Sphingosine-kinase-1 signaling promotes metastasis of triple-negative breast cancer

Abstract

Triple negative breast cancer (TNBC) is the most aggressive breast cancer subtype. To identify TNBC therapeutic targets, we performed integrative bioinformatics analysis of multiple breast cancer patient-derived gene expression datasets and focused on kinases with FDA-approved or in-pipeline inhibitors. Sphingosine kinase 1 (SPHK1) was identified as a top candidate. SPHK1 overexpression or downregulation in human TNBC cell lines increased or decreased spontaneous metastasis to lungs in nude mice, respectively. SPHK1 promoted metastasis by transcriptionally upregulating the expression of the metastasis-promoting gene FSCN1 via NFκB activation. Activation of the SPHK1/NFκB/FSCN1 signaling pathway was associated with distance metastasis and poor clinical outcome in TNBC patients. Targeting SPHK1 and NFκB using clinically-applicable inhibitors (safingol and bortezomib, respectively) significantly inhibited aggressive mammary tumor growth and spontaneous lung metastasis in orthotopic syngeneic TNBC mouse models. These findings highlight SPHK1 and its downstream target NFκB as promising therapeutic targets in TNBC.

Introduction

Breast cancer, which arises mainly from mammary ducts or lobules, is the leading cause of cancer-related death and most commonly diagnosed cancer in women worldwide (1). Approximately 10%-20% of breast cancers are triple-negative, i.e., they do not express estrogen receptor (ER), progesterone receptor (PR), or human epidermal growth factor receptor 2 (HER2) (2, 3). Triple-negative breast cancer (TNBC) tends to occur at higher frequency in young women and is particularly aggressive, with high recurrence and metastasis rates (4). Compared with patients having other subtypes of breast cancer, TNBC patients have a poor overall prognosis, e.g., the 5-year survival rate for patients with stage IV TNBC is about 22%, mainly due to early-onset of metastasis (4). It has been reported that TNBC tumors are about 2.5 times more likely to metastasize within 5 years than are breast tumors of other subtypes (5). Since TNBC tumors lack expression of hormone and HER2 receptors, i.e., negative for therapeutic targets, TNBCs do not respond to, and patients cannot benefit from, currently available hormonal and HER2-targeted therapies.

In contrast to the successful development of therapies for hormone receptors positive, and/or HER2 positive breast cancers, little progress has been made in identifying positively expressed molecular targets in TNBC that are druggable (6). Clearly, there is an imposing need to discover positive druggable targets in TNBC instead of accepting its triple negative non-targetable status. Kinases play central roles in cancer cell signaling pathways and are druggable targets for effective targeted therapies (7). In the past decade, numerous efforts have led to successful development and FDA approval of inhibitors of various cancer-promoting kinases (8). Therefore, we set out to identify activated and/or overexpressed kinases, as positive and druggable molecular targets, in TNBC with high potential for quick and efficient clinical translation.

Our bioinformatics analysis of multiple patient-derived datasets identified that sphingosine kinase 1 (SPHK1), a lipid kinase, was expressed at significantly higher levels in TNBC than in other breast cancer subtypes. SPHK1 catalyzes phosphorylation of sphingosine, an amino alcohol, to generate sphingosine-1-phosphate (S1P), a novel lipid signaling mediator with both intracellular (as a second messenger) and extracellular (as a ligand for G-protein-coupled receptors) functions (9). S1P regulates various cellular processes in mammalian cells, such as growth, survival, and migration. Exogenously overexpressing SPHK1 in 3T3 fibroblasts led to transformation in vitro and tumor formation in vivo, suggesting that SPHK1 acts as an oncogene (10). SPHK1 is shown to be overexpressed in various cancers including breast cancer (11-14). Importantly, a SPHK1 inhibitor, safingol, can effectively inhibit SPHK1 activities and is currently under multiple clinical trials (, ).

In this study, we systematically tested the function of SPHK1 in TNBC progression and metastasis using multiple TNBC spontaneous metastasis models that recapitulate the entire cascade of biological steps of metastasis in patients and found that SPHK1 has a critical function in enhancing TNBC spontaneous metastasis. Mechanistically, SPHK1 upregulates FSCN1 (also known as fascin) transcription via activation of the NFκB transcriptional factor, and FSCN1 promotes metastasis. Clinically, SPHK1/NFκB/FSCN1 signaling pathway activation in patients’ TNBC tissues correlates with poor patient survival and increased metastases. To test the validity of SPHK1 pathway as druggable targets in TNBC, we therapeutically targeted SPHK1 by safingol and/or NFκB with a clinically-applicable inhibitor bortezomib. Strikingly, combinatorial treatment with both SPHK1 and NFκB inhibitors significantly inhibited both TNBC primary tumor growth and lung metastasis compared to either single agent treatment. These data demonstrated that SPHK1/NFκB pathway can serve as a positive therapeutic targets for effective inhibition of TNBC and metastasis. These preclinical findings could be fast-track translated to the clinic for the treatment of TNBC and metastasis in patients.

Materials and methods

Cell culture.

Human cancer cell lines (MCF-7, T47D, BT474, HCC1954, HCC70, Hs578T, MDA-MB-231, MDA-MB-435, and MDA-MB-436) and a mouse breast cancer cell lines (4T1), were obtained from the American Type Culture Collection. Mouse breast cancer cell line E0771 and Met-1fvb2 were purchased from CH3BioSystems and Lonza respectively. These cell lines were verified by the MD Anderson Cancer Center Cell Line Characterization Core Facility. BC3-p53KD were provided by Dr. H. Piwnica-Worms (15). Cells were cultured in Dulbecco modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) and 0.1% penicillin-streptomycin in 5% CO₂ at 37°C. All cell lines were tested for mycoplasma contamination by using the MycoAlert Mycoplasma Detection Kit (Lonza) and were negative.

Antibodies and reagents.

Rabbit polyclonal antibody against Sphk1 (HPA022829) and FSCN1 (HPA005723) were purchased from Sigma-Aldrich. Mouse polyclonal antibody against FSCN1 1 (sc-46675) was bought from Santa Cruz Biotechnology, and rabbit monoclonal antibody against SPHK1 (ab109522) was bought from Abcam. Mouse monoclonal anti-β-actin antibody (A5441) was from Sigma-Aldrich, and anti-Ki67 antibody (M7240) was from Dako. Normal rabbit IgG (2729), NFκB p65 (8242), and H3K4me3 (9727) were all from Cell Signaling Technology. p- NFκB p65 (S536) (ab86299) was purchased from Abcam. The in situ cell death detection kit (TUNEL technology, 11684817910) was from Roche. The horseradish peroxidase–linked secondary antibodies against mouse (NA931) and rabbit (NA934) were from GE Healthcare. Actinomycin D (A9415) was from Sigma-Aldrich. A cell permeable peptide (CAS 213546-53-3) that inhibits translocation of the NFkB active complex into the nucleus and its corresponding NFkB control peptide (sc-3060) were brought from Santa Cruz Biotechnology. Safingol (CAS 15639-50-6) was purchased from Cayman Chemical. Bortezomib (CAS 179324-69-7) was purchased from EMD Millipore. CAPTISOL^® (20g) was kindly provided by Cydex Pharmeceuticals.

Generation of stable cell-lines.

To overexpress SPHK1, retroviral vector pWZL-Neo-Myr-Flag-DEST containing the SPHK1 open reading frame (ORF) under the control of CMV promotor with G418 (100 μg/ml) as selection marker was used (kindly provided by Dr. J. Zhao). Empty vector was used as a control. To stably knock down SPHK1 in MDA-MB-435, Hs578T, and BC3-p53 KD cells, we used two small hairpin RNA (shRNA) constructs, targeting the SPHK1 3’ untranslated region, cloned into the pGIPZ lentiviral vector (RefSeq NM_001142601, Open Biosystems) with puromycin (2 μg/ml) as selection marker. To stably knock down FSCN1 in MDA-MB-435 cells, we used three shRNA constructs, targeting the FSCN1 3’ untranslated region, cloned into the pGIPZ lentiviral vector (RefSeq NM_003088.3, Open Biosystems) with puromycin (2 μg/ml) as selection marker. Non-silencing shRNA was used as a control for both above mentioned shRNA knockdown experiments. To overexpress FSCN1, retroviral vector pLenti6/V5-DEST containing the FSCN1 ORF under the control of CMV promotor with blasticidin (3 μg/ml) as selection marker was used (plasmid #31207, Addgene). Lentiviral vector with mCherry sequence was used as a control. Lentiviral vectors (with ORFs or shRNA) were transfected into the packaging cell line 293T, together with a packaging DNA plasmid (psPAX2) and an envelope DNA plasmid (pMD2G), through Lipofectamine transfection. After 48 h, viruses were collected, filtered, and incubated with target cells in the presence of 8-10 μg/mL Polybrene for 24 h. The infected cells were selected with suitable selection markers, with concentration mentioned above, to generate the stable clone.

siRNA knockdown.

To knockdown FSCN1 in MDA-MB-231 cells, SMARTpool: ON-TARGETplus human FSCN1 siRNA (Dharmacon™) was used to transfect the cells. Lipofectamine^® RNAiMAX transfection reagents (Invitogen™) was used for siRNA transfection and the protocol was followed per manufacturer’s instruction.

Site-directed mutagenesis.

Nucleotides within the NFκB transcription factor binding sites of the FSCN1 promoter were altered by site-directed mutagenesis by using Q5 site-directed mutagenesis kit (NEB) according to the manufacturer’s instructions. Wild-type FSCN1 promoter (−333/+147 bp) in the pGL3luc (basic) vector was used as a template to introduce mutation at three nucleotides that affect both binding sites. The sense and anti-sense primers used were 5’- GTCCGAGGTGATGGACATCAGGGG-3’ and 5’- ACCCCGACCCCAAGCCTC -3’, respectively. Mutated promoter fragments were sequenced to verify the presence of mutations.

Western blotting.

Western blot analysis was performed as previously described (16). Densitometry analyses were performed by evaluating band intensity mean gray value of indicated protein and normalizing it with the mean gray value of corresponding lane’s loading control using ImageJ software.

RNA extraction, reverse transcriptase–polymerase chain reaction (RT-PCR), and quantitative real-time PCR (qPCR).

RNA extraction and RT-PCR were performed as described previously (16). For the SYBR green-based qPCR assay, 1 μL of cDNA was used as a template for quantitative real-time PCR with iQ™ SYBR Green Supermix (Biorad) and the StepOnePlus™ (Applied Biosystem) instrument according to the manufacturer’s instruction. The RNA expression rate was quantified by the relative quantification (2-Δ ΔCt) method, and 18S expression was used as the internal control. The primers that were used are listed below:

Genes	Primers
SPHK1	F: 5’-AACTACTTCTGGATGGTCAG -3’ R: 5’-TCCTGCAAGTAGACACTAAG -3’
FSCN1	F: 5’-CCAGGGTATGGACCTGTCTG-3’ R: 5’-CGCCACTCGATGTCAAAGTA-3’
18S	F: 5’-AACCCGTTGAACCCCATT-3’ R: 5’-CCATCCAATCGGTAGTAGCG-3’

ChIP assay.

Procedures for chromatin isolation and immunoprecipitation were performed as previously described (16). Normal IgG, NFκB p65, and H3K4me3 antibodies were used at 2 μg per reaction in immunoprecipitation. Co-precipitated DNA (2 μl) was analyzed by quantitative PCR. The forward and reverse primers used for amplification of the NFκB binding region in the FSCN1 promoter (−333/+147 bp) were as follows: forward: 5’- CTCAAACCTCGCTCGTCCTT-3’ and reverse: 5’- CATCACCCCTCACAACCCC -3’.

Three-dimensional (3D) cell culture.

3D culture was performed in either an 8-well chamber slide (BD Falcon) or in Costar 6-well plate with ultra-low attachment surface (Corning). For the 8-well chamber, 100 μl of Matrigel was added to the bottom of each chamber and incubated at 37°C for 20 min. Cells of interest were mixed in culture medium with 5% Matrigel and added to each well to a final concentration of 1500 cells/well. For 6-well plates with low attachment, 2 × 10⁵ cells of interest that were mixed into the culture medium with 5% Matrigel were added to each well.

Quantification of 3D invasiveness.

Indicated cells were grown in 3D culture in 8-well chamber culture slides. Images of the spheroid structures at multiple fields were obtained with use of a microscope at the indicated time, and invasive structures per field were counted. Structures that had projections coming out from the main spheroid body were counted as invasive structures.

mRNA stability assay.

Equal numbers of 435 shSCR and 435 shSPHK1.1 cells were plated in a 6-well low-attachment plate with 5% Matrigel and incubated for 3 d at 37°C. Cells were treated with 5 μg/mL actinomycin D for 1, 2, 4, 8, 12, and 16 h. Total RNA was extracted after each time-point by using TRIzol reagent (Invitrogen), and quantitative PCR was performed to determine the relative mRNA level of FSCN1.

Immunohistochemical analysis.

The excised MFP tumors were fixed in 10% neutral buffered formalin and embedded in paraffin for immunohistochemical (IHC) staining. IHC staining was performed similarly as previously described (16). TUNEL staining was done in paraffin sections with use of an in situ cell death detection kit, POD (11684817910, Roche), according to the manufacturer’s instructions.

SPHK1 kinase activity.

SPHK1 activity in cytosol was determined as described previously (17). The intracellular level of S1P in SPHK1 modulated cells were quantified using S1P ELISA kit (Echelon Biosciences, K-1900) and protocol was followed as per manufacturer’s instructions.

Animal experiments.

All procedures and experimental protocols involving mice were approved by the Institutional Animal Care and Use Committee at The University of Texas MD Anderson Cancer Center.

Female nude mice (6 weeks old, 4-7 mice per group as indicated in figures and/or figure legends) were orthotopically injected with human cancer cells (2 × 10⁵ cells for MDA-MB-435 and MDA-MB-231 cells, 1 × 10⁶ cells for BC3-p53KD cells; cells were re-suspended in 50:50 mixture of Matrigel in PBS) into mammary fat pads (mfps), and tumors were allowed to develop for an indicated number of days. Tumor sizes were measured with digital calipers twice a week, and tumor volumes were calculated with use of a modified ellipsoidal formula: 1/2 × (length × width²). MFP tumors were surgically excised with survival surgery, and the mice were further monitored for an indicated number of weeks for spontaneous metastasis. All mice were euthanized at indicated times, and lungs were harvested, fixed and paraffin embedded. After lungs were H&E stained, the number of metastatic lesions were enumerated by pathologist using brightfield microscopy. Detailed description of in vivo treatment experiments in mice is provided in Supplementary Methods.

cDNA microarray and analysis.

Unbiased platform, HumanHT-12_v4 (Illumina), was applied for gene profiling of MFP tumors and matched spontaneous lung metastasis formed by control and Sphk1 knockdown MDA-MB-435 cells in collaboration with the cDNA microarray core facility at MD Anderson Cancer Center. The raw and normalized microarray data have been deposited in the GEO database under accession number GSE128624. Gene cluster maps for MFP and lung metastasis samples were generated by using sequence analysis of microarray (SAM) analysis. To identify SPHK1-regulated genes in 435 cells, R software and limma software packages were used to identify differentially expressed genes using a 1.5-fold change threshold and an adjusted p value cutoff at 0.01. Ingenuity Pathway Analysis (IPA) software (http://www.ingenuity.com) was used to perform the functional annotation and pathway analysis of the differentially expressed genes. Gene set enrichment analysis was performed on MFP microarray data with use of an online tool (http://software.broadinstitute.org/gsea/index.jsp), as described previously (18).

Case Selection, Tissue Microarray (TMA) Construction and analysis.

We obtained archival, formalin-fixed and paraffin-embedded (FFPE) material from surgically resected breast cancer specimens from the Breast Tumor Bank at M. D. Anderson Cancer Center from 2001 to 2013 (Houston, TX). Tumor tissue specimens obtained from 117 triple negative breast cancers were histologically examined, classified using the World Health Organization (WHO) classification of Breast Tumors and selected for TMA construction. After histologic examination, tumor TMAs were prepared using triplicate 1-mm-diameter cores per tumor. All the archival paraffin-embedded tumor samples were coded with no patient identifiers. Detailed clinical and pathologic information, including demographic, pathologic TNM staging, overall survival, and time of recurrence were collected. Detailed descriptions of IHC staining/quantification and survival analysis are provided in Supplementary Methods.

Bioinformatics, statistics, and survival analysis.

GEO2R analysis, a Web-based application for analyzing gene expression in Gene Expression Omnibus (GEO) data sets, was performed as described elsewhere (19). The Kaplan-Meier plotter (20), a Web-based tool, was used to assess the effect of the SPHK1 and the FSCN1 genes on survival of TNBC patients. To select the TNBC patients, following selection criteria were used: ER negative, PR negative, HER2 negative, grade 3 and intrinsic subtype basal. Survival rates were compared by using the log-rank test, and hazard ratios were calculated by using a multivariable Cox proportional hazards model. For correlation analysis, expression values of SPHK1 and FSCN1 from patient samples were downloaded from The Cancer Genome Atlas (TCGA) and Curtis breast dataset (21). GraphPad Prism (Prism 6; GraphPad Software Inc.) was used to generate a correlation graph and calculate the Pearson coefficient (r) from the downloaded data. All statistical analyses were performed by using GraphPad Prism. The data were analyzed by either one-way analysis of variance (multiple groups) or a t test (two groups). Differences with p < 0.05 (two-sided) were considered statistically significant. *p<0.05, **p<0.01, ***p<0.001 by two-tailed t-test.

More methods (Plasmids construction; Cell proliferation assay; Migration assay and invasion assay; Transient transfection and luciferase reporter assay; and Flow cytometry) with detailed descriptions are provided in Supplementary Methods.

Results

SPHK1 is highly expressed in TNBCs and promotes spontaneous lung metastasis.

To identify kinase gene(s) that are particularly overexpressed in TNBC to serve as potential therapeutic targets, we performed GEO2R analysis (GSE27447) between TNBC and non-TNBC tumor samples from patients (22). Only four kinase genes were among the top 100 differentially expressed genes, of which SPHK1 was the only kinase that was overexpressed in TNBC tumors compared with non-TNBC tumors (Supplementary Fig.S1A-S1B). Our further analysis of breast cancer microarray data of TCGA validated that expression of SPHK1 is significantly higher in basal subtype when compared with normal breast tissues and other subtypes of breast cancers (Fig.1A).

Figure 1: Sphk1 is overexpressed in TNBC tumors and cell-lines and SPHK1 overexpression promotes spontaneous metastatic spread to lungs.

(A) Box-and-whisker plot showing the expression of SPHK1 in normal breast tissue and various subtypes of breast cancers. (B) Kaplan-Meier plotter was used to generate a survival curve of TNBC patients (total n = 112), which were stratified based on the SPHK1 expression. (C) Quantitative reverse transcriptase PCR (qRT–PCR) showing the relative expression of SPHK1 in various human breast cancer cell-lines. (D) Western blotting analysis showing the SPHK1 expression in various human breast cancer cell-lines as indicated. (E) qRT-PCR showing the relative expression of SPHK1 in mRNA level in MDA-MB-231 cells transduced with empty vector or SPHK1 overexpressing vector. (F) A western blot showing the expression of SPHK1 in protein level in MDA-MB-231 cells transduced with empty vector or SPHK1 overexpressing vector. (G) ELISA of intracellular S1P Levels in MDA-MB-231 cells transduced with empty vector or SPHK1 overexpressing vector. (H) In vitro kinase assay for the detection of sphingosine-1-phosphate (S1P) in indicated cells. (I) In vivo mammary fat pad (MFP) tumor growth in nude mice with orthotopic injection of 231vec and 231SPHK1 cells. (J) H&E stained lung sections were quantified for the number of spontaneous metastatic lesions from nude mice with orthotopic injection of 231vec and 231SPHK1 cells (left). Representative images of H&E stained lung sections are shown (right). Scale bar: 200 μm. Lum, Luminal; HR, Hazard ratio; NC, negative control; PC, Positive control. Data are represented as mean ± SD. *, p < 0.05, **, p < 0.01 and ***, p < 0.001, n.s., non-significant.

Survival analysis by Kaplan-Meier (KM)-plotter indicated that expression of SPHK1 is significantly associated with poorer relapse-free survival in TNBC patients (Fig.1B. SPHK1 expression is also higher in TNBC-derived cell-lines than in cell-lines of other subtypes at both mRNA (Fig.1C and Supplementary Fig.S1C-S1D) and protein (Fig.1D and Supplementary Fig.S1E) levels. Thus, SPHK1 is mostly overexpressed in TNBC tumors and cell-lines.

To determine the function of SPHK1 in TNBC progression and metastasis, we first performed gain-of-function studies using MDA-MB-231 human TNBC cell-line that express an intermediate level of endogenous SPHK1 compared with other breast cancer cell-lines, and generated control (231vec) and SPHK1-overexpressing (231SPHK1) stable sublines (Fig.1E-1F). 231SPHK1 cells showed increased levels of intracellular S1P compared to 231vec cells (Fig. 1G). 231SPHK1 cells showed increased phosphorylation of sphingosine leading to increased S1P as detected by a semi in vitro kinase assay compared to 231vec cells (Fig.1H). 231SPHK1 cells also showed increased migration (Supplementary Fig.S2A) and invasion (Supplementary Fig.S2B) potential in vitro, but no significant differences in cell proliferation in vitro (Supplementary Fig.S2C) compared to 231vec cells.

To examine whether SPHK1 promotes TNBC tumorigenesis and spontaneous metastasis, 231vec control and 231SPHK1 cells were orthotopically injected into mammary fat pads (mfp) of nude mice. The mfp tumors were surgically excised 28 days post-injection, and the mice were monitored for about 10 more weeks for development of spontaneous metastasis (Supplementary Fig.S2D). There was no significant difference in mfp tumor size between 231vec and 231SPHK1 groups by day 28 (Fig.1I). Remarkably, the number of metastatic lesions in the lungs was significantly higher in mice bearing 231SPHK1 mfp tumors than in control mice bearing 231vec mfp tumors (Fig.1J). These data showed that despite having no significant effect on mfp tumor growth, SPHK1 overexpression enhanced spontaneous lung metastasis of MDA-MB-231 human TNBC cells.

SPHK1 knockdown decreases spontaneous lung metastasis.

To determine whether SPHK1 is required for TNBC progression and metastasis, we stably knocked down SPHK1 in TNBC patient-derived xenograft (PDX) cells (BC3-p53KD) (15) using lentiviral vector expressing two distinct SPHK1-targeting shRNAs (shSPHK1.1 and shSPHK1.2). Lentiviral vector expressing non-targeting scrambled shRNA (shScr) was used to generate the control cell line. Additionally, since PDX cells are not amendable for some in vitro assays, we also knocked down SPHK1 in two other TNBC cell-lines [MDA-MB-435 (23) and Hs578T] expressing relatively high levels of endogenous SPHK1. SPHK1 knockdown by shSPHK1.1 was highly effective at the mRNA level detected by qRT-PCR (Supplementary Fig.S3A) and at the protein level determined by western blotting (Fig.2A and Supplementary Fig.S3B). Knocking down SPHK1 had no significant effect on cell proliferation in vitro in any of the three cell-lines (Supplementary Fig.S3C). SPHK1 knockdown in MDA-MB-435 cells showed decreased levels of intracellular S1P compared to control (Supplementary Fig.3D). SPHK1 knockdown in MDA-MB-435 and Hs578T cells led to decreased phosphorylation of sphingosine leading to decreased S1P as detected by a semi in vitro kinase assay (Supplementary Fig.S3E), and inhibited the migration and invasion potential in vitro (Supplementary Fig.3SF-SH).

Figure 2: SPHK1 knockdown decrease spontaneous metastatic spread to lungs.

(A) A western blot showing the expression of SPHK1 in BC3-p53KD and MDA-MB-435 cells transduced with lentiviral vectors expressing control shRNA (shScr) or SPHK1 targeting shRNA (shSPHK1.1 or shSPHK1.2). (B) MFP tumor growth curve in nude mice with orthotopic injection of BC3-p53KD.shScr and BC3-p53KD.shSPHK1.1cells. (C) MFP tumor growth in nude mice with orthotopic injection of 435.shcr, 435.shSPHK1.1 and 435.shSPHK1.2 cells. (D) H&E stained lung sections were quantified for the number of spontaneous metastatic lesions from nude mice with orthotopic injection of BC3-p53KD.shScr and BC3-p53KD.shSPHK1.1cells (left). Representative images of H&E stained lung sections are shown (right). (E) H&E stained lung sections were quantified for the number of spontaneous metastatic lesions from nude mice with orthotopic injection of 435.shScr, 435.shSPHK1.1 and 435.shSPHK1.2 cells (left). Representative images of H&E stained lung sections are shown (right). Data are represented as mean ± SD. *, p < 0.05, **, p < 0.01 and ***, p < 0.001, n.s., non-significant.

To determine whether SPHK1 is required for TNBC tumorigenesis and spontaneous metastasis, SPHK1 knockdown BC3-p53KD.shSPHK1.1 and MDA-MB-435 (shSPHK1.1 and shSPHK1.2) cells along with respective control cells were orthotopically injected into mfp of nude mice. The mfp tumors were resected by survival surgery on day 54 for mice injected with BC3-p53KD sublines, and on day 28 for mice injected with MDA-MB-435 sublines (Supplementary Fig.S4A-S4B). There was no significant difference in primary tumor size (Fig.2B-2C) or in tumor cell proliferation (Supplementary Fig.S4C-S4D) between control and SPHK1 knockdown mfp tumors in either models, although the mfp tumors of SPHK1 knockdown cells were more apoptotic as detected by TUNEL staining than did tumors formed by control cells (Supplementary Fig.S4E-S4F).

To evaluate the impact of SPHK1 knockdown on metastasis, lungs were harvested from mice bearing BC3-p53KD mfp tumors at 187 days after mfp tumor resection, and from mice bearing highly aggressive MDA-MB-435 mfp tumors at 47 days after primary tumor resection (Supplementary Fig.S4A-S4B). A significant decrease in the number of spontaneous lung metastases were observed in mice injected with SPHK1 knockdown cells compared with mice injected with control cells in both models (Fig.2D-2E). Together, these data indicate that SPHK1 is essential for TNBC’s development of aggressive spontaneous lung metastasis and the SPHK1 overexpression may serve as a positive therapeutic target for inhibition of TNBC lung metastasis.

FSCN1 upregulation contributes to SPHK1-driven metastasis.

SPHK1 is a critical lipid kinase with pleiotropic effects on various cellular functions (9), however, little is known about how SPHK1 promotes metastasis. To attain insights on the molecular mechanism underlying SPHK1-driven spontaneous metastasis, the primary mammary tumor tissues and matched spontaneous lung metastasis tissues from mice injected with 435.shScr versus 435.shSPHK1.1 cells (see Fig.2C and 2E) were profiled for SPHK1-modulated genes (Supplementary Fig.S5A-S5B). The list of SPHK1-regulated differentially expressed genes (435.shSPHK1.1 vs 435.shSCR) was generated for mfp tumor samples and was subjected to Ingenuity Pathway Analysis (IPA). IPA identified FSCN1 as a top SPHK1-regulated gene involved in cancer cell migration, invasion, and metastasis (Fig.3A and Supplementary Fig.S5C). Indeed, the FSCN1 gene expression was high in SPHK1 high-expressing tissue sample (i.e. 435.shScr) and low in SPHK1 low-expressing tissue samples (i.e. 435.shSPHK1.1) of both primary tumors and spontaneous lung metastases as validated by qRT-PCR (Fig.3B and Supplementary Fig.S5D) and by western blotting (Fig.3C and Supplementary Fig.S5E). Similarly, other SPHK1 high-expressing mfp tumor samples (i.e. 231SPHK1 and BC3-p53KD.shScr) also had higher FSCN1 expression than SPHK1 low-expressing mfp tumors (i.e. 231vec and BC3-p53KD.shSPHK1.1, respectively) as shown by western blotting (Fig.3D-3E). Furthermore, in two different breast cancer patients’ datasets [TCGA and Curtis breast datasets (21)], SPHK1 gene expression correlated with FSCN1 gene expression, and high expressions of both SPHK1 and FSCN1 genes was observed in TNBC patients’ compared to other subtypes (Fig.3F and Supplementary Fig.S5F). FSCN1 expression, like SPHK1 expression, was also upregulated in basal subtype compared with other subtypes of breast cancer in TCGA dataset (Fig.3G); moreover, FSCN1 expression correlated with a poor survival rate in TNBC patients (Fig.3H).

Figure 3: Positive correlation between expression of SPHK1 and expression of FSCN1 in orthotopic MFP tumors and TNBC patients.

(A) Venn diagram representing the number of differentially regulated genes in MFP tumor samples from 435.shScr that are involved in migration, invasion and metastasis pathways as identified by IPA (Ingenuity pathway analysis) analysis. Fifteen differentially regulated and common genes that are involved in metastasis related properties are listed according to the fold change difference between MFP tumor samples from 435.shSPHK1.1 vs. 435.shScr cells. (B) qRT-PCR showing the relative expression of FSCN1 in mRNA level in MFP tumor samples from 435.shScr and 435.shSPHK1.1 cells. (C) A western blot showing the expression of FSCN1 in protein level in MFP tumor samples from 435.shScr and 435.shSPHK1.1 cells. (D) A western blot showing the expression of FSCN1 and SPHK1 in protein level in MFP tumor samples from 231vec and 231SPHK1 cells. (E) A western blot showing the expression of FSCN1 and SPHK1 in protein level in MFP tumor samples from BC3-p53KD.shScr and BC3-p53KD.shSPHK1.1 cells. (F) Correlation between SPHK1 expression and FSCN1 expression in primary tumor samples of breast cancer patients from TCGA dataset. Gray circle represents TNBC patients and black circle represents patients with other subtypes. (G) Box-and-whisker plot showing the expression of FSCN1 in normal breast tissue and various subtypes of breast cancers. (H) Kaplan-Meier plotter was used to generate a survival curve of TNBC patients (total n = 112), which were stratified based on the SPHK1 expression. Data are represented as mean ± SD. *, p < 0.05, **, p < 0.01 and ***, p < 0.001, n.s., non-significant.

The FSCN1 proteins organize F-actin into parallel bundles and are required for the formation of actin-based cellular protrusions (24). It plays a critical role in cell migration, motility, and adhesion and in cellular interactions (25, 26). To test whether FSCN1 can recover metastasis from SPHK1 knockdown cells, FSCN1 was ectopically expressed in 435.shSPHK1.1 and Hs578T.shSPHK1.1 cells (Supplementary Fig.S6A-S6D). Clearly, ectopic expression of FSCN1 in SPHK1 knockdown cells increased their migration (Fig.4A and Supplementary Fig.S6E-S6F) and invasion (Fig. 4B and Supplementary Fig.S6G-S6H) potentials in vitro, but it had no significant effect on cell proliferation in vitro (Supplementary Fig.S6I-S6J). Importantly, ectopic expression of FSCN1 in 435shSPHK1.1 cells rescued the spontaneous metastatic potential in vivo (Fig.4C). On the other hand, knocking down FSCN1 gene in SPHK1 overexpressing 231SPHK1 cells (Supplementary Fig.S7A) and in MDA-MB-435 cells having high endogenous SPHK1 expression (Supplementary Fig.S7B-S7C) significantly reduced their migration (Fig.4D and Supplementary Fig.S7D) and invasion (Fig.4E and Supplementary Fig.S7E), indicating that the SPHK1-induced metastasis-related functions are mediated, at least partially, by FSCN1.

Figure 4: Fascin might mediate SPHK1 metastasis function.

(A) Quantification of transwell migration assay for indicated cells. (B) Quantification of transwell invasion assay using matrigel for indicated cells. (C) H&E stained lung sections were quantified for the number of spontaneous metastatic lesions from nude mice with orthotopic injection of 435 shSPHK1.1 cells transduced with either mCherry or FSCN1 (left). Representative images of H&E stained lung sections are shown (right). (D) 231SPHK1 cells were transfected with either siCtrl or siFSCN1 (pooled) and were subjected to transwell migration assay. Cells migrating through transwell membrane were imaged and representative images are shown (right) along with quantification (left). (E) 231SPHK1 cells were transfected with either siCtrl or siFSCN1 (pooled) and were subjected to transwell invasion assay. Cells invaded through transwell membrane were imaged and representative images are shown (right) along with quantification (left). (F) BC3-p53KD cells, transduced with either control shRNA (shScr) or SPHK1 targeting shRNA (shSPHK1.1) were grown in three dimensional (3D) culture system and cell colonies with invasive structures and projections were quantified (left) on both cell-lines and representative brightfield images are shown (right). (G) 231SPHK1 cells, transduced with either control siRNA (shCtrl) or FSCN1 targeting pooled siRNA (siFSCN1) were grown in 3D culture system and cell colonies with invasive structures and projections were quantified (left) on both cell-lines and representative brightfield images are shown (right). Data are represented as mean ± SD. *, p < 0.05, **, p < 0.01 and ***, p < 0.001, n.s., non-significant. Scale bar: 200 μm.

Regarding FSCN1-mediated metastasis-related functions, various studies have shown that FSCN1 is important for invadopodia assembly to promote protrusive invasion of cancer cells (27, 28). Thus, we tested whether SPHK1 high-expressing cells with high FSCN1 expression may induce more protrusive invasion in three dimensional (3D) culture. Indeed, SPHK1 high-expressing BC3-p53KD.shScr and 435.shScr cells with high FSCN1 expression formed many protrusive structures projecting into the surrounding matrix after 10 days in standard 3D culture; whereas BC3-p53KD.shSPHK1.1 and 435.shSPHK1.1 cells with low FSCN1 expression mostly formed round shaped structures with very few protrusions (Fig.4F and Supplementary Fig.S7F). Importantly, knocking down FSCN1 gene in SPHK1-overexpressing 231SPKH1 cells (Fig.4G) and MDA-MB-435 cells (Supplementary Fig.S7G) reduced their protrusive structures projecting into the surrounding matrix in standard 3D culture. These data indicate that FSCN1 upregulation could contribute to SPHK1-driven invasion and metastasis by inducing protrusive invasion.

SPHK1 upregulates FSCN1 transcription via activation of NFκB.

The critical function of FSCN1 in SPHK1-driven invasion and metastasis impelled us to dissect how FSCN1 is upregulated by SPHK1. Interestingly, we found that knockdown SPHK1 down-regulated FSCN1 expression in standard 3D culture but not in two dimensional (2D) culture at both mRNA (Fig.5A and Supplementary Fig.S8A) and protein (Fig.5B and Supplementary Fig.S8B) levels (29). Next, we tested whether SPHK1 regulates FSCN1 mRNA via modulating mRNA stability or transcription. The control and SPHK1 knockdown MDA-MB-435 cells in 3D culture were treated with actinomycin D to block transcription and were detected for FSCN1 mRNA stability, which showed no significant difference (Supplementary Fig.S8C), indicating that SPHK1 does not modulate FSCN1 mRNA stability.

Figure 5: SPHK1 upregulates FSCN1 gene expression at transcriptional level through NFkB transcription factor.

(A) qRT-PCR showing the relative mRNA expression of FSCN1 in BC3-p53KD (left) and MDA-MB-435 (right) cells transduced with either shScr or shSPHK1.1 in two dimensional (2D) and three dimensional (3D) culture system. (B) A western blot showing the expression of FSCN1 in BC3-p53KD (top) and MDA-MB-435 (bottom) cells transduced with either shScr or shSPHK1.1 in 2D and 3D culture system. (C) Relative promoter activity assay in 435.shScr and 435.shSPHK1.1 cells transduced with luciferase reporter plasmid containing FSCN1 promoter (−1376/+147 bp) cultured in 2D and 3D culture system. Firefly luciferase activity is reported after normalizing to renilla activity, which is used as internal control for transfection variability. (D) Promoter activity assay in MDA-MB-435 (left) and Hs578T (right) cells transduced with either control shRNA (shScr) or shRNA targeting SPHK1 (shSPHK1.1) in 3D culture system. Various 5’ deletion constructs were transduced into these cell-lines as indicated. NT, Non-transfecting. (E) Top: Schematic diagram showing various transcription factor binding sites in FSCN1 promoter region (−333/−93 bp). Bottom: nucleotide sequences of wild-type NFκB binding site (NFκB BSwt) and NFκB binding site with mutation (NFκB BSmut). Mutated nucleotides are shown in gray. (F) Relative promoter activity assay in MDA-MB-435 and Hs578T cells in 3D culture, transduced with luciferase reporter plasmid containing −333/+147 bp FSCN1 promoter with NFκB BSwt or NFκB BSmut sequence. (G) Relative promoter activity assay in MDA-MB-435 cells in 3D culture, transduced with luciferase reporter plasmid containing −333/+147 bp FSCN1 promoter with NFκB BSwt. Cells were treated with either NFκB control (C) peptide or NFκB inhibitor peptide (I). (H) MDA-MB-435 (left) and Hs578T (right) cells transduced with either shScr or shSPHK1.1 in 3D culture system. Chromatin immunoprecipitation (ChIP) was performed with antibodies against IgG (negative-control), NFκB and H3K4me3 (positive-control). Binding to the FSCN1 promoter region was quantified by qPCR from immunoprecipitated DNA, and fold-enrichment was calculated relative to IgG. Data are representative of at least three independent experiments and are represented as mean ± SD. *, p < 0.05, **, p < 0.01 and ***, p < 0.001, n.s., non-significant.

To investigate whether SPHK1 regulates FSCN1 mRNA transcription, we cloned the −1376 to +147 base pair (bp) FSCN1 promoter region into the pGL3 basic vector expressing a luciferase reporter gene, and transfected into either 435.shScr or 435.shSPHK1.1 cells. FSCN1 promoter-driven luciferase activities were reduced by SPHK1 knockdown in 3D cultured, but not in 2D cultured 435.shSPHK1.1 cells (Fig.5C), indicating SPHK1 is essential for FSCN1 mRNA transcription. Subsequently, all studies on SPHK1-induced FSCN1 transcription were performed in the 3D culture system.

To determine the FSCN1 promoter region responsible for SPHK1-mediated transcriptional upregulation, we made a series of 5’ deletions in the −1376/+147 bp FSCN1 promoter construct in the pGL3 basic vector and generated −333/+147 bp and −93/+147 bp constructs (Supplementary Fig.S8D). We transfected these deletion constructs of FSCN1 promoter into either 435.shScr or 435.shSPHK1.1 cells and compared their activities in driving luciferase reporter gene expression. The −333/+147 bp FSCN1 promoter region was sufficient to induce transcription in SPHK1 high expressing cells, whereas both basal transcription and SPHK1-induced transcription were significantly inhibited when FSCN1 promoter was deleted to −93/+147 bp region (Fig.5D). The data indicate that the −333/−93 bp region in FSCN1 promoter is important for basal and SPHK1-mediated FSCN1 transcription. Using online software (PROMO), we identified that the −333bp to −93 bp region of FSCN1 promoter harbors binding sites of various transcription factors, one of which was NFκB (Fig.5E), that can be activated by SPHK1 (30). Indeed, NFκB pathway was activated more in SPHK1-overexpressing 231SPHK1 tumor lysates and 435.shScr cells growing in 3D culture compared to respective controls (Supplementary Fig.S8E-S8F). Intracellular S1P, downstream of SPHK1, was known to act as a cofactor of TRAF2 leading to IKKα/β activation resulting in IκBα phosphorylation and degradation, consequently, NF-kB activation upon TNFα stimulation (31). Therefore, we examined the phosphorylation status of IKKα/β and IĸBα in our SPHK1-modulated MDA-MB-435 cells in both 2D and 3D cultures. Compared to SPHK1-overexpressing 435.shScr control cells, phosphorylation of IKKα/β and IĸBα are decreased in SPHK1 knocked down 435.shSPHK1 cells in 3D culture, not in 2D culture, confirming that SPHK1/S1P regulates NF-kB activation via IKKα/β/IĸBα pathway (Supplementary Fig.S8G). Additionally, gene set enrichment analysis (GSEA) of our cDNA microarray data revealed that tumor necrosis factor (TNF) signaling was upregulated in mfp tumors of SPHK1-high expressing 435.shScr cells relative to the SPHK1-low expressing 435.shSPHK1.1 mfp tumors (Supplementary Fig.S8H) (32). Thus, SPHK1 may induce FSCN1 upregulation via activation of the NFκB transcription factor.

To investigate whether the NFκB binding site in the FSCN1 promoter is critical for SPHK1-induced FSCN1 upregulation, we generated NFκB binding site mutant (NFκB BSmut) from the −333/+147 bp wild type FSCN1 promoter-driven luciferase reporter construct (NFκB BSwt) (Fig.5E and Supplementary Fig.S8I). The NFκB BSwt and the NFκB BSmut constructs were transfected into SPHK1 high-expressing (MDA-MB-435 and Hs578T) cells and compared for promoter activities. Compared to NFκB BSwt construct, NFκB BSmut construct with mutation of the NFκB binding site had significantly reduced FSCN1 promoter activity in both cell models (Fig.5F). Additionally, a cell permeable peptide that inhibits translocation of the NFkB active complex into the nucleus significantly inhibited NFκB BSwt FSCN1 promoter activities compared to the control peptide in MDA-MB-435 cells (Fig.5G). These data indicate that nuclear NFκB binding to the NFκB binding site of FSCN1 promoter is critical for FSCN1 transcriptional upregulation in SPHK1 high-expressing TNBC cells. Next, we examined whether NFκB transcription factor binding to the NFκB binding region (−263/−253 bp) of FSCN1 promoter are higher in SPHK1 high-expressing TNBC cells than that in SPHK1 low-expressing cells. We performed a ChIP assay using NFκB antibody to bring down NFκB by immunoprecipitation (IP) and followed by qPCR with primers flanking NFκB binding region (−263/−253 bp) of FSCN1 promoter. Clearly, the binding of NFκB is more enriched in the −263/−253 bp region of FSCN1 promoter in SPHK1 high-expressing (435.shScr and Hs578T.shScr) cells than in their corresponding SPHK1 knockdown cells, whereas the binding of tri-methylated H3K4 showed no significant difference (Fig.5H). Together, these data indicate that SPHK1 upregulates FSCN1 gene expression at the transcriptional level via activation of NFκB transcription factor.

SPHK1/pNFκB/FSCN1 expressions in patients’ TNBC tissues correlate with poor survival.

To determine the clinically relevance of our above findings, we examined whether SPHK1/NFκB/FSCN1 pathway activation in patients’ TNBC tissues is associated with increased metastasis and poor clinical outcome. We performed immunohistochemistry (IHC) analyses of SPHK1, p-NFκB and FSCN1 expressions in tissue microarrays (TMAs) of patients’ TNBC tissue samples. Patients (N=117) whose TNBC tissues are included in the TMAs have various clinical and pathological features (Supplementary Table S1). IHC analyses revealed high expression of SPHK1, pNFκB and FSCN1 in 59.2% (68/115), 67% (71/106) and 62.7 % (69/110) of TNBC tissue samples, respectively (Figs.6A-6B). To examine whether SPHK1/pNFκB/FSCN1 signaling pathway activation is critical for TNBC progression and metastasis in patients, we examined the relationships of high expressions of SPHK1, pNFκB and FSCN1 with TNBC progression and metastasis. Since IHC staining of phospho-proteins in archived tissues of patients may be unreliable, we analyzed whether co-expressions of SPHK1 and FSCN1 may correlates with increased metastasis and poor clinical outcome. Patients were divided into two groups: 1) low expression of one or both markers (i.e. SPHK1 and/or FSCN1 low), and 2) high expression of both markers (i.e. SPHK1 and FSCN1 high) in their TNBCs. Compared to group-1 patients, high expressions of SPHK1 and FSCN1 in group-2 patients are significantly associated with poorer distance metastasis-free survival (DMFS) (p=0.045) as well as worse overall survival (OS) (p=0.003) (Fig.6C-6D). Thus, clinically, high expression of SPHK1 and FCSN1 in TNBC tissues correlate with increased distant metastasis and poor survival in TNBC patients.

Figure 6: Combined Expression of SPHK1, pNFκB and FSCN1 in TNBC patients correlates with poor survival.

(A) Representative images of immunohistochemical staining showing the high and low expression of SPHK1, pNFκB and FSCN1. Figure objective magnification: 10X and inset objective magnification: 20X (B) Graph showing the percentage of TNBC patients with high and low expression of SPHK1, pNFκB and FSCN1. (C-D) Kaplan–Meier survival plot showing distance metastasis free survival (C) and overall survival (D) among TNBC patients classified into two different groups based on the combined expression of SPHK1 and FSCN1. Group 1: patients with low expression of one or both markers (SPHK1 and/or FSCN1 low, n=66); and Group 2: patients with high expression of both markers (SPHK1 and FSCN1 high, n=42).

Targeting SPHK1 and the NFκB axis impedes tumor progression and spontaneous lung metastasis.

The above data from TNBC animal models and TNBC patients’ tissues prompted us to test whether therapeutically targeting SPHK1/pNFκB/FSCN1 axis could inhibit TNBC progression and metastasis in an aggressive TNBC animal model. Since SPHK1 regulates FSCN1 via the NFκB transcription factor, and clinically applicable inhibitors of both SPHK1 and NFκB are available, we tested the efficacy of SPHK1 and NFκB inhibitors, either as a single agent or in combination, for deterring TNBC progression and metastasis. SPHK1 were targeted by safingol, a SPHK1 inhibitor tested in multiple clinical trials (, ), and NFκB were targeted with bortezomib, an FDA-approved proteasome inhibitor drug for multiple myeloma, but shown to inhibit NFκB activity (33). Safingol treatment decreased SPHK1 protein expression in both human MDA-MB-435 cells and 4T1 mouse TNBC cells in vitro (Supplementary Fig.S9A-S9B). Similarly, bortezomib treatment in 4T1 cells decreased phospho-NFκB expression in 2D and, more prominently, in 3D culture (Supplementary Fig.S9C).

To test whether safingol alone, bortezomib alone, or their combination could deter TNBC progression and metastasis, we orthotopically injected highly aggressive 4T1 mouse TNBC cells (50,000 cells/mice) into BALB/c mice to induce mfp tumors. On day 7, when the primary tumors were palpable, we randomized the mice into four groups and treated them by intraperitoneal (i.p.) injection of one of the following for 3 weeks: (a) vehicle (n=10), (b) safingol (5 mg/kg, every 3 days, n=11), (c) bortezomib (0.5 mg/kg, every 3 days, n=11), or (d) safingol plus bortezomib (n=15). Tumor growth and body weight of mice were monitored every three days. The single treatment of either safingol or bortezomib had no significant effects on primary tumor growth, whereas the combination treatment significantly reduced primary tumor growth (Fig.7A and Supplementary Fig.S9D). Interestingly, safingol only, or bortezomib only significantly prolonged overall survival of mice compared with vehicle treatment (median survival of 37 days, or 40 days, versus 31 days, respectively) (Fig.7B). Strikingly, safingol plus bortezomib combination treatment dramatically increased the median survival of mice from 31 days to 47 days (Fig.7B). There was no significant difference in mouse body weight, and no dramatic difference in their blood counts of lymphocytes and myeloid cells among four treatment groups, suggesting that drug treatment, either as a single agent or in combination, did not induce any acute toxicity (Supplementary Fig.S9E-S9F).

Figure 7: Targeting SPHK1 and NFκB signaling pathway delayed tumor progression and reduced spontaneous metastasis to lungs.

(A) MFP tumor growth curve in BALB/c mice with orthotopic injection of 4T1 cells. At day 7, when tumor reached the palpable size, mice were randomized into 4 groups for treatment: vehicle, safingol only, bortezomib only and combination of both drugs (saf + bor). (B) Survival curve of the mice on each treatment groups as mentioned in (A). (C) MFP tumor weight at day 28, from mice on each treatment groups as mentioned in (A). At day 28, five mice were euthanized from each group as mentioned in (A) for time matched experiment and tumor weight were taken. (D - H) Quantification of immunohistochemical staining of SPHK1 (D), nuclear pNFκB (E), FSCN1 (F), Ki67 (G) and TUNEL (H) on mfp tumor samples of mice from time matched experiment as mentioned in (C). (I) H&E stained lung sections were quantified for the number of spontaneous metastatic lesions from of BALB/c mice euthanized for time matched experiment as mentioned in (c). Data are representative as mean ± SD. *, p < 0.05, **, p < 0.01 and ***, p < 0.001, n.s., non-significant.

To evaluate the inhibitory effects of safingol, bortezomib, or their combination on tumor growth and metastasis, five mice under each of the four treatments were euthanized at day 21 post treatment and their mfp tumors as well as lungs were harvested for examination. Consistent with tumor volume detected above in mice, primary tumor weights were significantly reduced by combination treatment, but not the single-treatments (Fig.7C). IHC analyses of SPHK1 and pNFκB in these primary tumor samples showed that safingol treatment decreased the SPHK1 expression and bortezomib treatment decreased the nuclear pNFκB compared to vehicle treatment, suggesting that the drugs were effectively inhibiting their molecular targets (Fig.7D-7E and Supplementary Fig.S9G). Safingol treatment also decreased its downstream nuclear pNFκB compared to vehicle treatment (Fig.7E). Either safingol or bortezomib treatment reduced FSCN1 expression (Fig.7F and Supplementary Fig.S9G). Remarkably, combination treatment significantly decreased SPHK1 expression, nuclear pNFκB level, and FSCN1 expression compared to vehicle or single treatments (Fig.7D-7F and Supplementary Fig.S9G). Furthermore, combination treatment decreased tumor cell proliferation detected by Ki67 staining and increased tumor cell apoptosis as shown by TUNEL staining, respectively (Fig.7G-7H and Supplementary Fig.S9G). More importantly, safingol alone and bortezomib alone significantly reduced spontaneous metastasis to the lungs compared to vehicle treatment (Fig.7I and Supplementary Fig.S9H). Combination treatment further reduced lung metastasis significantly compared to each single treatment, inducing ~80% inhibition of spontaneous lung metastasis in mice compared to vehicle treatment (Fig.7I and Supplementary Fig.S9H).

To evaluate whether or not the inhibition of metastasis by combination treatment resulted from reduced primary tumor growth, we examine the effect on metastasis after primary tumors were removed at similar sizes in vehicle and combination treatment groups. (Supplementary Fig.S10A-S10B). . The lungs of these mice were harvested for examination of metastatic lesions 11 days post primary tumor resection in both vehicle and combination treatment groups. Combination treatment significantly reduced spontaneous metastasis to the lungs compared to vehicle treatment when the primary mfp tumors were removed at similar sizes, suggesting that inhibition of metastasis by combination treatment is not only the consequences of reduced primary tumor growth (Supplementary Fig.S10C).

Discussion

TNBCs are negative of therapeutic targets and do not respond to current hormonal and HER2-targeted therapies (4). In this study, we embark on identifying positive therapeutic target(s) in TNBCs to develop effective targeted therapies for TNBC, especially TNBC metastasis. Our integrative bioinformatics analysis of breast cancer patient-derived gene expression datasets revealed that expression of SPHK1 is significantly higher in TNBCs and is associated with poor clinical outcome of TNBC patients. Further, we found that SPHK1 plays a prominent role in promoting TNBC spontaneous metastasis in multiple mouse models. Mechanistically, SPHK1 upregulates the expression of FSCN1 at the transcriptional level via activation of NFκB. Clinically, high expression of SPHK1 and FCSN1 in patients’ TNBC tissues correlates with increased distant metastasis and poor survival. Furthermore, we demonstrated that targeting SPHK1 and NFκB by clinically-applicable inhibitors effectively inhibited tumor progression and spontaneous metastasis to the lungs in a highly aggressive TNBC mouse model. Together, our data indicate that SPHK1 and NFκB could serve as positive therapeutic targets for inhibiting TNBC metastasis.

We demonstrated that SPHK1 expression was higher in TNBCs compared to other subtypes of breast cancer. Abnormal expression of SPHK1 has been strongly associated with the development and progression of various cancers (9, 14). Several studies of breast cancer and other cancer types have also shown that SPHK1 plays a role in cancer cell proliferation (34-36). In contrast, our data showed that SPHK1 cannot drive TNBC cell proliferation in vitro nor tumor growth in vivo in multiple TNBC models and in both gain- and loss-of-function studies. Consistent with our findings, recent studies of SPHK1-specific inhibitor, PF-543, showed that inhibiting SPHK1 had no effect on the proliferation of various cancer cells, including a TNBC cell line MDA-MB-231 (37, 38). Notably, although SPHK1 knockdown had no significant impact on cell proliferation, MFP tumors of SPHK1 knockdown cells had increased apoptosis compared to MFP tumors of control cells, similar to a previous report (39). Most importantly, our data from both gain- and loss-of-function studies showed that SPHK1 functions to promote metastasis-related properties of TNBC cells in vitro and TNBC spontaneous metastasis in vivo. Furthermore, we dissected the FSCN1 upregulation as a critical molecular mechanism of SPHK1-induced TNBC cell invasion and metastasis.

In our study, we used spontaneous metastasis models to determine whether SPHK1 functions in metastasis, since the spontaneous metastasis model recapitulates all of the steps involved in the multistep process of the metastatic cascade (40). Exogenous expression of SPHK1 significantly increased, whereas knocking down SPHK1 significantly decreased spontaneous metastasis to the lungs in nude mice, suggesting SPHK1 expression is vital for the metastatic potential of TNBC cells. Despite some reports suggesting that the SPHK1/S1P axis modulates the MAPK/ERK pathway and EGFR pathways to promote invasion and metastasis in some cancer types, the mechanism by which SPHK1 promotes TNBC metastasis was unclear (41, 42). Understating molecular pathways driving TNBC metastasis process, with fast-track clinical translation potentials, is needed for developing effective therapeutic intervention of TNBC metastasis. In this study, using unbiased gene expression profiling of in vivo tumor samples, we identified FSCN1 as a top candidate gene modulated by SPHK1 and involved in TNBC cell migration, invasion, and metastasis. FSCN1 is a cytoskeletal actin bundling protein that binds and packages actin filaments into tertiary structures to enhance cell motility, migration, and adhesion (24, 25, 43). FSCN1 is overexpressed in various cancers (44-46), whereas its expression is either absent or very low in normal epithelial cells (43). Tumor cells with high expression of FSCN1 have increased cell membrane protrusions such as filopodia and invadopodia, which help tumor cells’ migration and extracellular matrix invasion, critical steps for metastasis development (25, 27, 43, 46). Although it is known that FSCN1 high-expression in cancer cells facilitates metastasis, the upstream regulators of FSCN1 were not well known. Here, we identified that SPHK1 high expression in TNBC cells increases FSCN1 mRNA transcription by activating NFκB. Additionally, we demonstrated that FSCN1 is a novel downstream effector of SPHK1 in promoting TNBC cell migration, invasion, and metastasis.

Previous studies have shown that the SPHK1/S1P signaling can induce activation or inhibition of various transcription factors, including NFκB, E2F, c-Myc, and Sp1, and consequently impact on cell proliferation, apoptosis, and/or inflammation (36, 47, 48). SPHK1 was reported to induce NFkB activation via intracellular S1P that serves as a cofactor of TRAF2 to activate IKKα/β, then IKKα/β phosphorylates IĸBα resulting in its degradation to allow NFkB activation upon TNFα stimulation (31). Consistently, in 3D-cultured 435.shSPHK1 cells, both IKKα/β and IĸBα phosphorylations were decreased compared to SPHK1-highexpressing 435.shScr cells. Thus, SPHK1-induced NFkB activation is mostly mediated by intracellular S1P function, although it may also involve S1P’s extracellular function.

Here, we made the unique finding that SPHK1 activates NFκB to upregulate FSCN1 expression. Our data showed that FSCN1 promoter activity was significantly decreased in SPHK1 knockdown cells, indicating SPHK1 upregulates FSCN1 at the transcriptional level. We identified a 240-bp region within FSCN1 promoter and a NFκB binding site herein that is responsible for this SPHK1-induced FSCN1 upregulation. We also showed that binding of NFκB to the FSCN1 promoter region was inhibited in SPHK1 knockdown cells compared with control cells (Fig. 5H). Interestingly, SPHK1 activated NFκB more effectively in 3D than 2D cultures (Supplementary Fig. S8F and S9C) and we readily detected SPHK1 upregulation of FSCN1 expression in 3D culture. Together, SPHK1 transcriptionally upregulates FSCN1 expression via activation of NFκB.

The SPHK1/S1P axis is known to regulate immune responses by regulating lymphocytes trafficking, innate immune response, and inflammation (49). For these reasons, we used a syngeneic mouse model for testing therapeutic efficacy of SPHK1 targeting in vivo. Also we decided to use drugs that are already in clinical trials or have been FDA-approved for fast track clinical translation. Specifically, we targeted SPHK1 by safingol, which is currently used in clinical trials and target NFκB by bortezomib, which is a FDA-approved drug against multiple myeloma. Although a few drugs have been developed against FSCN1, none have been tested so far in a clinical setting (50). Even tough, single agent treatment of safingol or bortezomib had no significant effect on primary tumor growth, combination treatment significantly delayed tumor progression (Fig. 7A), suggesting a complex interaction between these two drugs that warrants further investigation. Excitingly, treatment with single agent of either safingol or bortezomib significantly decreased the spontaneous metastasis of aggressive TNBC to the lungs, which was further decreased with combination treatment and the safingol plus bortezomib also significantly increased mice survival (Fig. 7B). Our data indicate that combining safingol and bortezomib may be effective in treating TNBC tumors and metastasis, which warrants further clinical testing. Additionally, understanding of the impact of these drugs on the tumor microenvironment in vivo will enable us to further improve these drugs’ efficacy in the clinic. It would also be interesting to test the benefits of using these drug combination as an adjuvant treatment regimen after surgery to prevent or intervene with metastasis and/or recurrences, especially for those patients whose primary tumor express high levels of SPHK1 and/or FSCN1. Our study also suggests that expression of SPHK1/pNFκB/FSCN1 axis in TNBC primary tumors could be a predictive biomarker for development of metastasis, which needs to be further validated. Taken together, we have shown that the SPHK1/pNFκB/FSCN1 axis is activated in TNBCs and can serve as positive therapeutic targets that can be inhibited by clinically applicable kinases inhibitors, and our preclinical testing demonstrated that targeting SPHK1/pNFκB effectively inhibited metastasis from highly aggressive TNBC tumors. We foresee that these findings may be speedily translated into the clinic to benefit TNBC patients in great need of effective therapies.

Significance: Sphingosine kinase 1 (SPHK1) is overexpressed in TNBC and promotes metastasis, targeting SPHK1 or its downstream target NFκB with clinically available inhibitors could be effective for inhibiting TNBC metastasis.