Aberrant lipid metabolism reshapes the immune landscape in bone metastasis of nasopharyngeal carcinoma Authors: Yuanyuan Feng, Hongmei Wang, Ziyan Zhu, Wenxiang Deng, Xiaoyue He, Yingying Yao, Miao Song, Xiaohong Peng, Yanling Lin, Longmei Cai Published: 2025 ABSTRACT AbstractBackgroundBone metastasis (BM) drives therapeutic resistance and mortality in nasopharyngeal carcinoma (NPC). Tumor metabolites are crucial for NPC metastasis; however, the mechanisms by which these metabolites synergistically alter the immune microenvironment to promote BM remain unclear.MethodsIn this study, limited immune infiltration was observed in the NPC BM tumor microenvironment. Multiomics analysis has identified sphingosine kinase 1 (SPHK1) as a pivotal mediator driving BM and immune evasion in NPC, orchestrating the production of 1-phosphorylated sphingosine (S1P), which is critical for NPC pathogenesis.ResultsThe aberrant buildup of lipid metabolites, along with immune microenvironment shifts, serves as a critical driver of NPC BM. Mechanistically, S1P enhanced osteoclast recruitment via S1PR3 binding and activated the Hippo pathway, worsening bone colonization and facilitating immune evasion by expanding the exhausted CD8+ T cell population.ConclusionsThe synergy between the SPHK1 inhibitor PF543 and anti-programmed cell death protein 1 therapy amplified treatment effectiveness beyond standalone approaches. Overall, the SPHK1/S1P pathway advances NPC growth and aids in suppressing immune defense. Regulation of lipid metabolism may be a therapeutic target against BM in NPC and may improve the effectiveness of immunotherapy. WHAT IS ALREADY KNOWN ON THIS TOPIC Bone metastasis (BM) drives therapeutic resistance and mortality in nasopharyngeal carcinoma (NPC). Tumor metabolites are crucial for NPC metastasis; however, the mechanisms by which these metabolites synergistically alter the immune microenvironment to promote BM remain unclear. Bone metastasis (BM) drives therapeutic resistance and mortality in nasopharyngeal carcinoma (NPC). Tumor metabolites are crucial for NPC metastasis; however, the mechanisms by which these metabolites synergistically alter the immune microenvironment to promote BM remain unclear. WHAT THIS STUDY ADDS We identify sphingosine kinase 1 (SPHK1) as a pivotal mediator driving BM and immune evasion in NPC, which orchestrates the production of 1-phosphorylated sphingosine (S1P). We uncover a mechanism whereby S1P enhances osteoclast recruitment via S1PR3 binding and activates the Hippo pathway, worsening bone colonization and facilitating immune evasion by expanding the exhausted CD8+ T cell population. Additionally, we demonstrate that the synergy between the SPHK1 inhibitor PF543 and anti-programmed cell death protein 1 (PD-1) therapy amplifies treatment effectiveness beyond standalone approaches. We identify sphingosine kinase 1 (SPHK1) as a pivotal mediator driving BM and immune evasion in NPC, which orchestrates the production of 1-phosphorylated sphingosine (S1P). We uncover a mechanism whereby S1P enhances osteoclast recruitment via S1PR3 binding and activates the Hippo pathway, worsening bone colonization and facilitating immune evasion by expanding the exhausted CD8+ T cell population. Additionally, we demonstrate that the synergy between the SPHK1 inhibitor PF543 and anti-programmed cell death protein 1 (PD-1) therapy amplifies treatment effectiveness beyond standalone approaches. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY This study reveals the SPHK1/S1P pathway as a key regulator of NPC metastasis and immune suppression, providing new mechanistic insights into lipid metabolism-mediated tumor immune microenvironment alteration in BM. The identification of SPHK1 as a therapeutic target and the synergistic effect of PF543 with anti-PD-1 therapy could guide the development of combination strategies to improve immunotherapy efficacy in patients with NPC with BM. Our findings suggest that targeting SPHK1/S1P signaling may represent a promising approach to address therapeutic resistance and mortality associated with BM in NPC, potentially influencing clinical trial design and personalized treatment decisions. This study reveals the SPHK1/S1P pathway as a key regulator of NPC metastasis and immune suppression, providing new mechanistic insights into lipid metabolism-mediated tumor immune microenvironment alteration in BM. The identification of SPHK1 as a therapeutic target and the synergistic effect of PF543 with anti-PD-1 therapy could guide the development of combination strategies to improve immunotherapy efficacy in patients with NPC with BM. Our findings suggest that targeting SPHK1/S1P signaling may represent a promising approach to address therapeutic resistance and mortality associated with BM in NPC, potentially influencing clinical trial design and personalized treatment decisions. Introduction Diverse cancer types metastasize to the bone in 50–80% of patients.13 Patients diagnosed with nasopharyngeal carcinoma (NPC) presenting with bone metastases commonly exhibit unfavorable prognoses. Skeletal-related events, including cancer-induced hypercalcemia, chronic bone discomfort, fractures from minor stress, and spinal cord or nerve pressure, have a heavy toll on patients’ daily lives and mental health, often triggering physical suffering and emotional distress. These complications significantly increase the risk of mortality in individuals with NPC.4 Bone homeostasis depends on the intricate equilibrium between bone formation by osteoblasts and resorption by osteoclasts. Tumor cells secrete several factors that disrupt this equilibrium by promoting osteoclast activity. This progression is facilitated through processes such as an increase in the release of receptor activator of nuclear factor kappa-B ligand (RANK-L) and a decrease in osteoprotegerin generation from osteoblasts. Consequently, when RANK-L attaches to RANK receptors when developing osteoclasts, it triggers their maturation into active bone-resorbing cells, facilitating accelerated bone breakdown.2 5 Immunotherapy is a promising therapeutic approach for various cancers, although few studies have characterized the immune landscape of patients with bone metastasis (BM).68 One defining feature of the tumor immune microenvironment (TIME) in NPC is the higher presence of tumor-infiltrating lymphocytes than that observed in many other solid tumors. Although these lymphocytes display effector markers such as interleukin-2 (IL-2) and interferon-gamma (IFN-γ), their ability to perform effector functions is often compromised or dampened.912 Most studies have concentrated on non-specific organ metastasis, leaving a gap in our understanding of BM in NPC13 which continues to hamper progress in refining treatment protocols. This gap highlights the urgent need to investigate the unique interactions and functional profiles of immune cells within the BM microenvironment to inform better clinical decisions. During cancer progression, metabolic reprogramming orchestrates a complex interplay with the TIME, profoundly affecting tumor metastasis and immune evasion.1416 Through metabolic alterations, tumor cells generate a TIME rich in immunomodulatory metabolites such as lactate, tryptophan metabolites, and arachidonic acid derivatives. These metabolites not only fuel tumor growth but also hinder immune cell function, fostering an immunosuppressive microenvironment characterized by inhibited T cell activation and proliferation, as well as the expansion of regulatory T cells (Tregs).1719 Moreover, metabolic reprogramming extends its influence to diverse cell populations within the TIME, polarizing resident macrophages toward an immunosuppressive M2 phenotype.20 21 This intricate interplay underscores the potential of targeting metabolic reprogramming as a pivotal strategy to dismantle the immunosuppressive microenvironment, augment the efficacy of immunotherapy, and revolutionize cancer treatment outcomes. In this study, using imaging mass cytometry (IMC) and immunohistochemistry (IHC), we determined that NPC tissues with BM exhibit an immunosuppressive microenvironment. Furthermore, by establishing NPC cell lines with a high propensity for BM, we delineated the characteristics of the immunosuppressive microenvironments of these cells. Transcriptomic and lipidomic analyses revealed the dysregulation of lipid metabolism in NPC cells with a high propensity for BM. Among these, sphingosine kinase 1 (SPHK1)/1-phosphorylated sphingosine (S1P), a member of the sphingolipid metabolic pathway, has emerged as a crucial molecule that promotes BM and mediates an immunosuppressive microenvironment. S1P can promote the recruitment of osteoclast precursor (OP) cells by upregulating the expression of S1PR3 on the surface of osteoclasts and acting on the immunosuppressive microenvironment. Additionally, targeting the SPHK1/S1P/S1PR3 axis with drugs can reduce BM and improve the immunosuppressive microenvironment, thereby enhancing immunotherapy efficacy. This is the first study to investigate how altered lipid metabolites affect the immune microenvironment in BM of patients with NPC, suggesting a novel therapeutic strategy for targeted treatment. Results Mapping the NPC tumor microenvironment with IMC To dissect the spatial organization and cellular composition of the TIME of patients with NPC with and without BM, we used tumor specimens to perform highly multiplexed IMC panels (25 protein markers and DNA), which included pancytokeratin+ epithelial, CD31+ endothelial, and viral immune cells (figure 1A,B). The heatmap shows the expression of each marker in the 18-cell clusters, including T cells (CD3+), B cells (CD20+), CD163− macrophages, natural killer cells (CD57+), and Tregs (FOXP3+), using canonical identity markers (online supplemental figure S1a). T and B cells were highly enriched in M0 (primary tumor sites) compared with BM (figure 1C). Comprehensive cell subgroup analysis showed that BM had fewer B, CD8+, and CD4+ T cells than M0, presenting a “cold” tumor immune environment (figure 1D–F). Imaging mass cytometry pipeline based on a 25-marker panel. (A) Flowchart of longitudinal tumor sampling for high-parameter imaging. (B) Single-color staining of 25 markers. (C) Pancytokeratin (green), CD68 (yellow), CD20 (red), and CD8 (gray) were used to represent the structure of NPC tissue (upper image). All cell types (lower images). (D–F) Box plots showing the ratio of B cells (D), CD8+ T cells (E), and CD4+ T cells (F) between M0 (primary tumor sites) and the BM samples (M0, n=14; BM=10). (G) Identification of 10 distinct CN species and their respective abundances (column scaled) within each CN. (H) Bar plots showing the abundance of 10 CNs between the M0 and BM samples (M0, n=14; BM, n=10). BM, bone metastasis; CN, cellular neighborhood; NK, natural killer; NPC, nasopharyngeal carcinoma; PD1, programmed cell death protein 1; PDL1, programmed death-ligand 1; Tem, effector memory T cell; Tex, exhausted T cell; Treg, regulatory T cell. We compared the multicellular interactions between M0 and BM. Two variables affected the cellular neighborhood (CN) assessment: the number of interacting cells within a neighborhood (N) and the number of total CNs. We identified 10 CNs across M0 and BM images, including epithelial-enriched (CN7) or endothelium-enriched (CN1), epithelial–macrophage-enriched (CN2), epithelial-T cell-enriched (CN10), B cell-enriched (CN8), T cell-enriched (CN3), T cell-immunoenriched (CN9), macrophage-enriched (CN5), granulocyte–macrophage enrichment region (CN4), and fibroblast–macrophage-enriched (CN6) (figure 1G). Our analysis of cellular distribution across CN subtypes revealed that bone marrow specimens exhibited stark contrast to M0 tumors, with diminished proportions of B cell-dense clusters (CN8) and T cell-dominated immune hubs (CN9) (figure 1H). Furthermore, the interaction between T cell subsets was more complex, and more immune cells interacted with local T cells in M0 than in BM (online supplemental figure S1b). In summary, the structural makeup of tissues, types of stromal cells, and various immune cell populations in NPC microenvironments observed using IMC revealed a distinct landscape of M0 and BM encompassing epithelial, stromal, and diverse immune cells. Sparse immune cell infiltration in the BM NPC microenvironment A total of 120 NPC tissue specimens were obtained for immunohistochemical examination. There was a contrast in immune cell presence, with BM samples exhibiting substantially diminished infiltration of key defense cells, including CD4+ and CD8+ T lymphocytes, B lymphocytes, and CD68+ macrophages, compared with M0 and M1-others (distant metastases excluding bone) (figure 2A,B). Low immune cell infiltration in the bone metastasis NPC microenvironment. (A, B) Immunohistochemical and statistical analyses of CD4+ T cells, CD8+ T cells, B cells, and CD68+ macrophage infiltration in the NPC tissues of M0 (n=10), M1-others (distant metastases excluding bone) (n=10), and BM (A) (n=10). Scale bar: 50 µm. Quantitative analysis (B). (C) From left to right: BLI of mice, μCT images, TRAP staining, and H&E staining. Scale bar: 100 µm. (D) Quantification of bone parameters, osteoclast numbers, and the osteolytic lesion area. TRAP+ osteoclasts and osteolytic area. (E) Schematic diagram: BM0 and BM3 subcutaneous tumors were generated in mice with a humanized immune system. (F, G) Flow cytometric analysis of the proportions of PD-1+, TIM3+, KI67+, GZMB+, and IFN-γ+ CD8+ T cells in subcutaneous tumors (F) and quantitative analysis (G). Data are presented as mean±SD. Statistical significance was determined with *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. BLI, bioluminescence imaging; BM, bone metastasis; BS/BV, bone surface/bone volume ratio; BV/TV, bone/tissue volume ratio; µCT, micro-CT; IFN-γ, interferon-gamma; NPC, nasopharyngeal carcinoma; PBMCs, peripheral blood mononuclear cells; PD-1, programmed cell death protein 1; Tb, trabecular number; Tb. n, trabecular number; TBPf, trabecular bone pattern factor; Tb. sp, trabecular separation; Th, trabecular thickness; TRAP, tartrate-resistant acid phosphatase. To explore the microenvironment of BM in NPC and the potential mechanisms underlying BM, highly metastatic 5–8F-BM3 cells were established after three rounds of intracardiac (i.c.) injection (online supplemental figure S2a). In the third round, i.c. injection of 5–8F-BM3 cells in mice displayed a high predisposition toward BM for less than 2 weeks, whereas the 5–8F-BM0 group showed a lower incidence of metastasis to the bone after 5 weeks (online supplemental figure S2b) but a longer survival time than the 5–8F-BM3 group (online supplemental figure S2c and d). Migration assays illustrated that 5–8F-BM3 cells significantly enhanced their migration ability after the third round of selection (online supplemental figure S2e and f). Researchers have developed a cardiac tumor model using genetically engineered BM0 and BM3 cell lines to produce a firefly reporter consistently. Bioluminescence imaging (BLI) was used to evaluate cancer growth and dissemination, and micro-CT scans tracked structural shifts in the bone tissue at the microscopic level (figure 2C). Mice treated with BM3 cell injections exhibited not only substantial metastatic tumor growth in bone tissue but also deterioration in bone structure, as evidenced by the increased incidence of osteolytic lesions, widened gaps between trabeculae, altered trabecular patterning, and reduced trabecular density and thickness compared with the BM0 controls. Bone resorption zones and tartrate-resistant acid phosphatase (TRAP)-positive osteoclast counts, visible in histological evaluations, were elevated near the trabecular surfaces in BM3-treated mice, highlighting intensified bone breakdown activity (figure 2C,D). These results suggested that BM3 cells have a high propensity for BM. To investigate the bone metastatic microenvironment in vivo, we generated NPC xenografts using BM0 or BM3 cells from humanized M-NSG mice (figure 2E). We analyzed CD8+ T cells, considering their role as key effector cells in tumor cytotoxicity. The percentage of CD8+ programmed cell death protein 1 (PD-1)+ and TIM3+ cells increased, whereas that of CD8+ Ki67, granzyme B+, and IFN-γ+ cells decreased in the BM3 group (figure 2F,G). IHC also showed a decrease in CD8+ T cells in BM3 tissues (online supplemental figure S2g and h). These results suggest an immunosuppressive microenvironment characterized by decreased effector T cells in BM3 tumor cells. Tumor-derived S1P induces T-cell exhaustion in BM To identify the key mechanisms behind elevated BM, we conducted RNA sequencing of the BM0 and BM3 cell populations. Kyoto Encyclopedia of Genes and Genomes pathway mapping indicated significant differential gene expression enrichment in sphingolipid signaling (online supplemental figure S3a). The upregulated genes identified using the Reactome database revealed that sphingolipid metabolism was the most enriched pathway (online supplemental figure S3b). Based on the RNA-seq results, we performed a lipidomic analysis of BM0 and BM3 cells. Lipid profiling revealed differential upregulation of sphingomyelin (SM), ceramides (CER), hexosylceramides, and lactosylceramides in BM3 cells (online supplemental figure S3c). Gene ontology analysis and tree mapping also showed that the majority of differentially expressed lipid metabolites were involved in glycerophospholipid and sphingolipid metabolism (online supplemental figure S3d and e). We also performed a lipidomic analysis of M0 (non-BM, n=4) and BM (n=4) frozen tissues. Our lipidomic analysis of frozen NPC tissues revealed distinct metabolic differences between M0 and BM groups (online supplemental figure S9a), with the most prominent change being a significant accumulation of CER species in BM tissues (online supplemental figure S9b and c). CER is the immediate precursor of S1P, and their accumulation in BM tissues strongly indicates increased flux through the “ceramide→S1P” metabolic axis, consistent with our prior in vitro findings and confirming conserved sphingolipid dysregulation in NPC BM cells. We measured sphingolipid pathway-metabolizing enzymes and found that these enzymes contributed to CER and sphingomyelin synthases (SGM), such as CER1, CER2, CER5, and SGM2, which were upregulated in BM3 cells (figure 3A). CER produces sphingosine (SPH), which is phosphorylated by SPHK to produce S1P. SPHK1, which is responsible for S1P generation, was significantly upregulated in BM3 cells (figure 3A). In accordance with the changes in metabolic enzymes, the levels of S1P significantly increased in BM3-conditioned media (CM) (figure 3B). S1P is a lipid mediator that is linked to cancer progression and metastasis. To evaluate whether S1P is a critical factor in promoting the BM of NPC, osteoclast concentrations of S1P were treated with different precursors of OPs to test their recruitment, proliferation, and differentiation. S1P recruited OPs in a dose-dependent manner (figure 3C,D) but did not affect OP proliferation (online supplemental figure S4a) or differentiation (online supplemental figure S4b and c). To determine whether S1P facilitated BM, we i.c. injected BM0 cells in the two groups (online supplemental figure S3d). Mice administered intravenous S1P exhibited an elevation in BM burden, with a concurrent expansion of osteolytic lesions and elevated TRAP+ osteoclast counts at the bone-tumor interface (online supplemental figure S3e). Tumor cell-derived S1P promotes bone metastasis and T-cell exhaustion.(A) qRT-PCR analysis of the expression of 17 sphingolipid metabolic enzyme genes. (B) Quantitative analysis of S1P in BM0 and BM3 cell supernatants by ELISA. (C, D) Effects of S1P on recruitment of OP cells at different concentrations (C) (nM refers to nanomoles) and statistical analysis (D). (E) From left to right: BLI, μCT imaging, TRAP staining, and H&E staining of the mice. Scale bar: 100 µm. (F) Schematic representation of S1P treatment of CD8+ T cells. (G) Effect of different concentrations of S1P on the mRNA expression of GZMB, TNFα, TIM3, PD-1, and LAG3 in CD8+ T cells. (H, I) Flow cytometric analysis of the proportion of TNF-α+, PD-1+, and TIM3+ CD8+ T cells after treatment with different concentrations of S1P (H) and statistical analysis (I). Data are presented as mean±SD. Statistical significance was determined with *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. BLI, bioluminescence imaging; BM, bone metastasis; CER, ceramides; µCT, micro-CT; mRNA, messenger RNA; ns, not significant; OPs, osteoclast precursor; PD1, programmed cell death protein 1; qRT-PCR, quantitative reverse transcription-PCR; SGM, sphingomyelin synthase; S1P, 1-phosphorylated sphingosine; SPHK1, sphingosine kinase 1; TNF-α, tumor necrosis factor-alpha; TRAP, tartrate-resistant acid phosphatase. S1P is an immune metabolite that regulates tumor-related immune responses and inflammatory reactions. Therefore, we investigated the in vitro effects of S1P on the immune microenvironment. We extracted CD8+ T cells from the blood samples of healthy individuals exposed to various concentrations of S1P (figure 3F). We observed that although S1P increased the levels of PD-1 and Tim3, it led to a reduction in tumor necrosis factor-alpha (TNF-α) expression (figure 3G). Flow cytometry demonstrated that CD8+ T cells exposed to S1P exhibited higher levels of PD-1 and Tim3 than those in the control group. In contrast, TNF-α concentrations measured in the control group T cells were higher than those in samples treated with S1P (figure 3H,I). Overall, following S1P exposure, CD8+ T cells exhibited diminished cytotoxic activity and upregulated expression of exhaustion-associated markers. SPHK1-dependent S1P mediates antitumor immune suppression in BM. SPHK1 levels were elevated in BM3 cells. IHC analysis revealed higher SPHK1 expression in BM tissues than in M0 and other metastatic sites, indicating a potential link to NPC-related BM (figure 4A,B). To verify this hypothesis, we deleted SPHK1 from BM3 cells using CRISPR/Cas9 gene editing and found that SPHK1 expression and S1P secretion were significantly decreased after SPHK1 knockout (KO) (online supplemental figure S5a and b). Conditioned medium from SPHK1 knockout diminished the recruitment of OPs (online supplemental figure 5c and d). To examine the role of SPHK1 in antitumor immunity, WT, and SPHK1-KO CM were cultured alongside CD8+ T cells for 72 hours (online supplemental figure S5e). Quantitative reverse transcription-PCR (qRT-PCR) analysis revealed that SPHK1 knockout reduced the expression of CD8+ T cell exhaustion markers PD-1 and TIM-3 and increased TNF-α levels (online supplemental figure S5f). Flow cytometry also showed reduced levels of PD-1 and TIM3 in SPHK1 KO T cells treated with CM (online supplemental figure S5g and h). SPHK1 causes bone metastasis and antitumor immune suppression dependent on S1P. (A, B) IHC analysis of SPHK1 expression in NPC tissues from M0 (left), M1 (middle), and BM (right), n=10/group (A), and relative quantitative analysis (B). Scale bar: 100 µm. (C) From left to right: First-week and third-week bioluminescence imaging, TRAP staining, μCT, and H&E staining. Scale bar: 100 µm. (D) Growth trend of subcutaneous tumors in NSG mice (n=5). (E) Statistical graph of weight of subcutaneous tumors. (F) Quantitative analysis of S1P in tumor tissues using ELISA (n=4). (G–J) Flow cytometry analysis of the proportions of PD-1+, TIM3+, and GZMB+ cells among CD8+ T cells in subcutaneous tumors (G) and statistical analysis (H–J). Data are presented as mean±SD. Statistical significance was determined with *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. BM, bone metastasis; µCT, micro-CT; IHC, immunohistochemical; NPC, nasopharyngeal carcinoma; ns, not significant; PD1, programmed cell death protein 1; S1P, 1-phosphorylated sphingosine; SPHK1, sphingosine kinase 1; TRAP, tartrate-resistant acid phosphatase. We used an in vivo experimental model to examine the effects of SPHK1 on the BM. We i.c. injected BM3 WT and BM3 SPHK1-KO1 cells into the mice. Moreover, we used a “rescue” strategy for the tail vein injection of S1P (online supplemental figure S6a). SPHK1 KO effectively reduced the BM burden, osteolytic bone lesions, BM onset, and a number of TRAP+ OPs. However, intravenous S1P reduced the effects of SPHK1 to a certain extent (figure 4C and online supplemental figure S6b). These results reveal that the SPHK1/S1P axis is a key regulator of BM. We used humanized M-NSG mice to assess the impact of SPHK1 on NPC cell growth and microenvironment (online supplemental figure S6d). We observed a tumor reduction in SPHK1 KO mice, whereas intravenous administration of S1P did not affect tumor growth or weight (figure 4D,E). We measured S1P in tissues from a mouse model. Mice that received injections of SPHK1 knockdown cells had significantly lower S1P levels as compared with the control group (figure 4F). Furthermore, serum S1P was significantly elevated in mice with bone metastases compared with non-metastatic tumor-bearing mice (online supplemental figure S6f). SPHK1 knockdown decreased PD-1+ and TIM-3+ CD8+ T cells while elevating GZMB+ CD8+ T cells. Intravenous S1P diminished SPHK1’s impact on the TIME (figure 4G,H). IHC analysis showed that SPHK1 expression was lower in the SPHK1 knockout group (online supplemental figure S6d and e). Collectively, these findings indicate that SPHK1/S1P affects BM development and anticancer immunity. S1P activates the Hippo pathway to promote osteoclast recruitment via S1PR3 In addition to exploring the effect of S1P on the immune microenvironment, we aimed to clarify its role in the bone microenvironment. The extracellular effects of S1P rely on its binding to five unique G-protein-coupled receptors termed S1PRs 1–5. To explore whether S1P promotes OP recruitment, we added S1P to the CM of bone marrow monocytes (BMM) and RAW264.7 cells. PCR assays revealed upregulation of S1PR3 in both cell lines (figure 5A,B). Western blotting also revealed an increase in S1PR3 protein levels after S1P treatment (figure 5C). We selected TY-52156, a specific S1PR3 antagonist, to investigate the effects of the pharmacological S1PR3 blockade on mouse OPs. TY-52156 effectively reduced the recruitment of S1P to mouse OPs (figure 5D,E). To explore this mechanism, we subjected the mouse OPs to RNA-seq. In total, 5230 DEGs were identified (|logFC|≥1, Q≤0.05), comprising 2457 upregulated and 2773 downregulated genes. The volcano plot and heatmap show that the DEG wwrt1 was upregulated after S1P treatment (figure 5F, online supplemental figure S7a). Pathway analysis, specifically gene set enrichment analysis (GSEA), examined the Hippo signaling and cellular migration genes (online supplemental figure S7b and c). Western blotting (figure 5G), immunofluorescence (figure 5H), and quantitative PCR (figure 5I) confirmed that S1P treatment increased YAP1 and TAZ expression. Treatment with the S1PR3 antagonist reduced the expression of YAP1 and TAZ, suggesting that S1P couples with S1PR3 to activate the Hippo signaling pathway (figure 5H,I). S1P activates the Hippo pathway to promote osteoclast recruitment via S1PR3. (A, B) qRT-PCR analysis of the expression changes of S1PR1, S1PR2, S1PR3, S1PR4, and S1PR5 in BMMs (A) and RAW264.7 cells (B) after S1P treatment (μM refers to micromole). (C) Western blotting analysis of S1PR3 protein expression in RAW264.7 cells. (D, E) Transwell assay to evaluate recruitment induced by different treatments on OP cells (D) and quantitative analysis (E). (F) Volcano plot showing the differential gene expression between the control and S1P-treated groups. (G) Western blot analysis of YAP1 and TAZ protein expression in osteoclast precursor cells treated with or without S1P. (H) Immunofluorescence analysis of YAP1 expression in osteoclast precursor cells treated with or without S1P, as well as the influence of the S1PR3 antagonist TY52156. Scale bar: 50 µm.(I) qRT-PCR analysis of YAP1 and TAZ expression in osteoclast precursor cells treated with or without S1P, as well as the influence of the S1PR3 antagonist TY52156. Data are presented as mean±SD. Statistical significance was determined with *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001 BMM,bone marrow monocytes; DAPI,4',6-diamidino-2-phenylindole;mRNA, messenger RNA; ns, not significant; OPs, osteoclast precursor; qRT-PCR, quantitative reverse transcription-PCR; S1P, 1-phosphorylated sphingosine. SPHK1 inhibition combined with PD-1 antibody enhances immune checkpoint therapy efficacy, and inhibition of SPHK1 and S1PR3 blockade reduces NPC BM PF543, a potent inhibitor of SPHK1, was demonstrated to be effective in vivo.2224 To investigate how SPHK1 knockout affects T cell function, we hypothesized that combining PF543, with an anti-PD-1 antibody would enhance the effectiveness of anti-PD-1 therapy by increasing the antitumor activity mediated by T cells. To test this hypothesis, humanized M-NSG mice were created and categorized into four cohorts: (1) a control group, (2) a group treated solely with PF543, (3) a group receiving only the anti-PD-1 antibody, and (4) a combination group treated with both PF543 and the anti-PD-1 antibody. Each group consisted of four mice (figure 6A). The integrated treatment cohort exhibited less neoplasm expansion and a lower mass than the controls (figure 6B–D). Flow cytometry analysis revealed that, in mice treated with PF543 or anti-PD-1 antibodies, the presence of exhausted CD8+ T cells within the tumor microenvironment was higher in their untreated counterparts. The synergistic effect of combining both treatments surpassed the outcomes observed with either therapy alone (figure 6E–J). The treatment group showed lower SPHK1 expression (figure 6K,L) and more CD8+ T cells than those in the untreated group (figure 6M,N). SPHK1 targeting enhances ICI efficacy in NPC cancer models. (A) Schematic diagram: BM3 cells were subcutaneously implanted into humanized mice with reconstituted immune systems to establish tumors. The effects of PF543, an anti-PD-1 antibody, and the combination therapy on tumor growth were assessed. (b–d) Tumor size (B), tumor growth trend (C), and survival curve of tumor weight (D). (E–J) Flow cytometry analysis of the proportion of PD-1+ (E, J), TIM3+ (F, H), GZMB+ (G), and KI67+ (I) CD8+ T cells and statistical analysis. (K–N) IHC shows the expression of SPHK1 (K) and CD8 (M) positive cells in tumor tissue and statistical analysis (L, M). Scale bar: 100/50 µm. Data are presented as mean±SD. Statistical significance was determined with *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. BM, bone metastasis; ICI, immune checkpoint inhibitor; IHC, immunohistochemical; i.p., intraperitoneal; mAb, monoclonal antibody; NPC, nasopharyngeal carcinoma; ns, not significant; PBMCs, peripheral blood mononuclear cells; PD-1, programmed cell death protein 1; SPHK1, sphingosine kinase 1. We investigated the effects of SPHK1 inhibition on BM. We i.c. injected BM3 cells into nude mice and treated them with PF543 (figure 7A). According to BLI and micro-CT (µCT), PF543 effectively reduced bone metastatic burden and osteolytic bone lesions. In the PF543 group, the TRAP+ osteoclast number and the area of osteolytic bone lesions decreased, the onset of BM was prevented (figure 7B,C), and the survival time was prolonged (figure 7D). Conversely, intravenous S1P every 3 days in the PF543 group impaired the effect of PF543. We hypothesized that S1PR3 promotes BM. Furthermore, the selective binding of TY-52156 to S1PR3 diminished the BM burden and osteolytic bone lesions. We observed fewer TRAP+ osteoclasts and osteolytic bone lesions in the TY-52156 group than in the control group. These results suggested that the SPHK1/S1P/SIPR3 axis is a promising target for the treatment of NPC BM. Overall, our data indicated that PF543 combined with anti-PD-1 therapy is a new strategy for treating patients with NPC and BM (online supplemental figure S8). Inhibition of SPHK1 and S1PR3 reduces NPC bone metastasis. (A) Schematic diagram: Nude mice were intracardially injected with BM3 cells to establish a metastasis model for assessing the effects of PF543 and TY52156 on bone metastasis. (B) From left to right: bioluminescence imaging of mice during the first and third weeks, μCT images, TRAP staining, and H&E staining. Scale bar: 100 µm. (C) Quantification of BV/TV, and Tb. n, Tb, Th, Tb. sp, TBPf, BS/BV, osteolytic area; BLI image and day of onset of bone metastasis. (D) Survival curve of mice. Data are presented as mean±SD. Statistical significance was determined with *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. BLI, bioluminescence imaging; BM, bone metastasis; BS/BV, bone surface/bone volume ratio; BV/TV, the bone/tissue volume ratio; µCT, micro-computed tomography; i.p., intraperitoneal; i.v., intravenous; NPC, nasopharyngeal carcinoma; S1P, 1-phosphorylated sphingosine; SPHK1, sphingosine kinase 1; Tb, trabecular number; Tb. n, trabecular number; TBPf, trabecular bone pattern factor; Tb. sp, trabecular separation; Th, trabecular thickness; TRAP, tartrate-resistant acid phosphatase. Discussion Distant metastases are a primary contributor to therapeutic inefficacy and mortality in NPC cases. Preventing BM in NPC and improving treatment efficacy are significant challenges for current therapeutic approaches. Immunotherapy has become a vital strategy for NPC treatment. However, it remains unclear whether a subgroup of patients with NPC and BM would benefit from immunotherapy. In this study, using clinical samples and IMC, we initially confirmed that NPC tissues with BM exhibit characteristics indicative of an immunosuppressive microenvironment. To investigate this, we established a highly BM-prone NPC cell line, BM3, and observed a significant enhancement in the recruitment of osteoclasts along with their capacity to facilitate the formation of an immunosuppressive microenvironment. By integrating transcriptomics and lipidomics, we revealed the accumulation of the sphingolipid metabolite S1P within BM3 cells, prompting an exploration of its role in BM and the immune microenvironment. Using qRT-PCR and western blotting, we found that SPHK1 served as a pivotal regulator of BM and the immunosuppressive microenvironment and that its effect was mediated through S1P. Subsequently, we developed an SPHK1 knockdown cell model and validated its impact on BM and the immune microenvironment through comprehensive in vitro and in vivo experiments. Furthermore, we elucidated the downstream signaling pathways involved in this process. SPHK1/S1P axis upregulates S1PR3 expression to promote BM. Finally, using in vivo experiments, we thoroughly explored the effects of SPHK1 inhibition on BM and the immune microenvironment. The SPHK1/S1P pathway plays a pivotal role in driving cancer growth and spread, orchestrating key mechanisms such as cell multiplication, cell movement, tissue infiltration, attachment to new sites, blood vessel formation, and controlled cell death, which fuel tumor development.2527 Nevertheless, the link between SPHK1/S1P signaling and BM remains poorly understood. Our results revealed that S1P interacts with S1PR3 receptors in mouse OPs, driving their migration and facilitating BM progression by enhancing osteoclast activity. Targeting the SPHK1/S1PR3 axis with drugs can reduce BM, thereby providing a new theoretical basis for studying BM in patients with NPC. We have previously investigated whether S1P affects osteoclast differentiation. In addition to bone cells, marrow-derived components (immune, endothelial, adipocyte, and nerve cells) shape the bone microenvironment and affect tumor progression.28 Tumor cells can engage in diverse, interconnected cellular networks within the bone marrow. S1P triggers downstream signaling cascades by interacting with various S1P receptors (S1PR1–5). IL-22 can induce the expression of S1PR1 on the surface of tumor cells, thereby promoting increased MMP9 secretion by breast cancer cells, enhancing invasiveness, and facilitating metastasis.13 Furthermore, S1PR3 targeting boosts chimeric antigen receptor-T cell antitumor efficacy by preventing T cell exhaustion and mobilizing proinflammatory macrophages to remodel the TIME.29 S1P also mediates T cell migration by binding to S1PR1, which plays crucial roles in T cell maturation, homing, and activation.30 31 The upregulation of S1PR1 expression in newly generated effector T cells enables rapid migration from lymph nodes to effector sites. The S1PR1 agonist FTY720 can downregulate S1PR1 and inhibit the migration of autoreactive T cells from the lymphatic system. It is currently used clinically to treat relapsing multiple sclerosis.32 S1PR1–3 exhibits broad tissue distribution.33 Understanding the function of S1PRs offers a critical framework for creating more selective and effective S1PR-targeting drugs and for treating associated conditions. In our study, S1PR4 was upregulated in RAW264.7 cells but not in BMM. Therefore, in the present study, we focused on S1PR3. In the future, we plan to conduct more comprehensive studies to clarify whether other S1PRs are involved in BM regulation. Osteolytic BM primarily stems from an imbalance in bone homeostasis within the microenvironment, which triggers increased osteoclast activity and suppresses osteoblast function.34 Our results indicate that S1P primarily promotes BM by enhancing the recruitment rather than the differentiation of osteoclasts. S1P stimulates Hippo pathway activation in osteoclasts, which play a key role in breast cancer BM, primarily by regulating the balance of bone metabolism.35 In addition to its individual functions, the Hippo signaling pathway serves as a key regulator of the cellular environment governing bone metabolism, operating through intricate crosstalk with multiple biological pathways. For example, it exhibits a strong functional interplay with the RANKL/RANK signaling axis, which underscores its pivotal role in maintaining skeletal equilibrium.36 The RANK/RANKL interaction is critical for osteolytic lesion formation.37 Ajuba, a crucial player within the Hippo signaling network, interacts directly with TRAF6 to amplify its functional activity. This action triggers a cascade of downstream effects that stimulates osteoclast activation and accelerates bone breakdown, thereby driving the overall process of bone remodeling and metabolic turnover. S1P can enhance RANKL secretion by osteoblasts, potentially activating osteoclasts and promoting bone metabolism.5 In addition to its effect on osteoclasts, Hippo signaling significantly affects osteoblast function. Snail, Slug, YAP, and TAZ interplay regulates bone marrow mesenchymal stem cell self-renewal and differentiation, which are critical for bone growth and formation. These elements can pair with YAP or TAZ to create binary complexes, effectively fine-tuning the expression of essential downstream genes such as alkaline phosphatase, RUNX2, and osterix. This intricate interplay underscores the significance of this pathway in skeletal development.3840 This process promoted osteogenesis. Therefore, targeting the Hippo signaling pathway may be a feasible strategy for treating BM. We plan to conduct a more comprehensive study to explore the specific role of the Hippo pathway in BM regulation in the future. Targeting SPHK1 in melanoma and ovarian cancer can ameliorate immunosuppressive TIME and enhance antitumor responses. Additionally, inhibition of SPHK1 expression with PF543 can achieve similar effects, and when combined with anti-PD-1 antibodies, PF543 can enhance therapeutic efficacy. Through in vitro and in vivo validation, we confirmed that SPHK1 and S1P promote the expansion of exhausted CD8+ T cells. Targeting the SPHK1/S1P axis could improve the immunosuppressive microenvironment mediated by BM in NPC. S1P can upregulate the expression of programmed death-ligand 1 in tumor cells,22 41 thereby reducing the cytotoxicity of T cells and promoting the generation of the cytokines TNF-α and IL-6, thereby promoting tumorigenesis.31 42 Correlation analysis revealed a positive correlation among SPHK1 expression, various immune checkpoints, and immunosuppressive cytokine expression. Furthermore, using previous transcriptomic sequencing data from BM0 and BM3 cells, we conducted GSEA to assess upregulated genes in BM3 cells. This analysis revealed the enrichment of several signaling pathways associated with immune responses (online supplemental figure S7d), accompanied by the upregulation of the immunosuppressive cytokines IL6, CXCL1, CXCL8, INHBA, INHBE, and CCL2 in BM3 cells (online supplemental figure S7e), potentially contributing to immune suppression within the BM microenvironment. BM0 cells were cultured in CM with or without S1P to identify the specific cytokines regulated by S1P, and changes in the RNA expression of inhibitory cytokines were assessed. Treatment with S1P promoted the expression of CCL2 in BM0 cells (online supplemental figure S7f). Subsequently, we examined the messenger RNA expression of the inhibitory cytokines in BM3-WT and SPHK1-KO1 cells. These results indicate a decrease in CCL2 expression following SPHK1 knockdown (online supplemental figure S7g). Additionally, we assessed the levels of CCL2 in BM0, BM3, and BM3-SPHK1-KO1 cells using ELISA. These findings suggested that CCL2 levels were higher in BM3 cells but lower in BM3-SPHK1-KO1 cells than in BM0 cells (online supplemental figure S7h). The SPHK1/S1P pathway is intricately linked to the production of CCL2 within tumors. This chemokine governs the movement and infiltration of monocytes and macrophages. CCL2 drives polarization toward M2 macrophages in the TIME, fostering an immunosuppressive environment.43 Thus, our findings underscore the critical involvement of the SPHK1/S1P pathway in driving immune suppression across diverse cancers and offer evidence to bolster the development of precision therapies aimed at disrupting this biological mechanism. In the future, we plan to expand the sample size to characterize NPC bone metastatic macrophages and explore their relationship with CCL2 expression. These results, taken together, lend credence to the idea that BM plays a role in disrupting metabolic balance. The SPHK1/S1P pathway appears to affect BM and the ability of tumors to evade the immune system. In the future, whether therapeutic approaches that combine SPHK1 inhibitors with immune checkpoint inhibitors could enhance treatment outcomes in patients with NPC and BM should be investigated. This dual strategy might increase patient response rates. Materials and methods Patient samples Cancer samples from patients with NPC were obtained from Southern Medical University, Nanfang Hospital. Detailed patient information is provided in online supplemental file 2 for reference. The biopsy specimens were definitively diagnosed as NPC through pathological examination. These samples subsequently underwent staining and were evaluated to assess the potential associations between their clinical features and relevant patient data from medical records. Informed consent was obtained from all patients, and the study adhered to the Declaration of Helsinki. IMC image preprocessing and cell segmentation pipeline The IMC image preprocessing and cell segmentation pipeline consists of four essential steps: compensation of spillover signals, denoising, contrast enhancement, and cell segmentation. Initially, spillover signals were compensated in each channel using pre-established spillover matrices. For the denoising process, median filtering was applied to reduce noise, where a 3×3 window size was used, and each output pixel was assigned a median value calculated from its 3×3 neighborhood. After denoising, image contrast was enhanced using the “imadjust” function in MATLAB,44 45 which optimized the intensity distribution across the full range of values (0–255), thereby improving image clarity. To perform cell segmentation, a connectivity-aware method, as described in a previous study,46 was used to delineate cell boundaries accurately. Following segmentation, the “regionprops” function in MATLAB is used to identify and label the connected components within the image. For the membrane channels, artifacts were removed if their distance from the nearest nuclear centroid exceeded 15 pixels. All processing steps were carried out in MATLAB R2017a, and the relevant code is publicly available on GitHub at https://github.com/shaoweinuaa/NC2022_Ulcerativecolitis. Cell culture and treatment Professor Mu-Sheng Zeng kindly supplied the 5–8F human NPC cell line at the Sun Yat-sen University Cancer Center (Guangzhou, China). Bone marrow nucleated cells (BMNCs) were harvested from 6-week-old to 8-week-old healthy C57BL/6 wild-type mice. The murine macrophage cell line RAW264.7 was acquired commercially from Procell Biotechnology. All cellular models were maintained under standard culture conditions using either RPMI-1640 or Dulbecco’s Modified Eagle Medium (DMEM) growth media (Gibco), each fortified with 10% heat-inactivated fetal bovine serum (FBS) and an antibiotic cocktail containing 100 U/mL penicillin-streptomycin. The cell culture procedures were conducted under sterile conditions. A highly bone-metastatic NPC cell model was generated by adapting a modified version of a method previously reported in the literature.35 Osteoclast migration and differentiation assay To assess S1P-triggered OP migration, graded S1P concentrations diluted in a control medium were loaded into the lower compartment of a Transwell system, whereas RAW264.7 cells (1×10⁵) were suspended in the upper chamber. The cultures were then incubated for 16–20 hours. For tumor-conditioned medium experiments, TCM (tumor-CM) was generated by concentrating supernatants from 1×10⁶ cultured cells, followed by sterile filtration (0.22 µm). The chemoattractant layer contained TCM blended with fresh medium (3:2 ratio), whereas matching quantities of RAW264.7 cells were plated in the upper inserts. This co-culture configuration underwent an extended incubation period of 28–30 hours prior to the analysis. BMNCs were extracted from fresh murine bone marrow and grown in DMEM supplemented with 10% FBS and antimicrobial agents to trigger osteoclast formation. Once the cells reached 80% confluence, they were seeded into 6-well culture plates and treated with a differentiation cocktail composed of α-MEM, 10% FBS, and 50 ng/mL RANKL. Cell cultures were maintained at 37°C in a humidified incubator with 5% CO₂, and the medium was refreshed every 2–3 days. Following a 6-day to 8-day induction period, osteoclast generation was verified by TRAP staining. CD8+ T cell isolation CD8+ T cells were isolated using a MojoSort Human CD8+T Cell Isolation Kit. Peripheral blood mononuclear cells (PBMCs) were extracted from whole blood via density gradient centrifugation. The isolated PBMCs were incubated with CD8-specific antibodies conjugated to magnetic beads. Following incubation, the cell suspension was filtered through a magnetic column, where magnetically labeled CD8+T cells were retained, and non-target cells were washed away. After washing, magnetically captured CD8+T cells were eluted from the column, resulting in a highly pure CD8+T cell population. The purity and yield of isolated CD8+T cells were evaluated using flow cytometry. Tumor models and treatment All experimental protocols involving animals were reviewed and approved by the Animal Ethics Review Board of the Nanfang Hospital, Southern Medical University. Approval ID: IACUC-LAC-20231017–001. Female BALB/c nude mice aged 3–4 weeks were procured from Guangdong Laboratory Animal Research Center. A suspension containing 8×10⁵ tumor cells was administered via i.c. injection into the left ventricle of each mouse. For the treatment of BM, the mice were treated with vehicle control (Ctrl), PF543 (50 mg/kg via intraperitoneal injection every 3 days), TY52156 (10 mg/kg orally every 3 days), or S1P (3 µg via tail vein injection every 3 days). M-NSG mice were obtained from Shanghai Nan Model Organisms and injected with 5×106 PBMCs through the tail vein to establish a humanized M-NSG mouse model. This model was validated by detecting more than 1% human CD45+ cells in peripheral blood 1-week postinjection. Additionally, 5×106 tumor cells were injected subcutaneously to create a subcutaneous tumor model, followed by treatment with either vehicle control (Ctrl), PF543 (50 mg/kg via intraperitoneal injection every 3 days), or anti-PD-1 (100 µg per mouse on days 7 and 10). Immunohistochemistry The initial processing involved clearing paraffin from the tissue samples through sequential xylene washes, followed by rehydration using increasingly diluted alcohol baths. Heat-mediated antigen unmasking was achieved by boiling the slides for 10 min in 10 mM citrate buffer (pH 6.0) in a pressurized heating chamber. Endogenous peroxidases were neutralized through a 10 min exposure to 3% hydrogen peroxide. The specimens underwent a 60 min pretreatment with a 5% bovine serum albumin (BSA) solution under standard laboratory conditions before primary antibody application to minimize background interference. Tissue sections were incubated overnight at 4°C with primary antibodies suspended in an antibody solution prepared with 1% BSA in phosphate-buffered saline (PBS). Specific antibody types and concentration ratios are provided in the online supplemental Materials. After thorough rinsing with PBS to remove unbound antibodies, the samples were treated for 1 hour at room temperature with species-matched secondary antibodies (eg, horseradish peroxidase (HRP)-conjugated anti-rabbit or anti-mouse IgG). Following multiple PBS washes to eliminate excess reagents, antibody binding was detected by applying a chromogenic 3,3′-diaminobenzidine DAB substrate solution for approximately 5 min before the reactions were halted. Tissue samples were subjected to a final hematoxylin counterstain before being dehydrated and secured onto coverslips using an appropriate mounting solution. For comparison, control samples were generated without the primary antibody incubation phase. All specimens were evaluated using a bright-field microscope, and digital images were obtained and processed using specialized analytical software. TRAP and H&E staining Mouse bone samples were preserved in a 4% paraformaldehyde fixative before undergoing a 2-week decalcification process with 0.5 M EDTA at 4°C; the solution was refreshed every 7 days. Thin 6 µm paraffin slices were stained using H&E or TRAP staining protocols. The prepared slides were analyzed using an Olympus BX63 automated microscope (Tokyo, Japan). Quantitative assessments, including osteoclast numbers, eroded bone surfaces, and tumor size measurements, were performed using ImageJ (National Institutes of Health), following standardized protocols. Throughout specimen processing and data collection, the researchers worked without knowing which samples belonged to specific experimental groups to prevent bias. ELISA The concentration of CCL2 in tumor tissues was quantified using ELISA. Tissue specimens were homogenized in lysis buffer to isolate the proteins. A BCA protein assay kit was used to determine total protein content, and absorbance was measured at 450 nm using a microplate reader to obtain final readings. CCL2 concentrations were calculated using a standard curve (r²>0.99) generated using known CCL2 concentrations. Additionally, S1P levels in the mouse peripheral blood serum were measured using ELISA. Flow cytometry A live/dead dye (BD Biosciences) was used to distinguish viable cells to examine immune cells from tumors. Subsequently, the surface markers were stained using the following antibodies: APC-CY7 CD3 (BioLegend), FITC CD8 (BD Pharmingen), APC CD4 (BioLegend), PE LAG-3 (BioLegend), BV421 TIM3 (BioLegend), and PD-1 (BioLegend). For intracellular cytokine staining, including IFN-γ, TNF-α, and granzyme B, the Leukocyte Activation Cocktail containing BD GolgiPlug was added 4–6 hour prior to staining. Following surface marker labeling, the cells were fixed and permeabilized using the Fix/Perm Buffer Kit (BD Pharmingen) according to the manufacturer’s guidelines. They were subsequently incubated with fluorochrome-conjugated antibodies (PE-Cy7-tagged IFN-γ, BV510-linked Granzyme B, APC-labeled TNF-α, BV650-PD-1, and BV421-TIM3) to identify cytokine expression profiles (BioLegend). For Ki67 intracellular staining, cells were fixed and permeabilized, followed by staining with PERCP-CY5.5 Ki67 (BioLegend). Measurement of cellular lipids For liquid chromatography-mass spectrometry (LC-MS), 107 BM0 or BM3 cells were placed into 3 mL of a mixture of chloroform and methanol [CHCl3:MeOH (2:1)] overnight. Then, the tube was centrifuged at 2700 rpm for 15 min at room temperature, and the lipid extract was transferred into a clean 16×100 mm glass tube. The lipid extract was dried down under a N2 stream on a heater at 60°C and resuspended in 3 mL of the chloroform:methanol mixture before phase separation by addition of 0.6 mL of diluted H2SO4 (0.05%). The sample tube was centrifuged at 2000 rpm for 15 min at room temperature. The top aqueous phase and the middle protein phase were aspirated. The total volume of the bottom phase was recorded and transferred to a new 16×100 mm screw cap tube containing 1 mL of 1% Triton-X-100 in CHCl3. This mixture was dried and then dissolved in 0.5 mL of water for lipid measurements by high-performance liquid chromatography (Biotree, Shanghai, China). Targeted lipidomics S1P in blood was quantified using high-performance LC-tandem MS with external standardization. S1P standard was dissolved in methanol to prepare a 100 µg/mL stock solution, which was serially diluted with methanol-water (80:20, v/v) to working concentrations of 0.1, 0.5, 2, 10, and 50 ng/mL. Standards and internal standards of S1P were purchased from Avanti Polar Lipids, Alabaster, Alabama, USA. A calibration curve was constructed by plotting peak areas against standard concentrations (weighted 1/C², R²≥0.99), and S1P levels in samples were calculated using the external standard method. Statistical analysis Results are presented as mean±SD. Statistical analyses were performed using GraphPad Prism V.8. For studies with three or more experimental groups, one-way analysis of variance (ANOVA) was applied to assess differences, with Dunnett’s post hoc analysis used for comparisons against a shared control group. The Tukey method, or least significant difference test, was used when multiple comparisons involved groups with equal sample sizes. Comparisons between groups were conducted using a two-tailed Student’s t-test with unpaired samples. Bone marrow flux dynamics were evaluated using two-way ANOVA, whereas survival outcomes without bone marrow involvement were visualized using Kaplan-Meier curves. The significance of the differences between survival trends was determined using the log-rank method. Statistical significance was set at p<0.05. Supplementary material