Inflammation-driven prognostic model and immune landscape profiling in osteosarcoma

Abstract

Background

Osteosarcoma is the most common primary malignant bone tumor in adolescents and young adults, and its prognosis remains poor, particularly in metastatic cases. Chronic inflammation within the tumor microenvironment promotes disease progression and immune evasion, yet few prognostic models incorporate inflammation‑related molecular features.

Methods

Bulk RNA-seq data and clinical annotations of osteosarcoma patients were obtained from TCGA, and a curated inflammation gene set (top 500 genes by relevance) was defined. LASSO and Cox regression analyses identified prognostic genes, from which we built a risk‑scoring model; optimal cut‑offs were set by maximally selected rank statistics. Model performance was evaluated using Kaplan–Meier survival curves and time‑dependent ROC analysis. We then constructed and calibrated a nomogram combining key genes and metastasis status. Single‑cell RNA‑seq data (GSE1624554) were processed in Seurat to map inflammation gene expression across cell types. Immune infiltration differences between risk groups were assessed via ESTIMATE and ssGSEA. Differentially expressed genes underwent GO and KEGG enrichment analysis, and potential drug repurposing candidates were explored through cMap and molecular docking with Temozolomide.

Results

The resulting 11‑gene signature stratified patients into high and low risk with markedly different overall survival (p < 0.001), achieving AUCs of 0.808, 0.883, and 0.879 at 1, 3, and 5 years, respectively. The nomogram demonstrated excellent calibration and discriminative ability. Single‑cell analysis revealed macrophage‑ and myeloid‑specific enrichment of CD163 and SAMHD1. Low‑risk tumors exhibited higher immune and stromal scores, increased CD8⁺ T‑cell and APC activity, and enrichment of cytokine‑related pathways. Pan‑cancer assessment highlighted context‑dependent roles for PPARG, TERT, and VEGFA. Molecular docking predicted a favorable binding energy (-6.8 kcal/mol) between TERT and Temozolomide.

Conclusions

This inflammation-related risk model provides a novel prognostic tool for osteosarcoma, elucidates the interplay between tumor inflammation and immune infiltration, and suggests potential therapeutic targets and drug repurposing strategies.

Introduction

Osteosarcoma stands as the most prevalent primary malignant bone tumor, predominantly affecting adolescents and young adults [1, 2]. Its aggressive nature and high propensity for metastasis, particularly to the lungs, render it a formidable challenge in oncology [3, 4]. Despite advancements in multimodal treatments combining surgery with neoadjuvant and adjuvant chemotherapy, the prognosis for patients, especially those with metastatic or recurrent disease, remains dismal [5–7]. Over the past few decades, the survival rates for osteosarcoma have plateaued, highlighting a critical need for novel therapeutic strategies and prognostic models to improve patient outcomes [8, 9].

The complexity of osteosarcoma, characterized by its heterogeneous genetic landscape, underscores the necessity for a deeper understanding of its molecular underpinnings [10, 11]. Recent breakthroughs in high-throughput sequencing technologies have unveiled a wealth of genomic alterations and dysregulated pathways in osteosarcoma, offering new vistas for targeted therapies [12, 13]. However, the translation of these molecular insights into clinical practice demands rigorous validation and a comprehensive understanding of their implications on disease progression and patient prognosis [14].

Inflammation, a hallmark of cancer, plays a pivotal role in the initiation, progression, and metastasis of various malignancies, including osteosarcoma [15, 16]. The tumor microenvironment, rich in inflammatory cells, cytokines, and chemokines, not only supports tumor growth but also facilitates immune evasion and resistance to therapies [17, 18]. In this context, the exploration of inflammation-related genes and pathways in osteosarcoma could unveil novel biomarkers for prognosis and targets for therapeutic intervention. Furthermore, the crosstalk between tumor cells and the immune system within the tumor microenvironment presents a compelling avenue for immunotherapeutic strategies, which have revolutionized the treatment landscape for several cancers but remain underexplored in osteosarcoma.

The advent of single-cell RNA sequencing (scRNA-seq) has provided an unprecedented resolution to dissect the cellular heterogeneity within tumors, including osteosarcoma [19, 20]. This technology enables the identification of distinct cell populations, their functional states, and interactions within the tumor microenvironment, offering insights into the mechanisms driving tumor progression and immune evasion [21, 22]. Moreover, scRNA-seq facilitates the delineation of the immune landscape in osteosarcoma, providing a foundation for the development of immunotherapies tailored to target specific immune cell populations or modulate their functions to enhance anti-tumor responses [23–25].

Given the intricate relationship between inflammation, immune infiltration, and osteosarcoma progression, this study aims to construct and validate an inflammation-related risk scoring model to predict patient prognosis. By integrating bulk and single-cell RNA-seq data, the study seeks to unravel the molecular signatures associated with inflammation and immune cell dynamics in the osteosarcoma microenvironment. The development of such a prognostic model is predicated on the hypothesis that inflammation-related genes and their expression patterns can serve as robust biomarkers for patient stratification and risk assessment. Furthermore, the analysis of immune cell infiltration patterns in high and low-risk groups aims to shed light on the immune evasion mechanisms employed by osteosarcoma and identify potential targets for immunotherapy.

The rationale for this study is anchored in the pressing need for personalized medicine approaches in osteosarcoma treatment. The heterogeneity of the disease, both at the genetic and immune levels, necessitates the development of prognostic models that can guide therapeutic decision-making and identify patients who might benefit from targeted or immunotherapeutic interventions. By leveraging the power of high-throughput sequencing and sophisticated bioinformatic analyses, this study endeavors to contribute to the evolving landscape of precision oncology in osteosarcoma.

Method

Data acquisition

This study included bulk RNA-seq gene expression data and comprehensive clinical annotations of osteosarcoma patients from the TCGA cohort, with chip and high-throughput data downloaded accordingly. Expression and clinical information for each group were organized, retaining samples with survival information. Additionally, single-cell scRNA-seq data from the GEO public database (GSE1624554) were included for further analysis.

To identify inflammation-related genes, we searched the GeneCards database (https://www.genecards.org) using the keyword “inflammation” and selected the top 500 genes ranked by relevance score. The GeneCards relevance score integrates multi-omics and literature-based data to prioritize genes functionally associated with the queried term. This approach has been widely adopted in previous bioinformatics studies exploring inflammation-related signatures [26, 27]. The resulting gene set was used for downstream univariate Cox regression and model construction.

For model construction, we used gene expression and clinical data from 84 osteosarcoma patients available in the TCGA cohort. These patients included both male and female individuals aged between 6 and 88 years (median age: 17.5), encompassing localized and metastatic cases. All included samples had complete survival information, and patients were not stratified by treatment regimens. The use of this relatively homogeneous cohort enabled consistent survival analysis and minimized confounding variables in gene selection and model validation.

Construction and validation of a prognostic model

Lasso analysis was conducted using the glmnet package, followed by the calculation of each patient’s risk scores using the model formula: Risk score = Σi Coefficient (mRNAi)×Expression(mRNAi). The cut-off value was determined by the “surv_cutpoint” function of the R package “survminer”, which computes statistics based on maximally selected rank statistics. The function’s principle for determining the optimal cutoff value is to identify two groups with the most statistically significant difference in survival rates through multiple simulations. Patients were categorized into high and low-risk groups based on the cutoff value. The Kaplan-Meier (K-M) method was used to compare the overall survival (OS) between these groups. The prognostic model’s predictive reliability and validity were assessed via time-dependent receiver operating characteristic (ROC) analysis.

To guard against overfitting during cutoff determination, we employed both cross‑validation and bootstrap resampling. First, the TCGA cohort was randomly divided into 10 equal subsets for 10‑fold cross‑validation: in each iteration, nine subsets were used to compute the maximally selected rank statistic and derive a candidate cutoff, and the remaining subset served for validation. The most frequently occurring cutoff across all folds was adopted as the final threshold. Second, we performed 1,000 bootstrap resamples of the entire cohort, recalculating the cutoff and corresponding time‑dependent AUCs in each resample to estimate the stability of the threshold and model performance. The cutoff value and AUCs varied by less than 3% across resamples, indicating minimal overfitting and robust prognostic stratification.

Construction and evaluation of nomograms

Nomograms were constructed using the “rms” R package, based on the expression levels of key genes, to predict the incidence of the disease. The calibration of the model was evaluated using calibration plots with the ggDCA package, and the clinical utility of the model was assessed through clinical decision curve analysis. The predictive accuracy and validity of the prognostic model were evaluated using time-dependent ROC analysis with the pROC package.

Differential gene expression and functional enrichment analysis

The FindAllMarkers function of the Seurat package was used to identify differentially expressed genes (DEGs) between high and low-risk groups, with adjusted P-values < 0.05 and an absolute log fold change (logFC) > 0.585. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed on DEGs between these groups using the R package “clusterProfiler” (version 4.0.5), with a false discovery rate (FDR) < 0.05 considered significant for enrichment.

Single-cell data analysis

The Seurat package was utilized to analyze single-cell sequencing data (GSE1624554), with cell quality control criteria set to: nFeature_RNA > 100 & nFeature_RNA < 8000 & percent.mt < 25 & nCount_RNA > 1000. This ensured the acquisition of high-quality single-cell data, which was then normalized using the LogNormalize method.

Immune infiltration analysis

The ESTIMATE algorithm, through the R package “estimate”, assessed differences in stromal scores, immune scores, and ESTIMATE scores. Single-sample gene set enrichment analysis (ssGSEA) was performed using the “GSVA” R package to compare the enrichment scores, indicative of the relative abundance of tumor-infiltrating immune cell types. The results were visualized using the “limma”, “ggpubr”, and “reshape2” R packages, employing box plots, heat maps, and violin plots, respectively.

Drug function annotation and molecular docking

We uploaded the list of differentially expressed genes to cMap (http://www.broad.mit.edu/cmap) for comparison with reference gene sets in the database. Correlation scores were obtained based on the enrichment of DEGs in the reference gene sets, with values ranging from − 100 to 100. Positive scores indicate a positive correlation between DEGs and reference gene sets, while negative scores indicate a negative correlation.

Results

Establishment of the inflammation-related risk scoring model

We conducted a univariate Cox regression analysis based on the inflammation gene set, and Fig. 1 illustrates our streamlined workflow for deriving the 11‑gene prognostic signature in a single integrated view: Fig. 1A presents the univariate Cox analysis of 13 inflammation‑related candidates (all p < 0.01), with hazard ratios indicating both protective (e.g., ITGAM, HR 0.862) and risk (e.g., TERT, HR 1.291) factors; Fig. 1B simplifies these candidates’ co‑expression network via a chord diagram, highlighting strong positive links (e.g., SAMHD1–ABCB4) and inverse relationships (e.g., PPARG–CD36); Fig. 1C tracks each gene’s LASSO coefficient across log(λ), emphasizing in bold the 11 genes that retain nonzero weights at the optimal λ_min (log(λ)=–3.20); Fig. 1D plots the 10‑fold cross‑validated partial‑likelihood deviance, with dashed lines marking λ_min and the more parsimonious λ_1se, thereby transparently defining the cutoff that yields our final, robust inflammation‑driven signature; and Fig. 1E presents GO and KEGG enrichment of those 11 genes across Biological Process (e.g., regulation of endothelial cell development, negative regulation of mRNA‑mediated gene silencing), Cellular Component (e.g., endocytic vesicle, platelet alpha granule), Molecular Function (e.g., scavenger receptor activity, transcription coregulator binding), and KEGG pathways (e.g., adipocytokine signaling, PPAR signaling, lipid and atherosclerosis), with bubble size denoting gene count and color indicating adjusted p‑value. Multivariate Cox regression analysis incorporating clinical traits like age, metastasis status, and gender indicated that metastasis, PPARG, TERT, and VEGFA were closely linked to osteosarcoma prognosis (Fig. 2A). A nomogram incorporating these four factors was constructed to predict the 1, 3, and 5-year prognosis of osteosarcoma patients (Fig. 2B). Calibration curves for 1, 3, and 5 years (Figs. 2C-E) demonstrated a close match between observed and predicted survival probabilities, indicating the nomogram’s strong prognostic value.

Fig. 1.

Inflammation-Related Genes and Their Prognostic Significance in Osteosarcoma. A: Forest plot illustrating the hazard ratios of 13 inflammation-related genes identified through univariate Cox regression analysis, highlighting their association with osteosarcoma prognosis. Genes such as CD163, ITGAM, PPARG, and WAS are depicted as favorable prognostic factors, while CBS, TERT, ABCB4, CD36 are shown as adverse factors. B: Chord diagram representing significant positive correlations among selected genes (WAS, CD163, ITGAM, SAMHD1), indicating their potential synergistic roles in osteosarcoma prognosis. C-D: Lasso regression analysis outcomes, showcasing the selection process of 11 critical genes from the inflammation-related gene set. These genes’ expression patterns were further analyzed for their prognostic implications in osteosarcoma

Fig. 2.

Construction and Validation of a Prognostic Nomogram for Osteosarcoma. A: Multivariate Cox regression analysis forest plot indicating the prognostic importance of clinical traits and selected genes (PPARG, TERT, VEGFA) alongside metastasis status in osteosarcoma patients. B: Prognostic nomogram incorporating metastasis status, PPARG, TERT, and VEGFA expression levels for predicting the 1-, 3-, and 5-year survival probabilities of osteosarcoma patients. C-E: Calibration curves for the nomogram predictions at 1, 3, and 5 years, demonstrating a close alignment between the predicted and observed survival probabilities, affirming the nomogram’s predictive accuracy

Single-cell sequencing analysis

Based on the GSE1624554 cohort, single-cell sequencing analysis was conducted, with results in Fig. 3A and B showing the smallest standard deviation at PC = 7 and a chosen clustering tree resolution of 1.2. UMAP plots in Figs. 3C-D illustrated the distribution of cell types such as myeloid cells, macrophages, stromal cells, NK cells, megakaryoblast, fibroblast cells, and endothelial cells. The heatmap in Fig. 3E displayed the enrichment levels of inflammation-related genes in these seven cell clusters, with genes like BCL2A1, S100A9, CCL3L1 significantly enriched in myeloid cells, and ACP5, SIGLEC15, ATP6V0D2 in macrophages, SFRP2, MEG3, CXCL14 in stromal cells. Figures 4A-B showed the enrichment levels of 11 inflammation-related genes associated with osteosarcoma across the seven cell types, with significant enrichment of CD163 in macrophages and myeloid cells, SAMHD1 in macrophages and myeloid cells, and TNFRSF1A in endothelial and fibroblast cells.

Fig. 3.

Single-cell RNA Sequencing Analysis of Osteosarcoma Tissue. A-B: Analysis of single-cell RNA sequencing data from the GSE1624554 cohort, highlighting the optimal principal component (PC = 7) for minimal standard deviation and the chosen resolution (1.2) for cluster analysis. C-D: UMAP plots depicting the distribution of various cell types within the osteosarcoma microenvironment, including myeloid cells, macrophages, stromal cells, NK cells, megakaryoblasts, fibroblast cells, and endothelial cells, illustrating the cellular heterogeneity. E: Heatmap showing the enrichment levels of inflammation-related genes across seven identified cell clusters, revealing significant gene expression patterns in myeloid cells, macrophages, and stromal cells, which may contribute to the tumor microenvironment’s complexity

Fig. 4.

Distribution of Inflammation-Related Genes Across Identified Cell Clusters. A-B: Analysis of the expression levels of 11 key inflammation-related genes within the seven cell types identified in the osteosarcoma microenvironment. The figures highlight the significant enrichment of CD163 and SAMHD1 in macrophages and myeloid cells, and TNFRSF1A in endothelial and fibroblast cells, underscoring their potential roles in modulating immune responses within the tumor microenvironment

Prognostic and functional enrichment analysis between high and low-risk groups

Osteosarcoma patients were divided into high and low-risk groups based on their risk scores. The Kaplan-Meier (KM) curve in Fig. 5A indicated significant prognostic differences between the groups, with the low-risk group showing significantly better prognosis than the high-risk group. The ROC curve in Fig. 5B demonstrated that the risk scoring model had AUC values of 0.808, 0.883, and 0.879 for predicting 1, 3, and 5-year survival, respectively. The survival status scatter plot in Fig. 5C revealed that patients in the low-risk group tended to have a survival status, while those in the high-risk group were more likely to be deceased. Decision Curve Analysis (DCA) in Figs. 5D-F showed that the risk scoring model had a higher net benefit for predicting 1, 3, and 5-year overall survival (OS) when the threshold probability exceeded 0.1, outperforming other factors like metastasis, PPARG, TERT, and VEGFA. Furthermore, differentially expressed genes (DEGs) between high and low-risk groups were identified and subjected to GO and KEGG enrichment analyses. GO enrichment analysis in Fig. 6A revealed that DEGs were mainly enriched in pathways like positive regulation of cytokine production, external side of plasma membrane, and leukocyte-mediated immunity. KEGG enrichment analysis showed DEGs were predominantly enriched in pathways such as B cell receptor signaling pathway, osteoclast differentiation, and primary immunodeficiency (Fig. 6B).

Fig. 5.

Prognostic Analysis and Risk Assessment in Osteosarcoma Patients. A: Kaplan-Meier survival curves comparing overall survival between high and low-risk groups, demonstrating significant prognostic differences with better outcomes in the low-risk group. B: Receiver operating characteristic (ROC) curves for the risk scoring model, showing area under the curve (AUC) values for 1, 3, and 5-year survival predictions, indicating high predictive accuracy. C: Scatter plot of survival status against risk score, highlighting the distribution of survival outcomes within high and low-risk groups, with deceased patients predominating in the high-risk group. D-F: Decision curve analysis (DCA) for 1, 3, and 5-year overall survival predictions, illustrating the net benefit of the risk scoring model across different threshold probabilities and its superior predictive power compared to other clinical factors

Fig. 6.

Functional Enrichment Analysis of Differentially Expressed Genes Between high and low-Risk Osteosarcoma Groups. A: Gene Ontology (GO) enrichment analysis of DEGs showcasing significant enrichment in biological processes related to the positive regulation of cytokine production, the external side of the plasma membrane, and leukocyte-mediated immunity, highlighting the immunological underpinnings that may contribute to the differences in prognosis between the high and low-risk groups. B: Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis illustrating the predominant enrichment of DEGs in critical pathways such as B cell receptor signaling, osteoclast differentiation, and primary immunodeficiency, suggesting potential mechanisms through which these genes influence osteosarcoma progression and patient outcomes

Immune infiltration analysis between high and low-risk groups

We analyzed the level of immune cell infiltration between high and low-risk groups. The results in Figs. 7A-B showed that compared to the high-risk group, the low-risk group exhibited higher levels of immune cell infiltration and enrichment in immune-related pathways, such as APC co-inhibition, APC co-stimulation, B cells, CCR, CD8 + T cells, checkpoint, cytolytic activity, and DCs. Heatmaps in Figs. 7C-E displayed the differences in immune cell and immune-related pathway enrichment levels between high and low-risk groups, based on the expression levels of TERT, VEGFA, and PPARG.

Fig. 7.

Immune Infiltration Analysis Between high and low-risk groups. A-B: Comparative analysis of immune cell infiltration and immune pathway enrichment between high and low-risk groups, showing higher levels of immune cell presence and pathway activity (e.g., APC co-inhibition, CD8 + T cells, cytolytic activity) in the low-risk group, suggesting a more robust immune response against the tumor. C-E: Heatmaps displaying the differential enrichment levels of immune cells and immune-related pathways based on the expression of TERT, VEGFA, and PPARG, illustrating the variation in immune landscape between high and low-risk groups

Pan-cancer analysis

We further investigated the expression differences of TERT, VEGFA, and PPARG genes in pan-cancer and normal tissues, as well as the relationship of these genes with immune scores and prognosis. Results in Figs. 8A-C showed that PPARG was expressed at low levels in cancers such as BRCA, CESC, COAD, HNSC, and at high levels in KIRC, KIRP, LIHC, STAD, among others. The TERT gene was highly expressed in most cancer types, while VEGFA was expressed at low levels in BRCA, CHOL, COAD, HNSC, and at high levels in PRAD, THCA, KIRP, and others. The correlation results between TERT, VEGFA, PPARG expression levels and immune scores (Figs. 8D-F) indicated that PPARG was positively correlated with immune scores in cancers like ACC, BRCA, DLBC, GBM, and negatively correlated in BLCA, COAD, STAD, UVM. VEGFR showed negative correlations with immune scores in cancer types such as ACC, BLCA, BRCA, CESC, and positive correlations in PCPG, LGG, DLBC. In contrast, TERT was negatively correlated with immune scores in most cancer types. The correlation results between TERT, VEGFA, PPARG expression levels and prognosis (Figs. 8G-I) showed that PPARG acted as an oncogene in GBM, LGG, PAAD, LIHC, and as a tumor suppressor in KIRC, KIPAN, UVM. Similarly, TERT acted as an oncogene in GBM, LGG, PAAD, LIHC, and as a tumor suppressor in THYM. VEGFA acted as an oncogene in most cancer types, such as GBM, LGG, KIPAN, KIRP, CESC.

Fig. 8.

Pan-Cancer Analysis of TERT, VEGFA, and PPARG. A-C: Expression patterns of TERT, VEGFA, and PPARG across various cancer types and their correlation with immune scores, highlighting the diverse roles of these genes in different cancer contexts, with implications for their involvement in immune regulation and cancer progression. D-F: Correlation analysis between the expression of TERT, VEGFA, PPARG, and immune scores in different cancers, showing varying patterns of positive and negative correlations, underscoring the complex interplay between these genes and the immune microenvironment. G-I: Analysis of the prognostic impact of TERT, VEGFA, and PPARG expression across multiple cancer types, revealing their oncogenic or tumor-suppressive roles in specific contexts, which could inform their potential as therapeutic targets

Molecular docking analysis

We annotated drug functions based on differentially expressed genes between high and low-risk groups (Fig. 9) and performed molecular docking analysis between TERT and the top-ranked drug Temozolomide, concluding with a binding energy of -6.8 kcal/mol.

Fig. 9.

Molecular Docking Analysis and Drug Repurposing. Drug function annotation based on DEGs between high and low-risk groups, followed by molecular docking analysis between TERT and Temozolomide, identifying a significant binding energy of -6.8 kcal/mol, suggesting the potential for repurposing Temozolomide in the treatment of osteosarcoma based on molecular compatibility

Discussion

The establishment of the inflammation-related risk scoring model, predicated on a carefully curated gene set, represents a significant step forward in stratifying osteosarcoma patients based on prognosis. This model’s reliance on genes such as CD163, ITGAM, PPARG, and WAS as favorable prognostic indicators aligns with previous findings that highlight the anti-tumor properties of these genes in various cancers. For instance, PPARG’s role in inhibiting tumorigenesis through the modulation of cellular metabolism and immune response has been documented across multiple cancer types [28–30]. Similarly, CD163 and ITGAM, markers of M2 macrophages, have been associated with a more regulated inflammatory response, potentially curbing tumor-promoting inflammation.

Conversely, the identification of genes like CBS, TERT, ABCB4, and CD36 as adverse prognostic factors underscores the dual role of inflammation in cancer progression. The association of TERT with poor prognosis, in particular, echoes the extensive body of research linking telomerase activity with cancer cell immortality and aggressive disease phenotypes. The nuanced roles of these genes in osteosarcoma progression emphasize the complexity of the tumor microenvironment and the need for a multifaceted approach to treatment and prognosis.

The differential immune infiltration patterns observed between high and low-risk groups further illuminate the intricate relationship between the immune system and tumor progression. The enhanced infiltration of immune cells and the upregulation of immune-related pathways in the low-risk group suggest a more active immune surveillance mechanism, potentially contributing to better prognosis. This finding aligns with the growing consensus that a robust anti-tumor immune response, characterized by the presence of CD8 + T cells and the absence of immune checkpoints, correlates with improved outcomes in various cancers [31, 32].

The single-cell RNA sequencing analysis offers an unprecedented view of the cellular heterogeneity within osteosarcoma, revealing distinct immune cell populations and their associated gene expression profiles. The significant enrichment of inflammation-related genes in specific cell types, such as macrophages and stromal cells, underscores the pivotal role these cells play in modulating the tumor microenvironment. These insights are particularly relevant in the context of emerging immunotherapies that target specific components of the immune system to enhance anti-tumor responses.

Comparing these findings to existing studies highlights both the advancements made and the challenges that persist. While previous research has underscored the prognostic value of immune infiltration in osteosarcoma, this study further refines our understanding by delineating the specific cell types and gene signatures associated with favorable and unfavorable outcomes. However, the variability in immune responses across patients and the dynamic nature of the tumor-immune interaction necessitate personalized approaches to immunotherapy, which remain a significant challenge in the clinical setting.

The pan-cancer analysis of TERT, VEGFA, and PPARG expression not only validates the relevance of these genes in osteosarcoma but also underscores their broader significance in cancer biology. The dual roles of these genes, acting as oncogenes or tumor suppressors depending on the cancer type, reflect the complexity of gene function in the oncogenic process. This duality presents both opportunities and challenges for targeted therapies, as the same gene could potentially have opposing effects in different tumor contexts. Mechanistically, PPARG exerts tumor‑suppressive effects in osteosarcoma: for instance, the soy isoflavone genistein triggers G₂/M arrest and apoptosis in MG‑63 cells by activating PPARγ—upregulating PTEN and inhibiting PI3K/AKT—and these effects are reversed by the PPARγ antagonist GW9662 [33, 34]. TERT sustains “stem‑like” properties in osteosarcoma: sphere‑derived MG‑63 cells exhibit elevated hTERT-driven telomerase activity, enhanced sphere formation, invasiveness and chemoresistance, and targeting telomerase with the inhibitor MST312 preferentially depletes these TEL⁺ subpopulations [35]. VEGFA drives angiogenesis and survival via VEGF/PI3K/AKT signaling: adenoviral VEGF‑siRNA suppresses VEGF expression, inhibits proliferation, induces apoptosis, and impairs neovascularization both in vitro and in Wistar rat xenografts [36]. Collectively, these studies establish PPARG, TERT and VEGFA as mechanistically distinct yet clinically actionable regulators of osteosarcoma progression.

The molecular docking analysis, particularly the interaction between TERT and Temozolomide, offers promising avenues for targeted therapy. While Temozolomide’s efficacy in glioblastoma is well-documented, its potential in osteosarcoma, as suggested by the docking analysis, warrants further exploration [37, 38]. This finding exemplifies the potential of repurposing existing drugs for new therapeutic targets based on molecular compatibility. Further integration of our docking results with existing telomerase biology underscores a plausible therapeutic angle for TERT in osteosarcoma. Osteosarcoma tissues and cell lines overexpress TERT, driving aberrant telomere maintenance, promoting replicative immortality, and conferring resistance to conventional chemotherapy. Beyond sustaining telomere length, TERT has been shown to interact with Wnt/β‑catenin and NF‑κB signaling axes, thereby enhancing proliferation, invasion, and anti‑apoptotic programs in sarcoma cells. The favorable − 6.8 kcal/mol binding energy we observed between Temozolomide and the TERT catalytic pocket suggests that Temozolomide—or its derivatives—could sterically hinder TERT enzymatic activity in osteosarcoma. Indeed, telomerase inhibitors such as imetelstat (GRN163L) have demonstrated preclinical efficacy in diverse solid tumors and are under clinical evaluation, while TERT‑based peptide vaccines have elicited immune responses against TERT‑expressing malignancies. Translating these insights into osteosarcoma will require rigorous in vitro assays of telomerase activity following Temozolomide treatment, assessment of telomere shortening and cell viability in osteosarcoma cell lines, and ultimately in vivo validation in orthotopic or patient‑derived xenograft models. Such studies will clarify whether Temozolomide can be repurposed as a dual‑function agent—both damaging DNA and inhibiting telomerase—in this highly aggressive bone tumor.

Clinical applicability of the inflammation‑driven model. Beyond prognostication, our 11‑gene inflammation signature and accompanying nomogram offer a practical tool for individualized risk stratification and treatment planning in osteosarcoma. By inputting a patient’s gene expression and metastasis status, clinicians can generate real‑time 1‑, 3‑ and 5‑year survival probabilities at the bedside, identifying high‑risk individuals who may benefit from intensified neoadjuvant/adjuvant chemotherapy or enrollment in clinical trials, as exemplified by a nomogram predicting distant metastasis in osteosarcoma (Construction and validation of nomogram to predict distant metastasis in different patients with osteosarcoma) [39]. Conversely, low‑risk patients could be spared overtreatment and its attendant toxicities, in line with an autophagy‑related gene model shown to guide chemotherapy and immunotherapy decisions (Autophagy‑related gene signature model guides clinical decision making in osteosarcoma) [40]. The development of web‑based interfaces further enables seamless integration into electronic health records for shared decision‑making, as demonstrated by a risk stratification system and web‑based nomogram for postoperative survival (Risk stratification system and web‑based nomogram constructed for postoperative OS prediction in osteosarcoma) [41]. Finally, our immune‑landscape profiling suggests that low‑risk tumors—characterized by high CD8⁺ T‑cell and APC activity—may be ideal candidates for immune checkpoint blockade.

In conclusion, this study contributes significantly to our understanding of osteosarcoma’s molecular and immunological landscape, offering new insights into prognosis and potential therapeutic targets. However, the translation of these findings into clinical practice requires further validation in larger cohorts and diverse populations. The complexity of the tumor microenvironment and the dynamic nature of tumor-immune interactions underscore the need for personalized and adaptive therapeutic strategies. Future research should focus on integrating molecular and immunological markers into clinical decision-making processes, alongside exploring the therapeutic potential of targeting specific immune cell populations or pathways. The journey towards precision medicine in osteosarcoma treatment is fraught with challenges, but the insights garnered from this study illuminate a path forward, promising improved outcomes for patients afflicted with this aggressive malignancy.

Limitations

This analysis is retrospective and relies primarily on publicly available osteosarcoma transcriptomic datasets of modest size, which constrains statistical power and the stability of multivariable estimates. External validity is limited because the risk model and nomogram were trained and evaluated within the same cohort without an independent validation set; although cross-validation and bootstrap resampling were applied to mitigate overfitting, prospective, multi-center validation is required. The inflammation-related feature set was curated from database relevance scores, which may introduce selection bias; different curation strategies could yield alternative signatures. Clinical covariates (e.g., treatment intensity, surgical margins, performance status) were incomplete or unavailable and therefore not modeled, leaving potential residual confounding. Immune and stromal inferences from bulk RNA-seq (e.g., ESTIMATE, ssGSEA) are indirect and sensitive to tumor purity, and scRNA-seq annotations were drawn from a limited public dataset, which—together with batch/platform differences and lack of spatial context—may restrict generalizability of cell-type findings. Enrichment analyses (GO/KEGG) depend on existing ontologies and over-representation statistics; despite FDR control, they do not establish causal pathway activation. Pan-cancer associations involving genes such as PPARG, TERT, and VEGFA remain correlative, and in silico drug-repurposing signals (e.g., connectivity mapping and docking) do not account for pharmacokinetics, target engagement, or toxicity. Finally, mechanistic experiments to validate the biological roles of signature genes were beyond the scope of this work. Taken together, these constraints do not negate the analytical conclusions but emphasize the need for independent, prospective validation; orthogonal immunologic profiling (e.g., multiplex IHC/flow cytometry, spatial transcriptomics); and targeted experimental studies to confirm causality and therapeutic relevance before clinical translation.

Author contributions

C.S. conceived and designed the study, supervised the project, and critically revised the manuscript. X.Z. performed bulk RNA-seq data processing, prognostic model construction, survival and ROC analyses, nomogram development, and drafted the manuscript. R.Z. carried out single-cell RNA-seq analysis, immune-infiltration assessment, functional enrichment analyses, and molecular docking studies. All authors reviewed and approved the final manuscript.

Funding

This work was supported by the Zhejiang Provincial Medical and Health Science and Technology Project (Grant No. 2022504276) and the Hangzhou Municipal Health Commission Science and Technology Program (Grant No. A20210086).

Data availability

Data Availability StatementThe datasets analyzed in this study are publicly available. Bulk RNA-seq data and clinical annotations of osteosarcoma patients were retrieved from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). Single-cell RNA sequencing data were obtained from the Gene Expression Omnibus (GEO) under accession number GSE1624554 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE1624554). All data used in this study are open-access and no new datasets were generated.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All authors have read and approved the final version.

Competing interests

The authors declare no competing interests.