Identification and prognostic evaluation of differentially expressed long noncoding RNAs associated with immune infiltration in osteosarcoma

Abstract

Osteosarcoma is a malignant bone cancer that originates from the bone with the strongest invasiveness. Tumor formation strongly correlates with immune cell infiltration into the tumor immune microenvironment (TIME). Therefore, we aimed to identify TIME-related biomarkers as potential prognostic markers of osteosarcoma. The mRNA and long noncoding RNA (lncRNA) transcriptome data of 88 patients with osteosarcoma and the expression profile of GSE99671 were downloaded from The Cancer Genome Atlas and Gene Expression Omnibus, respectively. Immune infiltration scores and types were evaluated using ESTIMATE and CIBERSORT. A linear model was established to identify the differentially expressed genes (DEGs) and lncRNAs (DElncRNAs). Functional enrichment analysis of DEGs was conducted by Gene Ontology, Kyoto Encyclopedia of Genes and Genomes, gene set enrichment analysis, and gene set variation analysis. DElncRNAs were analyzed using a weighted gene co-expression network. Least absolute shrinkage and selection operator regression was applied to screen for prognostic markers. Patient survival was predicted by the risk score and analyzed by receiver operating characteristic curve. Clinical features affecting patient survival were assessed. Immune infiltration positively correlated with osteosarcoma patient survival. Different immune cell infiltrates in patients with osteosarcma may serve as prognostic indicators and targets for immunotherapy. In total, 1125 DEGs, 80 DElncRNAs, and 11 pairs of co-expressed lncRNA-mRNAs were identified. DEGs in the three modules were associated with immune infiltration into the TIME. Four DElncRNAs, namely AC015819.1, AC015911.3, AL365361.1, and USP30-AS1, showed good prognostic ability for osteosarcoma and were positively correlated with the immune score. Tumor metastasis and risk scores alone were good prognostic indicators, and a combination of the two variables can better predict the prognosis of osteosarcoma. We identified four lncRNAs, AC015819.1, AC015911.3, AL365361.1, and USP30-AS1, as potential biomarkers for osteosarcoma prognosis.

1. Introduction

Osteosarcoma is one of the most prevalent malignant bone cancers affecting children and adolescents [1]. Osteosarcoma incidence in different age groups shows bimodal changes, with the first peak occurring in children and adolescents, and the second peak occurring after 60 years of age [2]. Osteosarcoma is usually caused by the inability of osteoblasts to differentiate into mature osteoblasts, and its typical pathological manifestations include spindle cells and osteoid formation [3,4]. Further, osteosarcomas exhibit a high degree of local invasion and early metastasis [5]. Patients with osteosarcoma are at higher risk of developing a second malignancy [6]. Osteosarcoma treatment typically involves a combination of chemotherapy and surgical resection [7,8]. Nevertheless, survival has stagnated in recent years and the prognosis of patients with metastatic or recurrent osteosarcoma remains unsatisfactory, with a 5-year overall survival rate of 20% [2,9]. Therefore, an urgent need exists to understand the exact mechanism of osteosarcoma and identify potential prognostic biomarkers and new treatment targets to improve survival rates.

The tumor immune microenvironment (TIME) comprises endothelial and mesenchymal cells, extracellular matrix molecules, tumor-infiltrating immune cells, and inflammatory mediators [10]. TIME is involved in the occurrence, metastasis, and prognosis of tumors [[11], [12], [13]]. In the TIME, tumor-infiltrating immune cells constitute the main non-tumor component, which is valuable for the prognostic evaluation of osteosarcoma [14]. As reported, osteosarcoma patients with a higher content of immune cells in the TIME have a better prognosis [15]. Therefore, systematically evaluating the immune characteristics of the TIME, determining the distribution and function of tumor-infiltrating immune cells, and improving the immunotherapeutic effect on osteosarcoma are crucial.

Long noncoding RNAs (lncRNAs) are RNA molecules with more than 200 nucleotides that do not encode proteins [16]. LncRNAs, which account for approximately 80% of the human transcriptome, play an important regulatory role in various cancers by interacting with DNA, RNA, and proteins [17]. Moreover, lncRNAs are involved in various biological processes such as proliferation, apoptosis, and invasion of tumor cells [18]. Specifically, immune-associated lncRNAs influence the prognosis of osteosarcoma and are correlated with the TIME [15]. These lncRNAs regulate the TIME by regulating immune-related gene expression and disturbing immune cell infiltration into the TIME [19]. Therefore, lncRNAs play key roles in the pathogenesis of osteosarcoma.

Integrated bioinformatics analysis of microarray data is beneficial for uncovering functions and signaling processes associated with various physiological and pathological conditions [20]. Recent studies have identified osteosarcoma signaling pathways through bioinformatics analysis based on the Gene Expression Omnibus (GEO) database, providing potential diagnostic biomarkers for osteosarcoma, including serum miRNA panels [21,22]. However, public databases with extensive genomic profiling data from patients with osteosarcoma have not been thoroughly analyzed. Additionally, studies on the discovery of abnormal lncRNA expression in osteosarcoma through comprehensive bioinformatic analyses have not been reported. Therefore, this study sought to identify potential TIME-related biomarkers of osteosarcoma through bioinformatics analyses.

2. Materials and methods

2.1. Data extraction and preprocessing

The transcriptome data (mRNA and lncRNA) from sarcoma tissues of 88 osteosarcoma patients in the TARGET project were downloaded from The Cancer Genome Atlas (TCGA). Additionally, clinical characteristics and follow-up outcomes of the patients were obtained. The original count matrix and annotation files of the RNA sequencing results from transcriptome dataset GSE99671, containing osteosarcoma and normal bone tissues from reliable samples, were downloaded from GEO [23]. GSE99671 includes the mRNA transcriptome data of 18 osteosarcomas and 18 normal bone tissues.

2.2. Analysis of immune infiltration score

The ESTIMATE algorithm was employed to calculate the infiltration score of immune cells in tissues using transcriptome data. Specifically, we used the Estimate package (version 1.0.13) to calculate and compare the immune cell infiltration score based on osteosarcoma mRNA data from TCGA database across different clinical subgroups [24]. We applied the Maxstat package (version 0.7–2.5) and Cox regression to determine the cutoff value for immune scores. Subsequently, osteosarcoma patients were categorized into high and low expression groups of immune invasion based on the established cutoff value for further downstream analyses [25].

2.3. Analysis of immune cell infiltration

We used the gene expression matrix to compute the immune cell infiltration of samples using the CIBERSORTx online analysis tool, and samples wit significance level of P < 0.05 were retained for further analysis [26]. Differences in the degree of immune cell infiltration between the high and low immune score groups were analyzed using the Wilcoxon test. In instances where immune cells exhibited significant differences, we conducted a correlation analysis to evaluate the degree of infiltration among various immune cells.

2.4. Analysis of immune-related differentially expressed mRNAs/lncRNAs

The limma package (version 3.58.1) was used to evaluate differentially expressed genes (DEGs) and lncRNAs (DElncRNAs) in the high and low immune infiltration groups, employing a linear model [27]. The screening criteria for DEGs/DElncRNAs were set at P < 0.05 and |log₂FC| > 1. We employed the ggplot2 package (version 3.4.4) to generate a volcanic map of DEGs/DElncRNA to present the outcomes of differential expression analysis. Furthermore, a Wayne diagram was constructed to illustrate the immune-related DEGs obtained from TCGA and GEO datasets [28].

2.5. Co-expression analysis of DEGs and DElncRNAs

We conducted co-expression analysis of immune-related DEGs and DElncRNAs, and calculated the Pearson correlation coefficient (PCC) between lncRNAs and mRNAs. Subsequently, we selected the lncRNA-mRNA pairs with an absolute PCC value ≥ 0.9 and P < 0.05 to construct the lncRNA-mRNA co-expression network. This network was then visualized by importing it into the regulatory network visualization software, Cytoscape (version 3.10.1) [29].

2.6. Functional enrichment analyses

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed on DEGs following established protocols [[30], [31], [32]]. Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were also performed [33]. Finally, differential pathways were screened using the limma package, with a significance threshold set at P < 0.05 for screening.

2.7. Weighted gene co-expression network analysis (WGCNA)

The identified DElncRNAs were assessed using WGCNA. The DElncRNAs were analyzed for calculating the PCC of each lncRNA-mRNA pair using the WGCNA package (version 1.72–5) [34]. All the selected lncRNAs were annotated using LNCipedia (version 5.2) [35].

2.8. Least absolute shrinkage and selection operator (LASSO) analysis

Following the identification of immune-related DElncRNAs, univariate Cox regression was used to screen lncRNAs significantly associated with the survival of patients with osteosarcoma. Subsequently, LASSO regression analysis was used to further screen for prognostic markers [36]. The glmnet package (version 4.1–7) facilitated the screening process, considering combinations of diagnostic markers with a minimum coefficient of variation [37].

2.9. Risk score (RS) and prognosis prediction model construction

The formula for calculating the RS of each case was $RS=\sum _{i=1}^n{Coef}_i\times {\mathit{Exp}}_i$, where Coef represents the regression coefficient obtained from LASSO analysis, and Exp is the expression level of corresponding lncRNA (log2 converted). We used the Maxstat package (version 0.7–25) to determine the optimal cutoff value of RS for predicting osteosarcoma survival rate. Subsequently, patients with osteosarcoma were assigned to high and low-risk groups, and the survival curve was generated using the Kaplan-Meier method. We also predicted the survival time (1, 3, and 5 years) of patients using survivalROC package (version 1.0.3.1), plotted the predicted receiver operating characteristic (ROC) and calculated the area under the curve (AUC) value [38]. We used the Cox proportional hazard model to evaluate the impact of other clinical features (age, sex, and metastasis) on prognostic survival and the forestmodel package (version 0.6.2) to generate a forest plot [39]. Next, the clinical features that had significant effects on prognosis were incorporated into multivariate Cox regression analysis to evaluate the independent predictive ability of RS. Finally, we employed the rms package (version 6.7–1) to generate the nomograms and calibration curves for visualizing the multivariate model. Additionally, the consistency index (C-index) was calculated to evaluate the survival prediction ability of the nomogram.

2.10. Experimental validation

2.10.1. Cell culture and transfection

Human osteosarcoma cells (143 B, MG63, and SW1353) and human osteoblasts (hFOB 1.19) were obtained from the National Collection of Authenticated Cell Cultures (Shanghai, China) and cultured in specific media. Cells were cultured at 37 °C in a 5% CO₂ atmosphere for all experiments. Small-interfering RNA (siRNA) and a negative control (siCON) were synthesized by YouBio (Hunan, China). Lipofectamine™ 3000 (Thermo Fisher Scientific, Waltham, MA, USA) was used for cell transfection [40]. The target sequences for transient silencing are listed in Table S1.

2.10.2. Cell viability and apoptosis

Cell viability was assessed using a Cell Counting Kit-8 (CCK-8; Solarbio, Beijing, China) [41]. Apoptotic cells were assessed using the Annexin V-FITC/PI Apoptosis Detection Kit (Vazyme, Nanjing, China) [42].

2.10.3. Enzyme-linked immunosorbent assay (ELISA)

Cell supernatants were harvested and the level of TNF-α was determined by an ELISA kit (ELK1190, ELK Biotechnology, Wuhan, China) as described previously [43].

2.10.4. Quantitative real-time polymerase chain reaction (qRT-PCR)

Total RNA was extracted with TRIzol Reagent (Vazyme). Subsequently, the extracted RNA was reverse-transcribed into cDNA using the cDNA Synthesis Kit (Yeasen, Shanghai, China), and quantified with Universal SYBR Green qPCR Supermix (UElandy, Suzhou, China), sequentially [44]. The primers utilized for these procedures were synthesized by Generay (Shanghai, China) and are listed in Table S2.

2.10.5. Transwell assay

Transwell migration assays were conducted to measure cell invasion as previously described [40].

2.11. Statistical analysis

All data calculations and statistical analyses were performed using the R software. The Benjamini-Hochberg test was used for multiple test corrections. For comparisons between two groups, the independent Student's t-test or Mann-Whitney U test was applied as appropriate. All statistical P values were obtained from bilateral tests, and P < 0.05 indicated statistical significance.

3. Results

3.1. Immune microenvironment assessment and immunization grouping

We used the ESTIMATE algorithm to calculate the immune invasion score of osteosarcoma patients in TCGA database, with a median score of 121.9 and a quartile range of 80.9–160.9, and compared the differences in immune invasion scores across clinical subgroups. No difference was observed in immune infiltrations among the groups with different ages, sex, and degrees (Fig. 1A–C), whereas significant differences were observed in the death/survival groups (Fig. 1D). The cutoff value of the score calculated using the Maxstat package was 125.3. Patients were divided into high and low immune infiltration groups based on the cutoff value, and a survival curve was generated. The results showed a difference in survival time between the groups (Fig. 1E).

Fig. 1.

Comparison of immune infiltration score in different clinical subgroups and its relationship with prognosis. A–C: Comparison of immune infiltration scores among the patients of different ages, sex, and metastasis; D: Comparison of immune infiltration scores between the survival and death groups; E: The K-M survival curve of the high/low immune score groups and prognosis.

CIBERSORT was employed to examine the infiltration of different immune cells. The infiltration of 22 types of immune cells is shown in Fig. 2A, among which M1 macrophages, M2 macrophages, and monocytes showed greater infiltration in the high immune score group. However, the degree of infiltration of resting dendritic cells, resting NK cells, naïve CD4 T cells, and gamma delta T cells was relatively higher in the group with low immunity scores. We analyzed the correlation between immune cells with significant differences in the degree of infiltration between the two groups; the correlation matrix is shown in Fig. 2B. The correlation between resting NK cells and naïve CD4 T cells was positive, and the absolute value of the coefficient was large. The correlation between M1 macrophages and naïve CD4 T cells was negative, and the absolute value of the coefficient was large.

Fig. 2.

Analysis of immune cell infiltration. A: Comparison of infiltration levels of different immune cells in the high/low immune score groups. Kruskal–Wallis test was used for comparisons between groups, and "*" indicated statistically significant differences. B: Correlation matrix between immune cells with the significant difference in the degree of infiltration in the high/low immune score groups; red represents positive correlation, blue represents negative correlation (darkness of the color reflects the degree of correlation), and the cells without statistical significance are represented by black X. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.2. Differentially expressed mRNA/lncRNA analysis and mRNA-lncRNA co-expression analysis

According to high/low immune infiltration grouping, 80 DElncRNAs were screened from TCGA lncRNA expression matrix using limma package, of which 57 were upregulated and 23 were downregulated. A total of 1125 DEGs were screened from TCGA mRNA expression matrix, of which 775 were upregulated and 350 were downregulated. The volcanic diagrams of the DElncRNAs and DEGs in TCGA dataset are shown in Fig. 3A–B. In the GEO dataset, differential expression analysis was also conducted according to the high/low grouping of immune infiltration (the median immunity score was used as a cutoff value), and 176 DEGs were identified, among which 68 were the same as those identified in TCGA dataset (Fig. 3E–F).

Fig. 3.

Immune-related DEGs/DElncRNAs screening and lncRNA-mRNA co-expression analysis. Volcanic maps of DEGs (A) and DElncRNAs (B) in TCGA dataset; C: Co-expression analysis network diagram of DEGs and DElncRNAs; the yellow hexagon represents lncRNAs, and the blue circle represents mRNAs; D: Correlation matrix between co-expressed lncRNAs and mRNAs; the colored square indicates the correlation coefficient; red represents positive correlation, blue represents negative correlation (the darker the color is, the larger the correlation coefficient band); correlation coefficient and P values are marked in the box; E: Volcano map of GEO data machine GSE99671 DEGs; F: Venn diagram of intersections of DEGs in TCGA and GEO datasets. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Co-expression analysis of 1125 immune-related DEGs and 80 DElncRNAs obtained from TCGA was performed. The PCC test was used to verify the relationship between mRNA and lncRNA levels (log2 transformation) in the pairs. Finally, a total of 11 pairs of lncRNA-mRNA were selected with an absolute value of PCC >0.9 and P < 0.05. The correlation results are presented in the form of a network (Fig. 3C), and the correlation matrix of the 11 lncRNA-mRNA pairs is shown in Fig. 3D.

3.3. Functional enrichment analysis of immune-related DEGs

KEGG analysis showed that the DEGs enrichment pathways mainly included hematopoietic cell lineage, allograft rejection, viral protein interaction with cytokines and cytokine receptors, graft-versus-host disease, and cytokine-cytokine receptor interaction (Fig. 4àD). GO analysis showed that DEGs were primarily related to T-cell activation, positive regulation of cell activation, positive adjustment of leukocyte activation, regulation of immune effector processes, leukocyte cell-cell adhesion, and other biological processes associated with cell components, such as immune receptor activity, antigen binding, cytokine activity, peptide antigen binding, and MHC protein complex binding, and associated with molecular function, such as the external side of the plasma membrane, tertiary granule, MHC protein complex, secretory granule membrane, and endocytic vesicle membrane (Fig. 5A&B). The pathways enriched by GSEA mainly involved the chemokine signaling pathway, hematopoietic cell lineage, Leishmania infection, cytokine receptor interaction, T-cell receptor signaling pathway, natural killer cell-mediated cytotoxicity, systemic lupus erythematosus, and cell adhesion molecule cascades (Fig. 6A–H). The GSVA results are presented in Fig. 6I–J. The DEGs enrichment pathways of the two datasets mainly included pathogenic Escherichia coli infection, vasopressin-regulated water reabsorption, adherens junctions, neurotrophin signaling pathways, and amyotrophic lateral sclerosis.

Fig. 4.

KEGG analysis of immune-related DEGs. A: Bubble diagram of KEGG results; the closer the color is to red, the smaller P value is, and the larger the bubble is, the more DEGs are enriched in the pathway; B Bar graph of KEGG results; the closer the color is to red, the smaller the P value is, and the horizontal axis represents the number of genes enriched in the pathway; C: Network diagram of KEGG results; D: Cnetplot circle diagram of KEGG results. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 5.

GO enrichment analysis of immune-related DEGs. A: The enrichment results of biological process, cellular component, and molecular function analyzed by GO were presented in the form of a bubble graph; B: The enrichment results of biological process, cellular component, and molecular function in GO analysis were presented in a bar graph.

Fig. 6.

GSEA and GSVA analysis of immune-associated DEGs. A–H: GSEA of enrichment pathways of immune-related DEGs; P value was determined by Kolmogorov–Smirnov test; I–J: the different pathways of immune-associated DEGs in GSVA analysis are shown in the volcanic map (I) and heat map (J).

3.4. WGCNA

WGCNA was performed on the selected DElncRNAs (Fig. 7àE), and three modules that showed an association with immune cell infiltration were identified. Among these, module brown contained 30 genes, module blue contained 27 genes, and module turquoise contained 21 genes.

Fig. 7.

WGCNA. A–B: Power parameter screening process; C: Cluster analysis tree diagram of samples and corresponding immune cell infiltration; D: Gene cluster analysis tree diagram; E: The correlation matrix between the module score and the degree of differential immune cell infiltration.

3.5. Prognostic marker screening

Univariate COX regression was used to identify the correlation between DElncRNAs and patient prognosis and survival of patients, and obtained 16 lncRNAs associated with prognosis were identified. Subsequently, LASSO regression analysis was performed and four lncRNAs were identified as prognostic markers (Fig. 8A–B), namely AC015819.1, AC015911.3, AL365361.1, and USP30.AS1. Details of these lncRNAs are listed in Table 1. The predictive ROC curve of LASSO regression showed an AUC value of 0.744, indicating a good prognostic ability (Fig. 8C). We then conducted a correlation analysis of the expression of these four lncRNAs and the immune infiltration score, which revealed a significant positive correlation between the candidate lncRNAs and the immune score (Fig. 8D–H). Finally, we evaluated their association with different immune cells and found that AC015911.3, AL365361.1, and USP30.AS1 were significantly and positively correlated with M1 macrophages.

Fig. 8.

Screening of prognostic lncRNAs by LASSO regression model. A–B: LASSO logistic regression model was used to screen prognostic markers; C: ROC curves; D–G: Correlation analysis of candidate prognostic lncRNAs and immune infiltration score; H: Correlation matrix between significantly different immune cells and candidate prognostic lncRNAs.

Table 1.

Details of four lncRNAs annotated in LNCipedia.

lncRNA	Ensembl Gene ID	LNCipedia gene ID	Location (hg38)	Strand	Class	Transcript size
AC015819.1	ENSG00000273669	lnc-ZNF407-2	chr18:75073543–75074205	+	intergenic	663 bp
AC015911.3	ENSG00000267074	lnc-SLFN14-3	chr17:35499690–35510270	–	intronic	849 bp
AL365361.1	ENSG00000259834	lnc-KCNA3-3	chr1:110653560-110657040	–	intergenic	3481 bp
USP30-AS1	ENSG00000256262	USP30-AS1	chr12:109051791–109053971	–	antisense	1234 bp

3.6. RS and prognostic prediction model construction

According to the LASSO regression model, the coefficients of the candidate prognostic markers were determined, and the RS was calculated. The formula was as follows: RS = (−0.1203) * AC015819.1 + (−0.0644) * AC015911.3 + (−0.0186) * AL365361.1 + (−0.0067) * USP30.AS1. Subsequently, the Maxstat package determined that the best cutoff for predicting the survival time of patients with osteosarcoma by RS was −0.6760. According to the cutoff value, patients with osteosarcoma were allocated to high/low-risk groups, and those without survival information were excluded. Finally, 29 and 56 patients were included in the low-and high-risk groups, respectively. The prognostic survival of patients with a high RS was lower than that of patients with a low RS (Fig. 9A). The ROC curves of 1-, 3-, and 5-year survival predicted by the RS are shown in Fig. 9B, among which the predictive ability for 5-year survival was the best (AUC = 0.769). Univariate Cox regression analysis showed that tumor metastasis had an impact on patient survival in addition to RS/grouping (Fig. 9C).

Fig. 9.

Univariate and multivariate Cox regression analysis of RS for predicting clinical prognosis and survival in patients with osteosarcoma. A: Survival curves of the high/low RS groups; B: ROC curve; C: Univariate Cox regression analysis; D: Multivariate Cox regression analysis.

Based on tumor metastasis and RS, a COX regression model was applied to the multivariate prognostic prediction model. The forest plot is shown in Fig. 9D. The rms package was used to create a nomogram and calibration curve to predict the probability of survival at 1, 3, and 5 years in patients (Fig. 10àD). These variables were good predictors, as shown by the nomogram constructed using tumor metastasis (C-index = 0.701) or RS (C-index = 0.698) alone. The c-index of the survival prediction model constructed using the combination of metastasis and RS was 0.800, indicating better prognostic prediction ability.

Fig. 10.

Multivariate Cox regression model analysis of RS. A: Nomogram; and B–D: calibration curves.

3.7. AC015911.3 and AL365361.1 promote osteosarcoma

We first detected the expression of four lncRNAs in different osteosarcoma cells, and found that AC015911.3 and AL365361.1, had the highest expression in 143 B cells (Fig. 11A and B, Fig. S1A). There is no doubt that the content of TNF-α, one of the markers of the immune response, in 143B cells was significantly higher than that in hFOB 1.19 cells (Fig. 11C, Fig. S1B). We then focused on the molecular mechanisms of AC015911.3 and AL365361.1 in osteosarcoma. siRNA with three different sequences of AC015911.3 and AL365361.1, significantly reduced their expression in 143 B cells; among them, siAC015911.3–2 (siAC015911.3) and siAL365361.1–3 (siAL365361.1) had the most obvious effect (Fig. 11D and E). After knockdown of AC015911.3 and AL365361.1, cell viability decreased significantly in hFOB 1.19 cells (Fig. 11F) and 143 B cells (Fig. 11G). Similarly, after the knockdown of AC015911.3 and AL365361.1, cell invasion ability decreased (Fig. 11H) and apoptosis increased (Fig. 11I).

Fig. 11.

AC015911.3 and AL365361.1 promote osteosarcoma. A: AC015911.3 was upregulated in the 143B cells; B: AL365361.1 was upregulated in the 143B cells; C: TNF-α was upregulated in the 143B cells; D: Verification of knockdown efficiency of AC015911.3; E: Verification of knockdown efficiency of AC015911.3; F: Knockdown of AC015911.3 and AL365361.1 decreased hFOB 1.19 cell viability; G: Knockdown of AC015911.3 and AL365361.1 decreased 143B cell viability; H: Knockdown of AC015911.3 and AL365361.1 decreased cell invasion ability; I: Knockdown of AC015911.3 and AL365361.1 increased cell apoptosis.

4. Discussion

Despite improvements in treatment strategies, most patients with metastatic or recurrent osteosarcoma have a poor prognosis [45]. Immunoreconstitution can inhibit osteosarcoma recurrence and improve the survival rate of metastatic osteosarcoma [46]. A previous study showed that abnormal lncRNA expression in osteosarcoma is highly linked to poor prognosis [47]. lncRNAs are conducive to the early diagnosis of osteosarcoma and contribute to improving the survival probability of patients with osteosarcoma [48]. Our study illustrates that lncRNAs (AC015819.1, AC015911.3, AL365361.1, and USP30.AS1) are potential prognostic biomarkers of osteosarcoma.

We downloaded the transcriptome data (mRNA + lncRNA) of patients with osteosarcoma (N = 88) from TCGA database. The transcriptome dataset GSE99671 (mRNA) for osteosarcoma (N = 18) and normal bone tissue (N = 18) was downloaded from the GEO database. The TIME is typically associated with immune cell invasion [49]. As expected, differences in immune infiltration scores were found in the TIME of patients with a death/survival outcome, suggesting that the TIME plays a role in the outcome of patients with osteosarcoma. Additionally, we found that the higher the degree of immune infiltration, the longer the survival time. In line with our conclusion, the immune score of patients with osteosarcoma was associated with their survival outcomes, and a high immune score indicated that patients had an advantage in survival time [50]. Patients with osteosarcoma and reduced immune cell infiltration frequently have high metastasis rates and poor clinical outcomes [51]. Immune cells in the TIME are associated with tumor treatment and prognosis [15]. The TIME of osteosarcoma consists primarily of macrophages, T lymphocytes, and other subsets, such as B lymphocytes and mast cells [52]. The CIBERSORT algorithm further determined the proportion of infiltrating immune cells in the TIME of osteosarcoma. In our study, macrophages (M1, M2) and monocytes infiltrated more in the TIME of patients with a high immune score, whereas dendritic cells, resting NK cells, naïve CD4 T cells, and gamma delta T cells infiltrated patients with a low immune score. The proportions of memory B cells, naïve T cells, M2 macrophages, and activated NK cells are enhanced in high/low immune subsets [53]. Gomez-Brouchet et al. showed that the higher the tumor-associated macrophage infiltration, the lower the metastasis and the better the prognosis [54]. Hence, different immune cell infiltrates in osteosarcoma may serve as prognostic indicators and immunotherapy targets.

Some lncRNAs are important regulators of TIME [55,56]. Abnormal expression is highly correlated with the occurrence and metastasis of tumors [57,58]. Several studies have shown the significant role of immune-related lncRNAs in the assessment of cancer patient prognosis [59,60]. A total of 80 DElncRNAs and 1125 DEGs were screened from TCGA database. To analyze their biological functions and explain their role in osteosarcoma, we subsequently conducted a co-expression analysis of DEGs and DElncRNAs and screened 11 lncRNA-mRNA pairs. Similarly, Shi et al. identified that lncRNA-C3orf35 and HMGB1 were linked to poor prognosis in osteosarcoma patients, and high levels of lncRNA-C3orf35 and HMGB1 were correlated with a low proportion of macrophage infiltration and low immune scores [61]. Next, we performed immune-related DEGs functional enrichment analysis to determine the functional categories and biological pathways of the DEGs. KEGG, GO, GSEA, and GSVA analyses revealed that the DEGs were involved in chemokine signaling pathways and cytokine-cytokine receptor interactions. Functional enrichment analysis revealed that immune responses and T-cell receptor cascades represented the main functions of immune-related genes in DEG between the high- and low-risk groups [62]. These deregulated immune genes may be the basis of TIME changes, and these findings provide a direction for further studies on immune response mechanisms.

WGCNA can identify a set of co-expressed genes (modules), and the modules can be associated with phenotypic data for analysis to mine potential marker genes [63]. Next, we performed WGCNA on the selected DElncRNAs and identified three modules that correlated with immune cell infiltration. Several studies have shown that lncRNAs are essential for predicting the prognosis of individuals [59,60]. To evaluate the association between DElncRNAs and the prognosis and survival of patients with osteosarcoma, the identified DElncRNAs were analyzed using univariate COX regression and LASSO regression. Four lncRNAs (AC015819.1, AC015911.3, AL365361.1, and USP30-AS1) were identified as prognostic markers (AUC = 0.744). Moreover, these four lncRNAs were positively correlated with immune infiltration scores. Most importantly, we found a significant positive correlation between M1 macrophages and AC015911.3, AL365361.1, and USP30.AS1. These genes are regulated by USP30.AS1 is mainly related in genetic regulation and the immune system [64]. The level of the immunophenotype-related lncRNA biomarker USP30-AS1 is correlated with immune cell infiltration in glioblastoma multiforme [65]. Five immune-related lncRNA signatures, including USP30-AS1, could predict the prognosis of cutaneous melanoma and contribute to immunotherapy [66]. Based on bioinformatics analyses, Zhang et al. revealed that certain lncRNAs (e.g., AL365361.1) with strong correlations with immune scores may modify the TIME of patients with high immune scores [67]. Taken together, the expression patterns of four lncRNAs (AC015819.1, AC015911.3, AL365361.1, and USP30-AS1) were associated with immune cell infiltration in osteosarcoma. The results of our experiments verified this hypothesis.

Metastasis and RS are two independent prognostic factors associated with overall survival in osteosarcoma patients [68]. Prior research has built an RS model to examine the prognosis of patients with osteosarcoma and lung adenocarcinoma based on tumor-infiltrating immune cells, and its prognostic value is superior to that of the TNM staging system. The immune RS model can be used to evaluate patients with recurrence risk [69,70]. The TNM staging system can be applied to stage-adapted therapy and prognostic prediction, showing more clinically relevant differentiation than the modified Masaoka staging system [71]. RS significantly correlated with metastasis, and high-risk patients were more likely to have tumor metastasis. A relevant study using Cox regression analysis illustrated that metastasis and RS were independent prognostic factors for osteosarcoma [72]. Furthermore, we constructed RS-and/or metastasis-dependent survival prediction models to assess patient outcomes. Interestingly, the results revealed that the two variables of tumor metastasis and RS alone could be used as good prognostic indicators, and the combination of the two variables showed better prognostic ability. The prognosis of patients at a high risk of osteosarcoma is worse than that of patients at a low risk, and tumor metastasis is another factor that affects the prognosis of patients [50,68]. Overall, the survival prediction model combined with metastasis and RS showed potential application for clinical prediction of the prognosis of patients with osteosarcoma.

However, this study has some limitations. First, the risk-scoring model was not validated in multicenter clinical trials or prospective studies. Second, the functions and mechanisms of these four immune-related lncRNAs remain unclear. Therefore, in future studies, we need to validate the risk-scoring model in multicenter clinical trials and prospective studies. Additionally, the functions and mechanisms of these four immune-related lncRNAs require further study.

5. Conclusions

Here, we identified abnormally regulated lncRNAs in osteosarcoma and lncRNAs (AC015819.1, AC015911.3, AL365361.1, and USP30.AS1) as potential biomarkers for OS prognosis.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article.

Funding

This work was sponsored by Henan Provincial Medical Science and Technology Research Project [grant number LHGJ20220203].

CRediT authorship contribution statement

Bangmin Wang: Writing – original draft, Software, Formal analysis, Conceptualization. Xin Wang: Writing – original draft, Software, Methodology, Data curation. Xinhui Du: Writing – original draft, Visualization, Validation, Resources. Shilei Gao: Writing – original draft, Resources, Methodology, Investigation. Bo Liang: Writing – review & editing, Methodology, Formal analysis, Conceptualization. Weitao Yao: Writing – review & editing, Supervision, Methodology.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.