Sphingosine-1-Phosphate Produced by Sphingosine Kinase 1 Promotes Breast Cancer Progression by Stimulating Angiogenesis and Lymphangiogenesis

Breast cancer is the most commonly diagnosed cancer among women accounting for 30% of all new cancer cases, and close to 40,000 breast cancer deaths are expected to occur among U.S. women in 2011 (1). The majority of deaths of breast cancer patients occur after it metastasizes and becomes a systemic disease (2). To date, most systemic therapies, including cytotoxic chemotherapy, target the tumor itself and not the host tumor microenvironment, which is known to play an important role in the progression of cancer (3, 4).

It is well established that hemangiogenesis is one of the key tumor microenvironmental factors in cancer progression. Tumor growth beyond a diameter of a few millimeters requires hemangiogenesis which contributes to the development of metastatic disease (5, 6). Although hemangiogenesis is known to be an acquired capability of cancer (7), results of multiple randomized clinical trials that targeted hemangiogenesis using bevacizumab, an anti-VEGF monoclonal antibody (mAb), failed to show an overall survival benefit in breast cancer (8–10).

Breast cancer first metastasizes to its sentinel lymph node, and the level of lymph node metastasis is a major determinant of staging and prognosis of breast cancer (2, 11). Recently, lymphangiogenesis, a process that generates new lymphatic vessels from preexisting ones, was found to be mediated by several factors including VEGF-C, D, and angiopoietins (Ang1 and Ang2), which provided new insights into how the lymphatic vessels grow and affect metastasis (12, 13). Although clinical and experimental evidence suggest a role for lymphangiogenesis in lymph node metastasis, this process is far less understood than hemangiogenesis (13, 14).

The pleiotropic bioactive lipid mediator sphingosine-1-phosphate (S1P) is now emerging as a key regulatory molecule in cancer through its ability to promote cell proliferation, migration, invasion, and hemangiogenesis (15, 16). S1P is generated intracellularly by 2 sphingosine kinases, sphingosine kinase 1 (SphK1) and SphK2, and is exported out of the cells where it regulates many functions by binding to and signaling through a family of 5 G protein-coupled receptors (S1P_1–5). This process known as “inside-out” signaling explains the autocrine and paracrine actions of S1P (15). We previously showed that SphK1, but not SphK2, is involved in S1P export from breast cancer cells mediated by the ATP-binding cassette transporters, ABCC1 and ABCG2 (17). Expression of SphK1 is upregulated in breast cancer (18, 19), correlates with poor prognosis (20), and is associated with resistance to chemotherapy (18). Furthermore, abundant evidence implicates the SphK1/S1P/S1P₁ axis in hemangiogenesis and vasculogenesis (21). In contrast, only a few studies so far have examined the involvement of SphK1 and S1P in lymphangiogenesis. S1P has been shown to induce in vitro lymphangiogenesis via the S1P₁ receptor (22), and it was recently reported that SphK1 and S1P in lymphatic endothelial cells (LEC) are required for the proper development of lymphatic vessels (23). However, nothing is yet known of the role of SphK1 and S1P in tumor-induced lymphangiogenesis in vivo.

In this study, we utilized an improved syngeneic breast cancer cell implantation method, which models human breast cancer biology better than conventional xenograft subcutaneous implantation, to explore the role of SphK1 and S1P in hemangiogenesis and lymphangiogenesis. We found not only that circulating S1P levels correlated with tumor burden but also that targeting SphK1 with a specific inhibitor reduced tumor growth, lymph node and lung metastasis, and decreased hemangiogenesis and lymphangiogenesis around the primary tumor and in the lymph nodes. Our results suggest that targeting SphK1 and S1P may be a useful additional modality for the treatment of metastatic breast cancer.

4T1-luc2 cells, a mouse mammary adenocarcinoma cell line that has been engineered to express luciferase (PerkinElmer), were cultured in RPMI Medium 1640 with 10% FBS. Green fluorescent protein (GFP)-expressing human umbilical vein endothelial cells (HUVEC) and GFP-expressing human lymphatic endothelial cells (HLEC) purchased from Angio-Proteomie were maintained in endothelial cell medium supplemented with 5% FBS and endothelial cell growth supplement (ScienCell Research Laboratories).

Three sense sequences of siRNA cassettes specifically targeting the nucleotides of mouse SphK1 (accession number: NM_011451.1) were designed using siRNA Target Finder (Ambion). The specificity of the selected sequences was confirmed by BLAST search. The lentiviral short hairpin RNAs (shRNA) were constructed using a pLL3.7 expression vector as described previously (24). The sequences of the shRNAs for mouse SphK1 are as follows: shRNA 1, 5′- GCACCCAAACTACCTTTGGAT-3′; shRNA 2, 5′- GCACCTTCTTTCGCCTAGCAA-3′; shRNA 3, 5′-GAGGCAGAGATAACCTTTAAA-3′. Recombinant lentivirus was produced by cotransfection of 293FT cells with lentiviral vector and the packaging vectors using calcium phosphate. The viruses were collected from the culture medium and purified by polyethylene glycol precipitation. The transduction efficiency of lentivirus in 4T1 cells was determined by fluorescence microscopy. The silencing efficiency was determined using quantitative PCR (qPCR) and Western blot analysis.

All animal studies were conducted in the Animal Research Core Facility at VCU School of Medicine in accordance with the institutional guidelines. All procedures were approved by the VCU Institutional Animal Care and Use Committee (IACUC) that is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). Female BALB/c mice, 8 to 12 weeks of age, weighing approximately 20 g were obtained from Harlan. 4T1-luc2 cells (1 × 10⁵ cells in 10 μL RPMI) were surgically implanted in the upper fat pad under direct vision. One day after implantation, mice were randomized by tumor size determined by bioluminescence with the IVIS Imaging System (Xenogen). SK1-I in PBS was injected intraperitoneally, as indicated, every day at a dose of 20 mg/kg. At the indicated times, the animals were sacrificed by exsanguination, blood was collected, tumors excised, weighed, fixed in formalin, and embedded in paraffin or frozen in liquid nitrogen.

d-Luciferin (0.2 mL of 15 mg/mL stock; PerkinElmer) was injected intraperitoneally into mice previously implanted with 4T1-luc2 cells. Living Image Software (Xenogen) was used to quantify the photons/sec emitted by the cells. Bioluminescence was measured and quantified at 5-minute intervals over 30 minutes using a subject height of 1.5 cm, medium binning and an exposure time of 0.5 seconds to 1 minutes. Bioluminescence was then determined by the peak number of photons/sec calculated over this time frame. Axillary lymph node metastasis was quantified in vivo after intraperitoneal injection of luciferin and primary tumor resection. Lung metastasis was quantified ex vivo after the lungs were removed.

Lipids were extracted from serum and tissues and sphingolipids quantified by liquid chromatography–electrospray ionization–tandem mass spectrometry (LC-ESI-MS/MS; 4000 QTRAP, AB Sciex) as described (25).

Angioreactors [directed in vivo angiogenesis assays (DIVAA) kit; Trevigen] were implanted into the subcutaneous layer of the back of 8-week-old nude mice under anesthesia according to the manufacturer's instructions. Mice were euthanized with CO₂ at day 11 after implantation and the angioreactors were removed under stereomicroscopy (SZ51, Olympus).

Single-cell suspensions were obtained from lymph nodes as described (26). For DIVAA/FACS (fluorescence-activated cell sorting) analyses, Matrigel and fibrotic reactive tissue in the open end of the angioreactors were removed with Dulbecco's modified Eagle's medium and digested with a mixture of 9 mL of Cell Sparse (Trevigen) containing 0.1% collagenase type 2 (Gibco), 1 mL of 0.25 U/mL dispase (Gibco), and 75 μL of 0.1% DNase (Invitrogen) for 30 minutes to obtain single-cell suspensions. Tumors were minced and digested similarly. Cell suspensions were blocked with anti-CD16/CD32 (Mouse BD Fc Block; BD Biosciences) and then stained with the following antibodies as indicated: Alexa 488–conjugated LYVE-1 (eBioscience); PE-conjugated podoplanin, PerCP-Cy5.5–conjugated CD45, APC-conjugated CD31, Alexa 700–conjugated TER-119 (Biolegend); or appropriate matched, fluorochrome-labeled isotype control mAb. The LIVE/DEAD Viability Assay kit (Invitrogen) was used to eliminate dead cells. Cells were analyzed by FACS using BD FACSCanto II and BD FACSAria II (BD Biosciences) and corollary data assessed with BD FACSDiva Software version 6.1.3 (BD Biosciences).

Immediately after the sacrifice of the animals, tumor samples were fixed in 10% neutral buffered formalin for immunohistochemical analyses of cell proliferation and apoptosis. 4T1-luc2 cell proliferation in the tumor was determined by staining histologic sections with mAbs against Ki67 (Dako), a nuclear protein expressed in proliferating cells. Apoptosis was determined by terminal uridine deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay using the ApopTag Peroxidase In Situ Apoptosis Detection Kit S7100 (Millipore).

The tumors were also embedded in optimal cutting medium (OCT 4583; Sakura Finetek) and frozen for immunofluorescent analysis for hemangiogenesis and lymphangiogenesis. The sections were stained with the following antibodies as indicated: CD31 (BD), LYVE-1 (Abcam) and F4/80 (kindly provided by Bin-Zhi Qian in Jeffrey Pollard laboratory at Albert Einstein College of Medicine; ref. 27). The stained sections were examined with LSM510 laser scanning confocal microscopes (Zeiss) and microvessel density was determined as previously described (28).

All studies using patient samples were conducted in accordance with the institutional guidelines after approval by VCU Institutional Review Board. Human sera were collected by the Tissue and Data Acquisition and Analysis Core of VCU and S1P levels determined in serum from stage IIIA breast cancer patients and age/ethnicity-matched healthy volunteers.

Cell number of 4T1-luc2 cells was determined by measurement of luciferase activity with the luciferin substrate (PerkinElmer) using a VICTOR X4 Multilabel Plate Reader (PerkinElmer). qPCR (15) and tube formation assay were conducted as described previously (29).

Results were analyzed for statistical significance with a 2-tailed Student's t test, with P < 0.05 considered significant. Experiments were repeated at least 3 times in triplicate with consistent results. In vivo experiments were repeated 3 times and each experimental group consisted of at least 6 mice.

Previous studies showed that S1P and SphK1, the enzyme that produces it, regulate many processes important for breast cancer (15). To examine their involvement in breast cancer progression in vivo, we utilized an enhanced syngeneic mouse metastatic breast cancer model in which 4T1-luc2 murine mammary cancer cells were orthotopically implanted under direct vision into the chest mammary fat pad of immune-competent mice. We have found that this chest orthotopic model more accurately mimics human breast cancer progression than subcutaneous models. Indeed, 4T1-luc2 cells produced large tumors in the chest mammary fat pad that rapidly metastasized and increased total tumor burden (Fig. 1A). Metastatic spread to the regional axillary lymph nodes and the lungs was quantified by bioluminescence after removal of the 4T1-luc2 tumors from the implantation sites (primary tumors). As can be seen in Fig. 1B, lymph node metastases emerged earlier than lung metastases, with a rapid increase in the incidence of metastasis to the lymph nodes, similar to human breast cancer progression.

Expressions of both SphK1 and SphK2 were very low in naive mammary fat pads, whereas 4T1-luc2 cells express much higher levels of SphK1 mRNA than SphK2 (Fig. 1C). However, after implantation in mammary fat pads, SphK1 mRNA levels in breast tumors were significantly increased (Fig. 1C), with concomitantly increased S1P levels in the tumors (Fig. 2E). Interestingly, levels of S1P in serum from these mice were also significantly elevated (Fig. 1D), suggesting that overexpression of SphK1 in the tumors may be responsible for the increased circulating S1P.

Because it has been shown that overexpression of SphK1 correlates with poor prognosis of breast cancer patients (20), it was of interest to measure serum levels of S1P in breast cancer patients. We found that the serum S1P levels were significantly elevated in stage IIIA breast cancer patients who have lymph node metastases, compared with age/ethnicity-matched healthy volunteers (Fig. 1E).

We next examined the effect of inhibition of SphK1 with SK1-I [(2R,3S,4E)-N-methyl-5-(4-pentylphenyl)-2-aminopent-4-ene-1,3-diol (BML-258); ref. 30]. SKI-I is a potent, water-soluble, isoenzyme-specific inhibitor of SphK1 that in contrast to pan-SphK inhibitors does not inhibit SphK2, protein kinase C, or numerous other protein kinases (30). Consistent with previous studies in other types of cancer cells (30, 31), SK1-I inhibited growth of 4T1-luc2 cells in a dose-dependent manner (Fig. 2A). A significant effect was observed at a concentration of 3 μmol/L and complete inhibition of growth at 10 μmol/L (Fig. 2A). We confirmed that SK1-I decreased the enzyme activity of SphK1 and that downregulation of SphK1 with specific siRNA, similar to SK1-I, also suppressed the growth of these cells (data not shown). We first examined circulating levels of SK1-I following a single intraperitoneal injection of 20 mg/kg. Plasma SK1-I levels reached a maximum concentration of 0.6 μmol/L within 2 hours but still could be detected up to 12 hours. Concomitantly, plasma levels of S1P were significantly reduced up to 12 hours (Supplementary Fig. S1). We then investigated the effect of SK1-I on the growth of 4T1-luc2 tumors in mouse mammary fat pads. Intraperitoneal injections of SK1-I significantly reduced both tumor volume and weight (Fig. 2B and C). The tumors from animals treated with SK1-I also showed reduced mitotic activity compared with those from vehicle-treated animals, as shown by Ki67 immunohistochemistry (Fig. 2D). Conversely, TUNEL staining revealed a large increase in apoptotic cells in tumors from SK1-I–treated mice (Fig. 2D). Consistent with the upregulation of SphK1 in tumors (Fig. 1C), S1P levels in tumors gradually increased compared with naive or sham operated mammary fat pads as measured by LC-ESI-MS/MS. This increase of S1P in the tumors was prevented by treatment with SK1-I (Fig. 2E). Notably, SK1-I was taken up by the mammary tumors and its levels gradually increased with time (Fig. 2F).

The total tumor burden determined by the IVIS Imaging System was also significantly suppressed by SK1-I treatment (Fig. 3A). Serum S1P levels, which are elevated in the tumor-bearing mice, were also reduced by SK1-I treatment (Fig. 3B). Serum levels of SK1-I in these mice increased with time (Fig. 3C). Importantly, SK1-I significantly reduced lymph node and lung metastasis (Fig. 3D and E). Taken together, these results suggested that S1P produced by SphK1 in the mammary tumors is involved in lymph node and lung metastasis.

Because hemangiogenesis and lymphangiogenesis are considered to play critical roles in lymph node and lung metastasis (28), it was of interest to next examine the role of S1P generated by SphK1 in these processes. We first determined the effect of S1P on in vitro hemangiogenesis and lymphangiogenesis. In agreement with previous studies (22, 32), we found that S1P is a potent hemangiogenic and lymphangiogenic factor for human blood and lymphatic endothelial cells, respectively, as shown by enhanced tube formation (Fig. 4A and B). This reflects cell migration and organization of a network architecture, key aspects of both hemangiogenesis and lymphangiogenesis. Ang2, whose levels in breast cancer correlate with metastasis to lymph nodes and poor prognosis (33), was used as a positive control. SK1-I inhibited in vitro hemangiogenesis and lymphangiogenesis (Fig. 4A and B), suggesting that the effects of Ang2 are mediated at least in part by SphK1 and formation of S1P.

We next sought to determine whether high expression of SphK1 and S1P production in 4T1-luc2 cells contribute to enhanced lymphangiogenesis and hemangiogenesis. Whereas supernatants from 4T1-luc2 cells significantly stimulated tube formation of HUVECs and HLECs, shRNA knockdown of SphK1, which decreased SphK1 expression (Supplementary Fig. S2A and S2B) and the release of S1P (Supplementary Fig. S2C), markedly reduced them (Supplementary Fig. S2D and S2E). Moreover, treatment with S1P did not affect expression of either VEGF-A or VEGF-C mRNA in 4T1-luc2 cells (Supplementary Fig. S3). These results indicated that S1P produced by SphK1 in 4T1-luc2 mammary cancer cells is an important contributor to hemangiogenesis and lymphangiogenesis.

DIVAA provides a simple and quantitative method to measure hemangiogenesis in vivo (34). In contrast, one of the challenges of studying lymphangiogenesis in vivo is the lack of assays for objective quantification (12). To this end, we developed a similar approach to simultaneously quantify both hemangiogenesis and lymphangiogenesis by combining DIVAA with FACS (DIVAA/FACS).

We initially modified the FACS gating scheme established by Pham and colleagues (23) to separate lymph node blood endothelial cells (BEC) and lymphatic endothelial cells (LEC) from other cells. TER-119 and CD45 were used to gate out red blood cells (RBC) and leukocytes, respectively (Fig. 5A). LECs and BECs were quantified using gp38 (podoplanin) and lymphatic vessel endothelial hyaluronan receptor (LYVE-1), specific markers for LECs, and CD31, a specific marker for both BECs and LECs. LECs were identified as CD45⁻CD31⁺gp38⁺, BECs as CD45⁻CD31⁺gp38⁻, and fibroblast reticular cells (FRC) as CD45⁻CD31⁻gp38⁺. As expected, the selected population of LECs showed a higher expression of LYVE-1 than BECs or FRCs (Fig. 5B).

In the DIVAA approach, mice were implanted subcutaneously with silicone cylinders (angioreactors) that were closed at one end and filled with 20 μL of Matrigel (34). The addition of stimuli enables directional migration of BECs and LECs into the angioreactors, which proliferate and form appropriate vessels. In agreement with previous studies (34), hemangiogenesis was readily observable by the appearance of blood vessels (Fig. 5C). Both S1P and Ang2 significantly induced hemangiogenesis 11 days after implantation of the angioreactors (Fig. 5C). To simultaneously quantify hemangiogenesis and lymphangiogenesis, cells were isolated from the angioreactors by collagenase digestion, followed by FACS analysis to separate the BECs from LECs and other cells as described above. Interestingly, both S1P and Ang2 enhanced the migration of BECs and LECs into the angioreactors, although hemangiogenesis was more pronounced than lymphangiogenesis (Fig. 5D and E). Similar to its effect on in vitro hemangiogenesis and lymphangiogenesis (Fig. 4B), the inclusion of SK1-I in the angioreactors together with Ang2 completely abolished its ability to stimulate both of these processes (Fig. 5E).

To establish that S1P generated by SphK1 plays a critical role in tumor-induced hemangiogenesis and lymphangiogenesis in vivo, we next examined the effect of SK1-I on hemangiogenesis and lymphangiogenesis in the tumors themselves and in the draining lymph nodes. FACS analysis of mammary pad site tumors revealed that both BECs and LECs were greatly reduced by treatment of mice with SK1-I (Fig. 6A and B). Interestingly, the number of BECs and LECs in lymph nodes, the host tumor microenvironment, was significantly increased in tumor-bearing animals compared with sham animals, and treatment with SK1-I inhibited the increase of both BECs and LECs in lymph nodes (Fig. 6C and D). Morphometric analysis confirmed that SK1-I reduced both tumor-induced hemangiogenesis and lymphangiogenesis as determined by densities of CD31⁺ LYVE-1⁻ blood vessels (BVD) and CD31⁺LYVE-1⁺ lymphatic vessels (LVD), respectively (Fig. 6E and F).

Previous studies have shown that SphK1 is overexpressed in breast cancer, and its level of expression correlates with resistance to treatments and poor patient outcomes (16, 18, 20). However, several factors have hindered progress in determining the role of S1P, the product of SphK1, in tumorigenesis and tumor-induced hemangiogenesis and lymphangiogenesis in vivo. First, because S1P has such a profound role in immune function (35), studies of traditional in vivo metastatic breast cancer models with xenografts in immune-compromised mice that ignore the host immune response to cancer may be difficult to interpret. Second, it has not been possible to accurately quantify S1P in tumors until recently with the development of high sensitivity mass spectrometric assays (25). To this end, we utilized our newly established syngeneic breast cancer model, which not only retains the host immune responses but also more closely models progression of the human disease by forming lymph node and lung metastasis leading to metastatic spread and death. Our data indicate that SphK1 is the predominant isoenzyme expressed in 4T1-luc2 cells in culture dish, and is further upregulated in tumors, in contrast to SphK2. This is in agreement with previous reports correlating expression of SphK1, but not of SphK2, with poor prognosis in breast cancer (20). We found that levels of S1P gradually increased both in tumors and in serum determined by LC-ESI-MS/MS and correlated with tumor burden. Moreover, treatment of tumor-bearing animals with the specific SphK1 inhibitor SK1-I reduced S1P levels in the tumor and in circulation, and greatly reduced the size of the primary tumor, lymph node, and lung metastasis. This is in agreement with a previous report that blood and tumor S1P levels were increased in mice with colon cancer (36).

Quantification of tumor-induced hemangiogenesis and lymphangiogenesis has remained a challenge (12). This is usually examined by histologic determinations of microvessel density or LVD, which rely on selective morphometric analysis (e.g., vessel counts, vascular morphology, etc.; ref. 12). The strengths of the morphologic approach are that it can evaluate the location of the vessels in relationship to the tumor and identify/quantify morphologic changes that lymphatic vessels undergo during tumor progression. The limitations include variable sites of tissue sectioning, variable immunostaining techniques, different vessel density quantification methods, and the lack of standardization in the estimation of hemangiogenesis and lymphangiogenesis (12). To compliment this approach and overcome some of these limitations, we developed a flow cytometry method to quantify both BECs and LECs in the same sample, to simultaneously evaluate both hemangiogenesis and lymphangiogenesis. Flow cytometric analysis can provide supportive data to quantify the changes in BECs and LECs.

Consistent with both the in vitro tube formation and DIVAA/FACS assays, we showed that exogenous S1P enhanced hemangiogenesis and lymphangiogenesis. In this regard, Anelli and colleagues showed that downregulation of SphK1 did not change the levels of VEGF-A or VEGF-C secretion in glioma cells, and downregulation of SphK1, but not VEGF-A or VEGF-C, suppressed glioma-induced hemangiogenesis (37). Utilizing shSphK1, we confirmed that tumor-derived S1P mediates tumor-induced hemangiogenesis and lymphangiogenesis without altering VEGF-A or VEGF-C (Supplementary Fig. S3), which is in agreement with Anelli and colleagues (37). Taken together, S1P is an important stimulator of hemangiogenesis and lymphangiogenesis in vitro.

SK1-I decreased hemangiogenesis and lymphangiogenesis not only around the primary tumor but also in lymph nodes that are distant from the tumor. Furthermore, we have observed by F4/80 immunofluorescent staining that SK1-I also reduced macrophage recruitment surrounding the tumor (Supplementary Fig. S4). Macrophages are cellular components of the tumor microenvironment that promote hemangiogenesis and lymphangiogenesis (38, 39). Thus, in addition to direct effects on hemangiogenesis and lymphangiogenesis, S1P can indirectly affect the tumor microenvironment to further enhance these processes.

In sum, S1P generated by SphK1 is important not only for tumor progression but also for tumor-induced hemangiogenesis and lymphangiogenesis and, therefore, targeting SphK1 and its product S1P would be a multipronged attack against breast cancer.

No potential conflicts of interest were disclosed.

The authors thank the Tissue and Data Acquisition and Analysis Core (TDAAC) of VCU for collection of human samples, and the Anatomic Pathology Research Services (APRS) Director, Dr. Jorge A. Almenara, and the histotechnologists for technical assistance with tissue processing, sectioning, and staining. The authors also thank Dr. Bin-Zhi Qian in Jeffrey Pollard laboratory at Albert Einstein College of Medicine, Bronx, NY, for anti-F4/80 antibody.

This work was supported by NCI grant R01CA61774 to S. Spiegel and NIH grant K12HD055881 and Susan G. Komen for the Cure Research Foundation Career Catalyst Research Grant KG090510 to K. Takabe. M. Nagahashi was supported by the SUMITOMO Life Social Welfare Services Foundation.

The Flow Cytometry and Confocal microscopy Shared Resource Cores were supported in part by NIH Grant P30CA16059 to the Massey Cancer Center and NINDS Center core Grant 5P30NS047463, respectively.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.