Sphingosine Kinase 2 in Stromal Fibroblasts Creates a Hospitable Tumor Microenvironment in Breast Cancer

Breast cancer is one of the most common malignant diseases in women. Most breast cancer-related deaths are not caused by the primary tumor itself, but are due to metastasis with a poor survival rate that has not changed much over the past 30 to 40 years. However, the molecular mechanisms underlying breast cancer metastasis have not yet been completely elucidated. The triple-negative breast cancer (TNBC) subtype, characterized by the absence of the estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (HER2), is aggressive with high recurrence, metastasis, and mortality rates. The inability to effectively predict, prevent, and treat metastatic breast cancer is a major problem in breast cancer care, and the development of additional effective targeted therapies is urgently required (1). It is well known that reciprocal interactions between cancer cells and the tumor microenvironment (TME) composed of the extracellular matrix (ECM) and various cell types, including immune cells, fibroblasts, and endothelial cells, are important for disease progression, metastasis, and prognosis (2). The TME continually changes throughout cancer progression, underscoring the need for a better understanding of the cellular and molecular mechanisms governing tumor–stroma interactions that could initiate the development of additional effective treatments to suppress breast cancer progression, metastasis, and recurrence (3).

Sphingosine-1-phosphate (S1P) is a pleiotropic bioactive sphingolipid metabolite that regulates numerous fundamental biological processes important for breast cancer progression and metastasis, including cell growth and survival, motility, invasion, and immune cell trafficking (4). S1P is produced intracellularly from sphingosine, the backbone of sphingolipids, by two closely related sphingosine kinases (SphK), SphK1 and SphK2. It is exported out of cells by specific transporters and exerts many of its effects via the activation of specific cell surface S1P receptors (4). There is abundant evidence linking SphK1 to breast cancer (5, 6). SphK1 expression is elevated in patients with breast cancer and correlates with poor prognosis and chemotherapy resistance (5). Moreover, increased circulating levels of S1P produced by tumor SphK1 in mice prime distant premetastatic niches and are correlated with breast cancer progression and pulmonary metastasis (6). Nevertheless, it has been suggested that cancer cells communicate with the host organism to regulate lung colonization and metastasis via S1P generated systemically by SphK1, rather than by the tumor (7). As nothing is yet known about such a role for SphK2, and previous studies have suggested that increasing circulating S1P leads to increased metastasis, we examined the effect of deletion of endogenous SphK2 on syngeneic breast tumor growth and lung metastasis. Our study uncovered a previously unappreciated role of SphK2 in regulating tumor suppressor p53 in stromal fibroblasts involved in the activation of cancer-associated fibroblasts (CAF). Deletion or inhibition of SphK2 restrained the production of tumor-promoting factors, creating an active anti-TME. Targeting SphK2 in the stromal compartment may lead to the development of new therapeutic strategies that interfere with cancer progression and metastasis.

E0771.LMB (mCherry+) cells were obtained from Robin L. Anderson (University of Melbourne, Melbourne, Australia). E0771.luc cells were kindly provided by Paula D. Bos (Virginia Commonwealth University School of Medicine, Richmond, VA; ref. 8). MMTV-Wnt-1 cells were derived from spontaneous mammary tumors from MMTV-Wnt-1 transgenic mice and were kindly provided by Emily J. Gallagher (School of Medicine at Mount Sinai, New York, NY). Mouse embryonic fibroblasts (MEF) were isolated from E14 wild-type (WT) and SphK2^−/− mouse embryos. Cells were cultured in DMEM (Gibco, 4.5 g/L glucose, # 11960–044) supplemented with 10% FBS and penicillin/streptomycin (pen/strep), sodium pyruvate, and GlutaMAX (Gibco, # 15140122, 11360070, 35050061). Primary human mammary fibroblasts isolated from normal human breast tissue from Cell Biologics (#H-6071) were cultured in complete fibroblast medium (Cell Biologics, #M2267). Primary murine lung fibroblasts were cultured in fibroblast growth medium FM containing 2% FBS, 1% fibroblast growth supplement, and pen/strep (ScienCell, # 2301). Primary murine lung epithelial cells were cultured in epithelial cell medium (EpiCM-a, ScienCell, #4131) containing 2% FBS, 1% epithelial cell growth supplement-animal (EpiCGS-a), and pen/strep. Primary murine lung endothelial cells were cultured in endothelial cell medium growth medium containing 5% FBS, 1.5% endothelial cell growth supplement, and pen/strep (Cell Biologics, # M1168). E0771.LMB and E0771.luc cells and not MMTV-Wnt-1, MEFs, or primary cells were tested for the absence of Mycoplasma contamination by PCR amplification detection kit (Applied Biological Materials, #G238). Experiments with cell lines were performed for up to 20 passages, primary cells were used for up to 5 passages.

All animal experiments were conducted according to protocols approved by the Virginia Commonwealth University Institutional Animal Care and Use Committee. Breeding pairs of WT and Sphk2^−/− mice were on a C57BL/6NJ background and obtained from the Jackson Laboratory. The colonies were maintained in the vivarium at Virginia Commonwealth University. All experiments were performed with littermates from breeding heterozygous mice. Tail snips for genotyping for background strain determination showed that 100% of the genome scan SNP panel tested was matched.

To study tumor progression and lung metastasis, 8 to 9 weeks old female mice were bilaterally implanted in the 4th mammary fat pads with 20 μL of DMEM (glucose- and phenol red-free; Gibco, #A14430–01) containing 0.1×10⁶ E0771.LMB murine breast cancer cells expressing mCherry, or MMTV-Wnt1 or E0771.luc murine breast cancer cells. Palpable tumor size was measured using calipers, and the volume was calculated using the formula V = (W²× L)/2, where V is the tumor volume, W is the tumor width, and L is the tumor length.

In some experiments, E0771.LMB cells were inoculated into the mammary pads, either alone or together, with WT or SphK2 null MEFs in a 1:1 ratio. To determine metastatic pulmonary colonization, 8- to 9-week-old male mice were injected intravenously with 0.5×10⁶ E0771.LMB (mCherry+) cells in 100 μL of Hanks' Balanced Salt Solution (VWR, #02–0121–0500). In some experiments, mice were injected intraperitoneally with the SphK2 inhibitor SLM6031434 (2 mg/kg) or the vehicle every other day. Mice were sacrificed by 5% isoflurane inhalation, blood was collected, the mouse perfused as indicated, tissues were removed for analyses, tumors were excised and weighed, fixed in formalin, and embedded in paraffin or frozen in liquid nitrogen.

Female Rag2^−/−Il2rg^−/− mice (C57BL/6NTac.Cg-Rag2^tm1Fwa Il2rg^tm1Wjl) and C57BL/6NTac control mice were purchased from Taconic Farms. 0.1×10⁶ E0771.LMB murine breast cancer cells in 20 μL of DMEM were bilaterally implanted in the 4th mammary fat pad. When tumors reached a palpable size (approximately 20–30 mm³), usually within 5 to 7 days or 10 to 12 days for Rag2^−/−Il2rg^−/− mice and C57BL/6NTac controls, respectively, the mice received intratumoral injections of SLM6031434 (2 mg/kg) into tumors on the left side and vehicle into tumors on the right side every other day until the humane endpoint. In some experiments, bilateral tumor-bearing C57BL/6NTac mice were injected intraperitoneally with SLM6031434 (2 mg/kg) or vehicle every other day.

Conditioned media (CM) from MEFs or primary fibroblasts was prepared by culturing 0.5×10⁶ cells in 6-well plates to confluency in DMEM containing 10% FBS or in fibroblast growth medium containing 2% FBS and 1% fibroblast growth supplement, respectively. The medium was replaced by DMEM or fibroblast medium, both containing 0.5% FBS, without growth supplements. CM was collected 2 days later and filtered through a 0.2-μm filter. In some experiments, MEFs were treated with 10 μmol/L SLM6031434 (gift from Dr. Kevin Lynch) or 10 μmol/L ABC294640 (#10587, Cayman) for 24 hours, washed extensively 3 times to remove the inhibitor, and CM prepared as above. DMEM containing 0.5% FBS or fibroblast medium containing 0.5% FBS incubated in empty wells for 2 days were used as controls.

E0771.LMB cells were labeled with 5-μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE) dye for 5 minutes in serum-free medium (Invitrogen, CellTrace CFSE). Labeled E0771.LMB cells were cultured in 24-well plates in DMEM containing 10% FBS. MEFs, primary lung fibroblasts, or primary lung endothelial or epithelial cells were plated separately on gelatin-coated 24-well Sarstedt transwell permeable supports (0.4-μm pores) at a density of 50,000 cells/well in DMEM containing 10% FBS, or in fibroblast or epithelial growth medium, respectively. Twenty-four hours later, the transwell permeable supports were placed on top of wells with E0771.LMB cells. Plates were incubated for the indicated days, cells in the lower compartments were then detached, fixed in 3.7% paraformaldehyde, and analyzed by flow cytometry with a BD LSRFortessa-X20. The fluorescence values were analyzed by FlowJo software to calculate the proliferation index. Unlabeled and labeled E0771.LMB cells collected at the start of coculturing were used as controls.

Spheroids were cultured according to the manufacturer´s protocol of the Cultrex 3D Spheroid Fluorometric Proliferation/Viability Assay Kit (Trevigen). Briefly, 4,000 E0771.LMB cancer cells in 45-μL cell culture medium per well were mixed with 5 μL cold 10X spheroid formation ECM. Next, 50 μL of the single-cell suspension in 1X spheroid formation ECM was added to each well of a three-dimensional (3D) culture qualified 96-well spheroid formation plate. The plate was centrifuged at 200 × g for 3 minutes in a swinging bucket rotor and incubated at 37°C in a tissue culture incubator for 72 hours to promote spheroid formation. Then, 50 μL of CM from primary cells or MEFs was added to each well, and the plate was incubated for 5 days. Spheroid viability was measured by the amount of reduced resazurin with a TECAN Infinite M1000 fluorescence plate reader with excitation 544 nm and emission 590 nm or the size was visualized microscopically and quantified by image analysis software. The Cultrex 3D Spheroid BME Cell Invasion Assay Kit (Trevigen) was used for invasion assays. After spheroid formation was established as described above, 50 μL of cell invasion matrix consisting of a basement membrane extract (BME) from the purified from Engelbreth-Holm-Swarm tumor was added to each well. The plate was centrifuged at 200 × g for 3 minutes in a swinging bucket rotor and incubated at 37°C for 1 hour. 100 μL of CM from primary cells or MEFs was added to each well, and the plate was incubated for an additional 5 days. Spheroid invasion was visualized microscopically and quantified by image analysis software.

Data are representative of the results obtained from at least three independent experiments to ensure reproducibility. For the in vivo lung metastasis and primary tumor growth experiments, the number of mice in each group is indicated in the figure legends. Statistical significance was determined using unpaired two-tailed Student t test with Welch´s correction for comparison of two groups unless indicated otherwise, or by ANOVA followed by post hoc tests for multiple comparisons using GraphPad Prism 7.0. For all experiments, the normality of the data from each group was first checked using the Shapiro–Wilk statistical test. For non-normally distributed data, the Mann–Whitney U test was used. Densitometric quantification of immunoblot bands normalized to loading controls was performed using the ImageJ software. The following designations for significance levels are: *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

Additional Materials and Methods used in this study can be found in the Supplementary File S1.

RNA sequencing (RNA-seq) data generated in this study have been deposited in the NCBI Gene Expression Omnibus with the accession code GSE136655.

All data needed to evaluate the conclusions in the paper are presented in the paper and/or Supplementary Materials. Additional data is available upon request from the corresponding author.

High levels of circulating S1P have been linked to increased tumor growth and metastasis (5–7). Unexpectedly, we found that genetic loss of SphK2, which greatly increases circulating levels of S1P in mice (9), markedly decreased, rather than increased, tumor growth in orthotopically implanted E0771.LMB murine TNBC compared with that in WT littermates (Fig. 1A and B). Consistent with the profound effects on tumor size in SphK2^−/− mice, there was a significant decrease in tumor cell proliferation as determined by Ki-67 staining (Fig. 1C). In agreement with previous studies (10), E0771.LMB tumor cells spontaneously metastasized to the lungs of the WT mice. Lung metastatic spread was also greatly reduced in SphK2^−/− mice (Fig. 1D and E). To validate and broaden our findings, other transplantable syngeneic breast cancer models were used. Growth of mammary tumors from implanted E0771.luc or MMTV-Wnt1 cells were significantly decreased in SphK2^−/− mice compared with WT (Supplementary Fig. S1A–S1E), as well as a reduction in lung metastases (Supplementary Fig. S1F). In addition, treatment of WT mice bearing E0771.LMB tumors with the potent isozyme-specific SphK2 inhibitor SLM6031434 (11) greatly reduced tumor growth and lung metastases (Supplementary Fig. S1G–S1J).

Tumor progression is not only dictated by the cancer cells themselves, but is also profoundly influenced by interactions between cancer cells and their microenvironment, which includes stromal and immune cells. The TME shapes the immunologic landscape and plays a vital role in determining whether the primary tumor is eradicated or metastasizes (3). To assess the involvement of SphK2 in the TME, immune cell infiltration into primary tumors was evaluated by flow cytometry (Supplementary Fig. S2A–S2F). Although there were no differences in the overall frequencies of CD45-positive cells, in the TME lacking SphK2, there were significant increases in the percentages of effector T cells (CD3⁺CD4⁺CD44^hiCD62L⁻ and CD3⁺CD8⁺CD44^hiCD62L⁻) and CD4 or CD8 T cells expressing IFNγ, as well as increased natural killer (NK) cells with high expression of IFNγ, a cytokine that promotes antitumor immune responses (Fig. 1F). Conversely, immunosuppressive regulatory T cells (Treg; CD3⁺CD4⁺FoxP3⁺CD25⁺) and polymorphonuclear myeloid-derived suppressor cells (PMN-MDSC, CD45⁺CD11b⁺Ly6G⁺Ly6C^lo), which represent more than 80% of all MDSCs and play a key role in immune suppression in cancer (12), were significantly decreased in SphK2^−/− mice (Fig. 1F). Furthermore, SphK2 deletion reprogrammed tumor-associated macrophages towards an M1-like phenotype (Fig. 1F). These ‘‘classically activated’’ M1 macrophages, with upregulation of MHC-II and CD86, are capable of cancer cell killing (13). Comparable changes in infiltrating immune cells in the lung microenvironment of tumor-bearing SphK2^−/− mice were noted (Supplementary Fig. S3), with a decrease in immunosuppressive populations of Tregs and PMN-MDSCs, increased percentages of CD4 and CD8 T effector cells, and increased NK cells with high expression of IFNγ and M1-like macrophages (Supplementary Fig. S3).

The experimental metastasis assay of tail vein injections of E0771.LMB cells that home into the lungs was used to directly examine the effects of deletion of SphK2 in mice on pulmonary colonization. Lung colonization was significantly decreased in SphK2^−/− mice compared with that in WT littermates (Fig. 2A-C). Similarly, treatment of mice with SLM6031434 to inhibit SphK2 after inoculation of E0771.LMB cells greatly suppressed metastatic colonization, as shown by histologic analysis of metastatic lung nodules and qPCR of the mCherry reporter gene present only in the tumor cells (Fig. 2D and E).

To distinguish the importance of infiltrating immune cells from other microenvironmental factors, bilateral tumors were generated in the mammary fat pads of Rag2^−/−Il2rg^−/− mice lacking functional T, B, and NK cells that developed after a shorter latency than their Rag2^+/+Il2rg^+/+ controls. The left tumors of each mouse were injected with SLM6031434, whereas the right tumors received vehicle, allowing meaningful comparisons of the effects of SphK2 inhibition in the same TME (Fig. 2F). Notably, SLM6031434 local administration induced marked tumor regression only at the injected site and not on the contralateral tumors in Rag2^−/−Il2rg^−/− mice as well as in Rag2^+/+Il2rg^+/+ controls (Fig. 2F). Moreover, systemic intraperitoneal injections of SLM6031434 inhibited tumor growth to a similar extent as did intratumoral delivery (Fig. 2F). Taken together, our results suggest that deletion or inhibition of SphK2 creates a hostile anti-TME that only partially depends on innate and adaptive immune cells.

It was of interest to determine how deletion or inhibition of SphK2 profoundly alters the TME. S1P levels in circulation are increased in SphK2 knockout mice or in mice treated with the SphK2 inhibitor SLM6031434 (9), probably due to reduced clearance by the liver (14). However, in perfused lungs, S1P levels, as expected, decreased with concomitant increases in the levels of its precursor ceramides, particularly C16:0, C20:0, C22:0, and C24:0 species (Fig. 3A). Similar tendencies were observed in SphK2^−/− MEFs and WT MEFs treated with SLM6031434, with a predominant increase in C16-ceramide (Fig. 3B).

We focused our attention on p53 because it was recently shown that C16-ceramide binds p53 and disrupts its interaction with the E3 ubiquitin ligase MDM2, leading to p53 accumulation (15). Second, numerous studies have shown that increased endogenous ceramides can activate p53 (16). Third, strong translocation of p53 to the nucleus was observed in SphK2 null fibroblasts compared with that in WT fibroblasts (Fig. 3C and D). To confirm that increased ceramides have the same effect on p53 as SphK2 deletion, WT MEFs were treated with C16-pyridinium ceramide. Consistent with a previous report (15), this water-soluble C16-ceramide analog significantly induced the translocation of p53 to the nucleus (Fig. 3D). We also took advantage of the known observation that treating cells with cell-permeable D-erythro-C6-ceramide, but not L-erythro-C6-ceramide, caused a significant increase in the levels of endogenous long-chain C16:0-ceramide and, to a lesser extent, C24:1-ceramide, because the generation of endogenous ceramides from C6-ceramide via deacylation/reacylation is highly stereospecific (17). Only D-erythro-C6-ceramide, and not L-erythro-C6-ceramide or S1P, induced strong translocation of p53 to the nucleus in WT MEFs compared with vehicle treatment (Fig. 3D). Moreover, p53 and phospho-p53 (Ser15), which increases its transcriptional activity (18), were also markedly increased in lungs with pulmonary metastatic nodules from SphK2^−/− compared with WT mice (Fig. 3E). Importantly, very little p53 staining was observed in the lungs of tumor-bearing WT mice compared with those of SphK2^−/− mice (Fig. 3F), where the increased staining of p53 was most evident in stromal fibroblasts, but not in the lung epithelium (Fig. 3G). Fibroblasts constitute a significant component of tumor stroma and play important roles in tumorigenesis and metastasis (2, 19).

As a first step in understanding how SphK2 and p53 act in the microenvironment, we compared the global transcriptomes of WT and SphK2 null MEFs by RNA-seq. The expression profile of SphK2 null fibroblasts differed substantially from that of WT, and at 0.05 FDR threshold of 46,983 tested transcripts, 11,582 genes (5,592 upregulated and 5,990 downregulated) were differentially expressed (Fig. 4A and B; Supplementary Dataset Table S1). As expected, numerous top genes that were previously identified in at least 10 genome-wide datasets as p53-activated targets (20), including CDKN1A, GDF15, GADD45A, TIGAR, and FAS, were upregulated (Fig. 4C; Supplementary Dataset Table S2). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes enrichment analyses of the transcriptional profiles identified potential processes affected by SphK2 deletion (Supplementary Dataset Table S3). The most downregulated genes in SphK2 null cells showed significant enrichment of functional pathways related to ECM organization and degradation, and regulation of cell adhesion and motility (Fig. 4D). As previous studies suggested a novel non–cell-autonomous function for p53 in normal fibroblasts to suppress their activation to CAFs, a major contributor in the microenvironment to cancer-associated changes in ECM architecture (21–25), we examined in more detail the effects of SphK2 deletion. The most commonly used genes/markers that identify CAFs, including alpha-smooth muscle actin (αSMA, ACTA2), collagen expression, fibroblast activation protein alpha, and fibroblast-specific protein 1, also known as S100A4, which regulates both cytoskeletal integrity and cell-cycle progression, integrin-alpha 11, and platelet-derived growth factor receptor α/β, were all markedly reduced by the deletion of SphK2 (Fig. 4E), consistent with the decrease in αSMA and type I collagen (Fig. 4F). Of the extensive list of 227 CAF-related genes compiled from literature searches (25, 26), 150 exhibited a statistically significant coordinate reduction in SphK2 null fibroblasts (Supplementary Dataset Table S4), and a hypergeometric test revealed a highly significant reduction in “CAFness” (P = 1.8⁻¹¹). Key ECM-remodeling proteases, matrix metalloproteinases (MMP) that promote motility and invasion of cancer cells (Fig. 4E), and growth factors, including HGF, PDGF, VEGF, FGF, and TGFβ, which are known to be produced and released by CAFs, were downregulated in SphK2 null fibroblasts (Fig. 4E). Moreover, CAF-derived cytokines and chemokines that recruit immunosuppressive cells into the TME to assist in immune evasion (2, 19) were significantly reduced (Fig. 4E).

Next, we examined the expression of these CAF-associated genes in the lungs, as tumor cells are not the predominant cellular subset in the lung microenvironment. Importantly, deletion of SphK2 greatly reduced the expression of CAF-associated genes in the lungs of mice with breast tumors (Fig. 5A) and metastatic nodules (Fig. 5B). As expected, deletion of SphK2 also significantly upregulated p53 target genes in tumors (Fig. 5C), lungs with metastatic nodules (Fig. 5D), and in mice treated with the SphK2 inhibitor (Supplementary Fig. S4A and S4B). Levels of the CAF marker αSMA protein in tumors and lungs with metastases were also markedly reduced by SphK2 deletion (Fig. 5E and F). Thus, consistent with p53 stabilization in SphK2 null stromal fibroblasts, gene expression profiling revealed a strongly decreased signature of CAF-related genes, which are key effectors that fuel cancer progression and promote a protumorigenic microenvironment (27, 28).

Cancer cells promote resident and recruited normal fibroblasts to acquire the features of CAFs (27, 28). In normal breast tissues, smooth muscle cells around the microvasculature are weakly positive for αSMA. However, the activated myofibroblast marker αSMA was highly expressed, particularly in fibroblasts with large spindle cell morphology, which were widely present in the stroma of breast carcinoma from WT mice and nearly absent in SphK2 knockout mice (Fig. 5G). Strong αSMA staining in the lungs of WT mice was also observed in the juxtaposition of pulmonary metastatic nodules, but not in SphK2^−/− mice (Fig. 5H) or in mice treated with the SphK2 inhibitor SLM6031434 (Supplementary Fig. S4C). Together, these results indicate that the lack of SphK2 or its inhibition in stromal fibroblasts suppresses the activation of CAFs in vivo.

Previously, it was suggested that cancer cells secrete factors that suppress p53 activity in fibroblasts and reprogram them into pro-tumorigenic CAFs (29). Indeed, we observed enhanced expression of the CAF activation marker αSMA and SphK2 as well as decreased phosphorylation of p53 on Ser15 in human primary mammary fibroblasts when cocultured with E0771.LMB cells seeded on transwell inserts (Supplementary Fig. S5A and S5B). Moreover, pretreatment of human mammary fibroblasts with the SphK2 inhibitor SLM6031434 increased phosphorylation of p53 and reduced αSMA in this coculture (Supplementary Fig. S5A and S5B).

Similarly, analysis of SphK2 expression in the publicly available gene expression database (GSE41678) revealed that coculturing of two human fibroblast lines (HFFF2 and HFF1, shown to promote tumorigenicity; ref. 30) with MDA-MB-231 TNBC cells increased SphK2 expression in these fibroblasts. In contrast, coculturing MDA-MB-231 cells with two other human fibroblast lines (Wi-38 and CCD1112Sk that were previously shown not to promote tumorigenesis in vivo; ref. 30), did not increase SphK2 expression (Supplementary Fig. S5C). To assess if upregulation of SphK2 in CAFs is relevant to human breast cancer, we analyzed publicly available data sets. SphK2 expression is increased in tumor-associated breast stroma microdissected by laser-capture from 53 patients with invasive ductal carcinoma (Supplementary Fig. S6A; ref. 31). Moreover, SphK2 was significantly overexpressed in 1,097 cases of breast cancer compared with 114 normal samples from TCGA Research Network (https://www.cancer.gov/tcga; Supplementary Fig. S6B). Higher SphK2 expression also correlated with poor prognosis, particularly with higher grades of breast cancer (Supplementary Fig. S6C).

As mentioned above, CAFs secrete MMPs, growth and angiogenic factors, and cytokines that promote tumor growth, invasion, and metastasis (27). They can also produce cytokines/chemokines that influence the recruitment and function of several types of innate and adaptive immune cells to create a more supportive TME (32). Because previous studies have suggested that in cultured mouse fibroblasts, p53 suppresses the expression of some of these protumorigenic factors (23–25, 33), we next examined the involvement of p53 in changes of the CAF-like secretome due to deletion of SphK2. Consistent with the reduction in CAF signature genes (Fig. 5A and B), the expression of MMP2, TGFβ, and IL6, which increased cancer cell proliferation and metastasis and negatively regulated NK and T-cell cytotoxic activity (32), was markedly decreased by deletion of SphK2 (Fig. 6A). Downregulation of p53 in SphK2 deleted fibroblasts with lentivirus expressing p53-specific short hairpin RNA (shRNA) reversed this suppression (Fig. 6A). Likewise, expression of the chemokines CCL2, CXCL1, CXCL12, and CSF1, which recruit MDSCs and influence M1 macrophage polarization, and the potent angiogenic factors VEGF-C and CXCL12, as well as CCL5, which also promote recruitment of Tregs (32), were all significantly reduced by SphK2 deletion, effects that were all negated by the downregulation of p53 (Fig. 6A). These data suggest that repression of the secretome in SphK2 knockout fibroblasts is mediated by p53. Changes in cytokines and chemokines were confirmed by LEGENDplex multiplex immunoassays (Supplementary Fig. S7A). Pretreatment of WT MEFs or primary human mammary fibroblasts with the SphK2 inhibitor also suppressed secretion of these cytokines and chemokines (Supplementary Fig. S7A and S7B). The effect on CXCL12 is particularly interesting, as the CAF secretome is maintained during tumor progression by a CXCL12 autocrine signaling loop (34). Furthermore, it has been shown that CAFs enhance CXCL12-dependent growth of breast cancer cells in vitro and in a mouse model (35), and p53 status in stromal fibroblasts modulates tumor growth in a CXCL12-dependent manner (24).

Numerous growth factors (e.g., IGF, HGF, FGF, PDGF, and SDF1/CXCL12) secreted by CAFs can strongly enhance the growth of the adjacent cancer cells themselves (19), and functional p53 suppresses the production of these tumor-promoting factors (24). To explore the consequences of the curtailed secretome due to the loss of SphK2 and increased p53, we compared the ability of CM from WT MEFs or primary lung fibroblasts to induce the growth and survival of E0771.LMB cells with that of CM from SphK2 null fibroblasts. Colony formation of E0771.LMB cells was greatly reduced by adding CM from SphK2 null MEFs compared with that from WT MEFs (Fig. 6B). Addition of S1P to the CM from SphK2 null MEFs or downregulation of p53 reduced this suppression (Fig. 6B). Similar effects were observed with CM from primary lung fibroblasts isolated from SphK2 knockout mice compared with those from WT lung fibroblasts or with CM from primary human mammary fibroblasts pretreated with SphK2 inhibitors (Fig. 6B). To further confirm this effect, the rate of proliferation of E0771.LMB cells was evaluated by CFSE staining when they were cocultured with fibroblasts in separate compartments of the transwells (Fig. 6C). Indeed, their proliferation was much greater when cocultured with WT MEFs compared with SphK2 null cells, and downregulation of p53 in SphK2 deleted MEFs reversed this suppression (Fig. 6C). Importantly, greater proliferation of these breast cancer cells was also observed when they were cocultured with primary lung fibroblasts isolated from WT mice compared with those isolated from SphK2 knockout mice (Fig. 6C). However, neither primary epithelial cells nor primary endothelial cells from SphK2 knockout mice showed this differential effect on breast cancer cell proliferation compared with their WT counterparts (Fig. 6C).

Next, we examined the 3D in vitro growth and invasion of multicellular tumor spheroids, which mimic aspects of dimensionality, cell–cell contact, and cell–matrix interactions of in vivo solid tumors (36). The growth and viability of E0771.LMB spheroids in 3D culture (Fig. 6D; Supplementary Fig. S8A) and invasion into the surrounding ECM (Fig. 6E; Supplementary Fig. S8B) were significantly enhanced only in the presence of CM from primary WT lung fibroblasts, but not from SphK2^−/− fibroblasts (Fig. 6D and E). Similarly, CM from SphK2^−/− MEFs or from WT MEFs as well as from primary human mammary fibroblasts treated with SphK2 inhibitors repressed spheroid growth or invasion into a matrix composed of basement membrane proteins (Fig. 6D and E; Supplementary Fig. S8A and S8B). Once again, knockdown of p53 or addition of S1P reversed the effects of SphK2 deletion or inhibition (Fig. 6D and E). Taken together, these results indicate that soluble factors secreted from fibroblasts in an SphK2/p53 axis–dependent manner play a role in enhancing the growth of cancer cells and their invasion.

CAFs promote tumor growth and cancer progression (2, 19). To examine the effect of SphK2 loss in fibroblasts with a reduction of the CAF-like secretome on tumor progression in vivo, we used a syngeneic murine breast cancer model in which E0771.LMB breast cancer cells were coinjected into the mouse mammary fat pad with WT or SphK2 null MEFs. Consistent with previous studies (21, 24, 37), WT fibroblasts enhanced tumor growth and increased tumor weight (Fig. 7A and B). In sharp contrast, SphK2 null fibroblasts did not enhance tumor growth or even decrease tumor weight (Fig. 7A and B). Accelerated tumor growth by WT fibroblasts was accompanied by increased proliferation, as determined by Ki-67 staining, whereas SphK2 null MEFs reduced it (Fig. 7C and D). The differential effects of WT and SphK2 null fibroblasts on tumor growth depended, at least in part, on p53, as downregulation of p53 in SphK2 null MEFs completely prevented the negative impact of SphK2 null MEFs on tumor growth (Fig. 7E and F) and correlated with changes in staining of the CAF marker αSMA (Fig. 7G). We speculate that the observed reduction in αSMA staining in the tumor sections from mice inoculated with E0771.LMB together with SphK2 null MEFs compared with those inoculated only with E0771.LMB (Fig. 7G) is because SphK2 deletion suppresses their activation to CAFs, which are the cells in the vicinity of the tumors. Alternatively, SphK2 deleted fibroblasts may also reduce the activation of nearby stromal fibroblasts into CAFs due to the reduction in autocrine signaling loops as we found that in the presence of E0771.LMB cells, coculturing SphK2^−/− MEFs with WT MEFs reduced the overall expression of αSMA (Supplementary Fig. S5A).

Finally, E0771.LMB cells were coinjected with WT or SphK2^−/− MEFs into the left and right mammary fat pads of Rag2^−/−Il2rg^−/− mice. Much larger tumors were formed in Rag2^−/−Il2rg^−/− mice or Rag2^+/+Il2rg^+/+ mice on the side coinjected with WT MEFs compared with those on the opposite side coinjected with SphK2 null MEFs (Fig. 7H), supporting the notion that direct CAF–tumor interaction is an important driver of tumor growth, independent of immune cells (38, 39). Hence, in agreement with our in vitro observations, the reduced expression of secreted protumorigenic factors in SphK2 null stromal fibroblasts resulted in cancer-suppressing effects in vivo to restrict tumor growth. Altogether, our work reveals a surprising role for SphK2 in stromal fibroblasts in creating a hospitable TME.

Bidirectional communication between cancer cells and their microenvironment influences breast cancer progression, metastasis, therapy resistance, and patient outcome. Targeting components of the TME, such as clinical success with cancer immunotherapy using checkpoint blockade (40), has stimulated great interest in understanding how the microenvironment influences tumor progression with the hope of developing novel therapeutic approaches that activate anticancer responses (3).

Here, we report that deletion or inhibition of SphK2 in mice suppressed TNBC progression and lung metastasis caused by reprogramming the TME. We found that the deletion or inhibition of SphK2 decreased S1P and increased ceramides, including C16-ceramide, which increased and activated the tumor suppressor p53 in stromal fibroblasts. In agreement with our studies, treatment of tumor-bearing mice with the SphK2 inhibitor ABC294640 also decreased S1P levels, increased ceramide levels in cells and plasma, and reduced tumor growth (41). Moreover, administration of C6-ceramide induces p53 activation (phosphorylation at Ser15 and increased protein levels) and inhibits xenograft growth (42). Consistent with our preclinical data in mice, a recent study reported that higher levels of ceramides, predominantly C16:0, C22:0, C24:1, and C24:0 species, and reduction in S1P in breast cancer tissue were associated with less aggressive cancer with lower nuclear grade (43). Although ceramide elevation is generally considered a proapoptotic event, different ceramide species in specific subcellular compartments have unique functions (4). It has recently been suggested that C16-ceramide is a natural regulatory ligand of p53 that disrupts its complex with the E3 ligase MDM2, leading to p53 stabilization, nuclear translocation, and activation of its downstream targets (15). Consistent with this finding, deletion or inhibition of SphK2 was also accompanied by increased ceramide and transcriptional activity of p53 in MEFs, as well as in interstitial fibroblasts in lungs with metastases.

While most of the p53 research has been primarily focused on its cell-autonomous functions as a tumor suppressor, there is growing evidence for the importance of its non–cell-autonomous functions in the microenvironment (23–25, 44–46). Consistent with our work, breast tumors containing p53-deficient fibroblasts developed faster and more aggressively than those containing WT p53 fibroblasts (21). Furthermore, deletion of p53 in resident normal fibroblasts augments their acquisition of a CAF phenotype (25) and promotes the expression of CXCL12 (23, 24), which enhances tumor invasion and malignancy. Conversely, stabilization of p53 by small-molecule MDM2 antagonists such as nutlin-3a attenuates the ability of fibroblasts to stimulate breast cancer aggressiveness (46). Indeed, we observed that deletion of SphK2, leading to p53 stabilization in stromal fibroblasts, markedly suppressed their activation to CAFs, which are key players in the cross-talk between tumor cells and their microenvironment (Fig. 8). Growth and angiogenic factors produced by CAFs, including CXCL12, TGFβ, IL6, and VEGF, which directly facilitate tumor growth, and MMPs that initiate remodeling of the ECM and enhance cancer cell migration and invasion, were all greatly attenuated in the absence of SphK2 in a p53-dependent manner. In addition, cytokines and chemokines produced by CAFs that favor the recruitment and differentiation of MDSCs and Tregs and influence M1 macrophage polarization, as well as immunosuppressive factors such as TGFβ, IL6, CXCL12, and FASL, which inhibit the functions of NK cells and CD8⁺ cytotoxic T cells in the TME (32, 47), were also markedly reduced. Therefore, preventing CAF activation by decreasing SphK2 and activating p53 mitigates all these processes, inducing switching to an active anti-TME and reducing tumor progression and metastasis.

Our observations that deletion of SphK2 in stromal fibroblasts results in p53 accumulation and concomitant lower expression of CAF markers shed light on the mechanism by which endogenous SphK2 regulates breast cancer growth, metastasis, and the TME. Importantly, silencing p53 in SphK2‐depleted fibroblasts completely rescued CAF-related markers, the protumorigenic secretome, 3D in vitro growth, invasion of multicellular breast tumor spheroids, and tumor growth in vivo. Thus, we uncovered a novel role of SphK2 in regulating the non–cell-autonomous functions of p53 in stromal fibroblasts involved in their activation to CAFs as a mechanism controlling tumor growth and metastasis. It is now recognized that CAFs in breast cancer are heterogeneous, with distinct subsets (48, 49). Although most CAFs exert tumor-promoting activities, others have tumor-suppressive functions, and the mechanisms underlying these opposing effects are of great interest (38, 39, 50, 51). Intriguingly, deletion or inhibition of SphK2 in stromal fibroblasts markedly reduced tumor-promoting activity in the TME without increasing tumor-suppressive functions.

Our data from in vitro cocultures, 3D multicellular tumor spheroids, and coinjection of SphK2^−/− MEFs with E0771.LMB breast cancer cells into Rag2^−/−Il2rg^−/− mice lacking functional T, B, and NK cells, as well as the activity of an SphK2 inhibitor in these immunodeficient mice, suggest that suppression of tumor growth and invasion by deletion or inhibition of SphK2 may only be partially dependent on innate and adaptive immune cells. Rather, our data suggest that direct fibroblast–tumor interactions are the predominant drivers of tumor growth and are regulated by non–cell-autonomous functions of the SphK2/p53 axis in stromal fibroblasts. Nevertheless, we cannot exclude the possibility that CAFs may exert some effects on metastatic growth by regulating antitumor immunity, in addition to direct interactions with cancer cells. While there has been a focus in the field of CAF-immune cell interactions [reviewed in (47)], our data are consistent with several recent studies demonstrating that CAFs promote tumor growth through direct interactions with and effects on the cancer cells themselves, independent of antitumor immunity (29, 38, 39). In this regard, it has been reported that depletion of all CAFs decreases tumor growth and mortality in desmoplastic colorectal and pancreatic metastasis mouse models. However, there had only minor effects on the composition of immune cells in the liver, except for a decrease in Tregs (38). Moreover, in agreement with our results, ablation of p53 in fibroblasts induced their activation to CAFs and enabled them to efficiently promote prostate and skin tumor growth and squamous carcinogenesis in immunodeficient NOD/SCID mice (24, 25, 29). Consistent with these studies and our data, downregulation of p53 in hTERT-immortalized normal fibroblasts enhanced the growth of H460 lung tumors in SCID mice. However, hTERT-immortalized CAFs derived from the same patient with lung cancer reduced the growth of H460 lung tumors, suggesting that p53 in CAFs may have the opposite effect on its role in normal fibroblasts (37).

Nonetheless, the effect of SphK2 in the microenvironment is independent of the p53 status of the cancer cells themselves as deletion of SphK2 in mice greatly reduced tumor growth of implanted E0771.LMB breast cancer cells that express mutant p53 as well as MMTV-Wnt1 breast cancer cells expressing WT p53. Altogether, our data suggest that repression of the CAF-related secretome explains the strong suppression of tumor progression and metastasis in SphK2^−/− mice or by the administration of a potent SphK2 inhibitor.

Previous studies have shown that SphK2 inhibitors reduce cancer cell growth and suppress tumor growth in preclinical animal models (52, 53). Clinical trials with one inhibitor of SphK2, ABC294640, are in progress (NCT02939807, NCT02229981, and NCT02757326). Our data with a more potent and specific SphK2 inhibitor indicate that inhibiting SphK2 is not only cytotoxic to the tumor itself but also prevents the activation of stromal fibroblasts to CAFs, creating an anti-TME for eradicating cancer cells. Thus, SphK2 inhibitors could be considered as a multipronged treatment modality to bolster clinical effectiveness.

Because CAFs do not harbor p53 mutations (54), CAF-directed therapy to reprogram them back to their normal resting phenotype is beginning to attract attention. However, identifying such an agent remains elusive (50). Our work suggests that the inhibition of SphK2 and p53 activation in stromal fibroblasts is a component of a mechanism restraining their activation to CAFs and driving tumor regression. Consistent with our findings, local p53 activation in the TME by nutlin-3a led to tumor regression and eradication (55). There already is proof of concept of the potential value of engaging stromal p53 in an anticancer response with the development of MDM2/X inhibitors including nutlin-3 now in clinical trials (NCT03654716, NCT03217266, NCT03654716, NCT01877382, NCT03217266; ref. 56). However, only some of these clinical trials stratified patients according to p53 status, and systemic delivery revealed hematological toxicity that is detrimental to immunotherapy (57). Our work suggests a new and different approach to harness the p53 network, which might pave the way for the development of a novel class of therapeutics targeting SphK2 and the microenvironment that interferes with breast cancer progression and metastasis.

S. Spiegel reports grants from NIGMS during the conduct of the study. No disclosures were reported by the other authors.

C. Weigel: Data curation, formal analysis, validation, investigation, visualization, methodology. M.A. Maczis: Data curation, investigation. E.N.D. Palladino: Data curation, investigation, visualization. C.D. Green: Data curation, formal analysis. M. Maceyka: Investigation, visualization. C. Guo: Validation. X.-Y. Wang: Validation, methodology. M.G. Dozmorov: Data curation, methodology. S. Milstien: Supervision, writing–original draft, writing–review and editing. S. Spiegel: Resources, data curation, supervision, funding acquisition, writing–original draft, writing–review and editing.

The authors are grateful to Dr. Webster Santos (Virginia Tech) and Dr. Kevin Lynch for their generous gift of SLM6031434. They thank Dr. Paula Bos for her helpful advice. The authors also thank Bin Hu, Julie Farnsworth, and Jeremy Allegood from the core facilities for their skillful technical assistance. This work was supported by the NIH under grant R01GM043880 (to S. Spiegel). M.A. Maczis was supported by the Ruth L. Kirschstein Individual Predoctoral National Service Award (F31 CA220798). S. Spiegel was also supported by DoD W81XWH2010434. The authors acknowledge the Virginia Commonwealth University Lipidomics/Metabolomics, Cancer Mouse Models, Flow Cytometry, and Microscopy Shared Resources, which are supported in part by funding from the NIH-NCI Cancer Center Support grant P30 CA016059.