Abstract

   The centrosome, a vital component in mitosis in eukaryotes, plays a
   pivotal role in cancer progression by influencing the proliferation and
   differentiation of malignant cells, making it a significant therapeutic
   target. We collected genes associated with centrosomes from existing
   literature and established a prognostic model for 85 osteosarcoma
   patients from the TARGET database. Genes associated with prognosis were
   identified through univariate Cox regression. We then mitigated
   overfitting by addressing collinearity using LASSO regression.
   Ultimately, a set of five genes was selected for the model through
   multivariable Cox regression. Model performance was assessed using ROC
   curves, which yielded a training set AUC of 0.965 and a validation set
   AUC of 0.770, indicating satisfactory model performance. We further
   identified genes with differential expression in high and low-risk
   groups and conducted functional enrichment analysis using KEGG, GO,
   Progeny, GSVA, and GSEA. Results revealed significant variances in
   various immune-related pathways between high and low-risk cohorts.
   Analysis of the immune microenvironment using ssGSEA and ESTIMATE
   indicated that individuals with unfavorable prognoses had lower immune
   scores, stromal scores, and ESTIMATE scores, coupled with higher tumor
   purity. This suggests that high-risk individuals have compromised
   immune microenvironments, potentially contributing to their unfavorable
   prognoses. Additionally, drug sensitivity and molecular docking
   analysis revealed increased responsiveness to paclitaxel in high-risk
   individuals, implying its prognostic value. The JTB-encoded protein
   exhibited a negative binding energy of − 5.5 kcal/mol when interacting
   with paclitaxel, indicating its potential to enhance the patient’s
   immune microenvironment. This framework enables patient prognosis
   prediction and sheds light on paclitaxel’s mechanism in osteosarcoma
   treatment, facilitating personalized treatment approaches.

Supplementary Information

   The online version contains supplementary material available at
   10.1038/s41598-025-99419-5.

   Keywords: Osteosarcoma, Centrosome, Tumor immune microenvironment,
   Prognostic model, Individualized therapy

   Subject terms: Cancer, Computational biology and bioinformatics

Introduction

   Osteosarcoma is a malignant neoplasm that originates in the bone or
   soft tissues of the musculoskeletal system, including muscle, adipose,
   and connective tissues^[34]1. Although it predominantly affects
   adolescents and young adults, cases have been reported across all age
   groups^[35]2. The clinical presentation typically includes localized
   swelling, progressive bone pain, and pathological fractures^[36]3. The
   current standard of care for osteosarcoma is a multimodal approach
   integrating surgical resection, chemotherapy, and, in select cases,
   radiotherapy. Treatment strategies are tailored based on tumor size,
   anatomical location, histological characteristics, and the patient’s
   overall health status^[37]4. Surgical excision remains the cornerstone
   of management, aiming for complete tumor removal with negative margins.
   Neoadjuvant and adjuvant chemotherapy significantly reduce recurrence
   risk and improve overall survival by targeting micrometastatic disease.
   Radiotherapy, though less commonly employed, may be beneficial in cases
   where complete surgical resection is unfeasible or for local control of
   residual disease^[38]5. Recent advancements in molecular oncology have
   led to the integration of targeted therapies and immunotherapeutic
   approaches in osteosarcoma management. Targeted agents aim to disrupt
   key oncogenic pathways, while immunotherapies, including immune
   checkpoint inhibitors and adoptive T-cell therapies, hold promise in
   enhancing anti-tumor immune responses^[39]6. Despite these innovations,
   osteosarcoma prognosis remains highly variable, influenced by factors
   such as tumor stage at diagnosis, histological grade, presence of
   metastases, and response to therapy^[40]7. Early detection and timely
   intervention are critical in improving clinical outcomes. Patients
   diagnosed at an early stage and receiving prompt treatment achieve a
   5-year survival rate exceeding 70%, whereas delayed diagnosis and
   advanced disease presentation correlate with significantly poorer
   prognoses^[41]8. Continued research into novel therapeutic targets and
   precision medicine approaches is essential to improving survival rates
   and quality of life for patients with osteosarcoma.

   The centrosome is a cellular organelle composed of multiple proteins
   and RNA, located at the center of the nucleus, and plays a critical
   role in various biological processes, including gene expression, RNA
   processing, and cell cycle regulation^[42]9. In recent years, a growing
   body of research has highlighted the significant involvement of the
   centrosome in cancer initiation and progression^[43]10. Various aspects
   of the centrosome, such as its number, size, shape, and composition,
   exhibit distinct alterations in numerous cancer cells^[44]10.
   Concurrently, several centrosome-associated proteins and RNA molecules
   have been found to undergo abnormal expression or mutations in cancer,
   underscoring the centrosome’s pivotal role in regulating cancer cell
   proliferation, metastasis, drug resistance, and immune evasion^[45]11.
   As a result, the centrosome has emerged as a promising target for
   cancer therapy, prompting the development of novel therapeutic
   strategies aimed at centrosome-specific interventions. Approaches such
   as siRNA, antibodies, and small-molecule compounds targeting
   centrosome-related proteins and RNA have been explored in cancer
   treatment, yielding promising preliminary findings^[46]12.
   Additionally, the expression of certain centrosome proteins has been
   identified as a potential prognostic biomarker, facilitating the
   assessment of treatment response and overall prognosis in cancer
   patients^[47]13.

   The immune microenvironment comprises various components surrounding a
   tumor, including tumor cells, immune cells, endothelial cells,
   fibroblasts, and a range of signaling molecules and pathways^[48]14.
   These components interact dynamically, collectively influencing tumor
   growth, metastasis, and treatment response. The therapeutic strategy of
   harnessing immune cells and immune-related molecules—such as immune
   checkpoint inhibitors, CAR-T cell therapy, and tumor vaccines—is known
   as immunotherapy^[49]15. In recent years, tumor research has
   increasingly focused on the immune microenvironment and immunotherapy.
   On one hand, accumulating evidence suggests that the immune
   microenvironment plays a crucial role in tumor initiation, progression,
   and response to therapy. By modulating the composition and function of
   the immune microenvironment, tumor cells can evade immune surveillance,
   ultimately leading to immune therapy resistance^[50]16. On the other
   hand, immunotherapy has demonstrated remarkable efficacy in certain
   tumor types, including melanoma, lung cancer, and breast
   cancer^[51]17–[52]19. Moreover, growing evidence indicates that the
   effectiveness of immunotherapy is closely associated with the state of
   the immune microenvironment^[53]20. In some tumor types, therapeutic
   response is linked to the degree of immune cell infiltration.
   Therefore, investigating the relationship between the immune
   microenvironment and immunotherapy is essential for developing more
   effective tumor treatment strategies.

   However, there remains a paucity of literature elucidating the
   significance and impact of the centrosome on the prognosis and immune
   microenvironment of osteosarcoma. This study aims to investigate the
   role of centrosome-associated genes in osteosarcoma. Furthermore, we
   developed a novel predictive model leveraging these findings to
   identify potential biomarkers and therapeutic targets for personalized
   immunotherapy. This approach is expected to enhance prognostic
   assessment and provide valuable guidance for clinical decision-making
   in osteosarcoma treatment.

Materials and methods

Data acquisition and preprocessing

   We obtained RNA sequencing (RNA-seq) data from 88 individuals diagnosed
   with osteosarcoma through the TARGET initiative, retrieved from the
   UCSC XENA database. The raw data was originally provided in
   log[2](FPKM + 1) format, but we converted it to log[2] (TPM + 1) to
   achieve a more accurate assessment of gene expression levels. TPM
   (Transcripts Per Million) normalization accounts for variations in
   sequencing depth and gene length, making it more suitable than FPKM
   (Fragments Per Kilobase of transcript per Million mapped reads) for
   cross-sample comparisons of gene expression. This transformation
   enhances the comparability of our dataset with external validation
   cohorts and strengthens the robustness of downstream analyses. Among
   the 88 samples, 85 had complete survival and clinical information. To
   externally validate our findings, we obtained three additional
   osteosarcoma datasets ([54]GSE16091, [55]GSE21257, and [56]GSE39055)
   from the Gene Expression Omnibus (GEO) database. These datasets were
   harmonized by removing batch effects before being used as independent
   validation cohorts.Additionally, we curated a set of 726
   centrosome-associated genes (CTRGs) based on previous
   literature^[57]21, of which 690 had available expression data for
   subsequent analysis.

Constructing and evaluating a prognostic signature based on CTRGs

   By integrating transcriptomic and survival data from the TARGET
   database, we randomly partitioned the samples into a training set and a
   test set at a 7:3 ratio. To further assess the robustness of our model,
   we incorporated independent osteosarcoma datasets from the GEO database
   ([58]GSE16091, [59]GSE21257, and [60]GSE39055) as external validation
   cohorts. Initially, we performed a univariate Cox regression analysis
   to identify centrosome-associated genes (CTRGs) significantly
   associated with prognosis. Next, we applied least absolute shrinkage
   and selection operator (LASSO) regression to eliminate collinearity
   among these genes, followed by multivariate Cox regression analysis to
   construct a prognostic model. These analyses were conducted using the
   glmnet package in R. The resulting model enables prognostic assessment
   by calculating the centrosome-related model score (CRMS) for each
   sample, computed using the following formula:
   graphic file with name d33e324.gif

   To evaluate the predictive accuracy of the model, we performed receiver
   operating characteristic (ROC) curve analysis using the survivalROC
   package in R, assessing performance across both the internal test set
   and the external validation cohorts. Finally, we stratified patients
   into two risk groups based on their CRMS: the high-risk group
   (HR-group) and the low-risk group (LR-group).

Development of a nomogram for predicting overall survival

   A nomogram was constructed using the rms package in R to predict the
   overall survival of osteosarcoma (OS) patients. The model incorporated
   key clinical variables, including age, sex, metastasis status, and risk
   score. To evaluate the predictive accuracy of the nomogram,
   time-dependent calibration curves were generated, assessing its
   performance in estimating survival probabilities over time.

Functional analysis of differentially expressed CTRGs

   We performed differential gene expression analysis between the HR-group
   and LR-group using the limma package in R. Genes were considered
   differentially expressed if they met the criteria of |logFC| > 1 and
   FDR < 0.05. To investigate protein-protein interaction (PPI) networks,
   we input the differentially expressed centrosome-associated genes
   (CTRGs) into the STRING database. The resulting PPI networks were then
   visualized using Cytoscape. Subsequently, we conducted functional
   enrichment analysis using Kyoto Encyclopedia of Genes and Genomes
   (KEGG)^[61]22–[62]24 and Gene Ontology (GO) to explore the biological
   pathways associated with the differentially expressed genes. The
   results were filtered using an FDR threshold of 0.05 and visualized
   using the enrichplot and ggplot2 packages in R.

Construction of a protein–protein interaction (PPI) network

   To investigate the interactions among differentially expressed genes
   (DEGs), we constructed a protein–protein interaction (PPI) network
   using the STRING database. The confidence score threshold was set to
   0.4 to ensure reliable interactions. The resulting PPI network was then
   visualized using Cytoscape software for further analysis. To identify
   key regulatory genes, we employed the cytoHubba plugin and utilized the
   Maximal Clique Centrality (MCC) algorithm to rank genes based on their
   topological importance within the network. Genes with the highest MCC
   scores were considered potential hub genes, playing crucial roles in
   osteosarcoma progression. Additionally, we performed network topology
   analysis, assessing parameters such as degree centrality, betweenness
   centrality, and closeness centrality, to further characterize the
   functional significance of key genes within the network. This
   comprehensive analysis allowed us to identify critical genes that may
   serve as potential biomarkers or therapeutic targets for osteosarcoma.

GSVA and GSEA analysis

   We obtained standardized gene sets in KEGG and GO GMT format from the
   MsigDB database and utilized them for Gene Set Variation Analysis
   (GSVA) and Gene Set Enrichment Analysis (GSEA). For GSVA, we used the
   expression data of individual genes across samples as input to assess
   pathway enrichment status. Only pathways with a P-value < 0.05 were
   retained for further analysis. For GSEA, genes were ranked in
   descending order based on their logFC values before performing the
   analysis. Similarly, we retained only enriched pathways with a
   P-value < 0.05. Additionally, we employed the Progeny software to
   compute scores for 14 well-established cancer-associated pathways in
   the samples and examined their differences between the HR and LR
   cohorts.

Evaluating immune infiltration and characterizing the immune microenvironment

   To comprehensively assess the immune microenvironment of the samples,
   we employed multiple analytical approaches. First, we applied
   single-sample Gene Set Enrichment Analysis (ssGSEA) to quantify immune
   infiltration levels based on a set of 29 immune-related gene sets
   curated from previous literature. Next, we conducted Spearman
   correlation analysis to evaluate the relationships between target gene
   expression, risk scores, and immune-related scores. Subsequently, we
   utilized the ESTIMATE algorithm to systematically assess the immune and
   stromal components of the tumor microenvironment. Finally, we compared
   immune microenvironment differences between the HR-group and LR-group
   and visualized the results for further interpretation.

Drug response analysis

   To evaluate drug efficacy and identify potential therapeutic targets,
   we conducted a drug sensitivity analysis on 198 compounds using the
   oncoPredict package in R. We then calculated the half-maximal
   inhibitory concentration (IC50) values for each drug and compared the
   IC50 distributions between the HR-group and LR-group. To prioritize
   potential therapeutic candidates, we selected drugs with an average
   IC50 value below 1 for further investigation.

Molecular docking and molecular dynamics simulation

   The Protein Data Bank (PDB) ([63]https://www.rcsb.org/) was used to
   obtain the 3D structure of proteins encoded by the model genes. To
   prepare the protein structures for molecular docking, we used PyMOL to
   remove water molecules and organic ligands. The 2D structure of drug
   molecules was retrieved from the PubChem database
   ([64]https://pubchem.ncbi.nlm.nih.gov/).Subsequently, AutoDock software
   was employed to assess the binding affinity and predict potential
   binding sites between drug compounds and target proteins. This analysis
   aimed to explore their molecular interactions and assess the
   feasibility of drug-protein binding.

Culturing of human osteosarcoma cell line

   The 143B osteosarcoma cell line was obtained from Wuhan Procell Life
   Science and Technology Co., Ltd. (Wuhan, China). The cells were
   cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with
   10% fetal bovine serum (FBS) and 1% penicillin–streptomycin. The
   cultures were maintained in a humidified incubator at 37 °C with 5%
   CO[2] to ensure optimal growth conditions.

Gene silencing

   SiRNA targeting the JTB gene and control siRNA were acquired from
   Beijing Tsingke Biotech Co., Ltd. The 143B cell line was seeded in a
   6-well plate and subsequently transfected with siRNA using
   Lipofectamine 2000. Following the attainment of cell confluency, they
   were utilized for subsequent experiments. The siRNA sequence used to
   knock down JTB gene expression is as follows: Sense strand:
   GGAAGAGUUUGUGGUAGCA, Antisense strand: UGCUACCACAAACUCUUCC. The
   knockdown efficiency was assessed using qRT-PCR.

qRT-PCR

   TRIzol reagent was used to extract the entire RNA from osteosarcoma
   cells, followed by reverse transcription using the TransScript
   First-Stand cDNA Synthesis Kit (TaKaRa, Japan). qRT-PCR was carried out
   on an ABI-7900HT Real-Time PCR System using SYBR Green MasterMix
   (TaKaRa, Japan). The primer sequences were as follows: GAPDH forward:
   5ʹ-CGCTCTCTGCTCCTCCTGT-3ʹ; reverse: 5ʹ-ATCCGTTGACTCCGACCTA-3ʹ. JTB
   forward: 5ʹ-AATAGGCAACTCCGGCCTTC-3ʹ; reverse:
   5ʹ-AGAGGGACCTACTCCACAGG-3ʹ.

CCK-8 assay

   Cells transfected with siRNA were seeded into 96-well plates at a
   density of 5000 cells per well. Once the cells adhered and stabilized,
   the original medium was removed, and the wells were washed with PBS.
   Subsequently, 100 µL of fresh medium containing 10 µL of CCK-8 reagent
   was added to each well, followed by incubation at 37 °C. After 1 h, the
   absorbance at 450 nm was measured using a microplate reader. Cell
   viability was calculated using the following formula:
   graphic file with name d33e393.gif

EdU assay

   The EdU assay was performed to assess cell proliferation. After siRNA
   treatment, cells were seeded onto glass slides and allowed to adhere.
   Staining was carried out using the EdU Kit (Beyotime) according to the
   manufacturer’s instructions. Following incubation, the medium
   containing EdU was removed, and cells were fixed with paraformaldehyde.
   After permeabilization, DAPI staining was applied to visualize the
   nuclei. The stained cells were then observed and imaged using an
   upright fluorescence microscope.

Cell apoptosis detection

   After treatment, 143B cells were harvested for apoptosis detection. The
   culture medium was first transferred to a new centrifuge tube to retain
   any detached cells. The adherent cells were then washed with PBS, and
   the remaining medium was removed. A total of 100,000 to 500,000
   suspended cells were collected and centrifuged at 300 × g for 5 min.
   The cell pellet was resuspended in 500 µL of Annexin V-binding buffer.
   Subsequently, 5 µL of Annexin V and 5 µL of nuclear staining solution
   were added to the suspension. The mixture was gently vortexed and
   incubated at room temperature for 15–20 min in the dark. Finally, flow
   cytometry was performed to assess apoptosis.

Would healing assay

   The 143B osteosarcoma cell line was cultured in 6-well plates and
   treated with si-JTB or si-NC. Once the cells reached confluence, a
   uniform scratch was introduced using a micropipette tip to simulate a
   wound. The medium was then replaced with serum-free DMEM to prevent
   cell proliferation from influencing migration results. Microscopic
   images of the wound area were captured at 0 h, 24 h, and 48 h
   post-wounding to assess cell migration. The wound area was quantified
   using ImageJ software, and the percentage of wound closure was
   calculated at 24 h and 48 h using the following formula:
   graphic file with name d33e410.gif

Transwell migration assay

   Following treatment, 40,000 cells suspended in serum-free medium were
   seeded into the upper chamber of a Transwell insert pre-coated with
   Matrigel. The lower chamber was filled with medium containing 10% fetal
   bovine serum (FBS) to serve as a chemoattractant. After 24 h of
   incubation, cells that migrated through the membrane were fixed with
   pre-cooled formaldehyde and subsequently stained with crystal violet.
   The stained cells were then imaged and quantified using an Olympus
   microscope (Japan).

Colony formation assay

   The colony formation assay was performed to evaluate the clonogenic
   potential of cells. siRNA-treated cells were seeded into 6-well plates
   at a low density (~ 500 cells per well) and cultured for 10–14 days
   until visible colonies had formed. The culture medium was refreshed
   every 3 days. After the incubation period, the colonies were fixed with
   4% paraformaldehyde for 15 min and subsequently stained with 0.1%
   crystal violet for 30 min. The number of colonies was then counted and
   photographed to assess cell proliferation capacity.

In vivo experiments

   Four-week-old male BALB/c nude mice were purchased from Beijing
   Huafukang Laboratory Animal Center (Beijing, China) and housed at the
   Animal Center of Renmin Hospital of Wuhan University. Following
   acclimatization, 5 × 10^6 control or siRNA-treated 143B cells were
   subcutaneously implanted into the mice. Tumor volume was measured every
   5 days as previously described. On day 25 post-injection, all mice were
   anesthetized with isoflurane and euthanized using CO[2] overdose.
   Tumors were subsequently excised and weighed. All animal care and
   experimental procedures were conducted in accordance with the ethical
   guidelines and approved protocols of the Animal Ethics Committee of
   Renmin Hospital of Wuhan University.

Statistical analysis

   The Wilcoxon rank-sum test is employed to assess if there exists a
   notable distinction between two groups, whereas the Kruskal–Wallis test
   is employed to assess if there exists a notable distinction among a
   minimum of three groups of data. The statistical analysis was performed
   using version 4.2.0 of the R package. Statistically significant results
   were considered for all analysis with a significance level of less than
   0.05.The symbol “*” indicated a significance level of p < 0.05, “**”
   indicated a significance level of p < 0.01, and “***” indicated a
   significance level of p < 0.001.

Results

Development of a centrosome-related prognostic model

   Among the 690 centrosome-related genes (CTRGs), we identified 67 genes
   significantly associated with prognosis through univariate Cox
   regression analysis (Supplementary Table [65]S1). To refine the model
   and minimize redundancy, we applied LASSO regression analysis
   (Fig. [66]1A,B), ultimately selecting five key genes: SPICE1 (Spindle
   and Centriole Associated Protein 1), HOOK2 (Hook Microtubule Tethering
   Protein 2), GNAI2 (G Protein Subunit Alpha I2), JTB (Jumping
   Translocation Breakpoint), and RPS27 (Ribosomal Protein S27). These
   genes were further validated through multivariate Cox regression
   analysis, confirming their independent prognostic significance
   (Fig. [67]1C). To quantify prognostic risk, we formulated the
   Centrosome-Related Model Score (CRMS). Using the median CRMS value in
   the training cohort as a threshold, we stratified patients into a
   high-risk group (HR-group) and a low-risk group (LR-group). In the
   training set, 30 patients were assigned to each group, and Kaplan-Meier
   survival analysis demonstrated that the HR-group exhibited
   significantly poorer overall survival (OS) compared to the LR-group
   (Fig. [68]2A). The same classification approach was applied to the test
   set and the external validation cohort (GEO database), both of which
   exhibited a consistent trend, underscoring the robustness of our model
   (Fig. [69]2A). To assess the predictive accuracy of CRMS, we conducted
   receiver operating characteristic (ROC) curve analysis at 1, 3, and 5
   years using the survivalROC package. The area under the curve (AUC)
   values were: Training set: 0.850, 0.941, and 0.96, Test set: 0.739,
   0.625, and 0.770 ,GEO validation cohort: 0.867, 0.751, and 0.795. These
   results indicate that CRMS is a reliable predictor of patient prognosis
   (Fig. [70]2B). Heatmap visualization of gene expression patterns
   revealed that SPICE1, HOOK2, JTB, and RPS27 were upregulated in the
   HR-group, whereas GNAI2 was downregulated, consistent with their risk
   profiles (Fig. [71]2C). Further survival analyses identified CRMS and
   metastasis as significant factors influencing OS in univariate Cox
   regression analysis (Table [72]1), and multivariate Cox regression
   confirmed their status as independent prognostic factors (Table [73]2).
   Finally, time-dependent ROC curve analysis demonstrated that CRMS
   outperformed other clinical variables in predicting patient survival
   (Fig. [74]3A,B), reinforcing its clinical utility. These findings
   collectively indicate that we have successfully developed a robust
   centrosome-related prognostic model for osteosarcoma, which may
   facilitate risk stratification and personalized treatment strategies.

Fig. 1.

   [75]Fig. 1
   [76]Open in a new tab

   Constructing a prognostic model based on CTRGs. (A) Selection of the
   optimal tuning parameter (λ) in the LASSO Cox regression model using
   tenfold cross-validation. The vertical dashed line represents the
   optimal λ value, which minimizes the partial likelihood deviance. (B)
   LASSO coefficient profiles of the candidate genes. As the tuning
   parameter λ increases, coefficients shrink toward zero, with the final
   selected genes identified at the optimal λ. (C) Forest plot displaying
   the hazard ratios (HR) and 95% confidence intervals (CI) of the five
   CTRGs (SPICE1, HOOK2, GNAI2, JTB, RPS27) in the multivariate Cox
   regression analysis. The HR values indicate the association of each
   gene with osteosarcoma prognosis, with SPICE1, HOOK2, JTB, and RPS27
   identified as risk factors and GNAI2 as a protective factor. The global
   p-value for the model is 1.515e-07, with an Akaike Information
   Criterion (AIC) of 87 and a concordance index (C-index) of 0.89,
   indicating strong predictive performance.

Fig. 2.

   [77]Fig. 2
   [78]Open in a new tab

   Validation of the prognostic model based on CTRGs. (A) Kaplan–Meier
   survival analysis of osteosarcoma patients in the training set, test
   set, and GEO validation cohort. Patients were stratified into high-risk
   (HR) and low-risk (LR) groups based on the prognostic model. In all
   three datasets, patients in the HR-group exhibited significantly poorer
   survival compared to those in the LR-group (p < 0.001). (B) Receiver
   operating characteristic (ROC) curves assessing the predictive
   performance of the model at 1, 3, and 5 years in the training set
   (left), test set (middle), and GEO validation cohort (right). The area
   under the curve (AUC) values indicate strong prognostic ability, with
   the external GEO validation set achieving an AUC of 0.795 for 5-year
   survival. (C) Heatmaps illustrating the expression patterns of the five
   prognostic genes (SPICE1, HOOK2, GNAI2, JTB, RPS27) in the training,
   test, and GEO validation cohorts. SPICE1, HOOK2, JTB, and RPS27 are
   highly expressed in the HR-group, while GNAI2 is downregulated in the
   HR-group across all datasets.

Table 1.

   Univariate analysis of clinical parameters.
   Id         HR       HR.95 L  HR.95 H  p value
   Gender     0.589957 0.220102 1.581308 0.294171
   Age        0.999822 0.999488 1.000157 0.296933
   Metastasis 6.165921 2.27241  16.73051 0.000355
   Riskscore  1.02179  1.01114  1.032552 5.52E−05
   [79]Open in a new tab

Table 2.

   Multivariate analysis of clinical parameters.
   Id         HR       HR.95 L  HR.95 H  p value
   Gender     0.812697 0.254226 2.597982 0.726503
   Age        0.999845 0.999402 1.000289 0.494573
   Metastasis 4.151228 1.374424 12.53812 0.011607
   Riskscore  1.018611 1.007414 1.029932 0.001076
   [80]Open in a new tab

Fig. 3.

   [81]Fig. 3
   [82]Open in a new tab

   Performance evaluation and clinical application of the prognostic
   model. (A,B) Receiver operating characteristic (ROC) curves comparing
   the prognostic model’s performance with clinical factors (gender, age,
   and metastasis status) in the training set (A) and test set (B). The
   risk score-based model exhibits superior predictive performance, with
   an AUC of 0.965 in the training set and 0.770 in the test set,
   outperforming other clinical variables. (C) Nomogram integrating the
   prognostic model with clinical variables (age, gender, metastasis
   status, and risk score) to predict 1-year, 3-year, and 5-year overall
   survival (OS) in osteosarcoma patients. The scoring system allows
   individualized survival probability estimation. (D) Calibration curves
   assessing the agreement between predicted and observed OS probabilities
   at 1-year, 3-year, and 5-year time points. The nomogram predictions
   closely align with actual survival outcomes, demonstrating high
   predictive accuracy.

Development of a nomogram for survival prediction

   A nomogram was developed by integrating age, sex, and metastasis status
   into a predictive model to estimate the overall survival (OS) of
   osteosarcoma patients (Fig. [83]3C). To assess the predictive
   performance of the nomogram, calibration curves were generated at 1, 3,
   and 5 years, demonstrating a high degree of alignment between the
   predicted and observed survival outcomes (Fig. [84]3D). These findings
   suggest that our nomogram serves as a highly effective clinical tool
   for assessing survival probability and providing prognostic insights
   for individuals diagnosed with osteosarcoma.

Differentially expressed genes (DEGs) and functional assessment in risk
groups

   After stratifying patients into high-risk and low-risk groups, we
   conducted a differential expression analysis based on the criteria
   outlined in the Methods section, identifying a total of 67
   differentially expressed genes (DEGs) (Fig. [85]4A,B). To further
   explore the biological significance of these DEGs, we constructed a
   protein-protein interaction (PPI) network using the STRING database,
   which revealed extensive interconnections among the DEGs, suggesting
   their potential involvement in shared biological pathways
   (Fig. [86]4C). To identify key regulatory genes within the network, we
   employed the Maximal Clique Centrality (MCC) algorithm in CytoHubba,
   ranking genes based on their centrality in the interaction network. The
   top-ranked genes, considered pivotal regulators influencing clinical
   divergence between risk groups, are detailed in Supplement Table [87]S2
   (Fig. [88]4D). To further elucidate the functional implications of
   these DEGs, we performed Gene Ontology (GO) and Kyoto Encyclopedia of
   Genes and Genomes (KEGG) enrichment analyses. GO analysis revealed
   significant enrichment in immune-related biological processes,
   including neutrophil activation, myeloid leukocyte activation, and
   macrophage chemotaxis, indicating that immune activity may serve as a
   crucial factor distinguishing high-risk from low-risk patients
   (Fig. [89]4F). KEGG pathway analysis identified 12 significantly
   enriched pathways, such as complement and coagulation cascades,
   osteoclast differentiation, phagosome formation, and acute myeloid
   leukemia, further supporting the involvement of immune-related
   processes in osteosarcoma prognosis (Fig. [90]4E). These findings
   collectively indicate that variations in immune-associated signaling
   pathways and biological processes contribute to differences in
   prognosis between the two risk groups, suggesting that immune
   dysregulation plays a critical role in osteosarcoma progression.

Fig. 4.

   [91]Fig. 4
   [92]Open in a new tab

   Differential expression analysis, protein–protein interaction (PPI)
   network, and functional enrichment analysis of CTRGs. (A) Heatmap of
   differentially expressed genes (DEGs) between the high-risk (HR) and
   low-risk (LR) groups. Upregulated and downregulated genes are indicated
   in red and blue, respectively. (B) Volcano plot displaying the
   distribution of DEGs. Upregulated genes are highlighted in pink,
   downregulated genes in blue, and non-significant genes in gray. The
   x-axis represents the log2 fold change, while the y-axis represents the
   − log10 adjusted p-value. (C) Protein–protein interaction (PPI) network
   of DEGs constructed using STRING. Nodes represent proteins, and edges
   indicate predicted interactions. (D) Visualization of hub genes in the
   PPI network, ranked by degree of connectivity. Nodes are color-coded
   based on their centrality, with highly connected genes in red and less
   connected genes in yellow. (E) KEGG pathway enrichment analysis of
   DEGs, showing significantly enriched pathways related to immune
   response, infectious diseases, and cancer. (F) Gene Ontology (GO)
   enrichment analysis, including biological processes (BP), molecular
   functions (MF), and cellular components (CC), highlighting key immune
   and signaling pathways associated with osteosarcoma prognosis.

GSVA analysis

   Using Gene Set Variation Analysis (GSVA), we evaluated pathway and
   biological process activity for each sample and analyzed differences in
   pathway scores between the high-risk and low-risk groups. Gene sets
   from both the KEGG and GO databases were examined, revealing 36
   significantly altered pathways in KEGG and 1216 in GO. For
   visualization, we presented the top 20 most significant pathways from
   each analysis. In the KEGG results, key immune-related pathways,
   including leukocyte transendothelial migration, complement and
   coagulation cascades, cell adhesion molecules (CAMs), and natural
   killer cell-mediated cytotoxicity, were notably downregulated in the
   high-risk group (HR-group), indicating a compromised immune environment
   (Fig. [93]5A). Consistent findings were observed in the GO database,
   where the most significantly altered biological processes were
   predominantly immune-related and exhibited marked downregulation in the
   HR-group (Fig. [94]5B). To further explore pathway alterations, we
   employed the Progeny package to assess 14 classic tumor-associated
   signaling pathways. The results revealed significant differences in
   EGFR, JAK-STAT, NFκB, TNFα, and TRAIL pathways, all of which
   demonstrated decreased activity in the HR-group (Fig. [95]5C).
   Additionally, we investigated the relationships between risk scores,
   JTB expression levels, and the five tumor-related pathways, revealing
   that higher risk scores were significantly correlated with lower
   activity in these pathways (Fig. [96]5D–H). However, when examining the
   correlation between JTB expression and pathway activity, we found that
   only the TRAIL pathway exhibited a significant inverse correlation with
   JTB expression (Fig. [97]5I–M). These findings suggest substantial
   immune function differences between the high-risk and low-risk groups,
   as identified by the centrosome-related prognostic model. Moreover,
   individuals in the high-risk group exhibit significantly impaired
   immune function, further reinforcing the potential link between
   centrosome-related gene dysregulation and immune suppression in
   osteosarcoma.

Fig. 5.

   [98]Fig. 5
   [99]Open in a new tab

   GSVA-based functional enrichment analysis and correlation between the
   risk score, key pathways, and gene expression. (A,B) Heatmaps showing
   the enrichment levels of KEGG (A) and GO BP (B) pathways between the
   high-risk and low-risk groups based on Gene Set Variation Analysis
   (GSVA). The color gradient represents the relative pathway activity.
   (C) Boxplot comparing pathway scores between the high-risk and low-risk
   groups for multiple oncogenic and immune-related signaling pathways.
   Statistically significant differences are indicated (**p < 0.05,
   **p < 0.01). (D–H) Correlation analysis between the risk score and
   immune-related pathways, including EGFR, JAK-STAT, NF-κB, PI3K, and
   TGF-β signaling, derived from GSVA. Pearson correlation coefficients
   (R) and p-values are indicated in each panel. (I–M) Correlation
   analysis between JTB expression and key immune-related pathways
   identified through GSVA, demonstrating the potential involvement of JTB
   in osteosarcoma progression. Pearson correlation coefficients (R) and
   p-values are indicated.

GSEA analysis

   The Gene Set Enrichment Analysis (GSEA) revealed that a majority of
   genes downregulated in the high-risk group (HR-group) were
   significantly enriched in immune-related pathways. Analysis using the
   KEGG standard gene set indicated that pathways such as antigen
   processing and presentation, B cell receptor signaling, chemokine
   signaling, and T cell receptor signaling were predominantly associated
   with the suppressed genes in the HR-group (Fig. [100]6A,B). Similarly,
   the GO standard gene set analysis demonstrated significant
   downregulation in key immune-associated processes, including adaptive
   immune response, antigen receptor-mediated signaling, and immune
   response-regulating signaling pathways (Fig. [101]6C,D). These findings
   suggest that immune dysfunction is a prominent feature of the HR-group,
   which may contribute to poor patient prognosis.

Fig. 6.

   [102]Fig. 6
   [103]Open in a new tab

   GSEA-based enrichment analysis of pathways in high-risk and low-risk
   groups. (A,B) KEGG pathway enrichment analysis using Gene Set
   Enrichment Analysis (GSEA). The enrichment scores indicate pathway
   activity differences between high-risk and low-risk groups. (A)
   Enriched pathways in the high-risk group, including antigen processing
   and presentation, hematopoietic cell lineage, and cytokine receptor
   interactions. (B) Enriched pathways in the low-risk group, such as drug
   metabolism, tight junction signaling, and TGF-β signaling pathway.
   (C,D) GO biological process enrichment analysis based on GSEA. (C)
   Pathways enriched in the high-risk group, predominantly associated with
   sensory perception, G-protein coupled receptor activity, and stimulus
   detection. (D) Pathways enriched in the low-risk group, including
   immune response activation, positive regulation of cell activation, and
   adaptive immunity.

Immune infiltration

   Given the significant differences in immune function between the two
   risk groups, we conducted immune infiltration analyses using
   single-sample Gene Set Enrichment Analysis (ssGSEA) and ESTIMATE to
   further investigate disparities in the tumor immune microenvironment.
   The ssGSEA analysis revealed that 25 out of 29 immune-related gene sets
   exhibited significant differences between the high-risk (HR) and
   low-risk (LR) groups, with all of these sets displaying reduced scores
   in the HR-group (Fig. [104]7A,B). Correlation analysis further
   indicated a negative correlation between risk score, JTB expression
   levels, and immune function, suggesting that patients with higher risk
   scores and increased JTB expression may experience impaired immune
   function, potentially contributing to poor prognosis (Fig. [105]7C,D).
   To complement these findings, we employed ESTIMATE to compute an
   overall immune score for each sample. The results indicated that
   ESTIMATEScore, StromalScore, and ImmuneScore were significantly lower
   in the HR-group, while TumorPurity was markedly higher compared to the
   LR-group (Fig. [106]7E–H). Similarly, individuals with higher risk
   scores and elevated JTB expression exhibited lower ESTIMATE scores and
   higher tumor purity, further supporting the hypothesis of immune
   dysfunction in the HR-group (Fig. [107]7I–P). These findings suggest
   that dysregulation of the tumor immune microenvironment plays a key
   role in differential patient prognosis. The observed immune suppression
   in the HR-group highlights the potential for targeted therapeutic
   interventions aimed at restoring immune function, which could improve
   patient outcomes through immunomodulatory drugs or alternative
   strategies.

Fig. 7.

   [108]Fig. 7
   [109]Open in a new tab

   Immune infiltration analysis and correlation between immune-related
   factors and risk score. (A) Heatmap showing the distribution of immune
   cell infiltration levels between the high-risk and low-risk groups. The
   color gradient represents the relative abundance of immune cell
   populations. (B) Boxplot comparing immune infiltration scores for
   various immune cell types between the high-risk and low-risk groups.
   Significant differences are indicated (**p < 0.05, **p < 0.01,
   ***p < 0.001). (C,D) Correlation heatmaps illustrate the relationships
   between immune-related factors and (C) JTB expression and (D) risk
   score, respectively. The color scale represents the strength of
   correlation coefficients. (E–H) Violin plots showing differences in
   immune score (E), stromal score (F), ESTIMATE score (b), and tumor
   purity (H) between high-risk and low-risk groups. Higher immune and
   stromal scores indicate a more immune-infiltrated tumor
   microenvironment. (I–L) Scatter plots depicting the correlation between
   the risk score and key immune-related features, including immune score
   (I), stromal score (J), ESTIMATE score (K), and tumor purity (L).
   Pearson correlation coefficients (R) and p-values are provided. (M–P)
   Scatter plots illustrating the correlation between JTB expression and
   key immune-related features, including immune score (M), stromal score
   (N), ESTIMATE score (O), and tumor purity (P).

Sensitivity to drugs

   To assess drug responsiveness in osteosarcoma, we conducted drug
   sensitivity analysis on 198 different compounds and calculated the
   half-maximal inhibitory concentration (IC50) values for each drug. The
   results revealed significant differences in IC50 values for 31 drugs
   between the high-risk (HR) and low-risk (LR) groups (Fig. [110]8A).
   Since a higher IC50 value indicates lower drug sensitivity, we focused
   on drugs with an average IC50 value below 1, identifying Sepantronium
   bromide, Daporinad, Paclitaxel, and Sabutoclax as promising candidates
   (Fig. [111]8B–E). Among these four drugs, Paclitaxel exhibited the most
   notable predictive efficacy, with the HR-group demonstrating increased
   sensitivity to this drug, as indicated by its lower IC50 value. This
   suggests that Paclitaxel may provide greater therapeutic benefits for
   high-risk patients classified by our centrosome-based model. Further
   investigation into Paclitaxel’s mechanism of action revealed a strong
   association with the centrosome, suggesting that it may exert its
   anticancer effects by interacting with genes identified in our model.

Fig. 8.

   [112]Fig. 8
   [113]Open in a new tab

   The results of drug sensitivity and molecular docking analysis. (A)
   There were significant differences observed in the IC50 values of 31
   drugs between the two groups. (B–E) There were four drugs with IC50
   values less than 1. (F) The molecular docking analysis revealed the
   approximate binding site of paclitaxel with the JTB-encoded protein.
   (G) The results display the specific binding mode of paclitaxel with
   the JTB-encoded protein, including the amino acid residues involved in
   the binding.

Molecular docking

   Building on our previous findings, we hypothesized a potential
   correlation between JTB and the anticancer effects of paclitaxel. To
   explore this, we conducted molecular docking analysis to examine their
   potential binding interactions. AutoDock analysis identified 20
   potential interaction sites between paclitaxel and the protein encoded
   by JTB. For further investigation, we selected the binding conformation
   with the most negative binding free energy (− 5.5 kcal/mol), indicating
   a stable interaction (Fig. [114]8F). A detailed structural analysis of
   the interaction site revealed that paclitaxel binds to the JTB-encoded
   protein via hydrogen bonding, suggesting a potential regulatory effect
   (Fig. [115]8G). Based on this observation, we speculate that paclitaxel
   may exert its therapeutic effects by binding to and inactivating the
   JTB-encoded protein, thereby disrupting its function. This mechanism
   may contribute to enhancing immune responses while simultaneously
   reducing cancer cell activity and migration, offering new insights into
   the therapeutic potential of paclitaxel in osteosarcoma treatment.

Knocking down JTB suppresses cell growth and enhances apoptosis in 143B

   To assess the effects of si-JTB, we first validated JTB mRNA levels
   using qRT-PCR, which confirmed a significant reduction in JTB
   transcription following knockdown (Fig. [116]9A). To further
   investigate the functional impact of JTB suppression, we evaluated cell
   viability and proliferation in siRNA-treated 143B cells. The CCK-8
   assay demonstrated that JTB knockdown led to partial growth arrest,
   indicating a reduction in cell viability (Fig. [117]9B). Consistently,
   EdU staining revealed a marked decrease in the percentage of
   EdU-positive cells, suggesting a significant inhibition of 143B cell
   proliferation upon JTB knockdown (Fig. [118]9C,D). Additionally, flow
   cytometry analysis showed that si-JTB induced apoptosis in 143B cells,
   further confirming its role in promoting programmed cell death
   (Fig. [119]9E,F). These findings collectively suggest that JTB is
   critical for 143B cell survival, as its inhibition directly suppresses
   cell growth and enhances apoptosis, reinforcing our previous analyses
   on its role in osteosarcoma pathogenesis.

Fig. 9.

   [120]Fig. 9
   [121]Open in a new tab

   JTB knockdown inhibited proliferation and promoted apoptosis in 143B
   cells. (A) qRT-PCR results show that si-JTB significantly reduced the
   transcription levels of JTB in 143B cells compared to control and si-NC
   groups. (B) Cell viability assay demonstrating that si-JTB treatment
   markedly decreased the relative viability of 143B cells. (C)
   Statistical analysis of EDU-positive cells revealed a significant
   reduction in proliferation capacity in 143B cells treated with si-JTB.
   (D) Representative immunofluorescence images of EDU staining showing a
   decrease in proliferating 143B cells in the si-JTB group compared to
   the control and si-NC groups. (E) Flow cytometry analysis illustrating
   an increase in apoptotic 143B cells following si-JTB treatment. (F)
   Quantitative analysis of apoptotic cells from flow cytometry results
   showing a significant elevation of apoptosis in the si-JTB group.

Knocking down JTB impairs colony formation, migration, and invasion, while
suppressing tumor growth in vivo

   To examine the effects of JTB knockdown on osteosarcoma cell behavior,
   we conducted a series of functional assays. Colony formation assays
   revealed that si-JTB significantly reduced the number of colonies
   compared to the control and si-NC groups, suggesting impaired
   proliferative capacity (Fig. [122]10A, B). Similarly, Transwell
   migration assays demonstrated a significant reduction in the number of
   migrating cells in the si-JTB group, indicating decreased migratory
   ability (Fig. [123]10C,D). To further assess the impact of JTB
   knockdown on cell migration, we performed scratch wound healing assays,
   capturing images at 24- and 48-h post-wounding. The si-JTB group
   exhibited significantly lower wound closure percentages than the
   control and si-NC groups, further confirming impaired migratory
   capacity (Fig. [124]10E–G). To evaluate the role of JTB in tumor
   progression, we utilized a xenograft mouse model. Tumors in the si-JTB
   group were visibly smaller than those in the control and si-NC groups
   (Fig. [125]10H), with a significant reduction in tumor volume observed
   over time (Fig. [126]10I). Collectively, these findings demonstrate
   that JTB knockdown suppresses colony formation, migration, and invasion
   of 143B osteosarcoma cells in vitro and inhibits tumor growth in vivo,
   highlighting the critical role of JTB in osteosarcoma progression and
   its potential as a therapeutic target.

Fig. 10.

   [127]Fig. 10
   [128]Open in a new tab

   Knockdown of JTB suppressed proliferation, migration, invasion, and
   tumor growth in 143B cells. (A) Colony formation assay images
   illustrating a reduction in colony numbers in the si-JTB group compared
   to the control and si-NC groups. (B) Quantitative analysis of colony
   numbers showed a significant decrease in the si-JTB group. (C)
   Representative images of the Transwell invasion assay showing fewer
   invaded cells in the si-JTB group compared to the control and si-NC
   groups. (D) Quantitative analysis of invaded cells demonstrated a
   significant reduction in invasion ability in the si-JTB group. (E)
   Wound healing assay images at 0 h, 24 h, and 48 h showing reduced
   migration in the si-JTB group compared to the control and si-NC groups.
   (F) Statistical analysis of the wound healing percentage at 24 h
   demonstrating reduced migration in the si-JTB group. (G) Statistical
   analysis of the wound healing percentage at 48 h confirming a further
   reduction in migration ability in the si-JTB group. (H) Representative
   images of tumors harvested from mice, showing smaller tumor sizes in
   the si-JTB group compared to control and si-NC groups. (I) Quantitative
   analysis of tumor volumes demonstrating a significant reduction in the
   si-JTB group.

Discussion

   Osteosarcoma remains a significant factor affecting the healthy
   development of adolescents^[129]25. Despite its low incidence, its poor
   prognosis continues to pose a major clinical challenge. Current
   treatment strategies still rely heavily on early diagnosis, making the
   identification of novel biological markers and therapeutic targets
   crucial for improving patient outcomes^[130]26. The centrosome, a key
   cellular organelle in eukaryotic cells, has been implicated in various
   cancers^[131]27. However, its precise role and mechanism in
   osteosarcoma remain largely unexplored. In this study, we curated
   centrosome-related genes to construct a prognostic model for
   osteosarcoma patients and evaluated their survival outcomes using this
   model. Simultaneously, we conducted mechanistic analyses, immune
   microenvironment characterization, and drug sensitivity profiling to
   identify potential immunotherapy targets and viable therapeutic
   options. These findings provide clinicians with valuable insights for
   refining treatment strategies and monitoring disease progression,
   ultimately contributing to personalized osteosarcoma management.

   By constructing a centrosome-related prognostic model, we identified
   five key model genes: GNAI2, SPICE1, RPS27, HOOK2, and JTB. GNAI2 (G
   protein subunit alpha i2) encodes an α-subunit of guanine
   nucleotide-binding proteins (G proteins), which plays a role in
   hormonal regulation of adenylate cyclase by binding guanine
   nucleotides. SPICE1 (spindle and centriole associated protein 1) is
   involved in metaphase plate congression, mitotic spindle assembly, and
   centriole replication regulation. RPS27 (ribosomal protein S27) encodes
   a C4-type zinc finger domain-containing protein, which may bind to zinc
   and nucleic acids, suggesting its role in ribosomal function. HOOK2
   (hook microtubule anchoring protein 2) is a cytosolic coiled-coil
   protein that interacts with microtubules via its conserved N-terminal
   domain, while its C-terminal domain binds to organelles, facilitating
   intracellular trafficking. JTB (jumping translocation breakpoint) plays
   a role in mitotic cytokinesis and the regulation of protein kinase
   activity, localizing to the cytoplasm, midbody, centrosome, and spindle
   apparatus. By evaluating the expression profiles of these five genes,
   we stratified patients into high-risk (HR-group) and low-risk
   (LR-group) categories. Among these genes, JTB exhibited the highest
   hazard ratio (HR), suggesting a strong influence on disease progression
   and prognosis in osteosarcoma patients. Given that JTB’s role in
   osteosarcoma remains unexplored, we focused subsequent investigations
   on elucidating its biological impact in this malignancy. To further
   validate the function of JTB, we conducted a series of experimental
   assays using osteosarcoma cell lines, aiming to clarify its mechanistic
   involvement in disease progression.

   Next, we conducted functional enrichment analysis to compare the
   biological differences between the high-risk and low-risk groups,
   revealing significant disparities in immune function and associated
   biological mechanisms. Additionally, pathway scoring of conventional
   signaling pathways identified five pathways with significant
   differences, among which only the TRAIL pathway exhibited a notable
   association with JTB expression. The TRAIL pathway, also known as the
   TNF-related apoptosis-inducing ligand (TRAIL) pathway, plays a critical
   role in cellular apoptosis^[132]28. TRAIL, a ligand that induces
   programmed cell death, binds to its receptors, TRAIL-R1 (DR4) and
   TRAIL-R2 (DR5), on the cell surface, thereby triggering apoptotic
   signaling cascades^[133]29. This pathway is essential in regulating
   cancer cell survival and apoptosis, and its activation can inhibit
   tumor growth and metastasis^[134]30. Our findings suggest an inverse
   relationship between JTB expression and the TRAIL pathway, implying
   that elevated JTB levels may suppress TRAIL signaling, thereby
   enhancing tumor cell proliferation. Furthermore, previous studies have
   established a complex interplay between the TRAIL pathway and immune
   function^[135]31. TRAIL has been shown to promote the activation of
   natural killer (NK) cells and cytotoxic T lymphocytes (CTLs) in tumors,
   contributing to tumor suppression^[136]32,[137]33. Based on this
   evidence, we hypothesize that impaired immune function is a primary
   driver of poor prognosis in high-risk patients. To further explore this
   hypothesis, we analyzed the tumor immune microenvironment of
   osteosarcoma patients. Prognosis is significantly affected by dynamic
   changes in the tumor stroma and immune cell populations, which are key
   components of the tumor microenvironment^[138]34. We performed immune
   infiltration analysis using ESTIMATE and ssGSEA, revealing that
   patients with poor prognosis exhibited significantly lower
   StromalScore, ImmuneScore, and ESTIMATEScore, while TumorPurity was
   notably higher in comparison to patients with better outcomes. The
   ssGSEA analysis specifically examined immune cell composition and
   activity, identifying 25 immune cell types that significantly
   influenced osteosarcoma initiation and progression. These findings
   collectively underscore the crucial role of immune dysregulation in
   osteosarcoma prognosis and suggest that targeting immune pathways,
   particularly the TRAIL pathway, may provide therapeutic benefits for
   high-risk patients.

   To further search for potential therapeutic targets, we conducted drug
   sensitivity analysis. The IC50 of a drug refers to its half maximal
   inhibitory concentration, which is a measure of its activity^[139]35.In
   vitro drug screening, IC50 refers to the drug concentration required to
   inhibit 50% of cell growth or metabolic activity in a cell or tissue
   culture^[140]36. A lower IC50 value corresponds to higher drug
   sensitivity. We prioritized drugs with an average IC50 below 1 and,
   unexpectedly, identified paclitaxel as a promising candidate. Notably,
   patients in the high-risk (HR) group exhibited significantly lower IC50
   values for paclitaxel, indicating enhanced sensitivity to this drug.
   Paclitaxel primarily exerts its anticancer effects by disrupting
   mitosis and inhibiting tumor cell proliferation^[141]37, Given that the
   centrosome plays a crucial role in mitosis^[142]38.Therefore, we
   believe that paclitaxel may exert its effects through
   centrosome-related genes. Hence, we conducted a molecular docking
   analysis to explore the potential interaction between paclitaxel and
   the protein produced by JTB gene. Through hydrogen bonding, we
   discovered that the interaction between JTB and paclitaxel has a
   binding affinity of − 5.5 kcal/mol. The presence and role of JTB in
   cancer are linked to the emergence, progression, and spread of
   different malignant tumors^[143]39.Several research studies indicate
   that the amount of JTB expression is notably increased in specific
   types of cancer, including breast cancer, liver cancer, and colon
   cancer^[144]40,[145]41. Moreover, JTB is involved in biological
   processes such as cell proliferation, invasion, and tumor metastasis,
   making it an attractive therapeutic target. Recent studies suggest that
   JTB also influences immune cell function, contributing to tumor immune
   evasion, positioning it as a focal point in cancer immunotherapy
   research^[146]42. However, its impact on osteosarcoma remains
   unexplored. Based on our findings, we propose that paclitaxel exerts
   its effects by interacting with the JTB-encoded protein, thereby
   modulating JTB’s biological activity. Given JTB’s dual role in tumor
   cell division and immune regulation, its interaction with paclitaxel
   may have profound implications for osteosarcoma progression and patient
   prognosis. These findings provide novel insights into potential
   therapeutic strategies, suggesting that targeting JTB with paclitaxel
   could simultaneously impair tumor proliferation and restore immune
   function, thereby improving clinical outcomes for high-risk
   osteosarcoma patients.

   To validate the findings from our bioinformatics analysis, we conducted
   laboratory experiments to examine the impact of JTB gene suppression in
   143B osteosarcoma cells. Following si-JTB transfection, we observed a
   significant reduction in cell proliferation, with a portion of the
   cells undergoing apoptosis. Additionally, JTB knockdown markedly
   decreased the migratory and invasive capacities of 143B cells, further
   highlighting its role in tumor progression. Collectively, these results
   suggest that JTB plays a critical role in osteosarcoma cell
   proliferation and metastasis, reinforcing its potential as a
   therapeutic target for osteosarcoma treatment. We speculate that JTB
   may influence key tumor cell and immune cell functions, leading to
   alterations in cellular growth and metastatic behavior. The
   effectiveness of paclitaxel against osteosarcoma may be attributed to
   its ability to bind to JTB, resulting in JTB inactivation and
   subsequent activation of the TRAIL pathway, which enhances tumor cell
   apoptosis. Leveraging this mechanism may improve the therapeutic
   efficacy of paclitaxel in osteosarcoma treatment.

   However, our study has certain limitations. Due to the limited
   availability of clinical samples, we were unable to comprehensively
   investigate JTB expression alterations in human osteosarcoma and
   adjacent tissues. Additionally, our study did not fully elucidate the
   molecular mechanisms underlying JTB’s role in osteosarcoma progression
   and immune regulation. Moving forward, we aim to expand our research to
   gain a more comprehensive understanding of these mechanisms, ultimately
   facilitating the development of targeted therapeutic strategies for
   osteosarcoma.

Conclusion

   In this study, we developed a centrosome-related prognostic model for
   osteosarcoma patients, providing a novel approach for risk
   stratification. Immune infiltration analysis revealed that high-risk
   patients exhibit impaired immune function, suggesting a potential link
   between centrosome-associated genes and immune dysregulation. Drug
   sensitivity analysis identified paclitaxel as a highly effective
   therapeutic agent, with the potential to improve prognosis by targeting
   both tumor cell mitosis and immune cell function. Our findings provide
   a robust predictive model for osteosarcoma prognosis while uncovering a
   previously unrecognized pathway and therapeutic target for
   paclitaxel-based treatment. These insights offer valuable guidance for
   clinical management, paving the way for personalized therapeutic
   strategies in osteosarcoma.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [147]Supplementary Material 1^ (11.8KB, xlsx)
   [148]Supplementary Material 2^ (724B, csv)

Author contributions

   Yj.Z. and K.T. designed and conducted the experiments, analyzed the
   data, and wrote the manuscript. Bh.G. was responsible for sample
   collection, data processing, and statistical analysis. L.L. provided
   technical support and assisted in experimental design. Qj.W. supervised
   the study, provided critical revisions to the manuscript, and guided
   data interpretation. Y.H. conceptualized the study, secured funding,
   and oversaw the overall research direction and execution. All authors
   reviewed and approved the final manuscript.

Funding

   The current research is funded by the Hubei province science Foundation
   for Young Scientists (Grant No.2023AFB255).

Data availability

   The study provides access to the datasets through online repositories.
   Below, you can find the repository/repositories’ names and
   corresponding accession number(s):
   [149]https://xenabrowser.net/datapages/.

Declarations

Competing interests

   The authors declare no competing interests.

Ethics statement

   This study was conducted in accordance with the ARRIVE guidelines
   ([150]https://arriveguidelines.org). All animal procedures were
   approved by the Animal Ethics Committee of Renmin Hospital of Wuhan
   University (No. 202300255) and performed in compliance with
   institutional and national ethical standards.

Footnotes

   Publisher’s note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

   Yujian Zhong, Bohua Gao and Kai Tong contributed equally to this work.

Contributor Information

   Qingjun Wei, Email: weiqingjun@gxmu.edu.cn.

   Yong Hu, Email: hyrm2017@126.com.

References