Correlations Among p53, Her-2/neu, and ras Overexpression and Aneuploidy by Multiparameter Flow Cytometry in Human Breast Cancer: Evidence for a Common Phenotypic Evolutionary Pattern in Infiltrating Ductal Carcinomas¹

Human solid tumors undergo multiple genetic changes as they progress from a near-normal state to aggressive malignancy. Some of these changes may occur in distinctive patterns, and there is reason to believe that some may develop in a preferred sequence. In a previous study, we performed multiple correlated cell-by-cell measurements by multiparameter flow cytometry on primary tumors from patients with breast cancer and found that the presence within a tumor of aneuploid cells that also overexpressed Her-2/neu and ras in the same cells was an adverse prognostic feature (1). We sought to determine whether this pattern of intracellular oncogene protein overexpression was the result of a specific sequence of evolutionary changes. In a subsequent study, we found that the overexpression of EGF receptor and/or the overexpression of Her-2/neu almost always preceded the development of ras overexpression (2). The present study was undertaken to examine the relationship between p53 abnormalities and previously studied intracellular patterns of Her-2/neu overexpression, ras overexpression, and aneuploidy in individual breast cancers.

Numerous studies in human solid tumors, and in breast cancer in particular, have demonstrated relationships between p53 abnormalities(or loss of heterozygosity at 17p) and aneuploidy, between p53 abnormalities and Her-2/neu amplification/overexpression, between Her-2/neu amplification and/or overexpression and aneuploidy, and between Her-2/neu amplification/overexpression and ras overexpression within individual tumors (reviewed in Ref. 3). Loss of wild type p53 function has been found to be associated with distinctive forms of genetic instability that are manifested by the development of gross abnormalities in chromosome number/cell (aneuploidy) and by structural abnormalities involving individual chromosomes, such as chromosomal breakage, deletions of chromosomal material, and gene amplification (3, 4). The Her-2/neu gene is often amplified and overexpressed in human breast cancer (5, 6), particularly in aneuploid tumors (7, 8, 9, 10). If Her-2/neu amplification were a direct or indirect consequence of p53-induced genetic instability, then one might anticipate that the development of p53 abnormalities might accompany or precede the development of Her-2/neu overexpression, aneuploidy, or both. The overexpression of ras protein has been found to be contingent on the loss of wild-type p53 function in a number of experimental systems(reviewed in Ref. 3). Thus, we hypothesized that p53 abnormalities might also precede the development of ras overexpression in human breast cancer.

Our methodological approach to extracting evolutionary sequence information from single tumor samples is based on the technical ability to perform multiple quantitative measurements on each cell by multiparameter flow cytometry. In previous studies, we found that breast cancer precursor populations generally persist in the background during and after the emergence of more advanced clones (those with larger numbers of genetic abnormalities/cell and/or higher levels of expression of individual oncogene proteins; Refs. 1, 2, and 11). Under such conditions, if there is a specific order in which different genetic abnormalities appear, the same early genetic changes among those measured would be found in the cells with the fewest abnormalities. In contrast, one might expect that abnormalities that consistently occur late would almost always be found to have accumulated in the same cells with abnormalities that had occurred early and had persisted.

In this study, multiparameter flow cytometric analyses were supplemented by FISH³ studies on the same tumors in which paired cell-by-cell measurements of a locus-specific fluorescent probe for p53 and an α satellite probe for chromosome 17 were performed to further explore the relationship between p53 abnormalities (manifested by allelic loss) and aneuploidy in the same cells. We also performed paired cell-by-cell measurements of a locus-specific probe for Her-2/neu and an α satellite probe for chromosome 17 in cells from the same tumors to examine the relationship between Her-2/neu gene amplification and aneuploidy in the same cells.

This study was approved by the Institutional Review Board of Allegheny General Hospital (Pittsburgh, PA). Fresh tumor samples were obtained with informed consent from 56 primary breast cancers of patients who underwent surgery at Allegheny General Hospital and other Pittsburgh area hospitals between November 1, 1995 and January 1, 1998. Table 1 summarizes the clinical characteristics of these patients. Four of these patients have had disease recurrences to date.

Freshly obtained breast tumor samples were finely scissor-minced in HBSS (Life Technologies, Inc., Gaithersburg, MD), filtered through 64μm nylon mesh (Small Parts, Miami, FL), and washed once with HBSS. A small aliquot of cells was fixed in methanol:glacial acetic acid(3:1) for FISH studies, and the remaining cells were then fixed either in methanol for DNA analysis or in paraformaldehyde plus methanol for multiparameter analysis, as described previously (12). A small aliquot of paraformaldehyde/methanol-fixed cells was set aside for cytological confirmation of the presence of tumor cells in each sample to provide assurance that diploid cell samples that did not exhibit p53, Her-2/neu, or ras overexpression did,in fact, contain monodispersed tumor cells.

Fluorescein-conjugated monoclonal antibody immunospecific for p53 protein (Clone DO-7) was purchased from Novocastra Laboratories, Ltd.(Newcastle upon Tyne, United Kingdom). Rabbit polyclonal antibody to c-erbB-2, purchased from Cambridge Research Biochemicals (Cambridge,United Kingdom), was used for indirect staining. Phycoerythrin-conjugated goat anti-rabbit IgG (1:20), purchased from Vector (Burlingame, CA), was used as a secondary antibody. Rat monoclonal antibody to human v-H-ras, which recognizes human c-H-ras,K-ras, and N-ras, was purchased from Oncogene Science (Cambridge, MA). This antibody was conjugated with Cy-5 (Amersham Life Sciences, Inc.,Pittsburgh, PA) and used for direct staining as described previously (2). JC 1939, a breast cancer cell line established in our laboratory, was used as a positive staining control for p53, Her-2/neu,and H-ras immunofluorescence and as a quantitative fluorescence staining reference for p53 and Her-2/neu (see below). Lymphocytes from healthy donors were used as low-level baseline immunofluorescence staining controls and as relative reference standards for the ras measurements.

Cells were stained with 4′,6-diamino-2-phenylindole (Sigma Chemical Co., St. Louis, MO) at a final concentration of 0.1 μg/ml. Multiparameter flow cytometry measurements were performed as described previously (2).

Methanol-fixed lymphocytes were used as diploid reference standards for ploidy analysis of methanol-fixed tumor cells (12). The criteria for tetraploidy and aneuploidy were as described previously (2).

Paraformaldehyde/methanol-fixed lymphocytes were used as reference cells for the multiparameter flow cytometry studies. Tumor cell ras levels were expressed as multiples of the baseline normal lymphocyte reference (by analogy to the DNA index). In this study, for convenience of graphical comparison of ras levels to levels of Her-2/neu in the same cells (see below), the normal lymphocyte reference was assigned a nominal reference value of 10,000 units. Thus, for example, 5 × 10⁴ units/cell would correspond to a 5-fold increase in ras content above that of normal lymphocytes, and 1 × 10⁵ ras units/cell would correspond to a 10-fold increase.

Passage 39 of cell line JC 1939 was found to contain 77,000 molecules/cell of HER-2/neu, quantitated by ELISA assay (Oncogene Science; Ref. 2), and a mean of 7,500 p53 molecules/cell, also quantitated by ELISA assay (Oncogene Research Products, Cambridge, MA). On the basis of this reference value, the mean p53 content in 23 samples of normal lymphocytes was 6600 molecules/cell, in general agreement with values reported by others for normal blood leukocytes (13, 14).

Using a computer program developed by one of the authors (S. E. S.),the logarithmic data for all three of the oncogene measurements per cell were scaled to a common origin. A subtractive correction for nonspecific antibody binding was applied to each measurement on a cell-by-cell basis, as described previously (2). In all data figures shown in this paper, nonspecific labeling has already been subtracted from each measurement on a cell-by-cell basis.

Tumor cells in single-cell suspension freshly fixed in methanol:glacial acetic acid (3:1) were applied to slides and processed in accordance with protocols developed by Vysis, Inc. (Downer’s Grove, IL) using the HYBrite hybridization apparatus (Vysis). Combinations of gene locus-specific and centromeric enumeration probes used in this study were Her-2/neu/chromosome 17 and p53/chromosome 17, obtained from Vysis. After hybridization, cells were counterstained with 4′,6-diamino-2-phenylindole (50 μg/ml), and slides were viewed on an Axiophot-II microscope (Carl Ziess, Inc). Standardized criteria for spot counting were used as recommended by Vysis. At least 50 cells on each slide were counted by each of two independent observers. Staining quality assurance controls consisting of normal lymphocytes were included in every staining batch for each probe set. Each bivariate data set was corrected for staining background using a bivariate template derived from at least five slides of normal lymphocytes, where at least 100 normal cells were counted on each slide for each probe set.

Brotheric et al. (13) estimated the staining intensity threshold for overexpression of p53 to be in excess of 9,000 molecules/cell for antibody DO-7 by flow cytometry. In this study, we used a staining intensity threshold for p53 overexpression of 10,000 molecules/cell.

Individual cells that overexpress Her-2/neu are generally detectable immunohistochemically when intracellular levels are in the range of 200,000–500,000 molecules/cell or higher (3). We adopted a mean level of Her-2/neu overexpression of >300,000 molecules/cell as a staining intensity threshold for overexpression to relate our studies to immunohistochemical studies. However, it was also apparent that flow cytometric measurements were capable of detecting Her-2/neu levels that were well below 300,000 molecules/cell but which were clearly higher than normal (∼50,000 molecules/cell). Therefore, in this study, we used both a >150,000 mean molecules/cell threshold and a >300,000 mean molecules/cell threshold. We view the former threshold as more inclusive for true overexpression and the latter as more appropriate for comparisons with immunohistochemical studies.

A relative mean staining intensity per cell at least four times higher than that of normal lymphocytes (i.e., 4 × 10⁴ units) was used as a threshold level for overexpression. This value exceeded the mean for normal lymphocytes(relative to a fluorescent bead reference) by more than 2 SDs.

For the FISH studies, allelic loss was defined by the presence of fewer p53 alleles/cell than chromosome 17 centromeres in the same cells, with a frequency threshold requirement that at least 15% of cells exhibit allelic loss, based on the studies of Kibbalaar et al. (15). True Her-2/neu gene amplification was defined by the presence of an excess of Her-2/neu gene loci over the number of chromosome 17 centromeres in the same cells. Absolute increases in the number of Her-2/neu loci/cell that were matched or exceeded by the number of chromosome 17 centromeres in the same cells may have been attributable solely to aneusomy and were treated separately. Frequency threshold requirements were similar to those for p53 allelic loss.

For each tumor sample analyzed, the cells were grouped by the number of abnormalities they contained. If these abnormalities had developed in an orderly sequence of inherited clonal changes, and if the precursor clonal subpopulations had persisted in the presence of more evolved clones, then the earliest abnormal change(s) would be identified among the cells that contain a single abnormality (Fig. 1,A). Once the earliest abnormality is identified, the next abnormal change in the sequence can then be deduced from the patterns observed in cells with two abnormalities, and the third abnormality in the sequence can be deduced, in turn, from the patterns observed in cells with three abnormalities. In the hypothetical example shown in Fig. 1,A,almost all cells with a single abnormality contained abnormality B;hence, among the abnormalities studied, B is likely to have occurred first. In this hypothetical example, among cells that contained two and only two of the abnormalities measured, almost all contained abnormalities A and B in the same cells. Having established that B occurred first, abnormality A must have occurred after abnormality B. In this example, because cells containing abnormality C almost always contained abnormalities B and A as well, the sequence of accumulation of all of these abnormalities must have been B → A → C. Illustrative examples based on actual data are given below in Figs. 3 and 4 and Tables 2 and 3.

It should be noted that the establishment of a regular sequence of occurrence for a group of measured abnormalities does not necessarily imply direct causal relationships among the members of the sequence,nor does it preclude the occurrence of other, unmeasured abnormalities within the sequence.

If precursor subpopulations had not persisted in sufficient numbers to be detected, then this approach to genetic evolutionary sequencing would break down (Fig. 1 B). However, in the present study,among 153 subpopulations with two or more abnormalities/cell that were detected among the 56 breast cancers (diploid and aneuploid cell subpopulations considered separately), at least one potential immediate precursor (i.e., a subpopulation with one abnormality fewer)was also present in the same tumor in all but one instance.

The presence of more than one candidate precursor subpopulation (Fig. 1,C) is not only possible but was actually quite common. Other variants included parallel pathways that converge after multiple steps. However, if all potential precursor populations are present in the same tumor, then this approach to genetic evolutionary sequencing would provide no useful information (Fig. 1 D). When the event frequency threshold for identifying a potential precursor populations was set at ≥1% of all cells analyzed in a given sample, well-defined evolutionary sequences could be discerned in all but one of the 56 tumors analyzed, and in that one case, an event frequency threshold of 1% of cells may have been too low.

Means of two groups were compared using Student’s t test. Associations between variables were assessed by the χ²test or by Fisher’s exact test, as appropriate.

An elevated overall sample mean level of p53 protein/cell (mean,>10,000 molecules/cell) was observed in 21 of 56 or 37.5% of samples,which is within the range reported in most studies using immunohistochemical techniques (6, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26). This is likely to be an underestimate of the proportion of tumors with p53 overexpression, because the presence of variable proportions of normal diploid cells would tend to lower the overall mean level of p53/cell in each sample.

To reduce the potential effects of dilution by normal diploid cells,mean p53 levels/cell were calculated separately for the diploid and aneuploid components of the aneuploid tumors. The results are shown in Fig. 2 A, together with p53 levels/cell in the purely diploid tumors. Mean p53 levels/cell were in the normal range in the majority of diploid tumors and in most of the diploid components of aneuploid tumors. In every aneuploid tumor, the aneuploid component exhibited higher mean levels of p53/cell than the diploid component, even when neither component exhibited a mean p53 level that exceeded 10,000 molecules/cell. When the presence of a separate aneuploid component with an elevated mean p53 level was considered in classifying tumors with respect to p53 status, the total number of samples exhibiting p53 overexpression increased to 29 of 56, or 51.8% of cases.

Increased mean p53 levels/cell were observed in some diploid tumors and in the diploid components of some aneuploid tumors, suggesting that although the transition to aneuploidy may be associated with an increase in intracellular p53 protein, the development of p53 abnormalities can occur prior to, and/or independently of, the development of aneuploidy. Because diploid tumors were not subdivided further in this study, true mean p53 levels/cell in diploid tumors are likely to have been higher than those observed because of the dilutional effect of normal diploid cells.

Among the 46 cases in which FISH studies involving the p53locus were performed, 26 showed allelic loss of p53 by FISH (56.5%). Fifteen of these 26 tumors also showed p53 overexpression by flow cytometry. In 11 cases, allelic loss of p53 by FISH was the only p53 abnormality noted. Overall, 37 of 46, or 80.4%, of tumors studied by both flow cytometry and FISH exhibited p53 abnormalities by either or both techniques.

Nineteen of the 26 tumors with allelic loss by FISH contained cells that were either monosomic for chromosome 17 or were disomic for chromosome 17 but exhibited only one p53 allele. Among the 19 cases that contained cells that exhibited disomy and p53allelic loss or monosomy for chromosome 17 in the same cells, 9 were diploid by flow cytometry. In six of these nine diploid tumors, p53 protein levels by flow cytometry were in the normal range, and allelic loss was the only p53 abnormality detected. Because hemizygosity for wild-type p53 has been reported to be associated with increased susceptibility to oncogenic transformation (27),p53 abnormalities that first became apparent in diploid cells in the present study may have been of clinical consequence.

FISH studies were performed on 17 of 23 diploid tumors. Nine of these 17 diploid tumors exhibited allelic loss of p53 (52.9%). All nine contained cells that exhibited disomy and p53allelic loss or monosomy for chromosome 17 in the same cells. Overall, p53 allelic loss, p53 protein overexpression, or both were documented in a total of 13 of 17, or 76.5%, of the diploid tumors studied by both techniques.

FISH studies were performed on 28 of 33 aneuploid tumors. Among these 28 tumors, 20 exhibited p53 overexpression (71.4%), and 17 exhibited p53 allelic loss (60.7%). Five of these 28 tumors exhibited p53 allelic loss as the only p53 abnormality, and 3 of the 28 tumors had no demonstrable p53 abnormalities at all. Overall, 25 of 28 aneuploid tumors (89.3%) studied by both flow cytometry and FISH exhibited p53 abnormalities by either or both techniques. The difference in frequency of p53 abnormalities between diploid and aneuploid tumors was not statistically significant(P = 0.399 by Fisher’s exact test).

Nineteen of 56 tumors (33.9%) had mean overall levels of Her-2/neu protein/cell that were considered to be potentially detectable by immunohistochemistry (>300,000 molecules/cell), a value that is within the range of reported frequencies of Her-2/neu overexpression in breast cancer (28). Overall mean levels of Her-2/neu/cell exceeding 150,000 molecules/cell were observed in 26 of 56, or 46.4%,of breast cancers. The higher apparent frequency reflects the greater sensitivity of detection of moderate increases in Her-2/neu expression by flow cytometry in comparison with immunohistochemical techniques. Even this value is still likely to underestimate the proportion of tumors with Her-2/neu overexpression because of the presence of normal diploid cells that have low intracellular levels of Her-2/neu.

To minimize the effect attributable to the presence of normal diploid cells in aneuploid tumors, mean Her-2/neu levels were calculated separately for the diploid and aneuploid components. The results are shown in Fig. 2 B, together with Her-2/neu levels in the purely diploid tumors. When the presence of a separate aneuploid component with an elevated Her-2/neu level was considered in classifying tumors with respect to Her-2/neu status, the total number of samples exhibiting Her-2/neu overexpression (>150,000 molecules/cell) increased to 31 of 56 tumors, or 55.4% of cases.

It is apparent from Fig. 2 B that high mean levels of Her-2/neu/cell were observed both among diploid tumors and among the aneuploid components of aneuploid tumors. The aneuploid component of each aneuploid tumor invariably exhibited higher mean levels of Her-2/neu/cell than the diploid component of the same tumor, even when neither component exhibited a mean Her-2/neu level that exceeded 150,000 molecules/cell. The diploid components of aneuploid tumors appear to fall into two easily separable groups, i.e., those with mean levels of Her-2/neu that were <150,000 molecules/cell and those with mean levels of Her-2/neu that were >150,000 molecules/cell,suggesting that in aneuploid tumors, the development of Her-2/neu overexpression can precede the development of aneuploidy.

True Her-2/neu gene amplification by FISH (where the number of gene loci/cell exceeded the number of chromosome 17 centromeres in each cell) was observed in 14 of 46 tumors in which Her-2/neu FISH studies were performed (30.4%). Concomitant Her-2/neu gene amplification and protein overexpression (>150,000 molecules/cell)were present in the same tumor in 9 of these 14 cases (64%). Six of these tumors exhibited mean levels of Her-2/neu/cell that exceeded 300,000 molecules/cell. Two of the tumors with Her-2/neugene amplification by FISH but no protein overexpression had unequivocally low/normal mean levels of Her-2/neu protein/cell (mean,<50,000 molecules/cell). Twelve of the 14 tumors that exhibited true Her-2/neu amplification were aneuploid (85.7%), and 12 of these 14 tumors exhibited p53 allelic loss, p53 overexpression, or both.

An absolute increase in Her-2/neu gene copy number/cell that was matched or exceeded by chromosome 17 centromere copy number in the same cells was observed in 12 additional tumors. Concomitant Her-2/neu protein overexpression was present in 6 of these 12 tumors (50%). These 6 tumors, together with the 9 tumors with true Her-2/neu amplification and mean levels of Her-2/neu overexpression exceeding 150,000 molecules/cell, accounted for only 65% of the 23 tumors with Her-2/neu overexpression in which FISH studies were performed,suggesting that transcriptional and/or posttranscriptional regulatory mechanisms, rather than increased gene dosage per se, may have played a role in increasing the intracellular levels of Her-2/neu in at least some of the tumors with Her-2/neu protein overexpression.

Elevated mean levels of ras/cell (mean, >40,000 units/cell) were observed in 14 of 56, or 25%, of tumors, in keeping with our earlier studies (2). This is likely to be an underestimate of the true proportion of tumors with ras overexpression because of the artifactual effects of normal diploid cells with low intracellular levels of ras. Hence, for aneuploid tumors, mean ras levels were calculated separately for the diploid and aneuploid components. The results are shown in Fig. 2,C, together with mean ras levels in the purely diploid tumors. When the presence of a separate aneuploid component with an elevated mean ras level was considered, the total number of samples exhibiting ras overexpression increased to 25 of 56 tumors, or 44.6% of cases. As Fig. 2 C shows, most of the aneuploid components of aneuploid tumors exhibited high mean levels of ras/cell, whereas most of the diploid components did not. However, a statistically significant relationship between ras overexpression and aneuploidy was not supported by χ² analysis(P > 0.09).

There was an association between mean p53 level/cell and mean ras level/cell in the same tumor. (In aneuploid tumors, the aneuploid component was used in the analysis.) Among 29 tumors with elevated mean p53 levels/cell, 19 tumors also had elevated mean ras levels/cell(>40,000 units/cell). Among 27 tumors in which mean p53 levels were not elevated, only 7 had mean ras levels/cell that exceeded 40,000 units/cell. This association was highly significant by χ²analysis (P = 0.007).

There was a strong association between mean Her-2/neu level/cell and mean ras level/cell in the same tumor (in aneuploid tumors the aneuploid component was used in the analysis). Among 31 tumors with elevated mean Her-2/neu levels/cell, 21 tumors also had elevated mean ras levels/cell; among 25 tumors in which mean Her-2/neu levels were not elevated, only 4 had mean ras levels/cell that exceeded 40,000 units/cell. The strength of this association was supported byχ ² analysis (P < 0.0002).

The foregoing analyses, which treated mean p53 overexpression, mean Her-2/neu overexpression, and mean ras overexpression as uncorrelated measurements, showed that there was an intricate pattern of interrelationships among them. To examine these interrelationships in greater detail, we took advantage of the fact that all of these measurements were made simultaneously on each of a large number of cells in each tumor and were, therefore, correlated on a cell-by-cell basis. Our approach was designed to explore the possibility that there may be preferred sequences in which these abnormalities developed in individual tumors, based on the premises that early changes can appear alone in individual cells, whereas late changes are likely to be accompanied by persistent early changes in the same cells.

Subpopulations of cells that simultaneously overexpressed p53,Her-2/neu, and ras in the same cells and represented 1% or more of the cells in a given tumor were detected in 45 of 56, or 80.4%, of the breast cancers studied. In most of these 45 tumors, the triple-positive cell subpopulations represented at least 5% of the cells present(1–1.9% of the cells in 6 of 45 tumors, 2–4.9% of the cells in 8 of 46 tumors, 5–9.9% of the cells in 13 of 46 tumors, 10–19.9% of the cells in 13 of 46 tumors, and ≥20% of the cells in 5 of 46 tumors). Triple-positive subpopulations were common among both diploid and aneuploid tumors (19 of 23, or 82.6%, and 26 of 33, or 78.8%,respectively).

Among the 45 tumors that contained triple protein-overexpressing cell subpopulations, 31 contained precursor populations that supported evolutionary sequences in which p53 overexpression was potentially the first abnormality to occur among those studied. In 13 of 31 cases,sequences in which p53 overexpression was the first detectable abnormality were the only sequences supported by the data. Data from one such tumor are presented in Fig. 3and Table 2.

Among the 45 tumors that contained triple protein-overexpressing cell subpopulations, 36 contained precursor populations that supported genetic evolutionary sequences in which Her-2/neu overexpression was potentially the first abnormality to occur among those studied. In 16 of 36 cases, sequences in which Her-2/neu overexpression was the first detectable abnormality were the only sequences supported by the flow cytometry data. However, among these 16 tumors, there were 10 for which p53 FISH data were available, and 4 of the 10 exhibited p53allelic loss in cells that were monosomic or disomic for chromosome 17. This suggests that p53 abnormalities were present in what may have been diploid cells, and that these abnormalities actually may have preceded,or occurred concomitantly with, Her-2/neu overexpression in at least 40% of tumors in which Her-2/neu overexpression might otherwise appear to have occurred first. The example shown in Fig. 4 and Table 3 illustrates a tumor in which p53 overexpression appears to have played a minor role in the early stages of its genetic evolution but in which early p53allelic loss was a prominent feature.

The overexpression of ras in more than 1% of the tumor cells was observed in 51 of 56 tumors. In 38 of these 51 tumors, ras overexpression was always accompanied by the overexpression of at least one other protein in the same cells. In 13 of these 38 tumors, ras overexpression first appeared in cells with p53 overexpression, in 18 of 38 tumors ras overexpression first appeared in cells with Her-2/neu overexpression, and in 7 of 38 tumors, ras overexpression was always accompanied by both p53 and Her-2/neu overexpression in the same cells. Only 13 of 51 tumors contained precursor populations that supported genetic evolutionary sequences in which ras overexpression was potentially the first detectable abnormality among those studied. Among the 13 tumors in which ras overexpression was one of the first detectable abnormalities, there were only four tumors in which sequences with ras first were the only sequences supported by the data;p53 FISH data were available in three of these four cases, and in two of the three tumors, p53 allelic loss was present. Taken together, these findings suggest that ras overexpression generally occurs late in the evolutionary sequence. This is readily apparent in the examples shown in Figs. 3 and 4 and Tables 2 and 3.

We compared the levels of overexpression of p53 protein in cells that overexpressed p53 alone with the levels of p53 protein overexpression in cell subpopulations from the same tumor that overexpressed p53 plus Her-2/neu, ras, or both. We reasoned that if intracellular p53 accumulation conferred no selective evolutionary advantage, then there would be no basis for expecting systematic differences in mean p53 levels/cell between cells that overexpressed p53 alone and cells from the same tumor that were further along in their evolutionary development. The data are shown in Fig. 5.

Among 31 tumors that contained cells with p53 overexpression alone, 15 also contained cell subpopulations that overexpressed p53 plus ras in the same cells. In all cases, the mean number of p53 molecules/cell was higher in cells that overexpressed p53 plus ras than in cells that overexpressed p53 alone in the same tumor. Furthermore, in 13 of these 15 tumors, the mean number of p53 molecules/cell was higher in cells that overexpressed p53, ras, and Her-2/neu than in cells that overexpressed only p53 and ras in the same tumor.

Among the 31 tumors that contained cells with p53 overexpression alone,16 also contained more advanced cell subpopulations that overexpressed p53 plus Her-2/neu in the same cells. In 12 of these 16 tumors, the mean number of p53 molecules/cell was higher in cells that overexpressed p53 plus Her-2/neu than in cells that overexpressed p53 alone in the same tumor. More striking increases in the mean number of p53 molecules/cell were seen in cells that overexpressed p53,Her-2/neu, and ras than in cells that overexpressed only p53 and Her-2/neu. Among the 14 of 16 tumors that contained cells that overexpressed all three proteins, the mean intracellular levels of p53 were higher among triple protein-overexpressing cells than among cells overexpressing only p53 and Her-2/neu in the same tumor.

The progressive increase in mean p53 levels/cell in cells that acquire increasing numbers of additional oncogene abnormalities, and particularly in cells that develop ras protein overexpression, suggests that the cells with high p53 protein levels may be favored with a survival advantage during the course of tumor evolution. Similarly,mean levels of Her-2/neu/cell increased progressively as Her-2/neu-overexpressing cells acquired increasing numbers of additional abnormalities (data not shown).

The majority of tumors in this study (43 of 56, or 76.8%) were pure infiltrating ductal cancers. Of these, 30 of 43 were aneuploid (70%). Thirty-one tumors (72.1%) exhibited early p53 abnormalities(p53 allelic loss in cells disomic or monosomic for chromosome 17 and/or overexpression of p53 protein alone). Thirty-six infiltrating ductal tumors (83.7%) contained subpopulations of cells that simultaneously overexpressed all three oncogene proteins. In 23 of 43 cases (53.5%), the triple-overexpressing cells represented >5% of the cells in the tumor.

Six tumors were classified as pure infiltrating lobular cancers. This group of tumors appears to have followed genetic evolutionary pathways that may have differed from those of most infiltrating ductal carcinomas. All six lobular breast cancers were diploid. All were positive for estrogen receptor, as compared with 27 of 43 (62.8%) pure infiltrating ductal carcinomas. Although this difference is not statistically significant (P = 0.069 byχ ² analysis), the number of lobular breast cancers analyzed was small. Three of the six lobular tumors contained cells that simultaneously overexpressed all three proteins; but in two of the three tumors, the triple-overexpressing subpopulation represented <5%of the cells in each tumor. In view of numerous reports that Her-2/neu overexpression is infrequent in lobular breast cancers, it is somewhat surprising that Her-2/neu overexpression was so common in the lobular tumors studied here. The greater sensitivity of flow cytometry over immunohistochemistry in detecting Her-2/neu overexpression may account in part for this discrepancy. Her-2/neu FISH data were available in four of the six pure lobular tumors, and none of the four exhibited Her-2/neu gene amplification. p53 FISH data were available for five of the six pure lobular tumors, and only one of the five exhibited p53 allelic loss.

Two tumors contained lobular breast cancer together with other histopathological elements. In both tumors, the patterns of ploidy and oncogene overexpression appeared to reflect the features of the nonlobular component. There were two pure mucinous (colloid) breast cancers among the tumors studied here. Both exhibited aneuploidy, early p53 abnormalities, and a subpopulation of triple protein-overexpressing cells that represented >5% of the cells in each tumor, features that were also characteristic of infiltrating ductal breast cancers. An additional tumor that contained both mucinous breast cancer and infiltrating ductal carcinoma exhibited similar features. A tubular breast cancer and a metaplastic carcinoma with squamous cell features were both diploid, and each contained a small triple protein-overexpressing cell subpopulation.

We have shown that p53 protein overexpression, Her-2/neu overexpression, and ras overexpression often occur together in the same tumor cells in various combinations in most breast cancers. Multiparameter flow cytometry studies demonstrated that it is uncommon for ras protein overexpression to occur in cells that do not contain p53 abnormalities and/or overexpress Her-2/neu protein. In contrast,simultaneous p53 protein and Her-2/neu protein overexpression occurs frequently in the absence of ras overexpression in the same cells,suggesting that they are likely to have developed earlier than the ras overexpression that is present in the triple-positive cells in same tumors. Overexpression of all three proteins, together with aneuploidy, were found to be prominent features of infiltrating ductal carcinomas but not of lobular breast cancers.

Supplementary FISH studies showed that tumors often contain cell subpopulations that exhibit p53 allelic loss by FISH in the absence of p53 overexpression, and that allelic loss of p53 without p53 protein overexpression often accompanies or may precede the development of Her-2/neu abnormalities. A comparison of FISH studies with flow cytometric findings in the same tumors also indicates that overexpression of Her-2/neu can occur in the absence of demonstrable Her-2/neu gene amplification. Taken together, these findings suggest that there is a common molecular phenotypic evolutionary pattern that is associated with many infiltrating ductal carcinomas of the breast, and that this pattern can develop by a variety of molecular genetic mechanisms that have equivalent or overlapping phenotypic consequences.

The flow cytometric methodology used in this study did not permit direct correlations between the cell-based measurements that were performed and the morphological appearance of the cells that were studied. Instead, the observed intracellular patterns of the abnormalities themselves were used to characterize the various cell subpopulations present in each tumor sample. In diploid samples in which none of the cells exhibited overexpression of the oncogenes studied, the presence of tumor cells in the monodispersed cell suspensions was confirmed by independent cytological examination. Direct correlations between morphological appearance and quantitative phenotypic measurements on a cell-by-cell basis must await the application of quantitative multiparameter fluorescence imaging technologies.

A relationship between the overexpression of p53, Her-2/neu, and ras proteins and the development of gross genetic abnormalities in the same cells is established by the finding that the levels of expression of each of these proteins were invariably higher in the aneuploid components of aneuploid tumors than in the diploid components of the same tumors (Fig. 2). However, the finding that levels of each protein increased progressively with the acquisition of new abnormalities (Fig. 5 and associated discussion) suggests that this progressive overexpression was the result of multiple events, some of which may have been attributable to effects of genes coding directly for the specific proteins involved, and others of which may have resulted from indirect effects, including the possibility of perpetuated epigenetic changes.

The relationship between p53 protein overexpression and the presence of genetic abnormalities that might affect p53 function directly or indirectly is complex. Wild-type p53 protein has a short half-life and does not ordinarily accumulate in normal cells in the absence of DNA damage. In experimental cell systems, sustained overexpression of the wild-type p53 protein commonly leads to cell cycle arrest, apoptosis,and/or inhibition or reversion of the transformed phenotype. In human tumors, p53 overexpression is commonly associated with missense mutations (29). The accumulation of mutant p53 protein in the cell nucleus is thought to be attributable to the failure of the mutant protein to induce MDM2 (30, 31), a regulatory protein that normally binds p53, inactivates it, and targets it for ubi- quitination and subsequent destruction.

p53 protein overexpression is not always associated with p53gene mutations. Wild-type p53 protein can also accumulate in human tumors and, like mutant p53 protein, is detected by the DO-7 antibody to p53 used in this study. When wild-type p53 protein is overexpressed in human tumors, it is thought to be maintained in a functionally inactive state by any of several possible mechanisms. In up to one-third of human breast cancers, wild-type p53 is excluded from the nucleus and accumulates in the cytoplasm (32), where it is transcriptionally inactive. Cytoplasmic p53 overexpression is associated with an adverse clinical outcome in breast cancer, even in the absence of nuclear overexpression (33). In the nuclear compartment, the level of MDM2 and the level of wild-type p53 are normally regulated mutually and reciprocally (30, 34, 35). However, the MDM2/p53 regulatory loop is modulated by additional factors, including the states of phosphorylation of the MDM2 and p53 proteins, and the level of p14^ARF, which regulates the level of MDM2, in turn (35). Simultaneous overexpression of both MDM2 and wild-type p53 proteins has been documented in a variety of human solid tumors, including a small proportion of breast cancers (36), which may reflect simultaneous inactivation of wild-type p53 by MDM2 binding and impairment of the ubiquitin pathway that leads to p53 degradation (30). Thus, direct or indirect interference with normal p53 transcriptional activity would appear to be a common underlying phenotypic link among the several known genetic mechanisms that can lead to wild-type or mutant p53 protein overexpression.

Abrogation of wild-type p53 function is not always accompanied by mutant or wild-type p53 protein overexpression. Frame shift and stop codon mutations, which represent ∼15% of all p53 mutations, often result in the absence or near absence of intracellular p53 protein. In the present study, 50% of all tumors were found to overexpress p53 protein, using sensitive flow cytometric techniques to detect p53 overexpression. With the addition of FISH studies to detect p53 allelic loss, p53 abnormalities were identified in 80.4% of all tumors studied by both techniques, suggesting that a substantial number of p53 abnormalities might be missed if one were to rely on p53 overexpression alone.

With regard to the molecular mechanisms underlying Her-2/neu overexpression, our FISH studies indicated that gene dosage effects attributable to gene amplification and/or increased chromosome copy number/cell could account for overexpression in ∼65% of cases. Interphase FISH techniques rely on the identification of multiple,spatially separated gene loci to identify gene amplification. It is conceivable that contiguous amplified loci, such as HSR regions,might be missed in FISH studies of interphase cells, possibly resulting in underestimation of gene amplification by this technique. The high level of sensitivity of flow cytometric techniques might also result in the categorization of tumors with moderately elevated levels of Her-2/neu expression as overexpressors, which might have been categorized otherwise in immunohistochemical studies. Apart from technical considerations, levels of Her-2/neu expression are known to be impacted by factors other than gene amplification, including down-regulation by estrogen through estrogen receptor-mediated pathways (37, 38) and down-regulation by Rb (39),either or both of which can be inactivated or lost in breast cancer.

Overexpression of ras is common in breast cancer, but ras is rarely mutated in this disease (40, 41, 42). Studies in experimental tumor systems have suggested that increased levels of the ras protein may be more closely associated with tumorigenicity, rapid growth rate,and acquisition of metastatic potential than the presence or absence of a mutated ras gene per se (reviewed in Ref. 3),suggesting that this phenotypic measurement may be a more reliable indicator of a hyperfunctional ras state.

There is substantial evidence linking the development of both gross numerical chromosomal changes and structural chromosomal abnormalities to the abrogation of wild-type p53 function (reviewed in Refs. 3, and 4). Although both types of chromosomal abnormalities often occur together, they probably arise by different p53-related mechanisms. Tetraploidy, an early step in the development of aneuploidy, has been attributed to the failure of p53-deficient or p53-inactivated cells to arrest at a mitotic checkpoint (43, 44, 45) and/or to the development of centrosome abnormalities (46). An association between p53 abnormalities and the development of structural genetic abnormalities, particularly in aneuploid cells, has been observed in a number of experimental studies (47, 48). Of special relevance to the present study, loss of wild-type p53 function has been found to be associated with increased levels of gene amplification, which has often been used as a marker for genetic instability in the experimental setting, as well as the development of aneuploidy (49). In the present study, p53 abnormalities commonly preceded or occurred simultaneously with Her-2/neu overexpression; p53 abnormalities were present in 12 of the 14 tumors that exhibited Her-2/neu amplification by FISH, and 12 of these tumors were aneuploid. These findings are consistent with experimental studies linking p53 abnormalities with genetic instability that may be manifested by gene amplification and/or aneuploidy in the clinical setting. However, one cannot conclude from the sequence information provided by our studies that p53 abnormalities are causally linked to Her-2/neu amplification, nor do our studies rule out other potential sources of genetic instability that might lead to Her-2/neu amplification and/or aneuploidy (50).

The frequent association of p53 abnormalities with ras overexpression in infiltrating ductal carcinomas is of particular interest, in view of the well-documented cooperation between mutant p53 and ras abnormalities in the transformation of normal cells in experimental cell systems (51, 52, 53, 54). Cooperativity might account for the statistically significant association between these two abnormalities,but it would not explain why ras overexpression is almost always accompanied by p53 overexpression in the same cells, whereas p53 overexpression can occur alone in individual cells in the same tumor. There is mounting experimental evidence that ras-mediated mitogenic signaling can be antagonized by p53-mediated apoptotic and growth-inhibitory responses (3). It has been reported recently that oncogenic ras transfected into mouse embryo fibroblasts induces an increase in p53 protein expression and a reduction in cell proliferative activity by a p19^ARF-mediated pathway that is separate and distinct from the DNA damage-induced p53 response pathway (which does not involve p19^ARF; Ref. 55). Our findings support the premise that for a strategy for neoplastic transformation that relies on sustained mitogenic signaling to succeed, prior abrogation of wild-type p53 function (with or without p53 protein overexpression) might be required. The present study suggests that both elements of this strategy may be embraced by many, and perhaps most, infiltrating ductal carcinomas.

Lobular breast cancers appear to pursue a different course from that followed by infiltrating ductal carcinomas. Our findings that lobular carcinomas are generally diploid are in keeping with other published studies (56), as are our findings that p53 abnormalities (16, 57, 58, 59), Her-2/neu amplification (60),and loss of estrogen receptor (58, 61) are not prominent features, even in advanced stages of disease. The present study and our previously published multiparameter study in human breast cancer (2) suggest that both epidermal growth factor receptor overexpression and normal to moderately elevated levels of Her-2/neu expression are more characteristic features of lobular breast tumors.