Breast cancer oestrogen independence mediated by BCAR1 or BCAR3 genes is transmitted through mechanisms distinct from the oestrogen receptor signalling pathway or the epidermal growth factor receptor signalling pathway

The oestrogen-dependent human breast cancer cell line ZR-75-1 was maintained in RPMI 1640 medium supplemented with 10% bovine calf serum and 1 nM 17 β-oestradiol (R/BCS/E2) as described previously [16]. The derived EGFR-transfectant cell line ZR/HERc(1A) [31], hereafter referred to as ZR/EGFR, a BCAR1-transfectant cell line (4A12) [18] and a BCAR3-transfectant cell line (B3-10) [17] were also maintained in R/BCS/E2 medium. For short-term induction experiments, cells were cultured for 4 days in regular medium lacking added oestrogen (R/BCS) in 162 cm² flasks, given fresh R/BCS medium 24 hours before manipulation, and cultured for 6 hours in the presence of 100 nM oestradiol or 1 μM ICI 164,384 (or ethanol vehicle alone) in R/BCS medium. Hormones and anti-hormones were provided by N.V. Organon (Oss, The Netherlands). For long-term induction experiments, cells were grown for 7 days in R/BCS/E2 medium or R/BCS medium containing 10 ng/ml EGF (Roche Diagnostics Nederland B.V., Almere, The Netherlands). Medium was replaced after 3 days and at 24 hours before harvest. After completion of the culture, the medium was removed and cells were lysed directly with 16–20 ml of RNAzol B solution (Campro Scientific, Veenendaal, the Netherlands). RNA was prepared as described by the manufacturer, quantified and checked for integrity on agarose gels. Poly(A)⁺ mRNA was prepared from pooled total RNA samples of two independent cultures by two cycles of binding to oligo(dT) with the use of OligoTex (Qiagen/Westburg B.V., Leusden, The Netherlands), and checked for integrity and contamination with ribosomal RNA by capillary electrophoresis (Lab-on-a-Chip, Agilent Technologies 2100 bioanalyzer, Amstelveen, The Netherlands).

Production of Cy5-labelled and Cy3-labelled cDNA from the purified mRNA, hybridisation of the UniGEMV cDNA microarrays and quantification of the signals were performed by Incyte Genomics (Mountain View, CA, USA) as described previously [32]. Two batches of UniGEMV2 microarrays were used for these experiments. All hybridisations (Table 1) were performed in duplicate with a fluor reversal to minimise possible bias caused by the molecular structure of the Cy3 and Cy5 dyes. Data analysis was performed with the Rosetta Resolver software package (v 3.2) with an Incyte/UniGEM microarray error model (Rosetta Inpharmatics Inc., Kirkland, WA, USA). Genes exhibiting at least once a significant difference (P ≤ 0.01) in expression in these experiments were used for further analysis (n = 2373). In addition to the actual measured gene expression ratios we calculated the expression ratios between different experimental conditions from two measurements that contained a common sample [33]. Because the calculated gene expression ratios were in good agreement with available actual measurements, we used these calculations as 'virtual experiments'.

For hierarchical clustering of the measured and calculated expression ratios, we used Resolver software (average linkage agglomerative clustering using Euclidean distance and weighted by error) and Spotfire Decision Site 7.1 analysis package (Spotfire Inc., Somerville, MA, USA) using the Unweighted Paired-Group Method with Arithmetic mean (UPGMA) and Pearson's correlation as a similarity measure. Information on the function of genes has been retrieved from various public databases (for example PubMed, OMIN, GENECARD, KEGG and GO) and from the LifeSeq Gold database (Incyte Genomics). Expression data publicly available from prostate cancer, breast cancer and cell lines were linked to our data by means of the Unigene cluster number.

For quantitative reverse transcriptase-mediated polymerase chain reaction (RT–PCR), 2.5 μg of total RNA (or 100 ng of poly(A)⁺ mRNA), 0.8 μg of oligo(dT)_12–18 (Invitrogen Corporation) and 0.5 μg of random hexamer (Pharmacia) in 20 μl of RNase-free water were heated for 5 min at 65°C and cooled on ice. The final reaction of 40 μl contained 0.4 mM dNTPs (Pharmacia), 60 units of RNAseOUT and 300 units of SuperscriptII Reverse Transcriptase (Invitrogen). Incubations were for 2 min at 0°C, 10 min at 25°C, 50 min at 42°C and 10 min at 55°C; the reaction was stopped by heating for 15 min at 75°C. RNA was destroyed by treatment with 2 units of RNAseH (Promega) for 30 min at 37°C. cDNA products were diluted to 100 μl with 10 mM Tris/HCl pH 7.5; these stocks were stored at -80°C. Forty cycles of amplification of 5 μl of cDNA stocks in distilled water (Invitrogen) diluted 1:19 were performed with an SYBR green PCR mix (Applied Biosystems or Stratagene) and 0.33 μM forward and reverse primers in a volume of 25 μl on a ABI Prism 7700 (Applied Biosystems) in accordance with the recommended protocol. Primer annealing was performed at 60 or 62°C. A dilution series (1:4 to about 1:10,000) of a reference cDNA pool (mixture of cDNA preparations of RNA derived from six different cell lines) was used for normalising gene expression. The intron-spanning gene primers used are listed in Table 2. The cycling conditions were as follows: denaturation for 10 min at 95°C; 40 cycles (15 s at 95°C, 30 s at 60°C (CTSD, TFF1 and MYC) or 62°C, 10 s ramping to 72°C, 20 s at 72°C, 10 s ramping to 79°C, 20 s at 79°C). Data were collected at 72 and 79°C and were analysed at 79°C. At the end of the amplification, the melting curve of the products was determined. PCR products showed discrete melting curves and specific bands of correct lengths on agarose gels after 40 cycles of amplification. We also used Assays-on-Demand™ (Applied Biosystems) for various genes, in accordance with the manufacturer's protocol. cDNA (5 μl, diluted 1:19 or 1:39) was measured in 25 μl reactions with the TaqMan Universal PCR master Mix. All cDNA samples were normalised for HPRT1 levels (four independent measurements) and are presented relative to the gene level in ZR-75-1 cells maintained in R/BCS medium.

To evaluate the effects of oestrogen, of EGF and of previously identified BCAR genes involved in oestrogen-independent growth of the human breast cancer cell line ZR-75-1, we determined the global gene expression in these cells by using UniGEMV2 cDNA microarrays. We performed direct comparisons of mRNA samples without the use of a general reference RNA sample (Table 1). Of the approximately 9000 sequences (about 88% represent known genes according to UNIGENE build no. 160) present on the microarray, the expression of 2373 genes (i.e. 26% of total sequences) was significantly (P ≤ 0.01) affected by either the oestrogen treatment or the EGF treatment or the BCAR transfections. Hierarchical clustering distributes the expression profiles according to the experimental culture conditions; that is, profiles of long-term oestrogen-treated cells were separated from short-term oestrogen-treated cells or BCAR-transfected cells (Additional file 1). The majority of the large changes in gene expression were observed in the oestrogen-stimulated or EGF-stimulated cultures. Hybridisation profiles of short-term oestrogen stimulation of ZR-75-1 cells and BCAR3-transfected cells are very similar and distinct from hybridisation profiles of long-term oestrogen-stimulated cell lines (Additional file 1). The EGF-stimulated ZR/EGFR cells as well as the BCAR1- and BCAR3-transfected cells exhibit clearly different profiles in comparison with the oestrogen-stimulated cultures. The comparison of BCAR-transfected cell lines with each other or with non-stimulated parental cells revealed modest changes in gene expression, indicating that gene expression differences in these BCAR cell lines are generally subtle. Below we discuss the expression profiles (average gene expression log(ratios) of the independent experiments) of a selection of 1006 genes exhibiting a |DE| ≥ 1.60 in at least one of the actual or virtual experiments.

ZR-75-1 breast cancer cells are completely dependent on oestrogen for growth. In standard medium without added oestrogen, growth is strongly reduced. Addition of anti-oestrogen completely abolishes the growth of these cells [16]. Oestrogen-induced cell proliferation is mediated by the transcription activation function of the ERα. To identify the early effects (that is, the transcription targets) of oestrogen stimulation, the expression profile was analysed after a 6-hour high-dose pulse of 100 nM oestrogen. From Fig. 1 and Additional file 1 it is clear that limited changes in gene expression (115 genes with |DE| ≥ 1.60) have occurred during this short treatment compared with mock-stimulated cultures. Over 75% of these genes seemed to be induced. Among these genes are well-known oestrogen targets such as TFF1 (PS2), CSTD (cathepsin D), CCND1 (cyclin D1) and PGR (progesterone receptor). These and several novel genes were rapidly induced by oestrogen both in the parental ZR-75-1 cells and in the BCAR3-transfected cells (Fig. 1 and Additional files 1 2 3). An extended picture emerges after continuous exposure to the regular dose of 1 nM oestrogen. About 400 genes exhibit consistent changes (at least 1.6-fold) in expression, of which about 60% of the sequences exhibit a significant decrease of gene expression (up to sevenfold) and 40% are increased (up to more than 10-fold). The genes specifically modulated by oestrogen comprise members of all functional compartments and processes in the cell. One-quarter of the early-induced genes remain expressed (DE > 1.60) during continuous exposure to oestrogen (see Fig. 1), whereas the expression of others is turned off (for example AMD1, BCL2, CCND1, GJA1, MEIS3, RIP140, RUNX1 and STC1) or even downregulated (for example HIF1A, IL6ST, MYB, PC4 and UGT2B7). Definitions of these and other and other gene names can be found in Additional file 3. Statistical analysis of all 1006 genes shows a clear positive correlation between early-induced gene expression and genes expressed after 7 days of oestrogen treatment (r _s = 0.36, P < 0.0001).

We have previously shown that the addition of EGF to the culture medium does not support growth of ZR-75-1 cells because of the absence of EGF receptors [31, 34]. The introduction of EGFR into ZR-75-1 cells (ZR/EGFR) permits a response to EGF and can support proliferation independently of oestrogen [31]. Gene expression of ZR/EGFR cells stimulated with EGF for 7 days was directly compared with that of ZR-75-1 cells stimulated with oestrogen continuously. It is clear from this direct comparison that 247 genes are specifically altered more than 1.6-fold (Additional file 1). From the virtual experiment, which compares the EGF stimulation of ZR/EGFR cells with unstimulated ZR-75-1 cells, we can conclude that EGF modulates a large cohort of genes (707) at least 1.6-fold (Fig. 1). Statistical analysis of all 1006 genes shows a strong positive correlation for gene expression regulated by EGF and long-term oestrogen (r _s = 0.67, P < 0.0001), but no significant association with genes induced early by oestrogen (r _s = 0.16). As expected, EGFR is one of the most prominently changed genes in our analysis as a consequence of the transgene expression. In addition, the expression of genes implicated in signalling processes, cell adhesion and structure, protein modification, transport and metabolic processes is specifically regulated by treatment with EGF or shows a pronounced alteration in comparison with that in oestrogen-treated cultures (Fig. 1).

We have previously shown that stable overexpression of BCAR1 or BCAR3 induces cell proliferation independently of oestrogen and anti-oestrogen [17, 18]. In an attempt to pinpoint the effects of these cytoplasmic signalling molecules on global gene expression in cultures without added oestrogen, we compared BCAR3 and BCAR1, and BCAR3 and ZR-75-1, directly on microarrays, and calculated the gene expression relation between BCAR1 and ZR-75-1 as a virtual experiment. A total of 79 genes exhibited consistent differences in expression of at least 1.6-fold (Fig. 1). Few genes are modulated solely by the overexpression of a BCAR gene; most are also a target for hormonal and/or EGF stimulation (Fig. 1 and Additional file 1). As expected, the largest observed difference (up to 15-fold) in these comparisons was derived from the BCAR3 transgene expression. No cDNA sequence corresponding to the BCAR1 gene was present on this microarray.

Changes in gene expression specifically caused by BCAR3 overexpression in ZR-75-1 cells were seen for CSTA, DLG7, FMOD, FOLR1, FOXJ1, HSPC242, IGFBP5, LGALS1, LIV1, NEDD4L, NELL2, PCDH7 and UBE2C. In addition, a clear induction of several genes involved in glucose metabolism (PGK1, LDHA, TPI1, and moderate induction levels of ALDOC, ENO1, ENO3 and PFKP; Fig. 1) was observed. This coherent change in gene expression is unlikely to represent a culture artefact because no expression change was observed in these genes in BCAR3-transfected cells after 6 hours of induction with oestrogen (Additional file 1). Specifically altered genes in BCAR1-transfected ZR-75-1 cells include APOD, ASS, CRIP1, ELL2, FOXM1, HSPG2, ID1, IL1R1, IL6ST, LGALS8, PDZK1, TK1, PPP3CA, TOMM20, VIPR1 and various genes encoding ribosomal proteins (Fig. 1). A moderately opposing direction of gene expression change in these two transfectant cell lines compared with ZR-75-1 cells was indicated by TFF3 and MGP (Fig. 1). A set of genes (including BF, BCL2, BIRC5, CDKN1A, INADL, KIAA0101, MAPKAPK2, MYB, P2RY10, PC4, PRCP, OCLN, RAB6KIFL (= KIF20A), SERPINA5, SLC7A2 and UGT2B11) exhibited differential expression in both BCAR-transfected cell lines when compared with the parental cell line ZR-75-1.

To establish that the measured differences in gene expression on the microarrays did indeed reflect the concentration of the respective mRNAs, we performed quantitative RT–PCR (Q-PCR) on a selection of 22 genes on the same RNA samples and on additional RNA preparations from different culture conditions. Standard quantities of intact total RNA or mRNA were reverse transcribed and subjected to Q-PCR. To normalise the cDNA samples we used HPRT1 (not present on the microarray) as a reference. Results have been presented relative to non-stimulated ZR-75-1 cells to facilitate direct comparison with microarray data (Table 3). In general, we found good agreement between the levels of HPRT1 and another housekeeping gene (HMBS) in our experimental samples. The expression of MYC was fairly constant in our series, with the exception of a slight decrease in EGF-stimulated ZR/EGFR cells (Table 3). This is in agreement with the results of our microarray hybridisations, which showed MYC expression to be significantly reduced only by EGF (Fig. 1). Oestrogen targets such as TFF1, PGR, PDZK1 and CTSD are indeed increased by treatment of our ZR-75-1 cells and BCAR1-transfected and BCAR3-transfected cells with oestrogen (Table 3). These genes are already elevated after 6 hours of treatment with oestrogen (not by the pure oestrogen antagonist ICI 164,384), but their levels increase further (TFF1 up to 30-fold) after prolonged oestrogen treatment, in agreement with our microarray data. After stimulation of ZR/EGFR cells with EGF, the expression of PGR and PDZK1 was completely abolished, whereas TFF1 and CTSD levels were induced under EGF (Table 3). TFF3 and MGP levels were clearly induced after long-term treatment with oestrogen and reduced after stimulation with EGF. The levels of ERα (not present on the array) show some decrease after treatment of ZR-75-1-derived cell lines with oestrogen (Table 3). A much stronger decrease in ERα levels (10-fold) is achieved after 7 days of treatment of ZR/EGFR cells with EGF. ERβ levels were found to be reduced about 1000-fold compared with ERα in our cells and not strongly affected by the culture conditions. The levels of HER2/Neu (ERBB2) were decreased after treatment with both oestrogen and EGF. The expression levels of BCAR3, EGFR and BCAR1 were not strongly affected in the various cultures, except for the cells containing the introduced transgene. The observed expression modulation of IGFBP5, IL1R1 and CSTA in our microarray experiments is very well reproduced by Q-PCR (Fig. 1 and Table 3). Although oestrogen treatment causes a moderate decrease in these genes in ZR-75-1 cells, treatment of ZR/EGFR cells with EGF causes a marked effect (100-fold decrease in IGFBP5 after 7 days). The Q-PCR data also support the microarray data that BCAR3 cells have decreased levels of IGFBP5 mRNA and increased levels of CSTA mRNA in comparison with the BCAR1 and parental cells. IL1R1 and APOD levels are slightly modulated in BCAR1 cells, in agreement with the hybridisation data. PSAT1 levels were found to be further decreased in BCAR1 cells than suggested by the array experiments.