A total of 188 familial breast cancer patients belonging either to families with at least three women affected with breast and/or ovarian cancer, one of them diagnosed before 50 years of age, or to families with women affected with breast and/or ovarian cancer and at least one case of male breast cancer. All patients were screened for point mutations and large rearrangements in the BRCA1/2 genes using standard methods 27 . A series of 277 sporadic breast cancer samples came from the Spanish National Cancer Center (Spain) (172 samples), the Cancer Center of Solca (Ecuador) (86 cases), and the University of Pennsylvania (USA) (19). Written consent for tumor analyses and publication was obtained from the patient or their relative. This research study has been performed with the approval of an ethics committee.

Genomic DNA isolation from formalin-fixed paraffin embedded (FFPE) tumors was carried out as previously described 28 . Briefly, two 30-μm sections were obtained from FFPE tumors, treated with xylene, incubated in Glycine Tris-EDTA and NaSCN, and finally digested with proteinase K and purified with phenol chloroform. All sections were previously examined and dissected with a scalpel to ensure at least 70% content of tumor cells.

Comparative genomic hybridization was done onto the 1 megabase (Mb) Bacterial Artificial Chromosome (BAC) clone array platform developed at the University of Pennsylvania. In few words, the platform is composed of 4134 BAC clones spaced at 1 Mb intervals, including a direct coverage of approximately 400 known cancer genes 29 . DNA probe labeling, aCGH protocol, and array data analysis have been described elsewhere 28 . Briefly, normalized aCGH data were analyzed using the Binary Segmentation algorithm implemented in the Insilico CGH software 30 . This algorithm depicts genomic segments showing estimative copy number values in log₂ratio, which are calculated from the mean log₂ratio of all the clones within that segment. As a result, it decreases both the dynamic range of the hybridization and the high noise usually present in hybridizations of DNA from FFPE tissues 31 32 . Therefore, we considered those segments with log₂ratio ≥ 0.1 as gains, whereas those with log₂ratio ≤ -0.1 were categorized as losses. Segments altered with DNA amplifications were considered when log₂ratio ≥ 0.4. A collection of genomic imbalances previously confirmed in breast cancer cell lines and breast tumor samples allowed us to set these thresholds.

FISH analysis was carried out on the original group of six tumor samples with Amp13q34 found in the aCGH studies to verify their amplification levels. Then, we performed the FISH study on two familial and three sporadic breast cancer tissue microarrays (TMA) containing 156 and 258 breast tumor samples respectively, in order to quantify the Amp13q34 rate in a larger collection. From the familial breast cancer series, 44 samples were already included in our aCGH analysis 28 . These TMA have been described elsewhere 9 33 34 . Briefly, tumor areas of the breast tumor samples were carefully selected from hematoxilin and eosin-stained sections and delimited on the individual paraffin blocks. Then, two tissue cores from the selected tumor area of each specimen were obtained and included in the TMA.

The test FISH probe was composed of three BAC clones mapping to the 13q34 chromosomal region, labeled with dUTP-SpectrumOrange (Vysis, Inc. Downers Grove, IL, USA): RP11-391H12 (AL136221.38, mid-position 113.96 Mb), RP11-102K13 (AL160251.29, 114.09 Mb) and RP11-230F18 (AL442125.13, 114.21 Mb). In addition, three BAC clones from the 13q12.11 chromosomal region were used as copy number reference for chromosome 13, labeled with dUTP-SpectrumGreen (Vysis, Inc. Downers Grove, IL, USA): RP11-301J16 (AL137001.23, 19.63 Mb), RP11-408E5 (AL139327.18, 19.77 Mb), and RP11-385E5 (AL356259.11, 19.92 Mb). FISH analysis was done according to Vysis' instructions, with slight modifications. An average of 110 (50 to 200) well-defined nuclei was analyzed by scoring the number of single copy 13q34 and 13q12.11 signals. We considered DNA amplification when > 50% of tumor cells showed more than three times as many 13q34 signals as 13q12.11 copy signals.

Total RNA was obtained from six breast tumors with Amp13q34 (three BRCA1-associated, one BRCA2-associated, one BRCAX-associated, and one sporadic breast tumor samples) and one breast cancer cell line that also carried the amplification (MDA-MB-157); as well as from 13 breast tumors that did not exhibit DNA copy number changes at 13q34. Importantly, 13q34 DNA copy number status for all tumoral specimens had been assessed by either aCGH or FISH analyses. All tumor samples were FFPE material, and RNA isolation was carried out as described before 35 . Briefly, a standard tissue sample deparaffinization was performed using xylene and alcohols. Then, samples were incubated in a digestion buffer (10 mM TrisHCl pH 8.0; 0.1 mM EDTA; 2% SDS and 500 μg/ml proteinase K). Nucleic acids were purified with standard phenol-chloroform-isoamyl alcohol procedures, followed by a precipitation with isopropanol, glycogen and sodium acetate. After DNase treatment (Ambion Inc., Austin, TX, USA) at 37°C for 30 minutes, RNA was finally processed for cDNA synthesis using M-MLV retrotranscriptase enzyme (Invitrogen, Carlsbad, CA, USA).

Assays were designed using the Roche Applied Science Universal Probe Library website 36 (Roche Diagnostics, Basel, Switzerland) for all target genes and endogenous control (see Additional file 1). qRT-PCR assays were set up in triplicates and performed using the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Relative expression was determined using the free access software qBase 37 based on a modification of the classic delta-delta Ct method that allows for PCR efficiency correction. Appropriate positive and negative controls including non-retrotranscribed RNA for each sample were run for the experiment.

To associate genomic copy number aberration at 13q34 with changes in protein levels of the candidate genes, we carried out antibody specific staining on the same TMA used in the FISH analyses.

IHC assays were performed by the Envision method (Dako, Glostrup, Denmark) with a heat-induced antigen retrieval step. TMA sections were immersed in 10 mM boiling sodium citrate at pH 6.5 for two minutes in a pressure cooker. Between 150 and 200 cells per core were scored to determine the percentage of cells with positive nuclei or cytoplasm, depending upon the marker. We evaluated the protein expression level for the two candidate genes within Amp13q34 that were significantly overexpressed in the qRT-PCR analyses: CUL4A and TFDP1. As CUL4A expression in the cytoplasm was found in most of cases, we delineated four levels of staining: 0 (null staining), 1 (low staining), 2 (medium staining), and 3 (strong staining), which we grouped into negative (0), mid-level (1, 2) and positive (3) in subsequent analyses. TFDP1 is localized in the cell nucleus, so the median percentage of positive nuclei in all cases was used as the threshold to discriminate between negative (< 25%) or positive expression (≥ 25%). Tumor samples with more than 60% of stained cells (the median percentage of stained cells in only positive cases) were further categorized as highly expressed for the association with Amp13q34.

Other proteins also were studied to establish potential associations. We had previously evaluated nuclear staining for estrogen receptor (ER), progesterone receptor (PR), p53, Ki-67, cyclins D1, D3, E, and A; p16, p27, p21, CDK1, CDK2, CDK4, Skp2, retinoblastoma protein (Rb), E2F1, E2F6, MDM2, topoisomerase IIα, survivin, and CHEK2; cytoplasmic staining for BCL2, vimentin, cytokeratin 5/6 (CK5/6), cytokeratin 8 (CK8), and cyclin B1; and membrane staining for E-cadherin, P-cadherin, B-catenin and G-catenin 34 . HER-2 expression was evaluated according to the four-category (0 to 3+) DAKO system proposed for the evaluation of the HercepTest, and HER-2 expression of 3+ was the only value considered positive 34 . Antibodies, dilutions, suppliers, and thresholds used for analyses are listed in Additional file 2.

In order to compare the mRNA expression levels between the different groups (tumors with or without Amp13q34), we applied the U-Mann Whitney test. We also determined associations between tumor groups with both clinical features and IHC markers using the chi square test, with the two-tailed Fisher's exact test correction when needed. The statistical software SPSS for Windows (SPSS, Inc., Chicago, IL, USA) was used to perform these statistical comparisons.

We have previously reported that about 5% of both familial and sporadic breast cancers contain a high-level amplification at 13q34 (Amp13q34) using chromosomal- and array-based comparative genomic hybridization techniques (cCGH, and aCGH, respectively) (Table 1 and Figure 1a) 21 28 . Now, we have extended the study of the rate of Amp13q34 to a larger cohort of 414 familial and sporadic breast cancer samples using FISH technique. We have found that Amp13q34 is present in around 4.5% of breast cancer samples (Figure 1b-c), although the rate differs slightly among tumor classes, being higher in BRCA1- (8.1%) than in BRCA2- or non-BRCA1/2- (< 3.0%) associated cancers (Table 1). However, these differences are not statistically significant, maybe due to the low number of cases (data not shown). Noteworthy, the level of amplification in this tumor cohort never exceeded a 13q34:13q12.11 copy ratio of seven, being the median ratio of 4.5 as many 13q34 as 13q12.11 copy signals (Figure 1c). This shows that 13q34 region, albeit that it is affected with amplification, is not altered with the same magnitude as other classical high-level DNA amplification sites, such as 17q12 (ERBB2). As expected, no tumor sample analyzed by both cCGH/aCGH and FISH showed discrepancies in copy number values.

To define the minimal common region of amplification, we used data from a previous aCGH analysis 28 . A genomic representation of those cases with Amp13q34 (three BRCA1-, one BRCA2-, one BRCAX-, one sporadic-breast cancers and one breast cancer cell line, MDA-MB-157) is shown in Figure 2a. Chromosome 13 has two different genomic profiles based on the start position of the genomic gain: a) loss from the centromere to the 13q31 or 13q32 chromosomal band, with gain or amplification extending from 13q31 to the telomere (cases 1, 3, 4, 6, and 7); or b) copy neutral from the centromere to 13q21 and gain or amplification from 13q21 to the telomere (cases 2 and 5). A detailed analysis of the genomic values on each of the studied cases allowed us to narrow down the minimal common region of DNA amplification to 1.83 Mb, mapping entirely within the 13q34 chromosomal band. Two samples delimited the Amp13q34 region: tumor #3 showed the Amp13q34 starting at RP11-520D2 (spanning 3.55 Mb, and 29 genes), whereas tumor #5 finally defined the minimal 1.83 Mb region from RP11-375A8 to the telomere that contains 22 genes (Figure 2b) (Table 2). As tumor material was limited to carry out a whole characterization of all the genes therein located, we selected those related to tumorigenesis or cell transformation according to either previous studies or their hypothetical function. Our final list of selected genes included: ARHGEF7, ATP11A, MCF2L, CUL4A, LAMP1, GRTP1, DCUN1D2, TFDP1, GAS6, RASA3, and CDC16 (bold genes in Table 2). Although ARHGEF7 is not located within the boundaries of the 1.83 Mb region, it was selected to check a possible correlation for those genes located outside the minimal region, but inside the 3.55 Mb defined by the case #3. Noteworthy, the 13q34 FISH probe used in this study covered CUL4A, LAMP1, GRTP1, DCUN1D2, and TFDP1 (Figure 2 and Table 2).

Once the minimal amplification site had been delineated, qRT-PCR was carried out to determine the mRNA expression level of each of the 11 candidate genes, comparing two cohorts of FFPE samples: tumors with Amp13q34 (six tumor samples and MDA-MB-157), and tumors without any genomic aberration at 13q34 region (13 cases). ARHGEF7 and ATP11A showed a non-statistically significant trend to be overexpressed in Amp13q34 tumors (P = 0.242 and P = 0.191, respectively); whereas GAS6 and RASA3 had non-significant tendencies to be downregulated in tumors with Amp13q34 (P = 0.285 and P = 0.052, respectively) (Figure 3). Nevertheless, two genes had significant association between amplification and overexpression: CUL4A (P = 0.007) and TFDP1 (P = 0.019) (Figure 3). At this point, we focused on these two genes as candidate drivers of the 13q34 amplification.

We performed immunohistochemical analyses of the two candidate genes (CUL4A and TFDP1) on a set of 414 breast cancer samples on TMAs, previously analyzed by FISH (see above). Interestingly, the 75% of the tumors with Amp13q34 showed a strong staining for CUL4A, compared with 34.3% of the tumors without Amp13q34, but overexpressing CUL4A (P = 0.010) (Figure 4a-c and Table 3). A significant association was also found between Amp13q34 and TFDP1, since the 92.3% of the tumors with Amp13q34 (12/13) presented TFDP1 overexpression (P = 0.004), whereas the 51.9% of the tumors without Amp13q34 overexpressed TFDP1 (Figure 4d-f and Table 3). Furthermore, when considering a higher percentage (> 60%) of TFDP1 positive cells as the threshold, 76.9% of Amp13q34 tumors had TFDP1 overexpression as compared with 20.1% of the tumors without Amp13q34 (P = < 0.001) (Table 3). Interestingly, 60% of tumors with Amp13q34 co-overexpressed both CUL4A and TFDP1 (6 out of 10 tumors), although it was not statistically significant maybe due to the low number of samples (data not shown). Altogether, these findings demonstrated that Amp13q34 was significantly associated with a higher expression level of both CUL4A and TFDP1, but given that there are cases without the amplification showing protein overexpression, Amp13q34 was not the exclusive mechanism causing overexpression of these genes.

In addition, we associated Amp13q34 with other immunohistochemical and clinical tumor features previously analyzed on a subset of the sample set (260 out of 414 samples) 33 34 . We observed that tumors with Amp13q34 were characterized by high histological grade; an absence of expression of ER, PR, CCND1, RB, p16, and CK8; and an overexpression of Ki-67, EGFR, P-Cadherin, G-catenin, CCNE, CCNB1, SKP2, survivin, vimentin, and CK5 (Table 4). As most of these markers are associated with basal-like phenotype of breast cancer, these associations may be due to the unique presence of the Amp13q34 in basal-like tumors as we previously reported 11 . CUL4A-overexpressing tumors had higher grade, absence of expression of PR and BCL2, and a higher percentage of cases expressing EGFR, P-Cadherin, CCNB1, SKP2, and CK5; when compared with tumors that did not overexpress CUL4A (Table 4). Last, TFDP1-overexpressing tumors were associated with overexpression of CCNE, CCNB1, and p16 (Table 4). However, no significant association was found between TFDP1 and CUL4A expression levels when studying the overall breast cancer sample set (data not shown). These associations support that the overexpression of CUL4A and TFDP1, even though it is correlated with Amp13q34, is not exclusive of basal-like tumors, but still relates to higher tumor aggressiveness.