Abstract
Several studies have shown that p53 ression has important clinical implications as an indicator of prognosis and response to chemotherapy or radiotherapy in different human tumor types. Determination of p53 expression by immunohistochemistry (IHC) has been incorporated into routine practice and its reliability has been consolidated. However, flow cytometric (FCM) analysis might represent an important objective and rapid approach. In the present study we determined p53 expression by IHC and FCM on a series of 118 human solid tumors. IHC determination was performed on histological sections and FCM analysis on cell suspensions. Low correlation coefficients (r5 from 0.22 to 0.57) were observed between IHC and FCM data from individual tumors. By considering the IHC approach as the gold standard, high sensitivity and low specificity were found for FCM in detecting p53 expression. The FCM analysis of p53 expression and DNA content showed p53-positive cells in all cell cycle phases. Moreover, in most breast, lung, and colon aneuploid tumors (77%), p53-positive cells were detected only in the subpopulations with abnormal DNA content. In conclusion, FCM-p53 expression cannot be used alternatively to IHC determination, and its clinical relevance remains to be validated. Nevertheless, FCM may provide important information about p53 protein expression in the different subpopulations and cell cycle phases.
I
Immunohistochemical (IHC) detection of p53 expression has been established as a relatively easy and straightforward method for fresh and archival tissues (Porter et al. 1992; Hall and Lane 1994; Soussi et al. 1994). Available monoclonal antibodies recognize both wild-type and mutant forms, but there may be a selective detection of the latter owing to the very short half-life of the former. However, IHC determination is subjective and time consuming because of the series of hand manipulations involved. Flow cytometric (FCM) analysis should offer rapid and objective quantitation of p53 expression and provide further information on the expression of the oncosuppressor gene in the different cell subpopulations and cell cycle phases (Clark et al. 1992; Remvikos et al. 1992; Camplejohn 1994).
In the present study, we analyzed the results of p53 expression obtained by IHC or FCM in the same series of different solid tumor types using the monoclonal antibody PAb1801.
Materials and Methods
Fresh tumor samples from 118 patients with primary breast (30 cases), colorectal (52 cases) or non-small-cell lung (36 cases) carcinoma were obtained at surgery. The tumor specimens were divided into two parts. One was immediately frozen in liquid nitrogen and stored in a freezer at − 80C for multiparameter FCM analysis, and the other was fixed in formalin and embedded in paraffin for IHC determination.
Multiparameter FCM
Tissue samples were thawed and minced with a scalpel in cold PBS. The samples were filtered through 70-μm nylon mesh. After washing in PBS and centrifugation, the cells were loaded at a concentration of 1 × 106. The cells were then washed once, centrifuged, and incubated with 500 μl of a solution containing 0.1% Tween-20 and 0.5% bovine serum albumin (BSA) for 10 min at 20C to permeabilize cell membranes and block nonspecific protein binding. After centrifugation and a wash in PBS, parallel samples were incubated, at a dilution of 1:10 at 20C for 60 min, with PAb1801 (Oncogene Science; Manhasset, NY), a human-specific anti-p53 antibody that recognizes an epitope between amino acids 32 and 79 and reacts with normal (wild-type) and mutated forms of p53. The negative control samples were incubated with an irrelevant mouse IgG1 antibody at the same concentration. The cells were then washed in PBS, centrifuged, and incubated with a secondary FITC-conjugated goat anti-mouse antibody (Sigma; St Louis, MO) at a dilution of 1:50 for 30 min at 20C in the dark. After immunofluorescence staining, the cells were centrifuged and resuspended in a solution containing propidium iodide (PI, 5 μg/ml), RNase (10 kU/ml; Sigma), and Nonidet P40 (0.005%). The samples were stored in the dark for 30–60 min before FCM analysis.
Cells were filtered through 30-μm nylon mesh and analyzed on a FACScan flow cytometer (Becton-Dickinson; San Jose, CA) equipped with an argon laser. Excitation light wavelength was 488 nm at 15 mW. Green (FITC) and red (PI) fluorescences were separated by a 560-nm dichroic mirror. In addition, the green and red photomultiplier tubes were guarded by a 530-nm bandpass and a 650-nm longpass filter, respectively. For each sample, 10,000 events were collected and stored in listmode (Lysys II software; Becton-Dickinson) for the later analyses. A peak width vs area cytogram was used to discriminate and gate out doublets from the analysis (Doublet Discrimination Mode; Becton-Dickinson). The p53-positive rate was defined as the percentage of p53-positive cells in the samples incubated with PAb1801 minus that detected in the negative control samples (FCM-p53) (Danova et al. 1990; Goukon et al. 1994).
Immunohistochemistry
Four-μm histological sections from formalin-fixed, paraffinembedded blocks were treated as previously described (Sil-vestrini et al. 1993) with a solution of 3% H2O2 for 5 min at room temperature (RT) to quench endogenous peroxidase activity. Sections were then incubated in a solution of 3% BSA for 20 min at 20C to block nonspecific protein binding. Finally, sections were incubated for 1 hr at RT in a humidified atmosphere with a 1:50 dilution of the same PAb1801 used for FCM. After incubation, the specimens were washed twice with PBS and then processed for the avidin-biotin-per-oxidase method (Vectastain ABC Kit; Vector Laboratories, Burlingame, CA) according to the manufacturer's recommendations. The slides were then incubated in diaminobenzidine and H2O2 chromogen substrate for 5 min at RT, washed in running tapwater for 2–3 min, counterstained in Meyer's hematoxylin, and mounted with a permanent mounting medium (Figures 1A-1C).
p53 immunohistochemistry with PAb1801 in formalin-fixed, paraffin-embedded sections of breast (
A set of breast cancers with high levels of p53 was used as positive control. Negative controls were obtained by omission of the primary antibody. Staining was considered positive if unequivocal brown staining was seen in the tumor cell nucleus. Consecutive areas were scored and positive tumor cells were quantified by two independent observers by evaluating at least 3000 cells from different specimens of the same tumor. Positivity was expressed as the percentage ratio of p53-positive cells over the total number of tumor cells (IHC-p53). The interobserver variability was less than 10%.
Analysis of Data
For each tumor type, FCM-p53 and IHC-p53 values were subjected to nonparametric regression analysis with estimation of Spearman's correlation coefficient (rs). Moreover, data were compared by the descriptive method as reported by Altman and Bland (1983). The average of p53 values obtained by FCM and IHC on each tumor was plotted against the difference between FCM and IHC p53 values. Such a descriptive analysis made it possible to evaluate the disagreement magnitude of two immunochemical estimates as a function of the average values and to spot outliers and possible trends.
Results
A wide variability in the percentage of p53-positive cells from tumor to tumor within the same histotype was observed by using both immunochemical approaches. In particular, in breast cancer the fraction of IHC p53-positive cells showed a slightly wider range (1.3–71%) than the fraction of FCM p53-positive cells (7.5–56.2%). Conversely, the fraction of p53-positive cells was similarly scattered but with a shift towards lower values in IHC than in FCM analysis in colon (2.3–49.7% vs 12.5–72.6%), rectal (5.1–74.2% vs 11–82.1%), and lung (1.8–59.3% vs 7.2–81.4%) cancer.
The concordance in p53-negative or p53-positive tumors, as defined by IHC and FCM approaches, was maximal for lung cancer (89%), intermediate for colon (69%) and breast (60%) cancers, and lowest in rectal adenocarcinoma (48%). Moreover, the correlation coefficients, obtained in paired comparisons of IHC and FCM percentages of p53-positive cells on individual tumors, were different for the various tumor types and, in particular, were lowest for breast (rs = 0.35) and lung (rs = 0.22) cancers and highest for colon (rs = 0.48) and rectal (rs = 0.57) adenocarcinomas.
A breakdown analysis (Table 1) as a function of DNA ploidy status showed a consistent agreement between p53-positive and p53-negative tumors, as detected by the two methodologies, in diploid and aneuploid tumors for lung and colon and a higher agreement in diploid than in aneuploid tumors in breast and rectal lesions. The correlation coefficients between the p53 values obtained by IHC and FCM approaches were very high or high for diploid breast, lung, and colon, but not for diploid rectal cancers, and were generally poor or very poor for aneuploid cancers of all histotypes. For lung cancer in particular, despite the 90% qualitative concordance between IHC and FCM determinations, the correlation coefficient was extremely low for the poor quantitative relation of the values provided by the two approaches in p53-positive tumors. In fact, the concordance between IHC and FCM was limited to intermediate values of positivity (20–35% p53-positive cells). Instead, the low IHC values were two- to ninefold lower than the corresponding FCM values, and the high IHC values were two- to fivefold higher than the corresponding FCM values.
Agreement between histochemistry and flow cytometry in detecting p53 expression as a function of DNA-ploidy
Ratio between the IHC/FCM p53+ + IHC/FCM p53 cases and the overall number of cases.
The sensitivity and specificity of the FCM approach compared to the IHC method, considered as a reference standard, are shown in Table 2. Sensitivity, i.e., the frequency of IHC-p53-positive tumors identified also by FCM as p53-positive, was generally high for all tumor types, ranging from 85.7% to 94.1%. Conversely, specificity, defined as the frequency of IHC-p53-negative tumors also identified by FCM as p53-negative, was still high for lung cancer and progressively lower for breast and colorectal carcinomas.
Ratio between true P53+ and IHC-P53+.
Ratio between true P53- and IHC-P53-.
To better understand the discrepancy between qualitative and quantitative estimates for c53-positive tumors, values obtained by FCM were compared to those obtained by IHC using a descriptive method (Figure 2). For each tumor, the mean value of the two immunochemical approaches was plotted against the difference between the p53 values obtained by IHC and FCM analyses. A general overestimation by FCM was observed for all the tumor types. Conversely, underestimation by FCM was observed in a limited number of cases, and lung cancers accounted for 50% of them.
Comparison between IHC- and FCM-p53 values by the Altman and Bland method.
In diploid lesions, FCM bivariate analysis of p53/NDA showed p53 expression in all S- and G2M-phase cells, with coexistence of p53-positive and p53-negative cells in the G0/1-phase (Figure 3). In aneuploid lesions, p53 expression could be detected in diploid and aneuploid subpopulations (Figure 4A), or only in the aneuploid subpopulation (Figure 4B). In particular, the FCM overexpression of p53 protein in the only aneuploid cell subpopulation was most frequently observed in breast (80%), colon (79%), and lung (74%) cancers, whereas p53 positivity in diploid and aneuploid cell subpopulations was observed in 71% of rectal cancers. With regard to p53 cell cycle distribution, as observed in diploid tumors, the G0/1-phase usually showed the coexistence of p53-positive and p53-negative cells.
A typical diploid tumor in which p53 protein accumulation was detected in G0/1-, S-, and G2/M phases. Bivariate DNA/immunofluorescence plots of cells stained with isotypic control (left) and PAb1801 (right).
Two aneuploid tumors in which p53 protein accumulation was detected in diploid and aneuploid subpopulations or only in the aneuploid population. Bivariate DNA/immunofluorescence plots of cells stained with isotypic control (left panels) and PAb1801 (right panels).
Discussion
Alterations in the p53 gene are the most common genetic events in human cancers, and their clinical implications have been repeatedly reported. Such evidence makes of utmost importance the search for methodologic approaches that would be highly feasible and able to give reliable information. Molecular detection of mutations through single-strand conformational polymorphism analysis and DNA sequencing should be reserved for basic or clinical studies on limited series for tumors. IHC detection is highly feasible on consecutive series of patients, provides clinically relevant information, and represents an ideal technique for determining the localization and distribution of a gene product in tumor tissues. In fact, human solid cancers are usually a mixture of tumor and inflammatory cells, extracellular connective tissue, blood vessels, and often adjacent normal tissue, and microscopic evaluation allows an accurate selection of tumor cells.
FCM analysis gives quick, objective, and reproducible results, and p53 determination by FCM has been found to be an important method in experimental systems. In fact, such an approach has allowed detection of very low levels of protein after viral infection within a cell population and in tumorigenesis (Agrawal et al. 1994). The dual parameter p53/DNA analysis is a useful tool to monitor variations of p53 expression after exposure to chemical or physical agents (Gazitt and Erdos 1994; Hickman et al. 1994) in the different cell cycle phases. In addition, bivariate analysis of the levels of p53 protein and the incorporation of bromo-deoxyuridine has provided information on the functional status of p53 in transiently transfected cells (Sharma et al. 1993).
The potential of FCM to detect p53 expression, as well as other oncogene products, in human tumors is well known, but it has not entered the clinical routine. Moreover, although p53 expression by FCM has been evaluated in different human solid tumors (Blount et al. 1991; Clark et al. 1992; M⊘rkve et al. 1992; Ramel et al. 1992; Remvikos et al. 1992; Zhu et al. 1993; Qiao et al. 1994), to our knowledge the only follow-up results available show a prognostic relevance of FCM-p53 for colorectal (Remvikos et al. 1992) and non-small-cell lung cancers (M⊘rkve et al. 1993) but not for hepatocellular carcinomas (Qiao et al. 1994).
Results from our study show a modulation of the agreement between FCM and IHC data as a function of the different tumor types. In fact, FCM appears highly sensitive and specific, compared to IHC, in detecting p53-positive lung cancers. Conversely, in breast, colon, and rectal cancers, the FCM approach still has a high sensitivity but a low specificity, because the incidence of undetectable expression by FCM in the presence of IHC reactivity was appreciable. In addition, a breakdown analysis as a function of DNA-ploidy status evidenced a generally good correlation between the two methodologies in diploid but not in aneuploid cancers for all types except rectal tumors.
The fundamental difficulty encountered in FCM detection of p53 expression, as well as of other antigens, is definition of the threshold for positivity. We standardized the methodology by subtracting for each tumor its control value from the p53-positive value, based on the assumption of a stoichiometric relation between the immunofluorescence signal measured on single cells and the relative gene product concentration. However, variability related to the staining methodology cannot be excluded. Taking into consideration all the possible drawbacks of each approach under investigation, the partial disagreement between FCM and IHC data could be tentatively explained by the low sensitivity of human observers to discrimination of low levels of immunostaining intensity and by the presence of strong FCM immunofluorescence signals in IHC p53 weakly expressing cells, as well as FCM-positive signals in cells from IHC p53-negative cases. Such explanations are supported by the high frequency of FCM p53 false-positive cases, as confirmed by the Altman and Bland method (1983), which showed a high level of sensitivity for FCM to evidence a few IHC-positive cells. The reasons underlying the tumor type-dependent modulation of agreement should be investigated, although such a finding could only be due to the relatively small number of cases analyzed within each histotype. Therefore, the FCM approach cannot yet be considered an alternative to IHC. However, in spite of the discrepancies between the results from the two methods, bivariate FCM analysis can give important information on p53 expression as a function of cell cycle phases. The presence of a p53-negative cell fraction in G0/1 of diploid and aneuploid lesions could be ascribed to the normal stromal population with a wild-type p53 protein not detectable for its short half-life.
In conclusion, our results demonstrate that FCM can be used to identify subpopulations expressing or not expressing p53 protein in human solid tumors. However, caution is recommended for its use as an alternative to immunohistochemical detection before its clinical relevance has been validated in large series of patients.
Footnotes
Acknowledgements
Supported in part by grant no. 96.00688.PF39 from the Consiglio Nazionale delle Ricerche, Rome, Italy.
We thank Ms B. Canova for typing and B. Johnston for editing the manuscript.