Abstract
The precursors of the blood group N and M-immunodominant structures, Tn and T (Thomsen–Friedenreich) epitopes (EPs) occur in ∼90% of carcinomas (CAs) but are masked in benign-diseased and healthy tissues. We determined quantitatively on 55 primary invasive ductal breast CAs, stages I to IV, the prognostic value of extent of Tn and T EP expression over an observation period exceeding 5 years postoperatively. Classical, established pathological and histological prognostic characteristic indicators associated with survival were subdivided by standard criteria into favorable and unfavorable categories. Tissue sections were reacted with monoclonal anti-Tn and -T antibodies, followed by the streptavidin–biotin–peroxidase–DAB procedure; counterstain was methyl green. Tn and T EPs were then quantitated by computerized image analysis. Of the 55 CAs, 51 clearly expressed Tn and T, and four had traces. Strong Tn EP expression was statistically significantly associated with shortened 5-year disease-free interval, increasing pTNM stages, positive lymph node status, and increasing combined histological grades. T EPs were usually well expressed but showed no significant association with prognostic factors. Our results suggest that quantitative immunohistochemistry–image analysis of Tn EPs of primary breast CAs may add new parameters to prognostication.
Keywords
C
Established prognosticators of the clinical course of breast CA include the following: TNM stage, histological grade (tubule formation, nuclear pleomorphism and hyperchromasia, mitotic rate) and, to some degree, steroid receptor status. Ancillary prognosticators such as C-erb B-2, cathepsin D, Ki67, p53, aneuploidy, and angiogenesis, as well as carcinoembryonic antigens, vimentin and epithelial membrane antigen are being investigated (Tavassoli 1992; Rosen 1996). The search continues for additional pathogenetic parameters which, in conjunction with established prognosticators, more accurately predict the eventual outcome of breast CA. There is strong evidence that the extent of expression (Springer and Desai 1977; Coon et al. 1982; Ghazizadeh et al. 1984; Yuan et al. 1986; Nishiyama et al. 1987; Springer et al. 1990, 1995; Springer 1984, 1995) and degree of clustering (Springer et al. 1983; Schlepper-Schαfer and Springer 1989; Nakada et al. 1991) of immunoreactive T EPs in CAs is of diagnostic importance, and the appearance of Tn is predominantly of prognostic import. Tn, T, and sialyl Tn EPs contribute actively to CA aggressiveness (Schlepper–Schαfer and Springer 1989; Inoue et al. 1990; Itzkowitz et al. 1990; Springer et al. 1983, 1990; Nakada et al. 1991; Takahashi et al. 1994; Springer 1995).
Immunohistochemistry (IHC) is a powerful tool for detection of these well-characterized EPs and may thus aid in understanding of some pathogenetic aspects of CA. Conventional evaluation of immunostaining permits only semiquantitative interpretation of the density and distribution of Tn and T EPs. Computerized image analysis allows quantitation of immunostaining (Mize et al. 1988), and hence more accurate assessment of some CA cell properties vs those of normal cells. The present study aimed to quantify immunohistochemically stained Tn and T EPs in invasive ductal breast CA tissues, which amount to about 75% of all breast CAs (Tavassoli 1992; Rosen 1996), by image analysis, and to correlate the results with established prognostic factors.
Materials and Methods
Tissues
Paraffin blocks of primary breast CA archived in Evanston Hospital (Evanston, IL) were used in this project. We studied 55 primary infiltrating ductal breast CAs, stages I–IV. Ages at surgery ranged from 34 to 77 years (median 56 years). Twenty-five CA patients had lymph node metastases. Staging by the surgical pathologists was according to the pathological TNM classification (Beahrs et al. 1992). Fifty-four patients were observed for 5 years, one only for 3.6 years (Stage IV), after which she was lost to follow-up. Also analyzed were 11 benign-diseased breast tissues based on the criteria of Rosen (1996), Tavassoli (1992), Page and Jensen (1994), and Page and Dupont (1993). At least three sections per block were obtained, separated from one another by more than 20 μ in thickness. There were no noteworthy variations in these sections per individual block in regard to analytical features. Three different specimens showed no increased risk (Page and Jensen 1994). They had mild ordinary ductal hyperplasia, sclerosing adenosis, or fibroadenoma. Six had hyperplasia without atypia (sclerosing adenosis one, intraductal papilloma one, focal papilloma one, moderate hyperplasia two, hyperplasia with florid papillomatosis and sclerosing adenosis one), and two had atypical hyperplasia. Ten of the 11 benign breast disease patients were followed for 1.5 years to 17 years 2 months (median 12 years 3 months). None developed breast cancer. One benign breast disease patient was lost to follow-up after surgery. An additional seven breast glandular tissue samples from different mammoplasties appeared to be normal histologically. These 18 cases served as controls and ranged in age from 34 to 61 years (median 48 years).
The largest diameter of the primary CAs ranged from 0.5 to 10 cm (median 2.0 cm). All tissues were routinely processed by fixation in 0.1 M phosphate-buffered 10% aqueous formalin for 24–36 hr at 20–23C and then embedded in paraffin. Serial tissue sections 5 μ thick were cut from the blocks, placed on coated glass slides (SuperFrost/Plus; Fisher Scientific, Pittsburgh, PA), and dried overnight in an oven at 50C. One section from each block was stained with hematoxylin–eosin to authenticate the tissue and confirm the pathologist's diagnosis. Histopathological grades were assessed by a modification of the Bloom–Richardson method (Elston 1987; Dalton et al. 1994).
Antibodies
Two pools each of three rodent monoclonal antibodies (MAbs) were used for Tn and T EP demonstration, respectively. The fine specificities of these MAbs have been described previously, together with their other properties: mouse MAb BaGS-3 elicited by isolated O Tn Ag and BaGS-5 raised against O Tn RBCs (Springer et al. 1988) were employed together with HB-Tn1 (DAKO; Carpinteria, CA) in assessing Tn EPs. Mouse anti-T MAb RS1–114 against human lung CA cell line A549 (Stein et al. 1989) and rat MAb HT-8 raised against O T RBCs, which crossreacts with goat anti-mouse Igs (Metcalfe et al. 1984), were used together with HB-T1 (DAKO) to detect T EPs. Our antibodies were affinity-purified (Springer et al. 1988). The optimal dilution of each MAb pool was predetermined using a poorly differentiated ductal breast CA as standard and healthy breast glandular tissue as negative control.
Immunohistochemistry (IHC)
All procedures were performed at room temperature (20–23C) and all conditions remained uniform throughout this study. The sections were deparaffinized in xylene, rehydrated in decreasing concentrations of ethanol/water and rinsed in Dulbecco's PBS (DPBS), pH 7.4. Endogenous peroxidase of tissues was inactivated by 30-min incubation in 3% H2O2 in 80% methanol with 0.1 M sodium azide. The sections were then rinsed with DPBS, treated with 0.3% BSA in DPBS for 30 min, and incubated with primary Ab. The fluids in the anti-Tn MAb pool were prediluted: BaGS-3 1:80; BaGS-5 1:80; HBTn-1 1:80. Similarly, the anti-T pool consisted of 3 MAbs diluted as follows: RS1–114 1:100; HT-8 1:200; HBT1 1:60. Incubation with primary Abs was 60 min, followed by washing in DPBS. Specimens were then incubated for 30 min with a 1:600 dilution of biotinylated goat anti-mouse Igs (DAKO). The slides were then washed in DPBS, incubated with a 1:300 dilution of streptavidin and biotinylated horseradish peroxidase complex (SABC) (DAKO) for 30 min, washed again, and reacted for 7 min in the dark with 0.05% 3–3′, 4–4′-diaminobenzidine tetrahydrochloride (DAB; Sigma, St Louis, MO) in DPBS containing 0.01% H2O2 (Wang et al. 1996). The sections were washed thoroughly in distilled water, counterstained with methyl green for 10 min, washed, dehydrated in increasing concentrations of ethanol, cleared in xylene, and mounted with Cytoseal 60 (Stephens Scientific; Riverdale, NJ). Negative controls, which ascertained specificity of the reactions, consisted (a) of at least two replicate immunostained sections in which primary Ab had been omitted or had been absorbed with RBCs containing its corresponding Ag and (b) benign-diseased and normal tissues surrounding the CA; the latter also served as built-in controls. Positive controls were serial sections from a primary breast CA known to be Tn- and T-EP-positive.
Quantitative Image Analysis
Computerized image analysis (Mize et al. 1988) was performed to quantitate the immunostained Tn and T EPs of primary CA epithelia. The image analysis system consisted of a Nikon Microphot-FXA microscope with a 100-W halogen lamp, × 4 and × 10 planapochromatic objective lenses, an intermediate × 2 magnification lens, and a 482-nm interference filter. A voltage regulator allowed operation at a constant voltage (11 V) for obtaining immediately stable illumination to ensure accurate measurements, which were periodically checked against a standard neutral density filter during each experiment. Additional components of the system were a power supply (PSM-4; Nikon), a Sony 3 CCD video camera with a control unit (DXC-760 MD), a Sony TV monitor (PVM-1943 MD), and an IBM PC-compatible Data Stor UM486 computer with integrated imaging CPU, a Matrox MVP frame grabber, and appropriate software (Image 1/AT; Universal Imaging, West Chester, PA).
Quantitative image analysis was essentially by the methods described by Larsson and Hougaard (1994). With all automatic functions disabled, specimens were illuminated with monochromatic light (482 ± 10-nm), corresponding to the absorbence peak of the wavelength of the DAB reaction product (Richmond et al. 1988). The percentage of immunostained epithelial cell areas in CA tissues was measured at low magnification with a × 10 objective lens. Image acquisition, shade correction, and image analysis were controlled by the Image 1/AT program. The immunostaining intensities of Tn and T Ags were measured in pixel values on a scale from 0 = black to 255 = white, i.e., very bright light. A threshold was determined using replicate sections of the tissues immunostained without primary Abs. The positively stained areas were highlighted (segmented) (Mize et al. 1988) by threshold and appeared green on the monitor. For each slide, the microscope fields to be measured were chosen as follows: if there were less than 20 CA foci per tissue section, then the entire CA regions were measured; if there were more, the largest 20 CA foci were selected and measured. When the glandular CA region to be measured was outlined with the computer mouse-controlled arrow on the screen (boundary setting), the luminal areas were excluded. All connective tissue and necrotic areas were also excluded, so that only the epithelial area of the CA was measured (Figure 1C). The total pixel number in positively stained areas was divided by the total pixel number in all measured areas to obtain the mean percentage of immunostained epithelial cell areas. A mean percentage of immunostaining areas of >33% vs ≥33% was considered as low vs high EP expression. To prove the reproducibility of the computerized image analysis system, the same CA area (e.g., the largest CA area of one section) was selected and measured by the same observer on different days or by different observers to compare the results. In addition, duplicate sections were stained on different days and measured to evaluate the technical reproducibility.
Statistical Analyses
Two basic parameters were assessed for their statistic significance: (a) association of prognostic factors, defined by pathological TNM stage (Beahrs et al. 1992), size of primary CA, lymph node status, and combined histological grade (Elston 1987; Dalton et al. 1994), with disease-free intervals in the 55 patients with primary breast CA (Table 1), and (b) correlation of extent of expression of Tn and T EPs with prognostic features such as pTNM stage (I and IIA vs IIB–IV), size of primary CA (>2 cm vs ≥2 cm), lymph node status (negative vs positive), combined histological grade (I and II vs III), and disease-free interval (Table 2).
Fisher's exact test (Mosteller et al. 1970) was used to calculate p values. A p value of >0.05 denotes significance. The degree of association was also measured by odds ratios, interpreting the association as strong if ≥3 or ≤0.4 (Bishop et al. 1975).
Results
Immunohistochemistry
Breast CA cells usually have large quantities of Tn and T EPs in cytoplasm and on cell membranes, demonstrable specifically with the anti-Tn and anti-T MAb pools used. Of the 55 primary invasive ductal CA studied, 51 expressed Tn and T EPs abundantly and four had only traces (1–5%). In the majority of cases, both Tn and T EPs were well expressed, a finding characteristic for CAs in contrast to normal and benign-diseased tissues (see Discussion). However, T EPs were mainly seen in lumina of well-differentiated CA and adjacent CA cells (Figure 1A), and T EP staining showed a wide variability in both well- and poorly-differentiated CAs, whereas staining for Tn EPs was in general more extensive in cytoplasm and on cell membranes than that of T EPs (Figures 1B–1D). Compared to primary CAs, metastatic CAs in lymph nodes usually showed increased Tn and decreased T EP expression (Figures 1E and 1F). Normal breast tissues and benign breast lesions did not bind anti-T and -Tn MAbs (see Discussion) and were found to be negative, even on replicate sections from the same paraffin block, except occasionally in the glandular lumina (Figure 2A) or in apocrine metaplasia, which frequently showed some staining in cytoplasm for Tn or T EP (Figure 2B).
Prognotic factors associated with disease-free interval of the 55 primary breast carcinoma (CA) patients studied
aSemiquantitative assessment of tubule formation, nuclear features, and mitotic index.
Histologically normal and benign glandular epithelial cells in the vicinity of the CA served as built-in controls (Figures 1C, lefthand side, lower quarter, and 1D, right side of photo). None of the negative controls (omission of primary Ab or its preabsorption) showed any staining for Tn and T EPs. Intra- and interobserver measurements of the same CA area on different days under constant conditions gave nearly the same results; the mean run-to-run variations were less than 8%.
Statistical Analyses
The results of statistical analyses of disease-free interval of our cohort in relation to established prognostic indicators of breast CA are listed in Table 1. The most highly significant prognostic parameters related to the disease-free interval are pTNM stage (p = 0.002), and combined histological grade (p = 0.002), followed by lymph node status (p = 0.016) and primary tumor size (p = 0.032). Degree of association for all parameters is strong.
Table 2 shows that percentage of Tn-immunostained CA area has a strong association with disease-free interval (p = 0.006) and combined histological grade (p = 0.001). High Tn EP expression is also associated with pTNM stage (p = 0.027), and lymph node metastasis (p = 0.013), but tumor size is without significant p value (p = 0.109) and only weak association (0.39). Therefore, in all instances, degree of association is strong except with tumor size. It is notable that there was no association of T EP expression with any of the prognostic indicators; therefore, the data are not shown in Table 2.
Immunohistochemical detection of T and Tn EPs in invasive ductal breast carcinoma by SABC-DAB method with anti-T (
Non-carcinoma breast glandular tissues, immunostained for Tn and T EPs as in Figure 1. (
Discussion
We and others have previously shown that both immunoreactive T and Tn EPs are almost always found in breast CA, as determined with human monospecific, polyclonal antibodies, and only rarely in small amounts in normal and benign tissues (Howard and Taylor 1980; Yuan et al. 1986; Stein et al. 1989; Nakada et al. 1991; Taylor–Papadimitriou 1994; Brockhausen et al. 1995; Springer 1995; Yamashita et al. 1995; Cao et al. 1996; Hakomori 1996) where their location usually amounts to being extracorporeal or immunoprivileged. These findings have been confirmed for secreted milk fat globules and the extracellular MUC1 antigen, although some findings remain controversial because of the anti-reagents employed, many of which are crossreactive (Howard and Taylor 1980; Seitz et al. 1984; Springer 1984; Springer et al. 1975, 1985; Hanisch and Baldus 1997). Increased density of Tn over T in primary ductal breast CAs correlates with poor prognosis (Springer 1989; Springer et al. 1990). CA-Tn and -T are autoimmunogens and their epitopes are essential carbohydrate structures involved in adhesion, invasion, and metastasis of CA cells (Springer et al. 1983; Schlepper-Schαfer and Springer 1989). However, immunohistochemical study of these EPs has thus far been semiquantitative only. In the present investigation, we quantified the immunohistochemically stained Tn and T EPs in ductal breast CA tissue using a computerized image analysis system. Computer-based video densitometry systems and image analyzers are being used to study the amounts and distribution of organic macromolecules in IHC-labeled tissues (Mize et al. 1988; Nabors et al. 1988). This tool is also being applied in more function-linked morphology and molecular cell biology studies (Larsson et al. 1991; Larsson and Hougaard 1994). Because immunostaining heterogeneities of CA areas sometimes make it difficult to obtain reproducible results by routinely used random selection of 5–10 size-fixed fields per section for quantitative analysis, investigators often face the problem of selection of appropriate microscopic fields to ensure representative and reproducible measurements. Ductal CAs usually grow as irregular nests and cords (Bartow 1994). The computer software enabled us to precisely outline, on the screen, irregularly-shaped CA foci with the mousecontrolled arrow (Figure 1C) and to obtain precise results by selecting the 20 largest CA foci, or total CA regions if there were less than 20 CA foci per tissue section. The results showed acceptable reproducibility by intra- and interobserver determination (two to three microscopists). It should be mentioned that in lobular CAs, which we did not investigate, single strands of CA cells often invade the stroma without cord or nest formation (Bartow 1994). In such cases, other region selection and calculation methods need to be developed.
Prognostic factors associated with the mean percentage of Tn epitopes immunostained on primary breast carcinoma (CA) cells of the 55 patients studied
aMean percentage of immunostained areas: high, ≥33%; low, <33%.
Recent advances in studies of mucin structure and glycosylation revealed that in breast and other CA cells the product of the MUC1 gene is aberrantly glycosylated and that Tn, T EPs, and core peptides are exposed (Devine and Mckenzie 1992; Brockhausen et al. 1995; Yamashita et al. 1995). It is these novel uncovered carbohydrate structures that distinguish CA cells from normal epithelial cells in immunohistochemical studies and that elicit specific autoimmune responses (Taylor–Papadimitriou and Epenetos 1994; Springer et al. 1995; Springer 1984, 1995).
Here, we employed MAb pools directed against uncovered Tn and T EPs in breast CA and demonstrated that the extent of Tn and T EP expression in CA can be quantitatively analyzed immunohistochemically, which may add to prognostication. The expression of Tn EPs predominated in invasive CAs over T EPs and the increasing expression of Tn EPs was highly significantly associated with established, prognostically unfavorable parameters of breast CA. Quantitative measurements of Tn EP expression provide important information on the metastatic potential of a CA and on patients' disease-free interval, provided that strict reproducibility of results is ascertained by the stringent controls as described in the Materials and Methods. Strict quantitation of biological phenomena is an overriding goal to comprehend and control patho-physiological phenoma (cf. Gordon–Taylor 1948).
In addition to aberrant glycoslation, there is another way, that may lead to the exposure of Tn and T EPs on cells. Previous studies have documented that sialic acid may be released from major sialoglycoproteins of erythrocyte membranes as the result of cell aging (Gattegno et al. 1979; Khansari et al. 1983). On desialylation of the RBC membrane, T Ag and some Tn Ag is unmasked (Springer and Desai 1982). In the present study we observed some positively immunostained Tn and T EPs in ductal lumina of normal breast glands (Figure 2A), but not within normal epithelial cells. It is likely that these Tn and T Ag reactivities result from degenerated cell components discharged from glandular epithelial cells and are not accessible to the immune system.
Our studies concerned overall densities of Tn and T EPs of carcinoma structures; single cells were not investigated. Therefore, we have no spatial resolution and no data on subpopulations. We plan to obtain information on the absolute concentration of both of these EPs per particular cell using a quantitative standard such as that described by Nabors et al. (1988).
Footnotes
Acknowledgements
Supported by US National Cancer Institute grant no. CA 22540 and by the Heather M. Bligh Cancer Research Fund.
We are grateful to M. Kaufman, MD for making the paraffin blocks of breast tissue available, to W. Rheault, PhD, PT, for guiding us in the statistical work and performing most of it, to M. Ghazizadeh, MD, for some preliminary investigations, and to J. Prasad and S. Kutsch for their excellent technical assistance.