Tumor grade is an established indicator of breast cancer outcome, although considerable heterogeneity exists even within-grade. Around 25% of grade III invasive ductal breast carcinomas are associated with a “basal” phenotype, and these tumors are reported to be a distinct subgroup. We have investigated whether this group of breast cancers has a distinguishing pattern of genetic alterations and which of these may relate to the different clinical outcome of these patients. We performed comparative genomic hybridization (CGH) analysis on 43 grade III invasive ductal breast carcinomas positive for basal cytokeratin 14, as well as 43 grade- and age-matched CK14-negative controls, all with up to 25 years (median, 7 years) of clinical follow-up. Significant differences in CGH alterations were seen between the two groups in terms of mean number of changes (CK14+ve − 6.5, CK14−ve − 10.3; P = 0.0012) and types of alterations at chromosomes 4q, 7q, 8q, 9p, 13q, 16p, 17p, 17q, 19p, 19q, 20p, 20q and Xp. Supervised and unsupervised algorithms separated the two groups on CGH data alone with 76% and 74% accuracy, respectively. Hierarchical clustering revealed distinct subgroups, one of which contained 18 (42%) of the CK14+ve tumors. This subgroup had significantly shorter overall survival (P = 0.0414) than other grade III tumors, regardless of CK14 status, and was an independent prognostic marker (P = 0.031). These data provide evidence that the “basal” phenotype on its own does not convey a poor prognosis. Basal tumors are also heterogeneous with only a subset, identifiable by pattern of genetic alterations, exhibiting a shorter overall survival. Robust characterization of this basal group is necessary if it is to have a major impact on management of patients with breast cancer.

Breast cancer is a heterogeneous disease with a disparate variety of histological types and a wide spectrum of responsiveness to different treatments, making clinical management difficult. The majority of breast carcinomas fall into the category of invasive ductal carcinoma, no special type, for which histological grade is one of the best predictors of behavior. Poorly differentiated grade III invasive ductal carcinomas are strongly associated with shorter recurrence-free and overall survival times than lower grade I and II tumors (1). Within these groups, however, considerable heterogeneity still exists, and delineation of the most aggressive subtypes within grades would be of considerable clinical benefit.

Invasive ductal carcinoma-no special type, as determined morphologically, is thought to arise exclusively from the luminal epithelial cells of the breast. It has been known for some time, however, that a proportion (2–18% of all invasive ductal carcinomas and up to 25% of grade III tumors) of these tumors have been demonstrated to show a basal/myoepithelial cell phenotype by immunohistochemical analysis using a range of markers including intermediate filaments cytokeratin (CK) 5 and 14 (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17). Recent microarray studies have also identified a “basal-like” group of breast tumors based on their patterns of gene expression (18, 19).

Although a comprehensive characterization and definition of basal tumors is lacking, there are a number of features reported to be associated with this phenotype of invasive ductal carcinoma-no special type. Morphologically, they appear to be predominantly grade III (14, 16) and are reported to frequently contain large central acellular zones composed of necrosis, tissue infarction, collagen, and hyaline material on their cut surfaces (15). Immunohistochemically, as well as in the expression of a number of myoepithelial markers, these tumors seem to be predominantly estrogen receptor (ER), progesterone receptor (PR), and ERBB2 negative (16), an immunophenotype resembling BRCA1 tumors (20). Microarray analysis has also demonstrated a similarity between sporadic, basal-like tumors and those familial tumors harboring a BRCA1 mutation, based on their patterns of gene expression (21).

The pathogenesis of such lesions appear to indicate a poor prognosis. As well as the association with high histological grade and hormone receptor negativity, the myoepithelial phenotype has been reported to be associated with a high risk of brain and lung metastases and of death by cancer independent of nodal status and tumor size (22). The basal-like phenotype identified by expression profiling experiments conferred a shorter survival time than the other tumor groups described (19). Tissue microarray analysis of basal keratins 5 and 17 showed a poorer clinical outcome in node-negative tumors expressing one or both of these markers (17).

Preliminary analysis has indicated that the pattern and number of alterations, as detected by comparative genomic hybridization (CGH), more closely resembles pure myoepithelial carcinomas than other grade III invasive ductal carcinoma (16). In this study, we have chosen to use CK14 to identify grade III invasive ductal carcinomas as exhibiting a basal/myoepithelial phenotype and have carried out CGH analysis on 43 CK14 positive and 43 age- and grade-matched CK14 negative tumors, in an attempt to investigate the molecular events associated with the different subtypes and to correlate these data to clinical outcome.

Tumor Specimens.

Patients diagnosed as having a grade III mammary carcinoma at the Hedley Atkins/Cancer Research UK Breast Pathology Laboratory, Guy’s Hospital between 1975 and 1992 were included in the study and have formed a cohort upon which studies have been published previously (23). Stained sections from the primary tumor were retrieved, and the tumor was regraded according to the modified Bloom and Richardson method (1). Any patient subsequently found to have a grade II tumor on review was excluded from the study. An additional group of patients was also excluded because insufficient tumor tissue remained in the diagnostic block for study. Therefore, a total of 418 patients were available. Of these, 89 were found to be either focally or diffusely positive for CK14 by immunohistochemistry, and 43 of these tumors proved amenable to microdissection, DNA extraction, and CGH analysis. Forty-three age-matched, CK14-negative tumors were chosen from the same cohort as controls and were subject to the same experimental difficulties. Matching for tumor size, stage, and nodal status were also carried out as closely as possible, with the two groups showing no significant differences in distribution of these markers.

Immunohistochemistry.

Staining was performed on 3-μm-thick sections of formalin-fixed, paraffin-embedded tissues. Sections were cut and placed onto Vectabond-coated slides and allowed to dry overnight at 37°C. Before staining, the sections were melted onto the slide by placing them in an oven at 60°C for 16 h. Sections were dewaxed, and endogenous peroxidase activity was inhibited using methanol and hydrogen peroxide. After thorough washing, heat-mediated antigen retrieval was used to expose sites of immunoreactivity. This was achieved by placing the slides in a pressure cooker containing boiling 0.01 M citrate buffer (pH 6.0) where they were kept under pressure (15 psi) for 2 min. Once the sections were cool, the remaining immunohistochemical staining was carried out using an Optimax Automated Immunostainer (A. Menarini) with a standard peroxidase-conjugated streptavidin-biotin complex method. CK19 (clone RCK108; 1:50), ER (clone 1D5; 1:100), and PR (clone PgR636; 1:400) were obtained from Dako Ltd., CK14 (clone LL002; 1:50) was obtained from Biogenex, and ERBB2 (clone 3B5; 1:1500) was obtained from Oncogene Science. Sites of peroxidase activity were detected using 3,3′-diaminobenzidine/hydrogen peroxide, which produced a brown reaction product. The slides were lightly stained with hematoxylin. A positive control, known to express the antigen, was included in every batch of staining.

ER and PR were assessed using the Quick score method (24) with a score of ≤3 considered negative. c-erbB2 was assessed in accordance with the Dako HercepTest guidelines with a score of ≤1 considered negative. Cytokeratin 14 and 19 were scored according to the presence or absence of expression in the invasive component.

CGH.

Microdissection, DNA extraction, and CGH analysis were carried out as described previously (25). Briefly, microdissection was carried out using the PixCell II Laser Capture Microdissection system (Arcturus, Mountain View, CA) from formalin-fixed, paraffin-embedded tissue and the DNA extracted with 0.5 μg/μl proteinase K. Amplification and fluorescent labeling of the DNA from microdissected tumor and normal tissue was carried out by degenerate oligonucleotide primed-PCR in two rounds as published previously (26) and hybridized to normal male metaphase spreads (Vysis United Kingdom Ltd., Richmond, England) for 2–3 days at 37°C. Metaphase chromosome preparations were captured using a Zeiss Axioskop microscope, Photometrics (Munich, Germany) KAF1400 CCD camera and Vysis SmartCapture software. Image analysis was performed using Vysis Quips CGH software. Between 5 and 10 representative images of high quality hybridizations were analyzed and the results combined to produce an average fluorescence ratio for each chromosome. Control experiments were carried out using normal:normal (microdissected normal lymph node) cohybridizations, of which the average red:green ratio levels and 95% confidence intervals were used to set the lower and upper limits for scoring losses and gains of genetic material as 0.80–1.20.

Statistical Analysis.

All of the changes in DNA copy number were reduced to “gains” or “losses” at the chromosomal arm resolution. The data were retabulated as +2 for gains and −2 for losses for supervised and unsupervised analysis. Unsupervised hierarchical clustering was carried out using “hclust” in R 1.7.0,6 and plotted using the “heatmap” function in the mva package in R. Hierarchical clustering was based on a Euclidean distance measure using Ward’s minimum variance method. Supervised analysis was carried out using the nearest shrunken centroid classifier (27) implemented in R 1.7.0. Survival analysis was carried out using the statistical platform S-Plus version 6.1 for Windows (Insightful) on our right-censored clinical follow-up data. Kaplan-Meier plots were generated using the function “survfit,” and the log-rank test was carried out to determine whether curves were significantly different from each other using the “survdiff” function. Multivariate analysis was carried out using the Cox proportional hazards model with “coxph” used to investigate the independence of the individual proteins on prognosis.

Of 418 grade III invasive ductal breast carcinomas in our study, 89 (21.2%) were either focally of diffusely positive for CK14. Of these, 43 were amenable to CGH analysis and 43 age-matched CK14-negative grade III tumors were analyzed as controls. Of these 86 tumors analyzed by molecular cytogenetics, 33 were ER positive (38.4%), 25 were PR positive (34.7%), 15 were ERBB2 positive (18.3%), and 83 were positive for CK19 (96.5%). Forty-nine showed positive lymph node metastasis (57.6%). Follow-up data were available for all of the patients who had a mean age at diagnosis of 47.8 years (median, 46.5), mean tumor size of 3.4 cm (median, 3.5 cm), a mean disease-free time of 8.7 years (median, 4.2), and mean survival time of 9.9 years (median, 7.0). Full details of clinicopathological data are available in Supplementary Table S1.

Fig. 1 shows a representative CK14-positive tumor taken for CGH analysis. Positive CK14 tumors were shown to be significantly associated with negative ER status (P < 0.00001), negative PR (P = 0.001), and negative ERBB2 (P = 0.0201), but not with nodal status (P = 0.191) or tumor size (P = 0.9479).

CGH analysis of the 86 tumors showed a mean number of alterations of 8.2 (median, 6.0). Full details of all CGH data for each case are available in Supplementary Table S1. The most frequent alterations over all samples were gains at 1q (21%), 17q (26%), and 20q (41%); and losses at 1p (21%), 16q (20%), 17p (20%), 17q (24%), 19p (21%), and 19q (21%). Differences in CGH profiles between the CK14-positive and -negative tumors are shown in Table 1. Statistical significance was determined by Fisher’s exact test. The CK14-positive tumors showed an increased number of losses at 16p (33% CK14 positive versus 5% CK14 negative; P = 0.016), 17q (37% versus 12%; P = 0.0110), 19q (33% versus 9%; P = 0.0155) and Xp (26% versus 7%, P = 0.0381). CK14-negative tumors displayed an increased number of gains at 17p (5% versus 21%; P = 0.0488), 17q (9% versus 42%; P < 0.0010), 20p (0% versus 21%; P = 0.0025), and 20q (21% versus 60%; P < 0.0004); and losses at 4q (5% versus 23%; P = 0.0261), 9p (0% versus 14%; P = 0.0259), and 13q (0% versus 21%; P = 0.025). Summary karyograms displaying all of the changes in DNA copy number are shown in Fig. 2.

Clinical follow-up data were available for all of the cases, and univariate analysis was carried out to determine which clinicopathological, immunohistochemical, and molecular variables were associated with prognosis (Table 2). In this grade III invasive ductal carcinoma cohort, the most significant prognostic indicator was nodal status, with positive lymph nodes predicting for shorter disease-free time (P = 0.00269) and overall survival (P = 0.00275). Trends were observed linking poor prognosis with ER negativity, ERBB2 positivity, and number of CGH alterations; however, these did not reach statistical significance. A number of chromosomal loci showed a significant association with disease-free and overall survival, including gain of 1q, loss of 3p, loss of 4p, gain of 6p, and loss of 8p. CK14 positivity showed no significant influence on prognosis in this cohort (disease-free survival P = 0.193; overall survival P = 0.385). CK14-positive tumors showed a significantly lower overall mean number of CGH alterations than the CK14-negative group (6.5 versus 10.3; P = 0.0012). Interestingly, the few CK19-negative tumors (3 of 83) showed a significant association with shorter disease-free time (P = 0.0209) and overall survival (P = 0.0402), despite these small numbers.

Because there were a number of alterations that were more or less prevalent in the different groups of grade III tumors, supervised cluster analysis was carried out to determine whether the CGH data would be able to predict for CK14 status on its own. Using the nearest shrunken centroid classifier, 76% of breast tumors were correctly assigned into the relevant group by CGH data alone using leave-one-out cross-validation (cross-validated error-rate, 0.24) after first training the classifier (Fig. 3,A). We next attempted to determine whether we could identify subgroups of these tumors by their CGH data, which may have different biological behaviors. Hierarchical clustering of the CGH data split the tumors into two large branches, containing 38 of 55 (69%) CK14-positive and 26 of 31(84%) CK14-negative tumors, respectively, giving an overall error rate of 0.256 (Fig. 3 B).

These branches could be additionally refined into 6 smaller clusters. When multivariate analysis was carried out using the Cox proportional hazards model, ER negativity, CK14 positivity, positive lymph nodes, and 1 cluster of tumors were found to be independent predictors of poor survival (Table 3). Within the large, predominantly CK14-positive branch, “cluster 1” was composed of 18 of 23 (78%) CK14 positive tumors and was found to indicate shorter overall survival (P = 0.0270) and disease-free time (P = 0.0503).

Univariate analyses showed that cluster 1 tumors predicted shorter overall survival for all of the tumors (P = 0.0414) and when the CK14-positive tumors were considered (P = 0.0475). Shorter disease-free times were also observed, but the differences were not statistically significant (P = 0.175, all tumors; P = 0.127, CK14-positive tumors). The Kaplan-Meier survival curves are shown in Fig. 4.

The cluster 1 tumors, as well as being predominantly CK14 positive (18 of 23; 79%; P = 0.003), were found to be associated with negative ER (P = 0.0231) and negative PR (P = 0.039) but not with ERBB2 (P = 0.7458), nodal status (P > 0.9999), or tumor size (P = 0.6237). The CGH alterations, which were found to be significantly associated with the cluster 1 tumors versus the rest, were gain of 1q (P = 0.004), and losses at 8p (P = 0.0199), 16p (P < 0.0001), 16q (P = 0.0065), 17p (P = 0.0488), 17q (P = 0.0001), 19p (P = 0.001), and 19q (P = 0.0001), and are shown in Table 4.

There is accumulating evidence to suggest that different histological grades of invasive ductal breast carcinomas may have distinct molecular origins and pathogenesis and do not typically progress from one grade group to another (28, 29, 30, 31). The different grades have different clinical behaviors, and within-grade studies to identify the more aggressive subgroups of these classes of breast tumors would be of great assistance in clinical management. The expression of basal/myoepithelial markers has been observed in a proportion of grade III invasive breast tumors, and the spectrum of basal-like tumors, also recognized by morphology (15, 32), molecular cytogenetics (16, 33), and expression profiling (18, 19), has been associated with poor prognosis (17). CGH has the advantage of being applied to archival pathology specimens with long-term follow-up as well as being amenable to microdissection strategies to profile the molecular genetic change occurring in a pure population of tumor cells.

In the cohort of grade III tumors in the present study, CK14 positivity was not associated significantly with prognosis. At the CGH level, the CK14-positive tumors showed fewer overall changes in DNA copy number than the CK14-negative group. This may in part explain the conflict with a previous study (33), which found a higher number of CGH alterations in the basal-like group of tumors as determined by CK5/6 positivity. This incongruity with published data suggesting that basal keratin expression confers a poorer prognosis must be interpreted in the light of the fact that we have focused only on poorly differentiated, high-grade malignancies. Within this cohort, ER negativity is not a significant prognostic indicator by univariate analysis, nor is tumor size or ERBB2 status. Interestingly, the few (3 of 83) tumors that were negative for the luminal epithelial keratin CK19 showed a very poor prognosis, statistically significant despite the small numbers.

The CK14-positive and -negative groups were clearly different in terms of their CGH profiles. In particular, the CK14-positive tumors showed an increased prevalence for losses at 16p, 17q, and 19q, all alterations associated with pure myoepithelial carcinomas (34). None of these alterations on their own conferred any prognostic information, although gain of 1q and 6p, as well as losses of 3p, 4p, and 8p did indicate shorter disease-free and overall survival times in the whole cohort. Taken as a whole, the CGH profiles of the tumors alone were able to predict the CK14 status in approximately three-quarters of cases by supervised analysis, using leave-one-out cross-validation. This demonstrates the inherent differences in the molecular evolution of the tumor groups.

Unsupervised hierarchical clustering has been applied to gene expression profiling data to identify subgroups of breast tumors with different clinical outcomes (19). Such a statistical approach may also be applied to CGH data, with the advantage in this instance that we are profiling pure populations of microdissected tumor cells. In our study, hierarchical clustering revealed two large branches, which predicted for CK status again with approximately three-quarters accuracy. Within the predominantly CK14-positive group, 4 separate clusters were identified, with the cluster 1 tumors accounting for 18 of 43 CK14-positive tumors. This subgroup of tumors had a worse prognosis than the rest of the tumors, both in terms of the whole data set, as well as that stratified purely by CK14 positivity. In multivariate analysis, this cluster of tumors was found to be an independent indicator of shorter overall survival in a model including ER, CK14, and nodal status, identified by the Cox proportional hazards test.

Substratification of CK14-positive breast tumors into two groups, based on their CGH profiles, which is reflected in their biological behavior, is of considerable clinical interest. This poor prognosis group was associated with negative hormone receptor status and exhibited a number of CGH alterations with higher prevalence than the better prognosis tumors, including gain at 1q, and losses at 8p, 16p, 16q, 17p, 17q, 19p, and 19q. The association of these loci with a poor clinical outcome group of tumors will provide clues for targeted studies hoping to unravel the underlying molecular events associated with the pathogenesis of these lesions.

The immunophenotype of these tumors, as has been pointed out in smaller studies (16), exhibits a resemblance to tumors with germ-line BRCA1 mutation (20), and this possible association has been postulated recently by gene expression analysis (21). Investigations into epigenetic mechanisms of BRCA1 inactivation in basal-like breast tumors seem warranted.

The clinical heterogeneity of grade III invasive ductal breast carcinomas is well-known. The idea that we could identify, at diagnosis, subgroups of these patients that will do badly (and thus require aggressive therapy) or relatively well (where the patient may be spared such treatment) is an attractive one. Clearly some form of basal/myoepithelial differentiation is apparent in these lesions, as determined by their expression of some basal markers, and a pattern of CGH alterations at loci associated with pure myoepithelial carcinomas. Accurate characterization of these lesions at the morphological level, along with an additional immunohistochemical refinement using an extensive panel of basal/myoepithelial markers will be necessary to produce a set of criteria that will allow the accurate diagnosis of these tumors, with the implications that will have on patient management. Unraveling the molecular pathways that drive the divergent groups of good- and poor-prognosis tumors will be required to identify potential targets for novel therapeutic strategies.

Fig. 1.

Photomicrographs of a grade III invasive ductal breast carcinoma CK272. A, H&E, low power (×10). B, H&E, high power (×40). C, diffuse immunohistochemical staining of cytokeratin 14 (×10). D, diffuse staining of cytokeratin 19 (×10).

Fig. 1.

Photomicrographs of a grade III invasive ductal breast carcinoma CK272. A, H&E, low power (×10). B, H&E, high power (×40). C, diffuse immunohistochemical staining of cytokeratin 14 (×10). D, diffuse staining of cytokeratin 19 (×10).

Close modal
Fig. 2.

Summary karyograms of comparative genomic hybridization alterations in grade III invasive ductal breast carcinomas. A, CK14-positive tumors. B, CK14-negative tumors. Red bars to the left of the chromosomal ideogram represent a loss at that locus in a single case; green bars to the right of the ideogram represent a gain in DNA copy number.

Fig. 2.

Summary karyograms of comparative genomic hybridization alterations in grade III invasive ductal breast carcinomas. A, CK14-positive tumors. B, CK14-negative tumors. Red bars to the left of the chromosomal ideogram represent a loss at that locus in a single case; green bars to the right of the ideogram represent a gain in DNA copy number.

Close modal
Fig. 3.

Statistical analysis of comparative genomic hybridization (CGH) profiles across all tumors. A, supervised analysis using the nearest shrunken centroid classifier. The cross-validated probabilities of class assignment are shown for all tumors separated by their CK14 status. The overall cross-validated error rate was 0.24. B, unsupervised hierarchical clustering of the CGH data. Agglomerative clustering identifies two major branches of tumors (horizontal) according to patterns of gain and loss on individual chromosomal arms (vertical), separating the cases according to CK14 status in 74% of samples. Seven additional subdivisions are identified, with cluster 1 composed of 18 of 23 CK14-positive cases. Green squares indicate a gain by CGH; red square a loss. CK14-positive tumors are highlighted in blue.

Fig. 3.

Statistical analysis of comparative genomic hybridization (CGH) profiles across all tumors. A, supervised analysis using the nearest shrunken centroid classifier. The cross-validated probabilities of class assignment are shown for all tumors separated by their CK14 status. The overall cross-validated error rate was 0.24. B, unsupervised hierarchical clustering of the CGH data. Agglomerative clustering identifies two major branches of tumors (horizontal) according to patterns of gain and loss on individual chromosomal arms (vertical), separating the cases according to CK14 status in 74% of samples. Seven additional subdivisions are identified, with cluster 1 composed of 18 of 23 CK14-positive cases. Green squares indicate a gain by CGH; red square a loss. CK14-positive tumors are highlighted in blue.

Close modal
Fig. 4.

Kaplan-Meier survival curves for subgroups of grade III breast tumors identified by cluster analysis of their comparative genomic hybridization profiles. A and C, disease-free survival; B and D, overall survival. A and B, cluster 1 tumors plotted against all others; C and D, CK14-positive cluster 1 tumors plotted against all other CK14-positive tumors. Statistical significance was calculated by the log-rank test.

Fig. 4.

Kaplan-Meier survival curves for subgroups of grade III breast tumors identified by cluster analysis of their comparative genomic hybridization profiles. A and C, disease-free survival; B and D, overall survival. A and B, cluster 1 tumors plotted against all others; C and D, CK14-positive cluster 1 tumors plotted against all other CK14-positive tumors. Statistical significance was calculated by the log-rank test.

Close modal

Grant support: J. Reis-Filho is the recipient of the Gordon Signy International Fellowship Award and is partially supported by a Ph.D. grant (Ref:SFRH/BD/5386/2001) from the Fundação para a Ciência e a Tecnologia, Portugal.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article may be found at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org).

Requests for reprints: Sunil R. Lakhani, The Breakthrough Toby Robins Breast Cancer Research Centre, Institute of Cancer Research, Fulham Road, London SW3 6JB, United Kingdom. Phone: 20-7153-5525; Fax: 20-7153-5533; E-mail: Sunil.Lakhani@icr.ac.uk

6

Internet address: http://www.r-project.org.

Table 1

Differences in CGH alterations between CK14-positive and CR14-negative tumors

Chromosome gainOverall (%)CK14 positive (%)CK14 negative (%)P                  *Chromosome lossOverall (%)CK14 positive (%)CK14 negative (%)P                  *
1p 5 (6) 4 (9) 1 (2) 0.2077 1p 18 (21) 9 (21) 9 (21) >0.9999 
1q 18 (21) 7 (16) 11 (26) 0.4271 1q 1 (1) 0 (0) 1 (2) >0.9999 
2p 7 (8) 2 (5) 5 (12) 0.4331 2p 11 (13) 3 (7) 8 (19) 0.1951 
2q 10 (12) 3 (7) 7 (16) 0.3134 2q 12 (14) 3 (7) 9 (21) 0.1171 
3p 8 (9) 3 (7) 5 (12) 0.7130 3p 10 (12) 3 (7) 7 (16) 0.3134 
3q 7 (8) 2 (5) 5 (12) 0.4331 3q 7 (8) 2 (5) 5 (12) 0.4331 
4p 4 (5) 1 (2) 3 (7) 0.6160 4p 7 (8) 1 (2) 6 (14) 0.1096 
4q 11 (13) 6 (14) 5 (12) >0.9999 4q 12 (14) 2 (5) 10 (23) 0.0261 
5p 3 (3) 0 (0) 3 (7) 0.2412 5p 6 (7) 1 (2) 5 (12) 0.2020 
5q 9 (10) 2 (5) 7 (16) 0.1561 5q 12 (14) 3 (7) 9 (21) 0.1171 
6p 7 (8) 1 (2) 6 (14) 0.1096 6p 5 (6) 2 (5) 3 (7) >0.9999 
6q 8 (9) 5 (12) 3 (7) 0.7130 6q 10 (12) 3 (7) 7 (16) 0.3134 
7p 6 (7) 4 (9) 2 (5) 0.7383 7p 6 (7) 1 (2) 5 (12) 0.2020 
7q 10 (12) 6 (14) 4 (9) 0.7383 7q 6 (7) 0 (0) 6 (14) 0.0259 
8p 7 (8) 3 (7) 4 (9) >0.9999 8p 7 (8) 2 (5) 5 (12) 0.4331 
8q 11 (13) 5 (12) 6 (14) >0.9999 8q 12 (14) 2 (5) 10 (23) 0.0261 
9p 4 (5) 2 (5) 2 (5) >0.9999 9p 6 (7) 0 (0) 6 (14) 0.0259 
9q 8 (9) 5 (12) 3 (7) 0.7130 9q 14 (16) 5 (12) 9 (21) 0.3816 
10p 2 (2) 1 (2) 1 (2) >0.9999 10p 2 (2) 1 (2) 1 (2) >0.9999 
10q 3 (3) 1 (2) 2 (5) >0.9999 10q 3 (3) 1 (2) 2 (5) >0.9999 
11p 2 (2) 0 (0) 2 (5) 0.4941 11p 7 (8) 3 (7) 4 (9) >0.9999 
11q 6 (7) 3 (7) 3 (7) >0.9999 11q 7 (8) 3 (7) 4 (9) >0.9999 
12p 6 (7) 2 (5) 4 (9) 0.7383 12p 6 (7) 1 (2) 5 (12) 0.2020 
12q 2 (2) 1 (2) 1 (2) >0.9999 12q 10 (12) 2 (5) 8 (19) 0.0887 
13q 5 (6) 2 (5) 3 (7) >0.9999 13q 9 (10) 0 (0) 9 (21) 0.0025 
14q 6 (7) 2 (5) 4 (9) 0.7383 14q 6 (7) 0 (0) 6 (14) 0.0259 
15q 4 (5) 1 (2) 3 (7) 0.6160 15q 9 (10) 2 (5) 7 (16) 0.1561 
16p 2 (2) 1 (2) 1 (2) >0.9999 16p 16 (19) 14 (33) 2 (5) 0.0016 
16q 1 (1) 0 (0) 1 (2) >0.9999 16q 17 (20) 11 (26) 6 (14) 0.2787 
17p 11 (13) 2 (5) 9 (21) 0.0488 17p 17 (20) 12 (28) 5 (12) 0.1025 
17q 22 (26) 4 (9) 18 (42) 0.0010 17q 21 (24) 16 (37) 5 (12) 0.0110 
18p 6 (7) 2 (5) 4 (9) 0.7383 18p 3 (3) 2 (5) 1 (2) >0.9999 
18q 2 (2) 1 (2) 1 (2) >0.9999 18q 14 (16) 4 (9) 10 (23) 0.1424 
19p 5 (6) 1 (2) 4 (9) 0.3600 19p 18 (21) 11 (26) 7 (16) 0.4271 
19q 5 (6) 0 (0) 5 (12) 0.0553 19q 18 (21) 14 (33) 4 (9) 0.0155 
20p 9 (10) 0 (0) 9 (21) 0.0025 20p 9 (10) 6 (14) 3 (7) 0.4833 
20q 35 (41) 9 (21) 26 (60) 0.0004 20q 4 (5) 0 (0) 4 (9) 0.1162 
21q 6 (7) 3 (7) 3 (7) >0.9999 21q 6 (7) 5 (12) 1 (9) 0.2020 
22q 4 (5) 0 (0) 4 (9) 0.1162 22q 14 (16) 8 (19) 6 (14) 0.7712 
Xp 1 (1) 0 (0) 1 (2) >0.9999 Xp 14 (16) 11 (26) 3 (7) 0.0381 
Xq 3 (3) 0 (0) 3 (7) 0.2412 Xq 13 (15) 7 (16) 6 (14) >0.9999 
Chromosome gainOverall (%)CK14 positive (%)CK14 negative (%)P                  *Chromosome lossOverall (%)CK14 positive (%)CK14 negative (%)P                  *
1p 5 (6) 4 (9) 1 (2) 0.2077 1p 18 (21) 9 (21) 9 (21) >0.9999 
1q 18 (21) 7 (16) 11 (26) 0.4271 1q 1 (1) 0 (0) 1 (2) >0.9999 
2p 7 (8) 2 (5) 5 (12) 0.4331 2p 11 (13) 3 (7) 8 (19) 0.1951 
2q 10 (12) 3 (7) 7 (16) 0.3134 2q 12 (14) 3 (7) 9 (21) 0.1171 
3p 8 (9) 3 (7) 5 (12) 0.7130 3p 10 (12) 3 (7) 7 (16) 0.3134 
3q 7 (8) 2 (5) 5 (12) 0.4331 3q 7 (8) 2 (5) 5 (12) 0.4331 
4p 4 (5) 1 (2) 3 (7) 0.6160 4p 7 (8) 1 (2) 6 (14) 0.1096 
4q 11 (13) 6 (14) 5 (12) >0.9999 4q 12 (14) 2 (5) 10 (23) 0.0261 
5p 3 (3) 0 (0) 3 (7) 0.2412 5p 6 (7) 1 (2) 5 (12) 0.2020 
5q 9 (10) 2 (5) 7 (16) 0.1561 5q 12 (14) 3 (7) 9 (21) 0.1171 
6p 7 (8) 1 (2) 6 (14) 0.1096 6p 5 (6) 2 (5) 3 (7) >0.9999 
6q 8 (9) 5 (12) 3 (7) 0.7130 6q 10 (12) 3 (7) 7 (16) 0.3134 
7p 6 (7) 4 (9) 2 (5) 0.7383 7p 6 (7) 1 (2) 5 (12) 0.2020 
7q 10 (12) 6 (14) 4 (9) 0.7383 7q 6 (7) 0 (0) 6 (14) 0.0259 
8p 7 (8) 3 (7) 4 (9) >0.9999 8p 7 (8) 2 (5) 5 (12) 0.4331 
8q 11 (13) 5 (12) 6 (14) >0.9999 8q 12 (14) 2 (5) 10 (23) 0.0261 
9p 4 (5) 2 (5) 2 (5) >0.9999 9p 6 (7) 0 (0) 6 (14) 0.0259 
9q 8 (9) 5 (12) 3 (7) 0.7130 9q 14 (16) 5 (12) 9 (21) 0.3816 
10p 2 (2) 1 (2) 1 (2) >0.9999 10p 2 (2) 1 (2) 1 (2) >0.9999 
10q 3 (3) 1 (2) 2 (5) >0.9999 10q 3 (3) 1 (2) 2 (5) >0.9999 
11p 2 (2) 0 (0) 2 (5) 0.4941 11p 7 (8) 3 (7) 4 (9) >0.9999 
11q 6 (7) 3 (7) 3 (7) >0.9999 11q 7 (8) 3 (7) 4 (9) >0.9999 
12p 6 (7) 2 (5) 4 (9) 0.7383 12p 6 (7) 1 (2) 5 (12) 0.2020 
12q 2 (2) 1 (2) 1 (2) >0.9999 12q 10 (12) 2 (5) 8 (19) 0.0887 
13q 5 (6) 2 (5) 3 (7) >0.9999 13q 9 (10) 0 (0) 9 (21) 0.0025 
14q 6 (7) 2 (5) 4 (9) 0.7383 14q 6 (7) 0 (0) 6 (14) 0.0259 
15q 4 (5) 1 (2) 3 (7) 0.6160 15q 9 (10) 2 (5) 7 (16) 0.1561 
16p 2 (2) 1 (2) 1 (2) >0.9999 16p 16 (19) 14 (33) 2 (5) 0.0016 
16q 1 (1) 0 (0) 1 (2) >0.9999 16q 17 (20) 11 (26) 6 (14) 0.2787 
17p 11 (13) 2 (5) 9 (21) 0.0488 17p 17 (20) 12 (28) 5 (12) 0.1025 
17q 22 (26) 4 (9) 18 (42) 0.0010 17q 21 (24) 16 (37) 5 (12) 0.0110 
18p 6 (7) 2 (5) 4 (9) 0.7383 18p 3 (3) 2 (5) 1 (2) >0.9999 
18q 2 (2) 1 (2) 1 (2) >0.9999 18q 14 (16) 4 (9) 10 (23) 0.1424 
19p 5 (6) 1 (2) 4 (9) 0.3600 19p 18 (21) 11 (26) 7 (16) 0.4271 
19q 5 (6) 0 (0) 5 (12) 0.0553 19q 18 (21) 14 (33) 4 (9) 0.0155 
20p 9 (10) 0 (0) 9 (21) 0.0025 20p 9 (10) 6 (14) 3 (7) 0.4833 
20q 35 (41) 9 (21) 26 (60) 0.0004 20q 4 (5) 0 (0) 4 (9) 0.1162 
21q 6 (7) 3 (7) 3 (7) >0.9999 21q 6 (7) 5 (12) 1 (9) 0.2020 
22q 4 (5) 0 (0) 4 (9) 0.1162 22q 14 (16) 8 (19) 6 (14) 0.7712 
Xp 1 (1) 0 (0) 1 (2) >0.9999 Xp 14 (16) 11 (26) 3 (7) 0.0381 
Xq 3 (3) 0 (0) 3 (7) 0.2412 Xq 13 (15) 7 (16) 6 (14) >0.9999 

NOTE: Numbers of tumors (and percentages) exhibiting a change in DNA copy number in all cases, as well as stratified by CK14 status are given along with the P value as determined by Fisher’s exact test to probe differences between the two groups. Significant (P < 0.05) differences in gains are in italics.

*

Determined by Fisher’s exact test.

Table 2

Univariate analysis of prognostic indicators

FactorNo. of casesDisease-free timePDeath from breast cancerP
Mean survival (y)SEMean survival (y)SE
Size of tumor    0.8   0.825 
 <2.0 cm 21 10.11 2.31  13.6 2.41  
 2.0–5.0 cm 54 11.76 1.49  13.2 1.47  
 >5.0 cm 10 9.39 3.2  10 3.04  
Nodal status    0.00269   0.00275 
 Positive 49 8.23 1.44  10.4 1.55  
 Negative 36 16.51 1.85  17.7 1.77  
ERa    0.151   0.139 
 Positive 33 12.5 1.73  13.6 1.59  
 Negative 53 10.5 1.57  12.5 1.59  
PR    0.471   0.588 
 Positive 25 9.68 2.08  11.1 1.99  
 Negative 47 12.03 1.63  13.7 1.59  
ERBB2    0.318   0.412 
 Positive 15 9.25 2.64  10.6 2.6  
 Negative 67 12.45 1.46  14 1.41  
CK14    0.193   0.385 
 Positive 43 12.7 1.73  14.2 1.67  
 Negative 43 10.5 1.72  12.5 1.71  
CK19    0.0209   0.0402 
 Positive 83 12.38 1.29  14.1 1.261  
 Negative 1.17 0.395  2.72 0.784  
CGH alterations    0.888   0.931 
 0 13.35 4.35  13.9 4.17  
 1–10 52 11.98 1.54  13.9 1.52  
 11–20 20 9.53 2.26  11.9 2.43  
 21 or more 9.53 3.54  10.5 3.21  
Gain of 1q    0.0469   0.043 
 Present 18 6.63 2.16  7.91 1.96  
 Absent 67 13.42 1.44  15.32 1.41  
Loss of 3p    0.00339   0.00244 
 Present 10 2.54 1.01  3.5 0.954  
 Absent 68 13.26 1.43  15.2 1.392  
Loss of 4p    0.00397   0.0423 
 Present 2.57 1.31  5.76 3.01  
 Absent 75 12.82 1.36  14.43 1.32  
Gain of 6p    0.0183   0.0408 
 Present 2.44 1.33  3.64 1.21  
 Absent 74 12.34 1.37  14.19 1.34  
Loss of 8p    0.0032   0.0186 
 Present 2.34 1.51  3.55 1.37  
 Absent 72 12.68 1.39  14.22 1.31  
CGH profile    0.175   0.0414 
 Cluster 1 23 8.19 2.06  8.83 1.98  
 All others 63 13.08 1.48  15.2 1.44  
FactorNo. of casesDisease-free timePDeath from breast cancerP
Mean survival (y)SEMean survival (y)SE
Size of tumor    0.8   0.825 
 <2.0 cm 21 10.11 2.31  13.6 2.41  
 2.0–5.0 cm 54 11.76 1.49  13.2 1.47  
 >5.0 cm 10 9.39 3.2  10 3.04  
Nodal status    0.00269   0.00275 
 Positive 49 8.23 1.44  10.4 1.55  
 Negative 36 16.51 1.85  17.7 1.77  
ERa    0.151   0.139 
 Positive 33 12.5 1.73  13.6 1.59  
 Negative 53 10.5 1.57  12.5 1.59  
PR    0.471   0.588 
 Positive 25 9.68 2.08  11.1 1.99  
 Negative 47 12.03 1.63  13.7 1.59  
ERBB2    0.318   0.412 
 Positive 15 9.25 2.64  10.6 2.6  
 Negative 67 12.45 1.46  14 1.41  
CK14    0.193   0.385 
 Positive 43 12.7 1.73  14.2 1.67  
 Negative 43 10.5 1.72  12.5 1.71  
CK19    0.0209   0.0402 
 Positive 83 12.38 1.29  14.1 1.261  
 Negative 1.17 0.395  2.72 0.784  
CGH alterations    0.888   0.931 
 0 13.35 4.35  13.9 4.17  
 1–10 52 11.98 1.54  13.9 1.52  
 11–20 20 9.53 2.26  11.9 2.43  
 21 or more 9.53 3.54  10.5 3.21  
Gain of 1q    0.0469   0.043 
 Present 18 6.63 2.16  7.91 1.96  
 Absent 67 13.42 1.44  15.32 1.41  
Loss of 3p    0.00339   0.00244 
 Present 10 2.54 1.01  3.5 0.954  
 Absent 68 13.26 1.43  15.2 1.392  
Loss of 4p    0.00397   0.0423 
 Present 2.57 1.31  5.76 3.01  
 Absent 75 12.82 1.36  14.43 1.32  
Gain of 6p    0.0183   0.0408 
 Present 2.44 1.33  3.64 1.21  
 Absent 74 12.34 1.37  14.19 1.34  
Loss of 8p    0.0032   0.0186 
 Present 2.34 1.51  3.55 1.37  
 Absent 72 12.68 1.39  14.22 1.31  
CGH profile    0.175   0.0414 
 Cluster 1 23 8.19 2.06  8.83 1.98  
 All others 63 13.08 1.48  15.2 1.44  

NOTE. Statistical significance was calculated for disease-free and overall survival between different stratifications of clinicopathological, immunohistochemical, and CGH data by the log-rank test. Significant factors (P < 0.05) are highlighted in italics.

a

ER, estrogen receptor; PR, progesterone receptor; CGH, comparative genomic hybridization.

Table 3

Multivariate analysis of prognostic indicators

FactorDisease-free timeDeath from breast cancer
Hazard ratio (95% CI)aP                  *Hazard ratio (95% CI)P                  *
ER negative 1.598 (1.417–1.853) 0.0072 1.682 (1.498–1.950) 0.0060 
Positive nodes 1.661 (1.213–2.276) 0.0016 1.789 (1.280–2.501) 0.0007 
CK14 negative 1.608 (1.426–1.865) 0.0071 1.570 (1.377–1.846) 0.0140 
Cluster 1 1.980 (0.991–3.957) 0.0530 2.231 (1.097–4.537) 0.0270 
FactorDisease-free timeDeath from breast cancer
Hazard ratio (95% CI)aP                  *Hazard ratio (95% CI)P                  *
ER negative 1.598 (1.417–1.853) 0.0072 1.682 (1.498–1.950) 0.0060 
Positive nodes 1.661 (1.213–2.276) 0.0016 1.789 (1.280–2.501) 0.0007 
CK14 negative 1.608 (1.426–1.865) 0.0071 1.570 (1.377–1.846) 0.0140 
Cluster 1 1.980 (0.991–3.957) 0.0530 2.231 (1.097–4.537) 0.0270 

NOTE. Independent factors that predict for disease-free and overall survival were calculated by the Cox proportional hazards model with simultaneous inclusion of all factors shown. Significant factors (P < 0.05) are highlighted in italics.

a

CI, confidence interval; ER, estrogen receptor.

*

Determined by Cox proportional hazards model.

Table 4

Association of specific immunohistochemical and CGH data with cluster 1 tumors

FactorCluster 1 tumors (%)All other tumors (%)P (Fisher’s exact)
CK14 positive 18 (78.3) 25 (39.7) 0.003 
ERa negative 19 (82.6) 34 (54.0) 0.0231 
PR negative 15 (65.2) 32 (50.8) 0.039 
1q gain 10 (43.5) 8 (12.7) 0.004 
8p loss 5 (21.7) 2 (3.2) 0.0199 
16p loss 13 (56.5) 3 (4.8) <0.0001 
16q loss 9 (39.1) 8 (12.7) 0.0116 
17p loss 8 (34.8) 9 (14.3) 0.0488 
17q loss 13 (56.5) 8 (12.7) 0.0001 
19p loss 12 (52.2) 6 (9.5) 0.0001 
19q loss 12 (52.2) 6 (9.5) 0.0001 
FactorCluster 1 tumors (%)All other tumors (%)P (Fisher’s exact)
CK14 positive 18 (78.3) 25 (39.7) 0.003 
ERa negative 19 (82.6) 34 (54.0) 0.0231 
PR negative 15 (65.2) 32 (50.8) 0.039 
1q gain 10 (43.5) 8 (12.7) 0.004 
8p loss 5 (21.7) 2 (3.2) 0.0199 
16p loss 13 (56.5) 3 (4.8) <0.0001 
16q loss 9 (39.1) 8 (12.7) 0.0116 
17p loss 8 (34.8) 9 (14.3) 0.0488 
17q loss 13 (56.5) 8 (12.7) 0.0001 
19p loss 12 (52.2) 6 (9.5) 0.0001 
19q loss 12 (52.2) 6 (9.5) 0.0001 

NOTE. Numbers of tumors (and percentage) exhibiting a change in DNA copy number in cluster 1 and other tumors are given along with the P value as determined by Fisher’s exact test to probe differences between the two groups. Significant (P < 0.05) differences in gains are highlighted in italics.

a

ER, estrogen receptor; PR, progesterone receptor.

* Determined by Fisher’s exact test.

1
Elston CW, Ellis IO Pathological prognostic factors in breast cancer.: I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up.“.
Histopathology
1991
;
19
:
403
-10.
2
Gusterson BA, Warburton MJ, Mitchell D, Ellison M, Neville AM, Rudland PS Distribution of myoepithelial cells and basement membrane proteins in the normal breast and in benign and malignant breast diseases.
Cancer Res
1982
;
42
:
4763
-70.
3
Dairkee SH, Blayney C, Smith HS, Hackett AJ Monoclonal antibody that defines human myoepithelium.
Proc Natl Acad Sci USA
1985
;
82
:
7409
-13.
4
Altmannsberger M, Dirk T, Droese M, Weber K, Osborn M Keratin polypeptide distribution in benign and malignant breast tumors: subdivision of ductal carcinomas using monoclonal antibodies“.
Virchows Arch B Cell Pathol Incl Mol Pathol
1986
;
51
:
265
-75.
5
Nagle RB, Bocker W, Davis JR, Heid HW, Kaufmann M, Lucas DO, Jarasch ED Characterization of breast carcinomas by two monoclonal antibodies distinguishing myoepithelial from luminal epithelial cells.
J Histochem Cytochem
1986
;
34
:
869
-81.
6
Dairkee SH, Ljung BM, Smith H, Hackett A Immunolocalization of a human basal epithelium specific keratin in benign and malignant breast disease.
Breast Cancer Res Treat
1987
;
10
:
11
-20.
7
Jarasch ED, Nagle RB, Kaufmann M, Maurer C, Bocker WJ Differential diagnosis of benign epithelial proliferations and carcinomas of the breast using antibodies to cytokeratins.
Hum Pathol
1988
;
19
:
276
-89.
8
Wetzels RH, Holland R, van Haelst UJ, Lane EB, Leigh IM, Ramaekers FC Detection of basement membrane components and basal cell keratin 14 in noninvasive and invasive carcinomas of the breast.
Am J Pathol
1989
;
134
:
571
-9.
9
Gould VE, Koukoulis GK, Jansson DS, Nagle RB, Franke WW, Moll R Coexpression patterns of vimentin and glial filament protein with cytokeratins in the normal, hyperplastic, and neoplastic breast.
Am J Pathol
1990
;
137
:
1143
-55.
10
Wetzels RH, Kuijpers HJ, Lane EB, et al C. Basal cell-specific and hyperproliferation-related keratins in human breast cancer.
Am J Pathol
1991
;
138
:
751
-63.
11
Heatley M, Maxwell P, Whiteside C, Toner P Cytokeratin intermediate filament expression in benign and malignant breast disease.
J Clin Pathol
1995
;
48
:
26
-32.
12
Santini D, Ceccarelli C, Taffurelli M, Pileri S, Marrano D Differentiation pathways in primary invasive breast carcinoma as suggested by intermediate filament and biopathological marker expression.
J Pathol
1996
;
179
:
386
-91.
13
Rejthar A, Nenutil R The intermediate filaments and prognostically oriented morphological classification in ductal breast carcinoma.
Neoplasma
1997
;
44
:
370
-3.
14
Malzahn K, Mitze M, Thoenes M, Moll R Biological and prognostic significance of stratified epithelial cytokeratins in infiltrating ductal breast carcinomas.
Virchows Arch
1998
;
433
:
119
-29.
15
Tsuda H, Takarabe T, Hasegawa T, Murata T, Hirohashi S Myoepithelial differentiation in high-grade invasive ductal carcinomas with large central acellular zones.
Hum Pathol
1999
;
30
:
1134
-9.
16
Jones C, Nonni AV, Fulford L, et al CGH analysis of ductal carcinoma of the breast with basaloid/myoepithelial cell differentiation.
Br J Cancer
2001
;
85
:
422
-7.
17
van de Rijn M, Perou CM, Tibshirani R, et al Expression of cytokeratins 17 and 5 identifies a group of breast carcinomas with poor clinical outcome.
Am J Pathol
2002
;
161
:
1991
-6.
18
Perou CM, Sorlie T, Eisen MB, et al Molecular portraits of human breast tumours.
Nature
2000
;
406
:
747
-52.
19
Sorlie T, Perou CM, Tibshirani R, et al Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications.
Proc Natl Acad Sci USA
2001
;
98
:
10869
-74.
20
Lakhani SR, Van De Vijver MJ, Jacquemier J, et al The pathology of familial breast cancer: predictive value of immunohistochemical markers estrogen receptor, progesterone receptor, HER-2, and p53 in patients with mutations in BRCA1 and BRCA2.
J Clin Oncol
2002
;
20
:
2310
-8.
21
Sorlie T, Tibshirani R, Parker J, et al Repeated observation of breast tumor subtypes in independent gene expression data sets.
Proc Natl Acad Sci USA
2003
;
10
:
8418
-23.
22
Tsuda H, Takarabe T, Hasegawa F, Fukutomi T, Hirohashi S Large, central acellular zones indicating myoepithelial tumor differentiation in high-grade invasive ductal carcinomas as markers of predisposition to lung and brain metastases.
Am J Surg Pathol
2000
;
24
:
197
-202.
23
Gillett CE, Miles DW, Ryder K, et al Retention of the expression of E-cadherin and catenins is associated with shorter survival in grade III ductal carcinoma of the breast.
J Pathol
2001
;
193
:
433
-41.
24
Barnes DM, Harris WH, Smith P, Millis RR, Rubens RD Immunohistochemical determination of oestrogen receptor: comparison of different methods of assessment of staining and correlation with clinical outcome of breast cancer patients.
Br J Cancer
1996
;
74
:
1445
-51.
25
Jones C, Damiani S, Wells D, Chaggar R, Lakhani SR, Eusebi V Molecular cytogenetic comparison of apocrine hyperplasia and apocrine carcinoma of the breast.
Am J Pathol
2001
;
158
:
207
-14.
26
Wells D, Sherlock JK, Handyside AH, Delhanty JD Detailed chromosomal and molecular genetic analysis of single cells by whole genome amplification and comparative genomic hybridisation.
Nucleic Acids Res
1999
;
27
:
1214
-8.
27
Tibshirani R, Hastie T, Narasimhan B, Chu G Diagnosis of multiple cancer types by shrunken centroids of gene expression.
Proc Natl Acad Sci USA
2002
;
99
:
6567
-72.
28
Buerger H, Otterbach F, Simon R, et al Different genetic pathways in the evolution of invasive breast cancer are associated with distinct morphological subtypes.
J Pathol
1999
;
189
:
521
-6.
29
Roylance R, Gorman P, Harris W, et al Comparative genomic hybridization of breast tumors stratified by histological grade reveals new insights into the biological progression of breast cancer.
Cancer Res
1999
;
59
:
1433
-6.
30
Buerger H, Mommers EC, Littmann R, et al Ductal invasive G2 and G3 carcinomas of the breast are the end stages of at least two different lines of genetic evolution.
J Pathol
2001
;
194
:
165
-70.
31
Roylance R, Gorman P, Hanby A, Tomlinson I Allelic imbalance analysis of chromosome 16q shows that grade I and grade III invasive ductal breast cancers follow different genetic pathways.
J Pathol
2002
;
196
:
32
-6.
32
Jimenez RE, Wallis T, Visscher DW Centrally necrotizing carcinomas of the breast: a distinct histologic subtype with aggressive clinical behavior.
Am J Surg Pathol
2001
;
25
:
331
-7.
33
Korsching E, Packeisen J, Agelopoulos K, et al Cytogenetic alterations and cytokeratin expression patterns in breast cancer: integrating a new model of breast differentiation into cytogenetic pathways of breast carcinogenesis.
Lab Invest
2002
;
82
:
1525
-33.
34
Jones C, Foschini MP, Chaggar R, et al Comparative genomic hybridization analysis of myoepithelial carcinoma of the breast.
Lab Invest
2000
;
80
:
831
-6.

Supplementary data