Formononetin suppresses osteosarcoma by targeting MYO1B and remodeling the tumor immune microenvironment

Abstract

Resistance to and associated toxic side effects of neoadjuvant chemotherapy remain major obstacles to improving the prognosis of osteosarcoma patients. Consequently, there is an urgent need to discover effective therapeutic agents with lower toxicity. In this study, the patient-derived xenograft (PDX) model was established and single-cell multi-omics sequencing was performed to comprehensively analyze changes in cellular heterogeneity and gene expression patterns of under formononetin treatment. We found that formononetin can significantly inhibit tumor growth in the osteosarcoma PDX model, on which the single-cell sequencing identified MYO1B as a key target mediating the anti-osteosarcoma effects of formononetin. In vitro experiments demonstrated that MYO1B overexpression enhanced the proliferation, invasion, and migration of osteosarcoma cells, while MYO1B silencing exhibited the opposite effects. Further investigation revealed that formononetin treatment markedly downregulated MYO1B expression, effectively suppressing the proliferative, invasive, and migratory phenotypes of osteosarcoma cells. Moreover, single-cell transcriptomic analysis of murine-derived cells showed that formononetin enhanced the cytotoxic activity of NK cells, promoted M1 macrophage polarization and inhibited M2 polarization, and reduced the proportion of senescent neutrophils, thereby alleviating the immunosuppressive state of the tumor microenvironment. Overall, our findings provide a comprehensive single-cell-level elucidation of the molecular mechanisms underlying the anti-osteosarcoma effects of formononetin, primarily involving downregulating the expression of MYO1B and remodeling the tumor immune microenvironment.

I. INTRODUCTION

Osteosarcoma is the most common primary malignant bone tumor, originating from bone marrow mesenchymal stem cells or osteoprogenitor cells, and is characterized by high malignancy, aggressive behavior, and poor prognosis.^1,2 Clinically, the standard treatment approach remains a combination of neoadjuvant chemotherapy and surgical resection. Cisplatin, doxorubicin, methotrexate, and ifosfamide are the primary agents used in neoadjuvant chemotherapy protocols. However, drug resistance and severe toxic side effects often lead to early termination of treatment and significantly compromise therapeutic outcomes.³ Therefore, the development of novel anti-osteosarcoma agents that are both effective and fewer side effects is an urgent clinical necessity.

Natural compounds are known for their lower toxicity and diverse molecular targets, making them a revolutionary strategy in anti-tumor therapy.^4,5 Formononetin (FMN), belongs to the class of 7-hydroxyisoflavones and characterized by the substitution of a methoxy group at the 4′ position on the 7-hydroxyisoflavone structure, also known as 7-hydroxy-3-(4-methoxyphenyl)-4H-chromen-4-one. Additionally, FMN is one of the primary flavonoid monomers isolated from the root and stem of Caulis spatholobi, exhibiting significant anti-tumor pharmacological activity and making it a promising therapeutic agent.^6,7 Increasing evidence has shown that FMN effectively inhibits the progression of various cancers, including multiple myeloma, hepatocellular carcinoma, and bladder cancer.^8–10 In osteosarcoma, FMN has also demonstrated notable anti-tumor effects.^11,12 Unfortunately, current studies on the anti-osteosarcoma mechanisms of FMN remain limited, with most research confined to in vitro experiments and cell line-derived xenograft (CDX) models, resulting in a lower level of credibility. Therefore, more advanced models and experimental techniques are urgently needed to investigate the anti-osteosarcoma effects of FMN comprehensively.

Currently, the primary in vivo models used to validate drug efficacy are CDX and patient-derived xenograft (PDX). CDX models are established using tumor cell lines that, during extended in vitro culture, may lose certain characteristics of the original tumors, limiting their ability to fully reflect the clinical features of tumors. In contrast, PDX models are constructed using tumor tissues directly obtained from patients, preserving the genetic mutations, tissue architecture, and tumor microenvironment (TME) of the primary tumor (PT). Consequently, PDX models demonstrate a higher consistency with clinical outcomes in terms of drug response and mechanism of action, making them a superior predictor of patient responses to therapeutics.^13–15

In recent years, single-cell RNA sequencing (scRNA-seq) has been widely applied in tumor. By precisely characterizing cellular subpopulations and spatial structures, scRNA-seq enables a deep analysis of cellular composition, regulation, evolution, and interactions, offering new insights into tumor heterogeneity, malignancy progression, and therapeutic responses.^16,17 Notably, scRNA-seq has been successfully utilized in PDX models to explore the therapeutic effects and underlying mechanisms of anti-tumor agents.^18,19 Similar to scRNA-seq, single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq), which profiles transposase-accessible chromatin, allows for high-resolution analysis of cellular composition and heterogeneity while providing additional insights into the chromatin accessibility landscape.²⁰ To date, there are limited studies utilizing PDX models in osteosarcoma research, and no reported studies have combined PDX models with single-cell omics technologies to investigate the anti-osteosarcoma effects and mechanisms of FMN.

In this study, we successfully established an osteosarcoma PDX model to evaluate the therapeutic efficacy of FMN. Furthermore, we employed scRNA-seq and scATAC-seq technologies to comprehensively elucidate the potential molecular mechanisms of FMN, as well as its impact on remodeling the tumor immune microenvironment (TIME). Our findings provide valuable insights that support the clinical translational application of FMN in osteosarcoma treatment.

II. RESULTS

A. FMN inhibited tumor growth in osteosarcoma PDX model

To evaluate the anti-osteosarcoma effect of FMN, we successfully established osteosarcoma PDX models and then treated with FMN, as outlined in Fig. 1(a). The results demonstrated that the tumor volume and weight were significantly reduced in the FMN group compared to the control group [Figs. 1(b)–1(e)]. H&E staining of liver sections revealed no evident pathological lesions in the FMN group with the control group as reference [Fig. 1(f)]. These findings suggested that FMN effectively suppressed osteosarcoma growth in vivo without causing significant hepatotoxicity.

FIG. 1.

FMN suppressed tumor growth in osteosarcoma PDX models. (a) Schematic diagram of PDX model establishment and intervention procedures. (b) and (c) General view of PDX models and excised tumor tissues. (d) and (e) Statistical analysis of tumor volume and wet weight. (f) H&E staining of mouse liver tissues.

B. Identification of FMN-responsive cell subpopulations

In order to investigate the potential mechanisms underlying the anti-osteosarcoma effects of FMN, we performed single-cell multiomics sequencing on samples from FMN-treated PDX models. For enhancing the robustness and accuracy of our analysis, we integrated data from the GSE237070 osteosarcoma dataset. After quality control, dimensionality reduction, and clustering, a total of 46 075 cells were identified and classified into nine major cellular subpopulations: myeloid cells, osteoblastic cells, tumor-infiltrating lymphocytes (TILs), fibroblasts, osteoclasts (OCs), endothelial cells (ECs), plasma cells, B cells (BCs), and mast cells, as shown in Fig. 2(a). These subpopulations were annotated based on the expression of canonical marker genes [Fig. 2(b)]. Among all identified cell types, osteoblastic cells constituted the largest proportion, accounting for 22.58% of the total cells [Fig. 2(c)]. To distinguish malignant osteoblastic cells from nonmalignant counterparts derived from PT and paracancerous (PC), we employed the “inferCNV” algorithm using B cells, plasma cells, and endothelial cells as reference cell types. The analysis revealed significant copy number variations (CNVs) in osteoblastic cells from osteosarcoma samples [Fig. 2(d)], with elevated CNV scores further supporting their malignant nature [Fig. 2(e)], indicating these cells as osteosarcoma cells. Next, to identify the cellular subpopulations potentially targeted by FMN, we analyzed the single-cell sequencing data from the FMN-treated group. A total of 13 408 human-derived PDX cells were clustered into four major subpopulations using unsupervised clustering [Fig. 2(f)]. To confirm the malignant status of these human-derived cells, we performed an inferCNV analysis using plasma cells, B cells, and endothelial cells from the GSE237070 dataset as reference cells. The analysis indicated significant CNVs abnormalities in the PDX human-derived cells, suggesting their malignant nature [Fig. 2(g)]. Gene set variation analysis (GSVA) indicated that the C0 subpopulation was associated with inflammatory response regulation and angiogenesis, the C1 subpopulation was linked to tumor initiation and metastatic phenotypes, and the C2 subpopulation was related to tumor proliferation and epithelial–mesenchymal transition pathways, while the C3 subpopulation played a key role in maintaining cell shape and signal transduction [Fig. 2(h)]. To identify FMN-sensitive osteosarcoma cell subpopulations, we quantified the proportional distribution of each subpopulation before and after FMN treatment. Notably, the proportion of the C1 subpopulation significantly decreased after FMN intervention [Fig. 2(i)]. Combined with the GSVA results, we identified the C1 subpopulation as the specific osteosarcoma cell cluster targeted by FMN.

FIG. 2.

Identification of cellular subpopulations targeted by FMN. (a) scRNA-seq data from PC and PT clinical samples of osteosarcoma identified nine major cell types. (b) Dot plot showed marker genes for each cell type. (c) Osteoblastic cells were the most abundant cell type. (d) and (e) Heatmap and boxplot from inferCNV analysis revealed significant CNVs in osteoblastic cells derived from PT, identifying them as malignant cells. (f) Human-derived cells in PDX tumors clustered into four major subpopulations. (g) Heatmap from inferCNV analysis demonstrated substantial CNVs among the human-derived subpopulations, confirming their malignancy as osteosarcoma cells. (h) GSVA illustrated functional enrichment of osteosarcoma subpopulations. (i) FMN treatment significantly reduced the proportion of C1 osteosarcoma cells.

C. Identification of FMN targets

For identifying potential targets of FMN, we first performed differential expression analysis between osteosarcoma cells and normal osteoblastic subpopulations in the GSE237070 dataset. A total of 1885 upregulated genes and 966 downregulated genes were identified in osteosarcoma cells [Fig. 3(a)]. Next, we conducted differential expression analysis within the C1 subpopulation, revealing 285 upregulated genes and 151 downregulated genes in the FMN group [Fig. 3(b)]. By intersecting the 1885 upregulated genes with the 151 downregulated genes, we identified 46 candidate genes as potential targets of FMN [Fig. 3(c)]. To further refine the core target genes, we applied three classical machine learning algorithms—LASSO, SVM-RFE, and Random Forest (RF-RFE)—using osteosarcoma cohort data from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database. This analysis yielded three intersecting genes: MYO1B, UBE2S, and MSMO1 [Figs. 3(d)–3(j)]. Kaplan–Meier survival analysis indicated that higher expression levels of MYO1B and UBE2S were significantly associated with lower overall survival rates in osteosarcoma patients, identifying these two genes as key candidate targets [Figs. 3(k) and S1(a)]. Integrative analysis of RNA sequencing data from the osteosarcoma TARGET cohort and normal samples from the GTEx database revealed that both MYO1B and UBE2S were upregulated in osteosarcoma tissues compared to normal tissues [Figs. 3(l) and S1(b)], and scRNA-seq analysis indicated that MYO1B and UBE2S expression was significantly higher in the osteoblastic C1 subgroup than in other subgroups [Figs. 3(m) and S1(c)]. Based on these findings, we speculate that MYO1B and UBE2S may function as oncogenes in osteosarcoma, contributing to its malignant progression and poor prognosis. These genes were likely critical targets through which FMN exerted its anti-osteosarcoma effects.

FIG. 3.

MYO1B as a potential target of FMN. (a) Volcano plot displaying differentially expressed genes (DEGs) in osteoblastic cells from PC and PT. (b) Volcano plot showed DEGs in human-derived cells from the PDX models. (c) Venn diagram of overlapping DEGs, with “PC-PT up” indicating genes upregulated in osteoblastic cells from PT and “Con-FMN down” denoting genes downregulated after FMN treatment. (d) and (e) Feature gene selection using the LASSO algorithm. (f) and (g) Feature gene selection via SVM-RFE algorithm. (h) and (i) Feature gene selection with the random forest algorithm. (j) Venn diagram identifying intersecting genes from the three algorithms, including MYO1B, UBE2S, and MSMO1. (k) Kaplan–Meier survival analysis reveals that osteosarcoma patients with high MYO1B expression exhibit lower overall survival. (l) MYO1B is significantly overexpressed in osteosarcoma tissues compared to normal musculoskeletal samples. (m) MYO1B is highly expressed in the C1 subgroup.

D. FMN inhibited proliferation, migration, and invasion of osteosarcoma cells via downregulation of MYO1B

Next, we assessed the expression of MYO1B and UBE2S in the human osteoblast cell line hFOB1.19 and five osteosarcoma cell lines (143B, HOS, U2OS, SAOS2, and MG63) using quantitative reverse transcription polymerase chain reaction (qRT-PCR). Compared to hFOB1.19, MYO1B mRNA expression was significantly elevated in HOS and MG63 [Fig. 4(a)]. In contrast, UBE2S was only upregulated in the 143B [Fig. 4(b)]. Based on these findings, we selected MYO1B as the key candidate target for further investigation and chose HOS and MG63 as the in vitro models for subsequent experiments. To confirm the treatment concentration of FMN on osteosarcoma cells, CCK-8 assays were used to determine the half-maximal inhibitory concentration (IC50) of FMN, of which the results were 420.3 μM in HOS cells and 410.6 μM in MG63 cells [Fig. S1(d)]. To confirm the regulatory effect of FMN on MYO1B expression, HOS and MG63 cells were treated with increasing concentrations of FMN (0, 10, 20, 40, 80, 160 μM) for 48 h. CCK-8 assays indicated that cell viability decreased in a dose-dependent manner with increasing FMN concentrations [Fig. S1(e)]. Based on these results and our previous study,²¹ we selected FMN concentrations of 80 and 160 μM for subsequent experiments. qRT-PCR and immunofluorescence revealed that MYO1B mRNA and protein levels were significantly reduced in a dose-dependent manner following FMN treatment, suggesting that FMN effectively inhibited MYO1B expression [Figs. 4(c)–4(g)]. Additionally, EdU and flow cytometry apoptosis assays demonstrated that FMN treatment significantly reduced the proliferation capacity and increased the apoptosis rate of HOS and MG63 cells compared to the control group, with a clear dose-dependent effect [Figs. 4(h)–4(m)]. These findings indicated that FMN significantly inhibited proliferation and promoted apoptosis of osteosarcoma cell, potentially through suppressing the MYO1B expression. Moreover, transwell assays showed that FMN significantly inhibited the migration and invasion abilities of osteosarcoma cells in a dose-dependent manner [Figs. S1(f)–S1(i)].

FIG. 4.

FMN and MYO1B modulated osteosarcoma cell proliferation and apoptosis. (a) and (b) MYO1B was highly expressed in multiple human osteosarcoma cell lines, while UBE2S was prominently expressed only in 143B cells. (c) and (d) FMN treatment downregulated MYO1B expression in a dose-dependent manner. (e)–(g) Immunofluorescence staining showed FMN inhibited MYO1B expression in HOS and MG63 cells in a dose-dependent manner. (h)–(j) EdU assays demonstrated FMN significantly suppressed HOS and MG63 cell proliferation in a dose-dependent manner. (k)–(m) Flow cytometry analysis showed FMN induced apoptosis in HOS and MG63 cells in a dose-dependent manner.

Furthermore, to elucidate the biological function of MYO1B in osteosarcoma, we designed three siRNAs targeting MYO1B. After transfection into HOS and MG63 for 48 h, qRT-PCR was performed to assess the silencing efficiency. The results showed a significant reduction of MYO1B expression in both cell lines, with siRNA-MYO1B-1 achieving the highest silencing efficiency [Figs. 5(a) and 5(b)]. Therefore, siRNA-MYO1B-1 was selected for subsequent experiments. The overexpression efficiency of the oe-MYO1B plasmid was also confirmed [Figs. 5(c) and 5(d)]. Using these transfected cells, we explored the regulatory role of MYO1B in proliferation, migration, and invasion of osteosarcoma cell. EdU assays indicated that MYO1B silencing led to a significant decrease in proliferation, while MYO1B overexpression resulted in the opposite effects [Figs. 5(e)–5(g)]. Additionally, transwell assays demonstrated a positive correlation between MYO1B expression levels and the proliferation, migration, and invasion abilities of HOS and MG63 [Figs. S1(i)–S1(l)]. In summary, these findings confirmed that MYO1B played a key role in osteosarcoma progression by mediating the proliferative, migratory, and invasive phenotypes of osteosarcoma cells.

FIG. 5.

FMN inhibited osteosarcoma cell proliferation by downregulating MYO1B expression. (a) and (b) Validation of MYO1B silencing efficiency via qRT-PCR. (c) and (d) Verification of MYO1B overexpression via qRT-PCR. (e)–(g) EdU assays revealed MYO1B silencing significantly inhibited tumor cell proliferation. (h) FMN significantly regulated MYO1B expression as shown by qRT-PCR rescue experiments. (i) and (j) Immunohistochemistry demonstrated FMN inhibited MYO1B expression in vivo. (k)–(m) EdU rescue assays showed FMN inhibited osteosarcoma cell proliferation through MYO1B downregulation.

It remains unclear whether FMN exerts its anti-osteosarcoma effects through the regulation of MYO1B expression. We performed a rescue experiment by transfecting MYO1B overexpression plasmids into HOS and MG63 osteosarcoma cells, followed by treatment with FMN. qRT-PCR results showed that MYO1B mRNA expression was significantly upregulated in the oe-MYO1B group, significantly downregulated in the FMN group, and was intermediate in the oe-MYO1B + FMN group [Fig. 5(h)]. Additionally, immunohistochemistry (IHC) analysis indicated that FMN significantly reduced MYO1B expression in vivo, further confirming the down-regulatory effect of FMN on MYO1B expression [Figs. 5(i) and 5(j)]. We then assessed the impact of FMN on the proliferative, migratory, and invasive phenotypes of osteosarcoma cells across the same experimental groups. EdU proliferation assays revealed that the oe-MYO1B group exhibited significantly enhanced proliferation compared to the control group, whereas the FMN group showed a marked reduction in proliferation. Notably, when MYO1B overexpression was combined with FMN treatment, the proliferative capacity was significantly suppressed [Figs. 5(k)–5(m)]. Similar trends were observed in the migration and invasion assays, where MYO1B overexpression enhanced migration and invasion, while FMN treatment significantly inhibited these phenotypes [Figs. S2(a)–S2(d)]. In summary, these results demonstrated that FMN inhibited the proliferation, migration and invasion of osteosarcoma cells by downregulating MYO1B expression.

E. ScATAC-seq analysis identified FOXP1 as a potential transcription factor of MYO1B

Transcription factors (TFs) play critical roles in regulating tumor proliferation, apoptosis, evasion, and metastasis. To identify upstream transcription factors of MYO1B, we integrated scATAC-seq and scRNA-seq data from the PDX models, annotating four distinct cell types and determining the sample origin for each cell type [Figs. S3(a) and S3(b)]. Analysis of MYO1B chromatin accessibility revealed a significant reduction in open peaks of chromatin at the MYO1B locus in the FMN group compared to the control group [Figs. S3(c) and S3(d)]. This indicates decreased chromatin accessibility and suggests that MYO1B may be less accessible to transcriptional regulation under FMN treatment, further supporting the down-regulatory effect of FMN on MYO1B expression. Next, we performed transcription factor motif analysis based on genes with differential open peaks of chromatin. The analysis identified FOXP1 as a potential specific transcription factor for MYO1B. We also identified potential motif sequences in the open region of chromatin of the MYO1B [Figs. S3(e) and S3(f)]. To verified the regulatory relation of the FOXP1/MYO1B axis, we transfected cells with a FOXP1-specific silencing plasmid. Subsequent PCR and immunofluorescence analyses revealed a significant downregulation of MYO1B expression at both the mRNA and protein levels following FOXP1 knockdown [Figs. S3(g)–S3(i)]. These findings suggested that FOXP1 may played a key role in the regulation of MYO1B, potentially mediating the anti-osteosarcoma effects of FMN by targeting MYO1B expression.

F. Impact of FMN on the immune landscape of mouse-derived cells and modulation of key immune subpopulations

We analyzed mouse-derived cells to investigate how FMN influences the remodeling of the osteosarcoma immune microenvironment. A total of 42 676 mouse-derived cells were included for analysis, with 27 604 cells (64.7%) from the control group and 15 072 cells (35.3%) from the FMN group. These cells were classified into 12 subclusters [Fig. 6(a)] and annotated based on the specific high-expression genes and canonical marker genes [Fig. 6(b)]. The t-SNE (t-Distributed Stochastic Neighbor Embedding) plot revealed seven major cell types: monocytes/macrophages (Mono/Mac), fibroblasts, tumor-infiltrating lymphocytes (TILs), endothelial cells, neutrophils, dendritic cells (DCs), and B cells [Fig. 6(c)]. Correlation analysis among cell subtypes indicated a high degree of correlation among immune cells (monocytes/macrophages, neutrophils, TILs, and B cells), as well as strong correlations among stromal cells (endothelial cells and fibroblasts) [Fig. S2(j)]. To characterize the effects of FMN on the tumor immune microenvironment (TIME), we quantified the proportions of each cell type before and after FMN intervention. Notably, the proportion of monocytes/macrophages decreased following FMN treatment, while the proportions of TILs, neutrophils, fibroblasts, and endothelial cells increased [Fig. 6(d)]. These changes in the proportions of non-tumor cell subpopulations suggested an initial indication of the remodeling effects of FMN on TIME.

FIG. 6.

FMN reshaped the TIME and macrophage heterogeneity. (a) Mouse-derived cells clustered into 12 distinct cell groups. (b) Bubble plot showed marker genes for mouse-derived cell clusters. (c) Annotation of seven major cell types based on marker genes. (d) Bar chart displayed changes in the proportions of different cell types after FMN intervention. (e) Reclustering of monocytes/macrophages identified six subtypes, including one monocyte and five macrophage subtypes. (f) Heatmap showed marker genes of monocyte/macrophage subtypes. (g) Heatmap depicted M1 and M2 polarization features of macrophage subtypes.

Given the high proportion of monocytes/macrophages among mouse-derived cells, we concentrated our analysis on these populations. Traditionally, macrophages are classified into two polarization states: tumor-suppressive M1 macrophages and tumor-promoting M2 macrophages.²² Based on canonical marker genes, we identified six subclusters: one monocyte subtype and five macrophage subtypes (TAM_Il1b, TAM_Trem2, TAM_Proli, TAM_Isg15, and TAM_Lyve1), with initial annotations indicating their polarization phenotypes [Figs. 6(e) and 6(f)]. To further characterize macrophage polarization, we utilized gene signatures from CIBERSORT. The analysis suggested that TAM_Il1b and TAM_Isg15 subclusters exhibit M1 phenotypes, while TAM_Trem2 and TAM_Lyve1 subclusters display M2 characteristics [Fig. 6(g)]. Moreover, GSVA and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses indicated that the functional states and polarization phenotypes of these macrophage subgroups play consistent roles in influencing tumor progression [Figs. 7(a) and 7(b)]. M1 macrophages are typically activated by stimuli such as interferon-γ and release inflammatory mediators like IL-1, promoting an inflammatory response that inhibits tumor growth. In contrast, M2 macrophages secrete anti-inflammatory cytokines, support angiogenesis, maintain cancer stemness, remodel tissue, and induce drug resistance, thereby activating Th2-type immune responses and facilitating tumor progression.^23–25 In theory, drug-responsive macrophages tend to shift toward an M1 phenotype, while M2 polarization is often suppressed. To clarify the impact of FMN on macrophage polarization, we quantified the proportion changes before and after treatment. Compared to the control group, the proportion of the M1-like TAM_Il1b significantly increased in the FMN group, while the proportion of the M2-like TAM_Lyve1 showed a decreasing trend, along with a reduction in the TAM_Proli [Fig. 7(c)]. Furthermore, M1 and M2 feature scores were assessed using marker genes and the AddModuleScore algorithm. The results indicated an increase in M1 scores and a decrease in M2 scores following FMN intervention [Figs. 7(d) and 7(e)]. To further investigate the effect of FMN on macrophage polarization, we first induced an M0 macrophage model using THP-1 cells. Microscopic observation showed that M0 macrophages adhered and extended pseudopodia [Fig. 7(f)]. Flow cytometry detection of CD11b and CD14 showed that both were elevated in the M0 macrophage group, indicating successful induction [Figs. 7(g) and 7(h)]. Enzyme-linked immunosorbent assay (ELISA) was used to detect the levels of cytokine TNF-α (mainly secreted by M1) and cytokine IL-10 (mainly secreted by M2). M0 macrophages were polarized to M1 using LPS + IFN-γ and to M2 using IL-4 + IL-13, with control, induced, and FMN intervention groups established accordingly. The results showed that FMN promoted M1 polarization and inhibited M2 polarization [Fig. 7(i)]. These findings suggest that FMN may promote the polarization of macrophages toward M1 phenotype while inhibiting their M2 polarization. In addition, monocle trajectory analysis revealed differentiation paths of the TAM_Proli toward both the M1-characteristic TAM_Il1b and the M2-characteristic TAM_Trem2 [Fig. 7(j)]. CellphoneDB analysis also indicated high levels of cell–cell interactions between TAMs and osteosarcoma cells, suggesting a potential regulatory role in tumor progression [Fig. 7(k)].

FIG. 7.

FMN promotes macrophage polarization. (a) GSVA enrichment analysis of macrophage subtypes. (b) KEGG pathway enrichment analysis of macrophage subtypes. (c) Bar chart showed changes in the proportions of macrophage subtypes after FMN treatment. (d) and (e) Polarization scores of M1 and M2 macrophages following FMN intervention. (f) The microscopic morphology of THP-1 and M0 macrophages was examined. Adherence to the culture surface and pseudopod extension were observed in M0 macrophages, confirming the successful establishment of the macrophage model. (g) and (h) Flow cytometry analysis revealed increased expression of CD11b and CD14 in M0 macrophages, further verifying the successful construction of the macrophage model. (i) The secretion levels of TNF-α (primarily produced by M1 macrophages) and IL-10 (mainly secreted by M2 macrophages) were measured using ELISA. M0 macrophages were polarized to the M1 phenotype using LPS + IFN-γ and to the M2 phenotype using IL-4 + IL-13. Experimental groups included control, induction, and FMN treatment groups. The results indicated that FMN promotes M1 polarization and suppresses M2 polarization. (i) Monocle trajectory analysis of macrophage development. (j) CellPhoneDB analysis revealed cell–cell communication mechanisms between macrophages and tumor cells. (k) CellPhoneDB analysis revealed cell–cell communication mechanisms between macrophages and tumor cells.

We also explored the impact of FMN on tumor-associated neutrophils (TANs), which play a dual role in shaping the fate of tumor cells, exhibiting phenotypic plasticity that can switch between anti-tumor and pro-tumor states depending on external stimuli from the TIME.^26,27 Anti-tumor TANs kill tumor cells through direct cytotoxic effects and by activating adaptive immune responses, while pro-tumor TANs are often linked with tumor proliferation, angiogenesis, and immune suppression.²⁸ In this study, we reclustered neutrophils into four distinct subgroups: Neu_Osm, Neu_Cd274, Neu_Cd74, and Neu_S100a8 [Fig. 8(a)]. A heatmap was generated to display the characteristic gene expression profiles of each neutrophil subgroup [Fig. 8(b)]. The Neu_Osm subgroup was characterized by high expression of Cxcr2, indicating a mature state, as well as elevated expression of granule-associated genes Osm and cytokine Ccl6, suggesting a role in promoting inflammatory responses. The Neu_Cd274 subgroup co-expressed Cxcr2 and Cxcr4, with relatively higher levels of Cxcr4, indicating a transition from a mature to a senescent state. This subgroup was marked by high expression of the pro-tumor gene Cd274 (PD-L1), reflecting an immunosuppressive phenotype that contributes to tumor progression. Previous studies have reported that Neu_Cd274 inhibits the activation of cytotoxic CD8⁺ T cells in primary liver cancer, contributing to immune suppression.²⁹ The Neu_Cd74 subgroup showed high expression of Cd74, MHC class II molecules (H2-Aa), and complement component genes (C1q), indicating a role in antigen presentation and complement activation. The Neu_S100a8 subgroup expressed high levels of granulocytic genes such as S100a8, S100a9, Mmp8, Mmp9, and Olfm4, which are involved in pro-inflammatory responses when released as granules. Gene Ontology (GO) enrichment analysis confirmed that the functional states of these subgroups aligned with the characteristic genes identified [Fig. 8(c)]. Trajectory analysis suggested that the Neu_S100a8 subgroup represents the starting point, with Neu_Osm occupying an intermediate stage. The Neu_Cd274 and Neu_Cd74 subclusters were identified as terminal states [Fig. 8(d)]. To examine the effects of FMN on neutrophil subpopulations, we compared their abundance between the control and FMN groups. The results showed an increase in the inflammatory response-related Neu_Osm and Neu_S100a8 following treatment, while the senescent, immunosuppressive Neu_Cd274 decreased in abundance [Fig. 8(e)]. Kaplan–Meier survival analysis revealed that high expression of Neu_Cd274 signature genes was associated with poor prognosis, consistent with previous findings [Fig. 8(f)]. Remarkably, after FMN treatment, the immunosuppressive Neu_Cd274 subgroup not only decreased in proportion but also showed reduced expression of the Cd274 [Fig. 8(g)]. GSVA and CellphoneDB analyses suggested potential crosstalk between Neu_Cd274 and tumor cells via ligand–receptor pairs such as LGALS9_CD44, LGALS9_LRP1, and HGF_CD44, which may contribute to the formation of an immunosuppressive environment. In conclusion, FMN appeared to mitigate the immunosuppressive state of neutrophils by reducing the proportion of the Neu_Cd274 and lowering Cd274 expression. This shift promoted an inflammatory activation response, enhancing the anti-tumor immune activity of neutrophils.

FIG. 8.

FMN reduces the infiltration of senescent neutrophils. (a) and (b) Reclustering of neutrophils identified four subpopulations, annotated based on marker genes. (c) KEGG pathway enrichment analysis of neutrophil subtypes. (d) Monocle trajectory analysis inferred the developmental trajectory of neutrophil subtypes. (e) Bar chart showed changes in the proportions of neutrophil subtypes after FMN intervention. (f) Kaplan–Meier survival analysis revealed lower overall survival rates in osteosarcoma patients with higher Neu_Cd274 feature scores. (g) Changes in the expression of immunosuppressive molecule Cd274 following FMN intervention. (h) CellPhoneDB analysis explored communication mechanisms among neutrophils, macrophages, and tumor cells.

Furthermore, we explored the effect of FMN on tumor-infiltrating lymphocytes (TILs), which are key immune components within the TIME and play crucial roles in tumor progression and therapeutic responses. TILs were reclustered and identified as six subclusters: three NK cell subtypes (NK_Xcl1, NK_Ncr1, NK_Mki67) and three T-cell subtypes (DPT, T_Cd74, T_Il7r) [Fig. S4(a)]. Each cell subcluster was annotated based on the expression of marker genes [Figs. S4(b) and S4(c)]. Notably, the double-positive T cells (DPT), which express both CD4 and CD8 markers, represent a unique T-cell population.^30,31 Previous studies have associated DPT cells with improved overall survival and relapse-free survival, indicating potential anti-tumor activity.³² However, in our analysis, the identified DPT cells also showed high expression of exhaustion-related genes such as Pdcd1 and Ctla4, suggesting that these cells may be in an exhausted state. Trajectory analysis revealed that the NK_Mki67 subtype, associated with proliferative activity, was positioned at the start of the differentiation trajectory, while NK_Xcl1 was primarily located at an intermediate stage, with some cells reaching the terminal differentiation state. NK_Ncr1 was situated at the end of the differentiation trajectory, representing a fully differentiated cytotoxic phenotype [Figs. S4(d) and S4(e)]. To evaluate the effects of FMN on TILs, we quantified the abundance of TIL subtypes before and after treatment. Compared to the control group, the FMN group showed slight increases in the NK_Xcl1 and NK_Ncr1, while the DPT exhibited a significant increase [Fig. S3(f)]. Heatmap indicated that DPT cells were in an exhausted state prior to treatment, but after FMN intervention, the expression of exhaustion markers Pdcd1 and Ctla4 was markedly reduced [Fig. S3(g)]. Furthermore, we assessed changes in cytotoxicity and exhaustion scores of TILs before and after FMN treatment. The results demonstrated a significant increase in cytotoxicity scores in the FMN group, while exhaustion scores showed no significant changes [Figs. S3(h) and S3(i)]. These findings suggested that FMN may enhanced the differentiation of the proliferative NK_Mki67 subtype into more cytotoxic NK_Xcl1 and NK_Ncr1 subtypes. Additionally, the reduction of exhaustion markers in DPT cells implied an improvement of their functional state. Collectively, FMN appeared to augment NK cell-mediated cytotoxicity and alleviate immune exhaustion, thereby inhibiting the progression of osteosarcoma. To further validate the effect of FMN on T cells, we isolated CD3+ T cells from human peripheral blood using magnetic bead positive selection and confirmed successful sorting by flow cytometry [Figs. S4(k) and S4(l)]. ELISA was used to detect the level of the T-cell exhaustion marker sPD-1. The control group was T cells without intervention, the exhaustion group was T cells induced to exhaustion using supernatants from HOS and MG63 cells, and the drug group was treated with FMN during exhaustion induction. The results showed that FMN could alleviate T-cell exhaustion [Fig. S4(i)]. Collectively, FMN appeared to augment NK cell-mediated cytotoxicity and alleviate immune exhaustion, thereby inhibiting the progression of osteosarcoma.

III. DISCUSSION

Neoadjuvant chemotherapy resistance and severe side effects remain major obstacles to improving the prognosis of osteosarcoma patients. Hence, there is an urgent need to identify effective therapeutic agents with reduced toxicity. FMN has been reported as a promising anti-tumor candidate derived from traditional Chinese medicine.^33,34 However, its precise anti-osteosarcoma effects and underlying mechanisms have not been fully elucidated. In this study, we validated the anti-osteosarcoma efficacy of FMN using a PDX model for the first time. Furthermore, we performed an in-depth exploration of its mechanisms of action through integrated scRNA-seq and scATAC-seq. Our findings provide novel insights and serve as a valuable reference for the development of new anti-osteosarcoma therapies.

In this study, analysis of osteosarcoma cells from PDX models revealed that the C1 subpopulation was particularly sensitive to FMN treatment. Combined with scRNA-seq data from clinical osteosarcoma samples, we identified MYO1B as a potential key target mediating the anti-osteosarcoma effects of FMN. MYO1B, a member of the class I myosin family, is located on chromosome 2q32.3 and contains 35 exons encoding the unconventional myosin 1B protein. It primarily participates in cytoskeletal organization and cellular movement, playing a critical role in maintaining cell morphology and motility. Increasing evidence suggests that dysregulated expression of MYO1B is a significant factor in tumor initiation and progression. Elevated MYO1B expression has been observed in several malignancies, including colorectal cancer,³⁵ glioma,³⁶ and head and neck squamous cell carcinoma,³⁷ where it is associated with aggressive tumor behavior and poor prognosis. For instance, Chen et al.³⁵ reported that MYO1B overexpression in colorectal cancer inhibited the autophagic degradation of HIF-1α, leading to its accumulation and subsequent enhancement of VEGF secretion, which promoted tumor metastasis. However, the potential role of MYO1B in osteosarcoma has not been previously reported. To validate MYO1B as a functional target, we performed gain- and loss-of-function experiments combined with FMN treatment in osteosarcoma cell lines. These findings reveal, for the first time, that FMN exerts its inhibitory effects on osteosarcoma malignancy via the direct targeting and suppression of MYO1B.

Notably, beyond directly targeting tumor cells, FMN treatment also resulted in significant remodeling of the TIME. Macrophages are the most abundant infiltrating immune cells and can be broadly classified into M1 and M2 phenotypes. Both M1 and M2 macrophages exhibit high plasticity and can undergo phenotype switching in response to changes in therapeutic interventions, making them promising targets for developing novel therapies.^38,39 Our findings indicated that FMN promoted macrophage polarization toward the M1 phenotype while inhibiting the shift toward the M2 phenotype, thereby exerting an anti-tumor immune effect. Beyond the observed shift in macrophage polarization, the underlying mechanisms may involve modulation of canonical signaling pathways such as NF-κB/STAT1 activation and suppression of STAT6/PI3K/AKT signaling,^40,41 thereby facilitating the reprogramming of macrophages toward a pro-inflammatory phenotype. Moreover, FMN may also reshape the metabolic profile of macrophages, favoring glycolysis over oxidative phosphorylation, which is characteristic of M1 polarization.⁴² Functionally, these changes are likely to diminish the immunosuppressive roles of TAMs, including angiogenesis and immune evasion, while enhancing antigen presentation and cytotoxic T-cell activation. Taken together, these macrophage-directed effects of FMN warrant further investigation to elucidate the underlying mechanisms and therapeutic implications.

Recent studies have highlighted the role of tumor cell-neutrophil crosstalk, where various chemokines increase the infiltration of TANs into the TIME. TANs can be classified into anti-tumor and pro-tumor subtypes based on their biological behaviors that influence tumor progression.^43–45 In our study, we observed a reduction in the pro-tumor Neu_Cd274 (PDL1⁺ neutrophils) population following FMN treatment. Cell–cell communication analysis indicated that Neu_Cd274 may engage in crosstalk with tumor cells through multiple immunosuppressive ligand–receptor pairs, which could be critical for recruiting pro-tumor neutrophils like Neu_Cd274 to the tumor site. Thus, FMN appeared to enhance host anti-tumor immunity by promoting M1 macrophage polarization, inhibiting M2 macrophage polarization, and reducing the infiltration of pro-tumor Neu_Cd274. This dual modulation of immune cell phenotypes highlighted the potential of FMN as a therapeutic agent in reshaping the TIME and enhancing anti-tumor immune responses.

However, this study has certain limitations. First of all, the sample size of the PDX model used in this study is relatively small. Although the data show significant differences, further expanding the sample size when conditions permit in the future will help confirm the generalizability of the efficacy and reduce potential bias. Second, this study primarily focuses on the pharmacological effects of formononetin and its regulatory mechanism on the FOXP1-MYO1B axis; its impact on the immune microenvironment was evaluated as a supplementary analysis. The lack of investigation and validation regarding the potential mechanisms underlying TIME remodeling warrants further in-depth exploration. Additionally, the findings of this study provide a proof-of-concept for targeting the FOXP1-MYO1B pathway. Subsequent translational research will require systematic pharmacokinetic and toxicological evaluations to advance FMN toward preclinical and clinical development.

IV. CONCLUSION

In summary, our findings reveal the gene expression and epigenetic regulatory features of osteosarcoma in the context of FMN treatment, highlighting the cellular diversity and heterogeneity within the TIME. This study provides valuable insights into the potential mechanisms of FMN's anti-osteosarcoma effects and supports its translational application as a therapeutic agent in osteosarcoma treatment.

V. METHODS

A. Acquisition of human osteosarcoma tissue samples

This study was conducted in accordance with the international ethical guidelines for biomedical research involving human subjects (CIOMS) and was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Approval No. 2019KY-E-097). Osteosarcoma tissue specimens were obtained under sterile conditions and immediately placed in pre-chilled complete culture medium containing 5% penicillin/streptomycin at 4 °C. The samples were promptly transported to the laboratory for PDX model establishment. Additionally, the tumor tissues were sent for pathological examination to further confirm the diagnosis.

B. Establishment and intervention of osteosarcoma PDX model

Ten female BALB/c-nu nude mice (4 weeks old, 13–16 g) and female NCG mice (4 weeks old) were purchased from GemPharmatech Co., Ltd. (Jiangsu, China). All mice were housed in the SPF-grade animal facility of Guangxi Medical University. To investigate the therapeutic response of FMN against human osteosarcoma, we established osteosarcoma PDX models to replicate the drug treatment response of the parental tumor.

Under sterile conditions, fresh tumor tissue samples were cut into small fragments (3 × 3 × 3 mm) and implanted into the right dorsal area between the spine and hip joint of NCG mice. Before the tumor volume reached 1500 mm³, tumors were harvested and passaged for subsequent model establishment. Finally, tumors from the third-generation NCG PDX models were transplanted into BALB/c-nu nude mice to establish the PDX models for drug intervention experiments.

Seven days after tumor implantation in the nude mice, the tumor volumes were measured. The mice were then randomly divided into two groups with no significant difference in tumor volume: the control group (control, n = 5) and the FMN treatment group (FMN, n = 5). The FMN group received FMN via oral gavage at a dose of 50 mg/kg, while the control group was given an equal volume of normal saline. Treatments were administered every two days for a total of 28 days. At the end of the experiment, all mice were sacrificed under excessive anesthesia. Tumor samples were collected and weighed, and their volumes were measured using the formula: tumor volume = (length × width2)/2. In the FMN group, three samples were randomly selected for scRNA-seq, and one sample was chosen for scATAC-seq. The remaining samples were stored for subsequent experimental validation.

C. scRNA-seq and scATAC-seq

Both scRNA-seq and scATAC-seq were performed by Genenergy Biotechnologies (Shanghai, China). The detailed experimental procedures followed the protocols provided by 10× Genomic (https://www.10xgenomics.com/cn/support).

D. Data acquisition

Gene expression quantitative data in HTSeq-FPKM format, along with clinical information from 84 osteosarcoma cases, were obtained from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database (https://ocg.cancer.gov/programs/target). The GSE237070 dataset was downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo), including primary tumor (PT) samples and matched paracancerous (PC) samples from four osteosarcoma patients, which were used for identifying pathogenic targets at the single-cell level. Gene expression data of musculoskeletal samples from 396 healthy individuals were retrieved from the GTEx database (https://gtexportal.org). Additionally, scRNA-seq and scATAC-seq data from three and one untreated osteosarcoma PDX model tumor samples, respectively, were mined from the Genome Sequence Archive in the National Genomics Data Center (https://ngdc.cncb.ac.cn/gsa, GSA: CRA017660).

E. Preprocessing and analysis of scRNA-seq data

Raw data in FASTQ format of scRNA-seq were processed using the Cell Ranger software package (Version 7.1.0). The reads were aligned to the GRCh38_and_mm10-2020-A reference genome, generating a gene expression matrix and distinguishing between human-derived and mouse-derived cells. Quality control and downstream analyses were performed using the Seurat package (Version 4.3.0) in R (Version 4.2.3). Potential doublets were identified and removed using the DoubletFinder package⁴⁶ (Version 2.0.3). After quality control, the gene expression matrix was normalized using the NormalizeData function and standardized with ScaleData, and highly variable genes were identified using the FindVariableFeatures function. Following principal component analysis (PCA) for dimensionality reduction, the Harmony package (Version 0.1.0) was applied to mitigate potential batch effects from different sample sources. Cell clustering analysis was performed using the FindNeighbors and FindClusters functions, and nonlinear dimensionality reduction and visualization were conducted with the RunTSNE function.

F. Cell type annotation

To annotate cell types, the “FindAllMarkers” function in the Seurat package was used to identify marker genes with high recognition across clusters. Subsequently, manual annotation of cell clusters was performed based on the expression of well-established marker genes.

G. Gene function enrichment analysis

Gene set variation analysis (GSVA) was conducted using the GSVA package (Version 1.32.0) in R to identify significantly enriched gene sets within each cellular subpopulation. The HALLMARK gene sets used for GSVA were obtained from the Molecular Signatures Database (MSigDB) (https://www.gseamsigdb.org/gsea/downloads.jsp). Additionally, metastasis-related gene sets were downloaded from the CancerSEA database (http://biocc.hrbmu.edu.cn/CancerSEA/). To further explore the biological functions and pathway enrichment characteristics of each cell type, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses using the clusterProfiler package.⁴⁷ High-expression genes identified for each cell type from the “FindAllMarkers” function were used as input to calculate the most likely enriched biological processes and involved signaling pathways.

H. Copy number variation analysis

To assess the copy number variations (CNVs) in osteosarcoma cells, we employed the inferCNV package (Version 1.2.1). Plasma cells, B cells, and endothelial cells were used as reference cells for CNV score calculation.

I. Trajectory analysis

We employed Monocle 2 package⁴⁸ (Version 2.20.0) to infer and reconstruct the differentiation trajectories of cells or the evolutionary relationships among different cell types based on changes in gene expression across various cellular subpopulations. The analysis was conducted in a completely unsupervised manner. Dimensionality reduction was performed using the “DDRTree” function, and the “plot_cell_trajectory” function was used to visualize the cell differentiation trajectories.

J. Cell–cell communication analysis

To infer potential molecular communication between different cell types, we utilized CellphoneDB.py⁴⁹ (Version 3.1.0). The analysis was based on the expression of ligand–receptor pairs to predict the strength of presumed cell–cell communication and construct interaction networks. Ligand–receptor pairs with a P-value < 0.05 were considered statistically significant.

K. Differentially expressed genes and identification of key feature genes

Differentially expressed genes (DEGs) were identified using the “FindMarkers” function in the Seurat package with default parameters. Specifically, differential analysis was performed between the control and FMN groups, as well as between PT and PC osteoblasts. DEGs were defined as those with |log fold change| > 0 and P-value < 0.05. Common DEGs identified from both comparisons were considered a set of tumor-promoting genes sensitive to FMN. To prioritize the most important genes within this set, we performed feature selection using osteosarcoma sequencing data from the TARGET database. Three machine learning algorithms—LASSO, SVM-RFE and Random Forest—were applied for variable selection.

L. Gene module enrichment analysis

The “AddModuleScore” function in the Seurat package was used to evaluate the functional characteristics of specific cell types. The related gene sets for tumor-infiltrating lymphocytes (TILs) and macrophages are listed in Table S1. Mouse macrophage M1/M2 polarization genes were sourced from the CIBERSORT reference dataset “mice.txt,” which includes markers for 25 immune cell types in mouse.⁵⁰

M. Data integration and survival analysis

Data from the TARGET osteosarcoma cohort and normal musculoskeletal samples from the GTEx database were integrated using the ComBat algorithm from the Survival package in R⁵¹ to mitigate batch effects and facilitate a comprehensive analysis of target gene expression patterns. Kaplan–Meier survival curves for candidate variables were plotted using the Survival package (Version 3.5-5) and the Survminer package (Version 0.4.9) in R.

N. ScATAC-seq data analysis

The raw data in FASTQ format of scATAC-seq were processed into output peak-barcode count matrix using Cell Ranger ATAC (v1.0.1) with the reference genome library “refdata-cellranger-atac-GRCh38-and-mm10-2020-A-2.0.0” provided by 10X Genomics. The Signac package (v0.2.5) was using for downstream analysis. Low-quality cells were filtered based on the following criteria: minFrags = 3000, maxFrags = 30 000, pctPF = 15, NS = 10, and TSS.ES = 1. Next, the “FindIntegrationAnchors” function was used to identify “anchors” between samples, which were then input into the “IntegrateData” function to create a unified, batch-effect-corrected expression matrix. High-dimensional data were visualized using t-SNE for dimensionality reduction, and cell clustering was identified using the “FindClusters” tool. Then, cell type annotations in scATAC-seq data were predicted by integrating anchor points between scRNA-seq and scATAC-seq datasets. To explore changes in chromatin accessibility of transcription factor (TF) motifs, we performed motif analysis using the chromVAR package.⁵² Differential TFs were identified based on genes with differential open chromatin peaks, and the footprinting signals of TFs were visualized using the “PlotFootprint” function.

O. Cell culture

The human osteosarcoma cell lines (143B, HOS, MG63, SAOS2, U2OS), the osteoblast cell line hFOB 1.19, and the Human monocytic leukemia cell line THP-1 were purchased from the Cell Bank of the Chinese Academy of Sciences. All tumor cell lines and THP-1 cells were incubated at 37 °C with 5% CO₂, while hFOB 1.19 cells were incubated at 33.5 °C with 5% CO₂. Details on the specific culture media conditions can be found in the relevant protocols provided by the Cell Bank of the Chinese Academy of Sciences (https://www.cellbank.org.cn/).

P. Isolation and sorting of primary T cells

This study was conducted in accordance with protocols approved by the Ethics Committee. Peripheral blood (10 ml) was collected from healthy volunteers and diluted with an equal volume of phosphate-buffered saline (PBS). The diluted blood was carefully layered onto 20 ml of human lymphocyte separation solution in a centrifuge tube using a pipette to maintain a clear interface. Centrifugation was performed at 500 × g for 30 min at 20 °C with slow acceleration and deceleration settings. Following centrifugation, the lymphocyte layer at the interface between the separation solution and plasma was carefully collected. The isolated lymphocytes were washed twice with PBS and resuspended in 1 ml of ImmunoCult™-XF T Cell Expansion Medium for counting and temporary storage. CD3⁺ T cells were subsequently isolated using the STEMSELL™ CD3 Positive Selection Kit according to the manufacturer's instructions. Briefly, the cell suspension was transferred to a 5 ml polystyrene flow cytometry tube, and 100 μl of antigen–antibody complex solution was added per milliliter of cells, followed by incubation at room temperature for 3 min. Then, 60 μl of magnetic beads per milliliter of cells was added, and the mixture was incubated for an additional 3 min at room temperature. The tube was filled with buffer to a total volume of 2.5 ml and placed in a magnetic stand for 3 min at room temperature. The supernatant was carefully aspirated, and the washing process was repeated twice. The magnetically retained CD3⁺ T cells were resuspended in 1 ml of culture medium. An aliquot of the sorted cells was analyzed by flow cytometry to determine CD3⁺ purity and for cell counting. For T-cell activation, the sorted CD3⁺ T cells were supplemented with recombinant IL-2 (50 U/ml) and CD3/CD28 T-cell activator at 25 μl/ml. The cell density was adjusted to 1–2 × 10⁶ cells/ml, and the suspension was transferred to 24-well or 6-well plates. Cultures were maintained at 37 °C under 5% CO₂ for subsequent experimental applications.

Q. Quantitative reverse transcription polymerase chain reaction

Total RNA was isolated and extracted using an RNA isolation kit following the manufacturer's instructions. RNA was then reverse transcribed into complementary DNA (cDNA) using the PrimeScript™ RT Master Mix kit (Takara, Japan). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed using the SYBR Green PCR Kit on a 7500 Real-Time PCR System. GAPDH was used as the reference gene, and the relative RNA expression levels of target genes were normalized using the 2 ^ ^(−ΔΔCt) method. The primer sequences used in this study are listed in Table S2.

R. Cell transfection

The MYO1B siRNA silencing reagent and overexpression plasmid were constructed by Sangon Biotech (Shanghai, China). Cell transfection was performed according to the manufacturer's protocol. The siRNA sequences targeting FOXP1 and MYO1B are provided in Tables S3 and S4. Details of the overexpression plasmid are as follows: vector: pEGFP-C1, cloning sites: BglII/ApaI, plasmid length: 3423 bp.

S. Cell viability assay

HOS and MG63 cells (5 × 10³ cells/well) were seeded in 96-well plates. When the cell confluency reached 60%–70%, the cells were treated with various concentrations of FMN (0, 10, 20, 40, 80, 160 μM) according to the experimental design and incubated at 37 °C for 48 h. Subsequently, 10 μl of CCK-8 reagent (Solarbio, Beijing, China) was added to each well and incubated for 1 h. The absorbance [OD (Optical Density) value] of each well was measured at 450 nm using a multifunctional microplate reader.

T. Immunofluorescence staining

HOS and MG63 cells were seeded in 24-well plates with glass coverslips (20 × 10³ cells/well). Once the cells reached the logarithmic growth phase, they were treated with FMN according to the experimental groups. After 48 h of treatment, the cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. The cells were then blocked with 5% BSA and incubated with the primary antibody at 4 °C overnight. Subsequently, fluorescence-conjugated secondary antibodies were applied and incubated at room temperature for 1 h. The nuclei were counterstained with DAPI, and anti-fade mounting medium was added. Fluorescent images were captured using an upright fluorescence microscope (Zeiss, Germany).

U. EdU cell proliferation assay

The BeyoClick™ EdU-488 Cell Proliferation Detection Kit (Beyotime, Shanghai) was used to assess cell proliferation. HOS and MG63 cells (20 × 10³ cells/well) were seeded in 24-well plates with glass coverslips and cultured. Following 48 h of drug treatment according to the experimental groups, 200 μl of EdU reagent was added to each well, and the cells were incubated at 37 °C for 2 h. The cells were then fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.3% Triton X-100 for 15 min. Click Additive Solution was applied, and the reaction mixture was incubated for 30 min. After washing, the cells were stained with DAPI for 10 min. Finally, cell images were captured using an upright fluorescence microscope (Zeiss, Germany).

V. Flow cytometry analysis for apoptosis detection

Flow cytometry was employed to assess apoptosis in HOS and MG63 cells. Cells were seeded in 6-well plates and treated with varying concentrations of FMN (0, 80, and 160 μM). After 48 h of treatment, the cells were harvested and resuspended in 500 μl of 1× Binding Buffer. Next, 5 μl of annexin V-APC and 10 μl of 7-AAD were added to the cell suspension and gently mixed. The mixture was incubated at room temperature in the dark for 5 min. Finally, the apoptotic status of the cells was analyzed using a flow cytometer (Thermo Fisher Scientific, America).

W. Flow cytometric analysis of cell surface marker expression

Flow cytometry was performed to assess the expression of specific surface markers for cell identification. For the detection of CD3 expression, the sorted T-cell suspension was labeled with 5 μl of CD3-FITC antibody and incubated at room temperature in the dark for 20 min. After incubation, the cells were washed twice with PBS via centrifugation and then subjected to flow cytometric analysis to determine the percentage of CD3-positive cells. For simultaneous detection of CD11b and CD14 expression, THP-1 cells were cultured in a 6-well plate, with experimental groups consisting of a control (untreated THP-1 cells) and an induced group (M0 macrophages differentiated by treatment with PMA for 48 h, followed by replacement with normal medium). Cells from both groups were harvested, resuspended in 500 μl of 1× binding buffer, and stained with 5 μl of CD11b-PE and 5 μl of CD14-APC antibodies. The samples were gently mixed and incubated for 20 min at room temperature in the dark, followed by two washes with PBS. The stained cells were then analyzed by flow cytometry to quantify CD11b and CD14 expression.

X. Cell migration and invasion assays

The migration and invasion abilities of osteosarcoma cells were assessed using transwell chambers (for migration assays) or matrigel-coated chambers (for invasion assays). For each assay, 50 000 cells/well were seeded into the upper chamber placed in a 24-well plate. The lower chamber was filled with 600 μl of complete medium containing 10% FBS as a chemoattractant. The plates were incubated at 37 °C for 48 h. After incubation, the cells that had migrated or invaded into the bottom surface of the upper chamber were fixed and stained. The stained cells were then visualized and counted under an upright microscope (Olympus, Japan).

Y. Enzyme-linked immunosorbent assay (ELISA)

THP-1 cells were seeded into 24-well plates and divided into experimental groups as follows: the control group (M0 macrophages differentiated with 320 nmol/l PMA for 18 h), the M1 group (treated with 320 nmol/l PMA for 12 h followed by induction with 100 nmol/l PMA, 100 ng/ml LPS, and 20 ng/ml IFN-γ for 48 h), and the M1 + FMN group; alternatively, cells were grouped into control, M2 (induced with 100 nmol/l PMA, 20 ng/ml IL-4, and 20 ng/ml IL-13 for 48 h after initial PMA treatment), and M2 + FMN groups. Meanwhile, isolated and sorted T cells were plated and assigned to a T-cell group, a T cell + OS cell group (treated with supernatants of HOS and MG63 cells for 48 h), and a T cell + OS cell + FMN group. For ELISA detection, after equilibrating reagents at room temperature for 20 min, standards and samples were applied to the plate—50 μl per standard well and 10 μl sample plus 40 μl diluent per sample well, with blank wells left untreated. Then, 100 μl of HRP-labeled detection antibody was added to each well except the blanks, followed by incubation at 37 °C for 60 min. After washing five times and drying, 50 μl of substrate A and B was added and incubated for 15 min at 37 °C. The reaction was stopped with 50 μl stop solution, and the OD values were measured at 450 nm to calculate protein concentrations based on the standard curve.

Z. H&E staining and immunohistochemistry

Liver of nude mice from the control and FMN groups was processed into tissue sections for staining. After deparaffinization, the sections were stained sequentially with H&E. To evaluate the changes in MYO1B protein expression levels before and after FMN intervention, PDX tumor from both groups was sectioned and subjected to immunohistochemistry (IHC) staining. The detailed IHC procedure followed a previously reported protocol.⁵³

AA. Statistical analysis

All statistical analyses were performed using R and GraphPad Prism (v8.0). P-value < 0.05 was considered statistically significant. Significance levels are indicated as follows: P < 0.05 (^*), P < 0.01 (^**), P < 0.001 (^***), and P < 0.0001 (^****). Non-significant results are denoted as “ns.”

SUPPLEMENTARY MATERIAL

See the supplementary material Fig. S1: FMN and MYO1B regulated the migration and invasion of osteosarcoma cell: [Fig. S1(a)] Kaplan–Meier analysis indicated higher UBE2S expression correlated with reduced overall survival in osteosarcoma patients. [Fig. S1(b)] UBE2S was upregulated in osteosarcoma tissues compared to normal tissues. [Fig. S1(c)] UBE2S is highly expressed in the C1 subgroup. [Fig. S1(d)] CCK-8 assays demonstrated FMN inhibited osteosarcoma cell viability in a dose-dependent manner. [Fig. S1(e)] The half-maximal inhibitory concentration (IC₅₀) of FMN was 402.3 μM in HOS cells and 410.6 μM in MG63 cells. [Figs. S1(f)–S1(i)] Transwell assays showed FMN significantly suppressed the migration and invasion abilities of osteosarcoma cells in a dose-dependent manner. [Figs. S1(j)–S1(m)] Transwell assays demonstrated that MYO1B silencing/overexpression inhibited/enhanced the migration and invasion of osteosarcoma cell, respectively. Mechanisms underlying FMN-mediated inhibition of the migration and invasion in osteosarcoma cells. [Figs. S2(a)–S2(d)] Transwell rescue experiments indicated FMN suppressed the migration and invasion through MYO1B inhibition. [Fig. S3(a)] Integrated scATAC-seq and scRNA-seq analysis identified four distinct cellular subpopulations. [Fig. S3(b)] t-SNE plot showed the sample origin of each subpopulation. [Fig. S3(c)] Open chromatin peaks of MYO1B in C1 cluster of osteosarcoma cells revealed reduced chromatin accessibility in the FMN group. [Fig. S3(d)] Differential chromatin accessibility of the transcription factor FOXP1 under FMN treatment. [Fig. S3(e)] Predicted binding sites of MYO1B and FOXP1. [Fig. S3(f)] Correlation heatmap depicting relationships among mouse-derived cell types. [Figs. S3(g) and S3(h)] Immunofluorescence analysis showed that the expression of MYO1B at the protein level decreased compared to the control group after FOXP1 silencing. [Fig. S3(i)] PCR results indicated that the expression of MYO1B at the gene level decreased compared to the control group after FOXP1 silencing. (Fig. S4) FMN enhanced cytotoxicity of NK/T cells. [Fig. S4(a)] Reclustering of TILs identified six subtypes. [Figs. S4(b) and S4(c)] Heatmap and bubble chart showed marker genes of TIL subtypes. [Fig. S4(d)] Monocle trajectory analysis inferred the developmental trajectory of NK cells, with darker colors indicating earlier developmental stages. [Fig. S4(e)] Pseudotime progression of subpopulations. [Fig. S4(f)] FMN treatment increased the proportions of highly cytotoxic NK_Xcl1 and NK_Ncr1 subtypes. [Fig. S4(g)] Heatmap showed downregulation of exhaustion markers (e.g., Pdcd1, Ctla4) in double-positive T cells (DPT) following FMN intervention. [Figs. S4(h) and S4(i)] Changes in cytotoxicity and exhaustion scores of TILs after FMN treatment. [Fig. S4(j)] The successful isolation of CD3⁺ T cells from human peripheral blood was confirmed by flow cytometry following positive selection using magnetic beads and staining with a CD3-FITC antibody. [Figs. S4(k) and S4(l)] The level of soluble PD-1 (sPD-1), a marker of T-cell exhaustion, was measured via ELISA. The control group received no treatment, while the exhaustion group was treated with supernatants from HOS and MG63 cultures to induce T-cell exhaustion. The treatment group received formononetin (FMN) intervention during exhaustion induction. The results demonstrated that FMN alleviated T-cell exhaustion; The related gene sets for annotating tumor-infiltrating lymphocytes and macrophages (Table S1); The primer sequences used in this study (Table S2); The siRNA sequences targeting FOXP1 (Table S3).

AUTHOR DECLARATIONS

Conflict of Interest

The authors have no conflicts to disclose.

Ethics Approval

Ethics approval for experiments reported in the submitted manuscript on animal or human subjects was granted. This study was conducted in accordance with the international ethical guidelines for biomedical research involving human subjects (CIOMS) and was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Approval No.: 2019KY-E-097).

Author Contributions

Yun Liu, Liang Xiong, Wenyu Feng, and Tianyu Xie contributed equally to this work.

Yun Liu: Writing – original draft (equal). Liang Xiong: Data curation (equal); Methodology (equal); Writing – review & editing (equal). Wenyu Feng: Writing – original draft (equal). Tianyu Xie: Writing – original draft (equal). Jiming Liang: Writing – review & editing (equal). Mingxiu Yang: Writing – review & editing (equal). Shanhang Li: Writing – review & editing (equal). Kai Luo: Writing – review & editing (equal). Feicui Li: Writing – review & editing (equal). Shengping Tang: Writing – review & editing (equal). Shangyu Liu: Methodology (equal); Writing – review & editing (equal). Qian Huang: Methodology (equal). Shijie Liao: Writing – review & editing (equal). Jianhong Liu: Methodology (equal); Writing – review & editing (equal). Yangjie Cai: Methodology (equal); Writing – review & editing (equal). Qingjun Wei: Data curation (equal); Funding acquisition (equal). Haijun Tang: Data curation (equal); Writing – review & editing (equal). Fuxing Tang: Data curation (equal).

DATA AVAILABILITY

The data that support the findings of this study are openly available in Genome Sequence Archive (GSA: CRA019120) at https://ngdc.cncb.ac.cn/gsa, Ref. 54.