Background
Breast cancer is the second most common cancer in the US among women, accounting for approximately 40,000 deaths in 2013 [
1]. The risk factors contributing to breast cancer initiation and progression include both environmental and genetic components, which interact in a complex and poorly understood fashion. In regard to the latter, genome wide association studies have been performed to identify novel disease risk alleles. This approach has identified high frequency and low penetrance disease-associated single nucleotide polymorphisms (SNP) in a number of genes, including that encoding the HMG-box nuclear protein TOX high mobility group box family member 3 (TOX3) [
2,
3]. Despite this association, little is known concerning the expression pattern or biological functions of TOX3 in breast cancer or in mammary epithelial cells.
The HMG-box superfamily is defined by a ~80 amino acid DNA-binding domain, and individual members function in regulation of gene expression, chromatin remodeling, genomic stability, DNA repair, and other DNA-dependent cellular processes (reviewed in ref. [
4]). The thymocyte selection-associated HMG box protein (TOX) subfamily of HMG-box proteins contains four evolutionarily conserved proteins: TOX, TOX2, TOX3, and TOX4. TOX family members share a near identical DNA-binding domain and are thought to interact with DNA in a structure-dependent but sequence-independent fashion [
5]. While the N-terminal domains of these proteins share some similarity and have transactivation activity ([
5], and J. Kaye data not shown) the C-terminal domains are distinct, suggesting the possibility of non-overlapping functions. The founding member of this protein family, TOX, plays a key role in the development of multiple aspects of the immune system [
6-
8] while the
in vivo function of TOX3 remains to be identified.
TOX3 risk-allele carriers have been reported to develop more lobular breast tumors, and patients with this SNP who develop luminal A (LumA) breast tumors have shorter overall survival [
9]. Rare allele homozygotes were also found to have a higher risk for distant metasteses [
10], although molecular subtype of the resulting tumors is uncertain. Recently, Lupien and colleagues [
11] used a bioinformatics approach to identify SNPs directly implicated in increased breast cancer risk. The
TOX3 SNP causative of increased cancer risk is located 18 kb upstream of the
TOX3 transcription start site. This SNP alters a FOXA1 binding site, with disease susceptibility associated with enhanced FOXA1 binding, disrupted enhancer function, and a decrease in
TOX3 gene expression [
11]. This was consistent with earlier work where a linked disease-associated SNP was correlated with lower
TOX3 mRNA in breast cancers [
9,
12]. The inverse association between TOX3 expression and disease risk has led to the suggestion that TOX3 may act as a tumor suppressor [
11]. In addition, rare mutations of TOX3 in breast tumors have been reported [
13]. However, some
TOX3 expressing tumors are associated with adverse outcome [
9], and increased expression of
TOX3 mRNA has been implicated in breast cancer metastatic to bone [
14]. Thus, whether TOX3 plays dual and opposing roles in cancer initiation and progression remains to be determined.
Here we show that TOX3 is specifically expressed in the estrogen receptor alpha positive (ER+) subset of murine mammary luminal epithelial cells, including a recently identified progenitor cell subset. Using a novel anti-TOX3 monoclonal antibody developed by our laboratory, we confirmed high expression of TOX3 in human breast tissue samples enriched for ER+, progesterone receptor positive (PR+), and FOXA1+ luminal epithelial cells. The TOX3 protein was also highly expressed in a subset of breast cancers, predominantly among histologically defined luminal B (LumB) and LumBHer2+ breast cancer. Since TOX3 overexpression is associated with poorer outcome in patients with LumB cancer, we also sought to identify genes whose expression would be influenced by expression of this nuclear protein. In the MCF-7 breast cancer cell line, TOX3 upregulates a subset of ER target genes in addition to genes involved in cell cycle, cancer progression and metastasis. The former includes TFF1, which was upregulated by TOX3 even in the absence of estrogen. In addition, TOX3 induces enhancer RNAs that have been implicated in TFF1 gene regulation. Conversely, loss of TOX3 from LumB BT474 breast cancer cells led to decreased proliferation. Together, this work implicates a role for TOX3 in tumor progression/metastasis as well as modulation of ER-dependent responses, potentially explaining the poorer outcome of the TOX3-high subset of LumB breast cancers. The apparent paradox whereby low TOX3 is associated with cancer risk and high expression is associated with poor outcome is discussed in relation to TOX3 expression in a subset of normal mammary epithelial cells.
Methods
Mice
All mice were bred at the Cedars-Sinai Medical Center and kept under specific pathogen free conditions, or purchased from the Jackson Laboratory (Bar Harbor, ME, USA). The CSMC Institutional Animal Care and Use Committee approved use of animals (IACUC#3376).
Cell culture and transfection
MCF-7, BT474, and MDA-MB-231 cells were generously provided by Dr. H. Phillip Koeffler (Cedars-Sinai). HEK293T cells were provided by Dr. D. Nemazee (The Scripps Research Institute). Cells were maintained in DMEM (Life Technologies, Carlsbad, CA, USA) containing 10% fetal bovine serum (FBS) (Atlanta Biologicals, Flowery Branch, GA, USA). For experiments involving estrogen depletion, media was replaced by phenol-free DMEM (Life Technologies) containing 5% charcoal/dextran-treated FBS (Atlanta Biologicals).
X-tremeGENE (Roche, Indianapolis, IN, USA) was used for the transfection of plasmids and Lipofectamine 2000 (Life Technologies) for transfection of siRNAs into MCF-7 and HEK293T cells. Lipofectamine 2000 was used for transfection of MDA-MB-231 cells. Two validated TOX3 or ER Stealth RNAi duplexes and Stealth RNAi negative control duplexes (Life Technologies) were tested. Depletion of mRNA expression was measured by quantitative real-time PCR (qRT-PCR), and the most efficient RNAi duplex was identified for further use. RNAi or negative control transfected BT474 cells were also processed for protein depletion as assessed by Western blot. The human ESR1 expression plasmid was purchased from Addgene (Cambridge, MA, USA).
In some experiments, MCF-7 cells were transfected and after 8 hours switched to estrogen-depleted conditions. After an addiitonal 40 hours of cullture, cells were assessed for gene expression by qRT-PCR, or GFP+ cells were sorted for microarray analysis.
The gene encoding the long form of TOX3 was cloned from a breast cancer sample, sequenced, and compared with publicly available sequences to rule out mutation. For stable transfection, MCF-7 cells were transfected with an IRES-GFP containing empty expression vector (V) or human TOX3 encoding vector (T). Stable transfectants were selected in 2 mg/ml active G418. Upon growth, cells expressing equivalent levels of GFP were isolated from each line by cell sorting. Expression of TOX3 was confirmed by qRT-PCR and Western blot.
MDA-MB-231 cells were cultured under estrogen-depleted conditions for 24 hours before transfection. Gene expression was measured 48 hours after transfection.
Generation and validation of anti-TOX3 monoclonal antibody
Anti-TOX3 monoclonal antibody was produced in conjunction with Epitomics (Burlingame, CA, USA). Rabbits were immunized with a mixture of N- or C-terminal peptides derived from the long form of TOX3. After multiple boosting with the peptide mixture, serum was isolated for analysis. To screen for production of specific antibodies against the native protein, protein lysates from HEK293T cells transfected either with empty vector or a TOX3-encoding expression plasmid were analyzed by Western dot blot using immune sera. Hybridomas were then produced from an animal with high titer. The resulting antibody-containing culture supernatants were screened for specific reactivity against TOX3 transfected HEK293T cells. Based on this analysis, one clone specific for the N-terminal TOX3 peptide was chosen for detailed analysis. Anti-TOX3 antibody, henceforth referred to as AJ-33, was purified by protein A affinity chromatography and was validated in multiple assays reported here.
Mammary cell isolation and flow cytometry
All reagents were from StemCell Technologies (Vancouver, BC, Canada) unless otherwise specified. Mammary glands from 8–20 week-old virgin female C57Bl/6 mice, were digested for >8 h at 37°C in EpiCult-B with 5% FBS, 300U collagenase and 100U hyaluronidase. After vortexing and lysis of the red blood cells in NH4Cl, a single cell suspension was obtained by sequential dissociation of the fragments by gentle pipetting for 1–2 minutes in 0.25% trypsin, and then 2 minutes in 5 mg/ml dispase II plus DNase I followed by filtration through a 40-mm mesh.
Antibodies were obtained from eBioscience/affymetrix (San Diego, CA, USA). Mouse mammary cells were preblocked with anti-CD16/CD32 and then incubated with the following primary antibodies CD31-biotin, CD45-biotin, Ter119-biotin, BP-1-biotin, EpCAM-AF647, CD49f-AF488, or CD49f-Pacific Blue, CD49b-PE and Sca1-PE/Cy7. CD45, Ter119, CD31 and BP-1 were used to identify contaminating haematopoietic cells, endothelial cells and a proportion of stromal cells, respectively (collectively termed Lin+ cells). Biotin-conjugated antibodies were detected with streptavidin-APC/Cy7. Cells were analysed using an LSRII and specific cell populations isolated using a FACSAria III (BD Biosciences, San Jose, CA, USA). Flow cytometry data were analysed using FlowJo™ software (Tree Star, Ashland, OR, USA).
For surface CXCR4 detection, 5 × 105 cells were incubated at 4°C for 45 min with 5 μg/ml of the specific monoclonal antibody to CXCR4 conjugated to PE. Cells were washed twice with PBS and resuspended in 500 μL of PBS for analysis.
Boyden chamber migration assay
For migration studies, cells were first grown in phenol red- and serum-free DMEM for 24–48 hours. Subsequently, 2 × 105 cells were seeded in 500 μl serum-free DMEM in the upper chamber of a 24 well transwell system (ThermoFisher Scientific, Waltham, MA, USA). Phenol red-free DMEM supplemented with 10% charcoal stripped dextran treated FBS was used as a chemoattractant in the lower wells. Phenol red- and serum-free DMEM was used as a negative control to assess basal migration rates. After 24 hours, membranes were scrubbed to remove nonmigrated cells and membranes were removed and stained using Diffquik (ThermoFisher Scientific). Migrated cells were visualized by microscopy and the number of cells in the center of each well was counted. Data are represented as number of migrated cells per field of view ± SD for triplicate samples.
Cell proliferation
Cells were seeded in 24-well plates (1 × 104 cells per well) and cultured for 10 days (MCF-7) or 14 days (BT474) in appropriate culture medium. Numbers of viable cells were determined using Trypan Blue (Sigma-Aldrich, St. Louis, MO, USA).
qRT-PCR
Primary breast cancer RNA samples were purchased from BioServe (Beltsville, MD, USA). Otherwise, RNA extraction was performed using the RNAeasy kit followed by cDNA production using the Quantitect Reverse Transcription kit (Qiagen, Valencia, CA, USA). qRT-PCR was performed using SYBR Green PCR Master Mix and commercially available primers (Qiagen) were used unless indicated.
GAPDH,
MRPL, or
ACTB expression served as housekeeping gene controls. Relative gene expression was analyzed using the 2-ΔΔCt method. Primer sequences to detect
TOX3 variants are available upon request. For pre-mRNA measurements, RNA was isolated and qRT-PCR performed as above, but using primers that specifically detect pre-spliced
TFF1 mRNA as previously described [
15].
Nuclear extractions and Western blot
Nuclear protein was isolated using a Nuclear Extract Kit (Active Motif, Carlsbad, CA, USA). 20 μg of protein lysate was subjected to Western blotting using AJ-33 or anti-Actin (EMD Millipore, Billerica, MA, USA) antibodies as detected by peroxidase-conjugated anti-rabbit IgG antibody (Bio-Rad, Hercules, CA, USA) and chemiluminescence.
Immunofluorescence
HEK293T cells were grown on eight-well chamber slides. After 48 hours of transfection, cells were fixed in 4% paraformaldehyde, blocked in 10% normal goat serum for 1 hour, and stained with primary antibodies overnight at 4°C. Goat anti-rabbit antibody conjugated to PE (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) was used to detect primary antibodies. Slides were mounted with 4',6-diamidino-2-phenylindole (DAPI) (Life Technologies) to visualize nuclei.
Tissue samples and immunohistochemistry
Immunohistochemical detection of TOX3 was performed on 4-μm sections of formalin fixed paraffin embedded tissue stained with AJ-33. Staining was done using an automated slide stainer (Dako, Carpinteria, California, USA). Antigen retrieval was performed using the Dako PT Link Module and low pH buffer. Staining was visualized using the Dako Envision + Rabbit Detection System. Slides were subsequently counterstained with Mayer’s hematoxylin (Sigma-Aldrich). Three commercially available tissue arrays (Pantomics, Richmond, CA, USA) that contained a total of 210 primary breast cancer samples in duplicate were analyzed for TOX3 expression. Some samples were destroyed during the processing and therefore removed from analysis. Thus, the TOX3 histological data shown here are derived from a total of 188 breast tumors. Other histological data that were associated with individual tumors on the tissue array were supplied by the manufacturer. For some histological analysis, the Department of Pathology and Laboratory Medicine at CSMC supplied normal breast tissue obtained following mammoplasty. Samples were obtained under a waiver of consent and provided in an anonymous fashion, so that the connection to individual patients was destroyed prior to their analysis. This work was performed under Cedars-Sinai Medical Center’s Institutional Review Board Study Number: Pro 00033387.
Molecular subtyping and microarray analysis
Publicly available (GSE12093, GSE11121, GSE7390, GSE2034) breast cancer microarray expression data was compiled and normalized using 820 patient samples run only on Affymetrix (Santa Clara, CA, USA) gene chips. Samples were molecularly subtyped using the CIT classification algorithm R package provided by the authors [
16]. Multiple probes of the
TOX3 gene were averaged. A box-and-whisker plot was generated demonstrating the levels of
TOX3 mRNA expression within each molecular subtype.
Human gene expression analysis was performed using Affymetrix microarrays. In brief, RNA was isolated using the RNAeasy kit (Qiagen), and RNA quality was assessed using a Nanodrop (Thermo Scientific, Wilmington, DE, USA) spectrometer and an Agilent 2100 Bioanalyzer. RNA was reverse transcribed and hybridized to HuGene 1.0 Affymetrix arrays and processed according to manufacturer’s recommendations. Data were normalized using Justplier algorithm by Affymetrix available in Bioconductor v2.0 and R v.3.0. Since microarray signals are not quantitative at lower and higher values, signal thresholds for floor and ceiling were set for all samples. Two-way unsupervised hierarchical clustering was performed to assess unbiased gene expression patterns associated with samples. Genes whose intensity varied at least two fold between any two arrays independent of treatment were analyzed. Dendrograms were then constructed from a distance matrix containing Pearson correlations calculated iteratively between the four samples and 223 genes.
Statistical analyses
Two-tailed Student t tests were used for significance testing; P < 0.05 was considered significant with * = P <0.05, ** = P <0.01, and *** = P <0.001. Error bars represent standard deviation of replicates.
Availability of supporting data
Raw binary CEL files as well as normalized data for microarray analysis were deposited in Gene Expression Omnibus at NCBI (
http://www.ncbi.nlm.nih.gov/geo/) and can be obtained using Accession # GSE57856.
Discussion
We show here that murine Tox3 is highly expressed in ER+ luminal progenitors and ER+ mature luminal cells, but not in ER− luminal progenitors or basal epithelial cell populations. Taking advantage of the recently reported expression profiling of the human counterparts of these cells, we found a similar expression pattern in the human mammary gland.
The function of TOX3 in the mammary epithelium remains to be determined. However, that TOX3 has the potential to regulate a subset of ER target genes (see below) raises the possibility that TOX3 might play a similar role in normal mammary epithelium. Interestingly, ER
+ progenitors have been shown to be resistant to
in vivo estrogen deprivation [
18]. Given our results of ligand-independent estrogen target gene upregulation by TOX3 in tumor cells, it is possible that TOX3 also contributes to the relative insensitivity of these progenitors to estrogen deprivation.
Two TOX3 transcripts encoding proteins with distinct N-termini have been reported, and we found predominant expression of mRNA encoding the long form of TOX3 in normal breast tissue, breast cancer cell lines, and primary breast cancer. Using a rabbit anti-TOX3 monoclonal antibody specific for the long-form of the protein, we showed that TOX3 is exclusively a nuclear protein that is expressed in a subset of luminal epithelial cells, but not in myoepithelial cells or stroma, consistent with the expression pattern of the gene. Expression of TOX3 protein was not uniform, and areas exhibiting high levels of expression were seen in luminal epithelial cells in some samples. Areas enriched for TOX3-expressing cells were also highly enriched for epithelial cells expressing ER, PR, and FOXA1. Whether these TOX3+ islands represent ER+ progenitors and/or an ER+ subset of mature epithelial cells is not known. TOX3 expression was not associated with highly proliferative regions of the mammary gland as assessed by Ki67 staining (data not shown).
The coexpression of TOX3 and FOXA1 in a subset of luminal epithelial cells is interesting. The expression of TOX3 is regulated at least in part by FOXA1, an ER pioneer factor that is involved in delineating the luminal lineage [
37-
39]. FOXA1 is thought to be a positive regulator of
TOX3 expression, mediated by binding to an upstream enhancer [
11], and knockdown of FOXA1 in a breast cancer cell line decreased
TOX3 expression [
40]. We also observed modest
TOX3 upregulation in MCF-7 cells following IGF-1 treatment. This growth factor has been shown to contribute to breast cancer progression and endocrine resistance [
29,
41] through stabilization of FOXA1 protein in MCF-7 cells [
42]. Moreover, FOXA1 alters the pattern of ER binding in ‘poor outcome/metastatic’ ER
+ breast cancer from that found in ER
+ ‘good outcome’ breast cancer [
43]. Identification of genes within 20 kb of these differential ER binding sites led to a gene expression predictor set that included
TOX3, with upregulation of
TOX3 associated with poor outcome patients. In normal cells, a FOXA1-TOX3 circuit may play a role during progenitor cell differentiation of the ER
+ luminal cell subset. However, the disease risk allele SNP increases the affinity of FOXA1 for the
TOX3 upstream enhancer, inhibiting the function of this regulatory sequence and leading to a reduction in
TOX3 expression [
11]. This suggests additional complexity in
TOX3 gene regulation, where a narrow range of FOXA1 binding may be key to appropriate enhancer function.
The
TOX3 gene was expressed across multiple molecular subtypes of breast cancers, and there was heterogeneity of expression within each subtype. The notable exception was basal tumors, which were generally
TOX3-/low. The TOX3 protein was expressed in a significant proportion of histologically defined LumB tumors, and high expression of the
TOX3 gene in patients bearing LumB tumors was associated with poorer outcome. Expression profiling comparisons between LumB tumors and normal mammary epithelial cell population has suggested that these tumors are most similar to ER
+ luminal progenitors and ER
+ mature luminal cells [
18,
25], the two cell populations that normally express TOX3. Thus, ER
+ luminal progenitors may be one origin of TOX3
+ LumB tumors.
MCF-7 cells poorly express endogenous TOX3 protein. This is likely due to heterozygosity for the
TOX3 SNP affecting enhancer function [
11]. Using MCF-7 cells, we found that TOX3 has the ability to acutely regulate key genes involved in cell cycle and metastases, two key features in breast cancer progression. Consistent with a role for TOX3 in proliferation of tumor cells, knockdown of TOX3 in BT474 inhibits growth in culture, while knockdown of TOX3 in ZR-75-1 led to poor tumor formation in nude mice [
44].
We also found upregulation of CXCR4 and an increase in migration of TOX3-expressing MCF-7 cells. In conjunction with upregulation of
TFF1, this may partly explain the finding that primary breast cancers that subsequently metastasize specifically to the bone are associated with upregulation of
TOX3 (
TNRC9 in [
14]). Although TOX3 expression led to consistent upregulation of
CXCR4 at the cell population level, expression of this chemokine receptor was only evident on a subset of cells that expressed TOX3 (Figure
6B and data not shown). This may reflect the intrinsic stochastic nature of gene regulation [
45,
46], and raises the possibility that factors such as TOX3 that interact with DNA in a non-sequence specific fashion to modify chromatin [
5], may alter the threshold of gene regulation and result in tumor cell heterogeneity that can then be acted upon by selective pressures.
The disease associated
TOX3 SNP has been shown to have an additive effect on disease risk in BRCA1 mutation carriers [
47]. Recently, knockdown of TOX3 was reported to cause upregulation of BRCA1 in MCF-7 cells [
44], thus implicating TOX3 as a BRCA1 repressor. In contrast, we observed upregulation of BRCA1 and BRCA2 (along with other genes related by network analysis) upon TOX3 expression, consistent with our observation that TOX3 primarily acts as a transcriptional activator. The reason for this apparent discrepancy is unclear, although our experiments were carried out under estrogen-depleted conditions, which might alter BRCA1 regulation.
We also noted significant overlap between genes whose expression was altered by TOX3 and genes regulated by estrogen in MCF-7 cells [
31]. This may reflect the direct action of TOX3 on estrogen responsive elements [
17]. Given that we performed our microarray analysis under estrogen-depleted conditions, this suggests that TOX3 can regulate a subset of estrogen responsive genes in a ligand-independent manner. Indeed, this was demonstrated for the well-characterized ER target gene
TFF1, which was induced by TOX3 under estrogen-depleted conditions and in MDA-MB-231 following co-transfection of TOX3 and ER. While the exact mechanism of action of TOX3 remains to be elucidated, TOX3 was able to induce
TFF1 eRNAs even in the absence of estrogen, possibly indicative of promotion of a similar looping mechanism to that of ER and its ligand. Moreover, although we could not extinguish
TFF1 expression in TOX3-expressing MCF7 by inhibition of ER, ER and TOX3 together were sufficient to induce TFF1 expression in MDA-MB-231 cells in the absence of estrogen. Thus, the relationship of ER and TOX3 in the absence of estrogen may be complex, as these data suggest that TOX3-mediated induction but not maintenance of
TFF1 gene expression is ER-dependent. Alternatively, other cofactors that differ between MCF7 and MDA-MB-231 cells may complicate the interpretation of these results. Nevertheless, unliganded ER has recently been shown to play a significant role in gene regulation in breast cancer cells [
48], and we would propose that TOX3 plays a modulatory role on this activity.
Additionally, we observed hyper-responsive
TFF1 gene expression following estrogen treatment in cells overexpressing TOX3. Expression of TOX3 may be involved in the resistance to endocrine therapy reported to occur in some LumB cancers [
41]. Consistent with this, cells overexpressing TOX3 were better able to survive under estrogen-deprived conditions.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AS and JK were responsible for the conception and overall design and supervision of the study, and were the primary writers of the manuscript. AS generated the majority of the experimental data. JK contributed to analysis of TOX3 mRNA in breast tumors, designed isoform analysis, and performed Ingenuity pathway analysis. AK contributed to characterization of the novel anti-TOX3 monoclonal antibody and was involved in acquisition of cell migration data. AW oversaw development of histological staining for TOX3 and aided in acquisition and analysis of tissue staining. BT contributed to cell migration analysis, analysis of TOX3 isoforms, and studies using mice as a source of cells. DB and SS mined publically available microarray data for subtype and gene expression data, and DB performed statistical analysis. PA contributed to development of flow cytometry analysis of cell populations. VF oversaw generation of microarray data and performed the analysis. All authors gave input into the writing of the manuscript and read and approved the final version.