfaty GA: Human kappacasein as a tumor marker: purification and properties. Cancer Detect Prevent 4: 193-204, 1981 Burchell J, Bartek J, Taylor-Papadimitriou J: Production and
characterization of monoclonal antibodies to human casein. A monoclonal antibody that cross-reacts with casein and alpha-lactalbumin. Hybridoma 4: 341-350, 1985 51. Earl HM, McIlbinney RA, Wilson P, Gusterson BA, Coombes RC: Immunohistochemical study of beta- and kappa-caseins in the human breast and breast carcinomas using monoclonal antibodies. Cancer Res 49: 6070-6076, 1989 52. Hall L, Craig RK, Campbell PB: m-RNA species directing synthesis of milk proteins in normal and tumor tissue of human mammary gland. Nature 277: 54-56, 1979
Report
I m m u n o h i s t o c h e m i c a l m a r k e r s in the identification of metastatic breast cancer
Paulo C. Cardoso de Almeida and Cynthia B. Pestana Centro de ReferOncia da Sauda da Mulher, Nutricdd, Alimentafdd e Desenvolvimento Infantil, Division of Pathology, Inst. Adolfo Lutz, Sgo Paulo, Brazil
Key words: breast neoplasms, immunohistochemistry, metastases, monoclonal antibodies, tumor antigens Abstract
A panel of nine monoclonal and polyclonal antibodies were tested regarding specificity for metastatic breast cancer. A hundred metastatic tumors were stained, 50 of breast origin and 50 of other origins. Antibodies used were anti-alpha-lactalbumin, anti-lactoferrin, anti-casein, E29 (Dako-EMA), anti-secretory component, anti-gross cystic disease fluid protein (GCDFP15), BRST1, BRST2, and MC5. Analyses of the results were performed using chi-square and logistic regression. Positivity for MC5, BRST1, BRST2, lactoferrin, EMA, and GCDFP15 was significantly higher in tumors of breast origin than in others (p < 0.05). Analyses of the whole panel indicated that GCDFP15 and MC5 were the best markers for identification of breast cancer metastases. When both were positive (58% of breast origin cases), the predicted probability of breast origin was 98%, compared to only 5% when both were negative. Comparison of anti-GCDFP15 with BRST2, a monoclonal antibody against the same protein, showed a slightly better sensitivity of the former, and a similar degree of specificity for breast tissue. In conclusion, a panel of antibodies can be used to securely differentiate metastatic breast cancer from other cancers in a large number of metastatic tumors of unknown origin.
Introduction
The introduction of immunohistochemistry and, more recently, of monoclonal antibodies, has brought a new tool in determining the origin of metastatic neoplasias, since several tissues seem to have specific antigens that allow their identification even after neoplastic transformation and spreading [1, 2]. Well known examples of tissuespecific antigens are thyroglobulin, prostate specific antigen, and prostatic acid phosphatase, which have been demonstrated to be reliable markers of thyroid and prostate origin, respectively, in metastatic carcinomas [3-5]. Although mammary
tissues are known to produce a great number of different proteins, few of them seem to be specific for breast tissue or neoplasia. Among the antigens used as markers of breast origin are alpha-lactalburain (ALA) [6-11], lactoferrin (LCF) [12-14], casein (CAS) [15-17], human milk fat globule membrane proteins (HMFGMP) [18-23], secretory component (SC) [24-26], and gross cystic disease fluid protein-15 (GCDFP15) [11, 27-29]. However, results using these markers are extremely variable due to different methods and antibodies employed. At least some of these discrepancies may also be related to tissue fixation and processing [5]. ALA, a milk protein reportedly synthesized in breast epi-
Address for offprints: RC. Cardoso de Almeida, R. Conselheiro Brotero, 1505, 01232 Sao Paulo, SP Brazil
202
PC Cardoso de Almeida and CB Pestana
thelial cells, is the most studied mammary antigen, but its value as a specific breast marker is presently controversial. Other monoclonal antibodies, such as MC5, produced against milk fat globule membrane, and BRST1, produced against a specific breast tumor cell line protein, have also been employed, but have seldom been tested in nonmammary tumors [20, 22, 30]. Most of the studies using breast markers were performed in normal or primary neoplastic tissue, and we believe that these results should be critically evaluated when metastatic tumor is concerned. Although there seems to be little variation in the immunophenotype of most of the proteins [31], tumor heterogeneity regarding a number of aspects such as morphology, DNA content, cytogenetic markers, growth rate, drug sensitivity, and also antigenic properties have been reported in neoplastic cellular populations [32, 33], including breast tumors [34, 35], and have been associated with cellular metastatic potential by several authors [36]. These findings suggest that at least some proteins may show phenotypic diversity in the metastatic lesions as compared to the primary tumors [37]. In the present study we have compared the immunohistochemical behavior of breast carcinoma metastases with non-mammary metastases, randomly selected, using a panel of nine antibodies: anti-ALA, anti-CAS, anti-LCF, anti-SC, antiGCDFP15, MC5, BRST1, BRST2, and E29. Statistical analyses were performed in order to evaluate their usefulness in the identification of breast origin in metastatic neoplasms.
Material and methods
Fifty metastatic ductal breast carcinomas were selected from the files of the Division of Pathology of the Silo Paulo University Medical School Hospital. As a control group, the last 50 metastatic tumors from other defined origins registered in these files were also analyzed. All breast tumors were lymph node metastases, except for one case which was an ovarian metastasis. Origin of the non-mammary tumors are presented in Table 1. All the speci-
mens had been fixed in buffered formalin and routinely processed. Immunohistochemical study was performed on 5 g m sections of the formalin-fixed and paraffinembedded tissue using the avidin-biotin-peroxidase method [38]. Polyclonal antisera and dilutions employed were: anti-ALA (1:100), anti-LCF (1:100), anti-SC (1:200), EMA absorbed anti CAS (1:300) [39], and anti-GCDFP15 (1:5000) [27]. All polyclonal antibodies were produced by Dako Corporation (Santa Barbara, CA) except anti-CAS and anti-GCDFP15 (kindly donated by Dr. G. Bussolatti, the latter antibody having been originally produced by Dr. D.E. Haagensen Jr and co-workers, Boston). Monoclonal antibodies employed were: E29 (Dako EMA, 1:50) [21, 40], BRST1 [41] (Cambridge Research Lab., Cambridge, MA; now commercialized by Signet Lab., clone CU-18), BRST2 [42], a monoclonal antibody against GCDFP-15 (Cambridge/Signet clone D6), and MC5 [34], a monoclonal antibody produced against HMFGMP (1:200, kindly donated by Dr. R. Ceriani). Briefly, after deparaffinization of histological sections, endogenous peroxidase was blocked with 0.3% H202 and non-specific binding with 10% nor-
Table 1. Origin and histologic type of the 50 non-mammary metastases
Neoplasia
Number of cases
Gastric adenocarcinoma Colon adenocarcinoma Pancreatic adenocarcinoma Thyroid papillary carcinoma Thyroid follicular carcinoma Lung adenocarcinoma Lung squamous cell carcinoma Larynx squamous cell carcinoma Hepatocarcinoma Bladder transitional cell carcinoma Kidney renal cell carcinoma Tongue squamous cell carcinoma Vulvar squamous cell carcinoma Skin Merckel cell carcinoma Ovarian serous papillary carcinoma Ovarian mucinous cystadenocarcinoma Adrenal cortical carcinoma Esophageal adenocarcinoma
12 11 5 2 2 2 2 2 2 2 1 1 1 l 1 1 1 1
ImmunohistochemicaI markers in breast cancer real goat serum or normal horse serum. Trypsinization improved positivity of control sections only for anti-E29, so that sections tested for this antibody were previously incubated with trypsin 0.25% (type II, Sigma Chemical Company, St. Louis, MO). The slides were incubated overnight with the primary antibody at 4 ° C and then consecutively incubated with biotinylated goat anti-rabbit IgG or horse anti-mouse IgG and the avidin-biotinperoxidase complex (Vector Laboratories, Burlingame, CA) for 30 min each at 37°C. Thorough washing in PBS was carried out between the various stages. The substrate used was 0.05% 3,3"-diaminobenzidine tetrahydrochloride (Sigma). Slides were counterstained with hematoxylin, dehydrated, and mounted in balsam. Positive controls consisted of known positive normal or neoplastic tissue. Negative controls were performed replacing the primary antibody with non-immunized sera. One to five slides were observed for each tumor, depending upon its size and availability of paraffin blocks. Observation of the slides was performed through an optical microscope without previous knowledge of the primary site of the neoplasia. Statistical analyses of the results were performed through a mainframe computer and the software packgage for biology BMDR using chi-square for simple comparison and stepwise linear logistic regression for interaction of all the various immunohistochemical markers.
Table 2. Tumor metastases positive for the nine markers employed
Marker
Mammary origin
Non-mammary origin
ALA LCF CAS E29 MC5 SC GCDFPI5 BRST2 BRST!
14 (28%) 19 (38%) 22 (44%) 49 (98%) 47 (94%) 28 (56%) 31 (62%) 17 (34%) 36 (72%)
15 (30%) 02 (04%) 15 (30%) 40 (80%) 27 (54%) 20 (40%) 01 (02%) 02 (04%) 12 (24%)
203
Results
All the cases that had at least some strongly immunostained neoplastic cells were considered positive. Faint and uniform staining in most of the cells was considered negative. Comparison of the different areas of the tumors was helpful in the interpretation of some otherwise questionable cases. The results are summarized in Table 2.
Alpha-lactalbumin (ALA ) Diffuse cytoplasmic staining in most of the cells in addition to background staining in the surrounding connective tissue, sometimes hampering interpretation, was observed in most of the positive cases. Variations in the methods (longer incubation with normal serum and use of bovine serum albumin) and in the antiserum had little effect in reducing background. Fourteen (28%) cases from the 50 breast metastases and 15 (30%) of the other tumors were positive. There was no statistical difference between the groups (22= 0.194, p> 0.25). Lactoferrin (LCF) Strong cytoplasmic staining was seen in the LCF positive tumors. While 19 (38%) of the 50 breast tumors showed immunoreactive cells, only 2 (4%) of the 50 other metastases were positive. This difference reaches statistical significance (Z2= 15.43, p< 0.0001). Casein (CAS) The positive tumors presented homogeneous cytoplasmic staining. Twenty-two (44%) of the mammary origin specimens were positive, while 15 (30%) from the control group showed immunoreactivity. This difference was not significant 0 ( = 1.54, p< 0.1). HMFGMP (E29 EMA) Cytoplasmic and membrane staining, especially around luminar spaces, was seen in most of the positive specimens. Forty nine cases (98%) from the breast origin group and 40 (80%) of the control group were positive for E29. Although close to the limit, this difference was statistically significant 0~2= 6.54; p< 0.025).
204
PC Cardoso de Almeida and CB Pestana
I m m u n o c y t o c h e m i c a l Markers in Metastatic C a r c i n o m a s percentage of positive cells
100%
l
80% 60%
40% 20%
0% ALA
LCF
[~
CAS
EMA
I
I
gO5
SC
Non breast origin
I
GCDFPI5 BRSTI
~
1
BRST2
Breast origin
Fig. 1. Immunocytochemicalmarkers in metastatic carcinomas.
HMFGMP (MC5) Neoplastic cells presented a membrane staining pattern in the positive cases, especially on the lining of intra or extra cellular lumens. Like E29, MC5 was positive in the majority of the metastases from the breast (94%). However, positivity in the other tumors was much lower, showing immunosmined ceils in only 27 (54%) of the 50 cases. This difference was highly significant 0 ( = 18.76, p< 0.0001).
the two groups was also extremely significant 0~2= 38.65, p< 0.0001). While reactivity for the mammary origin cases was 62% (31 cases), just a few stained cells could be observed in the only positive case (2%) of the control group, which was an hepatocarcinoma.
BRST2
0.1).
Positive neoplasms presented a staining pattern similar to the one observed with GCDFP15. Breast metastases were positive in 17 (34%) of the 50 cases, while only 2 (4%) showed immunoreactivity among the other tumors. The difference was significant (Z2= 12.73, p< 0.0005). The two positive metastases from other sites were the hepatocarcinoma, also positive for GCDFP15, and a papillary thyroid carcinoma, both with rare stained cells.
GCDFP15
BRST1
Great variability in the number of stained cells as well as in the intensity of the staining was seen with this antibody. Difference in the positivity between
Great variability in the cytoplasmic staining intensity of different cells in the same tumor was observed in the BRST1 positive cases. Thirty-six
Secretory component (SC) Positive neoplastic elements presented homogeneous cytoplasmic staining for SC. Twenty-eight (56%) cases from breast metastases and 20 (40%) cases from the other metastases were positive. This difference showed no significance 0~2= 1.76, p
0.2) LCF
MC5
GCDFP15
BRST1
Breast
Other
Observed
Predicted
Standard deviation
origin
origins
p r o p o r t i o n (%)*
probability (%)*
o f pred. prob.
-
-
-
+
0
2
0.00
9.80
8.69
+
-
-
-
0
1
0.00
17.83
18.46 3.58
.
1
20
4.76
4.17
-
. +
.
. -
-
6
16
27.27
27.48
8.75
-
+
-
+
9
9
50.00
48.60
11.19 23.65
+
+
-
-
1
1
50.00
65.38
-
+
+
+
8
1
88.89
97.17
3.00
-
-
+
+
1
0
!00.00
79.77
21.30
-
+
+
-
6
0
100.00
93.22
7.19
+
-
+
+
1
0
100.00
95.16
7.20
+
+
-
+
2
0
100.00
82.49
15.40
+
+
+
+
15
0
100.00
99.42
0.82
* O b s e r v e d p r o p o r t i o n a n d p r e d i c t e d probability o f breast origin
of breast origin in each of the combinations are shown in Tables 3 and 4.
(72%) of the mammary metastases, but only 12 (24%) of the other metastases were positive using this antibody. This difference is highly significant (;~2= 21.19, p< 0.0001).
Discussion
Assembly of results The importance of each antigen in differentiating mammary origin from non-mammary origin can be visualized in Fig. 1. Stepwise logistic regression analysis was performed using two degrees of restriction to select the group of markers which together could more efficiently predict breast origin of the neoplasia. The less restrictive analysis (removal limit p > 0.2) selected the antibodies anti-LCF, anti-GCDFP15, MC5, and BRST1, while a more restrictive one (removal limit p> 0.15) selected only MC5 and antiGCDFP15. The combined results obtained with these markers as well as the estimated probability
Alpha-lactalbumin (ALA ) Our results for ALA are within the broad range of positivity observed by different authors in primary and metastatic breast carcinoma, as well as in nonmammary tumors [6-11, 43-47]. Table 5 shows the variation of ALA positivity from 0 to 81% and illustrates the extremely discrepant conclusions found in the several studies. The most important cause of such a discrepancy seems to be antisera specificity. Lee et al. [10], testing primary and metastatic mammary tumors, could not verify absence or decrease of the staining when anti-ALA antisera were absorbed with purified ALA, sug-
Table 4. S u m m a r y o f statistical results o b t a i n e d t h r o u g h m o r e restrictive ' s t e p w i s e ' logistic r e g r e s s i o n ( r e m o v a l limit p > 0.15) MC5
G C D F P 15
B r e a s t origin
O t h e r origins
Observed proportion
P r e d i c t e d probability
Standard deviation of
(%)*
(%)*
pred. prob.
-
-
1
23
4.17
5.74
+
-
18
26
40.91
40.05
7.33
+
+
29
1
96.67
97.92
2.20
-
+
2
0
100.00
81.13
18.23
* O b s e r v e d p r o p o r t i o n and p r e d i c t e d probability o f breast origin
4.42
206
PC Cardoso de Almeida and CB Pestana
gesting that crossreaction may be occurring. This finding is also supported by Hall et al. [48], who, studying lactating and neoplastic mammary tissue, could detect mRNA for ALA in the former and not in the latter group, but were able to identify the protein in both using anti-ALA antibodies. The wide range of results observed in the literature as well as the difficult interpretation of our results due to strong background staining, indicate that ALA antisera commercially available have a great variability of specificity and sensitivity and should not be used as breast tissue marker. Production of monoclonal antibodies for ALA might clarify these conclusions and bring new results. Lactoferrin (LCF)
Few studies have dealt with the presence of LCF in human neoplasms. It was first described in breast carcinomas by Hurlimann etal. [12], who observed 29.5% positivity using immunofluorescence in paraffin sections. Employing the same methodology, Mason etal. [13] found 8.3% positivity. Similar results (7.1%) were reported by Charpin et al. [14] using immunoperoxidase. There are no studies demonstrating LCF in extramammary or metastatic breast tumors, except for the report of one positive gastric carcinoma. In our study, positivity in the breast metastases group was close to the value found by Hurlimann et al. [12]. Surprisingly, other metastases presented very low positivity,
suggesting that LCF could be an important tool in differentiating the two groups. These conclusions, however, should be confirmed for a greater number and variety of metastatic neoplasms. Casein (CAS)
Data concerning casein synthesis by human breast carcinomas is controversial in the literature. Several studies demonstrated presence of this protein in tissues and breast carcinomas using polyclonal antisera with variable positivity [15-17]. The specificity of the reagent has been questioned since contamination with other proteins such as EMA, ALA, and LCF has been described, probably due to the low immunogenicity of casein [49-50]. One study using two monoclonal antibodies against casein could detect this product in lactating but not in normal or neoplastic breast tissues [51]. These conclusions, however, do not exclude the possibility that mammary neoplasms synthesize modified casein which is not detectable by monoclonal antibodies but is reactive with polyclonal antisera. Other evidence for absence of casein in breast carcinomas is the study of Hall et al. [52], who was unable to demonstrate mRNA for casein in human breast neoplasms. We employed a polyclonal antiserum absorbed with EMA, but not with LCE The strong association between CAS and LCF (p< 0.0001) in our study suggests the possibility of cross-reaction. In any case, the difference in CAS positivity
Table 5. ALA positivity in human neoplasias. Summary of the studies using immunohistochemistry
Author
Method
Results
Bussolati et al. Walker
IF IP
71% breast carcinoma 51% ductal breast carcinoma
Bahu et al. Bailey et al. Clayton et al. Lloyd et al.
IF IP IP IP
Le Doussal et al. Lee et al. Doria et al. De las Mulas Wick et al. Our study
IP IP IP IP IP IP
53% 0% 77% 81% 76% 69% 19%
IF - Immunofluorescence. IP - Immunoperoxidase.
ductal breast carcinoma/63% breast metastases breast carcinoma ductal breast carcinoma/55% breast metastases breast carcinoma/0% other tumors breast carcinoma ductal breast carcinoma/62% breast metastases ovarian tumors
0% breast carcinoma and metastases 50% breast carcinoma/49% other carcinomas 28% breast metastases/30% other metastases
Immunohistochemical markers in breast cancer
between the mammary and non-mammary metastases groups was not significant, indicating that this antibody is not useful to identify metastatic breast cancer. HMFGMP
Studies with the monoclonal antibody E29 (Dako EMA) show a high frequency of positivity in all epithelial tissues, including neoplasms of various origins and different histological types [18, 19]. In the most extensive study using this antibody, Pinkus et al. [21] have reported 100% positivity for breast carcinomas, but 88% for adenocarcinomas from other sites. There are no reports of E29 behavior in metastatic tumors. In our study, positivity for breast origin neoplasms was 98%. High positivity (80%) was also found in the control group, but the difference between the two groups reaches significance, although marginally (p< 0.025). Therefore, E29 has limited usefulness in differentiating mammary from nonmammary metastases. If negative, it practically excludes the possibility of breast origin, but if positive, it only demonstrates the epithelial nature of the lesion. MC5, another monoclonal antibody prepared against HMGFMP, is not commercially available and is less extensively studied. Although Ceriani et al. [20] reported 98% and Russo [22] 100% positivity in breast carcinomas, no systematic analysis has been performed in other tumors or metastases. In our cases, MC5 positivity for breast metastases (94%) was similar to the results cited above. Other metastases showed lower immunoreactivity (54%) compared to the first group. This difference is statistically significant (p< 0.0001). Positivity in the control group, nevertheless, is too high to use MC5 alone as a definitive breast tissue marker. Secretory component (SC)
The first important study demonstrating SC in human breast neoplasms was performed by Harris et al. [24] using immunofluorescence. They found positivity for all breast carcinomas and metastatic adenocarcinomas, as well as for colonic, ovarian, and renal neoplasms. On the other hand, Walker [25] observed only 30% of positivity for breast carcinomas. Her explanation for the discrepancy be-
207
tween the results was the method of antibody preparation used by Harris et al. [24], which allowed contamination. Brooks et al. [26], using the immunoperoxidase technique, found 50% positivity for breast carcinomas and 75% for other epithelial tumors. Our results for breast metastatic tumors (56%) are in agreement with those of Brooks et al. [26], although immunoreactivity for other metastases was lower in our analysis. This difference was not significant, suggesting that SC is not useful in the identification of breast origin due to its low specificity and sensitivity. GCDFP15
Since its identification in 1976, several studies reporting GCDFP15 in human mammary carcinomas have been published. Mazoujian etal. [27] using anti-GCDFPI5 found 90% positivity for breast carcinomas and 75% for non-mammary tumors with apocrine features. Other tissues and neoplasms were negative. Their results were further confirmed by Eusebi et al. [28, 29]. Employing the same monoclonal antibody against GCDFP15 that we used (BRST2), Wick et al. [11] observed 72% positivity for breast carcinomas, including primary and metastatic tumors, and only 4% for other tumors. The studies above suggest high specificity of GCDFP15 for mammary origin neoplasia, in accordance with the present study where very few non-mammary metastases (2-4%) showed immunoreactivity. Positivity for breast tumors varied according to the antibody employed, with higher reactivity using the polyclonal (62%) than the monoclonal antibody BRST2 (34%). The difference in sensitivity may be due to the greater number of antigenic determinants detected by the polyclonal antiserum, allowing detection of modified proteins possibly synthesized by the neoplasms. GCDFP15 is the most reliable single marker of neoplasia of breast origin in this study. BRST1
The only immunohistochemical study using this antibody was performed by Mesa-Tejada et al. [30], who found 93% positivity for human breast carcinomas against 35% positivity for other neoplasms. In the present study, positivity for breast
208
PC Cardoso de Almeida and CB Pestana
origin tumors was lower (72%), but there was still a significant difference between the immunoreactivity of the mammary and non-mammary group (p< 0.0001). A possible explanation for this difference is that Mesa-Tejada et al. [30] tested primary neoplasms, while we analyzed metastatic tumors. Tissue fixation and processing may have also contributed to this difference. Summary
The linear logistic regression analyses using two degrees of restriction indicates that a panel of two or more antibodies increases the accuracy of the immunohistochemical method in the identification of breast origin carcinomas. Table 3 indicates that positivity for LCF, MC5, GCDFP15, and BRST1 corresponds to a 99.42% predicted probability of breast origin, compared to only 4.17% when all markers were negative. The more restrictive analysis (Table 4) (removal limit p> 0.15) showed that MC5 and GCDFP15 were the best markers for identification of breast metastases. When both were positive (58% of breast origin cases), the predicted probability of breast origin was 98%, compared to only 5% when both were negative. In this study, 54% of the specimens are part of one of these two groups. Therefore, in more than half of the cases, it is possible to predict with accuracy whether the metastatic neoplasia has originated from the breast. In conclusion, we suggest that a panel of antibodies can be used to securely differentiate metastatic breast cancer from other cancers in a large number of metastatic tumors of unknown origin. Anti-ALA, often employed as a breast specific marker, is unsuitable for this purpose. Anti-LCF, MC5, anti-GCDFP15, BRST1, and BRST2 are more often positive in mammary origin than in non-mammary origin carcinomas. MC5 and antiGCDFP15 used together are the most efficient for this identification. The most specific breast origin marker is anti-GCDFP15. Polyclonal antibody against this protein showed a slightly better sensitivity than the monoclonal BRST2 and a similar degree of specificity for breast origin tissue.
Acknowledgements We thank Prof. Dr. Thales de Brito for his suggestions and comments;Dr. G. Bussolati for his generous gift of antibodies anti-casein and antiGCDFP15, and for valuable advice during this study; and to Dr. R. Ceriani for kindly providing the MC5 monoclonal antibody. This study was supported by Fundag~o de Amparo a Pesquisa do Estado de Sad Paulo (FAPESP #87/0369-8).
References 1. DeLellis RA, Sternberger LA, Mann RB, Banks PM, Nakane PK: Immunoperoxidase techniques in diagnostic pathology. Am J Clin Pathol 71: 483-488, 1979 2. Krhler G, Milstein C: Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256: 495-496, 1975 3. Nadji M, Tabei SZ, Castro A, Ming Chu T, Murphy GR Wang MC, Morales AR: Prostatic specific antigen. An immunohistologic marker for prostatic neoplasms. Cancer 48: 122%1232, 1981 4. Albores-Saavedra J, Nadji M, Civantos E Morales AR: Thyroglobulin in carcinoma of the thyroid: an immunohistochemical study. Human Pathol 14: 62-66, 1983 5. Nadji M, Morales ZR: Immunoperoxidase Techniques: A Practical Approach to Tumor Diagnosis. American Society of Clinical Pathology Press, Chicago, pp. 50-51 and 64-65, I986 6. Bussotati G, Ghiringello B, DiCarlo F, Aimone V, Voglino GF: Lactalbumin synthesis in breast cancer tissue. Lancet ii: 352, 1977 7. Walker RA: The demonstration of alpha lactalbumin in human breast carcinomas. J Pathol 129: 37-42, 1979 8. Bahu 1LM,Mangkomkanok-Mark M, Albertson D, Fors E, Molteni A, Battifora H: Detection of alpha-lactalbumin in breast lesions and relationship to estrogen receptors and serum prolactin. Cancer 46: 1775-1780, 1980 9. Clayton F, Ord6fiez NG, Hanssen GN, Hanssen H: Immunoperoxidase localization of lactalbumin in malignant breast neoplasms. Arch Pathol Lab Med 106: 268-270, 1982 10. Lee AK, DeLeUis RA, Rose PP, Herbert-Stanton T, Tallberg K, Garcia C, Wolfe HJ: Alpha-lactalbumin as an immunohistochemical marker for metastatic breast carcinomas. Am J Surg Pathol 8: 93-100, 1984 11. Wick MR, Lillemoe TJ, Copland GT, Swanson PE, Manivel JC, Kiang DT: Gross cystic disease fluid protein-15 as a marker for breast cancer: immunohistochemical analysis of 690 human neoplasms and comparison with alpha-lactalbumin. Hum Pathot 20: 281-287, 1989 12. Hurlimann J, Lichaa M, Ozzello L: In vitro synthesis of immu-
Immunohistochemical markers in breast cancer
13. 14.
15.
16.
17.
18.
19.
20.
21.
22. 23.
24. 25.
26.
27.
28.
29.
noglobulins and other proteins by dysplastic and neoplastic human mammary tissues. Cancer Res 36: 1284-1292, 1976 Mason DY, Taylor CR: Distribution of tranferrin, ferritin, and lactoferrin in human tissues. J Ctin Pathol 31: 316-327, 1978 Charpin C, Lachard A, Pourrean-Schneider N, Jacquemier J, Lavaut MN, Andonian C, Martin PM, Toga M: Localization of lactoferrin and nonspecific cross-reacting antigen in human breast carcinomas. An immunohistochemical study using the avidinbiotin-peroxidase complex method. Cancer 55: 2612-2617, 1985 Bussolati G, Pich A, Alfani V: Immunofluorescence detection of casein in human mammary dysplastic and neoplastic tissues. Virchows Arch A (Pathol Anat Histol) 365: 15-21, 1975 Pich A, Bussolati G, Carbonara A: Immunocytochemical detection of casein and casein-like proteins in human tissues. J Histochem Cytochem 24: 940-947, 1976 Fortt RW, Gibbs AR, Williams D, Hansen J, Williams I: The identification of 'casein' in human breast cancer. Histopathology 3: 395-406, 1979 Heyderman E, Steele K, Ormerod MG: A new antigen on the epithelial membrane: Its immunoperoxidase location in normal and neoplastic tissue. J Clin Pathol 32: 35-39, 1989 Gatter KC, Alcock C, Heryet A, Pulford KA, Heyderman E, Taylor-Papadimitriou J, Stein H, Mason D: The differential diagnosis of routinely processed anaptastic tumors using monoclonal antibodies. Am J Clin Pathol 82: 33-43, 1984 Ceriani RL, Hill DL, Osvaldo L, Kandell C, Blank EW: Immunohistochemical studies in breast cancer using monoclonal antibodies against breast epithelial cell components and with lectins. In: Russo J (ed) Immunocytochemistry in Tumor Diagnosis. Martinus Nijhoff, Boston, 1985, pp. 233-263 Pinkus GS, Kurtin PJ: Epithelial membrane antigen - A diagnostic discriminant in surgical pathology. Hum Pathol 16: 929940, 1985 Russo J: Immunocytochemistry in Tumor Diagnosis. Martinus Nijhoff, Boston, 1985, pp. 207-232 Russo J, Russo IH: Immunocytochemical markers in breast cancer. In: DeLellis R (ed) Advances in Immunohistochemistry. Raven Press, New York, 1988, pp. 431-475 Harris JP, Caleb MH, South MA: Secretory component in human mammary cancer. Cancer Res 35: 1861-1864, 1975 Walker RA: Differentiation of human breast carcinomas: An immunohistological study of appropriate and inappropriate protein production. J Patho1135: 87-95, 1981 Brooks JJ, Ernst CS: Immunoreactive secretory component of IgA in human tissues and tumors. Am J Clin Patho182: 660-665, 1984 Mazoujian G, Pinkus GS, Davis S, Haagensen DE Jr: Immunohistochemistry of a gross cystic disease fluid protein (GCDFP-15) of the breast. A marker of apocrine epithelium and breast carcinomas with apocrine features. Am J Pathol I10: 105112, 1983 Eusebi V, Betts CM, Haagensen DE, Gugliotta P, Bussolati G, Azzopardi JG: Apocrine differentiation in lobular carcinoma of the breast: a morphologic, immunologic and ultrastructural study. Hum Pathot 15: 134-140, 1984 Eusebi V, Millis RR, Cattani MG, Bussolati G, Azzopardi JG:
30.
31.
32. 33.
34.
35.
36. 37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
209
Apocrine carcinoma of the breast. A morphologic and immunocytocbemical study. Am J Patho1123: 532-541, 1986 Mesa-Tejada R, Keydar I, Ramanarayan M, Ohno T, Fenoglio C, Spiegelman S: Detection in human breast carcinomas of an antigen immunologically related to group-specific antigen of mouse mammary tumor virus. Proc Natl Acad Sci USA 75: 1529-1538, 1978 Esteban JM, Battifora H: Tumor immunophenotype: comparison between primary neoplasm and its metastases. Mod Pathol 3: 192-197, 1990 Kerbel RS: Implications of immunological heterogeneity of tumors. Nature 280: 358-360, 1979 Hart IR, Fidler IJ: The implications of tumor heterogeneity for studies on the biology and therapy of cancer metastasis. Biochem Biophys Acta 651: 37-50, 1981 Ceriani RL, Peterson JA, Blank E: Variability in surface antigen expression of human breast epithelial cells cultured from normal breast, normal tissues peripheral to breast carcinomas, and breast carcinomas. Cancer Res 44: 3033-3039, 1984 Kim U: Pathogenesis and characteristics of spontaneously metastasizing mammary carcinomas and the general principle of metastasis. J Surg Oncol 33: 151-165, 1986 Fidler IJ: Review: Biologic heterogeneity of cancer metastases. Breast Cancer Res Treat 9: 17-35, 1987 Arklie J, Taylor-Papadimitriou J, Bodmer W, Egan M, Mitlis R: Differentiation antigens expressed in the lactating breast are also detected in breast cancers. Int J Cancer 28: 23-29, 1981 Hsu SM, Raine L, Fanger H: Use of avidin biotin peroxidase complex (ABC) in immunoperoxidase techniques. A comparison between ABC and unlabeled antibody (PAP) procedures. J Histochem Cytochem 29: 577-580, 1981 Ormerod MG, Bussolati G, Sloane JP, Steele K, Gugliotta P: Similarities of antisera to casein and epithelial membrane antigen. Virchows Arch (Pathol Anat) 397: 327-333, 1982 Cordell J, Richardson TC, Pulford KAF, Ghosh AK, Gatter KC, Heyderman E, Mason DY: Production of monoclonaI antibodies against human epithelial membrane antigen for use in diagnostic immunocytochemistry. Br J Cancer 52: 347-353, 1985 Mesa-Tejada R, Palakodety RB, Leon JA, Khatcherian AS, Greaton C J: Immunocytochemical distribution of a breast carcinoma associated glycoprotein identified by monoclonal antibodies. Am J Patho1130: 305-314, 1988 Mazoujian G, Parish T, Haagensen D: Immunoperoxidase localization of GCDFP-15 with mouse monoclonal antibodies versus rabbit antiserum. J Histochem Cytochem 36: 377-382, 1988 Bailey AJ, Sloane JP, Trickey BS, Ormerod MG: An immunocytochemical study of alpha lactalbumin in human breast tissue. J Patho1137: 13-23, 1982 Le Doussal V, Zangerle PF, Collette J, Spyratos F, Hacene K, Briere M, Gest J, Franchimont P: Immunohistochemical detection of alpha-Iactalbumin in breast lesions. Eur J Cancer Clin Oncol 20: 1069-1078, 1984 Lloyd RV, Foley J, Judd WJ: Peanut lectin agglutinin and alphalactalbumin. Binding and immunohistochemical localization in breast tissues. Arch Pathol Lab Med 108: 392-395, 1984 De las Mulas J, Morales C, Lopez A, Toro M: Value of alpha-
210
47.
48.
49.
50.
PC Cardoso de Almeida and CB Pestana lactalbumin as tissue marker for breast carcinomas. Breast Diseases-Senologia 2: 53-58, 1987 Doria MI Jr, Adamec T, Talerman A: Alpha-lactalbumin in 'common' epithelial tumors of the ovary. An immunohistochemical study. Am J Clin Pathol 87: 752-756, 1987 Hall L, Craig RK, Davies MS, Ralphs DNL, Campbell PN: AIpha-lactalbumin is not a marker of human hormone-dependent breast cancer. Nature 290: 602-604, 1981 Weir HM, MacKinlay AG, Nash AR, Sarfaty GA: Human kappacasein as a tumor marker: purification and properties. Cancer Detect Prevent 4: 193-204, 1981 Burchell J, Bartek J, Taylor-Papadimitriou J: Production and
characterization of monoclonal antibodies to human casein. A monoclonal antibody that cross-reacts with casein and alpha-lactalbumin. Hybridoma 4: 341-350, 1985 51. Earl HM, McIlbinney RA, Wilson P, Gusterson BA, Coombes RC: Immunohistochemical study of beta- and kappa-caseins in the human breast and breast carcinomas using monoclonal antibodies. Cancer Res 49: 6070-6076, 1989 52. Hall L, Craig RK, Campbell PB: m-RNA species directing synthesis of milk proteins in normal and tumor tissue of human mammary gland. Nature 277: 54-56, 1979
