This article is mentioned in:

Introduction

Osteosarcoma (OS) is an aggressive primary malignant bone tumor originating from abnormal osteogenic cells, which predominantly affects adolescents and young adults (1,2). Globally, it accounts for ~2% of pediatric cancers and 20% of primary bone tumors (3). Despite its relatively low incidence, the high mortality rate associated with OS is a cause for grave concern. For patients with localized OS, the 5-year survival rate is 60–70%, whereas for patients with metastatic OS, the 5-year survival rate is only 15–30% (4). Current treatment strategies primarily involve aggressive surgical resection, often necessitating limb-sparing (5) or amputation procedures (6), in conjunction with adjuvant chemotherapy (7). However, the toxicity of chemotherapy and the potential for drug resistance are formidable challenges in the management of this malignancy.

Cancer cells exploit changes in metabolic pathways as a key strategy to acquire the characteristics necessary for progression and metastasis (8). The tricarboxylic acid cycle is aided by propionate metabolism, which is a downstream process of lipid metabolism that serves as a source of energy (9). In atherosclerosis, propanoic acid, generated through intestinal cholesterol metabolism, has been shown to reduce low-density lipoprotein and total cholesterol levels in the blood, thereby alleviating atherosclerotic conditions (10). Research has suggested that flaxseed oil and vitamin E improve sperm quality and embryo development by modulating the 4-aminobutyrate aminotransferase (ABAT) gene, which plays a crucial role in propionate metabolism (11). In addition, the dysregulation of propionate metabolism in breast and lung cancer cells has been found to be associated with invasive characteristics and increased metastatic potential (12). However, the role and mechanisms of propionate metabolism in OS remain poorly understood.

The aim of the present study was to identify novel prognostic indicators for OS based on gene expression profiles and clinical data. First, bioinformatics methods were employed to explore propionate metabolism-related genes associated with OS. These genes were used to establish a prognostic model to gain deeper insights into the associations between OS and the immune microenvironment, as well as interacting pathways. Second, functional assays were performed to study the impact of key propionate metabolism-related genes in OS on cell proliferation, migration and invasion. Using this multifaceted approach, combining bioinformatics analysis with functional experiments, the present study sought to elucidate the role of propionate metabolism in OS.

Materials and methods

Data acquisition

The GSE12865 dataset from the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) was used for differential gene expression analysis (13). The GSE12865 gene expression data, comprising 12 tumor and 2 normal tissues, were derived from the GPL6244 platform. The GSE162454 single-cell RNA data, comprising primary OS tissues from six patients, were derived from the GPL24676 platform (14). In addition, clinical information and RNA sequencing results for patients with OS were obtained from the TARGET database (https://www.cancer.gov/ccg/research/genome-sequencing/target/about). Also, the MSigDB (https://www.gsea-msigdb.org/gsea/index.jsp) was searched using the keyword ‘METABOLISM’, and the ‘KEGG_PROPANOATE_METABOLISM’ collection was retrieved. The original datasets were formatted for analysis in R (version 4.0.4, r-project.org/), and the limma package was employed for data normalization and differentially expressed gene (DEG) identification within the GEO dataset. The DEG selection criteria were: Upregulation, fold change (FC) >1.3, P<0.05, and downregulation, FC <0.77, P<0.05.

Functional enrichment analysis

To elucidate protein-protein interactions (PPIs), the STRING database (version 10.5, string-db.org/) was used. Genes with interaction scores >0.4 were used to construct a PPI network among the propionate metabolism-related genes and DEGs, which were visualized and analyzed using Cytoscape software (version 3.4). DAVID (version 6.8; http://david.ncifcrf.gov/) was employed for Gene Ontology (GO) (15) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses, with the threshold set at P<0.05.

Cox proportional hazards model construction

Cox univariate regression was performed on the datasets of patients with OS from the TARGET database using the survival package in R. The glmnet package was then used to conduct regression analysis via the least absolute shrinkage and selection operator (LASSO)-Cox method, combining information on gene expression, survival status and time. The overall survival risk score was calculated as follows: Risk score=∑βi × expGenei, where expGene is the gene expression level and β is the regression coefficient obtained from the model. The median risk score was then determined to classify patients into high- and low-risk groups. The proportional hazards assumption of the Cox proportional hazards model was estimated using Kaplan-Meier analysis. Receiver operating characteristic (ROC) curves were drawn in R using the survival ROC package, and the area under the curve (AUC) was calculated. In addition, the data for each patient were analyzed and visualized using R software.

Clinical and pathological factor analysis

Using samples and clinical information of the OS cohort from the TARGET database, clinicopathological factors including the risk model, age and sex were incorporated into the risk model. Univariate and multivariate Cox analysis was subsequently performed using these clinicopathological parameters. The rms and survival packages were used to design a nomogram to predict the survival duration of of patients with OS. Finally, calibration curves and the Harrell concordance index (C-index) were evaluated to assess the predictive accuracy of the nomogram.

Characterization of the tumor microenvironment (TME)

The CIBERSORT algorithm was employed to compute the abundance of 22 types of tumor-infiltrating immune cells (TIICs) based on OS expression profiles from the TARGET database. The correlations of the levels of these TIICs with the expression levels of key genes were subsequently analyzed. In addition, the ESTIMATE tool was used to assess the stromal and immune cell infiltration of malignant tumors, and provide stromal, immunological and ESTIMATE scores for the OS TME.

Tumor immune dysfunction and exclusion (TIDE) analysis

The TIDE algorithm was applied to calculate TIDE scores (16). A lower TIDE score indicates a higher likelihood of responding to immunotherapy. Common immune checkpoint expression profiles were downloaded and examined for variations in expression among molecular subtypes (17).

Gene set enrichment analysis (GSEA) for risk subgroups

GSEA was employed to identify significantly enriched gene sets. The clusterProfiler package was used to perform the GSEA, with all candidate gene sets from the Hallmark database serving as the background set. The cutoff for significance was set at P<0.05.

Tumor immune single-cell hub (TISCH) database

The TISCH database is a repository for data on TME-related genes, which aggregates data from 27 tumor datasets across 76 cancer types, encompassing nearly 20,000 single-cell transcriptomes (18). TISCH was employed to comprehensively study the heterogeneity of the TME in diverse cell types and cancer categories.

Cell lines and maintenance

The osteocyte cell line hFOB1.19 (GNHu14) and four OS cell lines (U2OS, SCSP-5030; HOS, TCHu167; MG63, TCHu124; and SAOS-2, TCHu114) were purchased from The Cell Bank of Type Culture Collection of The Chinese Academy of Sciences and cultured in DMEM (cat. no. GNM31051-2; Gino Life & Health Holding Group Ltd.) supplemented with 10% FBS (Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml penicillin G (MedChemExpress) and 100 µg/ml streptomycin (Sigma-Aldrich). The cells were maintained at 37°C with 5% CO2 in a humidified incubator.

Reverse transcription-quantitative PCR (RT-qPCR)

Using TRIzol® reagent (Invitrogen, Beijing, China), cellular RNA was collected from the five cell lines. The RNA percentage was determined using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Inc.). According to the manufacturer's instructions, cDNA was synthesized from the RNA using a PrimeScript II 1st strand cDNA Synthesis Kit (6210A; Takara Bio, Inc.). qPCR was performed on an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems; Thermo Fisher Scientific, Inc.) with SYBR™ Green Universal Master Mix (4344463; Thermo Fisher Scientific, Inc.). The thermocycling conditions were 4 min at 94°C, followed by 40 cycles of 30 sec at 94°C, 30 sec at the annealing temperature and 30 sec at 72°C. The transcript levels of ABAT were normalized to those of GAPDH. The mRNA expression levels were calculated using the 2−ΔΔCq method (19). The primer sequences (Sangon Biotech Co., Ltd.) are listed in Table I.

Table I. Primers for reverse transcription-quantitative PCR.

Table I.

Primers for reverse transcription-quantitative PCR.

Gene	Forward (5′-3′)	Reverse (5′-3′)
ABAT	CTTCCGTCTTCATCAGAGGC	CAGCTTCCAGCACAGCTACC
GAPDH	CATGAGAAGTATGACAACAGCCT	AGTCCTTCCACGATACCAAAGT

[i] ABAT, 4-aminobutyrate aminotransferase.

Small interfering RNA (siRNA) transfection

siRNAs targeting ABAT (si-ABAT-1, si-ABAT-2 and si-ABAT-3; 10 nM) and a negative control siRNA (si-NC) were obtained from Guangzhou RiboBio Co., Ltd. The siRNA sequences are listed in Table II. Transfection of OS cells was performed using Lipofectamine® 2000 (Thermo Fisher Scientific, Inc) with 1 µg siRNA per well. After transfection for 6 h (37°C), the cell medium was replaced, and the cells were cultured for an additional 48 h before being collected for subsequent experiments. The transfection effectiveness was evaluated via RT-qPCR.

Table II. siRNA sequences.

Table II.

siRNA sequences.

Name	Sense (5′-3′)	Antisense (5′-3′)
si-ABAT-1	UGGAAGAGUUUGUGAAAGAUU	UCUUUCACAAACUCUUCCAUU
si-ABAT-2	CGGAGAACUUUGUGGAGAAUU	UUCUCCACAAAGUUCUCCGUU
si-ABAT-3	GGGAGGACCUGCUAAAUAAUU	UUAUUUAGCAGGUCCUCCCUU
NC	UUCUCCGAACGUGUCACGUTT	ACGUGACACGUUCGGAGAATT

[i] siRNA/si, short interfering RNA; ABAT, 4-aminobutyrate aminotransferase.

Protein extraction and western blotting (WB)

The transfected OS cells were lysed using RIPA buffer (P0013, Beyotime, Shanghai, China), and the total protein concentration of the lysate was determined using a BCA protein assay kit (Beyotime Institute of Biotechnology). The proteins (50 µg) were then separated by 10% SDS-PAGE and transferred to PVDF membranes (Bio-Rad Laboratories, Inc.). After blocking with 3% skimmed milk at room temperature for 1 h, the membranes were incubated with anti-ABAT (1:1,000; ab216465, Abcam) and anti-GAPDH (1:1,000; ab9485, Abcam) primary antibodies at 4°C overnight. After washing, the membranes were further incubated at room temperature for 1 h with secondary HRP-conjugated goat anti-rabbit antibodies (1:5,000; ab6721, Abcam). Odyssey imaging equipment (LI-COR Biosciences) was used to scan the blots, and Odyssey v2.0 software (LI-COR Biosciences) was used to evaluate the results using GAPDH as the internal control.

Cell viability assay

Following transient transfection for 24 h, the cells were placed in 48-well plates (4,000 cells/well). On days 1, 2, 3, 4 and 5 of culture, Cell Counting Kit-8 (CCK-8) solution (cat. no. P0038, Beyotime Institute of Biotechnology) was added to the cells and the plates were incubated at 37°C for 2 h The optical density was subsequently assessed at 450 nm using a Synergy 3 microplate reader (Biotek; Agilent Technologies, Inc.).

Transwell assays

Transwell chambers (pores, 8 µm; BD Biosciences) were used to evaluate the migration and invasion of the transfected OS cells. The transfected cells (5×104) were seeded into the upper chamber of the Transwell chamber. The upper chamber was supplemented with serum-free culture medium, whereas the medium in the lower chamber was supplemented with 10% FBS. For invasion experiments, a Matrigel (cat. no. 356234, Biolead, Beijing, China) coating was applied to the upper chamber. The Matrigel was allowed to solidify at 37°C for 30 min before the cells were added to the top chamber. The cells were cultured for 24 h at 37°C in a 5% CO2 atmosphere, fixed, stained with crystal violet (0.1%) for 15 min (room temperature) and observed under a light microscope (magnification, ×200) to assess migration and invasion. Quantification was performed using Image J software (version 1.53t).

Statistical analysis

All assays were repeated no less than three times to ensure reproducibility. Statistical analyses were performed using GraphPad Prism 8.0 (Dotmatics). Data are presented as the mean ± standard deviation. Comparisons between two groups were conducted using unpaired Student's t-tests, and multiple group comparisons were made using one-way ANOVA followed by Tukey's post hoc test. P<0.05 was considered to indicate a statistically significant result.

Results

OS-DEGs associated with propionate metabolism

In the GSE12865 gene set, the differential gene expression analysis revealed a total of 9,003 DEGs, including 4,050 upregulated and 4,953 downregulated genes (Fig. 1A). Using the MSigDB dataset, a set of 33 genes associated with propionate metabolism was obtained, and a PPI network was constructed (Fig. 1B). Venn diagram analysis identified 18 DEGs in the GSE12865 dataset that are associated with propionate metabolism (Fig. 1C). Since these genes displayed significant differences in expression between tumor and normal tissues and were associated with propionate metabolism, they were considered suitable for in-depth analysis. Using Cytoscape, a PPI network incorporating hub genes was constructed based on the aforementioned 18 genes. The hub genes comprised SUCLA2, SUCLG1, succinate-CoA ligase GDP-forming b subunit (SUCLG2), ACSS1, ACAT2, ACAT1, ECHS1, HADHA, aldehyde dehydrogenase 7 family member A1 (ALDH7A1), ALDH3A2, ACACB, aldehyde dehydrogenase 9 family member A1 (ALDH9A1), ALDH1B1, LDHA, ABAT, LDHAL6A and LDHAL6B (Fig. 1D).

Figure 1. Identification of DEGs associated with propionate metabolism. (A) Volcano plot of DEGs from the GSE12865 dataset. Blue represents downregulated genes and yellow represents upregulated genes. (B) PPI network of genes associated with propionate metabolism identified from IMSigDB. (C) Venn diagram revealing the intersection between DEGs from GSE12865 and propionate metabolism-related genes from IMSigDB. (D) PPI network of DEGs common to the GSE12865 and propionate metabolism-related gene datasets. DEGs, differentially expressed genes; PPI, protein-protein interaction; NS, not significant.

Functional enrichment analysis reveals insights into DEG functions and mechanisms

To elucidate the mechanisms of the DEGs, functional enrichment analysis was performed. KEGG analysis revealed significant associations between propionate metabolism-related DEGs and several pathways, including ‘carbon metabolism’, ‘metabolic pathways’, ‘pyruvate metabolism’ and ‘propanoate metabolism’ (Fig. 2A). GO functional enrichment analysis of the propionate metabolism related-DEGs highlighted their significant involvement in various biological processes, such as the ‘succinyl-CoA metabolic process’, ‘carbohydrate metabolic process’ and ‘fatty acid beta-oxidation’ (Fig. 2B). Additionally, these genes were associated with cellular components, including ‘succinate-CoA ligase complex’, ‘extracellular exosome’, ‘oxidoreductase complex’, ‘mitochondrial matrix’ and ‘mitochondrion’ (Fig. 2C). Furthermore, GO molecular function enrichment analysis revealed the involvement of these genes in functions including ‘identical protein binding’, ‘enoyl-CoA hydratase activity’, ‘ligase activity’, ‘aldehyde dehydrogenase (NAD) activity’ and ‘L-lactate dehydrogenase activity’ (Fig. 2D). These findings provide insights into the functional roles and mechanisms of the DEGs.

Figure 2. KEGG and GO enrichment analysis of propionate metabolism-related DEGs. Top 10 (A) KEGG, (B) GO-biological process, (C) GO-cellular component and (D) GO-molecular function enrichment analysis results of DEGs associated with propionate metabolism. KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, Gene Ontology; DEGs, differentially expressed genes.

Identification of prognostic genes and development of the risk model

Four genes were identified as prognostic genes using single-factor Cox analysis: ABAT, ALDH7A1, ALDH9A1 and SUCLG2 (Fig. 3A). The selection of these four genes with a λ-value of 0.05 was subsequently confirmed by LASSO Cox regression (Fig. 3B and C). Multivariate stepwise Cox analysis ultimately selected two genes, ABAT and ALDH7A1, for inclusion in a propionate metabolism-related gene signature. The formula for the computation of patient-specific risk scores was as follows: Risk score=(−0.2968) × ABAT + (−0.1483) × ALDH7A1. The median risk score was subsequently applied to classify patients into high- and low-risk groups (Fig. 3D). Survival curves and scatterplots of risk scores and survival curves demonstrated that deceased patients were predominantly concentrated in the high-risk group (Fig. 3D). Kaplan-Meier analysis revealed significant differences in overall survival outcomes between the two patient risk groups, indicating a worse prognosis in the high-risk group (Fig. 3E). Time-dependent ROC curves for 1-, 3- and 5-year overall survival demonstrated AUC values of 0.76, 0.75 and 0.75, respectively (Fig. 3F), highlighting the strong prognostic potential of the risk model.

Figure 3. Construction of the propionate metabolism-related gene signature based on data of patients with OS obtained from the TARGET database. (A) Cox univariate analysis of 18 propionate metabolism-related genes. (B) LASSO coefficient profiles of four propionate metabolism-related genes identified from the univariate analysis. (C) Relationship between partial likelihood deviance and log2(λ) in LASSO regression. (D) Patient risk scores, survival status and distribution of ABAT and ALDH7A1 gene expression levels. (E) Kaplan-Meier survival curve analysis of overall survival for patients with OS based on the risk score. (F) Receiver operating characteristic curve analysis of the 2-gene signature. OS, osteosarcoma; LASSO, least absolute shrinkage and selection operator; ABAT, 4-aminobutyrate aminotransferase; ALDH7A1, aldehyde dehydrogenase 7 family member A1; ALDH9A1, aldehyde dehydrogenase 9 family member A1; SUCLG2, succinate-CoA ligase GDP-forming b subunit; HR, hazard ratio; AUC, area under the curve; 95% CI, 95% confidence interval.

Immune infiltration and TIDE analysis

CIBERSORT was used to calculate the percentages of different immune cell types in the OS cohort from the TARGET database. Following profiling of the immune infiltration patterns of patients with OS, a marked difference in immune cell infiltration was detected between the high- and low-risk groups (Fig. 4A). In the TARGET-OS cohort, statistically significant differences in the infiltration of CD8+ T cells, resting and activated CD4+ memory cells, and follicular helper cells were observed between the two groups (Fig. 4B). Furthermore, ABAT levels were correlated with the infiltration of resting CD4+_memory_T cells, and ALDH7A1 was correlated with the infiltration of resting and activated CD4+ memory T cells, CD8+ T cells and follicular_helper cells (Fig. 4C). Patients in the high-risk group presented higher TIDE and exclusion scores and lower dysfunction scores compared with those in the low-risk group (Fig. 4D). In addition, higher stromal, immune and ESTIMATE scores were associated with the low-risk group (Fig. 4E).

Figure 4. Evaluation of tumor-infiltrating immune cells, immunotherapy and TME-related scores associated with the propionate metabolism-related gene signature. (A) Relative proportions of all 22 immune infiltrating cell types in the osteosarcoma cohort from the TARGET database as revealed by CIBERSORT analysis. (B) Box plots showing differences in all 22 immune cell types between high- and low-risk groups. (C) Correlation analysis of the expression levels of prognostic genes and the levels of immune infiltrating cells. (D) Differences in TIDE, dysfunction, exclusion and (E) TME-related scores calculated based on the propionate metabolism-related gene signature. *P<0.05 and **P<0.01 for high-risk vs. low-risk. TME, tumor microenvironment; TIDE, tumor immune dysfunction and exclusion.

Association of the gene signature with immune features

Following the initial demonstration of the relationship between prognosis and the constructed gene signature for OS, the impact of the gene signature on cancer immune features was further explored. GSEA revealed that the low-risk group was closely associated with the KEGG pathways ‘BUTANOATE_METABOLISM’, ‘GLYCEROLIPID_METABOLISM’, ‘LONG_TERM_POTENTIATION’, ‘FATTY_ACID_METABOLISM’, ‘SELENOAMINO_ACID_METABOLISM’ and ‘BASAL_TRANSCRIPTION_FACTORS’ (Fig. 5A-F). In addition, the levels of immune checkpoint molecules in the two groups were also investigated. The low-risk group presented increased expression of numerous genes, including BTNL2, ICOS and TNFRSF9 (Fig. 5G).

Figure 5. Propionate metabolism-related gene signature and immune checkpoint analysis. Gene set enrichment analysis of the association of the propionate metabolism-related gene risk score with the Kyoto Encyclopedia of Genes and Genomes pathways (A) ‘BUTANOATE_METABOLISM’, (B) ‘GLYCEROLIPID_METABOLISM’, (C) ‘LONG_TERM_POTENTIATION’, (D) ‘FATTY_ACID_METABOLISM’, (E) ‘SELENOAMINO_ACID_METABOLISM’ and (F) ‘BASAL_TRANSCRIPTION_FACTORS’. (G) Analysis of differences in the expression of immune checkpoint molecules between high- and low-risk groups in the osteosarcoma cohort from the TARGET database. *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001 vs. low-risk.

Prognostic factors and calibration of the risk score model

Cox regression analysis demonstrated that the risk score and primary tumor site were prognostic parameters for OS (Fig. 6A). A nomogram was subsequently generated based on the risk model and primary tumor site. According to a previous report, the c-index is 0.75, indicating a reasonable predictive value (20). Our results also demonstrate a reasonable predictive value (Fig. 6B). The calibration curves of the predictions at years 1, 3 and 5 closely aligned with the observed outcomes (Fig. 6C).

Figure 6. Independent prognostic analysis of risk scores. (A) Cox regression analysis of the osteosarcoma cohort from the TARGET database. (B) Nomogram showing the contribution of the risk score and primary tumor site to survival prediction (C-index, 0.75). (C) Calibration curves for 1-, 3- and 5-year survival.

Single-cell expression analysis of ABAT and ALDH7A1

The distributions of ABAT and ALDH7A1 expression in different types of cell in patients with OS from the GSE162454 dataset are shown in Fig. 7A and B. Further analysis of this dataset, showing the cell counts of seven cell types, is depicted in Fig. 7C, and the percentages and quantities of these TME-related cells are represented in Fig. 7D. t-SNE was applied to visualize the cell clusters, with cells having similar characteristics grouped together (Fig. 7D). These findings revealed that the predominant immune cells are monocytes/macrophages (mono/macro cells; n=16,682). In the dataset, ABAT was primarily concentrated in plasmocytes and CD4+ T conventional (CD4Tconv) cell clusters (Fig. 7E), whereas ALDH7A1 was mainly localized in mono/macro cells, endothelial cells, fibroblasts and CD8+ T exhausted cell clusters (Fig. 7F). This indicates that ABAT and ALDH7A1 may have functions in immune cells as well as in cancer cells.

Figure 7. Single-cell level analysis of prognosis-related genes. Violin plots showing the expression distribution of (A) ABAT and (B) ALDH7A1 in different types of cells in the GSE162454 OS dataset. (C) Cell types and (D) their distribution in the GSE162454 OS dataset. Distribution of (E) ABAT and (F) ALDH7A1 in different cell types in the GSE162454 OS dataset analyzed using single-cell resolution. ABAT, 4-aminobutyrate aminotransferase; ALDH7A1, aldehyde dehydrogenase 7 family member A1; CD4Tconv, CD4+ T conventional () cell clusters; CD8Tex, CD8+ T exhausted.

ABAT knockdown significantly promotes OS cell growth

To study the role of ABAT in OS, RT-qPCR analysis was performed on a panel of OS cell lines. As shown in Fig. 8A, ABAT expression was downregulated in the SAOS-2, HOS and U2OS OS cell lines compared with that in the hFOB1.19 osteoblast cell line, with the lowest expression observed in SAOS-2 cells. Therefore, follow-up experiments were conducted using the HOS and U2OS cell lines. Subsequently, RT-qPCR analysis indicated that ABAT expression in OS cells was silenced by ABAT siRNA, and WB results were consistent with these findings (Fig. 8B-E). siABAT-2 exhibited the highest transfection efficiency and so was used in subsequent experiments. CCK-8 analysis revealed that ABAT knockdown significantly increased OS cell viability compared with that in the si-NC group (Fig. 8F and G). In addition, Transwell assays demonstrated that ABAT knockdown markedly promoted OS cell migration (Fig. 8H) and invasion (Fig. 8I) compared with that in the si-NC group. In summary, these results indicated that ABAT knockdown promoted OS cell progression.

Figure 8. Silencing ABAT inhibits the proliferation, migration and invasion of OS cells. (A) Expression of ABAT in OS or osteoblast cell lines detected by RT-qPCR. (B) HOS and (C) U2OS cells were transfected with si-ABAT-1/2/3 or si-NC for 24 h, and the expression of ABAT was evaluated by RT-qPCR. (D,E) ABAT expression levels in the two cell lines after transfection were also tested by western blotting. Viability of (F) HOS and (G) U2OS OS cells measured by Cell Counting Kit-8 assay. (H) Migration and (I) invasion of HOS and U2OS cells as evaluated by Transwell assays. Magnification, ×200. *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001, vs. siNC. ABAT, 4-aminobutyrate aminotransferase; OS, osteosarcoma; RT-qPCR, reverse transcription-quantitative PCR; si, small interfering RNA; NC, negative control; OD450, optical density at 450 nm.

Discussion

Substantial advancements have been made in the comprehension of propionate metabolism, with researchers uncovering its pivotal role in tumorigenesis. For example, Gomes et al (21) reported that methylmalonic acid (MMA), a byproduct of propionate metabolism, was upregulated in the serum of elderly individuals and acted as a mediator of tumor progression. These researchers also demonstrated that MMA levels are increased via changes in propionate metabolism in cancer cells, particularly in aggressive cancers such as triple-negative breast cancers, enabling the cells to undergo prometastatic reprogramming (12). However, the role of propionate metabolism in OS cells has not been fully elucidated.

In the present study, 18 propionate metabolism-related DEGs were identified from the GSE12865 dataset. To substantiate the molecular functions of these aberrantly expressed DEGs, comprehensive functional enrichment analyses were conducted. These DEGs were found to be predominantly associated with processes associated with carbon, b-alanine and propanoate metabolism. Consistent with this, abnormal cellular metabolism in bone is known to be a key factor in the development of bone-related diseases such as OS (22). Furthermore, the DEGs were significantly enriched in GO categories including ‘tricarboxylic acid cycle’, ‘extracellular exosome’ and ‘mitochondrial matrix’. Notably, in OS research, the recruitment of bone marrow-derived macrophages has been shown to facilitate pulmonary metastasis (23). Researchers have also recognized that mitochondrial dysfunction is a crucial factor contributing to cellular metabolic disturbances in OS (24). These findings suggest that the dysregulation of energy metabolism is a prominent feature in patients with OS and is characterized by substantial downregulation of the tricarboxylic acid cycle (25).

To assess the associations between the propionate metabolism-related DEGs and clinical outcomes in patients with OS, the 18 DEGs were subjected to Cox univariate regression analysis, which led to the selection of four candidate genes. Further analysis led to the construction of a risk model comprising two of these genes: ABAT and ALDH7A1. Patients were classified into high- and low-risk categories according to this risk score, and Kaplan-Meier survival analysis revealed a striking separation in their survival curves. It is notable that among the multitude of analyzed genes associated with propionate metabolism, only two genes were identified as pivotal contributors to the construction of this gene signature. This signature has the potential to predict the prognosis and clinical outcomes of patients with OS, offering insights into the TME and therapeutic response.

Immunophenotypic analysis has been demonstrated to be valuable for the diagnosis, differential diagnosis and classification of OS cells (26). In the present study, activated CD4+ memory T cells and follicular_helper cells were present in elevated proportions in the low-risk group. Conversely, a greater proportion of resting CD4+ memory T cells was observed in the high-risk group. In addition, based on various scores calculated for the two groups and immune checkpoint marker levels, patients in the low-risk group were indicated to have a greater likelihood of responding favorably to immune checkpoint blockade therapy. Notably, previous research by Fritzsching et al (27) revealed an improvement in the survival period among patients with a relatively high CD8+ T cell/FOXP3+ T cell ratio in the OS microenvironment. Moreover, Lu et al (28) conducted an analysis of peripheral blood samples from patients with OS and found an association between follicular helper T cell activity levels and adverse prognosis.

At the single-cell level, the analysis performed in the present study revealed that mono/macro cells are predominant in OS. Prior research has elucidated the composition of the OS microenvironment, which is primarily composed of tumor-associated macrophages (TAMs) (29). TAMs play pivotal roles in the promotion of tumor growth and angiogenesis by upregulating tumorigenic and angiogenic factors (30). The enrichment of ABAT in plasmocytes and CD4Tconv cells suggests that ABAT may be involved in the regulation of immune responses and antibody production. Plasmocytes are key cells for antibody generation, whereas CD4Tconv cells are crucial for assisting immune responses and activating other immune cells (31,32). Downregulation of ABAT may impair the function of these cells, thereby weakening the host antitumor immune response. By contrast, the localization of ALDH7A1 in mono/macro cells, endothelial cells and fibroblasts suggests that it may be involved in TME remodeling and immune suppression. Macrophages in the TME often exhibit an M2 phenotype, which is associated with immunosuppression, the promotion of angiogenesis, tissue remodeling and the facilitation of tumor cell growth (33). Monocytes are recruited into the TME and differentiate into TAMs, which contribute to immunosuppression and the formation of a supportive microenvironment for tumor progression (34). Furthermore, endothelial cells and fibroblasts are critical in tumor progression and immune evasion (35). These findings highlight the potential of ALDH7A1 as a key regulator of immune modulation and suggest that it may serve as a target for novel immunotherapeutic strategies. On the basis of the findings of the current study, the prognosis of OS appears to be intricately associated with the expression levels of ABAT and ALDH7A1. ABAT has previously been implicated as a suppressor of cancer behavior, as its low expression level in hepatocellular carcinoma effectively restrains tumorigenesis (36). In addition, ABAT modulates lactate production, which inhibits the tumorigenicity of clear cell renal cell carcinoma (37). By contrast, ALDH7A1 is upregulated in pancreatic ductal adenocarcinoma and promotes tumor formation (38). ALDH7A1 also promotes the clonogenicity and migratory capabilities of human prostate cancer cells, playing a functional role in the formation of bone metastasis (39). In addition, the present study demonstrated that ABAT silencing promoted OS cell metastasis and growth, suggesting that ABAT is a suppressor gene in OS tumorigenesis.

The present study has several limitations. First, while key genes related to propionate metabolism in OS were identified through bioinformatics analysis, their regulatory roles in propionate metabolism require further experimental validation. Second, the role of ALDH7A1 in OS progression and its potential effects on tumor cell behavior were not experimentally verified. In future research, experiments on ALDH7A1 are planned to further validate the findings of the present study. In addition, the specific roles of BAT and ALDH7A1 in monocytes and macrophages within the TME warrant further exploration.

In summary, the present study provides a prognostic model based on propionate metabolism-related genes, which generates unique risk scores associated with clinical outcomes and the TME. This risk model exhibits potential for use in further research into the clinical prognosis and immunotherapeutic options for patients with OS. Two key genes, ABAT and ALDH7A1, associated with propionate metabolism in OS were identified and found to correlate with immune cell infiltration in patients with OS. These findings offer valuable insights for improving the overall management of OS.

Acknowledgements

Not applicable.

Funding

This study was supported by University Natural Science Foundation of Anhui Province (grant no, 2023AH053173) and Research Fund of Anhui Institute of Translational Medicine (grant no. 2023zhyx-C74).

Availability of data and materials

The data generated in the present study may be requested from the corresponding author.

Authors' contributions

WL, ZX, ML and NZ were responsible for conception and design of the study. SZ and JZ provided administrative support. WL and ML provided study materials. NZ, QZ, SZ and JZ collected and assembled the data. WL, ML and SZ performed data analysis and interpretation. WL, ML, NZ, SZ and JZ wrote the manuscript. WL, ZX, ZL and ML confirm the authenticity of all the raw data. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References