Regulation of hypoxia-inducible factor functions in the nucleus by sphingosine-1-phosphate

Abstract

Sphingosine kinase 2 (SphK2) is known to phosphorylate the nuclear sphingolipid metabolite to generate sphingosine-1-phosphate (S1P). Nuclear S1P is involved in epigenetic regulation of gene expression; however, the underlying mechanisms are not well understood. In this work, we have identified the role of nuclear S1P and SphK2 in regulating hypoxia-responsive master transcription factors HIF-1α/2α, and their functions in breast cancer, with a focus on triple-negative breast cancer (TNBC). We have shown SphK2 is associated with HIF-1α in protein complexes, and is enriched at the promoters of HIF target genes, including vascular endothelial growth factor VEGF, where it enhances local histone H3 acetylation and transcription. S1P specifically binds to the PAS domains of HIF-1α. SphK2 and HIF-1α expression levels are elevated in metastatic estrogen receptor-positive (ER+) and TNBC clinical tissue specimens compared to healthy breast tissue samples. To determine if S1P formation in the nucleus by SphK2 is a key regulator of HIF functions, we found using a preclinical TNBC xenograft mouse model, and an existing selective SphK2 inhibitor K-145, that nuclear S1P, histone acetylation, HIF-1α expression, and TNBC tumor growth were all reduced in vivo. Our results suggest that S1P and SphK2 in the nucleus are linked to the regulation of HIF-1α/2α functions associated with breast cancer progression, and may provide potential therapeutic targets.

INTRODUCTION

Breast cancer often manifests as solid hypoxic tumors which are associated with high levels of hypoxia-inducible factor (HIF)-α and a poor prognosis (1–4). The HIF protein consists of an oxygen-sensitive α-subunit and a constitutively expressed β-subunit that regulate the expression of target genes (5, 6). HIF-1α and HIF-2α are the best-characterized α subunits of HIFs. HIF-1α is ubiquitously expressed, while, HIF-2α is mainly detected in highly vascularized organs with hypoxic tissues, including kidney epithelial cells (7). Despite their extensive sequence similarity and co-expression in various cancer cell types, HIF-1α and HIF-2α have overlapping and non-overlapping roles in gene expression associated with tumor progression (8). Under hypoxia, HIFα subunits escape Von Hippel-Lindau (VHL)-mediated ubiquitin-proteasome (26S) pathway-mediated degradation and heterodimerize in the nucleus with HIF-1β for transactivation activity (5). Despite efforts in the past two decades, there are no specific inhibitors directly inhibiting HIFs heterodimerization that have been approved in the clinic for cancer treatment. Although CBP-p300 transcriptional coactivator proteins (9) and histone deacetylases (HDACs) (10, 11) have been identified as important regulators of HIF functions in response to hypoxia, regulatory mechanisms governing HIF transcription and cofactors mediating functional activity of HIF-1α/2α are not fully known.

Sphingosine-1-phosphate (S1P) is a bioactive sphingolipid mediator that regulates many cellular functions, including cancer cell survival and cancer progression (12–14). S1P is produced intracellularly by two closely related sphingosine kinases, SphK1 and SphK2. While we have learned much about the functions of cytosolic SphK1 in cancer progression, the functions of SphK2 in cancer remain vastly unknown (14–16). Because we previously described nuclear S1P produced from the nuclear-localized SphK2 that is involved in histone acetylation via inhibition of HDAC1, HDAC2 and epigenetic gene regulation (17–19), we sought to determine nuclear S1Ps direct role regulating HIF functions in breast cancer with a focus on TNBC. We found that nuclear S1P markedly enhanced proximal promoter histone acetylation and regulates the production of HIF-1α and HIF-2α transcripts in TNBC cells. S1P as an endogenous cofactor directly binds to the PAS domains of HIF-α and is involved in the activation of these transcription factors in response to hypoxia. SphK2 is associated with HIF-1α protein complexes, including p300, and is enriched at the HIF-1α direct target promoters, including HIF-2α or VEGF, for transcriptional activity. Nuclear S1P mechanisms in response to hypoxia-regulated transcription of HIF-α-direct target genes are associated with cancer stem cells, angiogenesis, and metastasis. SphK2 and HIF-1α expression levels are elevated in metastatic ER+, as well as in TNBC clinical tissue specimens compared to healthy breast tissues. Thus, to provide a proof of principle determination that S1P formation by SphK2 is a key regulator of HIF functions, in a TNBC xenograft mouse model, we used an existing selective SphK2 inhibitor, the K-145 compound, that reduced nuclear S1P, histone acetylation, and HIF-1α expression in tumor tissue, and also reduced human TNBC tumor growth in mice. Our data suggested that nuclear S1P is a direct regulator of HIF functions. Breast cancer subsets with elevated SphK2 and HIF-1α expression levels would be potential candidates for SphK2 targeted therapy.

MATERIALS AND METHODS

Cell culture and transfection

Human ER+ breast adenocarcinoma (MCF-7) cells, TNBC cell lines MDA-MB-231 and BT-549, and human cervical carcinoma (HeLa) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA), and cultured as described previously (17, 20, 21). Human LM2–4 cells, obtained from Dr. John Ebos, Roswell Park Comprehensive Cancer Center (RPCCC), are a metastatic variant of the triple-negative MDA-MB-231 breast cancer cell line derived after multiple rounds of in vivo lung metastasis selection in mice, and were grown as previously described (22). LM2–4 and MDA-MB-231 cells were maintained in phenol-red free RPMI (Gibco BRL, Grand Island, NY, USA) media containing 10% heat-inactivated Fetal Bovine Serum (FBS), 100 μg/ml penicillin, and 100 μg/ml streptomycin (Gibco) as described previously (21). Low passage cells (<10 passages) were used for all experiments. For hypoxia studies, cells were cultured under hypoxia at concentrations of 1% O2, 5% CO2, and 94.5 % N2 at 37°C using the Xvivo System (Biospherix). SphK1 and SphK2 were downregulated by transfection with On-TargetPlus SMARTpool small interfering RNA (siRNA) or control scrambled siRNA from Dharmacon (Lafayette, CO, USA), as described previously (17, 21). In some experiments, to confirm the lack of off-target effects, cells were transfected with siRNA against human SphK2 (5’-GGAUUGCGCUCGUCGCUUUCAU-3’) and control siRNA from Qiagen (Germantown, MD, USA), as described previously (17). Of note, knockdown of SphK1 and SphK2 genes were confirmed by qRT-PCR and Western blot analysis, as demonstrated before with the anti-SphK1 and anti-SphK2 antibodies (20, 21). In some experiments, as described in Figure legends, cells were transfected with vector, V5-SphK2, and catalytically-inactive (Ci) SphK2^G212E (in which glycine 212 replaced by glutamic acid) plasmids, as described previously (17, 18, 23).

Nuclear extracts

Nuclear extracts from cell pellets and tissues were isolated with a low-salt extraction procedure as described previously (17, 20, 23, 24). In brief, phosphate buffered saline (PBS) washed cells were re-suspended in low-salt buffer A containing 10 mM Hepes (pH 7.8), 10 mM KCl, 0.1 mM EDTA, 1 mM Na3VO4, 1 mM DTT, 0.2 mM PMSF, and 1:500 protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA). Human breast tissues and patient-derived xenograft (PDX) tissues were grounded into a fine powder in liquid nitrogen without thawing and suspended in ice-cold buffer A. Cells or tissue powder suspensions in buffer A were incubated for 15 min on ice, followed by vortexing for 10 seconds at room temperature (RT) with NP-40 (final concentration 0.75%), and then incubated on ice for 2 min. Crude nuclei pellets and nuclei-free cytoplasmic fractions were separated by centrifugation at 3000 RPM for 3 min at 4°C. Nuclear pellets were washed twice with two volumes of buffer A. Nuclei were then dissolved in high salt buffer containing 20 mM Hepes (pH 7.8), 0.4 M NaCl, 1 mM EDTA, 1 mM Na3VO4, 1 mM DTT, 0.2 mM PMSF, and 1:500 protease inhibitors cocktail and incubated on ice for 15 min. Nuclear extracts were isolated by centrifugation at 14,000 x g for 5 min at 4 ° C.

Immunoblotting and Immunoprecipitation

Immunoblotting was performed as described previously (17, 20, 23). Equal amounts of proteins were resolved by SDS-PAGE and transferred to nitrocellulose membranes with a Bio-Rad Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad, Hercules, California, USA). Membranes were blocked with 5% non-fat dry milk (Bio-Rad, # 170-G404) in Tris-buffered-saline (TBS) containing 0.1% Tween 20 for 1 hour at room temperature. The following primary antibodies were used for immunoblot analysis. Rabbit polyclonal antibodies to HIF-1α (#36169), HIF-2α (#7096), Histone H3 (#4499), Histone H4 (#14149), Lamin A/C (#2032), Acetylated-Lysine (#9441), His-tag (#2366), HDAC1 (#34589) from Cell Signaling Technology (Danvers, MA, USA); human SphK2 (#S3698) (21), human SphK1 (21) (#HPA022829); V5 (#V8137), β-actin (#A5316) from Sigma-Aldrich (St. Louis, MO, USA); H3-K9ac (# ab4441) (17–19, 23–26) from Abcam (Cambridge, MA, USA). Lamin B1 (#sc-374015), α-tubulin (#sc-8035), p300 (#sc-48343) from Santa Cruz Biotechnology (Dallas, Texas, USA). Immunopositive bands were visualized by incubating the blots with HRP-conjugated secondary antibodies (1:5000 dilution; Jackson ImmunoResearch, West Grove, PA, USA) for 1 h at room temperature before adding SuperSignal West Pico (Thermo Fisher Scientific) chemiluminescence substrate (17, 23). Representative blots from at least 3 independent experiments are shown.

Nuclear extracts isolated from naïve cells or cells transfected with vector or V5-SphK2, or V5-Ci-SphK2^G212E were immunoprecipitated with antibodies against HIF-1α or HIF-2α or V5, or control normal IgG, as indicated in Figure legends. Immunoprecipitates were captured with protein A/G-Sepharose beads (Santa Cruz). After washing the beads four times with lysis buffer, bound proteins were released by boiling in SDS-PAGE sample buffer and analyzed by Western blotting.

Pulldowns with S1P affinity matrices

Control or S1P-coated agarose beads (Echelon Biosciences, Salt Lake City, UT, USA) equilibrated with binding buffer containing 10 mM HEPES (pH 7.8), 150 mM NaCl, 0.5% NP-40 and 1:500 protease inhibitor cocktail were mixed with nuclear extracts preincubated with or without 10 µM S1P for 25 min and rocked overnight at 4 °C (17, 27, 28). For some experiments, recombinant wildtype-HIF-1α (# ab154478, >75 % pure, Abcam) and His-tagged deletion mutant HIF-1α with the PAS domains deleted (# 11977-H07E, delta M1-L574, >95 % pure, Sino Biological US Inc., Wayne, PA, USA) proteins pretreated with or without 10 µM S1P for 25 min were mixed with control or S1P affinity beads and incubated for overnight at 4 ° C. Beads were washed 4 times with 10 volumes of binding buffer, and protein-bound S1P beads and control beads were collected by centrifugation at 1000 RPM at 4 ° C. After removing the supernatant, the beads were boiled in SDS-PAGE sample buffer and bound proteins analyzed by Western blotting (17, 27, 28), as explained in the Figure legend.

[³²P]S1P binding assay

Recombinant wildtype-HIF-1α and His-tagged PAS-delta-HIF-1α were incubated with or without unlabeled lipids in the presence of [³²P]S1P (0.1 nM, 6.8 µCi/pmol) prepared as described previously (17, 29) in buffer containing 50 mM Tris-HCl (pH 7.5), 137 mM NaCl, 1 mM MgCl₂, 2.7 mM KCl, 15 mM NaF, and 0.5 mM NaV3O4 for 25 min at 30 °C. Proteins were then immunoprecipitated with antibodies against HIF-1α and His-tag for HIF-1α and PAS-delta-HIF-1α, respectively, or with control IgG, washed 4 times with the binding buffer, and radioactivity in the immunocomplexes was determined by scintillation counting (17, 27).

Quantitation of sphingolipids by LC-ESI-MS/MS

Sphingolipids were measured (17, 27) by liquid chromatography, electrospray ionization–tandem mass spectrometry (LC-ESI/MS/MS, 4000 QTRAP, AB Sciex, Framingham, MA, USA). Briefly, for some experiments, nuclei were isolated from the MDA-MB-231 xenograft tumor tissues, as described above. Internal standards were added (0.5 nmol each, Sphingolipid Mixture II/LM-6005, Avanti Polar Lipids), lipids extracted, and sphingolipids quantified by LC-ESI/MS/MS. For some experiments, nuclear extracts from MDA-MB-231 cells were immunoprecipitated with anti-HIF-1α and anti-HIF-2α antibodies, or with control IgG, as described above, and sphingolipids determined by LC-ESI-MS/MS (17, 27).

Quantitative real-time PCR

Quantitative real-time PCR (qRT-PCR) analysis was performed as described before (17, 20, 21). Briefly, total RNA was prepared with Trizol (Life Technologies, Carlsbad, CA, USA). RNA (2 μg) was reverse transcribed with the high-capacity cDNA Archive kit (Life Technologies). cDNAs were diluted 10-fold (target genes) and 100 fold (house-keeping genes, cyclophilin A or GAPDH, or RPLO, Large ribosomal protein), amplified with SYBR Green quantitative PCR on CFX96 cycler (Bio-Rad Laboratories). Gene expression levels were calculated by the 2^−(ΔΔCT) method (30, 31), and normalized to Cyclophilin A, GAPDH, or RPLO expression, as explained in the Figure legend. The sequences of the primers are shown in Supplemental Table S1.

Reporter gene assay

HeLa cells plated in 24-well plates were transfected with siSphK2 or siControl, as mentioned above. After 24 hours of siRNA transfection, cells were transfected for another 24 h using FuGENE HD (Promega Corporation, Madison, WI, USA) with 150 ng of human HIF-1α luciferase reporter vector and 50 ng of Renilla luciferase plasmid DNA, as an internal control. Human HIF-1α promoter plasmid (pGL2B-HIF-1α) was a gift from Alex Minella (Addgene plasmid # 40172). Briefly (32), a luciferase reporter plasmid containing the human HIF-1α promoter was made by amplifying nucleotides −2180 to +28 (relative to the transcriptional start position) from genomic DNA and subcloned into pGL2b vector (Promega). Cells were then lysed, and luciferase activities were measured using the Dual-Luciferase Assay Kit (Promega) following the manufacturer’s recommendations. Results were quantified with a Synergy HTX luminescence counter (BioTek Instruments, Winooski, VT, USA), and normalized values were expressed as the fold induction over control cells.

Chromatin immunoprecipitation (ChIP)

ChIP assays were performed as described previously (17–19). Briefly, 2 x 10⁷ MDA-MB-231 or MCF-7 cells were cross-linked with 1 % formaldehyde for 10 min at 37 °C before quenching with glycine (0.125 M). Cells were washed with ice-cold PBS and suspended in the lysis buffer (50 mM Tris-HCl pH 8.0, 10 mM EDTA, 1% SDS), sonicated 9 times with 10-second pulses on ice with 2 min intervals, and then centrifuged. Sonicated lysate supernatants were pre-cleared for 1 h at 4 °C in a 50% slurry of protein G-Sepharose 4 Fast Flow beads (GE Healthcare, Chicago, IL, USA) previously blocked with 25 μg/ml sonicated herring sperm DNA (Promega) in IP buffer containing16.7 mM Tris (pH 8.0), 16.7 mM NaCl, 1.2 mM EDTA, 0.01 % SDS, 1.1 % Triton X-100, and 0.05 mg/ml BSA. Soluble chromatin was immunoprecipitated with specific antibodies against HIF1-α, H3-K9ac, V5, or H4 total, or control normal IgG, as indicated in Figure legends. DNA-protein immunocomplexes captured by protein G-Sepharose beads were washed sequentially with low salt buffer (20 mM, Tris (pH 8.1), 2 mM EDTA, 150 mM NaCl, 1% Triton, 0.1% SDS), high salt buffer (20 mM Tris (pH 8.1), 2 mM EDTA, 500 mM NaCl, 1% Triton, 0.1% SDS), LiCl buffer (10 mM Tris (pH 8.1), 0.25 M LiCl, 1 mM EDTA, 0.5% NP40, 0.5% DOC), and TE buffer before eluting in 1% SDS in 0.1 M NaHCO3. DNA-protein cross-links were reversed by heating at 65 °C overnight in 0.3 M NaCl, then treated with proteinase K for 1 h at 55°C. 1/10th volume of input samples was also similarly treated. DNA was purified with QIAquick PCR Purification Kit (Qiagen). Inputs were eluted from the Qiagen-columns with a twice volume of TE buffer compared to the precipitated DNA samples. The input samples were diluted 1:100, and the precipitated DNA was analyzed by real-time PCR with primers amplifying the proximal promoter/enhancer sequences of the HIF-α/2α, VEGFA genes. HIF-1α (proximal) forward, 5’-CGCTAAACACAGACGAGCAC-3’ and reverse, 5’-GGGTTCCTCGAGATCCAATG-3’; HIF-2α (proximal) forward, 5’-GAGCTTTACACTCGCGAGCGGA-3’ and reverse, 5’-AGGACACTGCCGAGGATTGTAC-3’; VEGFA forward, 5’-AGACTCCACAGTGCATACGTG-3’ and reverse, 5’-AGTGTGTCCCTCTGACAATG-3’. Of note, since core histone acetylation at the distal promoter region of HIF-1α/2α (> − 8kb) is not regulated by hypoxia (33), so this region has not been examined for promoter analysis in this manuscript. ChIP-qPCR results were analyzed relative to input (unprecipitated chromatin) using the delta CT (ΔCT, normalized to the input for each sample) method. Relative binding to the promoter is expressed as fold enrichment compared with input. Input samples from each condition were serially diluted and used as a standard for the qPCR reactions to allow comparison among different primer sets. ChIP data represent the average of at least three independent experiments. For some experiments, PCR reactions were analyzed in agarose gels to visualize the PCR product after ethidium bromide staining.

Sphere formation assay

The sphere formation assay was performed as described previously (20, 34). Briefly, 20,000 cells suspended in DMEM F12 medium (Gibco/Thermo Fisher Scientific) were mixed thoroughly with low growth factor containing BD Matrigel^™ (#354234BD; Biosciences, San Jose, CA, USA) for plating (20,000 cells/well) in an ultra-low attachment 24-well plate (#29443-032; VWR, Radnor, PA, USA) in quadruplicate. The plate was maintained in a 5 % CO₂ incubator at 37 ° C for 7 days to obtain cancer cell spheroids. Spheroids were treated with the anticancer drug for another 2 days and imaged at x10 magnification using a Nikon Eclipse Ti-U Inverted Phase Contrast Microscope (Nikon Instruments Inc., New York, NY, USA).

Staining of live/dead cells

Live-dead cell assay was performed with the MCF-7 spheroids using a commercial kit (LIVE/DEAD Viability/Cytotoxicity kit; L3224; Life Technologies/Thermo Fisher Scientific), as described previously (20). Briefly, control and drug-treated spheroids stained with green‑fluorescent calcein AM as a measure of intracellular esterase activity of living cells and red‑fluorescent ethidium homodimer-1 to determine dead cells by the loss of plasma membrane integrity. Spheroids were visualized at x10 magnification using a Nikon Eclipse Ti-U Inverted Fluorescent Microscope (Nikon Instruments) and representative (N=3) viable (green), and dead cells (red) are shown as indicated in the Figure legend.

Source of breast cancer human tissues and establishment of xenografts

We obtained de-identified human breast tumor tissue clinical specimens from the Data Bank and BioRepository Shared Resources (BDDR) at RPCCC with Institutional Review Board approval (RPCCC, BDR # 074516). We also obtained healthy adjacent normal breast tissue specimens (NB1-3) from RPCCC under the approved protocol kindly provided for our study by Dr. Jianhong Chen, Department of Cancer Prevention and Control, RPCCC. Supplemental Table S2 shows the clinical characteristics (hormone receptor status, stage, tumor site, metastatic site) of breast cancer patients. Small breast tumor specimens (1–2 mm³) were established as xenografts into nine 8–10 week old female NOD SCID gamma mice (NSG mice) at either the 2nd or 4th mammary fat pad by a surgical procedure using an Institutional Animal Care and Use Committee (IACUC) approved animal protocol (IACUC #1333M). A detailed surgical protocol for the patient-derived xenograft (PDX) model was used as previously published (35), as well as that currently used by the laboratory of Dr. Kazuaki Takabe, Department of Breast Surgery, RPCCC (unpublished data). Successful engraftment was defined as generation of a palpable tumor using orthotopic implantation of tumors in mice. First passage tumors were collected by sacrificing the animals when the largest tumor-site reached 1.5 cm in diameter assessed by caliper measurements. Tumors were used for maintaining PDX lines for up to the third generation in NSG mice.

Animal studies

All animal studies were conducted in the Laboratory Animal Shared Resource Facility at RPCCC which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care. Animals were bred and maintained in a pathogen-free environment, and all procedures approved by the RPCCC IACUC with experiments performed under IACUC protocol 1333M. All mice were kept on a 12-h light/dark cycle with free access to food.

For the xenograft tumor model, TNBC MDA-MB-231 cells (1x10⁶) were surgically implanted in the upper-fat pads of NSG mice (8 to 12 weeks of age, RPCCC), as described previously (18, 36). At day 13 of implantation, tumor-bearing animals were randomized into two treatment groups (n=5/group): either vehicle alone or the SphK2 inhibitor, K-145 (Sigma-Aldrich). K-145 (30 mg/Kg) and vehicle control were administered via intraperitoneal (i.p) injection for 35 days. Every 2–3 days, tumors were measured via calipers and volume was calculated using the modified ellipsoid formula (37). Experiments were completed on day 35, mice sacrificed, tumors excised and weighed, and then processed for molecular analyses as mentioned in Figure legend 7H. After euthanasia, the presence of visceral metastasis was visually confirmed with dissection of organs.

Figure 7.

Selective inhibition of SphK2 with a pharmacological inhibitor inhibits tumor tissue nuclear bulk histone acetylation, HIFα, HIFα-target genes, and in vivo TNBC tumor growth. A) MCF-7 cells grown in low-attachment tissue culture plate for 7 days to form spheroids were treated with1 µM K-145 or vehicle for 2 days, as indicated. Light phase-contrast microscopy images were taken with 10X magnification; representative images are shown. B) Duplicate cultures were stained with LIVE/DEAD staining. For some experiments, nuclear proteins from MCF-7 spheroids treated with or without K-145 or naïve MCF-7 cells grown in tissue culture plates were used for Western blot analyses with the antibodies to H3-K9ac, HIF-1α, and histone H3, as indicated. D-G) Duplicate cultures of MCF-7 spheroids and cells were used for quantitative real-time PCR for the HIF-2α, VEGFA, and SphK2 gene levels normalized to GAPDH. Data are means ± SD, P<0.05 vs. control, by Student’s t-test. K-145 inhibited growth of MDA-MB-231 xenografts. H) MDA-MB-231 cells were surgically implanted under the fat pads of NSG mice. Mice bearing MDA-MB-231 tumors (~50–100 mm³) were separated into 2 groups randomly (N=5/group). Vehicle or K-145 (30 mg/Kg) was administered i.p for 35 days. Tumor volumes were recorded and calculated every 2–4 days until sacrifice as mm³ ± SD. *, P< 0.05 vs. vehicle, by Student’s t-test. Of note, visible breast cancer lung metastasis was not observed in our experimental setting with this xenograft model. I) Primary tumor weight was measured, data are means ± SD, P<0.05 vs. control. J) Tumor tissue nuclei were used for S1P and DH-S1P measurements using LC-MS/MS methods. K) Equal amounts of tumor tissue nuclear proteins were used for Western blot analyses with the antibodies to H3-K9ac, HIF-1α, and histone H3, as indicated. Histone H3 is used as a loading control for nuclear proteins. L-N) Tumor tissues treated with vehicle or K-145 were subjected to quantitative real-time PCR for the HIF-2α, VEGFA, and SERPINE1 gene levels normalized to GAPDH. Data are means ± SD. P<0.05 vs. vehicle.

Breast cancer patient cohorts

The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), a cohort (38) of 1904 patients, was used to investigate the association between the overall survival (OS) and the expression level of genes of interest. For each gene, patients were dichotomized into low and high groups based on gene expression. The dichotomization is defined using a running Cox proportional hazard statistics (39). Differences in the OS between the two groups were assessed at multiple candidate cutoff points, and the optimal cut point was chosen based on the statistical significance of the model. SphK2 mRNA expression data analysis (boxplot, median) was performed in the METABRIC cohort according to PAM50 breast cancer major subtypes (Basal, Her2+, Luminal A, and Luminal B). Distribution of SphK2 mRNA expression data in the human breast cancer TCGA (The Cancer Genome Atlas) cohort (40) was compared between matched breast cancer tissue and normal healthy breast tissues (N=114). SphK2 mRNA expression data were examined according to PAM50 breast cancer subtype (ER+) (N=78) vs. matched controls (N=114), as explained in Figure legends.

Statistical analysis

Statistical analysis was performed using an unpaired two-tailed Studenťs t-test for the comparison of 2 groups (using GraphPad Prism version 7.0a). For each experiment, the data set from at least triplicate samples were calculated and expressed as means ± Standard Deviation (SD) or ± Standard Error of the Mean (SEM). We used a one-way ANOVA test for datasets containing multiple group comparisons and Tukey's post hoc test for the family-wise error rate comparison. The experiments were repeated at least three times with consistent results and representative data are shown.

For patient cohort data, Chi-square and Fisher’s Exact tests and Two-tailed tests were used to investigate the PAM50 (breast cancer subtyping) data analysis. Mann–Whitney U test was used to test whether there was a difference in gene expression between normal and tumor samples. For all analyses, P values ≤ 0.05 were considered statistically significant.

RESULTS

SphK2 regulates expression of HIF-1α and HIF-2α in TNBC cells

It has been shown that cytosolic SphK1 regulates HIF-1α or HIF-2α activity in various cancer cells by S1P via extracellular signaling through S1P Receptor 1 (S1PR1) located on the membrane (41). Here we examine whether both SphK1 and SphK2 regulate HIFα transcripts in TNBC and ER+ breast cancer cells. Down-regulation of SphK2 with small inhibitory (si) RNA reduces basal HIF-1α and HIF-2α, but not HIF-1β transcripts in multiple TNBC cell lines (Fig. 1A–C and Supplemental Fig. S1A). In contrast, down-regulation of SphK1 does not affect basal HIF-1α, HIF-2α, or HIF-1β mRNA expression in TNBC cell lines (Fig. 1A–C and Supplemental Fig. S1A). As expected, siSphK2, but not siSphK1, reduces the HIFα direct target gene, VEGFA, in multiple TNBC MDA-MB-231 or BT-549 cell lines (Fig. 1D and Supplemental Fig. S1F). There is no noticeable compensatory mechanism of SphK1 observed when we knock down SphK2 with siRNA (knockdown efficiency >80%) in TNBC cells (Fig. 1H, and Supplemental Fig. S1B–E, and S1G–I). Low oxygen tension (1 % O2) transiently enhances the HIF-1α transcript and protein in various TNBC cells, including lung metastatic LM-2-4 cells (Fig. 1E, F, H, and Supplemental Fig. S1I), and HeLa cells (Fig. 1G); knock down of SphK2 with siRNA markedly reduced hypoxia-induced HIF-1α transcripts, nuclear HIF-1α protein (Fig. 1E, F, G, H, and Supplemental Fig. S1I), and HIF-2α protein in TNBC cells (Supplemental Fig. S1I). Efficient knockdown of both SphK1 and SphK2 with their respective siRNAs, as demonstrated by western blot analysis, significantly reduced hypoxia-induced HIF-1α transcript, HIF-1α nuclear protein, and direct target genes VEGFA and VEGFB, in ER+ MCF-7 cells (Supplemental Fig. S2A–D).

Figure 1.

SphK2 silencing prevents HIF-1α and HIF-2α expression. TNBC MDA-MB-231 cells (A, B, D, E, and H), BT-549, lung metastatic LM-2-4 (F), and HeLa cells (G) transfected with siControl or siSphK2 or siSphK1, for some experiments, were exposed to 1% O2 for 20–40 min (E, F, G, and H) as indicated or left at normoxia. mRNA levels of HIF-1α, HIF-2α, and VEGFA were determined by quantitative real-time PCR and normalized to GAPDH or Cyclophilin A or RPLPO (Large Ribosomal Protein, an endogenous control). Data are means ± SD, P<0.05 vs. control, by Student’s t-test. Of note, exposure of cells to 1% hypoxia beyond 40 min to 24 h did not express significant levels of HIF-1α or HIF-2α transcripts compared to normoxia. H) Nuclear fractions from the duplicate cultures of MDA-MB-231 cells were used for Western blot analysis with antibodies against HIF-1α, SphK2, and cytosolic fractions were immunoblotted with SphK1antibody. Lamin A/C or Actin were used as specific organelle markers for nuclei and cytosol, respectively, and to show equal loading and transfer.

These data suggest that SphK2, but not SphK1, regulates expression of both HIF-1α and HIF-2α in TNBC cells.

Nuclear S1P enhances HIFα promoter histone acetylation

Recent studies have suggested histone H3 acetylation is enhanced in response to hypoxia, however exposure of cells under normoxic conditions to the HDAC inhibitor, trichostatin A (TSA), induces rapid histone acetylation (42), suggesting HDACs may play essential roles in hypoxia. Since nuclear S1P generated from SphK2 acts as a natural inhibitor of HDACs, and in turn, enhances histone acetylation (17, 43), we propose that it enhances histone acetylation at the proximal promoter of HIF-1α/2α and hypoxia-responsive HIF-target genes. Since hypoxia induces bulk histone acetylation (42), we examined whether SphK2 regulates hypoxia-mediated histone acetylation in the nuclei of breast cancer cells. Remarkably, hypoxia-induced bulk histone acetylation (H3-K9ac) is markedly attenuated by the knockdown of SphK2, but not SphK1 (Fig. 2A), suggesting that the intracellular S1P-mediated HDAC1, 2 inhibition (17, 43) may regulate hypoxia-mediated histone acetylation. Down-regulation of SphK2 with multiple siRNAs reduces HIF-1α promoter activity (Fig. 2B). Acetylated histones bind to the proximal promoter and influence HIF-1α and HIF-2α transcription (33); however, the mechanism is not well established. Hypoxia enhances a 2.5 fold induction of proximal promoter histone acetylation of HIF-2α (Fig. 2C, D), and a 5 fold induction of the HIF-1α proximal promoter (Supplemental Fig. S3A–B). Down-regulation of SphK2 significantly reduces hypoxia-induced proximal promoter histone acetylation of HIF-2α (Fig. 2C, D) and HIF-1α (Supplemental Fig. S3A–B) in TNBC cells. Ectopic expression of SphK2 enhances 3 fold histone acetylation to the proximal promoter of HIF-2α (Fig. 2E, F) and 5 fold to the promoter of HIF-1α in normoxia (Fig. 2G, H).

Figure 2.

SphK2 enhances HIF-1α and HIF-2α promoter acetylation of histone H3. A) Effect of depletion of SphKs on histone acetylation. MDA-MB-231 cells transfected with siControl, or siRNA targeted to SphK1 or SphK2 were exposed to hypoxia or normoxia for 40 min. Equal protein concentrations from nuclear extracts were immunoblotted with a specific antibody against H3-K9ac. Lamin B1 was used as a specific marker for nuclei and equal loading. B) Human HIF-1α promoter sequence diagram is shown along with the histone acetylation site in the proximal promoter. HeLa cells were co-transfected with the HIF-1α HRE-driven luciferase reporter and Renilla luciferase vectors, together with siControl, or two different siRNAs targeted to human SphK2. Reporter activity is expressed as relative luciferase units normalized to Renilla activity. Data are means ± SD, P<0.05 vs. siControl.

C, D, E) MDA-MB-231 cells transfected with siControl or siSphK2 were incubated with hypoxia or normoxia for 40 min and subjected to ChIP analyses with antibodies to H3-K9ac or normal rabbit IgG, as indicated. The precipitated DNA was analyzed by real-time PCR with primers amplifying the core promoter sequence of the HIF-2α gene (C, D). PCR products were visualized on a 2% Agarose gel. Representative Agarose gel images of ChIP-qPCR are shown (E). MCF-7 cells (F, G) or MDA-MB-231 cells (H, I) were transfected with vector or V5-SphK2 and subjected to ChIP analyses with antibodies to H3-K9ac or total H4, or normal rabbit IgG, as indicated. The precipitated DNA was analyzed by real-time PCR with primers amplifying the core promoter sequence of the HIF-2α gene (F, G) or HIF-1α gene (H, I). ChIP-qPCR products for HIF-2α (G) or HIF-1α (I) were analyzed on a 2% Agarose gel, respectively, and representative graphs are shown. Relative binding to the HIF-1α/2α promoter is expressed as the fold enriched of input. Data are means ± SD. P < 0.05, relative to corresponding control, by Student’s t-test.

Together, these data suggest that nuclear S1P, and the kinase SphK2 that produces it, enhances proximal promoter histone acetylation and transcription of HIF-1α and HIF-2α in breast cancer cells.

Nuclear S1P directly binds to HIFα

HIFα is a multidomain protein, including the N-terminal basic Helix-Loop-Helix (bHLH) domain for DNA binding, tandem Per, Ahr/ARNT, Sim (PAS) domains for dimerization, and C-terminal transactivation domains. Amongst these, the conserved PASB domain in HIF-α is believed to play a pivotal role in the stabilization of HIF-α/β heterodimers (44). Protein stability is already a well-established mechanism by which the HIFα proteins are regulated physiologically (8). Identifying any natural ligands and their effects on the heterodimers would further propel our understanding of HIF functions in cancer, and drive the discovery of small molecule modulators.

Our molecular modelling data revealed S1P docks well into the PASB pocket of mouse orthologue HIFα (44) with a docking score of −11.9 Kcal/Mol compared to sphingosine, which is −7.3 Kcal/Mol (Fig. 3A, B and Supplemental Fig. S4A–B). To provide further evidence that HIFs are targets for S1P, we examined binding of endogenous HIF-1α in nuclear extracts to S1P immobilized on agarose beads. In Fig. 3C, we demonstrate using a S1P-Agarose bead experiment that endogenous HIF-1α specifically binds with S1P affinity matrices compared to control beads. Moreover, S1P: HIF-1α binding was almost completely blocked when the nuclear extract was pre-incubated with an excess amount of S1P. Furthermore, additional binding studies show S1P binds to bacterially expressed recombinant wildtype-HIF-1α protein, but not with deletion mutant HIF-1α recombinant protein with the PAS domains deleted (Fig. 3D). Indeed, wildtype HIF-1α, but not PAS-delta-HIF-1α, binds to [³²P]S1P in a specific manner, as the addition of excess unlabeled S1P abolished binding, but not by sphingosine (Fig. 3E, and Supplemental Fig. S4C). We then sought to examine whether endogenous S1P is bound to HIF-1α/2α in vivo. To this end, we measured the abundance of sphingolipids in HIFα immunoprecipitates from MDA-MB-231 nuclear extracts by LC-ESI-MS/MS. We found that S1P was not only present in the anti-HIF-2α nuclear immune complex, but that hypoxia significantly enhances S1P levels in that protein complex (Fig. 3F). Our data further suggests that down-regulation of SphK2 with siRNA, as we had reported earlier, not only reduced nuclear S1P (17), but also significantly reduces basal, as well as hypoxia-mediated S1P in the anti-HIF-2α immune protein complexes (Fig. 3F). Conversely, ectopic expression of SphK2 not only enhances nuclear S1P (17), but also significantly enhances S1P in the anti-HIF-1α immune complexes in normoxia and is further significantly enhanced under hypoxia (Fig. 3G).

Figure 3.

Nuclear S1P directly binds to HIFα and promotes activation in response to hypoxia.

Docking of S1P is into the pocket of the HIFα-PASB domain. The low energy conformations of S1P were calculated when docked in the model based on the crystal structure of mouse orthologue HIF-1α (44). A) The binding mode (82) of S1P (green) with the mouse HIF-1α PAS-B domain. B) The electrostatic surface of the S1P binding pocket of HIF-1α generated by PyMOL (83). Note that highly conserved His291 HIFs may stabilize the negatively charged phosphate of S1P through a salt bridge interaction. C) An association of HIF-1α with immobilized S1P. Nuclear extracts from MDA-MB-231 cells were preincubated with or without 10 µM S1P for 25 min, followed by pulldown with control (no lipid), or S1P affinity matrices for overnight at 4 °C, as indicated. Beads were washed extensively; bound proteins were separated by SDS-PAGE and then analyzed by Western blotting with anti-HIF-1α antibody. Input is shown separately in the leftmost lane. D) Direct interaction between recombinant HIF-1α and immobilized S1P. Bacterially expressed recombinant purified human HIF-1α and PAS-domain deleted HIF-1α proteins were incubated with S1P-beads or control beads for overnight at 4°C, washed extensively; bound proteins were resolved for Western blot analysis with the indicated antibodies. Wild-type HIF-1α was probed with anti-HIF-1α, and PAS-delta HIF-1α was probed with anti-His-Tag antibodies.

E) Wild-type and PAS-delta HIF-1α recombinant proteins were incubated for 30 min with [³²P] S1P (0.1 nM) in the absence (white bar) or presence (black bar) of 10 µM unlabeled S1P. Proteins were then immunoprecipitated with antibodies against HIF-1α or His-tag or with control IgG. Immunocomplexes were captured with protein A/G agarose beads and washed extensively, and radioactivity was determined by scintillation counting. Inserts, blots of bound HIF-1α and PAS-delta-HIF-1α are shown, respectively. Data are expressed as means ± SD. *P < 0.05. F, G) Detection of S1P is in the HIFα immunoprecipitates. F) Nuclear extracts isolated from MDA-MB-231 cells transfected with siControl or siSphK2 were incubated with 1% O2 or normoxia for 40 min and subjected to immunoprecipitation analyses with antibodies to HIF-2α or control normal IgG. Immunocomplexes were captured with protein A/G agarose beads, washed extensively, lipids extracted, and sphingolipids analyzed by LC-ESI-MS/MS. Quantification of S1P bound to IgG or HIF-2α immunocomplexes. Data are means ± SD, *P < 0.05 vs. siControl with or without hypoxia, as indicated, by Student’s t-test. G) MDA-MB-231 cells were transfected with vector or V5-SphK2 plasmids. SphK2 transfected cells were exposed to ± 1% O2 for 40 min. HIF-1α in the nuclear extracts was immunoprecipitated with anti-HIF-1α or control IgG, complexes captured with protein A/G agarose beads, washed extensively, lipids extracted, and sphingolipids analyzed by LC-ESI-MS/MS. Quantification of S1P bound to IgG, HIF-2α, and HIF-1α immunocomplexes. Data are means ± SD. *P < 0.05 vs. control, as indicated. H) Endogenous HIF-1α associates with V5-SphK2 in the nucleus. MDA-MB-231 cells were transfected with empty vector or V5-SphK2, and nuclear extracts were isolated. Nuclear extracts isolated from MDA-MB-231 cells transfected with vector or V5-SphK2 were immunoprecipitated with normal IgG or anti-V5 antibodies as indicated. Immunocomplexes were captured with protein A/G beads, washed extensively, and immunoblotted with antibodies against HIF-1α or V5. Inputs are shown separately on the left side. I) Active SphK2 is in the nuclear HIF-1α protein complexes, including p300 and HDAC1. HeLa cells transfected with vector, V5-SphK2, or catalytically inactive V5-SphK2^G212E were exposed to 1% O2 or left at normoxia for 20 min and subjected to immunoprecipitation from nuclear extracts with antibodies to HIF-1α or normal IgG, as indicated. Immunocomplexes were captured with protein A/G agarose beads, washed extensively, and immunoblotted with specific antibodies for HIF-1α, V5, HDAC1, p300, and acetylated-Lysine. Inputs are shown on the left side. Antibody against Lamin B1 was used as a specific organelle marker for nuclei, and to show equal loading and transfer.

Next, we examined whether SphK2 interacts with the HIF-1α protein in nuclei. Ectopically expressed SphK2, as expected, enhances endogenous HIF-1α in TNBC cells in normoxia (Fig. 3H, left). Immunoprecipitation of SphK2 from nuclear extracts of TNBC cells found endogenous HIF-1α bound to it (Fig. 3H, right). Hypoxia facilitates HIFα/β dimer formation in the nucleus that is a prerequisite event for HIFα-targeted transcription. Previously published data suggested that HDAC inhibitors repress HIF-1α expression by multiple mechanisms, including acetylation and proteasomal degradation in response to hypoxia (10).

Although class I HDACs, HDAC1 and HDAC3, have shown interactions with HIF-1α and enhance HIF-1α stability and transactivation functions under hypoxic conditions (10, 45, 46), HIF-1α protein stability is also increased by acetylation at lysine 709 via inhibition of HDACs (11). We hypothesize that when nuclear HIF-1α is present in a protein complex with SphK2 (this manuscript), along with the histone acetyltransferase (HAT)/p300-CBP coactivator family of proteins known to be a cofactor of HIF-1α (9, 47) and HDAC1 (48), S1P protects and stabilizes HIFα by acetylation via inhibition of HDACs, not by activation of HAT activity (17) for active transcription in response to hypoxia. Our data reveal that ectopically expressed nuclear wildtype SphK2, but not catalytically inactive SphK2^G212E (Ci-SphK2) (17, 18, 23), is markedly enriched in the anti-HIF-1α immunoprecipitate protein complexes containing p300 and HDAC1 in response to hypoxia (Fig. 3I, right panel and Supplemental Fig. S5, right panel). When we probe the blot with an anti-lysine antibody, we find that nuclear S1P generated from SphK2 enhances acetylated-HIF-1α in these cells (Fig. 3I and Supplemental Fig. S5).

Together, the data suggests S1P directly binds to the HIF-α pocket for active transcription in response to hypoxia.

Role of S1P in HIFα-mediated positive feedback regulation

While hypoxia elevates HIFα expression, and HIFα promotes positive feedback regulation for the expression of HIFα subunits (49), the precise mechanisms governing this regulatory pathway are lacking. We hypothesize that nuclear S1P epigenetically regulates HIFα-mediated HIFα transcription and protein synthesis in low oxygen tension. Our data suggests hypoxia significantly enhances HIF-1α binding to the HIF-2α promoter in both the TNBC (Fig. 4A, B) and ER+ cell lines (Fig. 4E, F). Our data further suggests that knock down of SphK2 reduces S1P in HIFα complexes and destabilizes the HIFα heterodimer that in turn inhibits HIF-1α binding to the HIF-2α promoter in both the TNBC and ER+ cells. Conversely, overexpression of SphK2 facilitates nuclear S1P-mediated HIFα heterodimerization and consequently, enhances HIF-1α binding to the HIF-2α promoter in TNBC (Fig. 4C, D) and ER+ cells (Fig. 4G, H), as shown by a chromatin precipitation (ChIP) assay. Thus, intracellular S1P mechanisms are involved in HIFα positive feedback regulation (Fig. 4I).

Figure 4.

SphK2 enhances HIF-1α protein binding to the HIF-2α promoter. MDA-MB-231 cells (A, B) or MCF7 cells (E, F) transfected with siControl or siSphK2 were exposed to hypoxia (black bar) or left at normoxia (white bar) for 40 min and subjected to ChIP analyses with antibodies to HIF-1α or normal rabbit IgG, as indicated. The precipitated DNA was analyzed by real-time PCR with primers amplifying the core promoter sequence of the HIF-2α gene. ChIP-qPCR products for HIF-2α were analyzed on a 2% Agarose gel (B, F). MDA-MB-231 cells (C, D) or MCF-7 cells (G, H) were transfected with vector (light gray) or V5-SphK2 (dark gray) and processed for ChIP analyses with antibodies to HIF-1α or normal control IgG. ChIP-PCR products were analyzed on a 2% Agarose gel, and representative images are shown (D, H). Relative binding to the promoter using ChIP-qPCR data is expressed as the fold enriched of input. Data are means ± SD, P<0.05, relative to siControl (A, E), or vector transfectants (C, G), respectively. I) A model is shown demonstrating the epigenetic role of nuclear S1P-mediated HIFα positive feedback regulation by hypoxia.

The role of S1P in regulating HIFα target gene VEGFA transcription

Bindings of HIF-1α on the angiogenic gene VEGF promoter increase during hypoxia (50). Since HIFα binds to the hypoxia-responsive element (HRE) domain of the HIFα promoter and promoters of HIF-1α targeted genes in response to hypoxia, we examined the HIF-1α binding to the promoter of human VEGFA regulated by nuclear S1P through epigenetic mechanisms. Transient expression of SphK2 enhances nuclear S1P (17) in the HIF-protein complex (Fig. 3F, G), and also enhances HIF-1α binding to the human VEGFA promoter (Fig. 5C), as shown by ChIP analysis. As expected, hypoxia enhances more than 2-fold HIF-1α binding to the human VEGFA promoter (Fig. 5A, B), as indicated by ChIP analysis. Knock down of SphK2 reduces nuclear S1P (17) and significantly attenuates hypoxia-mediated HIF-1α binding to the VEGFA promoter (Fig. 5A, B). Promoter activation in hypoxia enhances transcription of HIFα direct target genes. Hypoxia enhances more than 4-fold mRNA of human VEGFA, and knock down of SphK2 reduces hypoxia-mediated VEGF transcription in multiple TNBC cells (Fig. 5H and Supplemental Fig. S6D). SphK2 enhances nuclear S1P-mediated proximal promoter histone acetylation through inhibition of HDAC 1 and 2 (17), and enhances transcription of genes in TNBC cells (Fig. 5I). As a proof of principle determination, we have examined some known HIFα-targeted cancer-related genes to test whether nuclear S1P regulates their transcription under hypoxia. Hypoxia markedly enhances transcription of SERPINE1 (Fig. 5D), OCT4 (Fig. 5E), ALKBH5 (Fig. 5F and Supplemental Fig. S6A), and CD44 (Fig. 5G), while down-regulation of SphK2 significantly reduces hypoxia-induced transcription of these genes (Fig. 5D–H). Down-regulation of SphK2 also reduces the hypoxia-induced Dec1 gene in multiple TNBC cell lines (Supplemental Fig. S6B–C).

Figure 5.

SphK2 enhances HIF-1α binding to the VEGF promoter. MCF-7 cells (A, B) or MDA-MB-231 cells (C) transfected with siControl or siSphK2 were exposed to hypoxia (black bar) or normoxia (white bar) for 40 min. For some experiments, MCF-7 cells transfected with vector (light gray bar) or V5-SphK2 (dark gray bar) were incubated at normoxia (C). Cells were subjected to ChIP analyses with antibodies to HIF-1α or normal IgG. The precipitated DNA was analyzed by real-time PCR with primers amplifying the core promoter sequence of the VEGFA gene (A, C). For some experiments, ChIP-qPCR products were analyzed on a 2% Agarose gel (B). Relative binding to the VEGFA promoter is expressed as the fold enriched of input. Data are means ± SD. P < 0.05, relative to siControl (A) or vector transfectants (C), by Student’s t-test. D-H) MDA-MB-231 cells transfected with siControl or siSphK2 were treated with hypoxia (black bar) or normoxia (white bar) for 40 min and subjected to qPCR analyses with the specific primers to human SERPINE (D), OCT4 (E), ALKBH5 (F), CD44 (G), and VEGFA (H). mRNA levels of the genes, as indicated, were determined by quantitative real-time PCR and normalized to GAPDH. Data are means ± SD, P<0.05 vs. siControl, by Student’s t-test. I) A model is shown demonstrating the epigenetic role of nuclear S1P-mediated HIFα-direct target gene expression by hypoxia.

Together, these data suggest intracellular S1P regulates the expression of HIFα target genes in response to hypoxia in breast cancer cells.

Analysis of expression levels of SphK2 and HIF-1α in human patient tissue specimens

SphK2 expression has been found elevated in both colon cancer (51) and non-small cell lung cancer (NSCLC) (52). In NSCLC tumor tissues, high expression of SphK2 has been associated with cancer progression and has served as an independent prognostic factor for OS and disease-free survival (DFS) (53), supporting the idea of endogenous SphK2 as a potential cancer-promoting factor in other human cancers, including breast (21, 54–56). To determine the importance of SphK2 in breast cancer clinical tissue samples, we used TCGA, one of the two large breast cancer patient cohorts for gene expression data analysis, and found SphK2 transcripts were significantly elevated in breast cancer patients compared to matched controls (Fig. 6A, left). Data analysis using the same cohort according to PAM50 breast cancer subtyping analysis shows that ER+ (N=78) breast cancers have significantly elevated SphK2 expression compared to control (N=114) (Fig. 6A, right). TNBC (N=14) and HER2+ (N=22) subtype analysis did not show a statistically significant difference compared to normal, however the number of patients in this subtype was low (data not shown). Gene expression data analysis using the METABRIC cohort with PAM50 subtyping analysis data revealed SphK2 was expressed in all the major subtypes of breast cancer, including luminal A & B subtypes, which have significantly more expression compared to HER2+ and TNBC subtypes (Supplemental Fig. S7A).

Figure 6.

SphK2 and HIF-1α expression levels are elevated in BC patient samples. A) Boxplot showing the distribution of SphK2 mRNA expression in the TCGA BC cohort breast cancer vs. matched healthy control (N=114, left boxplots) and ER+ (N=78) vs. healthy control, based on PAM50 BC subtyping with the Chi-square and Fisher's Exact tests (right boxplots). The boxes represent the median (black middle line). Two-tailed tests were performed. P<0.05 is considered a significant difference. B) Tissue nuclear SphK2 and HIF-1α expression levels elevated in ER+ and TNBC patient-derived xenografts. Equal amounts of nuclear proteins isolated from ER+ (BC1-3), TNBC (BC4-9) PDXs, and healthy control breast tissues (Cont. breast1-3) were subjected to Western blot analyses with the specific antibodies to SphK2, HIF-1α, and histone H3 as indicated. Histone H3 is used as a specific organelle marker for nuclei. C) Tissue samples from (B) were used for quantitative real-time PCR. mRNA levels of HIF-1α, HIF-2α, and VEGFA were normalized to RPLPO. Data are expressed as means ± SD. ANOVA and Tukey’s ad hoc analyses, P<0.05 vs. control breast or between TNBC vs. ER+ tissue samples. D) TNBC patients with elevated HIF-1α mRNA levels are associated with poor survival. METABRIC cohort gene expression analyses, according to PAM50 BC subtyping, were analyzed for TNBC patient survival rates for HIF-1α mRNA expression levels. A Kaplan–Meier plot is showing the survival rate comparison between TNBC patients who have average or high levels of HIF-1 α mRNA expression. Chi-square and Fisher’s Exact tests, Two-tailed tests were performed. P<0.05 is considered a significant difference.

Since the bioactive sphingolipid S1P, and the kinase SphK2 that produced it, both localized in the nuclei of the breast cancer cells (17, 18, 23, 54, 57), we examine whether expression of SphK2 was elevated in the nuclei of the patient tissue specimens compared to control breast tissues. PDX models of human breast cancer were used. Early passage (<10) PDXs represent the clinical characteristics of human breast cancer subtypes. As previously mentioned (35, 58–60), PDX models of clinically defined ER+ (N=3) and TNBC (N=6) metastatic patient tissue specimens were developed (Supplemental Table S2) (Fig. 6B, C). Using a small number of PDXs, we found expression of SphK2 was elevated in passage-1 of breast tumor tissues of PDXs; the level of HIF-1α was also increased during carcinogenesis in breast tissue. Elevated HIF expression levels in primary tumor biopsies indicate increased risk for metastasis, which is the major cause of death in breast cancer (61). Our data revealed that similar to SphK2, expression of HIF-1α was elevated in breast cancer compared to healthy breast tissue specimens, and expression was more pronounced in the metastatic ER+ and TNBC samples (Fig. 6B, C). In agreement with the high expression of HIF-1α associated with a breast cancer patienťs prognosis (2, 3), METABRIC gene expression data analysis using PAM50 subtyping data analysis revealed that high HIF-1α expression is linked with worse survival of TNBC (p=0.0154) (Fig. 6D) compared to non-TNBC patients (p=0.0982) (Supplemental Fig. S7D). An elevated level of HIF-1α target gene VEGFB expression is significantly associated with the Overall Survival of TNBC (p=0.0429) and non-TNBC (p=0.0253) patients (Supplemental Fig. S7B–C). Thus, SphK2 and HIF-1α may be useful biomarkers for metastatic TNBC patients.

SphK2 generated S1P promotes expression of HIF-1α/2α and MDA-MB-231 xenograft tumor growth

A representative 3D culture method for cancer cells is the tumor-derived spheroid culture, which more closely mimics the natural environment found in vivo, and possesses a hypoxic tumor core (62). Hypoxia triggers epigenetic changes and cancer stem cell marker gene expression is associated with tumor progression (63). HIF-1α is a master regulator of breast cancer metastatic niche formation and lung metastasis. As a proof-of-concept, we examined whether inhibiting SphK2 with the existing SphK2 inhibitor, K145 (55), attenuates HIFα-mediated cell survival, and increases dead cells in ER+ breast cancer cell spheroids. Our data suggests that ER+ MCF-7 cancer cell spheroids markedly increase SphK2 (Fig. 7G), nuclear H3-K9ac, HIFα levels, and target gene VEGF mRNA compared to non-spheroid parental cancer cell (Fig. 7C, D, E, F). Conversely, the selective K145 inhibitor of SphK2, reduces nuclear S1P (18), and consequently almost completely blocked H3-K9ac, HIF1α associated spheroid formation, and enhanced dead cells (Fig. 7A, B, C, D, E, F). Our data suggests that endogenous SphK2/S1P is vital for cancer cell spheroid growth and survival. Endogenous SphK2 and the resulting S1P are essential for tumor growth (21, 55, 56, 64, 65). Notably, a selective SphK2 inhibitor, ABC294640, has shown promising results in a phase I clinical trial in patients with advanced solid tumors (66). To provide a further proof-of-concept, we investigated whether inhibiting SphK2 in vivo with K-145 (55) would block HIF-1α-mediated tumor growth in the TNBC xenograft model. We found that the anti-cancer K-145 compound was well tolerated in mice and did not alter body weight until 18 days into treatment in a mouse JC mammary adenocarcinoma syngeneic model (55). K-145 treatment showed in vivo activities at 20–35 mg/kg doses in the syngeneic mouse experiment (55), and we used a 30 mg/kg dose through i.p. injection for xenograft experiments using NSG mice. As illustrated in Figure 7H, mice bearing MDA-MB-231 tumors on day 13 when tumors reached a palpable size of approximately 50–100 mm3 were treated with a vehicle or K-145 till day 35. Animals treated with K-145 showed tumor growth inhibition in MDA-MB-231 xenografts. In the K-145 treatment group, on day 35, the mean volume of the MDA-MB-231 tumors in the treated mice had a reduced tumor volume >50% compared to vehicle-treated mice (Fig. 7H). Tumor weights of K-145-treated mice were also significantly less than that of vehicle-treated mice (Fig. 7I). Since, we previously had detected K-145 in the tumors by using ESI-MS/MS methods (55), our present data reveals S1P and dihydro-S1P levels in the tumor tissue nuclei were significantly reduced in the treated group compared to the vehicle group (Fig. 7J). Notably, H3-K9ac, HIF-1α, and HIF-2α levels were decreased in the K145-treated tumor samples compared to the vehicle controls (Fig. 7K, L), which is consistent with the results from breast cancer cells. Also, K145-treated tumors have significantly decreased HIFα-target genes VEGFA and SERPINE1 mRNA compared to the vehicle controls (Fig. 7 M, N). As expected, based on our previous mouse studies with the K145 compound (55), we also did not observe significant changes in body weights, nor in the major organs, such as heart, lung, liver, and kidney (data not shown), thus indicating a lack of general toxicity of K-145.

In sync with our published results (55), our data strongly suggests K-145 has in vivo antitumor activity, reduces intracellular S1P, and inhibits HIFα-mediated breast cancer progression.

DISCUSSION

Over the past two decades, detailed molecular insight has been sought to determine the complex role HIFs play in regards to oxygen–sensing pathways involving the canonical oxygen-dependent PHD-pVHL tumor suppressor protein axis dependent proteasomal degradation (5, 6, 8, 67). Here, we identify and characterize mechanisms through which intracellular bioactive sphingolipid metabolite S1P influences HIF-1α/2α transcription and activation in response to hypoxia, and assess the functional importance of intracellular S1P and one of the two isoenzymes that produce it, SphK2, in breast cancer progression, focusing on TNBC and ER+ subtypes. Proximal promoter histone acetylation with the histone mark (H3-K9ac) (68) is found in the HIF-1α/2α genes (33). Low oxygen tension is implicated in upregulating nuclear bulk histone acetylation (69), and the HDAC inhibitor, enhances histone acetylation in normoxia (70), suggesting that HDACs are the critical regulator of histone acetylation in response to hypoxia. Hypoxia has been implicated in the upregulation of SphK2 activity in cancer cells (71). Since nuclear S1P generated from the nuclear-localized SphK2 that inhibits HDAC1/2, in turn, enhances histone acetylation, we found that they regulate hypoxia-regulated bulk histone acetylation and proximal promoter histone acetylation of HIF-1α/2α genes in TNBC cells (Figs. 1, 2, Supplemental Figs. S1–2). In contrast, we found cytosolic SphK1 is not involved in nuclear, bulk histone acetylation (Fig. 2A), nor transcription of HIF-1α/2α and HIFα-target gene VEGFA in TNBC cells (Fig. 1 and Supplemental Fig. S1). Extracellular S1P generated from the cytosolic SphK1 and via S1PR1 signaling through activating AKT/GSK3β has been implicated in regulating HIF-1α stability and activity (41). In agreement with previous studies (41), the down-regulation of SphK1 with siRNA reduces HIF-1α expression and activity in ER+ breast cancer MCF7 cells (Supplemental Fig. S2). Transcriptionally active HIFs form a heterodimer via its PAS domains that comprise HIF-α and HIF-1β subunits (44). Our molecular modeling data revealed that S1P docks well (Fig. 3A, B and Supplemental Fig. S4) in the pocket of mouse orthologue HIFα PASB (docking score −11.9) compared to PASA (docking score −9.6). We have confirmed the HIF-1α-S1P interaction using several approaches (17, 23, 27). S1P-agarose beads and endogenous HIF-1α or recombinant purified HIF-1α binding data revealed the S1P-HIF-1α interaction is specific and dependent on PAS domains (Fig. 3C, D, E). Further studies with point mutation experiments are warranted to confirm our modelling data of S1P and HIF-1α.

As expected, nuclear S1P was enriched in the HIF-1α/2α immune protein complexes and hypoxia markedly enhanced S1P enrichment into the immune complexes (Fig. 3 F, G). Remarkably, however, these were dependent on the nuclear expression of SphK2, and the presence of SphK2, into the HIF-immune complexes. Interestingly, we observed that knockdown of SphK2 led to decreased hypoxia-mediated S1P enrichment in HIF-2α immune complexes (Fig. 3F). Conversely, overexpression of SphK2 led to enhanced basal and hypoxia-mediated S1P enrichment in HIF-1α immune protein complexes (Fig. 3 G). Our results are consistent with the idea that hypoxia enhances SphK2 activity (71), which induces nuclear S1P complexed with HIFα, promoting S1P binding to HIFα in response to hypoxia. We observed that ectopically expressed SphK2 enriched on the HIF-1α immune protein complexes, including p300 (9, 47, 72), HDAC1 (48), and enhances stability by lysine acetylation of HIFα in response to hypoxia. This novel observation is consistent with published data (11) and the idea that intracellular S1P acts as an endogenous ligand-agonist of HIFα-heterodimerization and activity in response to hypoxia. Further studies are warranted to dissect S1P-mediated HIFα acetylation via inhibition of HDACs in the nucleus.

In the canonical hypoxia-mediated transcription pathway, HIF-1α/2α-HIF-1β heterodimer binds to the HRE characterized by a conserved RCGTG DNA motif (73). HIF-1α/2α have been implicated in transcriptionally regulating themselves in a positive feedback manner (74), and canonical target genes associated with the promoter histone modification, along with the cofactor involvement, may represent additional layers of regulation (75). Indeed, our data revealed that ectopically overexpressed SphK2 led enhanced promoter occupancy of HIF-1α protein on the HIF-2α promoter, and conversely, knockdown of SphK2 led reduced hypoxia-mediated HIF-1α binding of the HIF-2α promoter in breast cancer cells (Fig. 4). This observation is consistent with the observation that SphK2/S1P regulates transcription of HIFα (Fig. 1, 2, Supplemental Figs. S1–2).

Recent studies suggested that target genes of HIF-1α and HIF-2α are different, but complementary (75); however, the canonical direct target genes of HIFα are shown to be involved in angiogenesis, metastasis, and activation of cancer stem cells. HIFα has also been shown to bind to the VEGF promoter and regulate its transcription in response to hypoxia (50). Our data revealed that ectopic expression of SphK2 led to increased binding of HIF-1α on the VEGF promoter, and conversely, knockdown of SphK2 reduced hypoxia-mediated HIF-1α promoter occupancy (Fig. 5A, B). This observation is in agreement with the observation that SphK2 regulates transcription of an angiogenesis gene (VEGFA), cancer stem cell-related genes (ALKBH5, CD44), a differentiation and proliferation-related gene (Dec 1), and metastasis-related genes (SERPINE1, OCT4), in various breast cancer cells (Fig. 5 D–H, Supplemental Fig. S6).

METABRIC cohort and PAM50 subtyping gene analysis data revealed SphK2 expressed more in the Luminal subtypes compared to the basal or the Her2+ breast cancer subtypes. TCGA gene expression analysis revealed ER+ breast cancer patients had elevated SphK2 transcripts compared to matched controls (Fig. 6A). SphK2 expression is also found elevated in other human cancers including NSCLC (52), and colon cancer (51), with targeting SphK2 with a selective inhibitor reducing lung cancer growth (76), papillary thyroid carcinoma (77), bladder cancer (78), colorectal cancer (79), and colon cancer (80), with just one exception (81), indicating SphK2 is a potentially good therapeutic target (65). Our previous work also revealed SphK2/S1P is essential for metastatic breast cancer cell growth and survival (21, 55). An anti-sphk2 drug (ABC294640) has been tested in clinical trials for patients with advanced solid tumors and presents a stable disease response (66). Together, our results highlight the importance of investigating the underlying mechanisms of SphK2-mediated metastatic breast cancer progression that will help to guide more focused trials. Elevated HIFα protein levels are associated with an increased risk of breast cancer metastasis and mortality in general (2–4), and worse outcomes for TNBC patients (Fig. 6D), those lacking conventional treatment options. Interestingly, using a small number of established PDXs, our data implied that similar to SphK2, expression of HIF-1α was elevated in breast cancer tissue specimens and was more pronounced in the TNBC (Fig. 6B, C, D).

Together, these suggest that similar to HIF-1α, expression of SphK2 is important in subsets of breast cancer, and more importantly, in metastatic TNBC. Further studies are essential with more patient tissue specimens to confirm SphK2 expression level as a biomarker for breast cancer metastasis.

Since SphK2/S1P regulates HIF functions both in ER+ and TNBC cells, as proof-of-concept validation, we examined whether it regulates HIF functions in in vivo models closely related to humans. 3D culture cancer cell tumorspheres maintain in vivo a hypoxic tumor core (62). We observed that tumorspheres of ER+ MCF-7 cells had elevated nuclear levels of histone mark H3-K9ac and HIF-1α/2α expression when cells were grown in matrigel matrices. Remarkably, inhibiting SphK2 with the selective inhibitor, K-145 (55), reduces nuclear S1P in the HIF complex, and almost completely abolished H3-K9ac, HIFα, and target genes VEGFA, B in our ER+ pre-clinical model (Fig. 7A–G). Since our primary focus is on HIF-1α regulation in TNBC, our data revealed TNBC MDA-MB-231 xenograft tumor tissue nuclei have a markedly elevated level of H3K9-ac histone mark accompanied by a high HIF-1α expression (Fig. 7H, K). In order to validate our concept that nuclear S1P regulates HIFs functions in vivo, we have used the SphK2 inhibitor K145, that in agreement (55), markedly reduced MDA-MB-231 TNBC xenograft tumor volumes and weight (Fig. 7H–I), accompanied by reduced tumor tissue nuclear S1P and dihydro-S1P, and reduced H3-K9ac histone mark, expression of HIF-1α/2α, and HIFα-target genes VEGFA and SERPINE1 (Fig. 7J–N).

Thus, our work highlights a critical role for intracellular S1P and SphK2 and their regulation in HIF-1α/2α functions associated with ER+, and particularly, TNBC subtype breast cancer progression. Our data further suggests a subset of breast cancers with elevated SphK2 and HIFα expression would be potential candidates for SphK2 targeted therapy.

Significance.

Our findings offer a preclinical proof of concept that signaling by an intracellular lipid mediator, S1P, may be an effective target to prevent HIF functions in aggressive breast cancers.

Abbreviations

ChIP

Chromatin immunoprecipitation

CSC

Cancer stem cell

DH-Sph

Dihydro-sphingosine

ER

Estrogen receptor

EMT

Epithelial-to-mesenchymal transition

HIF

Hypoxia-inducible factor

HRE

Hypoxia responsive element

HDAC

Histone deacetylase

METABRIC

Molecular Taxonomy of Breast Cancer International Consortium

OS

Overall survival

Sph

Sphingosine

SphK1

Sphingosine kinase1

SphK2

Sphingosine kinase 2

S1P

Sphingosine-1-phosphate

TNBC

Triple-negative breast cancer

TCGA

The Cancer Genome Atlas

VEGF

Vascular endothelial growth factor