To perform a comparative analysis of the GATA family members expressed in breast tissue, we analyzed 47 breast SAGE (serial analysis of gene expression) libraries: 4 normal breast epithelium, 8 ductal carcinoma in situ (DCIS), 33 invasive ductal carcinomas (IDCs), and the MCF7 and ZR75 breast cancer cell lines. To this end, we combined 29 breast cancer SAGE libraries generated by us at a resolution of 100,000 tags per library (Aldaz Laboratory) with 18 SAGE libraries (generated at the Polyak Laboratory, Dana-Farber Cancer Institute, Boston, MA, USA) downloaded from the Cancer Genome Anatomy Project – SAGE Genie database 23. SAGE data management and tag-to-gene matching for GATA1 (GCCTCCAGAG), GATA2 (GACAGTTGTT), GATA3 (AAGGATGCCA), GATA4 (TCTCTCCCCT), GATA5 (TCCTGGCATA), GATA6 (GAGAAGATCA), ESR1 (AGCAGGTGCC), and MUC1 (CCTGGGAAGT) were performed with a suite of web-based SAGE library annotation tools developed by us 24. To enable the visualization and illustration of our analyses, we used the TIGR MultiExperiment Viewer (MeV 2.2) software (The Institute for Genomic Research, Rockville, MD, USA). This tool was employed for normalization and average clustering of the SAGE data. Spearman's test was employed for a correlation analysis of transcripts. The CGAP-dbEST and Oncomine databases 25 were employed for collection and visualization of gene expression profiles of the previously mentioned transcripts from publicly available breast cancer ESTs and microarray data sets (Additional file 1).

For validation studies, total RNA was isolated from an independent set of 36 snap-frozen breast carcinomas (stages I and II: 13 ERα-negative tumors and 23 ERα-positive tumors) using TRIzol reagent (Invitrogen, San Francisco, CA, USA). Template cDNAs were synthesized with the SuperScript™ First-strand Synthesis System (Invitrogen). GATA3 and MUC1 primers and probes were obtained from Applied Biosystems (TaqMan Assays-on-Demand™ Gene Expression Products; Foster City, CA, USA). All the PCR reactions were performed with the TaqMan PCR Core Reagents kit and the ABI Prism^® 7700 Sequence Detection System (Applied Biosystems). Experiments were performed in duplicate for each data point and 18S rRNA was used as control. Results are expressed as means ± 2 SEM based on log₂ transformation of normalized real-time RT-PCR values of the assayed genes.

We used breast-specific tissue microarrays (TMAs) from two sources (Fox Chase Cancer Center, Philadelphia, and Cooperative Breast Cancer Tissue Resource, NCI-CBCTR) for GATA3 and MUC1 protein expression profiling. We analyzed a total of 263 cases (38 normal tissues (19 DCIS and 206 IDCs). Information on ER and progesterone receptor (PR) status was available for all IDC cases. Before immunostaining, endogenous peroxidase activity was blocked with 3% H₂O₂ in water for 10 minutes. Heat-induced epitope retrieve was performed with 1.0 mM EDTA buffer pH 8.0 for 10 minutes in a microwave oven followed by cooling for 20 minutes. To block non-specific antibody binding, the slides were incubated with 10% goat serum in PBS for 30 minutes. Primary monoclonal anti-GATA3 (HG3-31: SC-268; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and anti-MUC1 (VU4H5: SC-7313; Santa Cruz Biotechnology) antibodies were used at 1:50 and 1:100 dilutions respectively. Antibody detection was performed using diaminobenzidine (DAB). Staining intensity was determined by means of a Chromavision Automated Cellular Imaging System (ACIS^®) with the generic DAB software application, as described previously 26. This software determines brown intensity regardless of the area covered by the positive cells.

Multivariate analysis was performed by principal components analysis (PCA). Variables were codified and transformed as follows: negative staining (0) and positive staining (1) for ER, PR, GATA3 and MUC1 expression; normal tissue (0), DCIS (1) and IDC (2) for histology; lymph node-negative (0) and lymph node-positive (1) status; and low (1), moderate (2) and high (3) tumor grade. To enable visualization of the factorial analysis, we employed a three-dimensional representation of component plot in rotated space. The basic significance level was fixed at p < 0.05 and all data were analyzed with SPSS^® statistical software (SPSS Inc., Chicago, IL, USA).

The estrogen-dependent breast cancer cell lines MCF7 and T47D were grown in RPMI-1640 medium (Gibco, Gaithersburg, MD, USA) supplemented with 10% (v/v) FBS (Bioser, Buenos Aires, Argentina), 10 U/ml penicillin G, and 10 μg/ml streptomycin.

Trypan blue staining was used to assess cell viability. Cells (3 × 10⁶) were lysed with 1 ml of RIPA buffer (50 mM Tris pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 1% Triton X-100, 0.1% SDS) containing protease inhibitor. For western blotting, 50 μl of cell extract was separated by 10% SDS-PAGE and transferred to nitrocellulose membrane (Protran^®; BioScience, Dassel, Germany). Immunodetection was performed with a GATA3 mouse monoclonal antibody (HG3-31) at 1:300 dilution, followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibody (anti-mouse-HRP; Sigma, St Louis, MO, USA) at 1:1000 dilution. Immunostaining was performed with DAB as substrate.

MCF7 cells were also grown on coverslips; these preparations were washed twice with 1 × PBS and fixed with 10% formaldehyde in 1 × PBS. Immunodetection of GATA3 was assessed with HG3-31 monoclonal antibody. Primary antibodies were detected with biotin-conjugated goat anti-mouse IgG, followed by incubation with streptavidin-conjugated HRP and using DAB as a substrate. The cells grown on coverslips were counterstained with hematoxylin and examined by light microscopy.

The complete sequence of the human MUC1 promoter region was obtained from GenBank (accession no. NM_002456). We used the DNAMAN software (Version 4.15; Lynnon BioSoft, Vaudreuil-Dorion, Quebec, Canada) to identify putative GATA-binding sites (A/T-GATA-A/G) in this regulatory sequence.

MCF7 breast cancer cells were grown to 90% confluence; culture medium was then removed. Cells were washed and fixed with 1% formaldehyde in 1 × PBS for 10 minutes at room temperature. Fixed cells were washed twice with 1 × PBS, scraped off and incubated in 1 ml of lysis buffer A (20 mM HEPES pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 1 mM phenylmethylsulphonyl fluoride), followed by a second incubation in 300 μl of lysis buffer B (50 mM Tris–HCl pH 8.0, 1 mM EDTA, 150 mM NaCl, 1% SDS, 2% Triton X-100). The cell lysate was then sonicated with a Branson Sonifier 450 (30-second pulses at 40% output) and subjected to immunoprecipitation with mouse monoclonal antibody HG3-31 against GATA3, at 4°C for 1 hour; 50 μl of Protein A–Sepharose (1:1 slurry) was added to the reaction and incubated at 4°C for 4 hours. The unbound proteins were removed by washing the Protein A–Sepharose with Triton buffer (50 mM Tris–HCl pH 8.0, 1 mM EDTA, 150 mM NaCl, 0.1% Triton X-100) and 1 × PBS. The antigen–antibody complex was eluted with SDS-NaCl-dithiothreitol (DTT) buffer (62.5 mM Tris–HCl pH 6.8, 200 mM NaCl, 2% SDS, 10 mM DTT). The eluted protein–DNA complex was incubated overnight at 68°C. DNA was isolated with the phenol/chloroform extraction and ethanol precipitation protocol.

The PCR reactions were performed with 15 ng of DNA, 2.5 mM MgCl₂, each dNTP at 200 μM, 25 pmol of muc1p1 forward primer (5'-tagaagggtggggctattcc-3'), 25 pmol of muc1p1 reverse primer (5'-taggtcgaggtcctgtacag-3', flanking the GATA1-binding site), 25 pmol of muc1p2 forward primer (5'-tttggctgatttggggatgc-3'), 25 pmol of muc1p2 reverse primer (5'-aatattgcactcgtcccgtc-3', flanking the GATA3-binding site), and 1.25 units of Taq DNA polymerase (Promega, Madison, WI, USA), in PCR buffer (20 mM Tris–HCl pH 8.4, 50 mM KCl) in a final volume of 50 μl. The reactions were cycled as follows: 1 cycle of 94°C for 2 minutes and 25 cycles of 94°C for 1 minute, 56°C for 1 minute, and 72°C for 1 minute. HLA-DQα 1 amplicon (242 bp) was used as control (exon sequence without a GATA-binding site) 27. Separation and detection of the amplified fragments were performed by electrophoresis on a 6% minigel (19:1 polyacrylamide:bisacrylamide) and staining with silver.

Nuclear protein extract from the MCF7 cell line was prepared as described 28. The following double-stranded DNA probes (27 bp) were used: the wild-type GATA consensus oligonucleotide (5'-cacttgataacagaaagtgataactct-3') as positive control 4; the MUC1 promoter sequence encoding the putative GATA3-binding site (5'-ggcggatctttgatagactggagtgtc-3'); and a double GATA3-binding site mutated variant (with transitions G→C and A→T: 5'-ggcggatctttcttagactggagtgtc-3'). A non-isotopic electrophoretic mobility-shift assay (EMSA) was performed with 5 μg of nuclear cell extracts and 50 ng of the probe, 1 μg of salmon testes DNA (Sigma-Aldrich, St Louis, MO, USA) in a 5 × binding reaction buffer (100 mM Hepes pH 7.9, 250 mM KCl, 2.5 mM DTT, 0.25 mM EDTA, 5 mM MgCl₂, 25% glycerol). After 20 minutes the samples were separated on a 10% polyacrylamide (39:1 polyacrylamide:bisacrylamide) at a constant temperature of 4°C with 1 × Tris-borate-EDTA (45 mA for 4.5 hours). Gels were stained with a previously described silver staining technique 27. Similarly, for the supershift assay, 1 μg of specific antibody (anti-GATA3) or control antibody (anti-ERβ) was incubated for 20 minutes in the MCF7 nuclear protein extract before the addition of probes. Samples were size separated by electrophoresis in a 10% polyacrylamide minigel (39:1 polyacrylamide:bisacrylamide) at room temperature with 0.5 × Tris-borate-EDTA.

Anti-sense oligodeoxynucleotides were synthesized that matched the translational start region of GATA3 (5'-cgccgtcacctccatggcctc-3') 29. The GATA3 anti-sense oligodeoxynucleotide used was synthesized on a phosphorothioate backbone to improve resistance to endonucleases (IDT, Coralville, IA, USA). MCF7 and T47D cells were plated on 50 mm cultured dishes at 50% confluence in Opti-MEM^® I Reduced Serum Medium (Invitrogen) and transiently transfected with 3 μg of GATA3 anti-sense mixed with Lipofectamine in accordance with the manufacturer's protocol (Invitrogen). Cells were maintained at 37°C and harvested at two time points after transfection (48 and 72 hours). Cells were lysed in RIPA buffer, and total protein concentration was estimated by Lowry's method. Equal amounts of protein were used for the western blot analyses. GATA3 and MUC1 immunodetection were assessed with HG3-31 and CT2 (epitope corresponding to the carboxy terminus of MUC1) monoclonal antibodies, respectively. Primary antibodies were detected with biotin-conjugated goat anti-mouse IgG (for HG3-31) or biotin-SP-conjugated goat anti-Armenian hamster IgG (for CT2) (Jackson ImmunoResearch), followed by incubation with streptavidin-conjugated HRP and using DAB as a substrate. SDS-PAGE with silver staining was employed for the detection of total protein loaded.

In addition, we analyzed the expression of MUC1 protein and its response to GATA3 anti-sense by an ELISA assay with MCF7 cell culture. In brief, MCF7 cells were cultured on a 96-well microtitre plate and treated with 0.05 μg of GATA3 anti-sense (48 hours), and then blocked with 1% BSA. MUC1 immunodetection was performed with HMFG1 monoclonal antibody. The bound primary antibody was detected with peroxidase-conjugated goat anti-mouse IgG. Color was developed with ABTS (2,2'-Azinobis (3-ethylbenzthiazoline-6-sulfonic acid)) substrate solution and absorbance was measured at 405 nm with a microplate reader (SLT Spectra, SLT Labinstruments, Salzburg, Austria). MUC1 expression was expressed as mean optical density with a 95% confidence interval. Statistically significant difference was estimated by t test (p < 0.05).

Interestingly, among the GATA family members GATA3 was by far the most frequently expressed GATA transcription factor in breast cancer. Specifically, GATA3 was detected as being expressed in 90% of the breast cancer cases (42 out of 47) according to SAGE database analyses. In contrast, GATA family members GATA1, GATA2, GATA4, GATA5, and GATA6 were detected as expressed in only 13% (6 out of 47) of the SAGE libraries (Figure 1). Similarly, expressed sequence tag database and DNA microarray analyses showed that the GATA3 gene is expressed mainly in breast carcinomas compared with other GATA genes (Additional file 1). The MCF7 breast cancer cell line displayed the highest GATA3 expression level observed. We detected a significant positive correlation between ESR1 (the gene encoding ERα) and GATA3 tags (r = 0.702; p = 0.0001), between ESR1 and MUC1 (r = 0.46; p = 0.001), and between GATA3 and MUC1 (r = 0.45; p = 0.002) (Figure 1).

To validate the relationship of GATA3 and MUC1 in breast cancer epithelial cell further, we performed an mRNA expression analysis of 36 primary breast carcinomas by means of real-time RT-PCR. These data showed a significant positive correlation between the mRNA levels of both markers (r = 0.6; p < 0.0001) (Figure 2a). Furthermore, mRNA expression of GATA3 and MUC1 correlated positively with breast carcinomas ERα status (r = 0.492, p = 0.004; r = 0.608, p = 0.001, respectively) (Figure 2b,c).

PCA of 263 breast tissue samples (Figure 3a; Additional file 2) identified statistically significant positive correlations between GATA3 expression and ER/PR status (τ_b = 0.53; p = 0.001) (Figure 3a,b). It is interesting to note that GATA3 and MUC1 protein expression are positively correlated among normal breast samples as well as in breast carcinomas (τ_b = 0.44; p = 0.01). Additionally, MUC1 expression was correlated with lymph node-positive status in breast carcinomas (τ_b = 0.46; p = 0.006). We detected a strong correlation between the lack of GATA3 expression and higher histological tumor grade (poorly differentiated) demonstrated by multivariate (τ_b = -0.49; p = 0.002) and univariate (p = 0.004) analyses (Figure 3a,c). Whereas 54.3% (19 out of 35) of IDC grade I and 45% (40 out of 89) of IDC grade II tumors were GATA3-positive, only 27% (18 out of 67) of IDC grade III tumors were GATA3-positive (p = 0.013).

Overall, these studies showed a consistently statistically significant correlation between GATA3 and MUC1 expression at the mRNA and protein levels by SAGE profiling (p = 0.002), real-time RT-PCR (p < 0.0001), and TMA (p = 0.01) in breast carcinomas.

Analysis of the MUC1 promoter sequence indicated the presence of six putative GATA binding sites at -446/-451 (GATA1), -1444/-1449 (GATA-like), -1572/-1577 (GATA-like), -2398/-2393 (GATA3), -2475/-2470 (GATA-like), and -2602/-2607 (GATA-like) (Figure 4). These sites encode the conserved consensus sequence required for the binding of GATA factors. To determine whether GATA3 is capable of binding the -2398/-2393 sequence in vivo, we performed a chromatin immunoprecipitation (ChIP) assay with nuclear extracts obtained from the MCF7 breast cancer cell line. In brief, protein–DNA complexes were crosslinked in vivo, sonicated and then immunoprecipitated with anti-GATA3 monoclonal antibodies. DNA samples were obtained from three fractions: input, non-specific eluted, and immunoprecipitated. These DNA samples were subjected to PCR with two sets of primers that flanked the predicted GATA3-binding site (-2398/-2393) and the putative GATA1-binding site (-446/-451) in the MUC1 promoter region.

In addition, we employed a set of primers spanning the second exon of the HLA-DQA1 gene (242 bp), used as negative control because there are no GATA-binding motifs in its sequence. The amplified product corresponding to the GATA1-binding site (186 bp) was detected from input DNA and non-specific GATA3-eluted DNA fractions (Figure 5a). However, this GATA1-related amplified product was not detected from the anti-GATA3-immunoprecipitated DNA fraction (Figure 5a), indicating that the GATA3 protein does not recognize the -446/-451 site from the MUC1 promoter region.

Identical results were obtained for the HLA-DQA1-amplified products (negative control), demonstrating the specificity of the ChIP assay (Figure 5b). In contrast, the amplified product spanning the GATA3-binding site (292 bp in size) was detected in the input and anti-GATA3-immunoprecipitated DNA fractions (Figure 5c). These amplified products were detected only in the second immunoprecipitated DNA fraction (reChIP assay; Figure 5e). Furthermore, we employed an EMSA approach to validate the ChIP findings. Nuclear extract retarded the mobility of the GATA3 sequence-specific probe (positive control) and the probe encoding the putative GATA3-binding site (5'-ggcggatctttgatagactggagtgtc-3') found in the wild-type MUC1 promoter region (-2398/-2393) (Figure 6). In contrast, electromobility was not affected when a point mutant of the aforementioned MUC1 probe (T/CTTA/G instead of the wild-type T/GATA/G) was employed, indicating that no protein binding occurred (Figure 6). The presence of GATA3 in the retarded probe-protein complex was confirmed by supershift analysis (Figure 6b). The addition of anti-GATA3 antibody to the binding reaction showed a further decrease in the mobility of the probe-protein complex. Addition of the control antibody (anti-ERβ) had no effect.

Overall, these data indicate that endogenous GATA3 protein molecules from the MCF7 breast cancer cell line were able to associate with the putative GATA3-binding site of MUC1 promoter sequence, as demonstrated with in vivo and in vitro assays.

To investigate the effect of GATA3 on MUC1 gene expression, a GATA3 knock-down assay was performed with specific anti-sense oligonucleotides in MCF7 and T47D breast cancer cell lines. Cells were treated with optimal concentrations of GATA3 anti-sense oligonucleotides (3 μg/ml) and examined 48 and 72 hours after transfection. Immunoblotting analyses of the cells treated with MCF7 and T47D showed that GATA3 translation was blocked by the anti-sense oligonucleotide at both time points analyzed (namely at 48 and 72 hours after transfection; Figure 7a). Treated cells showed a significant decrease in MUC1 expression (Figure 7). Total protein staining showed that these results were not due to differences in protein loading. The same result was obtained in a separate experiment with a MCF7-cell ELISA assay with a different monoclonal antibody (HMFG1) against MUC1 (Figure 7b).