Abstract
Two opposing clusters of transcription factors (TF) have been associated with the differential risks of estrogen receptor positive or negative breast cancers, but the mechanisms underlying the opposing functions of the two clusters are undefined. In this study, we identified NFIB and YBX1 as novel interactors of the estrogen receptor (ESR1). NFIB and YBX1 are both risk TF associated with progression of ESR1-negative disease. Notably, they both interacted with the ESR1-FOXA1 complex and inhibited the transactivational potential of ESR1. Moreover, signaling through FGFR2, a known risk factor in breast cancer development, augmented these interactions and further repressed ESR1 target gene expression. We therefore show that members of two opposing clusters of risk TFs associated with ESR1-positive and -negative breast cancer can physically interact. We postulate that this interaction forms a toggle between two developmental pathways affected by FGFR2 signaling, possibly offering a junction to exploit therapeutically.
Significance: Binding of the transcription factors NFIB and YBX1 to the estrogen receptor can promote an estrogen-independent phenotype that can be reverted by inhibiting FGFR2 signaling. Cancer Res; 78(2); 410–21. ©2017 AACR.
Introduction
The estrogen receptor (ESR1) is the key driver and therapeutic target of breast cancer (1) and plays a critical role in determining the risk of developing this disease (2–4). Using a systems biology approach, we have examined transcriptional networks in breast cancer affecting ESR1 activity and have identified two distinct and opposing clusters of transcription factors (TF) associated with enhanced breast cancer risk (5). The “cluster 1” risk TFs are associated with estrogen receptor-positive (ER+) breast cancer risk and comprise TFs such as ESR1, FOXA1, and GATA3 whereas the “cluster 2” risk TFs appear to be associated with estrogen receptor-negative (ER−), basal-like breast cancer (BLBC). Two of the TFs located in the cluster associated with ER− disease are NFIB and YBX1. Here, we examine the molecular mechanisms underlying the opposing functions of the two groups of TFs by studying protein–protein interactions between TFs and their functional consequences. We also examine the effect of cell signaling, in particular by fibroblast growth factor receptor 2 (FGFR2), on the relative activity of the two groups of TFs.
The nuclear factor I (NFI) family of TFs consists of four members, NFIA, NFIB, NFIC, and NFIX, which can all bind DNA as homo- or heterodimers (6). They are particularly important during developmental stages (7, 8), and NFIB is crucial for normal lung and brain development (9). NFIB commonly has an increased copy number in small cell lung cancer, indicating a role as an oncogene (10). In BLBC, both copy number and expression levels of NFIB are also increased (11, 12). In addition, NFIB is important in the regulation of expression of mammary gland-specific genes, specifically those associated with lactation such as whey acidic protein and α-lactalbumin (13). NFIB has been shown to modulate androgen receptor target genes in prostate cancer cells via an interaction with FOXA1 (14, 15). An investigation into whether similar modulation of estrogen receptor (ER) occurs in the breast has yet to be carried out.
Y-box binding protein 1 (YBX1) is a member of a family of DNA- and RNA-binding proteins with an evolutionarily ancient and conserved cold shock domain. It is a multifunctional protein that certainly does not follow the classical “one protein-one function” rule, but rather has disordered structure, suggesting many different functions (16). It has been extensively studied in cancer, and its overexpression is associated with many hallmarks of the disease. It is expressed in many breast cancer cell lines regardless of subtype. However, there are higher levels of phosphorylated YBX1 in BLBC cell lines (17, 18). YBX1 expression is inversely correlated with ER, PR, and HER2 expressions and is positively correlated with the MAPK signaling cascade, a pathway important in BLBC (19, 20). YBX1 is highly expressed in 70% of BLBC cases and many of its target genes are associated with a basal-like signature (18, 20). Higher expression of YBX1 correlates with poor survival, drug resistance, and a high rate of relapse in all subtypes (18, 19, 21–23). Suppression of YBX1 reduces 2D cell growth and growth in mammospheres (18, 20). There is also evidence to suggest that YBX1 binds ESR1 in ER+ breast cancer cell nuclei (24, 25).
A locus within the second intron of the FGFR2 gene is consistently identified as the genetic locus most strongly associated with ER+ breast cancer risk by independent genome-wide association studies (GWAS; ref. 26). We have shown previously that the top three risk single nucleotide polymorphisms (SNP; refs. 27, 28) act to reduce FGFR2 gene expression and enhance the estrogen response (29). Increased FGFR2 stimulation repressed estrogen signaling in ER+ breast cancer cell lines. However, the underlying molecular mechanism remains unclear.
Here, we demonstrate that two members of the cluster 2 TFs, NFIB and YBX1, both physically interact with ESR1, repress its activity, and drive breast cancer cells toward a less estrogen-dependent cancer phenotype. FGFR2 signaling augments this interaction and subsequent repression of ESR1 target gene expression. Our evidence suggests that FGFR2 has wide-ranging effects on driving breast cancer cells toward a more basal-like phenotype and that inhibiting FGFR2 signaling in ER+ breast cancer sensitizes cells to antiestrogen therapies.
Materials and Methods
Cell culture
MCF-7 human breast cancer cells and HeLa cells were cultured in DMEM (Invitrogen) supplemented with 10% FBS and antibiotics. ZR751 human breast cancer cells were cultured in RPMI (Invitrogen) supplemented with 10% FBS and antibiotics. SUM52PE human breast cancer cells were cultured in Ham/F-12 (Invitrogen) supplemented with 10% FBS, 5 μg/mL insulin, 1 μg/mL hydrocortisone and antibiotics. All cells were maintained at 37°C, 5% CO2, obtained from the CRUK Cambridge Institute collection and authenticated by STR genotyping.
Quantitative RT-PCR
1 μg of total RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and qRT-PCR performed using cDNA obtained from 10 ng of total RNA. qRT-PCR was performed using a QuantStudio6 system (Life Technologies). Amplification and detection were carried out in 384-well Optical Reaction Plates (Applied Biosystems) with Power SYBR Green Fast 2× qRT-PCR Mastermix (Applied Biosystems). All expression data were normalized to DGUOK expression. The specificity of primers (Supplementary Table 1) was confirmed through generation of single peaks in a melt-curve analysis. Data analysis was performed using the 2−ΔΔCT method (30).
Western immunoblotting
Cells were grown in 10 cm Petri dishes, washed in PBS, and lysed on ice in RIPA buffer with complete Mini EDTA-free protease inhibitor cocktail (Roche). Resulting cell lysates were passed through a fine-gauge syringe needle several times, centrifuged at 10,000 g for 1 minute and left at −80°C at least overnight. Protein samples were separated by SDS-PAGE using 4% to 12% Bis-Tris gels (Novex) for 2.5 hours (30 minutes at 60 V, 120 minutes at 120 V) and transferred by electrophoresis using an iBlot (Novex) for 7 minutes onto a nitrocellulose membrane (iBlot Gel Transfer Stacks; Novex). Successful transfer of protein was confirmed using Ponceau S Solution (Sigma). Membranes were “blocked” at room temperature for 1 hour with 5% dried milk in Tris-buffered saline (TBS) with 0.1% Tween-20 (TTBS), washed 3× with TTBS and probed with the relevant primary antibody (Supplementary Table S2) in blocking solution at 4°C overnight. Membranes were then rewashed with TTBS 3× and incubated with appropriate HRP-conjugated secondary antibody (Supplementary Table S2) in blocking solution at room temperature for 90 minutes. Following further washing with TTBS, blots were treated with SuperSignal West Chemiluminescent Substrate (Thermo Scientific) and immunoreactive proteins detected by exposure to film (FUJIFILM).
Rapid immunoprecipitation mass spectrometry of endogenous proteins
Rapid immunoprecipitation mass spectrometry of endogenous proteins (RIME) was performed on the ESR1 protein (ERα) in MCF-7 and ZR751 ER+ breast cancer cells, as described previously (31–33). Briefly, cells were crosslinked for 8 minutes at room temperature in media containing 1% formaldehyde. Crosslinking was quenched by adding glycine to a final concentration of 0.2 M. Cells were washed with ice-cold PBS, harvested in PBS, and the resulting cell pellet was washed in PBS. The nuclear fraction was extracted from the samples by first suspending the pellet in 10 mL LB1 buffer (50 mmol/L HEPES-KOH pH7.5, 140 mmol/L NaCl, 1 mmol/L EDTA, 10% glycerol, 0.5% NP-40, and 0.25% Triton X-100) for 10 minutes at 4°C. Cells were then pelleted, resuspended in 10 mL LB2 buffer (10 mmol/L Tris-HCl pH8.0, 200 mmol/L NaCl, 1 mmol/L EDTA, and 0.5 mmol/L EGTA) and mixed at 4°C for 5 minutes. Cells were then pelleted and resuspended in 300 μL of LB3 buffer (10 mmol/L Tris-HCl pH8.0, 100 mmol/L NaCl, 1 mmol/L EDTA, 0.5 mmol/L EGTA, 0.1% Na-deoxycholate, and 0.5% N-lauroylsarcosine) and sonicated in a water bath sonicator (Diagenode). The resulting supernatant was incubated with protein A Dynabeads (Invitrogen) prebound with ESR1 antibody (Santa Cruz sc-543 X), and immunoprecipitation (IP) was performed at 4°C overnight. The beads were washed 10× in RIPA buffer and twice in 100 mmol/L AMBIC solution. Tryptic digestion of bead-bound protein and mass spectrometry was performed by the Proteomics Core Facility at The CRUK Cambridge Institute using an LTQ Velos-Orbitrap MS (Thermo Scientific) coupled to an Ultimate RSLCnano-LC system (Dionex). Full RIME data are given in Supplementary Table S3.
Coimmunoprecipitation
Cells from five 15 cm Petri dishes were harvested after washing with PBS. The cellular nuclear fraction was then obtained using a nuclear extraction kit (Affymetrix), according to manufacturer's protocol. The resulting nuclear fraction was precleared for 60 minutes with protein A Dynabeads (Invitrogen). IP was then performed with 5 μg of antibody prebound to protein A Dynabeads. Each IP was coupled with a corresponding IgG control of the same species. IP was performed overnight and the beads were washed with wash buffer (50 mmol/L Tris pH7.4, 140 nmol/L NaCl, 2 mmol/L EGTA, and 0.1% Tween-20). Beads were then boiled at 95°C for 15 minutes in LDS loading buffer and Western immunoblot analysis performed.
Molecular cloning
The plasmid constructs used for the fluorescence resonance energy transfer (FRET) were developed as follows from mCerulean3 (mCer3)-C1 and mVenus-C1 vectors kindly donated by Magdalena Grabowska (14). The ESR1-Cerulean construct was created by amplifying the gene encoding ESR1 (from RC213277; OriGene) and performing sequential digestion/ligation of the product and mCer3-C1 vector using NheI and AgeI restriction enzymes. The NFIB/YBX1-Cerulean and NFIB/YBX1-Venus constructs were created similarly (from plasmids RC231275 (NFIB) and RC209835 (YBX1); OriGene). The FOXA1-Venus construct was kindly donated by Magdalena Grabowska (14). All primer sequences are given in Supplementary Table S4. The orientation and sequence of all plasmids were confirmed by DNA sequencing (GATC Biotech).
FRET HeLa cells were transiently transfected with plasmid DNA encoding the tagged TFs described above. 15,000 cells were seeded into each well of a μ-Slide 8 Well-chambered coverslip (ibidi) and cultured for 24 to 48 hours. Samples were then fixed with 4% paraformaldehyde for 20 minutes at room temperature, washed in PBS, and stored in PBS. FRET imaging was performed using a Leica TCS SP5 confocal microscope (Leica Microsystems). Data were analyzed by FRET Acceptor Photobleaching (34) using the Leica LAS imaging software (Leica Microsystems). A total of 20 to 30 cells/well were quantified for FRET efficiency, and the experiments were repeated in at least three cellular preparations. FRET efficiency was calculated as follows:
Luciferase reporter assay
MCF-7 cells stably expressing a luciferase reporter gene under the transcriptional control of an upstream ESR1 and FOXA1 binding site, cloned from the human RARα gene (kindly donated by the lab of Jason Carroll), were plated at 50,000 cells/well in 24-well dishes and left in complete medium until 50% to 70% confluent. Cells were then transfected with the relevant siRNA/expression plasmids and a β-galactosidase construct using FuGENE HD Transfection Reagent (Promega), according to manufacturer's protocol (DNA:FuGENE ratio = 1 μg:4 μL). After 24 hours at 37°C, 5% CO2, cells were lysed with Reporter Lysis Buffer (Promega) and luciferase, and β-galactosidase assays were performed on a PHERAstar FS Microplate Reader (BMG LABTECH) using the appropriate assay kits (Promega), according to manufacturer's protocol. Each assay was performed in triplicate and a total of three assays were performed on three separate days.
Transient transfection of siRNA
MCF-7 cells were transfected with ON-TARGETplus SMARTpool siRNA (Dharmacon) directed against ESR1 (L-003401-00), FOXA1 (L-010319-00), NFIB (L-008456-00), YBX1 (L-010213-00), FGFR2 (L-003132-00), and a control nontargeting pool (D-001810-10) using Lipofectamine RNAiMax Reagent (Invitrogen), according to manufacturer's protocol. Following addition of the transfection complexes, cells were incubated at 37°C, 5% CO2 for at least 24 hours before experiments were performed.
Transient transfection of plasmid DNA
Cells were plated at 50,000 cells/well in 24-well dishes and grown in complete medium until 50% to 70% confluent, transiently transfected with plasmid using FuGENE HD Transfection Reagent (Promega), according to manufacturer's protocol (DNA:FuGENE ratio = 1 μg:4 μL), and maintained for 24 to 48 hours at 37°C, 5% CO2 in complete medium prior to conducting experiments.
Generation of stable cell lines
MCF-7 cells stably expressing FLAG-tagged NFIB (RC231275; OriGene) and YBX1 (RC209835; OriGene) were generated via transfection of the NFIB and YBX1 constructs, as described above. The day following cell transfection, cell culture medium was changed to fresh medium containing 1.5 mg/mL geneticin (G418; Invitrogen). Cells were grown and passaged, with media changed every other day until mass cell death was observed. Clonal populations of cells were selected by transferring well-isolated single clumps of cells into a 24-well plate. Cells were expanded under antibiotic selection.
Proliferation assay
Cells were plated at 4,000 cells per well into 96-well plates and cell numbers monitored in real time by in vitro microimaging using an IncuCyte incubator (Essen BioScience), allowing for monitoring of cell proliferation by observing cell confluence. Images were taken every 3 hours and data consisted of an average of four separate images taken for each well. Assays were performed in eight separate wells on three separate occasions.
RNA collection and RNA sequencing
Total RNA was extracted from cells using the miRNeasy Mini Kit (QIAGEN) and quality checked using an RNA 6000 Nano Chip on a 2100 Bioanalyzer (Agilent). mRNA-seq libraries were prepared from three biological replicates of each stable overexpression system using the TruSeq Stranded mRNA Library Prep Kit (Illumina), according to manufacturer's protocol. Single-end 50 bp reads generated on the Illumina HiSeq 4000 were aligned to the human genome version GRCh37.75. Read counts were then obtained using Subread v1.5.1 (35), normalized and tested for differential gene expression using the Bioconductor package DESeq2 (36, 37). Multiple testing correction was applied using the Benjamini–Hochberg method. The full mRNA-seq data set has been deposited in GEO under accession GSE95299.
Two-tailed gene set enrichment analysis
Two-tailed gene set enrichment analysis (GSEA; ref. 38) was performed as described previously (29). P values derived from DESeq analyses of the mRNA-seq data were −log10 transformed and then signed according to whether genes were up- or down-regulated compared with control samples. These values were then used for ranking and weighting of genes in subsequent GSEA analyses.
Survival analysis
Analysis of breast cancer patient survival stratified by YBX1 expression was carried out using the KM plotter (39).
Patient-derived xenograft analysis
A subset of breast cancer samples from Novartis’ patient-derived xenograft (PDX) dataset (40) was stratified according to YBX1, NFIB, and FGFR2 expression levels. Clinical tamoxifen response was assessed by comparison of tumor volume between treated versus untreated groups. P values were generated with Repeated Measures one-way ANOVA (RM-ANOVA) statistical test. ESR1 gene expression levels were also compared between groups using Kruskal–Wallis.
Stimulation of FGFR2 signaling
Cells in which FGFR2 signaling was stimulated were first left in complete medium overnight. Cell synchronization via estrogen-starvation was then carried out for 3 days in estrogen-free media (phenol red-free media supplemented with 5% charcoal dextran-treated FBS and 2 mmol/L L-glutamine), with media changes every 24 hours. Estrogen-deprived cells were stimulated with 1 nmol/L β-estradiol (E2; Sigma) or 100 ng/mL FGF10 (Invitrogen) in combination with 1 nmol/L E2, for 6 hours.
Results
ESR1 interacts with NFIB and YBX1
Previously, we have shown that FGFR2 signaling reduces estrogen responsiveness in breast cancer cells (29) but has little effect on ESR1 expression levels. We therefore tested whether FGFR2 signaling affects the interaction of ESR1 with its protein-binding partners. To this end, we performed a RIME analysis on the ESR1 protein (Table 1). Unique peptides for ESR1, as well as its known binding partners, FOXA1 and GATA3, were detected. The only other cluster 1 or 2 TFs for which unique peptides were detected in the RIME analysis were NFIB and YBX1. YBX1 has previously been reported to interact with ESR1 (24, 25), whereas NFIB appears to be a novel interacting partner. RIME cannot be considered a truly quantitative technique. Nevertheless, the number of unique peptides for NFIB and YBX1 detected by mass spectrometry increases when both MCF-7 and ZR751 are stimulated with FGF10, the most potent agonist of the FGFR2 receptor (41, 42). This suggests that FGFR2 signaling in ER+ breast cancer cell lines might augment the interaction of ESR1 with the two cluster 2 risk TFs.
. | E2 . | E2 + FGF10 . | IgG . | |||
---|---|---|---|---|---|---|
. | ZR751 . | MCF-7 . | ZR751 . | MCF-7 . | ZR751 . | MCF-7 . |
ESR1 | 8 | 8 | 7 | 7 | 0 | 0 |
GATA3 | 3 | 3 | 2 | 3 | 0 | 0 |
FOXA1 | 1 | 1 | 1 | 0 | 0 | 0 |
NFIB | 1 | 0 | 1 | 2 | 0 | 0 |
YBX1 | 2 | 4 | 4 | 7 | 0 | 1 |
. | E2 . | E2 + FGF10 . | IgG . | |||
---|---|---|---|---|---|---|
. | ZR751 . | MCF-7 . | ZR751 . | MCF-7 . | ZR751 . | MCF-7 . |
ESR1 | 8 | 8 | 7 | 7 | 0 | 0 |
GATA3 | 3 | 3 | 2 | 3 | 0 | 0 |
FOXA1 | 1 | 1 | 1 | 0 | 0 | 0 |
NFIB | 1 | 0 | 1 | 2 | 0 | 0 |
YBX1 | 2 | 4 | 4 | 7 | 0 | 1 |
NOTE: Values indicate the number of unique peptides identified by MS for the TFs listed in the left column, in ZR751 and MCF-7 ER+ breast cancer cells, following nuclear immunoprecipitation with an ESR1 antibody after treatment with 1 nmol/L E2 or 1 nmol/L E2 plus 100 ng/mL FGF10 (E2 + FGF10) for 90 minutes, or with an IgG control antibody after E2 treatment.
To confirm the exploratory RIME experiments, coimmunoprecipitation experiments were performed in order to test if NFIB and YBX1 could be confirmed as ESR1 binding partners by Western immunoblotting. Following IP of the nuclear fraction of both MCF-7 and ZR751 cells with an ESR1 antibody (Fig. 1A), ESR1, FOXA1, and GATA3 protein bands could be resolved by Western immunoblotting, as expected. Moreover, NFIB and YBX1 were also present in the ESR1 immunoprecipitates, while being absent in the IgG control pull downs, suggesting that both NFIB and YBX1 physically interact with the ESR1 protein in the nucleus of these ER+ breast cancer cells. As control experiments, blots were also performed for TFs that are not expected to bind to ESR1 (E2F2, SP1, and YY1), and no protein bands were detected. The inverse pull-down experiments were also performed, in which the nuclear fractions of MCF-7 and ZR751 cells were immunoprecipitated with an NFIB (Fig. 1B) or YBX1 (Fig. 1C) antibody. In both cases, the ESR1 protein was detected in the immunoprecipitate.
RIME data suggested that FGFR2 signaling in MCF-7 and ZR751 cells might increase the association of ESR1 with both NFIB and YBX1. Therefore, coimmunoprecipitation experiments were also carried out in MCF-7 cells that had been stimulated with estrogen alone or with a combination of estrogen and FGF10 (Fig. 1D and E). Densitometry analysis of the Western immunoblots against NFIB and YBX1 following pull down of ESR1 shows that stimulation of MCF-7 cells with FGF10 appears to augment the interaction of the two ER− risk TFs with ESR1, without affecting protein levels (Fig. 1F and G). Moreover, FGFR2 signaling in ER+ breast cancer cells increases the level of phosphorylated YBX1 (demonstrated in MCF-7 cells), while FGFR2 inhibition reduces it (demonstrated in SUM52PE cells, which carry an FGFR2 gene amplification; Fig. 1H and I). Our finding that YBX1 can bind to ESR1 is consistent with recent reports of an interaction between these two proteins (24, 25).
FRET, which is facilitated by tagging proteins of interest with fluorescent proteins as reporters, is an imaging technique useful for studying protein– interactions (43). FRET only occurs when the fluorescent proteins are within very close proximity of each other (<10 nm), thereby allowing for the measurement of the proximity of proteins of interest (Fig. 2A). Here, we tagged FOXA1, NFIB, YBX1, and ESR1 with either a donor (mCerulean3) or acceptor (mVenus) fluorescent protein and performed FRET in HeLa cells expressing the constructs (Fig. 2B; Supplementary Fig. S1). Consistent with previous reports of ESR1 and FOXA1 interactions (44), cotransfected ESR1-Cer and FOXA1-Venus emitted a strong FRET signal (Fig. 2B) with an efficiency of 0.139 ± 0.011 (Supplementary Fig. S2). To determine whether NFIB and FOXA1 are also able to interact directly, cells were cotransfected with NFIB-Cer donor and FOXA1-Venus acceptor constructs. The pairing resulted in a positive FRET signal with a FRET efficiency of 0.055 ± 0.007. On the other hand, the ESR1-Cer and NFIB-Venus pairing did not result in FRET (efficiency of 0), suggesting that these proteins do not interact directly. To test the hypothesis that FOXA1 can bridge the interaction between ESR1 and NFIB, we cotransfected cells with ESR1-Cer, NFIB-Venus, and untagged FOXA1. The FRET efficiency of ESR1-Cer and NFIB-Venus was increased to 0.018 ± 0.003. This result suggests that FOXA1 serves as an intermediary “bridge” to bring ESR1 and NFIB together. The same experiments were carried out with YBX1 FRET constructs, demonstrating that YBX1 is able to bind to ESR1 directly, without requiring FOXA1 (Fig. 2B; Supplementary Fig. S2).
NFIB and YBX1 suppress ESR1 activity
Having established that both NFIB and YBX1 interact with the ESR1/FOXA1 TF complex, we asked whether NFIB and YBX1 are able to influence the transcriptional activity of ESR1. When NFIB or YBX1 were transiently overexpressed in MCF-7 cells, expression of the ESR1-target gene, pS2, was significantly reduced compared with the control cells (Fig. 3A). Conversely, reduction of NFIB or YBX1 levels via siRNA transfection resulted in increased pS2 expression. The same results were also obtained for other ESR1-target genes (Supplementary Fig. S3). Similarly, when NFIB or YBX1 were transiently overexpressed in MCF-7 cells stably expressing a luciferase reporter gene under the transcriptional control of an upstream ESR1/FOXA1 binding site, luciferase expression was significantly reduced compared with control cells (Fig. 3B). These data suggest that both NFIB and YBX1 are able to inhibit ESR1-mediated transcriptional activity.
To investigate further the possible role of NFIB and YBX1 on ESR1 activity, MCF-7 cell lines stably overexpressing FLAG-tagged NFIB or YBX1 were generated (Supplementary Fig. S4). For each TF, three independent clones were expanded, mRNA-seq data generated, and the regulatory network examined. We previously defined regulons (set of target genes) for all TFs by measuring the similarities in gene expression patterns of the TF of interest and all possible target genes in gene expression data from breast tumor samples (5). Here, we carried out a two-tailed GSEA (5, 29) to assay the activity of the ESR1 regulon in the stably transfected cells. As a control, we show the behavior of the ESR1 regulon in response to estrogen stimulation. As expected, positive targets of ESR1 are induced and negative targets of ESR1 are repressed in the parental MCF-7 cells (Fig. 4A). Overexpression of both NFIB and YBX1 leads to a relative repression of the ESR1 regulon (Fig. 4B and C), with negative ESR1 targets being upregulated and positive targets showing lower expression. These experiments confirm that both NFIB and YBX1 are able to inhibit ESR1 function.
When the MCF-7 cells stably overexpressing NFIB or YBX1 were estrogen starved, they were able to proliferate faster than estrogen-starved parental MCF-7 cells (Fig. 4D and E). A study by Shibata and colleagues reported that YBX1 is able to reduce the stability of ESR1 protein (25). However, Western immunoblots of cell extracts demonstrate that full-length ESR1 protein levels are not altered by either NFIB or YBX1 overexpression in our system (Supplementary Fig. S4). Our results suggest that overexpression of these cluster 2 risk TFs is able to drive ER+ breast cancer cells toward a more ER−, basal-like cancer phenotype in which estrogen dependency is reduced.
FGFR2 signaling and breast cancer regulon activity
To further assess the shift from luminal to a more basal-like phenotype, we extended our two-tailed GSEA to all regulons and visualized the results in a tree and leaf diagram, where regulons are represented as leaves, and the branching between them is a measure of their relatedness (5). Using this approach, a gene signature derived from ER+ versus ER− tumors showed a positive enrichment in the regulons of cluster 1 risk TFs and a negative enrichment of cluster 2 risk TFs (Fig. 5A). A basal gene signature showed the inverse (Fig. 5B). Interestingly, we found that a FGFR2 signaling gene signature was able to activate the NFIB and YBX1 regulons, as well as almost all TF regulons that are associated with ER− disease (Fig. 5C), mimicking very closely the results obtained with the basal gene signature. A reduction of FGFR2 gene expression via siRNA transfection has the opposite effect, increasing the activity of ESR1 and other cluster 1 TFs (Fig. 5D), supporting and extending our earlier findings that FGFR2 signaling opposes estrogen signaling.
The fact that FGFR2 signaling inhibits estrogen signaling in ER+ breast cancer cells, possibly via an increased association of ESR1 with the ER− risk TFs, NFIB, and YBX1, led us to test the hypothesis that the inhibition of FGFR2 signaling in ER+ breast cancer cells sensitizes cells to antiestrogen therapies. When three different ER+ breast cancer cell lines (MCF-7, ZR751, and T47D), which all express NFIB and YBX1 (Fig. 6A), are treated with the FGFR2 inhibitors, AZD4547 and PD173074, their growth, as measured in an IncuCyte incubator, is more sensitive to the antiestrogen tamoxifen (Fig. 6B–D; Supplementary Fig. S5). This suggests that anti-FGFR2 treatments make breast cancer cells more reliant on estrogen signaling for growth and could therefore be used in combination with antiestrogen therapies to treat breast cancer. When MCF-7 cells stably overexpressing either NFIB or YBX1 are treated with siRNA against NFIB/YBX1, they become significantly less sensitive to the combined drug treatment when compared with nontransfected control cells (Fig. 6E and F; Supplementary Fig. S5), suggesting that NFIB and YBX1 do indeed play an important role in the FGFR2-driven estrogen activity/sensitivity of breast cancer cells. Much more work is needed to determine if the effect of FGFR2 signaling on a breast cancer cell's reliance on estrogen signaling is primarily mediated by NFIB and YBX1. However, it is interesting to note that overexpression of YBX1 in breast cancer is associated with poorer survival, even when tested just in ER+ breast cancer (Supplementary Fig. S6). Furthermore, in PDX models of breast tumors (40), we find that tamoxifen treatment is not very effective in PDXs with high YBX1 expression (Supplementary Fig. S7), although this group is likely to contain ER− tumors. In contrast, tamoxifen efficacy is greater in YBX1-low PDXs and even greater in PDXs that express both low FGFR2 and YBX1, further supporting the notion that inhibition of FGFR2 may increase a tumor's response to tamoxifen.
Discussion
In this study, we demonstrate that in ER+ breast cancer the TFs NFIB and YBX1 interact with ESR1, the key driver of luminal breast cancer. We examine the functional consequences of this interaction and find that NFIB and YBX1 are each able to repress transcriptional activation by ESR1. This can be observed in reporter assays, at the level of endogenous estrogen-regulated genes such as pS2, and also in the reduction of the overall activity of the ESR1 regulon. The interaction between YBX1 and ESR1 is direct, while NFIB requires FOXA1 as a bridging protein that allows the interaction. The complex formation we observe between NFIB or YBX1 and ESR1 may explain the opposing action that NFIB/YBX1 and ESR1 have on shared target genes (5). In addition to repressing ESR1, NFIB and YBX1 are also able to drive proliferation: while proliferation of parental MCF-7 cells is strictly dependent on the presence of estrogen and hence nuclear ESR1, MCF-7 cells overexpressing either NFIB or YBX1 are able to grow in estrogen-depleted medium.
To date, NFIB and YBX1 have primarily been associated with ER− breast cancer, where both factors contribute to increased aggressiveness and metastatic potential (12, 45). We now report that these two TFs repress ESR1 activity, suggesting that they may play a similar role in ER+ breast cancer. Although ER+ breast cancer has better patient outcomes, in large part driven by the effectiveness of hormone deprivation therapy, relapse and resistance to therapy are relatively common and can occur many years after the primary tumor was diagnosed and treated (46). Our previous work suggests that patient outcomes are strongly affected by the relative activity of TFs driving ER+ (cluster 1, e.g., ESR1, GATA3, and FOXA1) versus ER− disease (cluster 2, e.g., YBX1 and NFIB). We found that, in an ER+ patient cohort, patients with a repressed ESR1 regulon have worse prognosis (5). We now show that NFIB and YBX1 can both function to repress the activity of the ESR1 regulon. In line with this observation, we found that in clinical samples from patients with ER+ disease, higher YBX1 expression is associated with reduced survival. As a corollary, interventions that increase the activity of the ESR1 regulon may improve patient outcomes, since the tumor is likely to have increased sensitivity to estrogen deprivation therapy.
We have previously demonstrated that the risk gene FGFR2 can influence the way in which a cell responds to estrogen, with FGFR2 signaling leading to reduced activity of the ESR1 regulon (29). We have now extended our analysis and found that FGFR2 signaling not only affects the ESR1 regulon, but alters the activity of many TFs: the activity of TFs highly expressed in luminal A or B tumors is decreased, while the activity of TFs highly expressed in BLBC, such as NFIB and YBX1, is increased. A link between FGFR2 signaling and the activity of specific TFs has previously been reported. For example, in MCF-7 cells it causes degradation of the progesterone receptor, leading to increased proliferation and cell migration (47). FGFR2 mediated activation of TFs associated with ER− disease has not been studied directly, but indirect evidence exists. Signaling through FGFR2 leads to phosphorylation of RSK2, a mediator of anchorage independent growth and motility (48), which in turn activates YBX1 by phosphorylation (49). Our data here indicates that FGFR2 signaling also increases the affinity of YBX1 for ESR1. Taken together these observations suggest that FGFR2 signaling increases the ability of YBX1 to activate target genes associated with BLBC, while at the same time increasing its ability to repress ESR1 target genes.
A role for FGFR2 in promoting a basal-like phenotype is consistent with previous findings. Functional studies of FGFR2 risk variants have demonstrated that a decrease in FGFR2 expression is associated with an increased risk in ER+, but not ER− breast cancer (28). Conversely, FGFR2 amplifications, although infrequent (4%; ref. 50), occur primarily in ER− breast cancer. ER− breast cancer cell lines tend to express higher levels of FGFR2 than ER+ breast cancer cell lines (51) and are more sensitive to FGFR2 inhibitors such as PD173074. In clinical samples, FGFR2 expression was higher in ER− tumors and associated with poor patient outcome (51). However, inhibition of FGFR2 signaling may also be effective in ER+ tumors. We hypothesized that inhibition of FGFR2 signaling would make cells more dependent on estrogen (through upregulation of the ESR1 regulon) and therefore more sensitive to estrogen deprivation therapy. We tested this in cell lines and found that MCF-7, ZR751, and T47D cells treated with the FGFR2 inhibitors PD173074 or AZD4547 became more sensitive to treatment with tamoxifen.
FGFR inhibitors have been used effectively in the treatment of a variety of cancers, particularly those carrying FGFR amplifications (52, 53). In breast cancer, the FGFR1 gene is amplified in about 13% of all breast cancer cases, while other FGFR genes are only rarely amplified (FGFR2, 1.5%; FGFR3, 0.5%; FGFR4, 1.5%) and are not frequently mutated. In line with our findings for FGFR2, activation of both FGFR1 (by amplification) and FGFR3 (in vitro) is associated with a reduced response to endocrine therapy (54, 55). This observation led to clinical trials of FGFR inhibitors in combination with estrogen deprivation therapy. Not surprisingly, such trials have focused on patients with amplifications in the FGFR pathway and gave encouraging results, but were ultimately inconclusive due to the small number of patients carrying the relevant genomic alteration (56). Our work here suggests that rather than just focusing on FGFR amplification, alternative biomarkers such as the presence of activated YBX1 could be used to select patients that may benefit from FGFR2 inhibition. Consistent with this suggestion, we find that high expression of YBX1 in ER+ disease is associated with worse outcome. In the future, this link needs to be further explored and activated YBX1 protein measured in ER+ tumor samples. Alternatively, treatment could be focused on downstream events, preventing the interaction of YBX1 or NFIB with ESR1. If this interaction is dependent on posttranslational modifications, the inhibition of the relevant enzymes may be effective. As a first step toward moving our findings to the clinic, we envisage the use of PDX models of breast cancer to confirm synergy between FGFR2 inhibition and estrogen deprivation treatment in preventing tumor growth.
In conclusion, we demonstrate that signaling by FGFR2 pushes cells toward a more basal phenotype, which is at least in part mediated by facilitating the interaction between NFIB and YBX1, and ESR1. The regulatory loop between NFIB/YBX1 and ESR1 may be a promising target for developing new therapeutic strategies.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: T.M. Campbell, M.A.A. Castro, B.A.J. Ponder, K.B. Meyer
Development of methodology: T.M. Campbell, M.A.A. Castro, K.B. Meyer
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.M. Campbell
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.M. Campbell, M.A.A. Castro, K.G. de Oliveira, B.A.J. Ponder, K.B. Meyer
Writing, review, and/or revision of the manuscript: T.M. Campbell, M.A.A. Castro, K.G. de Oliveira, B.A.J. Ponder, K.B. Meyer
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.A.A. Castro, K.B. Meyer
Study supervision: M.A.A. Castro, B.A.J. Ponder, K.B. Meyer
Acknowledgments
This work was funded by The Breast Cancer Research Foundation (BCRF) and by Cancer Research UK (CRUK). We thank the Genomics, Proteomics, Bioinformatics, Light Microscopy and Research Instrumentation Core Facilities at The CRUK Cambridge Institute for their help and expertise. We are grateful to Magdalena Grabowska for the gift of plasmids for the FRET experiments, and Kelly Holmes for the gift of the luciferase reporter cell line. B.A.J. Ponder is a Gibb Fellow of CRUK.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.