Improved Grading of Breast Adenocarcinomas Based on Genomic Instability

Cyclin E, associated with cyclin-dependent kinase 2 during G₁-S-phase transition of the cell cycle, and cyclin A, interacting with cyclin-dependent kinase 2 during S phase, are found to be overexpressed in different types of cancer, e.g., colon, lung and endometrium (1, 2, 3, 4). In human breast tumors and breast cancer cell lines, cyclin E expression is increased at both the protein and mRNA levels (5, 6) and is furthermore encountered to a high extent in undifferentiated species with poor prognosis (7). Cyclin A protein expression is reported as being up-regulated in breast carcinomas and correlates with an unfavorable clinical course (8, 9).

Particularly in the case of cyclin E, a connection with oncogenesis is suggested. Although cyclin E and its associated kinase are under cell cycle control in normal tissue, in tumor cells, it remains in an active complex throughout the cell cycle (10). The constitutive expression of cyclin E may lead to aberrant regulation of S-phase progression (11). This could be a cause of karyotypic instability in addition to others such as the overexpression of centrosome-associated kinase STK15/BTAK or the malfunction of mitotic checkpoint genes (12, 13). In some authors’ opinion, karyotypic instability is directly proportional to the degree of aneuploidy. Thus, the further the distance of a cell’s stemline karyotype from either diploidy or tetraploidy, the greater the instability of its genome (14).

Closely involved in the process of generating aneuploidy seems to be disturbed regulation of the centrosome duplication. The centrosome as the microtubule assembling and organizing cellular compartment is responsible for correct chromosome segregation (15). During G₁-S-phase transition of the cell cycle, centrosome duplication initiates, and cyclins A and E appear to partake in this process. Blocking of either cyclin-dependent kinase 2 or of either cyclin will lead to an abolition of centrosome duplication (16, 17, 18). Throughout G₂-M-phase passage, replicated centrosomes separate and constitute the poles of the mitotic spindle (15, 18). When centrosome duplication and DNA replication are no longer synchronized, e.g., because of cyclin E overexpression, this may lead to interference with centrosome duplication, formation of multipolar mitotic spindles, and impaired distribution of chromosomes at cell division and, thus, to aneuploidy (19, 20, 21). Numerical and qualitative deviations of centrosomes, as well as aneuploidy, are common in human cancers (22, 23).

Aneuploid tumors prove to be more malignant than their diploid counterparts (24, 25), and in general, they are regarded as being genomically unstable (gu; Ref. 13). A type of genomic instability, described as chromosomal instability, was first characterized by Lengauer et al. (26). A tumor cell population with chromosomal instability attribute indicates, by definition, high clonal heterogeneity (27).

In the histograms obtained by us from image cytometry showing the nuclear DNA content of 154 primary breast tumors, we could clearly distinguish two subtypes of aneuploid species (see “Results”), as did previously Baldetorp et al. (28), using other criteria. Our measurements resulted from the detection of a genomically stable (gs) aneuploid type, essentially demonstrating clonal homogeneity and a gu one, demonstrating clonal heterogeneity (Fig. 1, A and B). As stated above, the cell cycle genes cyclin E and cyclin A seem to be closely involved in generation of karyotypic instability because they are crucial for G₁-S-phase transition and centrosome duplication. In this study, we have measured the mRNA levels of cyclin A and cyclin E in primary breast tumors of different DNA ploidy. We additionally determined cyclin A protein expression in these tumors. Chosen by random from those samples with evaluable mRNA levels, we investigated centrosomal aberrations in 10 breast carcinomas of different nuclear DNA content.

A total of 154 consecutive biopsies of breast adenocarcinomas was collected at Cell and Molecular Analysis, Karolinska Hospital, during 1998–1999. There was no radio- or chemotherapy before surgery. From each sample, imprints were taken for DNA ploidy measurements, and 4-μm histological sections were prepared for histopathological assessment and immunohistochemistry. Furthermore, total RNA was extracted from the same tumor sample for real-time quantitative PCR (qPCR) determinations. Additionally, we investigated four benign lesions (3 fibroadenomas and 1 phyllodes adenoma), and for the real-time qPCR, we also included three normal breast tissue samples.

Image cytometry was performed on imprints taken from the biopsies and Feulgen stained to measure the nuclear DNA content of the tumor cells. The staining, internal standardization, and tumor cell selection were based on previously described methods (29). All DNA values were determined in relation to a corresponding staining control, which obtained the value 2c, denoting the diploid DNA content. Our specimens could be divided into the already known groups: type I, with a diploid stemline (1.8c-2.2c), <5% of cells in S phase and no cells exceeding 5c; type II, with a tetraploid stemline (3.8–4.2c) and <5% of the cells exceeding 5c; type III or diploid proliferative, with a diploid stemline, but ≥5% of the cells being in S phase, called diploid proliferative; and type IV or aneuploid, with one peak or more outside the diploid or tetraploid region (29, 30).

Directly after surgery, a part of each tissue sample was fixed in 4% phosphate-buffered formalin and subsequently paraffin embedded. From each specimen, 10 contiguous sections (4 μm) were prepared for immunohistochemistry. All slides were deparaffinized with xylene, rehydrated, and microwave treated at 500 W for 2 × 5 min (1 × 10 min for γ-tubulin staining) in 10 mm sodium citrate buffer (pH 6.0).

After rinsing in Tris-buffered saline (pH 7.6), peroxidase activity was blocked by immersion of the slides in 0.5% hydrogen peroxide in distilled water for 30 min. Unspecific staining was prevented by using 1% BSA in Tris-buffered saline for 20 min, followed by overnight incubation with a monoclonal mouse antibody against human cyclin A protein from Novocastra Laboratories Ltd. (Newcastle-upon-Tyne, United Kingdom). We diluted the stock solution of the antibody 1:100 in 1% (wt/vol) BSA. Two independent investigators evaluated the results, and usually 10 areas of highest staining intensity were analyzed.

The antibody against cyclin E did not work for our purposes.

For γ-tubulin (centrosome) staining, we randomly chose 10 tumors from the 104 breast carcinomas that yielded evaluable real-time qPCR data. Additionally, we stained three fibroadenomas as controls.

After deparaffinizing, rehydration, and microwave treatment (as above), the sections were blocked in PBS, containing 10% normal goat serum, 3% BSA and 0.1% Tween 20 for 1 h. The specimens were then stained with an anti-γ-tubulin monoclonal antibody (Sigma-Aldrich, St. Louis; 1:300 in blocking solution, 1 h), followed by a fluorescent secondary antibody (Alexa Fluor 488; Molecular Probes, Leiden, The Netherlands; 1:200 in blocking solution, 1 h) and counterstained with 4′,6-diamidino 2-phenylindole DNA dye for 5 min. After each antibody incubation, the slides were washed in 0.1% Tween 20/PBS. The staining was evaluated with a fluorescence microscope (Zeiss Axioskop). The percentage of centrosomal aberrations was determined by counting centrosomes (γ-tubulin-stained spots) of increased size, i.e., those with a diameter larger than twice that of centrosomes from nontumor control cells, plus the number of centrosome clusters (more than two γ-tubulin-stained spots situated very closely to each other), and dividing them by the number of investigated cells. As an internal control for centrosome size, we used fibroblasts and normal breast cells in the sections. In malignant tissues, only areas with a change in normal tissue architecture were evaluated. Furthermore, the presence of aberrant mitotic spindles was determined. We examined at least 500 cells/tissue section.

A total of 154 biopsies from primary breast tumors, 4 benign lesions, 3 normal breast tissue samples, and the breast cancer cell line MDA-MB-231 (acquired from American Type Culture Collection) was used for RNA extraction, cDNA synthesis, and real-time qPCR analysis, using amplicons for cyclins A and E and 18S rRNA.

Total RNA was extracted using Trizol reagent, followed by treatment with DNase I (both Invitrogen-Life Technologies, Inc., Paisley, United Kingdom). Ten units of RNasin (Promega, Leiden, The Netherlands) were added to the master mix. After 15 min of incubation at room temperature, we inactivated DNase I in a 2.5 mm (final concentration) EDTA solution, followed by heating at 70°C for 10 min. We reverse transcribed 2 μg of total RNA by following the manufacturer’s instructions. The final reaction volume of 20 μl contained 1× first strand buffer (50 mm Tris-HCl, 75 mm KCl, and 3 mm MgCl₂), 100 ng random hexamers, 0.1 m DTT, 10 mm deoxynucleotide triphosphate mix (Applied Biosystems or ABI, Warrington, United Kingdom), 200 units of Superscript II RNase H⁻ Reverse Transcriptase and diethyl pyrocarbonate-treated water (all other reagents from Invitrogen-Life Technologies, Inc.).

Real-time qPCR enables the measurement of product accumulation during the exponential phase of PCR when all reagents are still abundant and limiting conditions do not interfere with precise quantitation. The instrument’s optic device detects an increase in fluorescent emission that occurs if Taq polymerase performs a 5′-3′-nucleolytic cleavage of the dual-labeled fluorogenic probe, which has to be hybridized to the template. The threshold cycle (C_T) is the fractional cycle number during PCR at which the amount of amplified target in each sample reaches a fixed threshold above background noise (31, 32). The logarithm of initial template copy number results from the measured C_T value in consequence of their negative linear correlation (33).

A validation experiment (34) confirmed comparable amplification efficiencies of the cyclin A, cyclin E, and 18S rRNA amplicons. This is the prerequisite for a relative quantitation without using standard curves. Relative quantitation of cyclin A and E mRNA levels was carried out using the comparative C_T method; 18S rRNA served as an endogenous external control, and the interrun standard was cDNA prepared from MDA-MB-231 cells. All PCR reactions were performed in triplicates on a 7700 Sequence Detector (ABI). From the mean C_T values of the respective triplicates, the mean C_T value of the 18S rRNA amplicon was subtracted to give a ΔC_T value for each sample and for each of the two cyclins. The mean C_T value of the interrun standard was then subtracted from the ΔC_T values to yield ΔΔC_T values for each sample and for each of the two cyclins. The relative amount of target mRNA was calculated as two to the power of -ΔΔC_T (34).

We designed primers and probes for the target genes cyclin A and cyclin E using Primer Express 1.0 (ABI). The probes were synthesized by ABI and the primers by Life Technologies, Inc. The amplicon of 18S rRNA was designed and synthesized by ABI. The sequences of primers and probes (all in 5′-3′ direction) were as follows: cyclin A—forward primer, GAAGACGAGACGGGTTGCA; probe, AGCCTGCGTTCACCATTCATGTGGA; reverse primer, GAGATTCAGCTGGCTTCTTCTG; cyclin E—forward primer, AAATGGCCAAAATCGACAGG; probe, CCCACACTGGTCCCTCGCCG; reverse primer, TGCATTATTGTCCCAAGGCTG; and 18S rRNA—forward primer, CGGCTACCACATCCAAGGAA; probe, TGCTGGCACCAGACTTGCCCT C; reverse primer, GCTGGAATTACCGCGGCT. To exclude amplification from genomic DNA, we situated one primer of each amplicon across an exon/exon junction (31) such that up to 6 bases of the 3′-end of the primer corresponded to one of the exons, and the remaining nucleotides corresponded to the adjoining one.

Each PCR comprised 1× TaqMan Buffer A, 6 mm MgCl₂, 200 μm dATP, dCTP, dGTP, and dTTP, and 0.625 units of AmpliTaq Gold DNA polymerase (all reagents ABI) in a final volume of 25 μl. We used the probes for cyclin A and cyclin E at a concentration of 100 nm and the probe for 18S rRNA at a concentration of 75 nm. Cyclin A and cyclin E primers had concentrations of 200 nm, the forward primer of 18S rRNA had a concentration of 50 nm, and the reverse primer a concentration of 40 nm. Each reaction contained 25 ng of cDNA. Thermal cycling conditions were as follows: 10 min at 95°C, followed by 40 cycles of 95°C, 15 s and 62°C, 1 min. Triplicate nontemplate controls were included for the endogenous control and all targets.

The Excel statistics package (Office 98) was used to perform the majority of the statistical calculations, e.g., the unpaired two-tailed Student’s t test and the Pearson correlation coefficient r. Spearman’s rank-order correlation was carried out with SPSS software and the logistic regression analysis with the R statistical package.⁷

Image cytometric measurements of 154 consecutive biopsies of breast adenocarcinomas resulted in 70 aneuploid (A), 61 diploid (D), and 23 tetraploid (T) tumors, as defined by the modal DNA content of cells from each individual cancer (29). The subjective classification of DNA histograms of all three ploidy categories indicated subtypes that differed clearly in their percentage of cells with nonmodal DNA content values (Fig. 1).

To define an objective measure, which allowed the identification of these subtypes, we added up (a) the percentage of cells with DNA content values in the S-phase region (S phase); (b) the percentage of cells with DNA content values exceeding twice the modal value plus 1c (G₂ exceeding rate or G₂ Exc); and (c) the coefficient of variation (CV) of the respective tumor stemline. The sum of these three DNA histogram variables was called the “stemline scatter index” (SSI), which is a measure of the clonal heterogeneity of the constituent tumor cells. We calculated each of these three values by the Ahrens image program. Logistic regression analysis on all SSI values of the 154 tumor samples showed S phase (z = 3.5, P = 5.3 × 10⁻⁴; Fig. 2) as contributing most to the discriminative strength of the SSI, followed by CV and G₂ Exc (z = 3.0, P = 0.003 and z = 2.1, P = 0.03, respectively; Fig. 2). None of the three summands was found as being sufficiently selective on its own. Using logistic regression, we could also determine the cut point of SSI = 8.8% (P = 0.03) for discrimination of all tumors showing significantly scattered DNA histograms (SSI > 8.8%) from those with insignificantly scattered ones (SSI ≤ 8.8%), independent of their cytophotometrically assessed modal DNA content values. Breast cancers with an SSI > 8.8% were termed gu, and those with an SSI ≤ 8.8% were termed gs. The cut point and the discriminative strength of the three summands of SSI have been validated on a data set of 890 breast cancer cases.⁸

To evaluate whether the mRNAs for cyclin A and E are differentially expressed in the two subtypes of each of the A, D, and T ploidy categories, we performed real-time qPCR on the 154 biopsies from primary breast carcinomas studied above on four biopsies from benign tumors and three biopsies from normal breast tissue. Of the 154 biopsies, 104 gave evaluable results. In the case of the remaining tumors, either the C_T value of 18S rRNA exceeded 23 (a sign of RNA degradation), or the C_T values of the cyclins were higher than 36, both of which render unreliable results (35). As a measure of intra-assay reproducibility, the CVs of triplicate reactions for cyclins A and E for each sample were on an average 2.2 and 1.9%, respectively. Interassay reproducibility, as demonstrated by the interrun standard CV, was 4.7% for cyclin A and 4.9% for cyclin E.

The two aneuploid lesion subtypes could not only be characterized according to DNA histogram and SSI value but also demonstrated statistically highly significant differences in cyclin A and E expression levels between them, according to the Student’s t test (Table 1). The transcript level of cyclin A was found to be four times higher in Agu cases than in Ags cases. Measuring the amount of cyclin E transcripts, we observed a 3-fold difference between the two types of aneuploid carcinomas (Fig. 3). On the other hand, the expression of these cyclins did not vary significantly in Ags versus Dgs tumors (P = 0.5, cyclin A; P = 0.9, cyclin E). The Dgu and Agu lesions also produced comparable expression levels for both cyclins, whereas transcript levels differed significantly in Dgs versus Dgu breast cancers (Table 1, Fig. 3). In the case of T lesions (Tgs and Tgu), cyclin A mRNA, but not cyclin E mRNA, gave a significant difference between the ploidy subtypes (Table 1). The mRNA expression of the two cyclins was ∼10–15 times lower in normal breast tissue and benign lesions than in Dgs tumors (data not shown). The best linear correlation between the mRNA levels of cyclin A and E were obtained from Dgs carcinomas (r = 0.76, P = 1 × 10⁻⁴; Fig. 4), followed by Ags lesions (r = 0.755, P = 2 × 10⁻⁴; Fig. 4). In Agu tumors, the coefficient of correlation was 0.44, with P = 0.009, and in Dgu lesions, we did not measure any linear relationship between the mRNA expression of these cyclins (r = 0.05, P = 0.9). At the protein level, as determined by immunohistochemistry for cyclin A, we found significant differences in the same tumor categories that differed significantly from each other at the mRNA level (Table 1). There was a good linear correlation between protein and mRNA expression of cyclin A in Agu lesions (r = 0.85, P = 1.0 × 10⁻⁴; Fig. 5), whereas this was not observed in Dgu and Tgu lesions (r = 0.16 and 0.1, respectively), nor in the stable ploidy subtypes (Ags, Dgs, and Tgs; r = 0.02, 0.25, and 0.05, respectively).

We also examined the clinicopathological data of this cohort of 104 women to observe how this related to the grouping of patients into those having a tumor containing cells with unstable genomes and an SSI > 8.8% and those having a tumor containing cells with stable genomes and an SSI ≤ 8.8%. The two groups did not differ significantly from each other concerning the clinical stage, nor was there any difference in the percentage of positive lymph nodes between them. The majority of the tumors were ductal (Table 2). Our follow-up data showed 76% of the patients who developed distant metastases and/or died as possessing low differentiated carcinomas. During the follow-up period of 2–3 years (average follow-up interval was 2 years and 5 months) in the group with an SSI > 8.8%, 25.6% of the patients died of breast cancer and 18.6% evolved distant metastases. In the group with an SSI ≤ 8.8%, 2% of the patients died and 4% had distant metastases.

Fifty-four percent of all patients with progression (distant metastases and/or disease-related death) appeared to have high cyclin A and E mRNA values, whereas 31% showed low mRNA levels for each of these two cyclins. A χ² test proved this difference to be significant (P = 0.03). The median cyclin A and E mRNA expression levels served as cutoff values and were 0.46 and 0.28, respectively. Cyclin A protein expression was high (>5% positive cells, general directions of the Department of Oncology-Pathology, Karolinska Hospital) in 62% of all patients with progression; in the remaining 38%, it was low (≤5%). This difference was not significant (P = 0.32).

In addition, we did not observe any relation between cyclin A and E mRNA overexpression and tumor size (r = 0.02, P = 0.8 and r = -0.2, P = 0.1, respectively), suggesting that overexpression of these cyclins may be an early event in cancer development.

In a pilot study, we examined three fibroadenomas, three Dgs and three Ags breast carcinomas and four Agu breast tumors for potential centrosome defects. We chose these samples at random from the 104 breast carcinomas that yielded evaluable real-time qPCR data. The tumor sections were immunostained with an antibody against γ-tubulin, a well-characterized marker of the centrosome (36) localized within the centrioles and the pericentriolar region (37). The aberrations observed included single centrosomes of increased size (diameters larger than twice that of centrosomes from nontumor control cells) and centrosome clusters with more than two γ-tubulin-stained spots situated in close proximity. The fibroadenomas had the lowest percentage of centrosomal aberrations, occurring on average in 0.4% of the cells. They were followed by Dgs and Ags breast carcinomas that exhibited similar frequencies of aberrant centrosomes in 2.6 and 2.4% of their cells, respectively. The Agu tumors showed the most centrosomal abnormalities, on average, in 9.6% of the cells (Table 3). The hypertrophies in Agu lesions proved to be of bigger size than those of the other samples. Furthermore, in three of the four Agu carcinomas, aberrant mitotic spindles could be observed, whereas this was not seen in any of the other cases (Table 3). Our results showed a positive linear correlation between the extent of centrosomal aberrations and the mRNA levels of both cyclin E (r = 0.83, P = 0.01) and cyclin A (r = 0.75, P = 0.01).

Image cytometrical measurement of diploid, tetraploid, and aneuploid primary breast adenocarcinomas resulted in two different subtypes of DNA histograms for each of the three ploidy categories. The histograms were clearly dissimilar in their percentage of cells with nonmodal DNA content values (Fig. 1), manifesting the clonal heterogeneity that is characteristic of cancer (27). By calculating the SSI as an objective measure of the nonmodal part of DNA content values or scattering of DNA histograms and then using a cut point of SSI = 8.8%, we were able to distinguish between lesions with insignificant scattering of the DNA histogram (SSI ≤ 8.8%), termed gs, and lesions with significant scattering (SSI > 8.8%), termed gu. Thus, for instance, an aneuploid stemline and a SSI ≤ 8.8% determined the stable aneuploid subtype of breast cancer. In general, this cut point value has validity independent of the stemline position of the particular tumor.

Aneuploid cancers are regarded as being more malignant than tumors of other ploidy categories (24, 38, 39), and they are characterized by increased genomic instability (13, 40, 41). Our measurements resulted in the observation of an aneuploid subtype of breast cancer that proved to be gs (Figs. 1,A and 2), although its modal DNA content value was, for instance, 2.7, 3.2, or 4.8c.

If cyclin A and/or cyclin E overexpression in combination with centrosome aberrations is diagnostic of genomic instability, then the measurement of these indicators should lead to different results in the two groups of aneuploid tumors. In the gu aneuploid subtype, the amounts of cyclin A and E transcripts were significantly higher than those in the gs aneuploid lesions. Furthermore, expression levels of these two mRNAs were not found to be significantly different in gs aneuploid and diploid tumors. The same was true for these expression levels in gu aneuploid and diploid lesions (Table 1, Fig. 3). Cyclin A protein expression was significantly up-regulated in all unstable ploidy subtypes, compared with the gs subtypes (Table 1).

We also investigated the presence of centrosomal aberrations in some of the ploidy subtypes and measured by far the most defects, in an average of 9.6% of cells, in gu aneuploid tumors. In accordance with the cyclin expression measurements, stable aneuploid and diploid cancers demonstrated comparable and much lower ranges of centrosomal aberrations. On average, 2.4 and 2.6% of the tumor cells, respectively, possessed centrosomal abnormalities (Table 3). Our results suggest greater affinity between tumors with comparable SSI values than between those of the same ploidy category.

A follow-up period of 2–3 years revealed a much better prognosis for patients with tumors having an SSI ≤ 8.8% than for patients with tumors having SSI values above this cut point. It appears to be of minor importance for clinical malignancy grade where the tumor’s stemline is situated but very critical how pronounced the scattering of the tumor’s DNA histogram proves to be as indicator of genomic instability. As described above, stable aneuploid and diploid cancer types show similarities in cyclin E and A expression and also in the degree of centrosome aberration. However, the representation of such tumors in the breast cancer population does deviate. In a data set derived from 890 consecutive primary breast adenocarcinomas, we measured 15% of the aneuploid cases as having SSI values ≤ 8.8%; among the diploid cancers, 62% had a SSI ≤ 8.8% (manuscript in preparation).⁸

We found a linear correlation between the mRNA levels of the cyclins and the extent of centrosomal aberrations. This is consistent with the proposal that cyclins A and E are involved in centrosome duplication (16, 18). However, this correlation has to be verified with a larger set of samples. There was no relation between the overexpression of cyclin A and E mRNAs and tumor size, which may indicate that disturbance of the expression of these cyclins is an early event in carcinogenesis. Aneuploidy and chromosomal instability are also early events in tumor development, as reviewed in Lengauer et al. (27), and furthermore, centrosome aberrations are detectable even in noninvasive breast cancers (23). This coincidence might emphasize connections between the above mentioned cell biological events.

Both Duesberg et al. (14) and Lingle et al. (23) have postulated the existence of stable aneuploid tumors, as characterized by fluorescence in situ hybridization. Duesberg et al. (14) found in cell lines a direct relation between genomic instability and the degree of aneuploidy that predicts relatively stable aneuploidy, with its stemline closely situated to the diploid or tetraploid stemline. These aneuploid cases were supposed not to involve chromosomes encoding mitosis-regulating genes. In contrast, we measured stable aneuploidy far from the diploid or tetraploid stemline peak. Aneuploid breast tumors with, e.g., a modal DNA content value of 3.2c, showed SSI values in the range of gs diploid and tetraploid tumors and proved to have a good prognosis. On the other hand, we detected gu diploid and tetraploid cancers with SSI values in the range of unstable aneuploid ones and an unfavorable prognosis. Lingle et al. (23) also investigated primary breast tumors and, according to their centrosome observations, could distinguish between stable and unstable aneuploid lesions. The unstable specimens exhibited significantly increased centrosome size and number in comparison to stable ones. This is consistent with our measurements where one criterion, in addition to others, to distinguish between gs and unstable aneuploid cancers has been the presence of differing extents of centrosome abnormalities.

We did not confirm yet whether the gs subtypes of the different ploidy categories remain stable with time. However, we have evidence from a 9-year follow-up study showing that the stable subtypes maintain a significantly better prognosis than the unstable ones (manuscript in preparation).⁸ Furthermore, there are studies concerning stable DNA ploidies of breast adenocarcinomas, which have been studied during the history of the disease. Modal DNA content values have been found to be the same in the primary tumor and its corresponding lymph node and distant metastases, occurring many years later (42, 43).

Until now, the occurrence of stable and unstable subtypes of ploidy categories has only been observed in breast cancer. It would be of interest to investigate the existence of this phenomenon in other malignancies.

In conclusion, we here describe a new classification of DNA ploidy profiles, which appears to be more relevant for estimation of the clinical malignancy potential of breast carcinomas than the traditional one. Aneuploid cancers are generally considered to be more dangerous and aggressive than their diploid counterparts. However, ∼15% of aneuploid tumors have a malignancy potential comparable with that found in gs diploid cancers. The findings reported herein may provide support for approaches aimed at improving the prognostic value of ploidy analysis in breast adenocarcinomas.