Background
Accumulation of somatic mutations is responsible for development of breast cancer, as 85% of affected women have no family history of the disease (
http://www.breastcancer.org). Nearly 31,000 point mutations and small insertions or deletions (indels) in at least 170 previously reported and novel cancer genes have been implicated in the development of breast tumors [
1]. Whole exome sequencing places the zinc-finger transcription factor
GATA3, with a 10% frequency of alterations, among the top three (together with p53 (
TP53) and phosphoinositide-3-kinase (
PIK3CA)) mutation driver genes in breast cancer [
1,
2].
On the basis of mutation pattern, Vogelstein and colleagues [
3] classify GATA3 as a tumor suppressor. Indeed, in mice xenograft studies GATA3 was positively correlated with survival and lack of metastasis [
4]. However, it has been also postulated that GATA3 defines a distinct class of cancer genes that are differentiation factors rather than conventional tumor suppressor genes, which affect the malignant phenotype by enforcing differentiation [
5‐
7]. Specifically, conditional deletion of GATA3 is not sufficient to promote malignant progression, and is not tolerated in early tumors [
5,
8]. GATA3 has been shown in mouse model of breast cancer to maintain tumor differentiation, suppress dissemination and inhibit metastasis [
8,
9]. While GATA3 has been intensively studied in the immune system, where it functions in development and differentiation of T-cells [
10], it is also an essential regulator of mammary-gland morphogenesis and luminal-cell differentiation [
11,
12]. It is frequently up-regulated in breast cancer and has been identified as a favorable prognosis marker [
13]. GATA3 is involved in a positive cross-regulatory loop with estrogen receptor-α (ERα) [
14] where they both serve as markers for luminal breast cancer [
15,
16].
The interplay of GATA3, ERα, and FOXA1 has been a topic of multiple functional genomic studies. Kong and co-authors defined an enhanceosome consisting of co-localizing ERα-FOXA1-GATA3 which recruits RNA Pol II and p300 [
17]. The triple conjoint binding sites are highly represented at the locations involved in frequent long-range chromatin interactions and associated with genes that are most responsive to estrogen. In turn, Theodorou and colleagues silenced GATA3 and observed a global redistribution of FOXA1 and p300 cofactors, and active histone marks prior to estrogen stimulation [
18]. These global genomic changes alter the ERα-binding profile that subsequently occurs following estrogen treatment, demonstrating that GATA3 can act upstream of FOXA1 in mediating ERα binding by modulating enhancer composition.
Haploinsufficiency of GATA3 in humans results in HDR syndrome, a rare condition inherited as autosomal dominant trait, characterized by hypoparathyroidism, deafness, and renal dysplasia [
19]. Genomic alteration of
GATA3 associated with HDR syndrome include large deletions removing the entire gene and flanking sequences, splice site mutations, indels, and point mutations resulting most often in frameshifts [
20]. Mutations in HDR patients localized in the second zinc finger (ZnF2) of
GATA3 or adjacent amino acids result in loss of DNA binding, whereas those in the first zinc finger (ZnF1) lead to loss of interaction with a cofactor, FOG2, or altered DNA-binding affinity [
20,
21]. Interestingly, while HDR
GATA3 mutations are spread throughout the gene, breast cancer mutations cluster around ZnF2 and C-terminal domain [
1,
22,
23]. Analysis of six different heterozygous
GATA3 mutations from eight breast tumors has demonstrated loss or reduction of DNA binding ability, aberrant nuclear localization, decrease in transcription activation, and alterations in invasiveness, but not proliferation [
22]. However, it is unclear how those functional modifications contribute to the oncogenesis process in breast cancer.
The aim of the present study was to evaluate the effect of a breast cancer-specific mutation in GATA3 on biochemical properties and genomic location of the protein. We utilized two luminal breast cancer cell lines, MCF7 harboring a heterozygous frameshift mutation in ZnF2, and T47D carrying wild-type version GATA3. We observed that mutant GATA3 was expressed at elevated levels relative to wild-type protein and it accumulated in nuclei. Surprisingly, the mutation led to enhanced protein stability following challenge with estrogen receptor agonist or antagonist. This increased stability led to increased levels, but not to global redistribution, of GATA3 binding in the genome as determined by ChIP-seq. The data collectively support the hypothesis that the carboxyl terminus of GATA3 contains protein regulatory information that ensures appropriate turnover following ligand binding by ERα.
Methods
Cell culture
Human breast carcinoma cell lines MCF-7 and T47D were obtained from the American Type Culture Collection (Manassas, VA, USA) and cultured in DMEM/F-12 medium supplemented with 10% FBS at 37°C in 5% CO2. Protein stability was evaluated in the normal growth medium and cells were treated with 1 μM cycloheximide (CHX) and/or 1 μM MG132 (MG) for up to eight hours. For estrogen starvation assays, cells were grown for 72 hours in MEM medium containing 5% FBS and then for 24 hours in phenol red-free MEM supplemented with 5% charcoal-dextran stripped FBS. Cells were treated with 50 nM 17β-estradiol (E2) for 24 hours. The effect of ERα inhibitor, ICI 182,780 (ICI) was tested in normal growth medium. ICI was added at 100 nM concentration and cells were harvested 24 hours later. MG (EMD Biosciences, San Diego, CA, USA) was dissolved in DMSO, CHX (Cayman Chemical, Ann Arbor, MI, USA) in water, ICI (Tocris Bioscience Ellisville, MS, USA) and E2 (Sigma, St. Louis, MO, USA) in ethanol.
Subcellular fractionation
Cells were grown in 10 cm tissue culture dishes until they were 70-80% confluent. The cells were washed with PBS, collected by scraping and resuspended in buffer containing 0.15 M NaCl, 10 mM HEPES, pH 7.4, 1.5 mM MgCl2, 10 mM KCl, 0.5% NP-40, 0.5 mM DTT and protease inhibitors. The cytoplasmic fraction was separated by centrifugation at 2500 rpm for 10 min. The pellet was resuspended in nuclear extraction buffer containing 0.1, 0.2, 0.4 or 0.8 M NaCl, 25 mM HEPES, pH 7.4, 0.15 mM spermidine, 0.5 mM spermine, 5% glycerol, 1 mM EDTA and protease inhibitors. Samples were rotated for 30 min at +4°C and spun down in Optima Max centrifuge (Beckman Coulter, Brea, CA, USA) at 38,000 rpm for 45 min at +4°C. The nuclear fraction was collected and remaining pellet was dissolved in lysis buffer (8 M urea, 1% SDS, 0.125 M Tris, pH 6.8).
Immunoblotting
Whole cell lysates were obtained using 8 M urea lysis buffer (8 M urea, 1% SDS, 0.125 M Tris, pH 6.8). Protein extracts (15 μg) were resolved on SDS–PAGE gels and immunoblotted using the following antibodies: GATA3 (D13C9; Cell Signaling Technology, Danvers, MA), FOXA1 (ab23738; Abcam, Cambridge, MA, USA), ERα (sc-543; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and actin (ab8226; Abcam). Signal intensity was analyzed using rectangular volume tool in Quantity One Analysis Software (Bio-Rad, Hercules, CA, USA) with global background subtraction.
Immunofluorescence staining
Cells were grown on glass coverslips in six-well tissue culture dishes. They were fixed with 4% formaldehyde in PBS for 10 min, washed with PBS, and permeabilized with 0.1% Triton X-100 for 2 min, washed with PBS, and blocked with 5% BSA in PBS. The coverslips were incubated with the anti-GATA3 antibody (Cell Signaling Technology) for one hour, washed with PBS, incubated with the secondary antibody (Alexa Fluor 568, Life Technologies, Grand Island, NY, USA) for one hour, washed with PBS, and mounted on glass slides with mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI). The slides were examined and photographed using a Zeiss Axiovert 200 M microscope equipped with an Axiocam MR digital camera controlled by AxioVision software (Zeiss, Thornwood, NY, USA).
Expression and purification of the DNA binding domain of GATA3
DNA binding domain (DBD) of GATA3 (amino acids 261 to 371) was cloned into the pET-15b vector to produce a hexahistidine tagged fusion protein. The expression vector was transformed into the E.coli BL21 (DE3) CodonPlus RIL cells, and the cells were cultured at 37°C. The bacterial cell lysate was centrifuged at 15,000 rpm for 20 min. The supernatant was mixed gently by the batch method with Ni-NTA beads (Qiagen, Valencia, CA, USA) at +4°C for 30 min. The beads were washed with 5 mM imidazole-containing buffer and GATA3-DBD was eluted with 500 mM imidazole-containing buffer. The fractions containing GATA3-DBD were subjected to MonoS column (GE Healthcare Life Sciences, Pittsburgh, PA, USA) chromatography. The binding domain was eluted with a 4-column volume linear gradient of 100–600 mM NaCl. The protein was further purified by Superdex 75 column (GE Healthcare) in a buffer containing 20 mM Tris–HCl pH 7.5, 0.3 M NaCl, 10% glycerol, 2 mM 2-mercaptoethanol, and 1 μM zinc sulfate. For the purification of GATA3 mutant (D336fs) DBD, the Ni-NTA beads were washed with the 20 mM imidazole-containing buffer. The fractions eluted from Ni-NTA beads were dialyzed against 20 mM Tris–HCl pH 7.5, 0.3 M NaCl, 10% glycerol, 2 mM 2-mercaptoethanol, and 1 μM zinc sulfate buffer, and concentrated with Amicon ultra-centrifuge filter (Millipore, Billerica, MA, USA).
Electrophoretic mobility shift assay (EMSA)
GATA protein (0.5, 1, 2 or 4 μM for the wild-type protein and 0.25, 0.5, 1 or 2 μM for the mutant protein) was incubated with 30 μM of 20 bp dsDNA (GATA3 recognition motif-containing oligonucleotide AATGTCCATCTGATAAGACG or GATA3 recognition motif-lacking oligonucleotide AATGTCAAACTTTTAAGACG) in 10 μl of a reaction buffer (28 mM Tris–HCl pH 7.5, 1 mM dithiothreitol, 0.8 mM 2-mercaptoethanol, 120 mM NaCl, 4% glycerol, and 0.4 μM zinc sulfate). After 10 min incubation at 37°C, the samples were analyzed by polyacrylamide gel electrophoresis, and the bands were visualized by ethidium bromide staining. In the competitive DNA binding assay, wild-type and mutated GATA3 DBDs were used individually or mixed in equimolar proportion. The reactions were performed with 15 μM of 20 bp GATA3 motif-containing oligonucleotide and 23 bp GATA3 motif-lacking DNA (CACTTTTTAACGTAATTTACTCT).
Heparin chromatography
T47D and MCF7 nuclear extracts were prepared as described above, using nuclear extraction buffer containing 0.4 M NaCl. The extracts were applied to a 1 ml HiTrap Heparin Sepharose (GE Healthcare Life Sciences). The column was eluted with a 10 ml linear gradient of NaCl concentration from 0.1 to 1 M in 20 mM Hepes, pH 7.9 containing 20% glycerol, 0.2 mM EDTA, 0.1 mM PMSF, and 0.5 mM DTT. Separated fractions were analyzed by Western blot directed against anti-GATA3.
Chromatin immunoprecipitation (ChIP) analysis
GATA3 antibody was generated in rabbits using recombinant 6x histidine tag-fused GATA3 full-length wild-type protein. ChIP was performed as previously described [
24] with the following modifications. T47D or MCF7 cells were cross-linked with 1% formaldehyde in DMEM F12 for 10 min at room temperature, quenched with glycine, and then sonicated using Bioruptor (Diagenode, Liège, Belgium) to generate 200 to 400 bp DNA fragments. Immunoprecipitation was performed with GATA3 serum, and normal rabbit serum (Santa Cruz Biotechnology, Dallas, TX, USA) was used as a control. The efficiency of the reaction was verified using SYBR-green (Bio-Rad) based Real-Time PCR and primers developed by Eeckhoute et al. [
14] for GATA3 binding sites at ESR1 locus. Quantitation of precipitated DNA was done using a standard curve with 10, 1, 0.1, and 0.01% of input DNA.
ChIP-seq library construction
DNA immunoprecipitated by GATA3 antibody in four to five individual reactions performed at the same time was pooled for T47D and MCF7 cells separately and purified using MinElute PCR Purification kit (Qiagen). Total 100 μg of ChIP or input DNA, quantified with Qubit Fluorometer (Life Technologies, Grand Island, NY, USA) and dsDNA High Sensitivity Assay kit (Life Technologies), was used for library construction with the help of TruSeq RNA Sample Preparation kit (Illumina, San Diego, CA, USA). The library was prepared following the manufacturer’s instructions, starting with the end repair step, and amplified with twelve PCR cycles. Two sets of libraries (ChIP and input) were prepared for each of the cell lines from samples immunoprecipitated on separate occasions. The libraries were sequenced on a Genome Analyzer IIx (Illumina) as single end 36mers.
ChIP-seq data analysis
To ensure that low quality reads were excluded from the analysis, the raw sequence reads were filtered to remove any entries with a mean base quality score < 20. Filtered reads were aligned to the human genome (Genome Reference Consortium build 37/hg19; excluding haplotype chromosomes) via Bowtie (v0.12.8 with parameters –m 1 –v 2) [
25]; only reads that were mapped to an unambiguous ‘best’ genomic location with no more than two mismatches were accepted. To limit PCR amplification bias, duplicate reads were removed using MarkDuplicates.jar from the Picard tools package (v1.62) (
http://picard.sourceforge.net). Replicate libraries were in good agreement and were merged prior to downstream analysis. All alignments were extended at the 3’ end to a length of 180 bases (the average expected genomic fragment size for these libraries). ‘bedGraph’ files were generated from these uniquely-mapped, non-duplicated, extended reads for visualization of aggregate genomic coverage. Peak calling for regions of enriched GATA3 binding was performed with HOMER (v4.1; with default parameters and “-style factor -tbp 0 -inputtpb 0”) [
26] using input (unchipped) data to model background.
Discussion
Large-scale genome sequencing projects have provided, and continue to provide, volumes of information on the mutational landscape of cancers. A current challenge for cancer biologists is to investigate the emerging genomic data in a mechanistic context, establishing the relationship of specific mutations to tumor biology and informing on clinical parameters including aggressiveness, response to therapy, and potential for metastasis. Here, we have initiated an attempt to address the mechanistic basis by which mutations in the transcription factor
GATA3 may provide a growth advantage to breast cancer cells. The Cancer Genome Atlas Network (TCGA) recently reported a comprehensive study of human breast cancer: tumors from 507 patients were analyzed on multiple high information content platforms: whole exome sequencing, DNA copy number arrays, DNA methylation, mRNA array and sequencing, microRNA sequencing and reverse-phase protein arrays [
1]. Somatic mutations in
GATA3 occurred in 58 cases (10.7%), predominantly in luminal A and B cancer subtypes, an additional 12 samples displayed copy number alterations (
http://www.cbioportal.org). Strikingly, while mutations of
GATA3 in the congenital disorder HDR syndrome are found throughout the protein [
22], breast cancer specific mutations occur almost exclusively in exons 5 and 6 (TCGA). This clustering suggests regulatory roles for the carboxyl terminus of GATA3 and that impairment of these functions can provide a growth advantage to cancer cells.
Careful scrutiny of the TCGA mutation data revealed that six mutations were localized in the second zinc finger and five of them were frameshifts, similar to the mutation in MCF7 [
23], making MCF7 a useful model to study a clinically relevant phenomenon. We confirmed the presence of a heterozygous guanine insertion in the fifth exon of
GATA3 in the MCF7 genome and showed that although both full-length and truncated proteins were expressed, the mutated protein was present in the cells at a higher level. The D336 frameshift does not affect the N-terminal and C-terminal sequences flanking ZnF1 that are required for nuclear localization [
22] and GATA3 proteins localized to the nucleus of MCF7 cells. Mutations in GATA3 ZnF2 impair DNA binding [
20‐
22] suggesting that the same effect could be expected for MCF7-specific mutation. The biochemical fractionation assay identified a pool of truncated protein very loosely associated with chromatin (Figure
2B). However, the gel shift assay demonstrated that truncated GATA3 could bind DNA selectively, albeit with decreased affinity compared to wild-type (Figure
5). Consistent with the documented capacity of GATA3 to self-associate and to dimerize on DNA [
32], we observed a pool of mutant protein that exhibited similar chromatin binding properties to wild-type GATA3. The data are consistent with formation of heterodimers between mutant and wild-type GATA3, potentially altering the association of the protein with its recognition elements in the genome.
ChIP-seq was utilized to assess the degree of overlap of GATA3 across the two cell lines used in our study. Surprisingly, the number of binding sites detected in MCF7 was substantially higher than in T47D cells. In spite of the large difference in genomic occupancy, detailed analysis of genes associated with GATA3 binding failed to identify any major functional differences between binding profiles in T47D and MCF7 cell lines (Additional file
1: Figure S6-S10). We speculated that the increased number of GATA3-enriched regions in MCF7 genome could have been due to compromised ability of the truncated protein to recognize the specific GATA binding motif, WGATAR. However, the proportion of GATA3 peaks containing the WGATAR motif was nearly identical in binding regions identified in T47D and MCF7 cells, as well as in cell-line specific regions (Table
2). This finding suggested that the heterozygous mutation did not affect binding specificity in MCF7 cells.
Although the number of GATA3 peaks was considerably lower in T47D than in MCF7 cells, progesterone receptor gene was an example of a locus featuring a greater number of bound regions in T47D than in MCF7. Remarkably, lack of PGR expression, as determined by immunohistochemical staining, was a common denominator for all five patients in the TCGA database carrying a frameshift mutation in ZnF2 of GATA3 (
http://www.cbioportal.org). Even though both T47D and MCF7 cell lines are classified as PGR and ERα positive, and belong to luminal A breast cancer subtype [
36], MCF7 has been also used as a model for luminal B subtype [
37]. The luminal B subtype is the more aggressive form of ERα-positive breast cancer that is less responsive to endocrine therapy [
38]. It is characterized by increased expression of proliferation-related genes and lower expression of ER-dependent genes, including PGR [
38,
39]. In our model system, PGR mRNA level was approximately 20-fold lower in MCF7 than in T47D cells (Additional file
1: Figure S11). Loss of PGR expression is often considered as a marker for the gain of hormone-independent growth properties by ERα-positive breast cancers, through increased cross-talk between ERα and growth factor signaling pathways [
38,
40]. In addition, the normal balance of the two known PGR isoforms, A or B, impacts biological properties of tumors [
41].
Comparison of the biochemical properties of mutated GATA3 with wild type protein present in the T47D cell line demonstrated an increased half-life of truncated GATA3 in normal growth conditions and in response to ERα agonist and antagonist (Figures
3 and
4). GATA3 levels were proteasome-dependent (Figure
3B), similar to ERα, where rapid turnover of the receptor upon ligand binding is based on the ubiquitin-proteasome pathway [
28]. GATA3 is required for estrogen stimulation of cell cycle progression in breast cancer cells [
14] and we showed that this truncating mutation present in MCF7 genome uncouples protein level regulation from hormonal signaling.
Competing interest
The authors declare no competing interest.
Authors’ contribution
Experimental design – ABA, MT, JKS, PAW. Performed experiments – ABA, CM, MT, JKS. Data analysis – SAG, ABA, PAW. Manuscript preparation – all authors. All authors read and approved the final manuscript.