PSF Promotes ER-Positive Breast Cancer Progression via Posttranscriptional Regulation of ESR1 and SCFD2

Breast cancer is one of the most common cancers in women, and approximately 70% of the tumors are estrogen receptor (ER)-positive. Estrogen is the primary sex hormone that regulates mammary gland development, whereas overexposure to estrogen is a risk factor for ER-positive breast cancer (1, 2). As a prototypical nuclear receptor, ER interacts with its cognate ligand estrogen, and ligand-bound ER regulates its target gene expression by binding to promoter/enhancer regions (3). Thus, endocrine therapy with selective ER modulators or aromatase inhibitors is a standard treatment for primary ER-positive breast cancer (4). Nevertheless, long-term endocrine therapy causes acquired therapy resistance, which remains to be conquered by the development of alternative diagnostic and therapeutic options for advanced breast cancers (5–8).

Alterations in RNA processing events would contribute to cancer pathophysiology. In RNA processing, RNA-binding proteins (RBP) primarily interact with RNA transcripts to form ribonucleoprotein complexes and control RNA quality via splicing, polyadenylation, modification, nuclear export, decay of RNA transcripts, and protein translation (9, 10). Recent studies revealed that RBPs play particular roles in cancer development and are expected as new therapeutic targets for cancers. Integrative profiles of RBP expression were recently shown in hepatocellular carcinoma, where 414 RBPs were upregulated compared with nontumor samples (11). In particular, an oncogenic role of RBP NELFE was exhibited in MYC-induced tumor growth by regulating the binding of MYC to its target promoters (11). In breast cancer, hnRNPM promotes CD44 exon skipping and increases the expression of a CD44 variant, which promotes metastasis (12). Stress granule–associated protein G3BP2 is involved in breast cancer initiation via interacting with and stabilizing SART3 mRNA, which upregulates pluripotency transcription factors such as Oct4 and Nanog (13).

Among RBPs, polypyrimidine tract-binding protein (PTB)-associated splicing factor (PSF), or splicing factor, proline- and glutamine rich (SFPQ), was originally identified as a component of large spliceosomal complex (14). PSF/SFPQ belongs to the family of Drosophila behavior/human splicing (DBHS) protein, whose structure is defined by RNA recognition motifs: NonA/paraspeckle domain and coiled-coil domain (14). PSF is involved in RNA splicing of its targets such as CD45 and microtubule-associated protein Tau (15). PSF is also reported as a component of nuclear speckles via interactions with other components such as long noncoding RNA NEAT1 or DBHS protein NONO (16). In addition to DBHS-defined motifs, PSF also has a DNA-binding domain and generally suppresses its target gene transcription by recruiting histone deacetylase proteins to its genomic-binding regions (17). PSF is involved in various biological phenomena such as innate immune responses (18) and neuronal development (19). Recent studies reported that PSF is also involved in tumorigenesis. In colon cancer cells, PSF physically associated with PPARγ and PSF knockdown suppressed cell proliferation (20).

We previously demonstrated that PSF modulates the expression of spliceosome genes and promotes prostate cancer metastasis and proliferation (21). In this study we aimed to clarify the clinical relevance and molecular function of PSF in breast cancer. Here, we performed pathophysiologic study of PSF in clinical breast cancers, as well as global transcriptome and interactome analyses in breast cancer cells by modulating PSF expression. Our findings reveal that PSF is a potential poor prognostic factor for ER-positive breast cancer and promotes the proliferation of ER-positive breast cancer cells by regulating mRNA expression of ERα (ESR1), Sec1 family domain containing 2 (SCFD2), transformer-2 protein homolog beta (TRA2B), and abnormal spindle-like microcephaly-associated protein (ASPM) at posttranscriptional level. We infer that PSF and its related pathways would be a new class of diagnostic and therapeutic targets for primary and hormone-refractory breast cancers.

Clinical samples in Fig. 1 were obtained from 73 Japanese female patients with breast cancer diagnosed between 1989 and 1998 with or without distant metastases during or after tamoxifen therapy in National Hospital Organization Shikoku Cancer Center (SCC, Matsuyama, Ehime, Japan). This study was approved by the ethical committees at SCC and Saitama Medical University (SMU, Hidaka, Saitama, Japan) and all the patients were with approval during the period. Clinical samples in Supplementary Fig. S1 were obtained from 114 Japanese female patients with breast cancer who underwent surgical treatment from 2006 to 2013 at Toranomon Hospital (TH, Minato-ku, Tokyo, Japan). No patients received chemotherapy or molecular target therapy before surgery. Standard adjuvant treatments were selected according to the clinical practice guidelines of the National Comprehensive Cancer Network (22). Staging was determined based on “TNM classification of malignant tumors” (23). This study was approved by the ethical committees at TH and SMU. All patients provided written informed consent to participate in this study. All tumor samples of the two cohorts were embedded by paraffin and stored at room temperature, and sliced and stored at room temperature for IHC. This study abides by the Declaration of Helsinki principles.

Details for IHC analysis of clinical specimens are described in the Supplementary Materials and Methods. PSF antibody (6D7; Sigma-Aldrich) and mouse IgG were used as primary antibody and negative control, respectively. Immunoreactivity (IR) for immunostained slides was determined as described previously (24), based on the evaluation by two well-trained pathologists, ranging from 0 to 200. Using a median IR value of the two independent examinations in each cohort as a cut-off threshold, PSF IR > 100 and >67.5 for SCC and TH cohorts, respectively, were defined as strong IR.

Details for cell culture and reagents are described in the Supplementary Materials and Methods. Short tandem repeat–based authentication of cell lines was verified by BEX Co., LTD. Mycoplasma testing was routinely carried out to ensure cell lines were Mycoplasma free. MCF7–derived 4-hydroxytamoxifen (OHT)-resistant OHTR cells were established as described previously (25). MCF7 cells stably overexpressing PSF were established by transfection with empty or PSF expression plasmid and selected with G418. SCFD2 stably overexpressing cells were established using lentiviral transduction system as described previously (26). For estrogen treatment, cells were cultured with phenol red-free DMEM (Thermo Fisher Scientific) supplemented with charcoal-stripped FBS (cFBS) for 48 hours and treated with 17β-estradiol (E₂).

Details for siRNA sequence are described in the Supplementary Materials and Methods. Cells were seeded at 3 × 10⁵ cells/well in 6-well plates and simultaneously transfected with siRNA (10 nmol/L) using Lipofectamine RNAiMAX (Thermo Fisher Scientific), and harvested 48–72 hours after transfection.

qRT-PCR analysis and Western blot analysis were performed as described in the Supplementary Materials and Methods. Primer sequences are listed in Supplementary Table S1.

DNA assay and cell-cycle analysis were performed as described in the Supplementary Materials and Methods. For DNA assay, each cell sample was stained with Hoechst 33258 Pentahydrate (Thermo Fisher Scientific) and its fluorescence was measured using ARVO5 Microplate Reader (Perkin Elmer). For cell-cycle analysis, DNA contents were measured using FACSCalibur (BD Biosciences).

All animal experiments were approved by the SMU Animal Care and Use Committee, and followed by the institutional guidelines and regulations. OHTR cells were mixed with an equal volume of Matrigel matrix (Corning) and injected subcutaneously into the side flank of 10-week-old female nude mice (BALB/cAJcI-nu/nu, Japan CREA, Inc.). When the tumor volume reached 100 mm³, mice were divided randomly into two groups. siControl or siPSF #A (5 μg each) prepared with GeneSilencer Reagent (Gene Therapy System) were injected into the generated tumors twice a week (every Monday and Thursday), as described previously (21, 27). Three dimensions of tumor were measured once a week and tumor volumes were estimated with following formula: 0.5 × first diameter × second diameter × third diameter.

Cells were transfected with indicated siRNAs (10 nmol/L) and cultured in cFBS-conditioned DMEM for 24 hours and treated with E₂ (10 nmol/L) or control ethanol for 12 hours. GeneChip Human Gene 1.0 ST Array (Affymetrix) was used according to the manufacturer's protocol. Data analysis was performed by Affymetrix Microarray Suite software. Microarray data are available in the Gene Expression Omnibus (GEO) database with the accession number GSE132743. Pathway analysis was performed using gene set enrichment analysis (GSEA; ref. 28) and gene ontology analysis (29).

Details for luciferase assay are described in the Supplementary Materials and Methods. Luciferase activities for cells transfected with reporter genes and treated with E₂ or vehicle were measured using Dual-Luciferase Reporter Assay System (Promega) on TriStar² S LB942 Microplate Reader (Berthold).

Cells were scraped and lysed with RNA immunoprecipitation (RIP) buffer (150 mmol/L KCl, 25 mmol/L Tris-HCl, pH 7.4, 5 mmol/L EDTA, and 0.5% NP40). Cell lysate was incubated with mouse IgG or anti-PSF (1 μg each) for 3 hours at 4°C with rotation. PSF–RNA complexes were precipitated using Protein G Sepharose 4 Fast Flow (GE Healthcare), and the bead-conjugated RNAs were isolated using ISOGEN reagent.

Cells collected from a 10-cm culture dish were incubated with nuclear isolation buffer (0.32 mol/L sucrose, 10 mmol/L Tris-HCl, pH 7.5, 5 mmol/L MgCl₂, and 1% Triton X-100) for 20 minutes at 4°C. Samples were vortexed and centrifuged at 500 × g for 10 minutes at 4°C. The supernatant was a cytoplasmic fraction, and the precipitant was a nuclear fraction.

Cells were cultured in cFBS-conditioned DMEM for 24 hours and treated with 10 nmol/L E₂ or control ethanol for 24 hours. Details for RIP-seq are described in the Supplementary Materials and Methods, as described previously (21). Fisher exact test was performed to statistically determine the difference between PSF-immunoprecipitated and input samples. The detailed data are submitted to the GEO as GSE133423.

Relapse-free survival curves for patients with breast cancer were acquired through Kaplan–Meier Plotter (http://kmplot.com/analysis/; ref. 30). PSF and SCFD2 expression in clinical samples were analyzed using Oncomine (https://www.oncomine.org). Coexpression analysis of PSF mRNA was performed using cBioPortal (http://cbioportal.org/; refs. 31, 32), based on RNA-seq data from METABRIC dataset (33).

Statistical analyses of in vitro and in vivo experiments were performed using Student t test and ANOVA, respectively.

To understand the clinical importance of PSF in breast cancer, we examined PSF protein expression using IHC of 73 ER-positive breast cancer specimens. Among them, 35 cancer tissues exhibited strong PSF IR (Fig. 1A) whereas 38 cancer tissues exhibited weak PSF IR (Fig. 1B). Benign breast tissues basically showed weak PSF IR (Fig. 1C). Analysis of PSF IR and patient clinical parameters revealed that PSF IR is significantly associated with lymph node status (P = 0.007) and ER-labeling index (P = 0.0012), but not with age or tumor size (Supplementary Table S2). We next examined the relationship between PSF IR and the clinical prognosis of patients with breast cancer. PSF strong IR was significantly associated with shorter distant disease-free survival (Fig. 1D) and overall survival (Fig. 1E), suggesting that high PSF expression is inversely associated with breast cancer patient survival. Univariate and multivariate analysis of distant disease-free and overall survival using Cox proportional hazard model demonstrated that PSF positivity could be an independent prognostic factor for distant disease-free and overall survival (Supplementary Table S3).

We also performed IHC study in another cohort of 114 ER-positive breast cancer specimens from another institute apart from the institute of the first cohort. Of the 114 samples, 57 cancer tissues exhibited strong PSF IR (Supplementary Fig. S1A). The rest of 57 cancer tissues exhibited weak PSF IR (Supplementary Fig. S1B), which was also observed in normal breast tissues (Supplementary Fig. S1C). ER-labeling index is significantly associated with PSF IR (P = 0.012) also in this cohort study (Supplementary Table S4). On the other hand, other clinical parameters including age, body weight, body mass index, stage, pathologic T factor, lymph node status, histologic grade, lymphovasucular infiltration, progesterone receptor status, and HER2 status are not associated with PSF IR. Prognostic analysis showed that PSF strong IR is significantly associated with shorter overall survival (Supplementary Fig. S1D; P = 0.010). Univariate and multivariate analyses of overall survival using Cox proportional hazard model demonstrated that PSF positivity could be an independent prognostic factor in patients with ER-positive breast cancer (Supplementary Table S5). Overall, these two independent cohort analyses showed that PSF strong IR is well correlated with ER-high labeling index and shorter overall survival.

Furthermore, Oncomine software–based expression analysis using The Cancer Genome Atlas (TCGA) dataset showed that PSF mRNA expressions in invasive ductal breast cancers (IDC; n = 389) and invasive lobular breast carcinomas (ILC; n = 36) were significantly higher than those in normal breast samples (n = 61; Fig. 1F). On the basis of a publicly available breast cancer dataset in Kaplan–Meier plotter platform, high PSF mRNA expression is also associated with a shorter relapse-free survival time in whole subtypes (Fig. 1G and H) and ER-positive subtype (Supplementary Fig. S1E and S1F).

On the basis of the potential clinical relevance of PSF in breast cancer, we next investigated PSF function in terms of tamoxifen resistance. We previously established tamoxifen-resistant model cells from MCF7 cells using long-term culture with OHT, named OHTR (25). ESR1 mRNA expression was elevated in OHTR cells compared with parental MCF7 cells (Supplementary Fig. S2A) and knockdown of ESR1 attenuates the proliferation of OHTR cells (Supplementary Fig. S2B and S2C), suggesting that this tamoxifen-resistant cell model maintains estrogen dependency. In the OHTR cells, expression levels of PSF mRNA (Fig. 2A) and protein (Fig. 2B) were both elevated compared with parental MCF7 cells. Next, we examined the effect of PSF siRNAs on breast cancer cell proliferation. Two distinct siRNAs against PSF (siPSF #A and #B) suppressed the expression of PSF in ER-positive MCF7, OHTR, and T47D breast cancer cells at the RNA (Fig. 2C and D; Supplementary Fig. S3A) and protein levels (Fig. 2E; Supplementary Fig. S3B). Cell proliferation analyses showed that the growth of ER-positive breast cancer cells was significantly repressed when they were transfected with PSF siRNAs compared with control siRNA (siControl; Fig. 2F and G; Supplementary Fig. S3C). Furthermore, cell-cycle analysis revealed that PSF knockdown decreased the proportion of S-phase cells (Fig. 2H and I; Supplementary Fig. S3D).

To dissect PSF-mediated signaling pathways in ER-positive breast cancer cells, we performed expression microarray analysis in MCF7 cells treated with PSF siRNA (Fig. 3A and B; Supplementary Fig. S4A; Supplementary Table S6). Pathway analysis using GSEA revealed that estrogen signaling pathway gene sets (“ESTROGEN_RESPONSE_EARLY” and “ESTROGEN_RESPONSE_LATE”) in Molecular Signatures Database were enriched among downregulated genes by siPSF in E₂-treated cells (Fig. 3C and D; Supplementary Fig. S4B and S4C). On the basis of the finding, we next examined whether PSF knockdown modulates estrogen responsiveness in ER-positive breast cancer cells. In MCF7, OHTR, and T47D cells, PSF knockdown significantly attenuated the expressions of estrogen-inducible genes such as GREB1, PGR, and STC2 (Fig. 3E and F; Supplementary Fig. S4D–S4J; ref. 34). Furthermore, PSF knockdown suppressed estrogen responsive element–driven reporter activity (Fig. 3G and H; Supplementary Fig. S4K). Notably, estrogen treatment did not change PSF mRNA levels (Supplementary Fig. S4L).

We hypothesized that ERα is a direct target of PSF because some RBPs were reported to target hormone receptors (21). In addition, we showed that ER-high labeling index is significantly associated with PSF strong IR in the two distinct clinical cohort studies (Supplementary Tables S2 and S4). In breast cancer cells, PSF knockdown significantly suppressed the expression of ESR1 mRNA (Fig. 4A and B; Supplementary Fig. S5A) and ERα protein (Fig. 4C and D; Supplementary Fig. S5B). As PSF is known to be involved in RNA splicing (35), we investigated whether PSF posttranscriptionally regulates ESR1 expression. We quantified premature ESR1 mRNA levels by qRT-PCR with primers specific for ESR1 intronic region (Supplementary Fig. S5C). PSF knockdown did not basically alter ESR1 intron levels (Fig. 4E and F; Supplementary Fig. S5D), suggesting that PSF knockdown downregulates ESR1 mRNA level via posttranscriptional regulation. We next assessed the interaction between PSF and ESR1 RNA. RIP assay using anti-PSF antibody revealed that PSF substantially interacts with ESR1 mRNA compared with negative control beta-2-microglobulin (B2M) mRNA (Fig. 4G and H; Supplementary Fig. S5E).

It has been reported that PSF could to be involved in short U snRNA nuclear export (36), whose mechanism remains unclear. We thus examined whether PSF contributes to ESR1 mRNA export from the nucleus in breast cancer cells. Lysates prepared from PSF siRNA-treated cells were fractionated into nuclear and cytoplasmic fractions and subjected to qRT-PCR. The results showed that PSF knockdown promotes nuclear accumulation of ESR1 RNA (Fig. 4I and J; Supplementary Fig. S5F).

We also established PSF stably overexpressing MCF7 cells (PSF #1 and #2) and control MCF7 cells stably transfected with empty vector (Vec #1 and #2; Supplementary Fig. S6A and S6B). Comparison studies of PSF and vector transformants showed that PSF overexpression substantially increases ERα expression at both RNA and protein levels (Supplementary Fig. S6C and S6D) and promotes cell proliferation (Supplementary Fig. S6E). We further showed that PSF-overexpressing cells exhibited increased viability compared with control vector-transfected cells in the presence of OHT (Supplementary Fig. S6F), suggesting that PSF contributes to tamoxifen resistance in hormone-dependent breast cancer cells.

To further understand the role of PSF in posttranscriptional regulation in ER-positive breast cancer cells, we explored PSF-interacting RNAs in MCF7 cells by RIP-seq using anti-PSF antibody (Fig. 5A). We showed significant interactions between PSF and its interacting RNAs including MALAT1, NEDD4L, and PPP3CA (Supplementary Fig. S7A–S7C and S7E), whereas not between PSF and B2M RNA, which was used as a negative control (Supplementary Fig. S7D; refs. 21, 37). We showed that ESR1 RNA was also significantly precipitated by anti-PSF antibody versus by input control (Supplementary Fig. S7F). RIP-seq analysis identified 2,264 and 2,329 anti-PSF–precipitated RNAs (fold enrichment ≥2; P < 1e-5), or PSF-associated RNAs, in MCF7 cells treated with vehicle (EtOH) and with E₂, respectively (Supplementary Fig. S7G) and 803 RNAs were common PSF-associated RNAs in both vehicle (EtOH) and E₂ treatment conditions. The rate of siPSF-downregulated RNAs in microarray analysis (≤0.8-fold) was significantly higher among PSF-associated RNAs than that among background RNAs, suggesting that the expression of PSF-associated RNAs could be more sensitive to PSF knockdown (Fig. 5B). To identify bona fide PSF target RNAs, we dissected 238 RNAs from the overlap of the 2,329 PSF-associated RNAs in E₂-treated cells and 1,831 siPSF-downregulated RNAs (≤0.8-fold) in microarray analysis (Fig. 5C). Among the 238 common RNAs, 34 mRNAs listed in Supplementary Table S7 positively correlate with PSF expression based on the coexpression analysis of PSF using breast cancer clinical data from the METABRIC dataset (Pearson correlation index > 0.2; n = 1,901; Fig. 5C and D; Supplementary Fig. S7H and S7I). Among these 34 RNAs, we validated that SCFD2, TRA2B, and ASPM mRNAs (Fig. 5E; Supplementary Fig. S8A) were downregulated by siPSF and bound to PSF protein as analyzed by microarray and RIP assay, respectively (Fig. 5F; Supplementary Fig. S8B). On the other hand, RT reaction–dependent pre-mRNA expressions (Supplementary Fig. S8C) of SCFD2, TRA2B, and ASPM were not significantly decreased by siPSF (Fig. 5G; Supplementary Fig. S8D), suggesting that PSF regulates the expressions of its target RNAs at posttranscriptional level. Similar to ESR1, PSF knockdown promoted nuclear accumulation of SCFD2 and TRA2B mRNAs (Fig. 5H; Supplementary Fig. S9A–S9C). ASPM mRNA, however, was not further accumulated by siPSF (Supplementary Fig. S9A–S9C), possibly because its basal ratio of nucleus versus cytoplasm was substantially higher than that of other RNAs.

Among the three validated PSF target RNAs, TRA2B and ASPM have been reported to promote the progression of breast cancer (38–40). Notably, Kaplan–Meier plotter platform showed that high mRNA expressions of TRA2B and ASPM are associated with a shorter relapse-free survival time in whole subtypes (Supplementary Fig. S9D and S9F) and ER-positive subtype (Supplementary Fig. S9E and S9G). Because the roles of SCFD2 in breast cancer have not been yet characterized, we next focused on SCFD2 activities. PSF knockdown decreased the expression of SCFD2 protein as well as mRNA (Fig. 5I; Supplementary Fig. S9H and S9I). In the mapping result of RIP-seq read tags in the vicinity of SCFD2 and adjacent gene loci, a substantial enrichment of anti-PSF–precipitated transcripts in SCFD2 locus was observed (Fig. 5J). SCFD2 mRNA expression was elevated in OHTR cells compared with parental MCF7 cells (Supplementary Fig. S9J) in parallel with PSF mRNA (Fig. 2A and B), and also in PSF-overexpressing MCF7 transfectants compared with control transfectants (Supplementary Fig. S9K). While PSF upregulates ERα expression, E₂ treatment did not further upregulate SCFD2 mRNA in MCF7 cells (Supplementary Fig. S9L).

To further define the association of SCFD2 with clinical pathogenesis, we analyzed publicly available breast cancer datasets. Oncomine-based expression analysis showed that SCFD2 mRNA is overexpressed in IDC and ILC samples compared with normal breast samples (Fig. 6A). In addition, SCFD2-high expression is associated with shorter relapse-free survival period in whole subtypes of patients with breast cancer (P = 1.6e-6; Fig. 6B) In patients with ER-positive subtype of breast cancer, cases with SCFD2-high expression have tendency to be associated with poorer prognosis (P = 0.062; Supplementary Fig. S10A). Notably, SCFD2 knockdown in MCF7, OHTR, and T47D cells significantly repressed cell proliferation (Fig. 6C–H; Supplementary Fig. S10B–S10D). We further evaluated the effect of SCFD2 overexpression on tamoxifen resistance. Stable SCFD2 overexpression in MCF7 cells (Fig. 6I and J) increased cell viability compared with control vector–transfected cells in the presence of OHT (Fig. 6K), suggesting that SCFD2 contributes to tamoxifen resistance in hormone-dependent breast cancer cells.

To examine the role of PSF in in vivo tumor growth of tamoxifen-resistant breast cancer cells, we generated OHTR-derived xenograft models and injected PSF siRNA or control siRNA twice weekly in OHTR-derived tumors. siPSF injection significantly suppressed OHTR-derived tumor growth without the alteration of mouse body weight (Fig. 7A and B; Supplementary Fig. S11A). The weights of tumors dissected 7 weeks after the first siRNA administration were significantly decreased in the siPSF-injected group compared with the siControl-injected group (Fig. 7C and D). Moreover, PSF, ESR1, and SCFD2 mRNA (Fig. 7E–G) and protein (Fig. 7H and I) expressions were decreased in siPSF-injected tumors compared with siControl-injected tumors, although the expressions of ESR1 and SCFD2 introns containing in the pre-mRNAs were not significantly altered (Supplementary Fig. S11B and S11C). IHC analysis showed that ER-labeling index was significantly (P = 0.010) decreased in siPSF-treated tumors (80.0 ± 4.5) compared with that in the siControl-treated tumors (86.7 ± 2.6; mean ± SD; n = 6 each; Supplementary Fig. S11D). In addition, a decrease in Ki67 IR was observed in siPSF-treated OHTR-derived tumors compared with siControl-treated tumors (Fig. 7J and K).

Taken together, PSF binds to ESR1, SCFD2, TRA2B, and ASPM RNAs and regulates their expression and nuclear export, suggesting that PSF would contribute to breast cancer progression (Fig. 7L).

In this study, we showed that PSF protein IR is significantly associated with poor prognosis of patients with breast cancer. PSF expression is higher in endocrine therapy–resistant breast cancer model OHTR cells compared with parental MCF7 cells. PSF knockdown could repress the proliferation of ER-positive breast cancer cells including OHTR cells as well as the growth of OHTR-derived xenograft tumors, suggesting that PSF could be a therapeutic target of primary ER-positive and endocrine therapy–resistant breast cancers. PSF posttranscriptionally regulates ESR1 mRNA but not pre-mRNA expression and promotes nuclear export of ESR1 mRNA. Integrative analyses of microarray-based transcriptome and RIP-seq–based interactome identified at least SCFD2, TRA2B, and ASPM as another PSF target. High SCFD2 mRNA expression is correlated with poorer breast cancer prognosis, and SCFD2 knockdown impairs breast cancer cell proliferation. Our results indicate that the PSF/ESR1 and PSF/SCFD2 axes would contribute to the pathophysiology of hormone-naïve and -refractory breast cancers.

We showed that PSF closely associates with estrogen signaling pathway via binding to ESR1 RNA and regulating its expression, suggesting that PSF modulates ESR1 RNA splicing. Oncogenic roles of some RBPs have been also reported in the context with hormone signaling. In breast cancer, MSI2 binds to the 3′-untranslated region of the ESR1 mRNA and maintains ERα protein stability (41). In prostate cancer, androgen receptor expression or splicing was modulated by RBPs such as Sam68 (42), HuR, poly(C)-binding protein (PCBP) 1, PCBP2 (43), and PSF (21). Notably, IHC analysis in patients with breast cancer showed that PSF IR and ERα-labeling index were significantly correlated with each other, suggesting that PSF would regulate ER expression in breast cancer tissues.

We showed that PSF promotes nuclear export of ESR1, SCFD2, and TRA2B mRNA. Because PSF knockdown did not modulate nuclear export of GAPDH mRNA, PSF may regulate nuclear export of specific target mRNAs. In previous study, PSF have been reported to facilitate nuclear export of short RNAs (36). In addition, PSF is involved in axon localization of several mRNAs in neuron cells to maintain axon viability (19), suggesting that PSF is a determinant of RNA subcellular localization as well as RNA splicing. Similar to PSF, another DBHS family RBP PSPC1 regulates adipose differentiation and development through promoting nuclear export of adipose mRNAs (44).

While we focused on the role of PSF in ER-positive breast cancer pathophysiology, other DBHS family members also have been reported to associate with cancer development through modulating transcription factor functions. PSPC1 promotes breast cancer metastasis through interacting Smad2/3 to enhance TGFβ signaling cascade (45). NONO physically interacts with lipogenic transcription factor SREBP-1A and regulates cell proliferation of breast cancer cells (46). As NONO forms a stable complex with PSF in vivo, the PSF–NONO complex also contributes to the repair of double-strand breaks and NONO attenuation leads to the repression of clonogenic survival of colon cancer cells following irradiation (47). PSF–NONO heterodimers are recruited to damaged DNA regions in the early stage of the double-strand break repair (48). In triple-negative breast cancer cells, IGFBP3 interacts with PSF and NONO in PARP-dependent double-strand break repair, which results in chemoresistance (49).

We newly identified SCFD2, TRA2B, and ASPM as PSF target RNAs in this study. TRA2B is a splicing factor that was reported to be involved in CD44 splicing (38). In addition, a recent study showed that TRA2B overexpression in untransformed MCF-10A cells promotes cell proliferation and metastasis via modulating alternative splicing of several RNAs related to cell proliferation and migration (39). ASPM is a centrosomal protein and also associated with the proliferation of breast cancer cells (40). As the upstream regulators of TRA2B and ASPM have not been well characterized, our findings in terms of PSF-mediated posttranscriptional regulation may provide new information for their gene expressions.

Notably, we demonstrated that SCFD2 could be a new regulator of breast cancer cell proliferation and a diagnostic and therapeutic target of breast cancer. SCFD2 is a member of the sec1-containing domain family (50) and was reported to be a p53-inducible gene (51). To date, no functional analysis of SCFD2 has been reported. Among sec domain proteins, STXBP1 was reported to play an important role in synaptic vesicle release (52), and STXBP1 gene mutation causes encephalopathy (53). SCFD1 associates with the trafficking of extracellular matrix like collagen and skeletal development in zebrafish (54). On the basis of the evidence, we assume that SCFD2 could contribute to breast cancer cell proliferation. Because PSF regulates expression of SCFD2 as well as ESR1, the attenuation of PSF–SCFD2 axis could be alternative option for advanced breast cancers with tamoxifen resistance.

As far as we examined in MCF7 cells, PSF and SCFD2 expression were not directly induced by estrogen treatment. Nevertheless, the transcription of both genes could be also modulated by ERα in some stages because the expression of PSF and SCFD2 was upregulated in tamoxifen-resistant OHTR cells. Our notion would be also supported by the existence of functional ERα−binding sites identified within and in the vicinity of SCFD2 gene based on chromatin immunoprecipitation database in UCSC Genome Browser. Further study will be required to know whether ERα plays a critical role in SCFD2 transcription in ER-positive breast cancer cells.

We demonstrated that the administration of PSF-specific siRNA significantly impaired in vivo tumor growth of tamoxifen-resistant breast cancer cells. Our results will highlight RBP-targeted small-nucleotide therapies using siRNAs or antisense oligonucleotides (ASO) as promising options for advanced breast cancer treatment. In a study of intravenous ASO administration, the specific ASO against eukaryotic translation initiation factor 4E significantly downregulated its target proteins such as VEGF, cyclin D1, survivin, c-myc, and Bcl-2, and suppressed the growth of xenograft tumors derived from human breast and prostate cancer cells without apparent toxicity based on the potential difference of ASO susceptibility in cancer and normal tissues (55). For the development of efficient small-nucleotide–based cancer therapy, improvement of drug delivery system would be also an important issue. In a study of transferrin receptor–targeted liposomal nanoparticle–based HuR siRNAs, the drug delivery system significantly repressed the growth of subcutaneous xenograft and metastatic tumors of human lung cancer cells (56).

In terms of in vivo tamoxifen resistance in xenograft tumors derived from in vitro tamoxifen-resistant cell models, previous studies showed that tamoxifen-resistant MCF7 cells could exhibit in vivo tamoxifen resistance also in xenograft models (57, 58). Further study will clarify whether PSF-mediated RNA regulation plays an important role in in vivo endocrine resistance of hormone-dependent breast cancers using other endocrine-resistant breast cancer models.

While it is the limitation of this study that we did not validate the reproducibility of microarray analysis with multiple replicated samples, we used the microarray results as exploratory data to dissect PSF target genes and further analyzed the statistical significance of the expression levels of the obtained PSF target genes in the validation study by qRT-PCR. Future pathway analysis based on transcriptomic studies with detail statistical analysis or with other cell systems will further highlight the functional properties of PSF-dependent mechanisms in breast cancers.

In conclusion, PSF plays a critical role in the cell proliferation of ER-positive breast cancer by regulating ESR1 and SCFD2 RNA expression at the posttranscriptional level. PSF and SCFD2 could be useful diagnostic and therapeutic targets of primary and endocrine therapy–resistant breast cancers.

No potential conflicts of interest were disclosed.

Conception and design: Y. Mitobe, K. Horie-Inoue, S. Inoue

Development of methodology: Y. Mitobe, K.-i. Takayama

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Mitobe, K. Iino, K.-i. Takayama, K. Ikeda, T. Suzuki, K. Aogi, H. Kawabata, Y. Suzuki

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-i. Takayama

Writing, review, and/or revision of the manuscript: Y. Mitobe, K.-i. Takayama, K. Horie-Inoue, S. Inoue

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.-i. Takayama

Study supervision: K. Horie-Inoue, S. Inoue

The authors thank Ms. Noriko Sasaki for technical assistance. The authors thank Dr. Keiichi Kinowaki for pathologic analysis of patients with breast cancer. This work was supported by Support Project of Strategic Research Center in Private Universities from the MEXT (to S. Inoue); grants from the Japan Society for the Promotion of Science (15K15353 to S. Inoue, 17H04205 to K. Horie-Inoue, 16K09809 to K. Iino, 17K18061 to Y. Mitobe, and 18J00252 to Y. Mitobe); by the Practical Research for Innovative Cancer Control (JP18ck0106194 to K. Iino) and the Project for Cancer Research And Therapeutic Evolution (P-CREATE; JP18cm0106144 to S. Inoue) from Japan Agency for Medical Research and Development, AMED; Takeda Science Foundation (to S. Inoue); and Mitsui Life Social Welfare Foundation (to S. Inoue).

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.