Functional characterization of the 19q12 amplicon in grade III breast cancers

Fresh/frozen primary invasive breast carcinomas from 313 patients were retrieved from independent consecutive cohorts after approval by local Research Ethic Committees from the authors' institutions. None of the patients received pre-operative chemotherapy and/or endocrine therapy. Tumours were either micro-dissected to ensure a tumour cell content > 75% as previously described [9], or one representative section of each tumour was stained with haematoxylin & eosin and only samples with > 70% of neoplastic cells were included, as previously described [19]. A description of the cohort analysed here is presented in Additional file 1 Table S1. Out of the samples included in this study, the aCGH profiles of 95, 24 and 8 cases were previously reported in Natrajan et al. [9], Turner et al. [19], and Hungermann et al. [30], respectively. The remaining samples comprise a series of 31 consecutive HER2-positive breast cancers retrieved from the tissue bank of The Netherlands Cancer Institute (NKI), and a collection of 119 consecutive breast cancers from the tissue bank of the NKI. RNA was available for 48 cases and the gene expression profiles were reported in Natrajan et al. [29]. RNA of sufficient quantity and quality was available for 48 cases and the gene expression profiles were reported in Natrajan et al. [29]. Immunohistochemistry for ER, progesterone receptor (PR) and HER2 was performed as previously described [9, 31]. Histological grade was determined by two pathologists (FCG and JSR-F) using the modified Scarff-Bloom-Richardson system [32]. ER and PR status was determined by immunohistochemical analysis using the 6F11 (1:150) and PgR636 (1:200) antibodies, respectively, as previously described [33]. The Allred scoring system was employed and tumours were considered positive when the score was ≥ 3. HER2 gene amplification was defined based on chromogenic in situ hybridisation analysis using a US Food and Drug Administrator approved probe (SpotLight HER2 amplification probe, Invitrogen, Carlsbad, CA, USA) and/or by inspection of the results of aCGH analysis, as previously described [20]. Tumours were classified into ER-positive/HER2-negative, ER-negative/HER2-negative and HER2-positive subgroups, given i) the results of recent studies demonstrating that the transcriptomic profiles of ER-positive/HER2-negative, ER-negative/HER2-negative and HER2-positive tumours are fundamentally different [34, 35], ii) the technical issues related to the assignment of tumour profiled with different platforms into the 'intrinsic' molecular subtypes [36, 37], and iii) that these subgroups are currently employed to define the systemic therapy for patients with breast cancer.

DNA and RNA were extracted as previously described [29]. DNA concentration was measured with Picogreen^® (Invitrogen, Paisley, UK) according to the manufacturer's instructions, and RNA was quantified using the Agilent 2100 Bioanalyser with RNA Nano LabChip Kits (Agilent Biosystems, Stockport, UK).

Fifty-six breast cancer cell lines were obtained from ATCC (LGC Standards, Teddington, UK) unless otherwise specified (Additional file 2 Table S2), maintained as previously described [16, 19, 38] and subjected to aCGH analysis as described below. Out of these cell lines, MDA-MB-157, HCC1569, MDA-MB-231, Hs578T, MCF7, ZR75.1, BT474 and JIMT1 were used as models for the functional characterisation of the 19q12 amplicon.

The 32K bacterial artificial chromosome (BAC) re-array collection (CHORI) tiling path aCGH platform was constructed at the Breakthrough Breast Cancer Research Centre, as described previously [33, 39]. This type of BAC array platform has been shown to be as robust as, and to have comparable resolution with, high density oligonucleotide arrays [40‐42]. DNA labelling, array hybridisations and image acquisition were performed as previously described [9]. aCGH data were pre-processed and analysed using an in-house R script (BACE.R) in R version 2.13.0, as previously described [29, 33]. In brief, raw Log₂ ratios of intensity between samples and pooled female genomic DNA were read without background subtraction and normalised in the LIMMA package in R using PrinTipLoess. Outliers were removed based upon their deviation from neighbouring genomic probes using an estimation of the genome-wide median absolute deviation of all probes. After filtering polymorphic BACs a final dataset of 31,367 clones with unambiguous mapping information according to the February 2009 build (hg19) of the human genome [43]. Log₂ ratios were rescaled using the genome-wide median absolute deviation in each sample, and then smoothed using circular binary segmentation in the DNACopy package as previously described [29, 33]. Loss was defined as a circular binary segmentation (cbs)-smoothed Log₂ ratio < -0.08. Low-level gain was defined as a cbs-smoothed Log₂ ratio in the range 0.08 to 0.45, corresponding to approximately three to five copies of the locus, whilst gene amplification was defined as having a Log₂ ratio > 0.45, corresponding to more than five copies [9, 33, 39]. Threshold values were determined as previously described [9] and validated empirically by means of in situ hybridisation methods in previous publications [9, 33, 39, 44‐46]. A categorical analysis was applied to the BACs after classifying them as representing amplification (>0.45), gain (>0.08 and ≤ 0.045), loss (< -0.08), or no-change according to their cbs-smoothed Log₂ ratio values [29, 33]. Threshold values were determined and validated as previously described [9]. aCGH data, the analysis history, script and code are publically available online at [47].

Fluorescence in situ hybridisation (FISH) analysis for 19q12 amplification was carried out as previously described [9, 48]. Briefly, a FISH probe mapping to 30.10 to 30.25 Mb on chromosome 19 was generated using the BAC RP11-372I05 and biotin labelled as previously described [48]. Pre-treatment and hybridisation were carried out as previously described [9, 48]. Cases were considered amplified if > 50% of neoplastic cells harboured large signal clusters or > 5 signals/nucleus. FISH performed with observers blinded to the results of aCGH analysis (Additional file 3 Figure S1).

The entire coding sequence of the TP53 gene (exons 2 to 11) was sequenced as previously described [39, 49]. PCR amplification and Sanger sequencing was carried out in the 16 cases with 19q12 amplification using 5 ng of tumour DNA [50] using the DNA Sequencing Kit BigDye Terminator v 3.1 Cycle Sequencing Ready Reaction Mix (Applied Biosystems, Warrington, UK), as described [39]. Sequences were analysed with Mutation Surveyor software (Softgenetics, State College, PA, USA).

Gene expression profiling of breast cancer samples in the 'Natrajan' dataset was performed using the Illumina human WG6 version 2 expression array as previously described [29]. Raw gene expression values were robust-spline normalised using the Bioconductor lumi package in R (R Foundation, Vienna, Austria) Genes were mapped to their genomic location using the lumiHumanAllv2 annotation database available from Bioconductor. Only Illumina transcript probes with detection P-values < 0.01 in > 25% of samples were included; this resulted in a dataset of 12,699 transcriptionally regulated probes with accurate and unequivocal mapping information. Gene-expression data are publicly available at ArrayExpress (accession number: E-TABM-543 [51]. The final dataset comprises 48 cases, whose clinicopathological characteristics are fully described elsewhere [29].

To identify genes whose expression levels correlate with copy number changes, cbs-smoothed Log₂ ratios from aCGH data were used to assign the aCGH states for each of the genes in the gene expression dataset using the median values for all BACs that overlap with the genomic position of each gene. This resulted in a 1:1 matrix of expression values and aCGH cbs values, which were used for downstream statistical analysis. To define genes that were overexpressed when amplified, a Mann-Whitney U test was performed using categorical aCGH states (that is, amplification versus no amplification) as the grouping variable and the expression of genes as the dependent variable, as previously described [16, 29]. P-values < 0.05 were considered significant.

Nine genes that mapped to the smallest region of amplification on 19q12 were selected for functional evaluation, namely CCNE1, UQCRFS1, POP4, PLEKHF1, C19ORF2, C19ORF12, VSTM2B, ZNF536, TSHZ3. Each gene was targeted with four distinct siRNAs (siGENOME SMARTpool) obtained from Thermo Scientific (Epsom, UK): siCCNE1 - pool M-003213-02; siUQCRFS1 - pool M-020100-01; siPOP4 - pool M-020046-00; siPLEKHF1 - pool M-018423-01; siC19ORF2 - pool M-017399-01; siC19ORF12 - pool M-014731-01; siVSTM2B - pool M-023625-01; siZNF536 - pool M-020506-01; siTSHZ3 - pool M-014119-01; and siCDK2 - pool M-003236-04. siGENOME Non-Targeting siRNA Pool #1 and #2 (siCON, D-001206-13 and D-001206-14) were used as controls. As a positive control (that is, a gene whose silencing is lethal) we employed siRNA pools for PLK1 (M-003290-01) as previously described [44]. To identify conditions that maximise transfection efficiency and minimise lipid-mediated toxic effects, multiple transfection conditions for each cell line were tested. Lipofectamine RNAiMax (Invitrogen, Paisley, UK) and Dharmafect 4 (Thermo Scientific, Epsom, UK) were identified as the most efficient and least toxic for the cell lines studied. Breast cancer cells were transfected with target and control siRNAs (50 nmol/L per well in 100 μL total volume) in 96-well plates, according to the manufacturers' instructions as previously described [44, 52]. A total of 1,000 to 5,000 cells were seeded per well that yielded 80 to 90% confluency in the controls at 10 days. Media were replaced every two days and cell viability was assessed using the CellTiter-Glo^® assay (Promega, Southampton, UK) as previously described [44]. The average cell survival fraction for each siRNA was calculated relative to that of cells transfected with non-coding control siRNA.

For assessment of drug sensitivity, cell lines were plated and transfected with target or control siRNAs in 96-well plates. At 48 hours post transfection, media were supplemented with serial dilutions (10^-4M to 10^-9M) of doxorubicin, cisplatin, paclitaxel, or the CDK1, CDK2 and CDK9 inhibitor AZD5438 (Tocris Biosciences, Bristol, UK), essentially as previously described [44]. All experiments were performed in triplicate. Survival was assessed with CellTiter-Glo^® cell viability assay after seven days of drug treatment and survival fraction compared to siRNA vehicle treated cells. Survival curves and estimated SF₅₀ (the drug concentration used following which 50% of cells survive) were calculated using non-linear regression with GraphPad Prism V5.0 (La Jolla, CA, USA).

FACS analysis for cell cycle assessment was performed with Propidium Iodide staining five days after transfection as described previously [52].

Total protein lysates (40 μg) from treated and untreated cells were separated by SDS-PAGE according to standard protocols, and immunoblotting was carried out using primary antibodies directed against CCNE1 (HE-12, Abcam ab3927, Cambridge, UK), CDK2 (2546, Cell Signaling Technology, Danvers, MA, USA), phospho-CDK2 (2561, Cell Signaling), PLEKHF1 (20389-1-AP, Proteintech, Chicago, IL, USA), POP4 (Rpp29, sc-23048, Santa Cruz Biotechnology, Santa Cruz, CA, USA), TSHZ3 (sc-134132, Santa Cruz), and CCNB1 (4135, Cell Signaling), and against β-tubulin as loading control (5346, Cell Signaling). Protein expression levels after siRNA silencing were assessed by western blotting 72 hours after transfection of the cells and protein bands quantified using ImageJ 1.44p software (NIH, Bethesda, MD, USA) [53].

First strand synthesis was performed as previously described [9], and qRT-PCR was performed using TaqMan chemistry on the ABI Prism 7900HT (Applied Biosystems), using the standard curve method. Assays for CCNE1, PLEKHF1, UQCRFS1, C19ORF12, C19ORF2, VSTM2B, ZNF536, TSHZ3, TFRC and MRPL19 were purchased from Applied Biosystems. Expression levels were normalised to those of TFRC and MRPL19 (Assay on demand ID: Hs00174609_m1-TFRC, Hs00608522_g1-MRPL19, Hs01026536_m1-CCNE1, Hs00759096_s1-PLEKHF1, Hs00705563_s1-UQCRFS1, Hs01026936_m1-C19ORF2, Hs00198357_m1-POP4, Hs01107514_m1-C19ORF12, Hs0416833_m1-VSTM2B, Hs00206981_m1-ZNF536, Hs01583885_m1-TSHZ3, Hs01548894_m1-CDK2, Hs99999188_m1-CCNB1). For the assessment of RNA levels after gene silencing, RNA was extracted using Trizol (Invitrogen, Paisley, UK), and transcript levels were assessed 48 hours after transfection. The level of siRNA silencing was measured relative to that of cells transfected with non-coding control siRNA.

To determine the clinicopathological correlates of the 19q amplification, 313 frozen breast cancers were subjected to aCGH analysis [9]. Amplification at 19q12 (27.8 to 32.9 Mb) was found in 16/313 (5.1%) of all tumours, and 7.8% (16/188) of grade III cancers (Figure 1A-C; Table 1). Amplification of 19q12 was significantly associated with histological grade III (Fisher's exact test P = 0.018, Table 1), lack of ER expression (Fisher's exact test, P = 0.0042) and ER-negative/HER2-negative subtype (Chi-square test P = 0.0056, Table 1). Given the small number of samples in the subgroup analysis of the clinical subtypes, we have compared the prevalence of 19q12 amplification in ER-negative/HER2-negative vs the other subtypes using the Fisher's exact test, which confirmed a significance association between 19q12 amplification and ER-negative/HER2-negative breast cancers (P = 0.0024).

Amplification of the 19q12 locus was shown to be complex and this amplicon displayed multiple core regions recurrently amplified in primary breast cancers (Figure 1C). Within these cores, nine genes were identified, including CCNE1 (Table 2). Although CCNE1 has been proposed as a driver of this amplicon in breast cancer [14, 26], CCNE1 was amplified in only 5 out of the 16 primary breast cancers harbouring amplification of the 19q12 locus (Figure 1C).

Genes within an amplicon that are overexpressed when amplified have been shown to constitute potential amplicon drivers and therapeutic targets [9‐12, 17‐19, 29, 54]. To determine which of the amplified genes in the 19q12 amplicon were significantly overexpressed when amplified, aCGH and microarray expression data from 48 microdissected grade III breast cancers were interrogated [29, 36]. This analysis revealed that out of all the genes mapping to the 19q12 amplification with detectable gene expression, UQCRFS1, POP4, PLEKHF1, C19ORF12, CCNE1 and C19ORF2 had expression levels that significantly correlated with gene copy number (Pearson's correlation r value = 0.65, 0.77, 0.44, 0.63, 0.66, 0.81, respectively; all P < 0.05) and UQCRFS1, POP4, C19ORF12, CCNE1 and C19ORF2 were significantly overexpressed when amplified (Mann-Whitney U test P < 0.05) (Figure 1D, Table 2). VSTM2B expression was undetected in any of the tumours analysed.

To identify models for the study of the genes overexpressed when amplified in the 19q12 amplicon, we first subjected 56 breast cancer cell lines to aCGH analysis and identified two cell lines (that is, HCC1569 and MDA-MB-157), which harbour amplification of 19q12 (minimal amplified region from 29.17 to 32.82 Mb). In a way akin to the majority of primary breast cancers harbouring 19q12 amplification [9], the only cell lines harbouring amplification of this locus were ER-negative/HER2-negative (MDA-MB-157) or ER-negative/HER2-positive (HCC1569) [16, 38]. A panel of phenotypically matched (in terms of ER, PR and HER2 expression) breast cancer cell lines were selected and used as controls (that is, MDA-MB-231, Hs578T, MCF7, ZR75.1, JIMT1 and BT474; Additional file 2 Table S2). Quantitative RT-PCR of cDNA from cancer cells harbouring 19q12 amplification or lacking amplification of this locus revealed that UQCRFS1, POP4, CCNE1 and C19ORF2 genes were overexpressed when amplified (P < 0.05, Mann Whitney U test, Figure 1E), whereas VSTM2B, ZNF536 and TSHZ3 expression could not be detected in the majority of cell lines at the mRNA level.

Taken together, our results suggest that UQCRFS1, POP4, C19ORF12, CCNE1 and C19ORF2 are overexpressed in primary breast cancers and/or breast cancer cell lines harbouring their amplification, and may constitute potential drivers of this amplicon.

There are several lines of evidence to suggest that one of the characteristics of amplicon drivers is that their expression is selectively required for the survival of cancer cells harbouring their amplification [9‐12, 19]. Given the complexity of amplification at this locus, we chose to assess the potential amplicon drivers by siRNA silencing of all genes mapping to the 19q12 amplicon (that is, UQCRFS1, VSTM2B, POP4, C19ORF12, PLEKHF1, CCNE1, C19ORF2, ZNF536 and TSHZ3) regardless of their expression levels. Silencing of CCNE1, PLEKHF1, POP4, ZNF536 and TSHZ3 expression selectively reduced cell viability in cancer cells harbouring 19q12 gene amplification but had a significantly lower impact on the survival of breast cancer cells lacking amplification of this locus (t-test P < 0.05; Figure 2). Upon siRNA pool deconvolution, the observation that silencing of CCNE1, PLEKHF1, POP4 and TSHZ3 is selectively lethal in cancer cells harbouring their amplification was validated (that is, two or more individual siRNAs were selectively lethal in cancer cells harbouring amplification of the gene silenced; Figure 3A). siRNA silencing of these genes was confirmed at the mRNA and protein levels (Figures 3B, C).

Assessment of the protein levels of POP4, PLEKHF1, CCNE1 and TSHZ3 was performed in the cell lines used in this study. Expression levels of POP4, PLEKHF1 and CCNE1 were significantly higher in cancer cells harbouring amplification of the genes encoding these proteins than in cancer cells devoid of amplification of these loci (Figure 4, Mann-Whitney U test P < 0.05). No significant association between TSHZ3 gene amplification and protein expression was observed, even when corrected for proliferation levels, as defined by CCNB1 expression levels (data not shown).

To test whether silencing of these genes affects cell cycle in cancer cells with or without 19q12 amplification, DNA content was measured upon gene silencing. Upon PLEKHF1, POP4 or TSHZ3 siRNA-mediated gene silencing, slight but not statistically significant increases in the subG1 fraction and decreases in the G2/M fraction were observed in cancer cells with (that is, MDA-MB-157 and HCC1569) or without (that is, MDA-MB-231, JIMT1, ZR75.1 and BT474) 19q12 amplification (Additional file 4 Figure S2). These observations suggest that the silencing of these genes does not significantly alter the cell cycle profile. On the other hand, in cancer cells harbouring 19q12 amplification, CCNE1 silencing resulted in a significant decrease in the fraction of cells in S/G2 and increase in the proportion of cells in G1 and subG1 cell cycle phases (P < 0.0001, Chi-square test, Figure 5A).

To assess whether amplification and overexpression of CCNE1, PLEKHF1, POP4 and TSHZ3 would alter the response of breast cancer cells to common chemotherapeutic agents, we subjected cancer cells with or without 19q12 gene amplification to treatment with doxorubicin, cisplatin and paclitaxel upon silencing of these genes. No significant difference in the sensitivity of these cells to the chemotherapy agents tested was observed upon silencing of PLEKHF1, POP4 or TSHZ3 compared to non-targeting siRNAs (data not shown). CCNE1 siRNA-mediated silencing, however, was found to reduce the sensitivity of cancer cells to doxorubicin, cisplatin and paclitaxel in cancer cells harbouring 19q12 amplification cells but not in cancer cells lacking amplification of this locus (Additional file 5 Figure S3). This is not surprising, given that these chemotherapy agents target cycling cells and, here, we demonstrate that CCNE1 siRNA-mediated silencing leads to G1 arrest.

Given that i) cancer cells harbouring CCNE1 gene amplification depend on CCNE1 expression for their survival, ii) that one of the main functions of CCNE1 is to activate CDK2 in the transition from G1/S phase [55], and iii) that CCNE1 RNAi-mediated silencing led to G1 arrest, we posited that cancer cells harbouring CCNE1 gene amplification would be dependent on CDK2 expression and kinase activity for their survival. siRNA-mediated CDK2 silencing resulted in a significantly higher reduction in cell survival of cancer cells harbouring CCNE1 gene amplification than in cancer cells devoid of amplification of this gene (t-test, P < 0.05, Figures 5B, C). Furthermore, cancer cells harbouring CCNE1 gene amplification displayed a higher sensitivity to the CDK1, CDK2 and CDK9 small molecule inhibitor AZD5438 [56, 57] than cancer cells lacking CCNE1 gene amplification (t-test, P = 0.019, Figure 5D). Moreover, primary breast cancers harbouring CCNE1 amplification were found to have higher levels of CDK2 mRNA expression, adjusted for proliferation using the mRNA levels of CCNB1 [58], than breast cancers devoid of CCNE1 amplification (P = 0.02789, t-test, data not shown).