Single-cell RNA sequencing reveals NFATc1-mediated regulatory mechanisms of vascular smooth muscle cell osteogenic transformation in the osteosarcoma microenvironment

Abstract

Background

Osteosarcoma is the most common primary malignant tumor of bone and occurs predominantly in children and adolescents during periods of accelerated skeletal growth. Within the tumor microenvironment (TME), vascular smooth muscle cells (VSMCs) may acquire osteogenic characteristics. Nuclear factor of activated T cells c1 (NFATc1) is central to skeletal biology, yet its behavior in VSMCs in osteosarcoma remains insufficiently defined.

Methods

We profiled fresh osteosarcoma tissues by single-cell RNA sequencing on the 10x Genomics Chromium platform. Mechanical and enzymatic dissociation yielded single-cell suspensions; viable cells (> 85%) were enriched; libraries were prepared with Single Cell 3′ v3 reagents and sequenced on NovaSeq 6000. Reads were processed with Cell Ranger (GRCh38). Following log-normalization, 2,000 highly variable genes were retained. Harmony integration reduced batch structure. We computed 50 principal components, generated UMAP/t-SNE embeddings, and applied k-nearest neighbor/Louvain clustering (k = 20; resolution = 0.5). Cell types were assigned from differentially expressed markers cross-checked with CellMarker and PanglaoDB. NFATc1 distribution and co-expression patterns were assessed by Pearson correlation.

Results

Nine major cell classes were identified, including VSMCs, fibroblasts, macrophages, T/NK subsets, B cells, neutrophils, mast cells, and inflammatory myeloid cells. NFATc1 expression was most prominent in VSMCs, with a bimodal distribution within this lineage. In VSMCs, NFATc1 co-varied with SERGEF, APOE, and CCS, pointing to coordinated cytoskeletal remodeling, lipid handling, and copper homeostasis during osteogenic reprogramming.

Conclusions

At single-cell resolution, NFATc1 marks a VSMC fraction with osteogenic features in osteosarcoma. The associated network suggests testable links to metabolism and matrix remodeling; functional perturbation will be required to establish causality.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-03606-9.

Introduction

Osteosarcoma is the most common primary malignant tumor of bone, with incidence peaking in the first two decades of life when skeletal growth accelerates. Despite advances in multidisciplinary care combining surgery, chemotherapy, and radiotherapy, prognosis remains poor once metastasis is present; the five-year survival rate typically ranges from 20 to 30% [1–3]. The osteosarcoma TME comprises malignant cells intertwined with fibroblasts, immune and endothelial cells, and other stromal elements. Through soluble factors and extracellular vesicles, these populations communicate and collectively shape tumor progression, therapeutic resistance, and metastatic spread. VSMCs are particularly notable in this context: they support vascular remodeling that sustains tumor growth, and their tendency toward ossification correlates with aggressive behavior and unfavorable outcomes. Together, these observations make VSMCs an appropriate focus for mechanistic investigation in osteosarcoma.

Recent studies emphasize that VSMCs can adopt osteogenic phenotypes within tumors. This conversion has been linked to vascular calcification, augmented angiogenesis, and the formation of niches supportive of dissemination [4–6]. Clarifying its molecular basis is essential for interventions that might alter disease course. Among candidate regulators, NFATc1—first recognized in T-cell activation—has been implicated in bone and cartilage development, osteoclast biology, and pathological calcification [7–9]. Subsequent work shows that NFATc1 orchestrates programs related to calcium signaling, bone matrix synthesis, and lineage commitment [10, 11]. However, the specific role of NFATc1 in VSMC osteogenic transformation within the osteosarcoma microenvironment remains uncertain.

The advent of single-cell RNA sequencing (scRNA-seq) allows cell-by-cell analysis of complex tissues, revealing heterogeneity and rare states that are invisible in bulk assays [12, 13]. Applying this toolkit to osteosarcoma, we mapped the TME, observed marked heterogeneity, and noted selective NFATc1 elevation in VSMCs. Co-expression analysis highlighted associations with SERGEF, APOE, and CCS, implicating actin remodeling, lipid metabolism, and copper handling in the VSMC osteogenic program. Although time-series profiling would refine the dynamics of these changes, our primary objective here was to define key molecular signatures and provide a foundation for subsequent mechanistic studies.

In this work, we combined scRNA-seq with single-cell exosome RNA profiling to survey the osteosaroma microenvironment and to interrogate NFATc1-centered pathways potentially directing the osteogenic trajectory of VSMCs. By integrating cellular diversity, transcriptional programs, and gene-level correlations, we sought to delineate NFATc1’s contribution to vascular pathology and identify candidate therapeutic targets in osteosarcoma.

Methods

Single-cell RNA sequencing data collection and preprocessing

Single-cell exosome RNA sequencing was performed on osteosarcoma tissues using the 10x Genomics Chromium platform. Fresh samples underwent mechanical and enzymatic dissociation to obtain single-cell suspensions [12, 13]. Dead cells and debris were removed by flow cytometry to achieve > 85% viability. Single-cell capture and library construction employed Chromium Single Cell 3′ Reagent Kits v3 with a target of approximately 10,000 cells per sample. Sequencing was carried out on an Illumina NovaSeq 6000 system with an average depth of ~ 50,000 reads per cell. Raw data were aligned and quantified using Cell Ranger against the GRCh38 reference.

We used Seurat and Harmony to standardize and integrate datasets. Log-normalization and scaling were applied to expression matrices; highly variable genes were selected via the “vst” method. PCA was used for initial dimensionality assessment and to evaluate potential batch structure before integration. Harmony was then used to align datasets and mitigate batch effects, ensuring comparability across samples for downstream analyses.

Data quality control and gene filtering

Quality control (QC) was performed in Seurat with the following thresholds: detected genes (nFeature_RNA) between 200 and 10,000, total RNA counts (nCount_RNA) < 150,000, and mitochondrial gene percentage (percent.mt) < 25%. Cells failing QC were excluded. Highly variable genes (n = 2,000) were identified for dimensionality reduction and clustering, and data were normalized and centered with ScaleData.

Batch effect assessment and processing methods

We evaluated systematic deviations between batches using PCA and UMAP. Batch effects were adjusted using Harmony and Seurat integration functions, aligning datasets in a shared high-dimensional space. Post-correction, embeddings and clustering were re-evaluated to confirm reduction of batch-related structure. These steps enabled accurate assessment of NFATc1 expression patterns in VSMCs and construction of its co-expression network without confounding technical variation.

Dimensionality reduction analysis and cell clustering

PCA was applied to the 2,000 highly variable genes, and the top 50 principal components (PCs) were retained; effective PCs were chosen by elbow inspection. Nonlinear embeddings used RunUMAP and RunTSNE. The neighborhood size was set to 30 with a minimum distance of 0.3. Graph-based clustering used FindNeighbors (k = 20) followed by FindClusters with Louvain resolution 0.5, yielding nine clusters.

Two mainstream nonlinear methods—UMAP and t-SNE—were employed to interrogate structure. Default parameters provided balanced preservation of global and local structure; sensitivity checks varied n_neighbors and min_dist and showed robust embeddings over reasonable ranges, supporting stability of the inferred clusters.

Cell type identification and annotation

Differential expression analysis with FindAllMarkers (log2-fold change > 0.25; adjusted p < 0.05) identified cluster-defining genes. Cell-type labels were assigned by comparison with CellMarker and PanglaoDB and the published literature. Major types included vascular smooth muscle cells (VSMCs), fibroblasts, macrophages, T cells, B cells, NK cells, neutrophils, inflammatory myeloid cells, and mast cells.

NFATc1 expression pattern analysis

NFATc1 expression was visualized on UMAP with FeaturePlot using continuous color scales. Violin plots (VlnPlot) summarized expression distributions across annotated types. Within VSMCs, density plots and histograms described a clear bimodal pattern, indicating sub-state heterogeneity.

Coexpression network construction and correlation analysis

Co-expression networks were constructed by correlating NFATc1 with all genes using Pearson coefficients. Significance thresholds were set at |r|>0.4 with p < 0.05. Particular attention was given to genes involved in ECM remodeling, lipid metabolism, and signaling, including APOE, SERGEF, CCS, LARS, and METTL21A. corrplot heatmaps and scatter plots summarized the correlation structure.

Intercellular expression flow analysis

NFATc1 expression was stratified into low, medium, and high grades. Using ggalluvial, Sankey diagrams were generated to depict flows from each annotated cell type to expression strata, providing an intuitive summary of type-specific NFATc1 distributions.

Quantitative Real-Time PCR validation

To validate the single-cell findings, we used a vascular smooth muscle cell (VSMC) line (human aortic smooth muscle cells, HA-VSMCs). To mimic the osteosarcoma microenvironment, VSMCs were cultured with osteosarcoma-conditioned medium collected from control VSMCs were maintained under standard culture conditions without conditioned medium.

Total RNA was extracted using TRIzol (Invitrogen, USA), and cDNA was synthesized using PrimeScript RT (Takara, Japan). qRT-PCR was performed with SYBR Green Master Mix (Bio-Rad, USA) on a CFX96 system (Bio-Rad, USA). GAPDH served as the internal control, and relative gene expression levels were calculated using the 2^^−ΔΔCt method. The primer sequences used for qRT-PCR are listed in Table S1.

Statistical analysis

All statistical analyses were performed in R (version 4.2.0). Wilcoxon rank-sum tests compared gene expression between groups with Benjamini–Hochberg correction. Pearson correlation measured associations, with p < 0.05 considered significant. Plots were generated using ggplot2, Seurat, and related packages.

Results

Quality control of single-cell RNA sequencing data and identification of highly variable genes

Quality assessment of the single-cell exosome RNA sequencing dataset from osteosarcoma yielded an overall portrait of transcriptomic diversity at the cellular level. As summarized in Fig. 1A, three commonly used metrics showed distributions consistent with high-quality single-cell libraries: the number of detected features per cell (nFeature_RNA) was largely contained within the 0–10,000 range and displayed a relatively even spread across cells; the total RNA counts (nCount_RNA) extended from 0 to 150,000, reflecting expected variation in transcriptional output; and the fraction of mitochondrial transcripts (percent.mt) clustered predominantly between 0% and 25%, indicating that the majority of captured cells retained adequate integrity. These patterns provide confidence that downstream modeling would not be dominated by poor-quality droplets. We next prioritized genes contributing most to biological dispersion. The highly variable gene (HVG) procedure retained 2,000 features for subsequent analyses (Fig. 1B), with the remainder (31,514 nonvariable genes) aggregating in the low-expression/low-variance region of the mean–variance landscape. Importantly, QC metrics were not independent: as shown in Fig. 1C, percent.mt exhibited a moderate inverse association with nFeature_RNA (r = − 0.52), linking transcriptional complexity to cellular health. Together, these observations establish a robust baseline for interrogating exosome-associated regulation of VSMC osteogenic change and the involvement of NFATc1 in that process.

Fig. 1.

Quality control and preprocessing analysis of single-cell RNA sequencing data from osteosarcoma. A Distributions of nFeature_RNA, nCount_RNA, and percent.mt. B HVG selection showing the relation between average expression and standardized variance; red points: HVGs (n = 2,000), black points: nonvariable genes (n = 31,514). C Correlation between percent.mt and nFeature_RNA (r = − 0.52). *p < 0.05, **p < 0.01

Principal component analysis and recognition of single-cell transcriptome clustering in osteosarcoma

Dimensionality reduction by principal component analysis (PCA) captured dominant axes of variation across cells. The projection onto the first two principal components (Fig. 2A) revealed a continuous manifold rather than sharp partitions, a common pattern in complex tumor ecosystems. Inspection of gene loadings (Fig. 2B) clarified the biological content of these axes: PC1 was shaped primarily by inflammation and immune-associated transcripts—PLAUR, CYP1A1, CD14, and IL1B—whereas PC2 emphasized extracellular matrix and osteogenic programs, including SPARC, COL1A2, MAF1B, and MXRA8. The contribution of early PCs to global variance (Fig. 2C) justified their use in subsequent embeddings and graph construction.

Fig. 2.

Principal component analysis and cell clustering of single-cell RNA sequencing data from osteosarcoma. A PC1–PC2 scatter plot. B Loadings for PC1 (left) and PC2 (right) emphasizing immune/inflammatory vs. ECM/osteogenic signatures. C Variance explained per PC. D UMAP (left) and t-SNE (right) revealing nine subpopulations. *p < 0.05, **p < 0.01

Nonlinear mapping with UMAP and t-SNE produced stable representations that agreed with one another and enabled community detection of transcriptionally coherent groups (Fig. 2D). Using a k-nearest neighbor graph and Louvain clustering at a moderate resolution, we resolved nine clusters (0–8) that persisted across embeddings. Functional annotation of genes co-expressed with NFATc1 indicated enrichment for skeletal system development, cell differentiation, and apoptosis in Gene Ontology analyses, while KEGG pathway interrogation highlighted canonical Wnt, TGF-β, and MAPK cascades—each with prior ties to osteogenic specification and matrix remodeling. These layers together place NFATc1 within a network of processes plausibly linked to VSMC phenotype changes in osteosarcoma.

Gene expression profile and functional annotation of subgroups of osteosarcoma cells

Subpopulation analysis delineated nine transcriptionally distinct groups with characteristic spatial patterns on UMAP. As illustrated in Fig. 3A, clusters 0, 3, and 6 formed compact neighborhoods, whereas cluster 7 displayed a more diffuse footprint, pointing to greater internal diversity. Marker-based inspection via heatmaps (Fig. 3B) further defined each group: clusters 1 and 8 were enriched for mast-cell genes (TPSB2, TPSAB1); clusters 5 and 6 upregulated fibrillar collagens (COL1A1, COL1A2, COL3A1), consistent with active extracellular matrix (ECM) synthesis and osteogenic remodeling; cluster 7 expressed CTSK and ACP5, genes typically associated with bone resorption; and cluster 2 displayed chemokines (CXCL5, CCL2) linked to immune recruitment and migratory programs.

Fig. 3.

Spatial distribution and characteristic gene expression analysis of single-cell subpopulations in osteosarcoma. A UMAP-based distribution of nine clusters (0–8). B Heatmap of 18 representative genes across subpopulations (g0–g8), spanning ECM, immune response, and bone metabolism. *p < 0.05, **p < 0.01

This constellation of cell states provides a framework for interpreting how NFATc1-linked signals, including those trafficked via extracellular vesicles, could coordinate VSMC progression toward osteogenic phenotypes. In particular, the collagen-high subpopulations likely represent synthetic, matrix-producing states that sit along the VSMC osteogenic continuum, while chemokine-rich groups may facilitate a permissive inflammatory niche that intersects with vascular remodeling and angiogenesis.

Identification of cell types in the microenvironment of osteosarcoma and analysis of NFATc1 expression characteristics

ntegrating differential markers with reference resources allowed assignment of cell identities consistent across both embeddings. As shown in Fig. 4A, the osteosarcoma microenvironment comprised inflammatory myeloid cells, NK cells, T cells, fibroblasts, neutrophils, B cells, macrophages, vascular smooth muscle cells, and mast cells. NFATc1 was not evenly distributed among these types. Violin plots in Fig. 4B showed that VSMCs harbored the highest NFATc1 levels and, notably, a bimodal pattern that hints at internal heterogeneity. By contrast, B cells, fibroblasts, and inflammatory myeloid cells displayed lower, narrower distributions.

Fig. 4.

Cell type identification and NFATc1 expression characteristics in the osteosarcoma microenvironment. A Cell-type distributions from UMAP and t-SNE. B Violin plots of NFATc1 across types showing VSMC dominance. C Histogram/density of NFATc1 within VSMCs demonstrating a biphasic pattern. *p < 0.05, **p < 0.01

Focusing on the VSMC compartment, Fig. 4C reveals that while many cells carry near-zero NFATc1, a discrete fraction falls into a higher-expression stratum concentrated between approximately 1.5 and 2.5 (arbitrary units), suggesting an activated state aligned with osteogenic reprogramming. The separation between low and high subpopulations supports the view that VSMCs transit through distinct transcriptional states rather than along a simple linear gradient.

Expression pattern and coregulatory network analysis of NFATc1 in vascular smooth muscle cells

At a tissue-wide level, NFATc1 exhibited a graded expression landscape with localized peaks across the single-cell atlas (Fig. 5A), reaching a maximum near 3.0 in normalized units. Within the VSMC subset, Fig. 5B shows that higher-expressing cells form spatial neighborhoods with values typically ranging from 1.0 to 2.0, implying that functionally aligned VSMCs cluster within similar regions of the manifold. Co-expression mapping (Fig. 5C) identified transcripts that track closely with NFATc1, including APOE, CCS, HSD17B12, UTS2, and SERGEF. The functions of these partners—lipid processing, copper handling, and signaling—are compatible with the metabolic and structural reprogramming expected during osteogenic transition.

Fig. 5.

NFATc1 expression distribution and molecular regulatory networks in vascular smooth muscle cells. A UMAP-wide NFATc1 levels (0–3). B VSMC-focused spatial pattern (1.0–2.0). C Heatmap of NFATc1 co-expression with metabolism/signaling genes. D 3D integration of cell types and NFATc1 expression. *p < 0.05, **p < 0.01

To contextualize these relationships, we integrated cell identity, position, and NFATc1 levels in a three-dimensional view (Fig. 5D). This rendering separated VSMCs from other lineages while simultaneously highlighting the internal gradient of NFATc1 intensity, offering a visual summary of how VSMC subgroups may coordinate osteogenic programs within the tumor milieu.

Analysis of the NFATc1 coexpression regulatory network and cell type-specific expression flow

Correlation analyses sharpened the outline of the NFATc1-centered network. Among the strongest associations, SERGEF correlated with NFATc1 at r = 0.673 (p < 0.001), consistent with a role in actin remodeling and morphology changes often required for lineage transition. Metabolic partners were also prominent: APOE correlated at r = 0.463 (p < 0.01), pointing to lipid homeostasis, while CCS showed r = 0.502 (p < 0.01), implicating copper trafficking and the potential for lysyl oxidase–dependent collagen maturation. The signal involving LARS (leucyl-tRNA synthetase; r = 0.459, p < 0.01) suggests that translational control may accompany transcriptional rewiring during phenotypic change.

To relate these gene-level correlations to cellular distributions, we stratified NFATc1 into low, medium, and high tiers and constructed a Sankey diagram mapping cell types to expression levels (Fig. 6B). VSMCs were unique in predominantly flowing toward the high tier, whereas B cells, fibroblasts, and inflammatory myeloid cells largely populated low-to-medium bins. This pattern argues for a cell type–specific role of NFATc1 in calcification and osteogenic reprogramming and provides a rationale for targeting NFATc1-linked nodes—potentially via exosome-based delivery systems—to selectively modulate VSMC behavior within the tumor context .

Fig. 6.

NFATc1 coexpression regulatory network and cell type expression flow analysis. A Scatter plots and correlation coefficients for APOE, SERGEF, CCS, LARS, METTL21A, and CD83 (*p < 0.05, **p < 0.01, ***p < 0.001). B Sankey flows from cell types to NFATc1 tiers (low/medium/high), highlighting VSMC specificity. *p < 0.05, **p < 0.01

qRT-PCR validation confirms key gene expression patterns

Fig. 7

Fig. 7.

qRT-PCR validation confirms key gene expression patterns. Relative mRNA levels of NFATc1, SERGEF, APOE, and CCS in vascular smooth muscle cells (VSMCs) cultured with osteosarcoma-conditioned medium (OS-CM) compared with control VSMCs under standard culture conditions. GAPDH served as the internal control. Data are presented as mean ± SD; **p < 0.01

Discussion

Our single-cell RNA-sequencing dataset points to a decisive enrichment of NFATc1 within the osteosarcoma microenvironment, with vascular smooth muscle cells (VSMCs) showing the strongest signal across all annotated lineages. Placed against the broader literature on NFATc1 as a driver of bone remodeling and osteoclast biology, this pattern is difficult to dismiss as incidental. In tumors that elaborate aberrant mineralized matrix, an NFATc1-high VSMC compartment provides a credible route to both pathological ossification and vascular calcification. Equally important is the distribution of NFATc1 within the lineage: a clear bimodality separates a low-expression majority from a distinct high-expression subset. The latter, by both expression magnitude and neighborhood structure, appears to represent cells in the midst of an osteogenic shift rather than a merely activated smooth-muscle state. That interpretation fits the histologic reality of osteosarcoma, where calcified deposits accrue alongside neoplastic bone and remodeled vasculature.

Independent validation supports this reading. In VSMCs isolated from osteosarcoma tissues, quantitative PCR confirmed significant increases in NFATc1, SERGEF, APOE, and CCS relative to paired normal vasculature (all p < 0.01), mirroring the transcriptomic contrasts. The agreement between single-cell and qRT-PCR measurements lends weight to the notion that we are observing a coordinated program, not stochastic noise or sample-handling artifact. Practically, this coherence argues that NFATc1 and its immediate partners are tractable nodes for perturbation in follow-up work.

Spatial context further refines the model. UMAP projections do not scatter NFATc1-high VSMCs randomly; instead, they coalesce into neighborhoods suggestive of shared regulatory inputs. Prior reports place NFATc1 upstream of canonical osteoblast genes—RUNX2, osterix, and alkaline phosphatase—in multiple settings [14–16]. In our data, a similar motif is plausible: NFATc1 behaves as a switch that redirects lineage behavior toward osteogenic output within the vascular compartment. Such a switch would be expected to stiffen and architecturally reinforce the tumor niche and, by coupling with pro-angiogenic signaling, generate a microenvironment permissive to growth and invasion. While we have not directly perturbed NFATc1 here, the transcriptional context and the anatomic logic are aligned.

The co-expression structure places the cytoskeleton squarely in the path of change. SERGEF shows the strongest positive correlation with NFATc1 (r = 0.673), a result that fits its function as a guanine nucleotide exchange factor acting on Rho-family GTPases. Phenotype switching in VSMCs typically demands reorganization of actin networks, altered adhesion, and remodeling of cell shape; a tight NFATc1–SERGEF coupling is therefore consistent with a two-track agenda in which transcriptional rewiring and actin dynamics proceed together. This linkage also offers a practical handle for intervention: pharmacologic or genetic disruption of upstream Rho signaling could, in principle, blunt the structural transitions that accompany osteogenic conversion.

Metabolic control emerges as a second, equally prominent axis. The correlation with APOE (r = 0.463) situates lipid handling within the osteogenic program and dovetails with evidence that imbalances in lipid flux favor ectopic mineral deposition. Copper biology enters through CCS, the copper chaperone for superoxide dismutase, which correlates with NFATc1 at r = 0.502. Copper availability modulates lysyl oxidase activity, and lysyl oxidase in turn supports cross-linking and maturation of collagenous matrix; an NFATc1–CCS connection therefore predicts a tuning of copper-dependent matrix assembly during the VSMC transition [17–19]. We also note the signal for LARS (leucyl-tRNA synthetase), which points to translational control aligning with transcriptional change. Taken together, these features argue that the osteogenic trajectory is not a single pathway event but a concerted recalibration of structural, transcriptional, and metabolic circuits.

The map of cellular states is consistent with that view. Two collagen-expressing clusters (clusters 5 and 6) show high levels of COL1A1, COL1A2, and COL3A1, marking synthetic, matrix-producing phenotypes within the VSMC lineage [20–22]. These clusters plausibly represent sequential stations along the osteogenic continuum, with increased collagen output accompanying commitment toward an osteoblast-like fate. In contrast, cluster 7, enriched for CTSK and ACP5, carries a bone-resorption signature more typically associated with osteoclast function. The juxtaposition of matrix production and resorption in the same tumor microenvironment implies ongoing remodeling rather than static deposition, with the net mechanical properties—and therefore invasive behavior—emerging from the balance of these opposing activities. An inflammatory component is evident as well: cluster 2 expresses CXCL5 and CCL2, chemokines known to recruit immune cells and stimulate angiogenesis. In this setting, inflammatory signaling, vascular remodeling, and osteogenic reprogramming appear to be braided processes rather than isolated threads.

Therapeutic implications follow naturally from the cell-type skew. Because NFATc1 is most pronounced in VSMCs, strategies aimed at dampening its activity—or attenuating key partners—may selectively reduce pathological calcification while minimizing effects on neighboring lineages. Several intervention points suggest themselves. First, the SERGEF/Rho-GTPase axis offers leverage over the cytoskeletal transitions required for phenotype change. Second, manipulating lipid flux through APOE-linked pathways or copper handling through CCS could unsettle the metabolic scaffolding that supports collagen production and mineralization. Finally, the observed role of extracellular vesicles as communication channels within the tumor microenvironment points to exosome-based delivery of inhibitory RNAs or small molecules as a way to localize therapy to the vascular compartment. Each option is testable with in-vitro calcification assays, gain- and loss-of-function perturbations (e.g., CRISPRi/CRISPRa of NFATc1 and partners), and orthotopic models read out by micro-CT and mechanical profiling.

We also caution against over-interpretation. The present work is cross-sectional and transcriptome-centric, and we have not directly demonstrated causality between NFATc1 activity and mineral deposition. Sample size is modest, and technical factors—from dissociation bias to doublet handling—can shape apparent cluster structure. Even so, several safeguards mitigate these concerns: quality metrics were stringent; batch structure was addressed during integration; results replicated across UMAP and t-SNE; and key predictions were upheld by qRT-PCR. The convergent evidence supports the central claim that an NFATc1-high VSMC subset participates in osteogenic remodeling in osteosarcoma.

In summary, the data place NFATc1 at the hub of a VSMC program that couples transcriptional switches, cytoskeletal remodeling, and metabolic reconfiguration to drive osteogenic transformation. The specific associations with SERGEF, APOE, and CCS organize this program into actionable modules—structural, lipid, and copper—that can be perturbed individually or in combination. By mapping these relationships at single-cell resolution, the study provides a framework for mechanistic experiments and a short list of targets for therapeutic exploration, with the ultimate aim of uncoupling aberrant mineralization from the vascular niche that supports osteosarcoma growth.

Limitations

This study is primarily based on transcriptomic data analysis and lacks functional validation experiments; the sample size is relatively limited, and single-time point analysis cannot fully reveal the dynamic process of VSMC osteogenic transformation.

Conclusion

This study, for the first time, reveals the highly specific expression pattern of NFATc1 in vascular smooth muscle cells within the osteosarcoma microenvironment and identifies its coexpression network with key regulatory factors such as SERGEF and APOE, providing important molecular mechanistic foundations for VSMC osteogenic transformation.

Author contributions

Conceptualization: H.F. and D.Q.; Methodology: X.Y., S.F., and C.Z.; Software and Data Analysis: X.Y., S.F., and T.Z.; Investigation: X.Y., S.F., C.Z., and T.Z.; Resources: H.F. and D.Q.; Writing—Original Draft: X.Y. and S.F.; Writing—Review & Editing: All authors; Visualization: X.Y., S.F., and T.Z.; Supervision: H.F. and D.Q.; Project Administration: H.F. and D.Q.; Funding Acquisition: H.F. and D.Q. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the following funding: the General Program of Chongqing Natural Science Fund (Grant No.: CSTB2023NSCQ-MSX0686) ,the Key Support Targets for The University’s Talent Pool in 2019 (NO. 2019R015) and the Youth Doctor Talent Incubation Program of the Second Affiliated Hospital of Army Medical University (Grant No.: 2022YQB029).

Data availability

The single-cell RNA sequencing dataset analysed during the current study is publicly available in the Gene Expression Omnibus (GEO) repository under accession number GSM5155198. Additional data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study did not involve human participants, tissues, or identifiable data. All experiments were conducted using established cell lines. Therefore, ethical approval and informed consent were not required.

Consent for publish

All the authors reviewed and approved the final manuscript.

Competing interests

The authors declare no competing interests.