Abstract
Recent reports have suggested that the immunodetection of tumor cells in bone marrow of small-cell lung cancer (SCLC) patients is by far more effective than traditional cytohistological methods and that this may be clinically relevant. This study aimed to evaluate whether the level of detection of tumor cells in bone marrow is affected by different immunostaining methods. Using two anti-NCAM monoclonal antibodies (MAbs), we compared four different “sandwich” methods on cytospin preparations of the N592 human SCLC cell line and of bone marrow aspirates from 37 SCLC patients. Our data indicate that the combination of the alkaline phosphatase-anti-alkaline phosphatase and streptavidin-biotin-alkaline phosphatase complex methods provides the best results in terms of sensitivity and specificity, and of intensity of immunoreaction and absence of staining background. Moreover, bone marrow micrometastases detected by this method were prognostically relevant and identified, among patients with apparently limited disease according to conventional staging procedures, a subgroup with shorter survival. We suggest that the choice of a sensitive immunostaining technique may significantly increase the detection rate of SCLC cells in bone marrow, mirroring the biological aggressiveness of the disease. (J Histochem Cytochem 47:1075–1087, 1999)
Keywords
S
Many studies on the detection of SCLC micrometastases in BMA have employed antibodies to NCAM (Frew et al. 1986; Berendsen et al. 1988; Hay et al. 1988; Moss et al. 1988; Trillet et al. 1989; Leonard et al. 1990; Beiske et al. 1992; Skov et al. 1992; Myklebust et al. 1993a,b; Pasini et al. 1994a, 1995). In fact, NCAM (CD56, NKH-1, cluster-1 antigen) is expressed by almost all SCLCs (Souhami et al. 1987, 1991; Stahel et al. 1994) and in a variety of normal tissues (Cunningham et al. 1987; Dalseg et al. 1989; Patel et al. 1989; Hida et al. 1991; Lanier et al. 1992), and in most tumors of neuroectodermal and mesodermal lineage (Patel et al. 1989, 1991; Moolenaar et al. 1990; Tome et al. 1991; Kern et al. 1992; Miettinen and Cupo 1993; Ledermann et al. 1994; Nakamura et al. 1995; Pasini et al. 1994a,1995). NCAM belongs to a family of membrane-bound, homophilic calcium-dependent glycoproteins involved in both cell-cell and cell-substrate interactions (Keilhauer et al. 1985; Cunningham et al. 1987; Rutishauer et al. 1988; Acheson et al. 1991). Various isoforms of the molecule have been described, deriving from alternative mRNA splicing (Cunningham et al. 1987). The expression of these isoforms is developmentally regulated in a tissue-specific manner, with prevalence of strongly sialylated isoforms of 200–250 kD during early embryogenesis (embryonic type) and less sialylated isoforms of 180, 145, and 129 kD in adult tissues (Pollerberg et al. 1986). The isoforms of NCAM present in SCLCs are mostly of the embryonic type (Moolenaar et al. 1990; Komminoth et al. 1991; Patriarca et al. 1997). Although the biological role of these molecules in SCLC is still unknown, there is some evidence that sialylated isoforms may be involved in the metastatic process owing to their reduced adhesive properties (Doyle et al. 1990; Moolenaar et al. 1990, 1992; Carbone et al. 1991; Komminoth et al. 1991; Patriarca et al. 1997).
A certain number of studies addressing the localization of SCLC micrometastases in BMA by use of anti-NCAM antibodied have used standard methods of immunostaining (Frew et al. 1986; Berendsen et al. 1988; Hay et al. 1988; Moss et al. 1988; Trillet et al. 1989; Leonard et al. 1990; Beiske et al. 1992; Skov et al. 1992; Myklebust et al. 1993a,b; Pasini et al. 1994a, 1995). However, SCLC cells may sometimes show reduced expression of NCAM (Doyle et al. 1990) to the point that standard immunostains may be inadequate to highlight minimal degrees of bone marrow contamination. To the best of our knowledge, no data are available concerning the influence of different immunostaining procedures in the detection of SCLC micrometastases in BMA.
This study compares four immunocytochemical staining “sandwich” methods to improve the detection rate of BMA micrometastases. We evaluated the immunoreactivity of two anti-NCAM antibodies in cytospin preparations of the N592 human SCLC cell line and of BMA from 37 SCLC patients. Moreover, we correlated the findings of bone marrow positivity with the clinicopathological features of the patients. Our study suggests that the choice of the immunocytochemical technique may affect the detection rate of metastatic tumor cells with anti-NCAM antibodies and that bone marrow micrometastases parallel the biological aggressiveness of the disease.
Materials and Methods
Strategy of Study, Selection of Patients and Adequacy of Samples
We designed and conducted a two-phase experiment to test four different nonlabeled antibody detection methods based on alkaline phosphatase (AP) with two purified mouse anti-human NCAM MAbs to examine the effectiveness of different labeling conditions on the detection of contaminating tumor cells in BMAs of SCLC patients. In the first phase, we tested two antibodies, clone NCC-LU243 (IgG2a) and clone NCC-LU-246 (IgG1) (Nippon Kayaku; Tokyo, Japan) on cytospin preparations of the N592 human SCLC cell line using the following detection systems: AP-anti-AP (APAAP) technique; APAAP-double bridge technique; combination of APAAP and streptavidin-biotin-AP (SABAP) complex techniques, as described by Davidoff et al. (1991) with some modifications; and Universal Large Volume DAKO LSAB Kit AP (Dakopatts; Glostrup, Denmark). The two MAbs react with adjacent epitopes of the 145- and 185-kD human isoforms of NCAM (Hirano et al. 1989; Moolenaar et al. 1990) and of the highly sialylated isoforms (Moolenaar et al. 1990). DAKO LSAB Kit AP was applied according to the manufacturer's instructions. The other three methods tested are detailed in Table 1. AP labeling was preferred to peroxidase labeling because blocking procedures for endogenous peroxidase activity were shown to dramatically decrease NCAM immunostaining intensity. On the basis of these preliminary findings, in the second phase we examined the immunoreactivity of the two antibodies developed by the four immunostaining techniques in a series of 74 BMAs obtained from bilateral posterior iliac crests of 37 patients. These patients were retrieved from our series of BMAs obtained from 145 SCLC patients at the Institute of Medical Oncology of the University of Verona from March 1990 to July 1998. The selection criteria included (a) the availability of enough unstained cytospin preparations to compare the diverse immunocytochemical techniques and (b) the presence in each cytological sample of high cellularity composed of a well-preserved monolayer of mononuclear cells. In fact, only the use of high-quality cytospin preparations containing at least 1 × 105 mononuclear cells allowed contamination of tumor cells to be estimated with reasonable confidence. According to the Veterans Administration Lung Cancer Study Group (Hyde et al. 1965), patients were classified as having limited disease (LD, 22 cases) if tumor was restricted to one hemithorax with regional metastases to hilar, ipsilateral, and controlateral mediastinal and/or supraclavicular lymph nodes and ipsilateral pleural effusion, or extensive disease (ED, 15 cases) if tumor was spread beyond these sites. According to the TNM staging system (Mountain 1997), in our study, the LD patients corresponded to Stage III and ED patients to Stage IV of disease.
Cell Line
The N592 human cell line, kindly provided by Dr. Sylvie Menard (National Cancer Institute; Milan, Italy), was grown in RPMI medium 1640 (MA Bioproducts; Walkersville, MD) supplemented with 10% heat-inactivated fetal calf serum and 80 mg/ml gentamycin at 37C in a fully humidified atmosphere of 5% CO2. Before use, the cultures were washed repeatedly in RPMI 1640 with 5% heat-inactivated fetal calf serum, and viable cells were counted by trypan blue exclusion at a final concentration of 1–1.5 × 106 cells/ml. Cytospin slides were then prepared for immunostaining using the same procedure as for BMA.
Immunostaining procedures a
Dakopatts, Glostrup, Denmark: Merck, Darmstadt, Germany: Sigma Chemical, St Louis, MO; Vector Laboratories, Burlingame, CA; Southern Biotechnology Associates, Birmingham, AL
Bone Marrow Aspirates
All BMAs (3–5 ml each) were collected in disposable heparinized syringes, sedimented onto a Ficoll-Hypaque (Pharmacia; Uppsala, Sweden) density gradient, cytospun, and fixed in acetone. Briefly, after density centrifugation (400 × g for 30 min at 19C), the mononuclear cell layer was washed three times in RPMI medium 1640 supplemented with 15% heat-inactivated fetal calf serum (350 × g for 7 min at 4C) and then resuspended at 1-1.5 × 106 cells/ml. Then the isolated mononuclear cells were cytospun (Cytospin 3; Shandon Scientific, Cheshire, UK) by introducing 75 μl of the cell suspension into each sample chamber (Cytofunnel) and centrifuging them at 500 rpm for 4 min at room temperature (RT) on 3-aminopropyltriethoxysilane-coated slides (Sigma Chemical; St Louis, MO). According to this procedure, a monolayered spot of 0.6 cm diameter containing 5–7.5 × 104 well-preserved mononuclear cells was obtained for each slide. Cytospins were then fixed in pure acetone for 5 min at 4C and stored at -20 or −80C in aluminum foil. Before immunostaining, frozen cytospin preparations were air-dried for 1 hr and rehydrated in Tris buffer with 5% human serum.
Immunocytochemistry
Both primary antibodies were used at a concentration of 20 pig/ml at RT for 1 hr. Briefly, after the primary antibody incubation in the APAAP technique, polyvalent rabbit anti-mouse immunoglobulin and APAAP mouse complex were applied in sequence. In the APAAP-double bridge technique, the second and third incubations of the previous procedure were repeated in the same sequence. In the combined technique, polyvalent biotinylated horse anti-mouse immunoglobulin, APAAP mouse complex, and again biotinylated horse anti-mouse immunoglobulin were applied in sequence and followed by labeling with the streptavidin-AP complex (Figure 1). Endogenous AP activity was blocked with 1% w/v levamisole (Sigma) in the chromogen solution. AP activity was developed with 2% New Fuchsin (Merck; Darmstadt, Germany) and 0.2% naphthol AS-BI phosphate (Sigma) in propandiol buffer, pH 8–9 to yield a red endproduct. Slides were lightly counterstained with 1% Harris' hematoxylin. The specificity of all reactions was verified by replacing primary antibodies with unrelated mouse IgG in buffer at a comparable dilution. Cytospins of N592 cell line stained in parallel were employed as external positive control, and myeloid cells of each BMA were used as internal negative control. Moreover, BMAs from healthy volunteers and patients with unrelated, nonmalignant diseases provided the control group for noncarcinoma patients. Finally, some BMAs from unrelated diseases and cytospins of the MCF-7 human breast cancer cell line provided the baseline for background staining.
Double Immunocytochemical Staining
Double immunostains for NCAM and low molecular weight cytokeratins, as well as for NCAM and leukocyte common antigen (CD45), were carried out in both BMAs from SCLC patients and artificial mixtures (see below) to differentiate, in doubtful cases, true cancer cells from other nonspecifically stained cells (Makin et al. 1984; Kurtin and Pincus 1985; Leader et al. 1986).
Diagram of the multilayer immunocytochemical assay for the NCAM detection in the N592 human cell line. All secondary antibodies are labeled with alkaline phosphatase (AP) through the soluble complexes APAAP mouse and streptavidin-AP to amplify staining intensity (more enzyme molecules are bound per antigenic site).
To co-localize NCAM, cytokeratins, and CD45, we used a protocol combining two APAAP+SABAP staining procedures at RT. For NCAM-cytokeratin or NCAM-CD45 reactions, the first APAAP+SABAP staining was carried out employing the NCC-LU-246 MAb (1:50) as a primary antibody reagent and New Fuchsin as a chromogen substrate to obtain a red endproduct. After incubation with human serum (5% in TRIS buffer), a second APAAP+SABAP procedure was added using the Cam 5.2 MAb to cytokeratin polypeptide 8 and weakly to 7 (Becton-Dickinson, Mountain View, CA; dilution 1:10) in combination with the 5D3 MAb to polypeptides 8, 18, and 19 (BioGenex, San Ramon, CA; dilution 1:60) or an anti-human leukocyte common antigen antibody (DAKO-LCA, mixture of clones PD7/26-2B11, Dakopatts; dilution 1:100), respectively, as a second primary antibody reagent and a 0.7% v/v solution of 5% w/v nitroblue tetrazolium (NBT; Sigma) and 2.5% w/v 5-bromo-4-chloro-3-indolyl phosphate (BCIP; Sigma) in Tris buffer as a second substrate to yield a blue-black endproduct (Unger et al. 1986). Slides were lightly counterstained with 1% Harris' hematoxylin.
Artificial Tumor Cell Mixtures for Assessment of Sensitivity
In a set of experiments we used BMAs from healthy volunteers and from patients with unrelated, nonmalignant diseases artificially contaminated with a known number of tumor cells to determine the sensitivity of our method in the detection of NCAM-reactive tumor cells. After sedimentation onto a Ficoll-Hipaque gradient density, the cell layer was resuspended at 1 × 106 cells/ml and contaminated logarithmically with the N592 cell line up to 1 tumor cell in 105 mononuclear cells, with additional contaminations at 1:2000, 1:4000, 1:8000, and 1:50,000 cells. Tumor cell dilutions were cytospun to 5–7.5 × 104 cells/slide, fixed in cold acetone, and immunostained for NCAM with the NCCLU-246 MAb (1:50) according to the combined APAAP+SABAP method. Experiments were repeated four to seven times for every dilution and the mean number of immunoreactive tumor cells over 1 × 105 mononuclear cells was recorded.
Evaluation of Immunostained Cells
All estimations of labeled cells, including double stains, tumor cell dilutions, and control group for noncancer patients, were performed independently and blindly by two observers (GP and FP) using an Olympus BH-2 light microscope at a magnification of x400 or x1000 oil immersion. BMAs from SCLC patients were analyzed without knowledge of the patients' identity or stage of disease. For NCAM, cells with thick and continuous membrane staining and morphologically consistent with tumor cells were regarded as immunoreactive. Five cytospins of N592 cell line were examined for each antibody and for each immunostaining method, scoring at least 500 tumor cells in every preparation. A simple semiquantitative grading system for NCAM immunoreactivity was devised: ≤s50% of positive cells +; 51–75% of positive cells ++; 76-100% of positive cells +++. For BMAs, two to four cytospins were examined for each antibody and for each immunostaining method. Quantitative estimation was thus obtained by counting the number of immunopositive tumor cells over 1 × 105 mononuclear cells. All cytospin preparations were also scanned for intensity of immunostaining (graded subjectively as mild +, moderate ++, or strong +++) and nonspecific background staining (evaluated subjectively as absent -, or present in the form of a granular precipitate of staining +). All scores were assessed by each observer for each antibody and for each immunostaining procedure by counting every case twice. The highest score for each antibody and the highest score between the two observers were thus reported. Positivity for cytokeratins was judged by the occurrence of a strong cytoplasmic staining, whereas positivity for CD45 was identified by a stained ring decorating the cell surface membrane.
Statistical Analysis
The statistical tests used were Mann-Whitney's test and Fisher's exact t-test. The intra- and interobserver reproducibility was evaluated by analysis of variance and Spearman's rank test, respectively. Survival curves were calculated with the Kaplan-Meier method (Kaplan and Meier 1958) and compared by the log rank test (Mantel and Haenszel 1959). Multivariate analysis was carried out by the proportional hazards Cox model (Cox 1972). The following variables were examined: stage of disease combined with bone marrow assessment (patients with LD and negative BMA vs patients with LD and positive BMA vs patients with ED); Karnofsky performance status (Karnofsky et al. 1948); bone marrow biopsy status; age; and sex. Confidence intervals (CI) were considered at the 95% level. In all tests, p was considered significant at ≤0.05, and all significance levels were of the two-sided type.
Results
N592 Cell Line
NCAM immunostaining results of N592 cell line cytospins are summarized in Table 2 and illustrated in Figure 2. The best results were obtained for both antibodies by using the combined APAAP+SABAP method, which provided the highest number of positive cells and the most intense labeling cell by cell, along with complete absence of nonspecific background staining. In addition, the universal large volume DAKO LSAB Kit AP method strongly stained the large majority of tumor cells with both antibodies, but this was associated with nonspecific granular background staining even after dilution of the reagents. On the other hand, the APAAP and double bridge APAAP techniques, especially with NCC-LU-243, resulted in the staining of fewer cells and in weaker intensity cell by cell, although the nonspecific background staining was very weak.
Bone Marrow Aspirates
The results of NCAM immunostaining of BMA cytospins are summarized in Table 3. For both antibodies, the combined APAAP+SABAP technique showed a high and comparable number of tumor cells in 24/37 patients (p=NS), strong intensity of immunostaining, and absence of nonspecific background. The DAKO LSAB Kit AP gave almost identical results for the number of stained cases, (23/37), positive cells (p=NS), and the intensity of immunostaining, but it was associated with nonspecific granular background staining. The APAAP method resulted in a smaller number of stained cases (12 to 22/37), with lower numbers of immunoreactive cells (p=NS) and weaker staining intensity. However, nonspecific background staining was absent. The APAAP-double bridge method stained 19/37 cases with LU-243 and 23/37 cases with LU-246 (p=NS), showing moderate staining intensity and variable occurrence of nonspecific granular background staining. Differences in the prevalence of positive cases (12/37 vs 24/37; p=0.005) and of NCAM-reactive cells (136 vs 710; p=0.019) were found only for the NCC-LU-243 antibody between the APAAP and the APAAP+SABAP techniques, respectively. For all methods of immunostaining, intraobserver reproducibility did not show statistically significant differences in the number of immunostained tumor cells, and a high grade of correlation between the two observers was found (p < 0.001; CI 97.0-100.0%).
With all methods of immunostaining, NCAM-positive tumor cells could be seen not only as single cells but also as cell aggregates of variable sizes (Figures 3a and 3b). Both individual and clustered tumor cells were generally larger than normal mononuclear cells. Clusters were formed by two to three to many tumor cells in direct cell-to-cell contact without recognizable intercellular space. The prevalence of clusters and their size increased with the absolute number of tumor cells in BMAs. In fact, 13 patients with a diffuse marrow involvement (more than 70-100 positive cells/1 × 105 cells total according to the combined APAAP+SABAP method) exhibited many clusters, each formed by a variable number of irregularly aggregated tumor cells.
Control experiments showed that normal bone marrow elements present on cytospin preparations of both clinical material and control group of noncancer patients were completely negative for both NCC-LU-243 and NCC-LU-246, although a few lymphoid cells (Figure 4a) showed a weak and often discontinuous decoration of the membrane, and rare plasma cells (Figures 4b 1 and 4b2 ) and megakaryocytes (Figure 4c) exhibited granular staining of the cytoplasm.
Results of double immunostaining for NCAM and cytokeratins confirmed a definite co-localization of the two markers in the same cells, morphologically consistent with neoplastic epithelial cells (Figures 5a 1 and 5a 2). Although no cytokeratin+/NCAM− tumor cells were noted, variable numbers of CD45+/NCAM−(Figure 5b, arrow) or CD45+/NCAM+ lymphoid cells (Figure 5c) were also seen. Differential features of NCAM-positive cancer cells vs NCAM-positive non-cancer cells are summarized in Table 4.
Results of NCC-LU-243 and NCC-LU-246 reactivity in N592 cell line as a function of four different immunostaining techniques
APAAP, alkaline phosphatase-anti-alkaline phosphatase method; APAAP+SABAP, combination of the APAAP and the streptavidin-biotin complex method; DAKA LSAB Kit AP, Universal Large Volume DAKO LSAB (labeled streptavidin-biotin) alkaline phosphatase.
Semiquantitative scoring system: +, ≤50% positive cells; ++, 51-75% positive cells; +++, 76-100% positive cells.
Intensity of immunostaining subjectively graded as light (+), moderate (++), and strong (+++).
Nonspecific background staining evaluated as present (+) or absent (−).
NCC-LU-243 and NCC-LU-246 reactivities in BMAs of SCLC patients as a function of four different immunostaining techniques
APAAP, alkaline phosphatase-anti-alkaline phosphatase method; APAAP+SABAP, combination of the APAAP and the streptavidin-biotin complex methods; DAKO LSAB Kit AP, Universal Large Volume DAKO LSAB (labeled streptavidin-biotin) alkaline phosphatase.
Positive tumor cells/100,000 mononuclear cells; ∗differences between APAAP and APAAP+SABAP as to number of LU-243-positive cases (p=0.005) and tumor cells (p=0.019).
Intensity of immunostaining subjectively graded as light (+), moderate (++), and strong (+++).
Nonspecific background staining evaluated as present (+) or absent (−).
Analysis of Artificial Mixtures
Results of the artificial contamination tests showed that NCAM-positive tumor cells were detected in expected quantities in all mixtures up to the final dilution of 1:105, even if standard deviations were sometimes high (Table 5). Moreover, the intensity of immunostaining of tumor cells was independent of the number of cells detected in every dilution. In doubtful cases, double immunostaining for NCAM and cytokeratins or CD45 was carried out to differentiate true tumor epithelial cells from NCAM-positive noncancer cells. Of note is that the percentage of contaminating cells was one- to 900-fold higher than the percentage of added cells and the enrichment fraction (% of detected cells/% of added cells) was inversely proportional to the percentage of added cells (Figure 6).
Clinical Implications
Positivity of BMAs evaluated by the combined APAAP+SABAP technique was associated with extensive disease [15/15 (100%) ED vs 10/22 (45.5%) LD] (p < 0.001) and positivity of conventional histological examination of marrow biopsy [8/25 (32%) positive BMA vs 0/12 (0%) negative BMA] (p = 0.035), whereas no significant relationship was found with age, sex, and performance status. The 22 patients with limited disease presented a significantly lower prevalence of BM involvement than the 15 patients with extensive disease (99 vs 1386 cells, respectively) (p = 0.006). The mean amount of marrow contamination in the sub-group of limited disease with positive BMAs was lower than in extensive disease patients (216 vs 1594 cells), although this difference was not statistically significant. The occurrence of BMA involvement significantly affected patients' survival (median survival 9 mo vs 17.5 mo) (p < 0.001) (Figure 7a). Other significant indicators were extensive disease (median survival 9 mo vs 16.5 mo) (p < 0.001), positivity of conventional bone marrow biopsy (median survival 9 mo vs 15.5 mo) (p < 0.001), and performance status lower than 70% (median survival 9 mo vs 16 mo) (p=0.003), whereas age and sex did not affect survival. Combining stage of disease and BMA immunoreactivity at diagnosis, three groups of patients with different prognoses could be identified: Group A, with limited disease and negative BMA (12 patients; median survival 16.5 mo) (“true limited disease”); Group B, with limited disease and positive BMA (10 patients; median survival 13 mo) (“untrue limited disease”); and Group C, with extensive disease and either positive or negative BMA (15 patients; median survival 9 mo). The likelihood of survival was significantly different among groups: A vs B, p=0.017; B vs C, p < 0.001; A vs C, p < 0.001 (Figure 7b). In multivariate analysis, both extensive disease (hazard ratio = 26.778; CI = 4.380-166.009; p < 0.001) and limited disease with positive BMA (hazard ratio = 7.399; CI = 1.655-33.079; p = 0.003) emerged as independent predictors of a dismal prognosis.
NCAM expression in the N592 human SCLC cell line with NCC-LU-243 (a-d) and NCC-LU-246 (e-h) MoAb according to APAAP (a,e), APAAP-double bridge (b,f), labeled streptavidin-biotin DAKO Kit AP (c,g), and combined APAAP+SABAP (d,h) methods. The best results for number of positive cells, intensity of immunoreaction, and negligible background staining were achieved by using the combined method (x1000, oil immersion).
Discussion
Our results can be summarized as follows. First, the choice of the immunostaining technique may influence the detection level of contaminating tumor cells in bone marrow of SCLC patients using anti-NCAM antibodies. Second, BMA contamination by SCLC cells is related to the extension of disease and predicts poor survival. Finally, the presence of tumor cells in BMAs of patients with limited disease appears to identify a subgroup of “untrue limited disease” with reduced life expectancy.
Differences between NCAM-positive SCLC cells and NCAM-positive noncancer cells
Evaluation of sensitivity of the APAAP+SABAP method in the detection of NCAM-positive tumor cells
Experiments were performed four to seven times for every dilution.
Positive tumor cells/100,000 mononuclear cells.
Only the combined APAAP+SABAP method provided for both LU-243 and LU-246 the highest sensitivity, specificity, and staining intensity, with a complete absence of nonspecific background staining on both BMAs and cultured cells. The other three methods tested yielded unsatisfactory results on both materials, with fewer tumor cells revealed, and/or weaker intensity of immunostaining, and/or variable occurrence of disturbing background staining. This observation emphasizes not only the usefulness and the diagnostic value of an immunocytochemical panel approach but also the need for adequate immunostaining techniques for detection of contaminating cells in bone marrow. The higher sensitivity of the combined APAAP+SABAP immunostaining technique is due to the increase in AP molecule numbers brought about at the antigen site through the reaction of the APAAP (two molecules of enzyme for each single molecule of antibody) and biotin-streptavidin-AP (a lattice structure containing several enzyme molecules) complexes. Whereas the original Davidoff method combined the APAAP and the avidin-biotin-AP complex techniques through a four-step procedure (Davidoff et al. 1991), in this study we adopted a five-step procedure by adding biotinylated antibody again before labeling with the streptavidin-AP complex to further enhance the sensitivity of immunostaining (Figure 1). The main advantage of this procedure, which uses biotinylated bridge immunoglobulins over other reported “sandwich” techniques employing direct enzyme-conjugated secondary antibodies (Tòth et al. 1994), lies in the greater numbers of enzyme molecules bound to the site of primary reaction.
NCAM expression in BMA using MAb NCC-LU-243 (a) and MAb NCC-LU-246 (b) according to the combined method of immunostaining. Strong membrane labeling is seen in both dispersed and clustered tumor cells with no relevant differences between the two antibodies (x1000, oil immersion).
NCAM membrane immunostaining in lymphoid cells, showing a weaker and discontinuous decoration of the membrane (a). Rarely, mononuclear cells resembling plasma cells (b1 and b2) and megakaryocytes (c) exhibit nonspecific staining of the cytoplasm (x1000, oil immersion).
Double staining for NCAM-cytokeratins (a) and NCAM-CD45 (b,c). Simultaneous staining for cytokeratins in the cytoplasm (blue) and for NCAM in the cell membrane (red) is seen in both clustered (a1) and dispersed (a2) tumor cells. Lymphoid cells exhibit strong expression for CD45 (b, blue, arrow), with an NCAM-positive tumor cell being completely negative for this marker (b, red, asterisk). Simultaneous expression for CD45 and NCAM on a lymphoid cell appears as a brownish color (blue for CD45 and red for NCAM) (c) (x1000, oil immersion).
To our knowledge, this is the first report in which such an immunostaining amplification system has been evaluated in cytological preparations of BMAs in SCLC patients. Furthermore, we have successfully tested the efficiency of this method in BMAs of gastric and breast cancer patients using several monoclonal antibodies recognizing either epithelial antigens, such as clones MLu-C1, MBr1, MBr8, MOv8, and MOv16 (kindly provided by Dr. Sylvie Menard) and clone BerEp4, or cell cycle-related antigens, such as Ki-67 (unpublished observations). This technique was particularly suitable in cases of false-negative results with conventional immunostaining techniques, such as the APAAP method. In fact, the formation of lattice structures with several enzyme molecules may overcome the presence of small amounts of antigen. The reliability of this amplification system was assessed not only by the morphological recognition of immunoreactive cells as tumor cells but also by the co-localization of NCAM and cytokeratins. Moreover, the complete absence of nonspecific background staining confirmed the high specificity of the method. Crossreaction with plasma cells and megakaryocytes was easily recognizable as nonspecific because it was confined to the cytoplasm, persisted in the negative control experiments using nonrelated IgG, and the cell morphology was clearly not consistent with that of epithelial cells. We believe that the morphology should be the basic principle of immunocytochemical detection of isolated epithelial cells in bone marrow because a potential contribution to nonspecific staining has been reported by normal mononuclear cells directly reactive to AP (Borgen et al. 1998). Although NCAM is also expressed by lymphoid cells, including natural killer cells (Hercend and Schmidt 1988; Di Giuseppe et al. 1997), potential misinterpretation of these cells as tumor cells could be minimized by using the combination of morphological and immunocytochemical criteria listed in Table 4. In fact, NCAM-positive noncancer cells showed discontinuous membrane labeling with a weaker intensity of immunostaining, a co-localization of CD45 (NCAM+/CD45+) but not cytokeratin (NCAM+/cytokeratins−), and a cell morphology different from that of epithelial cells. By combining two consecutive APAAP+SABAP procedures with different primary antibodies (NCAM, cytokeratins, CD45), it is possible to identify simultaneously the subcellular localizations of these reagents without remarkable color mixing or false-negative results. Therefore, we believe that the judicious use of double staining techiniques should minimize the risk of a false-positive detection of tumor cells by NCAM immunostaining and should also make it possible to localize additional markers on individual disseminated cells: a fraction of NCAM+ tumor cells also expressed the proliferation-associated molecule Ki-67(Gerdes et al. 1984) (not shown).
Relationship between detected and added cells in artificial mixtures of BMA from healthy volunteers and patients with unrelated, nonmalignant diseases. Vertical axis corresponds to the enrichment fraction, indicated by the ratio between percentage of detected cells and percentage of added cells.
(a) Actuarial curves of 37 patients according to the absence (BMA−) or presence (BMA+) of tumor cells at diagnosis: the patients with contaminating tumor cells in BMAs show a decreased percentage of cumulative survival (p<0.001). (b) Actuarial curves of the same patients according to the combination of stage of disease and BMA immunoreactivity: Group A, true limited disease (12 patients; median survival 16.5 months); Group B, untrue limited disease (10 patients; median survival 13 months); and Group C, extensive disease (15 patients; median survival 9 months). Group A vs Group B, p=0.017; Group B vs Group C, p<0.001; Group A vs Group C, p<0.001 (b).
The high sensitivity of the combined APAAP+SABAP method enable us to identify tumor cells in expected amounts in all dilution tests, even though the number of cells detected could be higher than that of added cells (Table 5). This can certainly be explained by the varying distribution of tumor cells during centrifugation within cytospin preparations, especially in cases with low NCAM-positive tumor cell counts, probably due to the different density of cancer cells compared with that of marrow cells (Molino et al. 1991: Allgayer et al. 1997).
Few studies have been aimed at increasing the sensitivity of AP complex-based signal systems for immunocytochemistry. These approaches included repeated applications of signal complexes (Ordronneau et al. 1981: Davidoff et al. 1991), new reagents (Universal Large Volume DAKO LSAB Kit: Dakopatts), or the new procedure of catalyzed reporter deposition (Bobrow et al. 1991). Moreover, a modified avidin-biotin complex procedure for detection of antigens at increased sensitivity did not consider AP as an enzyme molecule in the immunostaining sequence (Grumbach and Veh 1995). In our hands, the amplification systems based either on the repeated application of signal complexes, such as the APAAP-double bridge method, or the use of a high-sensitivity reagent kit, such as the Dakopatt's method, produced a variable and disturbing background staining both in clinical and in cultured material. Another recent approach for amplifying immunostaining is the high-temperature epitope retrieval method incorporating either microwave irradiation (Shi et al. 1991) or regular oven (Man and Tavassoli 1996) or pressure cooker (Miller and Estran 1995) heating. However, in our experience heating was not suitable because it almost completely abolished NCAM immunoreactivity in all cell preparations tested (not shown).
Regarding clinical implications, our data, although preliminary, show that BMA contamination by tumor cells is predictive of poorer survival (p<0.001) (Figure 7). However, this phenomenon might reflect the very poor prognosis of patients with extensive disease, because in our series positive BMAs were more common in patients with extensive disease (p<0.001). Consistently, we detected at least one order of magnitude more tumor cells in BMAs of patients with extensive disease (1386 cells) than in those with conventionally assessed limited disease (99 cells) (p=0.006). This observation also confirms previous reports showing that the persistence of residual marrow disease at clinical remission (Leonard et al. 1990), as well as the presence of tumor cells in BMAs at diagnosis (Bucher et al. 1994; Pasini et al. 1994b, 1995), are predicitive of early metastatic relapse and reduced survival rate, respectively. Of note is that our figure of 45.5% of bone marrow positivity in limited disease confirms previous articles reporting that 30-50% of limited disease patients have BMA contamination when assessed immunocytochemically (Berendsen et al. 1988; Hay et al. 1988; Moss et al. 1988; Trillet et al. 1989; Beiske et al. 1992; Bucher et al. 1994; Pasini et al. 1994b, 1995). The level of marrow involvement of these “untrue” limited disease patients was lower than that of extensive disease (216 vs 1594 cells), even though this did not achieve significance. This finding is in agreement with the median survival of this subset, which was intermediate between that of “true” limited disease and extensive disease patients (Figure 7b), as well as with the results of multivariate analysis, in which both the occurrence of limited disease with positive BMA (p=0.003) and the presence of ED (p<0.001) emerged as independent factors predicting a dismal prognosis. Our data confirm a growing body of literature that the immunocytochemistry of BMAs is an easy and reproducible method for selecting a subset of limited disease patients who are prognostically advantaged (Hay et al. 1988; Menard et al. 1988; Leonard et al. 1990; Bucher et al. 1994; Pasini et al. 1994a,b, 1995). Moreover, it has been claimed that leukapheresis products for the support of high-dose chemotherapy may be contaminated in those patients who exhibit bone marrow involvement on the basis of tumor cells detected immunocytochemically (Brugger et al. 1994). Although the clinical value of this marrow colonization is still unclear, it raises concern, and therefore immunocytochemical screening for bone marrow and autograft contamination has been included in multi-institutional studies of high dose chemotherapy (Brugger et al. 1996), and standard chemotherapy has been proposed as a method of in vivo purging (Elias 1995; Brugger et al. 1996). Independent of speculations of whether bone marrow is a sanctuary selectively and precociously colonized by tumor cells or simply a signal of disease activity at many other anatomic sites (Leonard et al. 1990), the patients with “true” limited disease, as defined in our study (i.e., patients with limited disease without marrow contamination), may become suitable subjects for innovative therapy, including high-dose chemotherapy or targeted immunotherapy (Myklebust et al. 1993a,b). We therefore emphasize the potential usefulness of immunocytochemical BMA assessment, especially by virtue of its easy standardization and negligible discomfort to the patient.
Footnotes
Acknowledgments
Supported by grants from the Ministry of University and Scientific Research (60%), Rome, Italy.